WELFARE EFFECTS OF TRADE BARRIERS ON MALAYSIAN INDUSTRY: AN ALTERNATIVE APPROACH

Wai Kun C Lau (1718460)

A Dissertation Submitted In Fulfilment Of The Requirements For The

Degree of

DOCTOR OF PHILOSOPHY

FACULTY OF BUSINESS & LAW

SWINBURNE UNIVERSITY OF TECHNOLOGY

April 2020

i

Abstract

Malaysian car industry has been heavily protected by tariff and non-tariff tools since it was founded in 1983. Despite excessive tariffs imposed on foreign , the demand for foreign cars increases after the Asian financial crisis 1997 while the demand for domestic cars declines. Partial equilibrium framework is applied in this research because the car industry’s contribution to GDP is very small and the focus of this research is specifically on the car industry.

Since cars are durable and differentiated, changes due to technological advancement may influence car demand. This research applies Discrete Choice model to account for car characteristics in addition to socio-economic factors for analysis of car demand in . Logistic regression analysis results show factors that influence car demand are: horsepower, fuel consumption, and car size that is measured by number of passengers. Results suggest that non-tariff barriers and government incentives given to the civil servants have significant influence on Proton cars’ demand, and foreign car makers that have been operating in Malaysia before the founding of Proton enjoy their reputation from their historical experience and performance. While it is often believed that European cars have ostentatious value in Malaysia, the results show otherwise.

Price elasticity of demand for major car makes is estimated based on the average horsepower, car size and fuel consumption. Results show that in general, the demand for car is elastic in Malaysia with Proton cars’ demand having the lowest price elasticity of demand. Price elasticity range from -1.73 to -4.79 for domestic cars and -3.1 to -8.2 for foreign cars in Malaysia compared to -3.9 to -24.3 for domestic cars and -3.1 to -27.5 for foreign cars in the United States. The differences in the range of price elasticity in the two countries imply that there are less variety of choices available in Malaysia than in the United States. Relatively low price elasticity of demand in Malaysia may also reflect relatively lower income level’s effect on highly differentiated cars. The cost of protectionism is approximately 16.7 per cent of real GDP and 73.6 per cent of manufacturing sector’s GDP contribution in 2014.

Finally, assessment on efficiency using Data Envelopment Analysis and ratio analysis show that protectionism has led to high inefficiency of domestic car makers. Although the cost of inefficiency cannot be completely identified and measured, the results show

ii that there is serious excess capacity problem associated with protectionism. The research also highlights other social costs highly likely to be associated with protectionism but, not easily traced and measured due to missing data.

iii

ACKNOWLEDGEMENT

To GOD be the glory. Thanks be to the Lord, my Saviour who is loving and knows my heart desire to pursue PhD with an Australian university. Without HIM, nothing is possible.

My heartfelt gratitude goes to my supervisors, Dr. Omar Bashar and Assoc. Professor Malcolm Abbott for their invaluable comments and suggestions, for being very understanding and helpful. Many thanks to the panel members namely: Dr Chee Jin Yap, Dr Jeremy Nguyen and Dr Mark Bowden for their time and constructive comments over the years.

The 2nd Adelaide PhD Summer Institute in International Trade has benefited me greatly during which I presented my dissertation. Many thanks to Dr. Frank Staehler, Dr. Martin Richardson, and Dr. Benedikt Heid for their invaluable comments and suggestions that helped improve my dissertation. Thanks to Dr. Calvin Lee, an ex-colleague from South Korea for his questions and suggestions that led me to case study of and South Korean car industries.

I am grateful to the late Mr. Raymond Lim, an ex-general manager of UMW Sdn. Bhd. (Kuching), Mr. Edmond Lim, the then administrative and sales manager of UMW Toyota Sdn. Bhd. head office in , Mr. Anthony Goh of Toyota Boulevard (Kuching) for their time, published information and unpublished information. Many thanks to Mr. Jonathan Teng, then with Road and Transport Department, Kuching for data.

I would like to thank managers of several car service centres from different states of Malaysia, heads of some government offices, principals of schools and higher learning institutions and, department heads of some business organizations for their approvals and assistance to distribute questionnaires in their offices. I also would like to thank Mr. Ambrose Tai of St. Joseph Private School for proof reading the dissertation.

Many thanks to anonymous external examiner and Associate Professor Kenneth Jackson for their comments and words of encouragement in their Examiner’s report.

iv

I deeply appreciate Reverend and Mrs. George Tay, brothers and sisters in Christ for their prayers, and Mr. Chan Kok Poh, my A level Economics teacher in , who has been encouraging me since the first day I was in his class until today. I would like to thank my mother and my sister who stood in when I was busy writing my dissertation and work, and my children, Scarlett and Perseus for putting up with my long work hours and made me laugh at home.

v

Praise the LORD for YOUR love endures forever

My LORD, You provide so that I have;

My GOD, You protect me and strengthen me;

My FATHER, You love me and comfort me.

I was searching, I found You waiting;

I tried to prove myself, I found Your grace;

I tried to quench my thirst, I found Your water flows.

I know I have everything because You are with me.

You have never forsaken me. Thank You LORD.

In Jesus’ name. Amen.

vi

Declaration

I hereby declare that this dissertation contains no material which has been accepted for the award to candidature of any other degree or diploma, except where due reference is made in the text of the examinable outcome. It is to the best of my knowledge that this dissertation does not contain material previously published or written by another person except where due references are made in the text of the examinable outcome.

Wai Kun Callie Lau

vii

LIST OF TABLES Table 2.1 Comparison of import duties (%) for CBU and CKD cars as in 20 2003 and 2011 Table 2.2 Comparisons of selected car made and their efficiency 29 Table 2.3 Price comparison across different car makes of 1.6 litre engine 34 Table 2.4 Production, domestic sales and exports: Year 2017 39 Table 3.1 Top 10 car manufacturers: Year 2017 66 Table 3.2 Summary of literature survey in car market: Models and 79-80 estimation of elasticities Table 4.1 Summary of selected literature on car market: Logistic 121 regression models and sample sizes (in chronological order), and comparison with studies in related areas Table 4.2 High range car makes - Quantity and relative market shares: 123 Year 2012 Table 4.3 Sample size determination: Low and medium range 124 Table 4.4 Ratio of civil servant to population: Selected countries 128 Table 4.5 Composition of civil servant: Ethnic groups 128 Table 4.6 Sample size by states 139 Table 4.7 Samples: Mean price, power and sizes 141 Table 4.8 Car types: Characteristics by size 142 Table 4.9a Demographic of car owners’ characteristics: Number and 143 percentage Table 4.9b Demographic of car owners’ income by range and group: 145 Number and percentage Table 5.1 Comparisons of prices of selected car makes - by segments 163 Table 5.2 Market shares in year 2012, 2015 and sample shares 165 Table 5.3 Light trucks: First or second choice 167 Table 5.4a Summary: Number of passengers (NS) and Income group 169 (YGRP) Table 5.4b Tests of association: Income group and car size (NS) 170 Table 5.5 Test of correlation: Car size and number of dependents 171 Table 5.6a Interaction of number of car in households (NOC) with car 172 buyers characteristics: Correlation, p-value, and covariance Table 5.6b Interaction of number of car in households (NOC) with car 174 characteristics of first and second choice: Correlation, p-value, and covariance Table 5.7 Characteristics of first (1st) and second (2nd) choice: 175 Correlation, p-value, and covariance Table 5.8 Estimation of β major car makes 178-9 Table 5.9 Comparison of price elasticity of demand (h) for selected car 184 makes: by horsepower (cc) Table 5.10 Cost of protectionism: Current tariff rates and 50 per cent 187 reduction viii

Table 5.11 Estimation of interaction term bo: car characteristics and car 188 buyer attributes Table 5.12 Income elasticity of demand: Domestic and foreign cars below 191 1800cc Table 5.13 Estimation of interaction term bu: unobservable car 192 characteristics by car makes Table 5.14 Estimation of interaction term bu: unobservable car buyer 194 characteristics Table 5.15 Accounting data and statistical data: Major car makes 196 Table 5.16 Ratio analysis 196 Table 5.17 Data Envelopment Analysis - Efficiency index 198 Table 5.18 Comparison of efficiency index - UMW Toyota: Before and 200 after 30 per cent tariff reduction Table 5.19 Production input multiples - Capital and labour: Japan and 201 Malaysia Table 6.1 Summary of elasticity findings in a priori studies and this 221 research and comparisons across studies

LIST OF FIGURES Figure 2.1 Number of new cars registered - Proton and others: 1980 - 2017 26 Figure 2.2 Malaysia GDP growth: Years 1984 - 2017 32 Figure 2.3 Growth of new car registered: Years 1984 - 2017 32 Figure 2.4 Market shares: 2001 - 2017 33 Figure 2.5 Top 15 car exporting countries: Dollar value (USD) and global 43 shares (%): Year 2017 Figure 3.1 Production and consumption in the Standard Trade Model 51 Figure 4.1 The effects of trade barriers removal 97 Figure 4.2 Market shares of major car makes 123 Figure 4.3 Car makes by income groups 140 Figure 4.4 Least cost combination and maximum output 150 Figure 5.1 Higher range cars: Unit sold in year 2012 168 Figure 5.2 Input (Ringgit) and output (sales in quantity): Toyota, , 195 and Proton

ix

LIST OF ACCRONYMS -2 Log L Logarithm of likelihood function multiplied by 2 ADF Automotive Development Fund AIC Akaike Information Criterion ANCAP Australasian New Car Assessment Program AP(s) Approval permit(s) ASEAN Association of Southeast Asian Nations AUD Australian Dollar BLP Berry, Levinsohn & Pakes CKD Complete knocked down CBU Complete built-up DEA Data envelopment analysis DEP Number of dependents DMU(s) Decision making unit(s) DRB-HICOM Diversified Resources Bhd - Heavy Industries Corporation of Malaysia Bhd ESCAP United Nations Economic and Social Commission for Asia and the Pacific GDP Gross Domestic Product GEN Gender GNI Gross National Income GNP Gross National Product HICOM Heavy Industries Corporation of Malaysia Bhd KML1, KML2 Fuel consumption: Kilometre per litre petrol of first choice, of second choice HP Horsepower HP1, HP2 Horsepower of first choice, of second choice LP1, LP2 Logarithm of price of first choice, of second choice MAA Malaysian Automotive Association MIDA Malaysian Industrial Development Authority MITI Ministry of International Trade and Industry ML Maximum Likelihood MPV Multi purpose vehicle NAP National Automobile Policy NEP National Economic Policy ND Dummy variable: National and non-national cars NOC Number of cars in a household NS Number of passengers NS1, NS2 Number of passengers of first choice, of second choice OTR On the road PEKEMA Persatuan Pengimpot & Peniaga Kenderaan Melayu Malaysia (Association of Malay Importers and Traders of Motor Vehicles Malaysia)

x

PERKASA Pertubuhan Pribumi Perkasa (Mighty Native Organisation) PERODUA Perusahaan Otomobil Kedua Sdn. Bhd. (Second Automobile Enterprise) PROTON Perusahaan Otomobil Nasional (National Automobile Enterprise) RM Ringgit Malaysia UD Dummy variable: Luxury cars and non-luxury cars UMNO United Malays National Organisation US United States USD U.S. Dollar RC Race Sdn. Bhd. Sendirian Berhad (private limited company) SPEC1, SPEC2 Dummy variable: Car specification of first choice, of second choice: full specification or non-full specification SEC Dummy variable: Public and private sector SUV WTO World Trade Organization YGRP, Y Income group, Income

xi

CONTENT Pages 1 Introduction 1 1.1 Malaysia Economy Background 1 1.2 The Development of 2 1.3 Performance of Proton 5 1.4 Problem Statement 7 1.5 Justification of Study 9 1.6 Objectives of Research 10 1.7 Research Questions and Hypotheses 11 1.8 Chapters Organization 13 1.9 Definition of Terms and Assumptions 14 1.10 Contributions of Research 15 1.11 Conclusion 16 Notes 16 2 Trade Barriers and Infant Industry Growth: Malaysia, Thailand 17 and South Korea 2.1 Introduction 17 2.2 Tariff barriers 20 2.3 Non-tariff Barriers and Their Social Impacts 21 2.3.1 Approval permits (APs) and Trade licenses 22 2.3.2 Guaranteed loan approval and hire-purchase rate 23 2.3.3 Car-scrapping scheme 24 2.3.4 Other non-tariff measure 24 2.4 Proton’s Performance 25 2.5 Infant Industry Growth 30 2.5.1 Domestic consumption 30 2.5.2 Prices 33 2.6 The Implication of Extensive Government Protectionism 35 2.7 Takeover of Proton 37 2.8 Case Study 1 - Thailand 37 2.9 Case Study 2 - South Korea 40 2.10 Conclusion 44 Notes 45

xii

3 Literature Review 46 3.1 Introduction 46 3.2 The Basic Trade Models 47

3.2.1 The Ricardian Model 47 3.2.2 The Heckscher-Ohlin Model 50 3.2.3 The Standard Trade Model 51 3.3 The Benefits of Free Trade 52 3.3.1 Economies of scale 53 3.3.2 Variety of choices 54 3.3.3 Competition and efficiency 56 3.3.4 Technology transfer and innovation 58 3.4 Protectionism - A double edged sword 59 3.4.1 Infant industry growth 60 3.4.2 Specialization in creation 65 3.4.3 Technology diffusion 67 3.5 The Costs of Protectionism 69 3.6 The Costs of Protectionism - Studies at firms level 71 3.7 The Impact of Protectionism of Domestic Car Maker 74 3.8 Car Demand 76 3.9 Conclusion 81 Notes 83 4 Methodology 84 4.1 Introduction 84 4.2 Research Philosophy 86 4.2.1 Metatheoretical assumption 1: Ontology 86 4.2.2 Metatheoretical assumption 2: Epistemology 87 4.2.3 Metatheoretical assumption 3: Research object 88 4.2.4 Metatheoretical assumption 4: Research method 88 4.2.5 Metatheoretical assumption 5: Truth 89 4.2.6 Metatheoretical assumption 6: Validity 90 4.2.7 Metatheoretical assumption 7: Reliability 91 4.3 Ethical Issues 91 4.4 Theoretical Framework 93 4.4.1 Welfare loss 95 4.4.2 Models 99 4.5 Literature Review - Logit model 100

xiii

4.5.1 Origin of logit 100 4.5.2 Development of logit model 102 4.6 Application of Logit Model - Overview 109 4.7 Models 113 4.7.1 Estimation 117 4.7.2 Sample 120 4.7.3 Data and variables 126 4.7.3.1 Selection of car characteristics 133 4.7.3.2 Missing data and false data 136 4.7.3.3 Demographic of data 138 4.8 Inefficiency - Sources and Definition 145 4.9 Data Envelopment Analysis (DEA) - Application 149 4.9.1 CCR Model 151 4.9.2 Sample 153 4.9.3 Data 155 4.10 Conclusion 156 Notes 159 5 Results and Analysis 160 5.1 Introduction 160 5.2 Preliminary Analysis 162 5.3 Estimation/Test Results and Analysis 177 5.3.1 Estimation and interpretation 177 5.3.2 Price elasticity of demand and deadweight loss 183 5.3.3 Cost of protectionism 186 5.3.4 Socio-economic factors influencing the demand for cars 188 5.3.5 Efficiency 195 5.4 Conclusion 201 5.4.1 Car characteristics and car buyer characteristics 201 5.4.2 Price elasticity of demand and cost of protectionism 204 5.4.3 Efficiency 206 6 Discussion 208 6.1 Introduction 208 6.2 Benefits of Trade 209 6.3 Costs of Protectionism 211 6.3.1 Allocative efficiency 212 6.3.2 Disappearance of products 215 6.3.3 Unproductive profit-seeking activities 216

xiv

6.3.4 X-efficiency 218 6.3.5 Other costs borne by car buyers and society as a whole 219 6.4 Elasticities and Car Characteristics 220 6.5 Other Welfare Effects of Trade Barriers 223 6.6 The Benefits of Protectionism 225 6.7 Tariffs and Deadweight Loss 226 6.8 Conclusion 227 Notes 228 7 Conclusion 229 7.1 Introduction 229 7.2 Welfare Effects of Trade Barriers 231 7.3 Strengths and Weaknesses of Research 234 7.4 Conclusion 237 7.5 Future Research 238

REFERENCES 240

APPENDICES 258 A. Statistics 258 B. SAS® generated output 259 C. Application for Ethics Approval of a Research Protocol 277 D. Application for Ethics Approval of a Research Protocol - 291 Additional notes E. Human Research - Modification/Additions to Approved Project 297

xv

xvi

Chapter 1 Introduction

1.1 Overview of Malaysia Economy

Malaya, a British colony prior to its independence in 1957 is ethnically heterogeneous in its population. The Malays who claim to be the “sons of the land” (Bumiputra) constitute to about 67 per cent of the population, the Chinese and Indian constitute to about 25 per cent and 7 per cent of the population respectively, and others including Eurasians constitute to about 1 per cent of the population (Background note: Malaysia’s 2012, 2012). During the colonial period, Chinese, Javanese and Indian were encouraged to migrate to Malaya. Majority of the Chinese involved in urban-based tin-mining activities while the Indians settled in the semi-rural plantation. The Malay peasants were mostly involved in fishery and paddy plantation.

During the colonial period in the 1940’s, Malaya’s economy relied heavily on the agricultural sector with the major exports being tin and rubber that were mostly controlled by the British, and to a smaller extent, the Chinese. As the exports of tin and rubber were lucrative, the British were more interested to invest in tin-mining and rubber plantation. Such movement towards specialization subsequently led the economy to greater income increase among the Chinese compared to the Indians and the Malays. By the early1950’s, the European companies controlled about 70 per cent of the export and import trade while the Chinese and Indian constituted to about 10 per cent and 2 per cent of trade respectively. The Malays virtually had no participation in the trade.

The Malaysian economy remained dependent heavily on exports of primary sector after independence. The primary sector contributed to 45 per cent of the GDP, while the secondary sector and tertiary sector’s contributions are 11 per cent and 44 per cent respectively in the 1960s. As there were volatile fluctuations in the prices of primary commodities together with the anticipation of the depletion of tin deposits and the downward trend in the prices of rubber due to the substitution by synthetic rubber, the Malaysian economy began to diversify its exports with the focus on palm oil and cocoa.

1

Although the economy remained export-oriented, Malaysian government began to direct the economy towards import-substitution. To promote import-substitution and labour intensive industries, the Pioneer Industries Ordinance was introduced in 1958 offering tax haven for pioneering firms.

The development of economy since the colonial period until the 1960’s saw the growing income disparities among the three ethnic groups due to the nature of their economic activities. Consequently, the New Economic Policy (NEP) was introduced in 1970 with the objectives to eradicate poverty and to achieve inter-ethnic economic parity between the Malays and the Chinese. The policy has been well received and considered a success because Malaysia enjoyed high GDP growth and the standard of living of the Malays has overall improved. However, the overall performance of the nation has not been impressive. This is because when compared to other East Asian economies such as Taiwan and Singapore that were in the same income range, they have been operating at high-income level since 1990 while Malaysia remain as a middle-income country.

It is believed that the economic performance and the standard of living of the Malays could have been enhanced if the policy has not been implemented (Gomez & Jomo 1999, p.25). However, there has not been statistical evidence to suggest that NEP has either led to high GDP growth or has choked the potential growth.

1.2 The Development of Automotive Industry

The idea to develop the auto industry and a home brand began in 1960’s as Malaysian government believed that the industry had strong linkages to the other industries, particularly the manufacturing and service industries. The development of the automotive industry started off with the entrance of UMW Toyota, a between UMW Holding Berhad and Toyota Tsusho Corporation, into the industry. By the late 1960’s, six assembly plants were approved by the Malaysian government.

The open Approval Permits (APs) system was introduced in the 1970’s as a tool to restrict the number of imported cars in the economy while serving as a tool for the government

2 to reduce income disparities by allocating the licenses to the Malay entrepreneurs. Although there were six assembly plants, Malaysia continued to depend on the imported cars while the local content began to increase up to 40 per cent. The demand for imported cars remained relatively strong possibly due to the increasing income level and relatively strong currency. During this period, the major players in the market were Toyota and Nissan while there were many other smaller Japanese and European players.

In the early 1980’s, Malaysian government introduced the Heavy Industrial Policy. The policy states that industrialization strategy of the country is geared toward diversification of the country’s exports to reduce Malaysia’s GDP dependency on a small number of activities. In addition, the strategy also aims to increase employment through manufacturing sector and service sector as technological capacities of heavy industry sector expands (The Ministry of Information, Communication and Culture, Malaysia, 2012).

The first phase of the auto industry’s development involved promotion of the growth of the parts manufacturing. However, it is found that firms could not reap economies of scale despite increased in local content requirement. This was attributable to the fragmented parts market that is, ‘too many different make of cars in a relatively small car market” (ESCAP 2002).

Subsequently, the second phase was launched in 1984, with the launching of the National Car Project, Perusahaan Automobil Nasional (PROTON) (meaning “National Automobile Enterprise”), that was a joint-venture programme with Corporation, Japan. The car-maker is heavily protected using tariffs and non-tariffs measures. Given the extensive protectionism measures, Proton managed to gobble up to 90 per cent of the market share in the 1990’s.

The government’s plan created an oligopolistic car market structure with Proton being the leader. Proton’s leadership in the market was made possible by excessive tariff and non-tariff trade barriers erected. However, the advantages and privileges Proton enjoys could not last because of its lack of competitiveness. Proton’s market shares started to shrink since the beginning of 2000’s. By 2005, Proton’s market share was merely 24 per

3 cent while the market share of non-Proton cars continued to increase despite their higher prices. This is partly because the second national car, Second Automobile Enterprise (Perodua) that was founded in 1992 to complement Proton has begun to compete with Proton in the lower-income niche market while economic growth continued to stimulate the demand for foreign cars.

In 2006, the Malaysian government introduced National Automotive Policy (NAP) aiming to promote GDP growth via the automotive industry. The objectives of promoting the automotive sector as stated by the policy are:

(a) To promote a competitive and viable domestic automotive sector, in particular the national car manufacturers;

(b) To promote Malaysia as an automotive regional hub;

(c) To promote a sustainable level of economic value added and enhance domestic capabilities;

(d) To promote a higher level of competitive exports of vehicles, components and parts;

(e) To encourage Malays’ participation in the domestic automotive sector; and

(f) To safeguard the interests of the consumers.

(Malaysia Automotive Association 2006).

Due to the trade agreement in ASEAN, Malaysian government began to liberalize the car industry in 2000s. There have been overall cuts in excise and import duties imposed on cars, and values of the imported cars are gazetted to prevent under-declaration of tax while the Approval Permit (AP) system that restricts the number of cars imported from overseas, is expected to phase out by 2010. However, at the time this research is

4 undertaken, the AP system remains effective. This is because there has been pressure from the Malay entrepreneurs, who benefit from the system lobby to maintain the status of the AP system on political ground. In addition to import licences, the government allocates funds for Proton to encourage research and development and to provide assistance for Malay-owned firms in the car industry to undertake joint-venture projects with either local or foreign firms.

1.3 Performance of Proton

There has been excess capacity in the car industry in the major car producing countries such as the United States, Europe and China. The global car demand was about 64 million units, far below the global supply of about 94 million units (The Economist 13/01/2012). In Malaysia, the domestic car maker Proton, had been utilising about 50 per cent of its production capacity in the mid of 2000s. To protect the domestic car maker, the government launched Automotive Development Fund (ADF) in 2006, and allocated RM400 million (USD 125.4 million) to improve the competitiveness of the automotive industry. In 2009, RM200 million (USD 62.7 million) was allocated for the fund. However, there is no further explanation by the authority how the funds are used and no formal evaluation of the plan to conclude if objectives of the funds are achieved.

During 2006, government’s acquisition of Proton cars went up to 2.2 per cent of Proton’s total sales compared to the average of 0.2 per cent over the period 2001 to 20101 (Road and Transport Department, Malaysia). The statistics show that in addition to financial support and bailout, government may have been assisting Proton by buying Proton cars when there is unexpected hike in stock accumulation.

In 2007, Proton launched its car-scrapping scheme. It gives owners of Proton cars of more than ten years old the entitlement to RM5000 discount if the old cars are exchanged for new national cars. The scheme was supported by the government in 2009 to benefit Proton car buyers because Proton cars have far lower resale value than foreign cars of similar age. This scheme was introduced to encourage repeating Proton car buyers so as to maintain their market shares. Although there is no further study to evaluate the success

5 of the scheme or information regarding the scheme, it can be deduced that the scheme has not been successful because Proton’s market share continues to shrink while Perodua and foreign car’s market share continue to rise.

Over the years since its founding, Proton has been through a series of joint-venture projects with foreign car makers that have vast experience and technology know-how in producing cars while the government continuously injecting funds into Proton so that the ailing car maker can remain in the industry. The arguments for protection of the domestic car maker are employment and, redistribution of income and wealth from the Malaysian Chinese to the ethnic groups that are considered natives to the land. Other argument for protectionism but, not supported by studies, is the argument for GDP growth. There are lack of studies in car industry’s linkages to other industries in the economy as well as studies on effective policies that may promote car industry growth instead of impeding it.

After more than 30 years of protection, Proton anchored itself to a new Chinese partner, as a condition for receiving government’s financial aid of RM1.5 billion (USD338.2 million) in 2016. Proton’s record of joint venture partners shows the following:

(a) Proton has not been able to acquire the knowledge from car makers from developed nations over the last 30 years. The problem is more an issue of the quality of labour than an issue of unaffordable technology cost because of different product cycle phases in Malaysia and in developed nations. As technology and knowledge are readily available, the cost of technology and knowledge are relatively lower in Malaysia than in the technology- and knowledge-exporting countries; and

(b) Malaysia has no comparative advantage in producing cars. Therefore, the economy can be better off if there is freer trade to allow imports of foreign cars or allow foreign car makers to operate freely in Malaysia.

6

Protection is justified if infant industry grew, achieved efficiency in production and led to long run falls of prices (Head 1994, p. 163). Prolonged protectionism of Malaysian car industry therefore, imposes high costs to the society.

1.4 Problem Statement

The infant industry argument is valid if protectionist measures are temporary and the cost of protectionism can be offset by the benefit generated through learning experience (Ederington & McCalman 2011; Sauré 2007; Merlitz 2005; and Krugman 1994). Although the ex-ante studies do not define optimum duration for infant industry growth, most studies evaluate the performance of infant industries after 10 years and 20 years of protectionism for light industries and heavy industries, respectively.

While the founding of domestic car makers improves the welfare of the lower and lower- medium income groups through employment and ownership of private transport, industry players in the import sector and car buyers who prefer foreign cars do not benefit from the protected car industry in Malaysia. The pressure from industry players begins to mount, urging the government to liberalize the car industry when the infant industry has entered its 30th year of heavy protection while there is increasing demand for foreign cars that are perceived to be of better quality over the years. The Malay entrepreneurs on the other hand, lobby for continuing protection on the ground that the infant industry’s assembly plants alone are providing 12,000 jobs. In addition, it is argued that the constitutional rights of the Malays are to be given the privilege in entrepreneurship. Therefore, the Approval Permits system that issues import licenses to the Malays entrepreneurs is to be retained for the holders’ benefits.

The implications of upward trend in the demand for foreign cars are lack of effectiveness of the trade barriers and increasing cost borne by the society. Trade barriers impose monetary cost in the form of higher car prices and maintenance cost to car buyers in foreign car segment. In the domestic car segment, excessive tariffs on foreign cars and controlled oil price make private transport attractive in the 1980s to 1990s. Government’s

7 action to cut oil price subsidies in the early 2000s caused drastic increases in households’ spending on private transport while reducing the demand for other goods.

The increases in private transport ownership have repercussion effects on the car industry as private transport ownership squeezes out public transport operators such as school bus and public buses. Consequently, the dependent on private transport increases, putting more upward pressure on oil price and car price, and in some parts of the country, tolls.

As technology advances, cars become more highly differentiated than they were decades ago. The cost of protectionism will increase if the demand for foreign car increases because of certain car characteristics not offered by domestic cars.

Three sources of protectionism cost identified in Hickok (1985) imposed on car buyers are:

(1) high price of the foreign cars assembled in Malaysia and imported cars;

(2) the switch-over cost incurred as consumers buy domestic cars instead of imported cars; and

(3) the higher price of the domestic cars due to reduced import competition.

Other costs borne by Malaysian society are the switch-over cost of substituting public transport with private transport and, the cost of resources misallocation following the expansion of Proton’s production plants.

An approach to measure the effects of tariff and non-tariff barriers on the economy is to estimate the demand for both domestic cars and foreign cars when there is removal or reduction in trade barriers.

8

1.5 Justification of Study

Studies such as Wemeisflder (1960) and Balassa & Kreinin (1967) find cost of protectionism relatively very small that is, less than 1 per cent of the countries’ GDP (Panagariya 2002, p. 5). However, there are studies that believe the cost is large (Krugman 1990 and Mundell 1962). Other studies suggest underestimation of cost may be due to omission of factors that are not taken into account (see Panagariya 2002). Cost of protectionism is found relatively large when factors such as linkages, X-efficiency, and product disappearance, are accounted for (Ibid, p. 10, 23). It is generally believed that the methodology taken is the reason for the gaps observed.

There have also been lack of studies in protectionism cost in small economies. This study bridges the gaps by studying a protected car industry in a small developing economy in which the car industry has been protected for more than 30 years.

Lancaster (1966) suggests that the demand for differentiated products is a derived demand for the product characteristics. Although Gragg & Uhler (1970) and subsequently, other studies such as Agarwal & Ratchford (1980) and Turnnovsky (1966) attempt to estimate demand functions for differentiated products such as cars, many have not been able to incorporate essential car characteristics that differentiate individual car makes possibly due to availability of data and tools. Berry (1994) proposes study of differentiated products using Discrete Choice Models (DCMs) of which has been used since the early nineteenth century for the study of population growth. This approach is not new in pure science but, it is relatively new in the study of social sciences.

Discrete Choice modelling allows more product characteristics to be accounted for and the statistics generated is less sensitive to sample size as Logistic Regression analysis of discrete choice is probabilistic. DCM is adopted to overcome the problems associated with missing data and lack of transparency in Malaysia.

Since private transport ownership takes up a relatively large percentage of households’ income, prolonged protectionism of the infant industry may have significant impact on the domestic car owners particularly the lower income group as well as the higher income

9 group that prefer foreign cars. Statistics show that there is an upward trend in the numbers of new foreign car registered per year despite excessive trade barriers erected. As such, factors that influence car owners’ choice have to be taken into account. These factors are socio-economic factors and car characteristics.

As the overall contribution of the car industry’s output to the economy’s GDP is less than 0.5 per cent based on the 2012 statistics, a partial equilibrium analysis is adopted in this research. However, due to the extensiveness of the protectionist measures implemented, both positive and negative “spillover” effects are identified but, their costs are not estimated as this will require different methodology and different set of data.

1.6 Objectives of Research

The objective of this research is to analyse if it is still worth developing home brand cars in Malaysia considering the foreign players who have been in the market for nearly two decades before the domestic brand was found. In addition, this research aims to investigate how misallocation of resources due to extensive government intervention may contribute to the cost of protectionism. Under normal circumstances, without government intervention, the cost of excess capacity is reflected by the financial statement of a firm. However, operation of the car industry in Malaysia is unique in that financial and non- financial assistance given are not fully reflected by the financial statement and not published.

This research specifically attempts to:

(a) Estimate the ex-ante cost of protectionism of the “infant” car industry as percentage of GDP, using a partial equilibrium framework, that is, to estimate the cost of protectionism based on the loss of consumer surplus in both domestic car and foreign car markets not captured by the gain in producer surplus. This takes into consideration the change in income spent on private transport of the lower and lower-medium income group that the domestic car maker is targeting;

10

(b) Identify the cost of excess capacity under which Proton has been operating; and (c) Identify spillover costs of protectionism. The spillover costs are costs imposed on other sectors of the economy but not accountable for.

1.7 Research Questions and Hypotheses

The research questions of this dissertation are:

(a) Do car characteristics that differentiate various car makes have significant influence on car demand in a developing economy such that car buyers are willing to buy foreign cars despite high prices?

(b) Has technology diffusion taken place in the infant car industry during the period of protection? If technology diffusion has taken place, the demand function for domestic cars will reflect the contribution of car characteristics to domestic cars. Technology diffusion is also reflected by the productivity level of labour and capital.

(c) What is the cost of protectionism, as percentage of manufacturing sector’s contribution to real GDP, borne by the domestic car and foreign car owners when car features are accounted for?

(d) Does competition in the industry lead to greater price elasticity of demand? It is difficult to determine if competition leads to greater price elasticity of demand because this will entail estimation of price elasticity of demand when trade barriers are reduced or removed. In this research, the estimated price elasticity of demand for cars in Malaysia will be compared with the results estimated by studies carried out in freer economies such as in the United States.

11

The hypotheses are this research are:

Hypothesis 1: Social economic factors such as income, prices, and some car buyers’ characteristics influence the demand for cars.

Although this research includes car buyers characteristics found significant in prior studies, additional car buyer characteristics are added to this research. They are: sector (SEC) in which car buyers are working in, and race (RC) of car buyers. As public sector is dominated by the natives (bumiputra Malays) and incentives are given to civil servant to buy domestic cars, it is expected that either SEC or RC but, not both may have statistical significant effect on car demand.

Hypothesis 2: Car characteristics do not influence the demand for domestic cars.

Hypothesis 2 is based on the observation that the demand for domestic cars has overall demonstrating downward trend despite their lower price than foreign cars.

Hypothesis 3: The cost of protectionism as a percentage of manufacturing sector’s GDP contribution is large.

Although there has not been any studies that define “large”, this study considers any percentage 10 per cent or more of car industry’s contribution to total manufacturing sector’s GDP contribution. This study chooses threshold of 10 per cent because many studies of protectionism cost find the cost to be less than 5 per cent of the economies’ GDP.

Hypothesis 4: Demand for cars is price inelastic.

It is assumed that the demand for both domestic and foreign cars are price inelastic because of limited choices when there are trade restrictions.

12

1.8 Chapters Organization

Chapter 2 presents the history and an overview of the Malaysian car market and the comparison of the Malaysian car market performance with the Korean car market’s performance, and how the car industry performs compared to the other protected industries in the economy. Subsequently, Chapter 3 surveys the literature of the areas related to infant industry’s growth, a priori studies of the demand for car and how trade liberalization benefits the infant industry. This chapter also identifies the research gap and lists the hypotheses of this research.

Chapter 4 describes philosophy of this research and methodology applied in this research. It describes how Discrete Choice Models are used to take into account the characteristics of cars. Since literature does not specifically recommend how car characteristics that differentiate cars are selected, selection in this research is based on their relative importance. The approach taken in this research is described in this chapter. It discusses how data is treated, and how relevant car characteristics are identified and selected from the list of characteristics stated in the survey questionnaires. This chapter also describes how Data Envelopment Analysis is used to examine the cost of protectionism in the market where there is very limited reliable data and limited access to data.

Chapter 5 presents the results and analysis of the research.

Chapter 6 discusses of the implication of the results in the context of taxes, and the effectiveness of the trade barriers. This includes identification of the benefits and/or cost of trade barriers if there are any benefits and/or costs that cannot be measured in this research. These benefits and/or costs are not measured in this research because of data constrains and requirement for different analytical tools. This chapter also highlights the directions for future research.

Chapter 7 is the last chapter summarises the entire dissertation and proposes future research areas.

13

1.9 Definition of Terms and Assumptions

The terms “national cars” and “domestic cars” are referring to Proton and Perodua cars. Malaysian government declared that “national cars” are: Proton, Perodua, and . KIA’s country of origin is South Korea, but since 2003 KIA cars are labelled by Malaysian government as national cars because they are assembled in Malaysia and they possess relatively high proportion of local content. However, since KIA’s market share is relatively very small and there has been very limited published information from the car industry and local government regarding KIA cars, “national cars” and “domestic cars” in this research are referring to both Proton and Perodua only.

In this research, the two national cars Proton and Perodua are specifically identified by their name because Proton enjoys more privileges than Perodua in which the latter’s majority shareholders are Japanese-based corporations. Therefore, it is reasonable to conclude that Proton’s sales are influenced by non-tariff barriers that are not necessarily published.

The assumptions made in this research are:

(1) all the foreign players in the car industry in Malaysia are operating at full capacity in Malaysia;

(2) all the foreign players are reaping constant economies of scales thus, will not expand production scales; and

(3) the multiplier effects of the car industry on the economy is negligible since contribution of the car industry to the economy’s GDP is less than 0.5 per cent based on the 2018’s statistics.

14

1.10 Contributions of Research

Theoretical contribution of this research are as follows:

(a) Microeconomic theory. This study contributes to Microeconomic theory in that the former shows that product characteristics that differentiate products from their competitors have influence on products’ prices and demand. Findings of this study also show that income levels are associated with product characteristics. As a result, price elasticity of demand and income elasticity of demand are also found associated with product characteristics. For example, engine size and car size by number of passengers are associated with prices of the car make. The price elasticity of demand for the car is the response of quantity demanded change to total price changes that comprise of price’s interaction with engine size, and interaction with car size instead of the change of price alone;

(b) International Trade theory. The results of this research provides evidence on the effects of competition on price elasticity of demand for differentiated goods. Comparisons of price elasticity of demand across studies using Discrete Choice modelling show that price elasticity of demand for cars is highly elastic in the U.S. than in Malaysia. This study therefore, supports the view that freer trade leads to competition and hence, greater price elasticity of demand.

The evidence also shows that protectionism without government’s commitment to lift protectionist tools will impede the growth of infant industry, if not destroy it. The results of Logistic Regression analysis show that domestic car maker Proton relies heavily on pricing strategy to capture their market while contribution of car characteristics to demand is statistically insignificant. The results suggest that prolonged protection may have caused complacency of the protected industry hence, unable to compete with foreign car makes that successfully differentiate themselves and experience upward trend in the demand for their cars despite excessive protectionist tools used by Malaysian government.

(c) Discrete Choice modelling. This study contributes to Discrete Choice modelling (DCM) in that the model is applied for study of the relative importance of product 15

characteristics to differentiated demand. DCM allows interaction of product characteristics and socioeconomic factors to study a market in the presence of prices distortion due to trade barriers. Application of DCM provides an alternative approach to study social sciences in an environment where factors that influence differentiated products are not as straight forward as they were when products are relatively simple and marketing strategies were more straight forward.

1.11 Conclusion

This research investigates the market of cars, highly differentiated products in an industry protected by tariff and excessive non-tariff trade barriers. Using Discrete Choice model to analyse car demand, this research contributes to microeconomic theory in that the results of this research suggest that car buyers derive satisfaction from car characteristics. This research also estimates price elasticity of demand for car demand functions derived from Logistic Regression model and subsequently, based on the price elasticity of demand, the cost of protectionism is estimated. As there is lack of studies in protected car market in Malaysia, comparisons of price elasticity of demand over time within a protected industry is not available but, comparisons across countries can be made. Comparisons show that freer trade leads to competition and hence, higher price elasticity of demand.

Notes 1 Figures are based on the number of Proton cars registered as government official vehicles at the Road and Transport Department of Malaysia. The average sales of Proton cars to government exclude sales to government in year 2006.

16

Chapter 2 Trade Barriers and Infant Industry Growth: Malaysia, Thailand, and South Korea

2.1 Introduction

The Malaysian car maker Proton, was established in 1985 after the idea was conceived for nearly two decades and pioneered by the then Prime Minister who believed that the industry has strong linkages in the economy. The production plant had capacity of 80,000 units per year in an area of 923,900 square metres.

While had been a player in Malaya since 1926, Toyota, and component parts manufacturing plants were established in the 1960s. By the end of 1960s there were six assembly plants that were joint venture projects between the European firms and the local firms, approved by the government. However, the local content of the locally assembled foreign cars were not more than 10 per cent. Effective 1970, the local content of the car assembled locally grown up to 40 per cent following the implementation of a localisation policy. There were eleven assembly plants operating in Malaysia prior to the founding of Proton. These foreign owned assembly plants were Citroën, Chrysler, , Fiat, Ford, , Land , Mercedes Benz, Mitsubishi, Peugeot and Volvo.

In 1970, the Approval Permit (AP) system was introduced to promote and provide opportunities for Malay entrepreneurship in the automotive sector, with the ultimate aims of increasing the number of Malay equity holders and the income level of the Malays in the country. Import licences were given free to the Malay entrepreneurs who are mostly politically linked individuals or firms, effectively giving the exclusive rights to import foreign cars to the elite group.

Heavy industry project began following the implementation of the National Automotive Policy. The Investment Act 1986 was introduced after the establishment of Proton in 1985. The Act aims to protect infant industries that obtain the “Pioneer” status to enjoy investment tax allowance. By the end of 2000s, there were 15 motor vehicle producers of which 6 were motor vehicle manufacturers and 9 were assemblers, including the 17 franchise holders of non-national car makers such as Toyota and Honda (Wad & Govindaraju 2011). However, during that period, the number of assemblers has decreased because a few car assemblers such as Ford and , were displaced. They ceased operation and diverted to other segment of the car market such as commercial vehicles and higher range cars (Athukorala 2014, p. 9).

The first national car project was approved in 1983 with the agreement struck between the Heavy Industries Corporation of Malaysia Bhd. (HICOM), Mitsubishi Motor Corporation and . The project was launched in 1985 and , the first national car was introduced as a result of the joint venture with Japanese car-makers. The alliance was threatened by conflicting objectives for instant, the Malaysian government wanted local parts to be used while Mitsubishi wanted their Japanese parts to be used on the ground that local parts were sub-standard.

Proton was privatized in 1995, with HICOM was holding 25.8 per cent of the interest (Motor Trader 2012), while Mitsubishi Motor Corporation and Mitsubishi Corporation each were holding 7.93 per cent of the interest. Although privatized, Khazanah Nasional1 retains its 17.96 per cent share in Proton, giving the government the opportunity to intervene in the operation of the firm. In year 2000, HICOM then, DRB-HICOM, sold their Proton shares to (National Oil Company), a state-owned entity. In the same year, Khazanah Nasional acquired about 43 per cent of the Proton’s share to become the largest shareholder. In 2004, Proton becomes fully Malaysian owned as Mitsubishi Corporation Malaysia and Mitsubishi Motors sold their shares to Khazanah Nasional. In year 2012, Proton is privatized again as DRB-HICOM acquires Khazanah Nasional’s 42.74 per cent shares to hold up to 98.6 per cent of the Proton’s shares (Choong 2012).

In October 1992, Perusahaan Otomobil kedua Sdn. Bhd. (Perodua) meaning “Second Automobile Enterprise”, was established. The second domestic car maker targeted lower middle income group. Unlike Proton that is fully Malaysian owned, Perodua’s shareholders (interest in brackets) are: UMW Corporation Sdn. Bhd. (38 per cent), Med- Bumikar Mara Sdn. Bhd. (20 per cent), Daihatsu Motor Co. Ltd. (20 per cent), PNB Equity Resource Corporation Sdn. Bhd. (10 per cent), Daihatsu (Malaysia) Sdn. Bhd. (5

18 per cent), Mitsui & Co. Ltd. (4.2 per cent) and Mitsui & Co. (Asia Pacific) Ltd. (2.8 per cent).

Although Perodua was expected to diversify the range of cars in the country and to support the components and parts manufacturing, Perodua subsequently competed with Proton. Perodua’s market shares increased from 31 per cent in 2009 to 34 per cent in 2012 while Proton’s share decreased from 27 per cent to 26 per cent during that period (autoworld.com.my 2009 and Malaysian Automotive Association 2012).

Selective intervention has been designed to promote the domestic car maker’s growth on the expense of existing foreign car makers who have been players in the industries for about five decades. While imposing tariffs and restricting imports of foreign cars, the Malaysian government encourages foreign direct investment in the industry. In 1996, Proton acquired Lotus technologies from ACBN Holdings that led to the introduction of Proton Gen-2 model.

As the car-maker’s performance continued to deteriorate and liberalization of the car market by end of 2000s is anticipated, Proton saw a need a partnership with an established car-maker for the technology know-how in order to compete with the foreign car-makers. Thus, Khazanah Nasional negotiated with AG for joint venture in 2004. This involved Volkswagen’s plan to improve parts quality and cost efficiency in parts distribution. As it is found that about 200 parts suppliers are politically well-connected Malay businessmen, the government disagree with VW’s plan to rationalise the vendor system (Azhar 2012). Following Proton’s failed attempt to collaborate with Volkswagen Aktiengessellchaft (VWAG) and the end of the joint venture between Proton and Mitsubishi in 2004, Proton attempted to collaborate with MG Rover in 2005 but the plan failed. In 2007, Proton collaborated with Mitsubishi Motors again.

Proton’s struggle to find an established partner and the failure of the plans imply that Proton has failed to acquire the technological know-how over the last 25 years. Proton’s complacency is due to government’s lack of commitment to lift the protectionism measures and the profit-seeking groups’ lobby for continuous protection on the ground that their interests are protected by the law. Consequently, the car maker is plagued with

19 the production problems related to the quality of the cars produced, management of cost, competitive pricing, and tarnished image. Despite the pressure to liberalize the car market and the belief that the car maker could have performed better if left to compete, the adviser of national car project and the labour who benefit from the protected market believe that protection is necessary (Baharom 2013, p.6-7).

2.2 Tariff Barriers

To protect Proton and to create competitive edge for Proton over other car makes, the Malaysian government imposes complicated tariff barriers. For example, the introduction of import duties for Complete Built-Up (CBU) and Complete Knocked Down (CKD) cars. Import duties imposed on the imported cars that is, CBU cars were higher than the duties imposed on CKD cars of which there was at least 40 per cent of local content. Although there was a cut in the import duties imposed on both CBU and CKD cars after tremendous pressure from the WTO and the U.S., the import duties imposed on CBU remained relatively high and higher than the import duties imposed on the CKD cars (Refer to Table 2.1 below).

Table 2.1 Comparison of import duties (%) for CBU and CKD cars as in 2003 and 2011 2003 After 2011 Engine capacity (cc) CBU CKD CBU CKD Less than 1800 140 42 30 10 1800 - < 2000 170 42 30 10 2000 - < 2500 170 60 30 10 2500 - < 3000 200 70 30 10 3000 and above 200 80 30 10

Sources: Malaysian Industrial Development Authority (MIDA). Malaysian Automotive Association, http://www.maa.org.my/info_duty.htm.

20

In addition to the import duties on cars, there are local taxes in the form of excise duties that range from 75 to 105 per cent and sales tax of 10 per cent imposed on both the CBU and CKD cars effective January of 2007 (Malaysia Automotive Association 2008).

Malaysian Industrial Development Authority (MIDA) intervened in the car market by restricting the range of models the foreign car makers might assemble locally shortly after Proton is founded. This might have effectively contributed to Proton’s success in increasing its market share in the 1980s. MIDA’s intervention was successful partly because the differences in car characteristics of domestic cars and foreign cars were not large. In addition, income level in Malaysia was relatively low as the economy was recovering from recession.

Apart from the tariffs on cars, there was import duty of 10 per cent imposed on car parts and was later increased to 13 per cent in 2001. Following cuts of import duties in 2011, the numbers of passenger car makes available in Malaysia including Proton and Perodua, was 26, compared to 16 in 2003 (autoworld.com.my 2004 and 2011), reflecting the increases of the consumers’ choice after the car market made a small step towards liberalization.

2.3 Non-tariff Barriers and Their Social Impacts

Malaysian government introduce many non-tariff barriers and provides assistance to domestic car maker to deter foreign competitors. These barriers are Approval Permits (APs), guaranteed loan approval and lower hire-purchase rate for civil servants who buy national cars, and product quality standards for imported cars although imported cars from industrialised economies possess higher safety standard and offer better quality standard than the domestic cars.

21

2.3.1 Approval Permits (APs) and Trade Licenses

The AP system that was introduced for the purpose of reducing income disparities between the Malays and the Chinese, allows the holders to import up to 10 per cent of the industry’s volume. The license approved imports of a certain range of foreign cars determined by the government. The system offers two types of APs namely, Open APs and Franchise APs. The Open AP holders are Malays while the Franchise AP holders are mostly Malays. These two groups of permit holders are allocated 6 per cent and 4 per cent of the industry’s total volume respectively (Azhar 2012, p. 50).

Beginning June of 2007, AP holders are allowed to import any range of foreign cars of the holders’ choice. Although this move can be seen as a move to liberalization, it is detrimental to the economy as it contributes to income disparities by giving exclusive rights to a small group of elites. Although it is stated that the AP system was to be abolished by 2010 in the National Automotive Policy (NAP), the government postponed the termination of the Open APs and Franchise APs to 2015 and 2020 respectively. Many industrial players believed that the government is not committed to abolish the system (Mahalingam 2011). To date, the AP system continues to operate.

According to Mahalingam (2011), many Malay entrepreneurs who obtained the APs free ended up selling them to third parties for profit instead of importing cars. Government’s deliberate action to protect Malays in this case, gave rise to market that trades import licenses as commodities. Beginning 2010, the government imposed a fee of RM10,000 for the issuance of new Open AP. The fee is expected to be used for setting up a fund to support the shift of Malay entrepreneurs to other business sectors. The distortion of the non-tariff barrier further extends to other sectors of the economy through government’s deliberate action to protect Malay entrepreneurs.

The Bumiputra Car Dealers Association (PEKEMA) and the non-governmental Malay Supremacy Organization (PERKASA) pose strong opponents to the government’s proposal to terminate the APs. Their objectives are to ensure that “rent-seeking” remains feasible to their members. Therefore, due to some political reasons, the directly unproductive profit-seeking activities will continue to contribute to the cost of

22 protectionism of the industry. Unfortunately, the cost associated with “rent-seeking” behaviour is not directly observable because there is no transparency in the number of APs issued and the list of firm that obtained APs.

In addition to the APs, trade licenses are also tools for achieving the political objective of redistributing wealth from the non-bumiputras to the Malays. According to Athukorala (2014, p. 17) most of the Proton car parts supplying firms are owned by Malays who contribute to the ruling party, UMNO’s electoral votes. As such, it is believed that the government does not have the incentive to lift protectionist measures.

2.3.2 Guaranteed loan approval and special hire-purchase rate

Government has been supporting Proton through guaranteed loan approval for civil servants who planned to buy Proton cars. In addition, civil servants are entitled to special hire-purchase rate that is subsidized by the government. Consequently, the public sector contributes to Proton’s growth on the expense of the efficiency of the loanable fund market.

On the other hand, commercial banks charge up to 2 per cent more on the hire and purchase of national cars than that of the foreign cars. Risk premium is charged on loans for national cars because of the perceived higher default risk associated with the income level of the general buyers of the national cars.

Government’s policy for civil servants’ car loans and private sector’s corrective action imply inefficiency has arised in this segment of the fund market. Relatively high default risk in the domestic car segment implies relatively higher depreciation rate of domestic cars than foreign cars and subsequently, lower resale value of the domestic cars than foreign cars.

23

2.3.3 Car-scrapping scheme

Proton initiated the car-scrapping scheme in 2007 to encourage Proton car owners to trade-in more than ten years old Proton cars for new Proton cars at a discount of RM5,000. In early 2009, Malaysian government proposed one-year car-scrapping scheme for the purposes of boosting car sales and rejuvenate the local motor vehicle industry. The scheme allowed individuals to trade-in their old cars of more than ten years old for new Proton and Perodua cars2.

Although there is no study and evaluation of the scheme, it is believed that the scheme benefited Proton car owners whose cars virtually had no resale value and the car-owners who had the financial ability to acquire new cars. The lower income group that predominantly bought Proton cars have not benefited from the scheme. The scheme was less attractive to the lower income group because the cost of private transport ownership takes up a large proportion of household income. Due to lax in road and transport regulation coupled with relatively high maintenance cost, the use of old cars unfit for roads by the low income group is common. As a result, the scheme effectively targeted the middle-income group of Proton repeating buyers. As foreign cars have better resale value due to their quality and brand name, the scheme was not attractive to the foreign car-owners.

In addition to the tariff and non-tariff barriers, government also supports Proton by acquiring Proton cars for government offices use. The numbers of new Proton cars registered yearly over the years from 2001 to 2010 are fluctuating around 0.1 to 0.6 per cent. The number increased drastically in 2006 up to 2.2 per cent3 when the sales of Proton plummeted.

2.3.4 Other non-tariff measure

Investigation through face-to-face interviewing of senior industry players reveals that there are informal and unpublished non-tariff measures used for protecting Proton. One of the informal and unpublished measure is the regulation of price discount and, time and

24 duration of offer. According to an industry player, non-Proton car makers are required to obtain approval from government agent if the non-Proton car maker wanted to offer price cuts. The approval takes into consideration amount of price cut, the date and duration of the offer. Such regulation aims to prevent non-Proton cars from competing with Proton.

Other informal and unpublished non-tariff measure is the regulation on launching of the new Proton models. That is, Proton has the priority to choose their date for launching new Proton cars while the other car makers are allowed to launch after Proton have launched their new cars.

2.4 Proton’s Performance

Following a series of multifaceted protectionist measures put in place, the car ownership ratio and real GDP per capita for Malaysia is 1:2.9 and USD7,278 respectively, compared to Thailand’s ratio of 1:6.7 in which the economy’s income level was about 54 per cent of Malaysia’s real GDP per capita. While ’s real GDP per capita was about 6 folds Malaysia’s, the former’s car ownership ratio was 1:1.5 (World Bank data, viewed 25/10/2012).

While the relatively high car ownership ratio may be attributable to the success of Proton’s strategies to penetrate into its market niche, the declining market shares over the years particularly since the mid of 2000s implies that Proton is losing its competitiveness in the market.

Proton’s market share was nearly 50 per cent when the Proton cars were launched in 1985. Proton’s sales reached its peak in 1997 and another peak in 2002 after the Asian Financial crisis, selling about 210,000 units in the domestic market. Although Proton gobbled up nearly 70 per cent of the market share in the 1990s such achievement was not due to efficiency. Consequently, the achievement was short-lived. Despite extensive government protectionist measures, Proton has not been able to maintain its market share. The figure hit its minimum in 25 years at 26 per cent in 2006. By 2011, Proton constituted to merely 29 per cent of the total new cars registered in the country although the real price

25 of the lower end range, Saga 1500cc declined from RM40,061.06 in 2003 to RM20,147.23 in 20103.

Figure 2.1 Number of new cars registered - Proton and others: Years 1980 - 2017

800000

700000

600000

500000

400000

300000

200000

100000

0

Years

All new Adj Proton

Notes: 1. The number of new vehicles includes passengers, commercial and 4x4 vehicles. 2. The numbers of Proton cars registered are adjusted for the numbers of Proton cars acquired by the government for government offices use. 3. The number of Proton cars registered in 2017 is not adjusted for government acquisition of Proton cars because information is not available.

Sources: Malaysian Automotive Association (MAA), various years. Ministry of International Trade and Industry, Malaysia (MITI), 2009. Road and Transport Department, Malaysia.

As Proton has lost much of its market share, Proton has not been able to reap economies of scale and produce profitably. Since Proton’s production capacity increased to a medium volume factory of 230,000 units per year in 1997, approximately 70 per cent of 26 the capacity was utilized in the early 2010s. While the numbers of new non-national cars registered increase over the years since the early 1990’s with a dive in 1998, the numbers of new Proton cars sold, adjusted for the number of Proton cars acquired by the government for offices use, has been showing a declining trend (See Figure 2.1).

Proton Waja and Perdana of middle and high range were discontinued. They were subsequently replaced by Inspira in 2010. Perdana model was rejuvenated and launched in 2013. The re-entering of Perdana show more of a political decision than a corporate decision based on economic justification.

Proton’s declining market shares in contrast with the overall upward trend of new cars registered implies the followings:

(1) The lower end of the market that is targeted by Proton, is already saturated;

(2) There is insignificant Proton repeating buyers;

(3) The middle and higher range Proton cars that is, Waja and Perdana respectively, could not compete with foreign cars of similar capacity and range;

(4) The competition between Proton and the foreign car makers is decreasing as the efficiency gaps between the former and the latter is widening (See Bernard, et al. (2004, p.1269) for discussion of American firms’ efficiency and productivity response to foreign competitors); and

(5) There is increasing negative welfare effect on non-Proton car buyers over the years as the numbers of new non-Proton cars registered are reflecting upward trend.

In 2006, there was an excess supply of 250,000 cars in Malaysia due to the tightening of lending policies by the financial institutions. Edaran Automobile Nasional Dealers Association (Malaysia) urged the government to intervene in order to prevent job loss. The lobbyist suggested that as the problem worsens, the number of members who would 27 be forced out of the industry by end of 2006 might be up to 40 per cent of their members who were distributors of Proton cars and parts. The lobbyist’s justification for intervention was to alleviate “unhealthy competition” (Azhar 2006, p. 3).

It is reported that Proton’s profit in 2009 was largely due to cuts in research and development expenditure. The reported profit was estimated to be about 78 per cent lower than it was actually reported if the amount of research and development expenditure is the same as it was in the year 2006 (Wrightreports 2011). This implies that prolonged protectionism may be leading to lack of innovation and hence, lower quality goods produced in this case.

Over the years Proton has been enjoying preferential treatment such as periodic capital injection on concessionary terms by the government through public funds and Petroliam Nasional Berhad (Petronas), a state-owned oil company. In 2006, public fund of RM400 million was allocated for the Automotive Development Fund (ADF) launched to improve the overall competitiveness of the automotive industry but, there was no further clarification of the given objective (.com, 9/7/2007). In 2009, RM200 million was allocated to the ADF. There was however, no information on how the funds have been used or what the government’s strategies were to improve competitiveness. Proton was reported seeking for up to RM3 billion from the government and Petronas again in 2014 for funding its development (Sidhu 2014, p. 1).

In the domestic market, trade barriers have not been able to stimulate growth in Proton after 25 years. The local content for the that was launched in 2010 was about 20 per cent (Kon 2010). Comparatively, the Korean car maker Hyundai, was able to produce cars with more than 90 per cent local content after 10 years of heavy protectionism (Greenbaum 2002).

Proton’s performance in the automobile export market has not reflected its growth. Proton’s exports to the United States failed in the late 1980s because the exported models were unable to meet the American road safety standard. Proton’s success in penetrating the English car market in the 1990s does not mark the car maker’s success in the Western car market. Although exports have increased from 7 per cent in 2004 to 14 per cent in

28

2009, total export is not more than 20 per cent of the total sales with Asia remains the major export market. While Proton’s exports grow over the years, the number of Proton cars exported has never been more than 26,000 units (MITI 2010). The South Korean car maker Hyundai however, was able to capture up to 4 per cent of the world automobile export market and was able to export up to 263,610 units of car to the U.S. after 25 years of establishment (Ebert & Montoney 2002).

Table 2.2 Comparisons of selected car made and their efficiency Holden Hyundai KIA Rio Ford Proton Barina Accent Fiesta Gen2 Fuel use (l/100km) 7.3 6.4 6.1 6.1 6.8

CO2 emission 174 151 145 146 187

Sources: The Saturday Age (Australia), 12/11/2011, p. 6, 10. The Age, Holiday edition (Australia), 30-31/12/2011, p. 14. http://www.proton.com.au/s16/specifications, viewed 22/5/2012.

In 2013, Malaysia’s exports of road vehicles including motorcycles and cushion vehicles, and other transport equipment amounted to RM6,096 million and RM3,532 million respectively (Department of Statistics, Malaysia 2013). The automotive industry’s contribution to Malaysia real GDP was approximately 1.3 per cent. Total number of employment in the car assembly industry and car parts and accessories industry are merely 47,574 in 2012 which is about 10 per cent the number of total employment in the Thai car industry in the same year (The German Chamber Network 2012, p.8).

From the perspective of technology, the Australian, South Korean and American car makes outperform Proton in the same category in Australian car market. Table 2.2 above, shows comparisons made between Proton cars and other made of cars with 1.6 litre engine sold in Australia. Statistics show that Proton’s counterparts such as Holden, Hyundai,

Kia and Ford are performing better than Proton in terms of their fuel used and CO2 emission within the price range of AUD16,000 to AUD19,000. 29

In conclusion, Proton is unable to compete in the overseas market due to its lack of technological competitiveness and its inability to meet the stringent road safety standard based on the occupant protection score given in the ANCAP crash test, for example. Declining domestic market share and insignificant export cause Proton continuously operates at excess capacity. Based on the number new Proton cars registered, the plant operates at 50 per cent of its capacity during the years 2014 to 2016.

2.5 Infant Industry Growth

2.5.1 Domestic consumption

Figure 2.1 shows the numbers of new car registered per year show upward trend starting 1985. Domestic demand for car increases since the founding of Proton is mainly attributable to government’s policy to promote private transport ownership. Trend shows faster rates of increase after 1994, implying that the founding of Perodua may have contributed to accelerated growth of private transport ownership in the country. There is a break in the trend in 1998, reflecting a tremendous fall in car demand amid the Asian Financial Crisis 1997.

Government’s policy to promote private transport ownership essentially targets the lower and lower-middle income groups, causes reallocation of household income from ‘other goods’ including consumption of public transport to private transport. Both Proton and Perodua’s penetration into the niche market not profitable for foreign players also lead to tremendous increases in the demand for related goods such as oil, roads and parking space in the early 1990s.

The demand for car by the lower and lower-middle income group is stimulated by domestic car prices that are made attractive through various promotional tactics such as low down payment, longer loan period, and guaranteed loan approval particularly for civil servants. Oil price control in the country during the period of 1980s to 1990s made private transport ownership artificially cheap hence, further supports the demand for car. Consequently, oil subsidy cuts have significant effect on the lower and lower-middle income group, because switching of private transport to public transport is costly. This 30 reduces the lower income and lower-middle income households’ spending on other goods.

Falls of households’ income allocation for other goods also pose threat and cost to the society when cars are used beyond their useful life. Subsequently, low car values coupled by budget constraint, the problems associated with car owners who drive unfit old cars without insurance become common in the country.

Protectionism has trickle-down effect on other industry mainly public transport that in turns, further causes repercussion effect on the society. Private transport ownership of the lower and lower-middle income groups leads to decreases in demand for public transport and undercut profitability of this industry. Falls in the supply of public transport in turns, put more pressure on the demand for private transport ownership and related goods.

The negative welfare effect on the lower and lower-middle income group is relatively larger than that of the higher income group because of the larger proportion of income spent on private transport by the formers than by the latter. However, this is not fairly reflected by welfare effect measured in absolute monetary terms because the higher income group is highly likely to spend more on transport than the lower income group due to the numbers and sizes of car owned per household.

Based on series of GDP growth statistics (see Figure 2.2) and annual growth of new cars registered (see Figure 2.3) in Malaysia over 34 years, correlation coefficient estimated is 0.65. This implies that there is a relatively strong correlation between income growth and the demand for new cars in Malaysia.

The inconsistencies observed suggest the lack of effectiveness of trade barriers to protect the domestic car makers as well as the presence of other factors that overwhelm foreign cars’ prices and their maintenance cost.

31

Figure 2.2 Malaysia GDP growth: Years 1984 - 2017

12

10

8

6

4

2

0

-2 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 GDP Growth GDP Growth (%) -4

-6

-8

-10 Years

Source: World Bank 2019.

Figure 2.3 Growth of new cars registered: Years 1984 - 2017

100.0

80.0

60.0

40.0

20.0

0.0

-20.0 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

-40.0

Growth of new cars registered (%) -60.0

-80.0 Years

Sources: Malaysian Automotive Association (MAA), various years.

32

When the number of new cars registered is disaggregated into two categories namely, domestic and foreign, it is found that the demand for domestic cars demonstrates gentle downward trend while the demand for foreign cars demonstrates upward trend (See Figure 2.4). Although average oil price has increased since the 2000s as government cut oil subsidies, higher oil prices and high tariffs on foreign cars have not made domestic cars more attractive than foreign cars.

Figure 2.4 Market shares: 2001 to 2017

90

80

70

60

50

40

30

20

10

0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

Proton Perodua Foreign Domestic

Sources: 1. Malaysian Automotive Association (MAA), Press Conference, various issues. 2. www.autoworld.com.my, Malaysian Car Sales Figures. Various publications. 3. www.motortrader.com.my

2.5.2 Prices

Both the tariff and non-tariff are contributing to high foreign car prices in Malaysia (see Table 2.3). Price differences estimated based on the prices in the domestic market and in Australian market shows that extensive protectionism measures cause price differences to range from about RM9,000 to RM68,800. 33

Table 2.3 Price comparison across different made of cars of 1.6-litre engine Car Made Price (RM) Price (AUD) Price difference (a) (b) (RM) (a) - (b) 1 Proton 50,000 16,000 -1,200 (32,000 - 40,000)* 13,000* -5,600 2 Perodua (49,000 - 53,000)* - - 3 Honda (80,000 - 90,000)† (20,500 - 23,000)† 15,400 4 Hyundai (70,000 - 74,000)† 17,000 17,600 5 KIA 70,000 - 82,000 19,000 15,200 6 Mazda 120,000 16,000† 68,800 7 Nissan 85,000 - 100,000 - - 8 Peugeot 73,000 20,000 9,000 9 Toyota 105,000 - - (93,000 - 108,000)† (15,000 - 21,000)† 42,900 10 Volkswagen 110,000 - 140,000 20,000† 61,000

Notes: (1) * 1.3 litre engine (2) † 1.5 litre engine (3) the exchange rate used is RM3.2 : AUD1

Sources: http://www.cardirect.com.my (viewed 31/1/2012) The Age (Australia), October - December, 2011.

For example, a Mazda price in Australia is approximately 50 per cent the price of a Mazda in Malaysia. However, the Proton car prices are higher in the foreign market than in the domestic market. This is may be due to transport cost, differences in features of the car in the Australian market to meet more stringent safety regulation in Australia and to create value for Australian car buyers.

Despite over-priced of foreign cars, the demand for foreign cars shows upward trend (See Figure 2.4). This observation suggests there are other factors that contribute to the 34 demand foreign cars. These factors may be car characteristics which add value to car buyers hence, have positive effect that overwhelms the negative effect of prices.

Foreign car makers’ succeed in their market shares expansion despite heavy trade barriers reflects Proton’s failure to acquire new technology and foreign car makers’ success in product differentiation. Saurè (2007) suggests an infant industry under protectionism may not have the incentive to acquire new technology if there is old technology. Thus, the infant industry is unlikely to mature and contribute to economic growth. Infant industry’s commitment to learn and behave strategically is a reaction to the government’s commitment to strategic trade policies (Leahy & Neary 1999, p. 447).

In the environment with rapid advancement of technology, the technology gap between protected industry and the world widens if capital and/or labour employed in the protected industry were not productive. The gap gives foreign car makers the leverage to alter the technology level for cars exported to economies in which the car industry is protected and the level of technology is low. Consequently, car buyers in the protected economies may pay higher price for technology already obsolete in the technology-exporting economies.

2.6 The Implications of Extensive Government Protectionism

Tariffs have direct impact on buyers of domestic and foreign cars, and domestic and foreign producers through prices while non-tariff barriers’ impact may not be easily quantifiable. For example, cars’ required standard and administrative procedures do not have direct effect on car prices, but influence supply which in turns, affect prices through interaction with the demand. Both tariffs and non-tariffs also have spillover effects on other industry in domestic economy. The impact on foreign economy may not be significant because Malaysia is a small economy.

The spillover costs imposed on other industries in domestic economy are as follows:

(a) Changes in the consumption pattern. There may be significant changes in the consumption of other goods in the lower income households after the founding of

35

domestic car makers. The lower income group’s acquisition of private transport leads to relatively large proportion of their income spent on the car and related goods such as petrol, road tax, insurance and maintenance cost. As a result, total expenditure on other goods declines;

(b) Misallocation of loanable funds. Loanable funds are reallocated from other industries to the car industry as the demand for cars increases. This causes higher cost of borrowing borne by borrowers and higher cost associated with default risk borne by the commercial banks;

(c) Misallocation of resources. Large amount of factors: land, capital and labour are allocated to the establishment of Proton. In addition to implicit safety net in the form of bailout, Malaysian government provided implicit protectionist measure in the form of acquiring Proton cars when their sales were low. Imperfect competition in the industry further led to inefficiency in Proton and other related industries such as car service industry and parts manufacturing and distribution industries;

(d) Decreases in the supply of public transport. Increases private transport ownership led to decline in the demand for public transport hence, eliminating economies of scales in the public transport industry. Falls in the supply of public transport caused job loss in this industry and worsened imperfect substitution of public transport for private transport; and

(e) Environmental costs. Large numbers of unfit cars on rural and urban roads are attributable to the lax in road and transport regulation in Malaysia. Lax in regulation coupled with increasing numbers of new cars registered contribute to increasing environmental cost associated with congestion, pollution and to certain extent, road accidents.

The spillover costs identified are not readily measurable using the methodology and data set applied in this research. In addition, the change in spending behaviour is not directly observable as it is difficult to determine households’ consumption behaviour over the years had they not owned private transport. 36

2.7 Takeover of Proton

Proton began partnership with Chinese car maker Geely since 2017 as part of Proton’s turnaround plan following Malaysian government’s financial assistance of RM1.5 billion (USD338.2 million) given in 2016 ( online, 24/05/2018). Zhejiang Geely bought 49 per cent interest in Proton for RM460.3 million (The Star online, 23/06/2017). Following the takeover, Geely injected RM170.3 million for launching of a new Boyue model instead of Proton’s new model. Geely also implemented a new plan to cut the number of employees, requested vendors to cut prices of car parts and to improve their service quality.

Association of Malay Vehicle Importers and Traders (Pekema) urged government to intervene because Geely’s plan is not in line with the Bumiputra policy that gives privileges to the Malays and other native groups for the objectives of redistributing income and wealth from Chinese to the Malays and other natives. The self-interest group argued that Geely’s plans force the Malay parts dealers to shut down (The Star online, 15/2/2018).

Proton resumed their export to the Middle East countries in 2018. It was reported that 453 Proton cars were exported to Jordan for distribution to other Middle East countries (, 7/9/2018). Proton reported a fall of sales from 70,991 units in 2017 to 64,744 in 2018 (Lye 2019). Geely’s takeover plan may has benefited Geely as their partnership with Proton allows Geely to access the ASEAN market of up to 630 million people in population.

2.8 Case Study 1 - Thailand

Similar to Malaysia, the Thai car industry is founded in 1961. However, instead of developing their own Thai brand, the Thai government encouraged foreign car makers to operate in Thailand. Thai government introduced import taxes of up to 150 per cent in the early 1970s to protect the car industry and subsequently, embargoes were imposed on

37 foreign cars in the late 1970s to ensure that the foreign car makers operating in Thailand were able to capture the domestic market.

Mitsubishi had the advantage of being the first foreign car maker to enter Thailand. Toyota and Nissan plants were later founded in 1962. Other Japanese car makers such as Honda, Isuzu and , and South Korean car makers also entered the market following Mitsubishi. All of these car makers focused on production of small and cheap cars affordable to the developing economy.

In addition to import taxes, supports provided by the government were provision of strategic investment site coupled by improved infrastructure and investment incentives both fiscal such as corporate income tax holidays up to eight years, and non-tax incentive such as land ownership rights for foreign investors and minimum bureaucracy in issuance of work permits for foreign labour. There was also equal treatment of all industry players and no requirement on local content. Liberalised trade foreign exchange market facilitated industry players’ access to the Southeast Asian economies.

Due to the liberalized capital and labour market, complete foreign ownership of firms and employment of foreign skilled labour, there was technology diffusion. As a result, growth of the infant industry stimulated the supporting industry that is, the auto parts manufacturing industry. The car industry enjoyed about 10 per cent growth until the onset of Asian Financial crisis 1997.

By 2012, there were 14 car makers and 7 motorcycle makers that were under the category of large scale enterprises which made up of mainly foreign joint-venture projects. These enterprises generated about 100,000 jobs. In the infant industry, the small and medium size enterprises were subdivided into Tier 1 of which more than 50 per cent of the 690 firms were foreign-owned while Tier 2 and 3 enterprises comprise of 1700 firms, generally local-owned. The small medium size enterprises generated up to 425,000 jobs (Thailand Board of Investment 2012).

The external environment in Thailand has been conducive for foreign direct investment. By 2015, there are 18 car makers in Thailand, including American car makers Ford and

38

General Motor, German car maker Mercedes Benz, Swedish car maker Volvo and Indian car maker TATA (Thailand Board of Investment 2017).

Thailand’s car exports began to gain their momentum after the Asian Financial crisis 1997 because of excess capacity. In 2013, Thailand was ranked nineth world largest car maker and the largest in the Southeast Asian region. Thailand exported 14,020 units of car in 1996 and the statistics surged up to 1,094,089 that is 7703 per cent in less than 20 years (thaiwebsite.com 2014). Its automotive industry contributed to 12 per cent of the economy’s GDP in 2013 (Tractus Asia Ltd 2014) and in 2016 (Thailand Board of Investment 2017). Thailand was ranked twelfth world car exporter and the largest car producer in Southeast Asia in 2016.

Table 2.4 Production, domestic sales and exports: Year 2017 Quantity Average growth (%) Growth (%) Production 157,207 15.8 2.3 Domestic sales 104,302 20.1 13.4 Exports (CBU) 95,834 11.3 -4.1

Source: www.marklines.com 2018, viewed 14/09/2018.

Statistics show that in 2017, the total number of complete built-up (CBU) cars exported was 1,139,696 (www.marklines.com 2018) that is, about 2516 times the size of Proton’s exports in 2018. Refer to Table 2.4, although the exports fell by 4.1 per cent from 2016 to 2017, the average car exports growth was 11.3 per cent (Ibid).

In conclusion, by liberalizing the industry, more jobs were created as foreign car makers were invited to operate in Thailand.

The advantages of liberal policies are as follows:

39

(a) Entries of foreign car makers offer more variety of choice to the domestic economy;

(b) Foreign car makers have control over their product quality hence, domestic economy benefits from foreign car makers’ goodwill built upon their history and experience;

(c) Economies of scales. Foreign car makers may reap economies of scale because of the greater access to domestic economy’s neighbouring countries at lower transport cost; and

(d) Relatively fair competition among foreign car makers improves efficiency.

The Thai car industry’s case shows that an economy is able to benefit from trade by allowing foreign firms to enter the industry in which the domestic economy does not have comparative advantage. By allowing entry of foreign firms into the industry, the Thai economy avoided the cost of protecting domestic firms while allowing technology diffusion to take place in the economy. Although it may be argued that Thailand’s car industry may be too heavily dependent on foreign car makers and may have negative implication on Thailand’s balance of payment, this argument is not be valid if the domestic economy enjoys the benefits of positive externalities such as improved infrastructure and productivity of domestic labour.

2.9 Case Study 2 - South Korea

The South Korean car industry was founded in the early 1960s with the aim to develop domestic car brands. In 2016, South Korea was ranked sixth, by production volume, in the world car manufacturing country (Statista 2018). Out of a total production of 4,228,509 vehicles produced, approximately 90 per cent of the total cars produced were passenger cars while 10 per cent was made up of commercial vehicles.

40

At its infant stage, the Korean car industry comprised three car makers namely: Kia, Ha- Dong-Hwan, and Saenara. Hyundai was later found in 1967 and acquired 51 per cent of Kia’s interest in 1998. Today, the industry comprises Hyundai-Kia, Daewoo, Samsung Motors that was formed after Samsung Motors took over Renault in 2000, and Ssangyong that specializes in sport utility vehicles.

The car industry began with assembly of knockdown kits from Japan and the United States. The protectionist measures adopted by the South Korean government were: embargo of foreign cars, tax exemptions and subsidized loans for assemblers, and duty free for parts and components. To reduce dependency on imported parts and kits, the Korean government gave subsidies to parts producers and encouraged foreign car maker to enter Korean market through joint ventures with the domestic car assemblers. Joint ventures with experienced foreign car makers provided opportunities for the domestic car assemblers to acquire skills and technology necessary for growth. In order to stimulate growth of domestic car market, the Korean government lifted the cap on yearly car registration introduced in 1957.

Hyundai Motor Company began their joint venture with Ford to produce Cortina in the 1960s. In the 1970s, Hyundai entered into joint ventures with Mitsubishi and Honda before entering into joint ventures with the British and Italian car makers. A series of joint venture saw Korean car makers acquired the technology and grew.

The Korean government took a step further in the 1980s to lift the embargo on imported cars was put in place for up to about 25 years. Other forms protectionist tools such as differential taxation rates and tax audit on potential foreign car buyers are introduced after the embargo was lifted. In addition, 60 per cent tariff was imposed on imported cars and non-tariff barriers such as administrative procedures that require extensive paperwork and regulations on quality or safety standard. Until 1990s, due to certain models’ lack of satisfactory crash tests results, the domestic brand is often perceived to be of low quality. Despite significant improvement in quality control, the Korean car maker has not been able to reverse the markets’ perception. Consequently, the car maker was lack the power to control their prices.

41

In addition to history, Korean car maker’s effort to penetrate world car market is weighed down by imprudent expansion of production capacity and a mismatch of industrial strategy. Prior to Samsung Group, a large chaebol’s enter to the car industry, the existing car makers then set up entry barriers by increasing investment in production capacity and through government lobbying. Hence when Samsung Motor Inc. added to excess production capacity in the industry when the chaebol was founded in 1995.

Extensive government intervention in the industry led to rent-seeking behaviour and corruption. Consequently, the Korean government’s techno nationalism strategy had not been successful when world economies were moving towards techno-globalism, making globalization in new technology and car market accessible at very minimum cost. It is believed that without rent-seeking behaviour and corruption, the protected industry might have been able to take advantage of relatively cheap new technology during the season of techno-globalism.

According to Ravenhill (2001, p. 2), the Korean car industry continues to depend on aggressive pricing strategy until the end of 90s. Car industry often makes very slim or negative profit in order to compete in the international car market. The Korean government and industry players saw the need to export their cars after the financial crisis as the domestic demand for cars fell by about 30 per cent in the market that was already saturated prior to the onset of the crisis.

Exports of Korean cars begin to take off after the Asian Financial Crisis. Greenbaum (2005, p. 17) suggests the major factor that contributes to Korean car industry’s growth are the foreign car makers’ direct investment, strategic alliances and imports of car parts. Ebert & Montoney (2007, p. 19) says that increasing demand for Korean cars’ in the U.S. market after the financial crisis was attributable to the U.S. real disposable income, favourable fuel cost, quality, and cost of borrowing.

In conclusion, South Korean car makers’ successful penetration into international car market is attributable to technology know-how acquired, which in turns, led to development of technological independence, and liberalization of trade opened. Access to more foreign markets enable the domestic car makers to reap economies of scale.

42

Figure 2.5 Top 15 car exporting countries: Dollar value (USD) and global shares (%): Year 2017

180 160 140 120 100 80 60 40 20 0

Dollar value (USD) Global shares (%)

Source: Workman 2018.

Statistics (see Figure 2.5) show that in 2017, South Korea is ranked seventh while Thailand is ranked fifteenth by value of total car exported in US dollar in the global car export markets. Although South Korea’s sales declined by 12.3 per cent in 2017, the Korean car makers remain major players in the international car market. This shows that technology diffusion has taken place and Korean car makers are competitive in the world car market after nearly three decades of protection compared to about two decades for Japan’s infant car industry.

However, Truett & Truett (2014, p.91) suggests that it cannot be concluded that the South Korean car industry has matured because barriers to trade still exist. Regulations restrict imports of foreign cars, substantially protecting the domestic car industry. Korean car industry’s success in the world car market implies that technology diffusion has been successful and that they are able to meet the minimum standard requirement in other countries. The issue in Korean car industry’s case is more of an issue of fair trade rather than maturity of the industry.

43

2.10 Conclusion

Market data of new cars registration shows that trade barriers have not been able to stimulate Proton’s growth. Malaysian government’s lack of commitment to lift protectionist measures sees Proton’s downward trend in yearly numbers of new car registered. Market data also shows that in the absence of privileges, Perodua outperforms Proton, reflecting the former’s incentive to learn and experience of growth.

Observation of market data movements also suggests that trade barriers may be rendered ineffective as income increases over the years. It also suggests that repeating buyers of domestic cars is insignificant as the niche market may be saturated. However, more substantial statistical evidence is needed to determine if the niche market is saturated.

Despite the large overall market shares captured by Proton over the years, government’s policies to protect the infant industry have not been successful and they are not justified. With reference to the objectives of National Automotive Policy in Section 1.2 (p.2), Malaysian government’s policies have failed to achieve most of the stated objectives.

As the numbers of domestic cars decline since Asian Financial Crisis and the foreign cars outperform domestic cars in terms of the number of new cars registered since 2013, the protectionist measures have failed to promote a viable domestic car makers. Protectionist measures also have not been able to promote Malaysia as an automotive regional hub because trade barriers restrict foreign car makers’ access to domestic market while domestic car makers are unable to compete with locally assembled foreign cars. Low vehicle export volumes relative to Thailand suggest that government’s objective to promote high level of competitive vehicle exports is not achieved.

Although the number of Malay participation in the domestic automotive sector is far greater than that of non-Malays, this objective is achieved on the expense of car buyers in the form of higher car prices. This in turns, makes the objective to safeguard the interest of consumer futile.

44

Thailand’s case has shown that domestic economy may gains significantly from foreign direct investment without establishing a domestic brand name. Foreign direct investment allows domestic economy to benefit from employment creation, trade creation, and improvement in labour productivity that results from technology diffusion. Due to the technology, reputation, and market of the foreign car makers, the benefits enjoyed by Thailand economy is highly likely to be greater than the benefits the economy may enjoy if Thailand developed their own home brand.

Although there are conflicting views on the growth of South Korean car industry, studies generally agree that the industry benefit significantly from trade liberalization. Imports of technology know-how through joint venture project, foreign direct investment, and imports of intermediate goods such as car parts are proven necessary for an infant industry. However, lack of commitment to lift protectionist measures bounds to take its toll on the growth of infant industry and subsequently, imposes cost on the society in the form of higher prices.

Notes

1. Khazanah Nasional (meaning “National Treasury”) is an investment holding arm of the Malaysian government that actively participate in new investment and markets, acts as a trustee to the state’s commercial asset.

2. Perodua is entitled to such privilege because it is considered as a national car-maker.

3. Estimation is based on the number of Proton cars registered as government official vehicles, recorded by the Road and Transport Department, Malaysia.

4. The estimation is based on the average market price for 2003 and “on-the-road” price for 2010. The nominal market prices are deflated using GDP deflator (2000=100) of which the statistics are 111 and 202.4 in 2003 and 2010 respectively.

45

Chapter 3 Literature Review

3.1 Introduction

The conventional trade theory postulated by David Ricardo suggests that economies benefit from free trade if they are able to specialize in producing goods of which they have comparative advantage in producing. In a simple framework, an economy may produce the goods of which the opportunity cost incurred is lower than the opportunity cost of production of other goods. The opportunity cost difference between the two economies generates economic value of trade. The benefit however, may dissipate if transport cost is sufficiently large or trade barriers such as tariffs are imposed by an economy to its imports. Non-tariff barriers such as quotas dissipate the benefit of trade by creating artificial shortages in the domestic economy hence, increasing the price of the imported goods.

International trade allows economies to specialise in producing goods in which the formers have comparative advantage in producing and trade the goods with other economies. Therefore, trade provides more choices to consumers and allows increasing return to scale even when two economies that possess identical characteristics trade with one another (Markusen 1992, p. 63). The gain from trade is large if there are great potential for economies of scale and products are highly differentiated (Krugman & Obstfeld 1994, p. 132). However, it is also shown that if there exist increasing return to scales in the production of a good, free trade allows net increases in the world’s total production of the good while, an importer of the good and large economy tends to lose out in real income (Markusen 1992, p. 69). In short, if there is strong increasing return to scales, protectionism that leads to increase in monopoly power may favour an economy.

Therefore, within a protected environment, if an infant industry learned and technology diffusion took place, the infant industry will be able to produce differentiated products and subsequently, able to grow and compete with foreign products. However, there remains a question of how long the infant industry is to be protected. This is because

46 infant industry’s commitment to learn is influenced by the government’s commitment to lift protection (Ederington & McCalman 2011, p. 38).

Protectionism is justified if the total benefit offset total cost and that the protected industry grow and contribute to economy’s GDP growth eventually without protection. There remains an issue of impact of protectionism on income distribution that has not been discussed and studied substantially in many studies of protectionism. Income distribution is an important issue in Malaysia’s case because the protected industry is linked to individuals who are politically connected. The self-interest group lobby for government’s protectionist policies for their rent-seeking purposes.

This chapter reviews the early basic trade models, followed by the literature related to protectionism. The literature in protectionism provides an overview of how cost of protectionism is estimated and the size of the cost of protectionism in some economies. It also provides evidence for and against the argument for protection of infant industry. Subsequently, this chapter reviews literature in car market. This review provides an overview of the development in studies of car market over the time, reflecting economic growth and the roles of technological changes on car demand.

3.2 The Basic Trade Models

3.2.1 The Ricardian Model

On the assumptions that there is a single input labour, two economies and two goods produced, the early conventional trade theory, Ricardian Model shows that simple economies benefit from trade when there is specialization. Economies specializing in producing goods are able to produce goods at relatively lower opportunity cost that is, producing goods that the economies have comparative advantage. In this model, the opportunity cost, the value of the alternative goods forgone when additional unit of the goods produced is assumed to be constant. Specialization allows each of the economies to produce more of their specialized goods, export to each other and to reap economies

47 of scale. When more goods are included in the model, the concept of comparative advantage still applies.

The terms of trade is reflected by the opportunity cost difference between the two economies. This model also shows that transport cost and tariffs may dissipate the benefit of trade, by narrowing down the opportunity cost difference. As a result, there exist non- traded goods that an economy finds not profitable to export.

Deardorff (2004, p. 44) concludes that when Ricardian model is extended to include transport cost, it will affect production cost and subsequently influence comparative advantage. This study implies that variation in transport cost across countries may dissipate opportunity cost differences among countries. As a result, their comparative advantage may be altered. As such, trade within a region is more attractive than trading across and hence, to reap economies of scale and reap the benefit of comparative advantage, foreign direct investment or joint ventures or other entry modes not incurring high transport cost are more efficient channels to market firms’ products outside its home country.

The sweatshop labour argument put forth in the early 1980s suggest that low wage rate in developing nations may harm developed nations where wage rates are high. This is particularly seemingly true for the case of trade between the United States and China for example. The Ricardian model shows that this argument does not hold because as long as opportunity cost difference exist, trade is likely to increase wage rate of the exporting economy while the importing economy is able to enjoy the goods at a price lower than its domestic price.

In the case of seemingly unfair trade between the United States and China, it may be issues of asymmetric information and preference. Low wage rates in China reflects low productivity in the economy than in developed nations like the United States. The differences in wage rates between the two nations may be reflected by the differences in quality of products produced in the two economies. The demand for cheaper imported Chinese products is showing a change in preference that is, consumers’ willingness to tolerate lower quality products at a lower price.

48

A counter argument put forth by the developing nations argues that trade may lead to exploitation of resources of developing nations as their resources tend to be cheaper than the resources in the developed nations. Similarly, the Ricardian Model shows that an economy that has cheaper resources may still benefit from trade because it may cost the economy more to produce importable goods. The argument against trade based and exploitation of resources may be an issue of mispriced resources due to market failure and/or corruption.

This model is extended to include specific factors such as land and capital. These factors are specific to the production of certain goods, assuming that production of each of the two goods uses labour and one of the two specific factors, either land or capital but, not both. The extended model also called the Specific Factors Model shows that trade may have an effect on income distribution.

In the Specific Factor Model, mobile factor is labour combined with immobile factor, capital or land. The law of diminishing returns operates in the short run implying that the mobility of labour is constrained by the amount of land and capital available. Therefore, the production possibility frontier is concave to the point of origin while the frontier is a straight line in the Ricardian Model. When specific factors are incorporated into the model, it shows that an economy exports goods that can be produced at a relatively low opportunity cost that is, goods that are made of factors that the economy has relatively abundant. Therefore, if an economy has relatively more land than capital, more labour will be channelled to combine with land for production of a good while leaving less labour for production of other goods.

In the Specific Factor Model, it shows that trade may increase the price of a good, thus causing an increase in the derived demand for the labour and subsequently, the domestic wage rate. In short, the wage differential between labour in different sectors is influenced by the prices of the goods which in turns, influenced by the demand for an economy’s exports. Krugman & Obstfeld (1994) argues that pattern of trade is not to be influenced by the domestic economy’s desired income distribution because the latter is not specific to international trade. Income distribution is an objective domestic economy that shall be pursued internally. In addition, as long as trade increases the overall income level of the

49 economy, income may be redistributed either with or without the government’s interventionist policies.

The income distribution effect in this model due to immobility of factor tends to be temporary. In the long run, technology change or increase in capital may reduce the price of the exportable goods hence, reducing the wage rate of labour in the sector.

3.2.2 The Heckscher-Ohlin Model

The Heckscher-Ohlin Model suggests that pattern of trade can be influenced by the natural endowment of the economies. This model assumes that factors of production are not close substitute hence, production of a good may be labour-intensive but, land- intensive for the other good. This model shows that it is relatively cheaper for an economy to produce goods that can be produced using the factor that the economy is relatively well endowed. Therefore, prices of the goods produced using scarce factor will be relatively more costly than the goods produced using the abundant factor. Consequently, owners of abundant factors will benefit from trade while the owners of scarce factors lose out.

The underlying concept of this model is consistent with the Ricardian Model. In the Heckscher-Ohlin Model, the definition of endowment is defined as the ratio of a factor to another factor for example, the ratio of land to labour. An economy is considered well- endowed with land if the amount of land is relatively large compared to the amount of labour. In other words, endowment of an economy may be a source of comparative advantage.

Deardorff (2006, p. 4) points out that this model is however, very sensitive to the numbers of variable included for example, cost of trade. That is, a small change in the cost may cause changes in the trade flow from an economy to another, thus causing some changes in economies’ specialization.

50

3.2.3 The Standard Trade Model

The Standard Trade Model takes into consideration the imperfect substitution of factors of production hence, an economy has a production possibility frontier that is concave to the point of origin. The economy’s choice of good to produce and to import are influenced by the relative price of the good that is, the good that has the highest value or the lowest cost.

Refer to Exhibit 3.1, the production possibility frontier of domestic economy is PP. The isovalue, II is the relative price of the goods that is, the ratio of agriculture goods price over the manufactured goods price. At the given prices of the manufactured goods and agriculture goods, an efficient combination of goods produced is at point Q where the isovalue is tangential to the production possibility frontier. Assuming the society is able to rank their preference, their satisfaction is maximised at point E, which is an infeasible set as it is lying outside the production possibility frontier, if there is no trade. Trade allows the economy to produce at Q and export some of the agriculture goods in exchange for more manufactured goods desired to be at point E.

Figure 3.1 Production and consumption in the Standard Trade Model

Manufactured Goods

I E P ● ● Q

I

P Agriculture Goods Source: Krugman & Obstfeld 1994.

51

All the basic trade models conclude that economies benefit from trade due to the differences in the opportunity cost of producing the goods within the economy and across the economies. However, all of the models implicitly assume that the economies are small therefore, are unable to influence other economies.

Trade allows economies to channel more of their resources to produce more of some goods and export the goods in exchange for imports that the domestic economies may have to produce at a relatively higher price if produced the goods themselves. However, opportunity cost is no longer the only factor for trade in this trade model except in the Ricardian model. This is because increasing opportunity cost as reflected by concavity of the production possibility frontier in Figure 3.1 may be equalized if there is free trade (Deardorff 2005, p. 7). Assuming plants are homogeneous in the perfectly competitive industry, free trade allows domestic economy to consume other good at a lower price than the domestic cost of producing the good.

3.3 The Benefits of Free Trade

The direct benefit of trade is generated through specialization and exchange of the domestic goods for the foreign goods that the domestic economy may have to produce at a higher cost than foreign economy. While the direct benefit that is often private to firms, technology investment generates positive externalities, the “spillover” effect on the society (Keller 2004, p. 753). In the conventional framework, the benefit of trade is often represented by the gain in consumer surplus, offsetting the loss of producer surplus and tariff revenue that could have been collected by the government. Using this conventional approach, the estimation of static gain from trade after liberalization of trade is simulated under the scenario of tariff being removed completely. The estimation is often found to be a very small percentage of the country’s GDP (Panagariya 2002, p. 175). Studies proposed that the cost of protectionism has been understated because spillover effects are often not taken into account.

52

3.3.1 Economies of scale

Harris (1984) shows that freer trade allows firms to achieve economies of scale due to larger foreign market and competitive pressure from foreign countries. Irwin (2009, p. 42) however argued that the importance of economies of scale is overstated. Studies show that the gain from economies of scale at plant level for manufacturing firms is relatively small compared to the size of the market. Subsequently, the average cost changes are insignificant (Ibid, p. 42).

There are also negative secondary effects of protectionism that have been neglected. These effects are the diminishing market power of the domestic firms and efficiency due to greater competition from abroad. These effects can be increasingly overwhelming as products are differentiated and technology changes at a relatively fast pace. Failure of the domestic firms to acquire the knowledge may result in the exit of the domestic firms from the industry.

Freer trade may not benefit an economy if there is an inverse relationship between diversity and production scale. While diversity in production may increase the variety of choices for consumers, it may reduce the market share of firms to the extent that firms are unable to produce at larger scale to reduce their average cost. Consequently, competition from foreign producers may lead to increases in the average production cost in the domestic economy (Helpman & Krugman 1985, p. 188). This implies that the domestic market size that domestic firms have immediate access to is essentially a factor that may influence the firms’ chance to reap economies of scale. Consumers may have benefit from lower prices but, forgoing choices.

Under normal circumstances where transport cost limits trade, larger countries, defined as countries that have larger GNP, tend to have the benefit of scales and of greater variety of choice than smaller countries. This is because when there is factor mobility, there is an incentive for factor to migrate to the larger countries that provide a larger market for the non-traded goods. Subsequently, imperfectly competitive industries tend to concentrate in larger countries than in smaller countries (Ibid, p. 209). Therefore,

53 depending on the size of the economy, economies of scale may not be significant for small economies.

In the case of the United States-Canada auto pact, Wonnacott & Wonnacott (1980, p. 3) suggest that Canada being a smaller economy than the United States, may reap economies of scale if the former is able to specialize in producing a smaller range of cars and export them to the United States in exchange for greater variety. Although differentiated, the cars produced in the two economies are assumed to embody insignificant difference in the level of their technology so that exchange may take place. This point will be discussed further in the subsequent section.

In conclusion, economies of scale can be reaped if the domestic market is sufficiently large or if domestic firms are able to tap into foreign markets. Trade barriers on one hand may allow domestic firms to reap economies of scale; consumers may have to forego choices. In the case when government does not have commitment to lift trade barriers, lack of competition may lead to inefficiency. Consequently, the demand for foreign cars may increase as domestic income increases and leads to greater cost of protectionism. Domestic firms are unable to reap economies of scale in the long run when inefficiency exacerbates.

3.3.2 Variety of choices

Intuitively, trade gives the benefit of larger variety of choices for consumers but, this benefit is rather difficult to quantify. Romer (1994) suggests that while tariff may encourage foreign firms’ direct investment to avoid tariff, tariff may cause import substitution policies to be effective and hence, reducing the amount of new goods that could have been made available to consumers.

The conventional method often does not estimate the cost associated with the disappearance of new goods because it is difficult to quantify utility of the goods that “do not exist”. This implies that there exists certain benefit of free trade that is overlooked. Irwin (2009, pp. 43-33) suggests such secondary benefit can be in the form of availability

54 of specialized final goods and producer intermediate goods that are excluded in the market due to the existence of high entry cost.

Romer (1994) studies the effects of trade barriers on disappearance of goods by revisiting the general equilibrium analysis on new goods and studies the effect using partial equilibrium analysis in the attempts to identify association of trade barriers with decreases in imported goods and welfare loss. The study finds that if trade barriers do not have an effect on the sets of goods available to an economy, income falls by 1 per cent when tariff is 10 per cent. However, if the set of goods change as a result of economy’s response to tariff, income falls about ten times the fall without change in goods available (Ibid, p. 34). When the tariff rate is changed, the study consistently finds that income fall is greater under the scenario of goods disappear than under the scenario of no change in goods available.

This implies that if there is free trade, the benefit of enjoying larger variety of goods, measured as the percentage increases in GDP, can be as high as ten times more than the benefit when there are trade barriers.

Corden (1997, p. 113) highlights the possibility of consumers desire for homogenous goods at minimum prices in a fragmented market selling seemingly differentiated products due to freedom of entry to the industry. In the case of cars, the products attributes are influenced by technology and thus, consumers may not know what to expect from new technology. At aggregate level, consumers’ desire for more advanced technology can be indicated by the short run profit. So long as the cars carrying new advanced attributes are priced more than the marginal cost, the demand for unknown future technology exist.

However, in a distorted market coupled by unreliable public transport, the demand for cars may be merely for the purpose of transport while there are demand for cars for status, aesthetic, as well as for the satisfaction of exploring advanced technological attributes of the cars. As such, a niche market for less technologically advanced cars may arise. The demand for cars will be discussed in Section 4.

55

Although there is a lack of studies that may suggest the size of the market being a factor that influences the variety of choices in the Malaysian car market, industry players in 2013 anticipated increases in the number of imported high-end foreign cars if excise duties are removed as the rates range from 60 to 105 per cent. The Malaysian Automotive Association car sales statistics show that the overall number of car makes available in the domestic car market increased from 17 in year 2002 to about 50 in year 2016 while Proton’s market shares shrank from 49.4 per cent to 12.5 per cent during the same period. This may imply that in Malaysia’s small car market, preference for diversity may be achieved at the expense of economies of scale.

3.3.3 Competition and efficiency

Free trade leads to a greater variety of choice therefore, is likely to lead to greater consumers’ sensitiveness to prices change thus, exerting pressure for efficiency (Krugman 1994, p. 79). Firms in less developed countries are unable to achieve allocative efficiency and X-efficiency due to trade barriers (Bergsman 1988, p. 420). Corden (1997, p. 112) suggests that X-inefficiency is due to product fragmentation that arises because tariff generates monopoly profits that attract new entrants and increases the number of producers. The resource loss as a result of product fragmentation is a cost borne by firms. In the case of tariff imposed on the products of foreign firms, the loss is not restricted to loss of revenue of the foreign firms but, also tax payment that could have been gained by the domestic country. In addition, the loss can potentially be the higher wages that could have been earned by employees when monopoly profit is made (Ibid, p. 112).

However, Noorzoy (1979, p. 53) shows that dynamic efficiency can be achieved when tariff is reduced because of the positive income effect and substitution effect when the importable goods are normal goods. Tariff cuts coupled by increase in income lead to fall in the demand for and supply of the domestic goods in the protected industry. As a result, there is incentive for domestic firms to improve in order to compete. However, protectionism may seem desirable for an economy where domestic goods are unable to compete with foreign goods if the domestic firm has no incentive to learn and government has no incentive to lift the control.

56

Irwin (2009, p. 47) argues that efficiency may be improved when there is freer trade. This is because when the foreign capital goods that embody advanced technology are imported, the productivity of the industry can be stimulated. Therefore, the benefit can be substantial when the spillover effects of the new technology are taken into account.

In addition to deterring technology transfer, trade barriers promote directly unproductive, profit-seeking (DUP) activities (Bhagwati 1982, p. 989). The DUP activities are defined as “activities that are directly unproductive, that yield pecuniary returns but do not produce goods or services that are enter a utility directly or indirectly via increased production or availability to the economy of goods that enter a utility function” (Ibid, p. 989). Quantitative restriction such as import license effectively limits the amount of imports, may become a priced commodity due to rent-seeking behaviour (Krueger 1974, p. 291). The price of commodity is contributed by the resources that are used for the competition to obtain the commodity, the shortages in the imported goods due to the restrain, and the allocation of resources to influence the outcome of government’s action such as bribes (Ibid, p. 292; Corden 1997, p. 127).

The conventional measure of the benefit of freer trade may be able to capture the gain from removal of import license but, unable to measure the loss of output due to resources channelled for obtaining the license and for lobbying for protection when there is lack of transparency in the number of licences issued. However, Corden (1997, p. 127) suggests that such loss is no longer the deadweight loss triangle but the part of consumer surplus loss that could have been taken up by the increase in producer surplus. Bhagwati (1982, p. 994) argues that in a first-best situation, misallocation of resources due to DUP may impose a cost on the society. Although quantitative restriction often leads to distortion, DUP may not necessarily lead to welfare loss in a second-best situation.

Costinot (2008, p. 23) concludes that if a labour’s current income is an important factor for protectionism, less productive labour is more likely to be protectionist. This implies that if there exist self-interest groups that lobby for protection and if there is a lack of commitment in the government to lift the protectionist measures, moral hazard may arise.

57

If trade barriers are erected to protect the small group of the society, the benefit of this group of the society will be gained on the expense of other groups of a different industry and finally the consumers. Due to the gain, the self-interest group has the tendency to lobby for the government’s protection (Weidenbaum 1983, p. 783).

Approval Permits (APs) that are used as protectionist tools and as a tool to redistribute income in favour of the Malays hinder competition in the Malaysian car industry. The Malay Vehicle Importers and Trade Association continues to lobby for issuance of permits to its members for self-interest on the expense of efficiency of the industry and on car buyers.

3.3.4 Technology transfer and innovation

Due to the dynamic nature of technology, there has been lack of studies that emphasizes technology’s role in international trade. Krugman (1990, p. 143) uses a simple model of two economies, assuming perfectly competitive industries, the wage rates are determined by productivity of the labour therefore prices of the goods. The study shows that technology transfer from developed economies to developing economies makes new goods old thus, shifting the derived demand for labour from developed economies to developing economies. Subsequently, wage differential between the developed and developing economies narrow as the wage of the developing economies increase, reflecting improvement in productivity.

In this model, successful technology transfer benefits the developing economies. Under a static setting, the terms of trade will deteriorate against developed economies. However, due to their advantage in knowledge, innovation allows the developed economies to continue to export new products and earn monopolistic income. Therefore, under a dynamic setting, technology transfer and innovation lead to increases in the world output after trade. So long as the developed economies benefit from temporary monopolistic power they can “renew” through induce innovation, the developed economies may continue to “export” technology to the developing economies (Ibid, p. 173).

58

The technology changes experienced by the developed and developing economies also show that history matters. As the developed economies invested in technology and education before the developing economies began doing so, the developed economies have written the history of technology and pioneered various disciplines of education all of which contribute to the comparative advantage of the developed economies in technology. This also lead to the “narrow moving band” argument that a developing economy may be able to shield itself from foreign competition while accelerating its speed of acquiring new technology to undercut the foreign competitors when the former is “mature”.

Corden (1997, p. 113, 140) argues that loss of profit due to new entrants or relatively large number of competitors to an industry may be unavoidable and may be seen as a cost of investment because there are needs for firms to enter the industry early to establish themselves through learning-by-doing. Irreversible economies or dynamic internal economies can be achieved when the time involved is sufficient to allow firms to improve their productivity through gaining of experience in production. As there are continuously increasing pressures for pollution control and , time available for infant industry to learn-by-doing may be short and the process of inviting competitors and learn from them is unavoidable.

The cost of protection borne by the society, is in the form of higher relative price of the domestic cars using old technology while new technology is available may be high.

3.4 Protectionism - A double edged sword

There are many arguments for protectionism but, the infant industry argument will be the focus of discussion in this research. The initial argument for the protection of infant industry is often based on the assumption that there is perfect competition in the industry. Therefore, firms do not have the incentives to acquire knowledge that is costly in terms of time and money.

59

The proponents of infant industry suggest that the argument is justified on the ground that temporary shield from competition enables the infant industry as a late-comer, to acquire the knowledge and thus, become more cost efficient as the firms learn and are able to compete with the established foreign firms at the level. This can be traced as far back as in Hamilton (1791) according to Baldwin (1969, p. 296). The study also says that Meade (1955, p. 256) however, believes that high production cost facing the infant industry does not justify protection as long as there is positive net present value in the firms’ investment (Ibid, p. 297).

3.4.1 Infant industry growth

Dasgupta & Stiglitz (1988, p. 266) concludes that in an industry that demonstrates myopic behaviour, small entry cost and powerful knowledge may give rise to a monopoly. Intervention by the government provides chances to make profits in the future and the society’s welfare is greater if domestic producers are able to grow than it is if there is free trade that may threaten the growth of the infant industry. Protectionism is therefore justifiable if the protected infant industry makes losses.

In a study of Pakistani infant industries, investigation was carried to find out if they have learned during the period of protectionism using the estimated coefficients of learning: the cumulated gross investment and cumulated output (Kemal 1979, p. 3). Based on the assumptions that increasing investment implies greater production and thus, income level, and greater output stimulates learning, the coefficients are used to measure infant industries’ growth and their potential economies of scale. The results of the study suggest that the infant industries have learned during the period of protection. Although the study suggests that protection is necessary for firms to learn and for labour to acquire necessary skills and that such cost may be offset by the gain in the future, the study is insufficient to make a strong proposition for protectionism.

By studying the relative growth of Turkish infant industries, measured in terms of the change in cost per unit of output, Krueger & Tuncer (1982, p. 1149) finds that statistically, there is no evidence of growth among the infant industries, assuming that only the growth

60 of the domestic total factor productivity influences the change in comparative advantage. The methodology and results of the study are unable to make conclusive remark about the infant industry growth due to the single condition of which the study is built upon. While the study uses input per unit of output as indicator for all the infant industries in is easy to measure and to understand, the study does not take into account the market structure and the age of the infant industries. The Turkish government protects all infant industries with the use of embargo, without targeting specific industries. As such, while the study believes that the lack of growth in the infant industries is due to inappropriate trade incentives, there is a possibility of lack of government’s commitment to lift the protectionist measures.

Similar approach is used in Lee (1997, p. 1272-8) in the study of the South Korean infant industries. However, this study identifies the trend of effective rate of protection of various industries over the years from 1970 to 1990. The results show that the South Korean infant industries have grown and matured during the period of protection. The positive performance of the industries is partly attributable to the trade policies that is, restricting imports while promoting exports. Other factors contributing to favourable outcome are competition among the domestic firms and government’s commitment to liberalize the market.

In the Brazilian microcomputer industry, the government’s protection of the domestic has led to high domestic price of the products and arisen of black market. Using hedonic prices as the proxy for growth, the study finds that although the speed of technology advancement in the domestic industry is comparable to the foreign industry, the prices of the domestic microcomputers are higher than the prices of the foreign microcomputers, implying the lack of efficiency in the domestic industry. Luzio & Greenstein (1995, p. 631-2) concludes that the cost of protectionism measured as the loss of industry output due to trade restriction is a third of the total expenditure on domestic microcomputers. The study also observes that the domestic firms begin to compete and to achieve efficiency when the government promises to lift the protection. This implies that the government’s commitment to promote competition and efficiency is essential for infant industry’s growth.

61

Das (1995, p. 123) concludes that the age of infant industry has a strong positive correlation with the growth of the industry using the Jovanovic growth model where the size of an industry is specified as a function of its lag size and its age. The study’s conclusion also suggests that consumers’ awareness has direct relationship with the firms’ age. This study is at best able to suggest that infant industry grows over time but, unable to provide more meaningful and substantial implication on the infant industry’s behaviour under protection or government policies. The study also has not taken into consideration foreign competitors’ prices.

The study also suggests that infant industry’s survival may be influenced by consumers’ learning of the new product and reputation of the firms (Ibid, p. 123). Although the study has not taken into account the availability of foreign products in the economy, the results may suggest that consumers’ experience in using the products may eventually contribute to the survival of the infant industry by putting pressure on the infant industry to learn despite being protected.

A study of the Indonesian car infant industry using decomposition methodology in which total factor productivity is decomposed into plants’ improvement and changes in output, the results show that productivity growth in the car industry has been appalling (Okamoto & Sjӧholm 2000, pp. 66-67). The negative growth of sub-sectors such as assemblers and car body makers were found unable to achieve growth. The study attributes the negative growth to the take-overs of the productive firms by the less productive firms in the industry. It is also pointed out that there is a failure of technology diffusion due to the firms’ inability to acquire the knowledge although technology is available, and there is a lack of incentive to achieve efficiency due to protectionism.

It is worth to investigate why more productive firms opted out of the industry and less productive firms enter the industry. If the more productive firms are the foreign firms that enter the industry earlier before the establishment of the new domestic firms, the formers’ decision to exit the industry may reflect the environment being less conducive for them to continue operation. The factors that contribute to conducive environment may include the availability of labour to support the need of production, government policies, and the availability of infrastructure.

62

The study highlights the importance of foreign firms’ roles in infant industries as a source of knowledge and the foreign firms’ production may have spillover effects on related industries, implying the possibility of reaping the benefit of having foreign firms operating in the domestic economy through competition.

All the studies of infant industry growth described above suggest that infant industries do grow to a certain extent but, these studies are not sufficient neither to argument for protectionism nor against protectionism because of the following reasons:

(a) Time matters when there are possible cost and benefit. Even when infant industries do learn, the duration taken to learn is essential because of the industries’ linkages to other industries and the cost implication on the consumers;

(b) The level of efficiency of the infant industries has implication of resources allocation in the industry and thus, the cost of resources imposed on other related industries;

(c) The incentive to acquire technology when it is available has strong implication on the welfare of the consumers. The cost of failure to acquire technology is higher if the industry is at the higher end of the continuum of production technology intensity; and

(d) Protectionism opens up opportunity for lobby by the self-interest group and subsequently leads to further misallocation of resources.

A study of the Canada-United States auto pact in 1965 concludes that rationalization does not lead to as much improvement in efficiency as it might have been if there was liberalization. This is because the Canadian car market is relatively small, making it difficult to exploit economies of scale and to exploit the benefit of specialization. In addition, although firms have the incentive to minimise cost, it is believed that more liberalized trade will exert more pressure on the then, oligopolistic firms to be more efficient (Fuss & Waverman 1986, p. 29).

63

In the study of protection of the United States’ steel rail industry, Head (1994, p. 162) concludes that the infant industry grew while under protection. Although the gains on producer surplus were small and the gains on consumer surplus materialized shortly before the industry was liberalized, there was an overall gain on welfare as protection led to long run decreases in the domestic prices. The positive outcome of the protective policy was attributable to the comparative advantage of the country, the growing domestic market size, and the pressure that protection was temporary.

Irwin (1998, p. 30)’s study of tinplate industry in the United States during the years from 1891 to 1900, concludes that the industry could have performed better without tariff. This is because there were no substantial static and dynamic scale of economies and technology know-how was readily available. As the factor costs were high, free trade would have granted the industry the advantage of costless technology diffusion and thus, cheaper production cost.

In another study of the United States iron industry during the years from 1867 to 1889, Irwin (2000, p. 25) concludes that the iron industry would have suffered a 15 per cent fall in the domestic output accompanied by the increase in the imported iron market of up to 23 per cent if tariff was not put in place. However, the domestic iron industry would still survive the competition (Ibid, p. 26).

The experience of infant industries in the United States can be summarised as follows:

(a) Infant industries need time to grow thus, protectionism may enable them to grow at low cost;

(b) Decreases in the infant industries’ output in a liberalized market may be unavoidable because the industries’ low productivity at infancy stage;

(c) Protectionism is costly if the technology level in domestic economy is lower than the technology level in the foreign economies because protectionism makes technology diffusion more costly in terms of both time and money; and

64

(d) Due to the gaps between the technology in domestic economy and in foreign economy, efficiency can be improved when there is freer trade even when the domestic firms have the incentive to minimise cost.

3.4.2 Specialization in creation

Krugman (1990, p. 107) suggests that due to externalities, industry may be producing at a low level of output while the economy may be demonstrating the characteristics for specialization. Temporary protection therefore, allows the industry to reap dynamic economies of scale as the industry moves towards the right side of its learning curve.

Studies show that pattern of specialization can be created and preserved (Krugman 1994, p. 112). The “narrow moving band” argument suggests the possibility of temporary protection of an infant industry such as the protection of the Japanese car industry that allows a permanent shift of comparative advantage by allowing sufficient time for the infant industry to improve its productivity (Ibid, p. 113). The essence is that the infant industry acquires the knowledge and improves its productivity at a fast pace to catch up with the foreign industry.

The “narrow moving band” argument is applicable to the case of the Korean car maker, Hyundai. After the founding of Hyundai the oldest and largest Korean car maker in 1968, the Korean government imposed an embargo on foreign cars for 25 years. During this period, Hyundai initially adopted Ford’s technology through learning-by-doing before acquiring Mitsubishi and Honda’s technology in the mid of 1970s. Within five years, the local content of the Korean cars has grown up to 60 per cent. By 1979, the industry was able to achieve more than 90 per cent local content (Greenbaum 2002, p. 4).

Although the industry was liberalized in the 1980s, other protectionist tools such as 60 per cent tariff on foreign cars, administrative procedures and regulation of safety standards were still in place. The Korean car market subsequently saw a few steps taken to liberalization in the mid of 1990s and in the early 2000s (Ibid, p. 6). However, the Korean car market remained relatively closed to foreign competitors.

65

In year 2012 (see Table 3.2), Hyundai Kia is the third largest car maker in the world (Maps of World 2013). Although Hyundai Kia’s position in the world car market may be achieved on the expense of the other car makers of the countries that adopt free economy, the former’s performance shows that the Korean car infant industry has successfully adopted new technology from their counterparts and subsequently, compete with their “teachers”.

Table 3.1 Top 10 car manufacturers: Year 2017

Rank Manufacturer (Country) Number of cars manufactured (Global sales in million units)

1 Volkswagen (Germany) 10.41 2 Toyota (Japan) 10.16 3 Hyundai Kia (South Korea) 7.28 4 (United States) 6.88 5 Honda (Japan) 5.86 6 Renault-Nissan (Japan) 10.12 7 Ford (United States) 6.25 8 PSA Peugeot Citroёn (France) 4.16 9 Suzuki (Japan) 8.15 10 Daimler (Germany) 2.67

Source: Statista 2018.

Similarly, according to Nishiwaki (2007, p. 5), the Japanese car makers were protected by their government during the period from 1955 to 1965 in which an embargo was enforced on foreign cars for the generally public’s use. During that time, Japanese cars were not competitive in terms of their prices and quality. Approximately ten years under protection, the Japanese car makers have made it to the world top ten exporters.

66

3.4.3 Technology diffusion

Studies found that foreign technology may contribute to productivity growth of as high as 90 per cent with the major channels being international trade and foreign direct investment (Keller 2004, p. 776). A common argument for protectionism of infant industry assumes that infant industry learns and eventually adopts the new technology and the adoption of new technology is exogeneous. Protectionism may contribute to external economies of scales of industries if the learning process generates spillovers across the industries regardless of the development state of the domestic economy (Melitz 2005, p. 179). Therefore, domestic economy may suffer dynamic losses due to specialization in a free trade setting (Sauré 2007, p. 104).

Ederington & McCalman (2011) assumes endogeneity of the state of technology and size of industry following Melitz (2003) and Bernard et al. (2003), suggests that endogeneity in industry size that is, barriers of entry and exit over a period of time have some implication on the firms’ productivity and technology diffusion. The study also suggests that while protectionism reduces the level of foreign competition, the former increases the level of domestic competition. Therefore, although the cost of adopting new technology may decline over the time, domestic firms may not adopt the technology because of the insignificant difference between the present cost of adopting new technology and the expected future profit that potentially eroded if the entry cost is not substantial.

It is also found that protectionism has no significant effect on technology diffusion in a contestable market but, has effect in a stable market that is, a market characterised by high entry cost (Ederington & McCalman 2011, p. 38, Baldwin, 1969, p. 298). That is to say infant industry may not learn during the period under protection if there is no pressure or incentive to learn.

The speed and probability of technology transfer will improve if there is commitment to growth and if there is imposition of termination date (Ibid, p.38; Leahy & Neary 1999, p. 43) implying that protection should be temporary (Melitz 2005, p. 178). However, there have been lack of studies that recommend an optimum duration for protection.

67

While the study does not specify the optimum duration of protection, Baldwin (1969, p. 298) suggests that the crucial time frame within which an infant industry is able to take advantage of protection is the “technologically fixed time lag.” It is defined as the time between the introduction of new technology and the time new firms are able to enter the industry at negligible cost. The speed of technology diffusion may then be influenced by the level of expected gain during the period of protection before competitors enter the industry.

Technology diffusion is also found to be inversely related to the distant between the location of origin of the technology and the location of beneficiary via trade and foreign direct investment (Keller 2001, p. 25-26). Therefore, in any industry where technology diffusion takes place as a result of close proximity, product differentiation may not be significant and competition is likely to be stiff. For technology diffusion to take place in the car industry where product differentiation is an important factor for competitive edge, it will benefit a developing economy to engage in joint venture with firms in developed economies.

Grossman & Horn (1987, p. 22)’s study of asymmetric information as barriers to entry by an infant industry finds that temporary tariff does not have a positive effect on the incentive for firms to provide quality product although the firms are facing competition from well-established foreign firms. This study explains very well the reason why Malaysia’s domestic car maker’s market share continues to shrink (Lee 2014, p. 4) and the firm is reported to seek up to RM3 billion funds to develop their new car models (Sidhu 2014, p. 1).

The domestic car maker suffered from possibly, a temporary competitive disadvantage of lack of reputation. Tariff barriers enabled Proton to gobble up to about 47 per cent of the market share in 1986. The domestic car might be able to take advantage of car buyers’ lack of information as they were unable to observe all the cars’ attributes initially after the new domestic cars were introduced to the market. If technology diffusion has been successful in the domestic economy, experience after using a domestic car and perhaps words-of-mouth might have over the time, help the domestic car maker to grow. However, domestic car buyers’ experience and words-of-mouth might have led car buyers

68 to recognition of moral hazard in car makers’ choice of quality and adverse selection arises from availability of different production technology. Consequently, unsuccessful learning-by-doing of the domestic car maker in Malaysia’s case, is likely to lead to lower domestic welfare, consistent with the conclusion in Grossman & Horn (1987, p. 22).

Keller (1996, p. 202) attributes the success of technology diffusion to human capital that is bounded by mobility. This can be explained by the brain drain problem facing Malaysia since the 1980s and the lack of government’s commitment to lift protection that leads to complacency in the protected labour.

Although Proton has been losing its market shares over the years, the ailing car maker remains the top three biggest players in the industry. This is nothing more than ‘home- bias in trade’ puzzle of which Obstfeld & Rogoff (2000, p. 4) finds attributable to trade costs. On the other hand, the market share for foreign cars as a whole has increased tremendously but, the increase of individual brands of foreign cars are slow. Increase in the demand for foreign cars in Malaysia generally are attributable to increases in the market shares of a small number of prominent foreign car makes.

Increases in income may contribute to multilateral imports of various brands of foreign cars at small volume as Havemen & Hummels (2004, p. 210) shows. Failure of domestic firms to acquire technology know-how during the period of protection will portray less favourable image on domestic cars hence, unable to compete with foreign cars. As income increases, the demand for foreign cars that are more highly differentiated will increase.

3.5 The Costs of Protectionism

The cost of protectionism is strongly linked to economic policy and the former is often large (Anderson & van Wincoop 2004, p. 691). The cost is however, often found small because studies do not take into account the cost imposed on the foreign firms in the form of loss of exports. The cost is expected to be larger if the protected economy is large because of its influence over the price and the loss in terms of the quantity. When the

69 foreign deadweight loss is taken into account, its size can be as large as 138 per cent the size of the United States’ deadweight loss, reflecting the pressure exerted by the trade barriers erected by a large economy on its trading partner (Feenstra 1992, p. 163). In an imperfectly competitive market, the cost of protectionism is higher if trade barriers reduce the variety of goods made available by the foreign producers (Ibid, p. 160).

In other cases such as voluntary export agreement on cars between the United States and Japan, trade barriers benefit the foreign producers (Coughlin, Chrystal & Wood 1988, p. 6). This can be due to the diversity of products attributes and thus, products’ marketability in other regions. A fall in the Japanese’s exports to the United States might be absorbed by the increase in exports to other countries. This study shows that trade barriers are unable to harm exporting firms if the exports possess characteristics that are generally acceptable in other economies.

Feenstra (1985, p. 67) studies the same case concludes that due to the trade barriers, the benefit of foreign currency depreciation has not been passed on to importers of the foreign cars through lower prices. In addition, there has been upgrading of quality of the imported cars. As a result, although increases in the availability of the quality cars benefit those buyers who desire better quality cars, cost is imposed on those buyers who desire simple models that may be cheaper. This study also shows that trade barriers dissipate the positive effect of depreciation of foreign currency.

Protection of the steel industry in the United States in the 1950s provides evidence against protection because of adverse effects of the government policies. The American firms were unable to reap economies of scales, dependent on their low-cost reserves. Consequently, coupled by Japanese firms’ acquisition of new and efficient technology, lower transport cost enabled Japanese firms to outperform American firms (Canto, Eastin & Laffer 1982, p. 44).

The long-term effect of protectionism tend to be defeating the purpose of protection. This is because protection leads to higher domestic price thus, there are incentives for foreign firms that are able to produce at a lower cost to enter the market after protection. In

70

Malaysia, foreign car makers enter the market using different strategies, mostly joint venture.

3.6 The Cost of Protectionism - Studies at firms level

Another area of study in international economics at plant level is the study of firms’ productivity in relation to international trade. International trade theory suggests that an economy is able to specialize and reap economies of scale, and finally, exports the output in exchange for output that the economy has no comparative advantage in producing. Following the availability of micro data, studies at firm level have become feasible. A pioneer study at producer level find that foreign markets may be less significant compared to the domestic markets for firms that have high productivity and efficiency. In other words, high productivity and efficiency within firms do not lead to higher exports Bernard et al. (2003, p. 1268).

The study finds that exporting firms have greater advantage in productivity than the non- exporting firms. Thus, it is concluded that differences in the underlying efficiency may be contributing to the coexistence of the exporting and non-exporting firms within and industry. Firms that are more efficient have greater chance of exporting their products than the less efficient firms and that with greater efficiency, the efficient firms have greater leverage in setting their prices and hence competing in the domestic market (Ibid, p. 1278).

Melitz (2003, p. 1716) shows that apart from the common mechanism in which domestic firms may be driven out of the market by more productive foreign competitors, domestic firms can be driven out through entry cost. Since high entry cost is associated with high expected future return hence, wages are bid up when the number of new entrants increases and subsequently, the least efficient firms are forced to exit.

Melitz & Ottaviano (2008, p. 312) presents mathematical model that shows stiffer competition exists in larger market thus, resulting in overall higher productivity and lower average mark-ups. The model also shows that trade may reduce the cost of imperfect

71 competition arises from trade barriers and trade may erode short run gain from trade barriers.

In a study of the Chilean manufacturing sector, Pavcnik (2002, p. 271) finds that trade leads to improvement of productivity in the import-competing industry and that innovation is positively correlated to competition and market share. Although it is found that the export-oriented firms are the least affected by freer trade in terms of firms’ survival and displaced workers, the study is unable to conclude that exporting firms are more efficient and more productive because the overall size of the exporting industry is relatively smaller than the non-traded goods and import-competing sectors. However, there is a causal link from productivity to exports but, not the other way round (Greenaway & Kneller 2007, p. 157; Wagner 2008, p. 178; Bombardini, Kurz & Morrow 2012, p. 610; Minondo 2010, p. 285).

A study of Malaysia’s manufacturing sector shows productivity of the exporters is significantly greater than that of the non-exporter and that exporters have relatively large percentage of foreign owners while the non-exporters are mostly locally owned (Lee 2011, p. 290). However, there is no causal relationship found between productivity and exports while freer trade reduces the probability of exports by firms that do not acquire technology. There is no statistical evidence of freer trade’s relation with probability of exports by firms that acquire technology (Ibid, p. 291). The study suggests that firms’ productivity is driven by capital intensity and human capital in the country.

While most of the a priori studies focus on the effect of tariffs on final output, Amiti & Konings (2007, p. 1621) suggests that the effect of tariff on intermediate goods may be greater than its effect on the final goods. This is because the foreign technology embodied in the intermediate goods allows improvement of productivity of the firms that import foreign intermediate goods. Keller (2004, p. 777) however states that although statistical evidence supports the view that imports contribute to technology diffusion, more substantial evidence and in-depth studies are essential.

The studies above may be summed up as follows:

72

(a) The cost of trade imposed on the domestic firms both exporting and non-exporting may be the cost associated with reshuffling of resources such as displacement of workers, and decreases in profit;

(b) Competition leads to improvement in firms’ productivity and hence, lower average mark-ups;

(c) Trade will increase welfare gain although the former may lead to reversal of short term gain that is due to trade barriers;

(d) Trade does not necessarily lead to expansion of the exporting sector but, productivity increases attribute to firms’ success in the foreign market;

(e) Trade may lead to higher probability of closure or downsizing; and

(f) Trade benefits economies that have lower level of technology and knowledge than foreign economies.

Studies however, did not further suggest factors that may contribute to the greater risk of closure apart from lack of efficiency. Sunk cost and characteristics of firms are cited to be the factors attributable to firms’ participation in the foreign market (Greenaway & Kneller 2007, p. 135). Participation in the international market exposes firms to various types of external risk such as exchange rate, political changes, administrative procedures and damages and lost in transport, and internal risk such as financial leverage. Consequently, productivity gain from trade may not necessarily lead to higher exports.

At aggregate level, Kunst & Marin (1989, p. 703) uses time series analysis for Austrian economy which concludes that exports and freer trade do not lead to productivity improvement. However, the study finds that productivity improvement leads to increases in the export market shares. Although there may be a need to have more studies to support this findings, the result nevertheless, shows consistency in the flow of the factor, productivity improvement to increases in the exports, both at disaggregate and at aggregate level.

73

Most studies measure the cost of protectionism from the perspective of success in exports. In the case of arguments for infant industry and job creation, the cost of protectionism may be more relevant if measured from firm’s efficiency perspective. Hence, this study attempt to compare efficiency level between protected firm and unprotected foreign firm that has been operating within the domestic industry.

3.7 The Impact of Protectionism of Domestic Car Maker

There is a large variety of protectionist tools introduced in developing countries as well as developed countries that have been favouring free trade. The traditional tools are tariffs and non-tariffs such as quota, administrative procedure, and voluntary restraint. Innovation of the protectionist tools has taken place and more tools are introduced based on the nature and the needs of the targeted industries. For example, to protect the car industry, tools such as standard requirement, domestic content and persuasive measures that encourage the locals to buy local cars are applied.

Measure such as standard requirement may be seen excessive and unnecessary especially in the case of Malaysia when it is generally agreed that imported cars possess higher technology attributes than the domestic cars and hence, imported cars are favoured.

In general, the cost of protectionism is often the portion of reduced producers’ benefit and the higher prices that consumers have to pay. However, due to the dynamic of technology and competition, the prices paid by consumers can be more sophisticated than what the conventional theory may imply. Due to trade barriers, the foreign products prices are forced to be higher than the domestic products, increasing the production or carrying cost to the foreign producers as the ‘non-tradable’ range of the products arises. Consequently, foreign producers upgrade their products to the range that yield greater profit margin effectively reducing the lower range products available (Hickok 1985, p. 2; Munger 1984, p. 57).

The cost associated with the disappearance of the lower range of products may be negligible if they can be replaced with domestic products of similar range at relatively

74 close prices. However, in the case of domestic products that are not up to the par, which may often be the case in the industry where the speed of technology changes is relatively fast, the cost can be significant. This is because consumers are forced to pay the price for products that are of a lower quality while consumers could have enjoyed better products if there were freer trade.

The cost of protectionism may increase exponentially across the continuum of technology intensity in production. The cost is expected to be higher in the car industry than in the textile industry for example, because consumption of cars takes up relatively a large proportion of the households’ income.

In addition, the failure to acquire the technology-know-how in production may impose high cost to the society in the form of defects that may endanger the safety of the consumers and other consumers who are not using the products for example, the other cars and other road users. The cost imposed on the latter is however, often not accounted for in the estimation of cost of protectionism. This is possibly due to the relatively negligible amount in a static model and possibly due to the difficulties in determining the cause of accident at a low cost particularly in a developing country where poor management of roads, poor law enforcement and weather may contribute to road accidents.

While protection of the car industry imposes cost to the society, the government and the self-interest group have complementing objectives. The self-interest group is predominantly the Malays who are politically linked to the government. They have the privileges to obtain import licenses and acquire shares in the domestic firms in return if they support the government in the election1.

Finally, the upward trend in the number of new car registration implies that the Malaysian car market is not saturated. The demand for car therefore, may have a positive correlation with the population growth in the long run. However, it is not the objective of this research to study the long run equilibrium in the Malaysian car market.

75

3.8 Car Demand

Cars are different from the “bundle of goods” that is often used in the explanation of consumer equilibrium analysis. Cars are durable goods of which their various services are used over a multiple period of time. Car owners derive satisfaction from the transportation services as well as comfort, aesthetic value, and for some brands of cars, their prestige. As Cragg & Uhler (1970, p. 387) suggests, valuation of cars by the owners therefore, render many statistical models insufficient to analyse cars’ demand. The conventional estimation of demand function for a good is a function of quantifiable factors but, car demand may not be linear and unlike the “bundle of goods” that are assumed homogeneous, cars are differentiated by brand and by models of each of the brands. In addition to the differentiated cars attributes, car owners’ demand for the same service of the same model may be different. For instant, a car owner may value the handling of a Toyota Camry while another Toyota Camry owner may value comfort. Such differences may be influenced by the car owners’ experience in driving different brands of cars or it can simply be through the word-of-mouth.

The complication of estimating the demand function for a car using statistical tools is further aggravated by the changes in technology that in turn, change the attributes of the cars, the ability and speed of the car makers to acquire the knowledge, and the possibility of positive disposal value of the old cars. Therefore, it can be seen that instead of attempting to quantify factors, estimation of the probability of how a certain set of factors may influence the demand for a brand of car and its attributes may generate more insightful results.

The studies of car market began in the United States in the 1950s. These studies use either cross section data or time series data, to estimate the demand function and supply function of the cars and subsequently, the price elasticity of demand for the car. The studies of American car market namely, Bennett (1957) and Bandeen (1957) and the study of car market, Turnovsky (1966), use cross section data but, different in their treatment of the definition of car demand of which depreciation is used as a proxy. Bennett (1957) defines depreciation as the change in the wholesale price for each car, Bandeen (1957) defines the depreciation of the new and used cars services while

76

Turnovsky (1966) uses average depreciation weighted by the total number of cars by their sizes.

Farrell (1954) and Suits (1958) study the American car market using time series data. Farrell (1954) takes into consideration new and used cars depreciation while Suits (1958) uses the quantity of new cars sold. All of these studies find that the most important factor that influence car demand is the disposable income. Other factors are prices of cars, age of car buyers, number of dependents, location, race, and preference factors.

Acknowledging the roles of products features in influencing consumers’ buying decisions, Cragg & Uhler (1970, p. 388) says that in addition to the basic function of a car, car buyers may derive aesthetic satisfaction and comfort from the car. Therefore, the value and age of a car essentially determine the demand for the car.

Subsequently, cars ‘attributes are beginning to be accounted in the 1980s. Agarwal & Ratchford (1980, p. 251) adopts the method used in Rosen (1974) to estimate the relationship between prices and the car characteristics using regression model. The car attributes accounted for in the model are: displacement, handling, ride, passing, luggage volume, and leg room.

The common car attributes selected in Feenstra (1984, p. 60) and Bresnahan (1987, p. 469) are length, weight, and horsepower. Other car attributes that are included in the model used in Feenstra (1984) are , power steering and air-conditioning while the other attributes included in Bresnahan (1987) are engine type and body-type. The difference in the car attributes chosen is mainly due to the times of which the studies were undertaken. Attributes such as transmission, power steering and air-conditioning are feasible to car buyers due to technology improvement. Therefore, Bresnahan (1987)’s study that uses sample drawn from the years 1953 to 1957 naturally reflects simpler technology.

Feenstra & Levinsohn (1995, p. 37), one of the pioneer studies of car market uses discrete choice model, finds that while weight, horsepower, status or safety of the cars and

77 consumer reports reliability index are significant factors that may influence the selected brands of car demand, width and length are found statistically insignificant. In this study, a dummy variable is used as a proxy for unobserved car attributes such as status and safety.

Following Berry (1994)’s proposition to use discrete choice models so as to include the unobserved demand factors that may consist unobserved consumers’ characteristics and unobserved product attributes, a few studies have begun to take this approach particularly for differentiated products.

Using data of a very large sample drawn from a population of car registered from the year 1971 to 1990 at industry level, Berry, Levinsohn & Pakes (1995, p. 869) finds that car characteristics influenced car demand. It is found that the ratio of horsepower over weight, air condition, miles per gallon dollar and size of the car, defined as a product of width and length, are statistically significant in influencing the demand for cars in the United States.

Subsequently, Berry, Levinsohn & Pakes (2004, p. 90) includes more attributes that are made available by technology and may be of car-buyers’ interest because of awareness of road safety. These attributes are the number of power accessories and the safety features. A new economic factor included is the payload in thousands of pounds. The other conventional attributes that included are the quantity of the different brands of cars, prices, horsepower, number of passengers, miles per gallon, and drive of the vehicles that is, either a four-wheel drive or none four-wheel drive.

Studies show that over the time, some attributes that were once important to car-buyers’ are no longer important. These attributes are air-conditioners and car sizes. Studies do not explain why these two conditions are not significant in the year the samples are drawn from but, it may be deduced that the technology for air-condition has reached its saturated level and there are other characteristics car buyers weigh more than the sizes of cars.

The trend observed in studies of car demand, in chronological order reflect technology as well as car-buyers’ expectation change. Therefore, micro studies of the differentiated

78 products of which their attributes and services are relatively heavily influenced by technology require careful selection of product attributes.

Table 3.2 Summary of literature survey in car market: Models and estimation of elasticities Studies Models | Countries1 | Time2 Elasticities of demand3 Farrell (1954) Regression analysis, time series Graphs show that income analysis | U.S. | 1947-1952. elasticity of demand increases as income increases. Bandeen (1957) Least square regression method Income elasticity: 0.90 developed in Duesenberry & Kistin (1953) | U.S. | 1940-1950.

Bennett (1957) Linear regression | U.S. | 1955- Income elasticity: 1.43 to 1957. 1.74 Suits (1958) Least square linear regression Price elasticity: -0.40 to -0.59 model | U.S. | 1929-1956. Income elasticity: 3.80 to 4.49 Turnovsky Ordinary least square, 3-pass least Price elasticity: -3.0 to -0.4 (1966) square | New Zealand | 1948- Income elasticity: Less than 5 1963. Cragg & Uhler Linear logit model | Canada | Not estimated (1970) 1960-1962. Agarwal & Rosen (1974) model: Log-log Not estimated Ratchford (1980) regression models - consumer level and product level | U.S. (New York) | 1976. Feenstra (1984) Hedonic regression model Not estimated Hickok (1985) Not specified | U.S. | 1985. Price elasticity: -0.18 (imported) Tarr & Morkre Linear regression model | U.S. | Price elasticity: (1984) 1981-1985. -1.1 (domestic), -3.4 (imported) Bresnahan Model used in Prescott & Price elasticity: -51 (1987) Visscher (1977), vertical product differentiation | U.S. | 1953-1957.

Levinsohn Log-linear regression model | U.S. Price (market) elasticity: (1988) | 1983-1985. -0.8 to -2.0

79

De Pelsmacker Structural model | Belgium | 1956- Price elasticity: slightly (1990) 1986. inelastic to slightly elastic. Income elasticity: Elastic Diewert & General equilibrium analysis Price elasticity: -0.29 Lawrence (1996) (industry level) | New Zealand | 1971-1991. Hufbauer & Partial equilibrium analysis Price elasticities: Elliot (1994) Log-log regression model | U.S. | -1.19 (domestic) 1990. -1.50 (imported) Feenstra & Oligopolistic pricing model | U.S. Not estimated Levinsohn | 1987. (1995) Berry, Levinsohn Discrete choice model: Simple Price elasticity: Inelastic in & Pakes (1995) logit and instrumental logit | U.S. both functions | 1971-1990. Berry, Levinsohn Discrete choice model | U.S. | Price elasticity: -3.58 to -11 & Pakes (2004) 1993

Notes: 1. Countries in which studies were carried out. 2. Time frame within which data was collected. 3. Studies did not report if the elasticities of demand estimated are constant elasticity of demand.

A summary of the studies of the car market, the models used and estimation of own price and cross price elasticity estimated are shown in Table 3.2 above. The table shows that study of the car market began as early as in the 1950s. However, due to the state of technology, the demand for cars was generally derived demand for transport. Statistical tools used were conventional econometrics tools. As technology advances and income level increases, the derived demand for car is no longer for transport but, also for other characteristics that differentiate cars. However, it is highly likely due to the availability of statistical tools that in depth study of car characteristics’ influence on car demand begin much later although studies such as Lancaster (1966) and Cragg & Uhler (1970) suggest that the demand for some goods can be influenced by product characteristics because consumers derive satisfaction from product characteristics.

80

Studies begin to incorporate car characteristics about 20 years later after literature propose the importance of product characteristics in influencing product demand. Studies of car market take a turn after Berry (1994) applies Discrete Choice model to estimate demand function for differentiated products. This study applies Discrete Choice model to estimate demand functions of different car makes in Malaysia in the attempt to find out if car characteristics influence the demand for domestic cars. In a contestable car market like in the United States, car characteristics influence the demand for each car makes at different degree, reflecting car buyers’ preference. However, in a protected car market, it is anticipated that protected car maker is highly likely to depend on pricing strategies instead of car characteristics that require competitive technology and competitive labour.

3.9 Conclusion

Benefits of free trade often cited and supported by evidence are economies of scale, efficiency and spillover effect of technology transfer. However, these benefits may not materialise in some economies because of the size of the economies and other factors such as characteristics of the product either homogenous or differentiated. Although variety of choice is also a benefit of trade, there is still lack of statistical evidence to measure the significance of the benefit. Other argument for free trade is tax revenue from more firms in a contestable market and income increases associated with improved productivity.

There are arguments that suggest economies of scale is not a significant reason for free trade. Factors that influence the significance of economies of scale are the learning curve of the protected domestic firm, diversity of products, and transport cost. Studies suggest that failure of domestic firms to learn, large diversity of differentiated product and high transport cost will dissipate the benefit of scales. Studies also say that a variety of choice decreases when firms exit the industry as a result of fierce competition, making freer trade undesirable. However, this argument is relevant if domestic firms produce less differentiated products and the economy is a developing economy where large proportion of the population demand cheaper and less differentiated products.

81

Cost of protectionism associated with efficiency is difficult to trace and measure because the sources of inefficiency are difficult to be identified specifically. Self-interest groups’ intention and government’s objectives may be political rather than economic.

Evidence of protectionism is mixed. Infant industries in some economies grow because of the firms’ desire to acquire knowledge and to compete while in some economies, infant industries do not grow because there is lack of growth despite low cost of acquiring the knowledge. Failure of infant industry may be attributable to government’s lack of commitment to lift protection, hence causing complacency in the labour. It is therefore concluded that learning is unavoidable and protectionism cannot be for long term.

The exports of Proton have been about 20,000 units each year since 2008 (MITI 2010, p. 1) reflects its lack of competitiveness in the international car market. Therefore, Malaysia is unable to experience the benefit of freer trade as in the case of the United States-Canada auto pact. Since Malaysia’s car market is small and the technology gap between the domestic car maker and the foreign car makers may be so large that freer trade may lead to excessive imports of foreign cars, Proton is unlikely to survive the competition both in the domestic market and the foreign market.

This study attempts to answer the first research question if car characteristics have significant influence on car demand such that buyers are willing to pay for characteristics foreign cars despite their higher prices than domestic cars through estimation of demand function using Discrete Choice modelling.

Literature review shows that evidence for infant industry argument is mixed. This study takes on a car industry that is heavily protected using tariff and non-tariff measures, attempts to find out if technology diffusion has taken place such that the protected domestic car maker is able to compete with other car makers in the domestic car market. It follows that if the domestic car maker is able to compete with other car makes, statistical evidence will show the contribution of car characteristics’ influence on the demand for domestic cars.

82

Literature also shows evidence of relatively small protectionism cost that is less than two per cent of the economies’ GNP in many studies while a small number of studies suggest that the cost can be high when high tariff or non-tariff measures restrict imports.

This study is unable to segregate the price increases due to tariff and non-tariff measures because of lack of information. As a result, the cost estimated is the aggregate cost of protectionism. The cost estimated is measured as percentage of manufacturing sector’s and as percentage of the economy’s GDP as a whole. This is because manufacturing sector’s contribution to the economy’s GDP is relatively small in a developing economy like Malaysia.

Literature generally argues that freer trade leads to greater price elasticity of demand due to competition, estimation of price elasticity of demand in this study allows the following:

(a) Comparisons across different car makes available in Malaysia. The results will give price elasticity of demand for domestic car makes, for foreign car makes of which their manufacturing and assembling plants are located in Malaysia, and foreign cars that are imported from their countries of origin; and

(b) Comparisons across studies in different economies over different period of time. Comparisons across studies over different period of time suggests technology changes and different marketing strategies adopted by car makers in highly competitive environment may have effects on the demand for cars.

Notes

1. See Jomo, KS 2004 for the rise of cronies and corruption, and emergence of Malay elites through the implementation of various policies and privatization. There is no specific study of wealth distribution and efficiency in Proton. The implicit but, possibly the main objectives of establishing Proton and protecting the infant industry are mentioned in reports not published in the local mainstream periodicals. For example, see Lee (2013).

83

Chapter 4 Methodology

4.1 Introduction

This research attempts to answer the stated research questions related to conventional microeconomic theories and estimate cost of protectionism using an alternative approach - Discrete Choice model. Lancaster 1966 argues that the utility of consuming a good is not always derived directly from consumption of the good itself but, the characteristics of the good when the good is not homogenous. For example, the utility of driving a car is not only the satisfaction derived from transport services of the car but, satisfaction derived from car characteristics such as comfort offered by high horsepower when driving long distant.

This research essentially attempts to estimate the cost of protectionism based on the estimated number of new cars registered in a specified year if there were no tariff, and based on a set of car characteristics and a set of car buyer characteristics. The approach taken is different from the conventional mainstream economics approach in that this research does not assume homogeneity in products.

In the discipline of marketing, car makers ought to identify the niche market for their products, that is target to certain groups of car buyers rather than to treat car buyers as a general group of consumers. As a result, based on the characteristics and preference of targeted groups of buyers, cars are made to possess a specific set of characteristics at certain price level that enable the car makers to penetrate their targeted market. Therefore, as technology advances and competition becomes stiffer, cars are highly differentiated.

Linear regression analysis using econometric tool may not be able to incorporate car characteristics and car buyer characteristics without causing spurious regression problem. By using Discrete Choice model, this research is able to quantify the contribution of car characteristics to certain car makes while accounting for car buyer characteristics that are often cited in conventional econometric models. For example, it may be found that

84 relatively more female car buyers buy larger cars measured by the number of passengers, while relatively more male car buyers buy cars with larger horsepower.

The alternative approach taken in this research does not attempt to quantify satisfaction but, specifies demand function as a probabilistic function. Therefore, estimation of price elasticity of demand is based on the estimators found in Logistic Regression analysis.

The approach taken in this research allows product differentiation to be accounted for their influence on the demand for cars. Logistic regression analysis applied for Discrete Choice modelling has the advantage of not relying on linear relationship between an independent variable and the dependent variable but allow interaction between independent variables.

This research, therefore, takes on epistemology approach in that samples of car buyers and car makes are drawn from across 13 states and the Federal Territory of Malaysia. It is assumed that the samples drawn are representatives of the population. This approach reflects the principles of positivism. Data collection will follow protocol prescribed in applied research literature to ensure minimum sampling bias. Several steps to achieve objectivity of data are the use of two languages, the English and Malay languages in questionnaires and keeping questions short and simple to reduce ambiguity.

Since partial equilibrium analysis framework is applied in this research, the costs of the spillover effects of protectionism cannot be estimated. A spillover effect of protectionism is an unrealistic expansion of the protected firm. Such expansion coupled with lack of competitiveness of the firm gives rise to excess capacity and misallocation of resources. Although non-parametric approach using the Data Envelopment Analysis (DEA) is applied to identify the gap arises from excess capacity, the cost of misallocation of resources is not measured in this research.

Other spillover effects that are associated with protectionism such as unproductive profit- seeking activities and economies of scale are not accounted for in this research due to lack of reliable data.

85

Approach taken in this research involve quantitative analysis of relationships among the variables identified. This approach entails quantifying the identified characteristics, testing and measuring the relationships between different combinations of the characteristics. Quantitative accounting data and market data are also analysed to measure efficiency of car makers. The latter is measured by the weighted output of firms that is, quantity of cars that can be produced.

Primary data collected is a pool of quantifiable observations for two sets of variables, namely: car buyer characteristics and car characteristics. All car buyer characteristics and car characteristics are variables identified in a priori studies. Since to date, most studies in related area that incorporate car characteristics are carried out in developed economies, precautionary steps are taken to ensure car characteristics that are not relevant in Malaysia. In addition, steps are also taken to elicit characteristics that are not identified in those studies but, relevant to Malaysia are incorporated in this research. As such, the data collected can sufficiently represent the car market in Malaysia.

4.2 Research Philosophy

Scientific approach taken in conventional microeconomic theory measures satisfaction derived from consuming goods. This approach entails identification of measurable variables, assumptions that allow theorizing of observation, specification of models, and hypotheses that are based on the observation relations among variables. As such, this research is based on the principles of positivism (Saunders, Lewis & Thornhill 2009).

4.2.1 Metatheoretical assumption 1: Ontology

The ontology of this research is dualistic in nature that is, there is separation between subject and the object. Although sample frame and parameters are based upon observation in the a priori studies, data collection procedures involve application of face- to-face interview and open-ended questions so as to allow respondents to provide information that reflects closely the phenomena this research attempts to investigate.

86

Information provided is comprised of objective answers such as “Yes” or “No”, and selection of measureable indicators such as income range and horsepower of the cars.

4.2.2 Metatheoretical assumption 2: Epistemology

The epistemology of the research is the attempt to develop knowledge of car market based upon the observe reality - car buyers are willing to pay more for foreign cars and car makers are increasingly spending huge sum on research and development, and marketing strategies. Therefore, in addition to the social-economic factors that are stated in mainstream Microeconomic theories, car characteristics that differentiate various car makes are included in the pre-test and pilot test questionnaires for identification of relevant cars and car buyers characteristics. This is to ensure that technology advances that make some characteristics insignificant are excluded while new characteristics if any, will be added to the models.

Discrete choice model using logistic regression analysis allows more variables to be incorporated in the model and allows interaction of the variables that is, collinearity which the conventional econometric models are to avoid in order to generate reliable estimators. Due to competition, products today are often differentiated, making it difficult to identify causal relationships between variables. Advertisements of cars suggest that there are certain car characteristics targeting gender and income groups. For example, BMW’s high horsepower models target male drivers who are well versed in high-technology equipment while Nissan’s MPV targets female drivers who have to fetch children from school and need large space for grocery.

In order gauge reliable information from car buyers, the model and analysis approach taken will allow car buyers to indicate their preference and decision through measurable indicators. Through logistic regression analysis, data of different variables collected will be allowed to “speak” for themselves.

87

4.2.3 Metatheoretical assumption 3: Research object

Although it is generally believed that research objects possess qualities that are independent of the researchers, it cannot be denied that researchers may have to a certain extent, effect on the quality of research object. The assumption of research object being independent of researcher does not hold in many research including this research. This is because the researcher may affect the research object through the data collection, cleaning and analysis processes.

In this research, this assumption may not hold because there remains some level of connection between researcher and research objects. The approach taken to collect data at car service centres, schools, offices, churches and universities to certain extent influenced by researcher through selection of places. However, design of questionnaires may maintain the independence of research object if there is no significant ambiguity in questionnaires and questionnaires are distributed by personnel, not known to researcher, in each location. The desired object independence from researcher can be enhanced through the application of analytical tools that do not require specification of functional form.

4.2.4 Metatheoretical assumption 4: Research method

This research requires collection of primary data as there are no institutions that record and store data of new cars sold every year. The sample size is determined based on the desired confidence level. Ideally, samples will be drawn proportionately according to the proportion of new cars registered in all 13 states and the Federal Territory. The underlying principle of the desired sample size is that the sample size is sufficiently large to represent the population.

Quantitative data collected will be filtered to remove irregularities prior to test and analysis for study of causal relationships observed and not observed. However, not all irregularities are removed but, irregularities that are seen personal and untrue. For example, a car buyer claimed that while making the buying decision, a Proton car of

88 which the niche market is the low income group, is compared to a Mercedes Benz car of which the niche market is the high income group. This sample will be omitted because the information is not reliable. However, other irregularity such as a car buyer with low income, acquired a BMW is not omitted from the sample. This is because the Chinese culture favours wealth accumulation for the future.

Although random sampling is desired to collect a representative sample, the collection process is planned to be broken into two rounds that is, second collection is carried out three months after the first collection for the purpose of achieving a sample that reflects market shares of major car makes. As a representative sample will reflect a balance spread between two genders, samples collected are not mapped to the observation that there is almost equal number of male drivers and female drivers. Similarly, a representative sample will reflect the ethnic proportions close to the actual market’s ethnic proportions.

4.2.5 Metatheoretical assumption 5: Truth

The notion that a sample can be used for the study of population is the known population parameters. While a few parameters such as number of new cars registered per year and market shares of each car makes are known, there are actual population parameters that are not known. These parameters are household income and cost of production that at best, can be represented by proxies published by Statistics Department and yearly audited and published financial statement. Therefore, while there is a certain level of truth in the information, the accuracy of the information may not be known.

This assumption further linked to the assumption that data collection tools do not present ambiguities that hinder their ability to gauge true data. Procedures taken to minimise ambiguities in and misunderstanding of questions are as follows:

(a) Application of commonly used terms in the questionnaires in designing pre-test questionnaire;

89

(b) Translation of English questionnaires to Malay language questionnaires;

(c) Performance of pre-test using English and Malay language questionnaires in face- to-face interview and self-administered survey;

(d) Questions raised by respondents during interview or answered questions in self- administered survey are reviewed and re-written;

(e) Revised questionnaires are distributed for pilot test;

(f) Through the unanswered questions and inappropriate answers given in the pilot test, amendments deemed necessary are made; and

(g) Final draft of English and Malay language questionnaires are distributed to different states and the Federal Territory.

The procedures taken are not able to overcome the weaknesses of this research due to unknown population parameters. It is contended that the actual population parameters are not fully known and therefore, published information are used either as benchmarks or as actual parameters. It follows that the conclusion and remarks derived from tests and analysis of data are fallible.

4.2.6 Metatheoretical assumption 6: Validity

Following the procedures to gauge true information from car buyers, the data collected is mapped to a priori studies and to market observation. Subsequently, the data collected is considered valid and therefore, conclusion based upon study of the data is also valid.

The conventional statistical tools strive to achieve validity of test results through investigation of the “residuals” not explained by functional forms specified. The results are considered valid so long as test results are consistent with theories and the residuals are random.

90

In this research, data of all variables collected are tested for their correlations with each other prior to logistic regression analysis. An additional step taken in logistic regression is testing the residuals for unobservable car and car buyer characteristics. Subsequently, conclusion is drawn based upon mapping of test results to the observed behaviour of the society.

4.2.7 Metatheoretical assumption 7: Reliability

The approach taken in this research can be replicated because of the followings:

(a) Parameters applied are traceable and measurable variables;

(b) Models are consistent with models specified in a priori studies;

(c) Primary data is collected from car buyers across Malaysia while secondary data is audited and published data; and

(d) Sources of secondary data are cited in the reference list of this research.

In conclusion, this research is highly structured, from design of questionnaires to analysis of data. Sample size chosen is based on the level of confidence level with reference to sample sizes of studies in the area of Discrete Choice modelling. Treatment of raw data such as treatment of outliers, missing data, and data coding prior to tests are consistent with underlying theories in quantitative analysis. SAS programme is applied for data analysis and for generating estimators that are then used for calculation of economic data such as elasticities and cost of protectionism.

4.3 Ethical Issues

The SUHREC (Swinburne’s Human Research Ethics Committee) project number of this research is 2015/071 for ethic clearance. Ethic approval was granted for the period from

91

August of 2015 to July of 2016 during which pre-test, pilot test and final data collection are carried out.

This research is broadly categorised as Social/Cultural/Humanities research, conducted entirely in Malaysia. It involves collecting samples from across all 13 states and the Federal Territory directly from car owners. The procedures involve obtaining lists of car service centres in Malaysia. There are lists of service centres for designated car makes such as Proton and Toyota 3S centres, as well as a list of general service centres that bear the name of local private entity but, provide services to designated new cars for example, KM Otomobil Sdn. Bhd. in Pahang that provide services to new Perodua cars. Managers of the service centres are informed of the nature of this research, how survey might be carried out and assistance they might be able to provide. Front desk of the service centres distributed questionnaires to car owners who waited at the centres. Front desk of the service centres would also collect completed questionnaires and sent them back to student investigator using the pre-paid, self-addressed courier envelope provided.

A few sensitive questions related to gender, salary group, and race are written. Most studies find that gender and salaries are important social-economic factors that influence the demand for car. For example, Bennett (1967, p. 844) includes race in the study of the U.S. car market because of the hypothesis of different spending habits of the whites from the other ethnic groups such as African Americans, Puerto Rican and other group categorised as non-whites. Bandeen (1957, p. 240) includes occupation although income is also specified in the model.

Malaysia is a multi-racial country and spending habit of the natives can be different from non-natives. In addition, natives particularly Malays dominate the public sector and incentives are given to civil servant to buy domestic cars. Therefore, it is reasonable to test for correlations between car demand and sensitive social factors.

Ethical issues arise in the process of collecting data because staff of service centres are volunteered to assist distributing questionnaires to their customers. In addition, sensitive questions are posted to obtain social-economic data. Respondents, the car buyers might see themselves exposed to risk of their personal information being abused for commercial

92 purposes such as mailing list of circulars. Due to offering of prize for respondents who sent in their completed questionnaires on time, names and addresses of respondents who wanted to participate in the draw are provided. However, respondents are not identifiable because the slips of which names and addresses of respondents are to be detached from the questionnaires by the respondents themselves before submission of completed questionnaires.

Although it is stated in the consent statement how data is analysed and that the data is used solely for academic purposes, there remained a certain level of risk facing car owners from their perspective. Apart from car owners’ perceived risk of abuse of personal information, there might be risk of hostility arise among domestic car owners or car owners who are employees of domestic car makers. Since hostility among car owners might lead to low response rate, the words “cost”, “protectionism”, and “government” are not used in the questionnaires.

Consent statement attached to questionnaires contain contact numbers and emails of both student investigator and Swinburne’s Human Research Ethics Committee (SUHREC) for respondents’ enquiries and/or complaints. There were neither enquiries nor complaints received throughout the period of data collection.

4.4 Theoretical Framework

Partial equilibrium analysis framework is often used for specific product, for examples Santerre & Vernon (2006) and Craft (2002) study the effects of price control on drugs and maximization of tax revenue from vanity license plates respectively. Such approach makes the assumption of ceteris paribus and examine how government intervention for example, may affect the price of the products concerned. Krugman, Obstfeld & Melitz (2012, p. 61) mentions that partial equilibrium analysis is sufficient if a study aims to focus on a single market. However, if it is also a study’s objective to investigate how an event affects different markets due to their linkages, general equilibrium analysis is appropriate.

93

Alvarez & Lucas (2007) uses general equilibrium analysis for patterns trade across countries after imposition of import tariff. Using general equilibrium framework, Harris (1984) shows that when accounting for internal economies of scale for about 30 industries in a small open economy, removal of tariff generates welfare gain of more than 2 per cent of base national income. It is found that the gain is higher when there is imperfect competition.

Partial equilibrium analysis is adopted in this study because this research aims to study the effects of import tariffs and quota imposed specifically on imported cars, and the protectionist measures’ effect on the domestic cars. In addition, the contribution of the manufacturing sector to the country’s GDP in 2012 is approximately 4.8 per cent of which the manufacturing of motor vehicles plus the manufacturing of the parts and accessories for motor vehicles and their engine contribute to less than 10 per cent of the sector’s output value (Statistics Department, Malaysia). Therefore, the effects of tariffs and quotas on imported cars on the other sectors that can be captured in the general equilibrium analysis, is expected to be negligible relative to the Malaysia’s GDP.

Studies conducted in the 1960s using the measure of static cost of protectionism, often underestimate the cost of protectionism that is, less than 1 per cent of the GNP (Panagariya 2002). The reason for under estimation is cited to be the “spillover” effect of protectionism that has not been taken into account. The “spillover” effect postulated in the alternatives approached in general equilibrium framework take into account economies of scale (Harris 1984), disappearance of imports (Romer 1994), X-efficiency (Bergsman 1974 and Balassa 1971), and unproductive rent- or profit- seeking activities (Bhagwati 1982 and Krueger 1974) all of which are on the supply side.

The model used in this study follows the study in Hufbauer & Elliot (1994) in that it extends partial equilibrium model to take into account the spillover effects of protectionism. The spillover effects in Hufbauer & Elliot (1994) are the possible fall in the domestic car production that leads to job loss and the possible change in the terms of trade.

94

As Malaysia is a small country, it is assumed that tariffs and quota do not influence the world price thus, terms of trade. Elimination of protectionist measures may cause a downward shift of the domestic car maker’s supply and therefore, increases job loss. However, the downward shift of the domestic car supply may be matched by the upward shift of the foreign car supply because the foreign-based car makers that have been operating in the industry may be able to reap economies of scale and new jobs may be created. As a result, liberalization of car industry has more contribution to GDP growth than a protected car industry.

However, instead of using representative consumer model for estimation of price elasticity of demand and cross price elasticity of demand, this study uses Discrete Choice Model (DCM) that is reminiscent of representative consumer model and characteristics model. Using this method, consumers’ preferences are the characteristics of the cars. Individual consumers with their characteristics are “correlated” to the prices and characteristics of the cars of their choices. As such, the model allows estimation of demand’s response to prices, correlation of cars’ characteristics to car prices, and cross substitution between sets of characteristics.

4.4.1 Welfare loss

A priori studies that evaluate performance of infant industries involve counterfactual analysis of a scenario without intervention. Head (1994, p. 142) suggests that it is desirable to have series of quantitative analysis of the industry over a period of time so as to have a true picture of technology evolution. The intuition is that if industries’ growth does not synchronize with technology evolution, the cost of failure in technology diffusion will arise. Successful technology diffusion at a time does not promise permanent success.

In this partial equilibrium model, the assumptions made are:

95

(a) Proton cars and the non-Proton cars are imperfect substitutes. Studies found that assumption of perfect competition leads to underestimation of the cost (Ibid, p. 22);

(b) the supply of Proton cars is positively correlated to their prices although Proton is heavily assisted through government’s allocation for capital and land;

(c) Malaysia is a small country, therefore, the supply of non-Proton cars is highly elastic; and

(d) all markets are contestable as large gap between Proton’s technology and the other car makers’ technology provide leverage to the foreign players to adjust their production level and product characteristics to achieve price competitiveness.

Due to the imperfect substitution of the Proton cars and the non-Proton cars, the welfare effects of two market categories have to be calculated separately. The framework applied in this study is initially used in Morkre & Tarr (1980) and subsequently applied in Tarr & Morkre (1984) and Hufbauer & Elliot (1994). The framework is illustrated in Figure 4.1.

The upper frame of Figure 4.1 shows that removal of trade barriers reduces the price of the imported cars in the domestic market and increases the demand for imported cars. The increases in demand from Qm to Qm’, at a lower price level at Pm’, can be reflected by the new demand curve Dm’ that implies where the demand could have been if there is no trade barriers. The consumer surplus loss due to the higher price is PmbedPm’, while the portion of consumer surplus loss that is gained by producer is the increases in the producer surplus represented by the area PmbePm’. The deadweight loss in the imported car market is represented by the area bed.

The lower frame of Figure 4.1 shows that if trade barriers are removed, the demand curve for the domestic cars will lower. The imposition of the trade barriers causes higher price and thus, loss of consumer surplus represented by the area PdnpPd’. The portion of lost consumer surplus is gained by producer is PdmpPd’ while mnp is the loss of consumer

96 surplus that is not gained by the society in the domestic car market. The deadweight loss represented by the area mnp.

Figure 4.1 The effects of trade barriers removal

Price

Pm a b

Pm’ e d c Sm

Dm’ Dm

Qm Qm’ Quantity

Import market

Price

Sd

Pd l m n

Pd’ o p

Dd Dd’

Qd’ Qd Quantity

Domestic market

Source: Hufbauer & Elliot (1994), pp. 32-33.

97

In this model, estimation of welfare changes due to import tariffs and quota is based on the estimated price elasticity of demand, cross price elasticity of demand, and price elasticity of supply. The cross price elasticity of demand allows estimation of the cost associated with the substitution of non-Proton cars with Proton cars due to tariffs and non- tariff barriers.

However, the price elasticity of world car supply is not available. Hufbauer & Elliot (1994) assumes that the price elasticity of world car supply is 3, without taking into consideration spillover effect of the trade barriers that is, the terms-of-trade losses while de Melo & Tarr (1992) assumes the figure to be 5, taking into account the terms-of-trade as the United States is a large economy. As Malaysia is a small country, tariffs and quota imposed by Malaysia is unlikely to cause terms-of-trade losses to the economy. However, as there is world excess supply of cars and Malaysia is a small country, it is reasonable to assume that the world supply of car to Malaysia is very elastic to price changes. Therefore, it may be assumed that the price elasticity of world supply of car to Malaysia is more than 3.

The welfare change that is, the cost of protectionism imposed on the car buyers are in the forms of:

(a) higher price paid by the domestic car buyers;

(b) higher price paid by the foreign car buyers because of tariff and quota;

(c) higher price paid for ownership of domestic car for substitute of public transport due to promotion of private transport ownership by the government; and

(d) substitution of expensive foreign cars for relatively cheaper domestic cars.

The implicit cost of tariff such as sales tax may be reflected by “income tax surcharge” equivalent that measure income distribution effect of sales tax imposed on the car buyers.

98

4.4.2 Models

At aggregate level of study, the demand for cars is defined as the quantity of cars purchased per year over the years. Feenstra (1985) for example, uses hedonic prices of cars to take into account the characteristics of the car and applies logarithmic regression model in the study of the effect of voluntary export restraint on Japanese cars on the U.S. car demand. The a priori studies done after the World War II, at disaggregate level, defined demand for car as the service consumed of which the proxy used is the value of depreciation assigned to each car (Cragg & Uhler 1970, Turnovsky 1966, Bandeen 1957, and Bennett 1957), with very limited account for car characteristics.

As cars in general, are not homogenous products, their attributes may have an effect on buyers’ decision (Berry, Levinshon & Pakes 1995; Earl 1995; Feenstra 1985; Berkovec & Rust 1985; Agarwal & Ratchford 1980; Cragg & Uhler 1970; and Lancaster 1966) and these attributes are often not directly observable and measurable using Econometrics models. According to Head (1994, p. 147), Anderson et al. (1992) shows that conventional approach using representative consumer model with the assumption of constant elasticity of substitution (CES), is able to generate equivalent results despite heterogeneity in the sample that would have been tackled well using discrete choice model. Heterogeneity in sample is often dealt with by segmentation so that sample in each segments are homogenous.

Cars are treated as continuous goods when the proxy for car demand is depreciation measured in monetary terms and influenced by the distant travelled. However, cars are discrete goods due to their durability and relatively large proportion of income spent on cars ownership. As income increases and technology improves, the derived demand for cars may vary across car owners who possess different sets of characteristics. In addition to transportation, the very basic function of cars, car owners’ derive their satisfaction from aesthetic value of the cars, the prestige of the car make, comfort and the perceived level of injuries when accident occurs.

99

4.5 Literature Review - Logit Model

4.5.1 Origin of logit

J.S. Cramer (1928-2014) wrote a detail account on the origin of logit model in his book titled: “Logit models from economics and other fields”, published by Cambridge University Press in 2003. Much of the content in this section is taken from Cramer (2004). According to Cramer, logistic function was first introduced in the 19th century to express organisms and population growth for autocatalytic chemical reactions course. The function time path of a population W(t) for example, can be expressed as a first order derivative of the function:

W⁰( ) = W(t) / t (1)

� � � that is, W⁰( ) = W(t) (2)

� β where β is a constant rate of growth.

Without constraints, the expression suggests exponential growth that can be expressed as:

W(t) = A exp βt (3) where A is any initial value of W, β is a constant growth rate.

Alphonse Quetelet (1795-1874) and his student Pierre-Franҫois Verhulst (1804-1849) agreed that indiscriminate extrapolation of exponential function generates undefined values. It must have been their observation that there are constrains to growth. Therefore, Verhulst added a component that represents the constrain to the growth function as:

W⁰( ) = W(t) – W(t) (4)

� β ϕ It is not written in what forms was experimented but, a simple quadratic form is described. The quadratic function is expressedϕ as:

100

W⁰( ) = W(t) [ -W(t)] (5)

� β � Where denotes the upper limit of W, and asymptotes as t approaches infinity. This quadratic� function reflects not only population growth that is constrained by resources, it also reflects economy’s adjustment to long run equilibrium in the labour market and output market for instance.

With the upper limits imposed on the function, growth is proportional both to W(t) and the remaining space for further expansion [ -W(t)]. Thus, the proportion of W(t) can be expressed as: �

P( ) = W(t)/ (6)

� � Therefore, P⁰(t) = β P(t) [1-P(t)] (7)

First order derivative of (7) gives:

( ) P(t) = ( ) (8) exp �+�� 1+exp �+�� which is the logistic function.

Logistic function used in bio-essay is expressed as:

P(Z) = ���� 1+���� where P is the probability of binary outcome Z, where Z = α + βX , and X is a continuous stimulus or exposure variable.

Logistic function has since been applied in the course of chemical reactions and being commonly known as autocatalytic function until 1920. Subsequently, it was also applied in the course of individual organisms and in the course of population growth.

101

Probit (probability unit) was introduced in the 1930s, often credited to Gaddum’s report of 1933 and Bliss (1934a,b), according to Cramer. Probit model began from Fechner’s (1801-1887) observation that human response to an identical stimulus is not uniform. Thus, he transformed the observed differences to equivalent normal deviates. Probit function is based on the principle that where stimulus is determinate, responses are random because of variance in individuals’ tolerance level. The function is defined as:

exp ( ƒ = 2 ) du (9) 1 � � � 2 √2� ∫−∞ − 2� that is, for any relative frequency ƒ, there is an equivalent cumulative normal deviate Z.

In 1944, Joseph Berkson postulated the inverse of logistic function, giving:

Logit (P) = log [P/(1-P)] = Z (10)

As a result, calculation of logit (equation 10) is more straight forward than calculation of probit (equation 9) although logit function cannot be related to the underlying concept of normal distribution like probit function does.

Berkson went on to write a paper titled: “Why I prefer logits to probits” and published it in 1951. Logit began to gain ground in the 1950s due to its simplicity in computation while probit is still in use for many studies. Examples of study using probit in the 1950 is Aitchison & Silvey (1957). Other studies using probit are like Hausman & Wise (1978), Imbens & Lancaster (1994) and one of the more recent studies in road accident is Fountas & Anastasopoulos (2018).

4.5.2 Development of logit model

This section reviews a few selected studies that show development of logit model and its application in social sciences. There are more studies in the early years in 1960 and 1970s that are not mentioned in this section.

102

Studies began to apply probability model to study travel mode probably as early as in the 1960s according to Stopher (1969, p. 57). A few early studies referred to by Stopher (1969) are Quinby (1961), London Traffic Survey (1964) and Quarmby (1967). The study applies simple linear regression model, expressing the probability of people using cars given an estimated function of cost and time of travel. In this model, a set of independent variables are specified as function for the probability of an event to occur. Due to possible positive or negative values of cost and time of travel, results generated may be outside the range of probability that is, from 0 to 1. As such, the function is transformed to generate desirable result. The simple logistic is expressed as:

p = � (11) � � 1+� The transformed logistic function can be expressed as:

loge ( ) = y (12) � 1−� where y is a linear function of economic parameters.

However, such simple form is a binomial function appropriate for studies observing 2 outcomes or 2 dependent variables such as car or bus. Theil (1969) proposes a remedy for studies that may observe more than 2 outcomes for example, car, bus, train, and motor bike by applying a multinomial linear logit model. The study concludes that multinomial linear logit model is capable of generating reliable coefficients of independent variables.

Following Lancaster (1966)’s proposition for study of goods of which their characteristics yield satisfaction to consumers, Cragg & Uhler (1970) for example, treats cars as durable goods and qualitative attributes of the cars constitute to car owners’ satisfaction. In addition to the qualitative attributes such as comfort and aesthetic satisfaction that are not easily observed and measured, the process of observation is difficult to control to obtain unbiased information (McFadden 1974, p.106).

Due to the nature of qualitative factors, their effect on car demand may be “lumpy” hence, systematic variation in choice at aggregate level within a set of qualitative factors (Ibid, 103 p. 106). In a study of choice behaviour, a conditional logit model is used to study how household make their choice of shopping destination, given the set of transport modes, a set of attributes such as walk time and travel time, and a set of household characteristics. This study shows logit model is gaining ground and its usefulness for more complicated studies that involve more than 2 sets of independent variables.

Studies in the 1970s were challenged by availability of disaggregate data and sophisticated statistical tools. During the 1970s, studies on United States’ car market use aggregate data for econometrics modelling. Lave & Train (1979, p. 1) argues that most of the studies earlier have not included sufficient car characteristics for tests of their significance in car demand and aggregate data is able to capture variation in independent variables and their influence on car demand. In addition to car price and fuel economy, the study suggests that other characteristics such as weight, external dimension, car size measured by number of passenger, and horsepower have to be taken into account.

Using disaggregate data of new car buyers and multinomial logit (MNL) model, Lave & Train (1979, pp.8-9) finds that taxes on gas have positive impact on smaller cars but have a negative impact on intermediate and large cars. Similar observation is made when 10 per cent excise tax was imposed on intermediate and large cars. Although the study shows that there is a substitution effect, the study has not shown car characteristics’ influence on car demand. The study suggests that unobservable variable being taken into consideration when research attempts to examine consumer choice. However, the study has not suggested if the unobservable variable is variable associated to car characteristics or to car buyers.

Over the years, studies begin to look at the dynamic nature of technological change. Individual car buyers’ demand for car is discrete because once a new car is acquired, car owners do not immediately response to changes in car market. Study of durable goods such as cars is unique because of the dynamic nature of car demand is influenced by technological changes. On one hand, technology changes and hence, car characteristics change, on the other hand, cars obsolete as the age of cars and distant travelled increases. In addition, technology advancement makes the old technology obsolete hence,

104 technological changes have impact on the value of cars and subsequently, the demand for car.

In addition to social-economic variables, Manski & Sherman (1980, pp. 351-352) specifies car characteristics in multinomial logit model. Three variables included in the study but not suggested in most studies are acceleration time, noise level, and disposal rate of the cars after used. Based on a sample of about 600, using combination of disaggregate and aggregate data, results of the study finds probability of buyers choosing cars of a certain set of characteristics. For examples, probability of young families prefer light cars is greater than young families that prefer heavier cars. The probability of large luggage space preference associated with large families is greater than the former’s association with small families. Results of the study show that lower household income is associated with lower car price and cost of car ownership is associated with education level of the car owners through the choice of cars. Influence of car performance characteristics such as noise level, turning radius, and braking distance are found insignificant. Although the study has taken into account variables associated with technological changes, the study has not been able to able to show the impact of technological changes on car demand.

Apart from technology, other factors that change over time may be car owners’ characteristics such as age and taste, and changes in market conditions (Mannering & Train 1985, p. 268). Although not suggesting how the dynamic may be treated in the study of car market, Mannering & Train (1985) highlights the need for studies have to take changes in the car market into consideration for logit modelling. Hensher & Le Plastrier (1985) applies multinomial logit (MNL) model on sample of 400 households, takes into account households’ accumulated experience and vehicle replacement. It is concluded that qualitative factors that influence buyers’ decision are brand loyalty and experience of previous car ownership. The study suggests the importance of accounting for unobservable variable as choice is continuous.

Berkovec & Rust (1985, p. 275) applies nested multinomial logit to study how households that own a car made their choice in a static model. It is called “nested” because households can be nested within a set of clusters. Sample of 237 chosen are made up of

105 new car, middle-age cars and old cars. Observable car characteristics are fuel economy, car size measured by number of passengers, and weight of cars. The study concludes that American car buyers who bought cars during the period 1967 to 1978 have the tendency to keep their cars instead of trading their old cars for new cars. This implies that technology changes may not contribute to “replacement” of cars and that transaction cost may influence new car demand. However, the study cannot conclude the significance of transaction cost but, suggests the presence of unobservable buyer-specific preference.

Nested logit model overcomes the interdependent from irrelevant alternative (IIA) property that basic logit model is subjected to. This property is often explained in related literature using an example of two modes of transportation: car or bus (red) bus. The probability of choosing car is a half. However, if a blue bus is added to the existing set of choice, the number of choice has become three and the probability of choosing a car has decreased from a half to a third. Nested logit model sees both red bus and blue bus as a choice and account for correlation between the two buses. The probability of choosing a car will not change if the negative correlation between a red bus and blue bus is taken into consideration. Since logit model is unable to overcome the IIA property, it leads to mutual interdependence of error terms.

According to Train (1998, p. 230), the weaknesses of logit model and nested logit model are their restrictions on data. The assumptions necessary but unrealistic, for these models are:

(a) Homogeneity in consumers’ preference. That is, the coefficients of variables are assumed applied to every individual sample used. This assumption implies that every car buyer’s value the same car characteristics similarly;

(b) A change in a choice leads to changes in the probability of other choices proportionately. This assumption suggests that the first choice has linear substitution with all other choices; and

(c) Unobserved variables are independent over time for cases of repeated choice. This implies that utility derived from goods consumed at a time does not influence consumers’ decision in the future. This assumption is not realistic 106

especially in consumption of durable goods that take up relatively large proportion of consumers’ income. In the case of differentiated goods such as cars, brand and reputation of car makers, and experience from ownership of previous cars may have influence over consumers’ decision to repeat their consumption.

As such, studies began to revisit and estimate models that allow relaxation of the assumptions. Up to the 1980s, studies of car market have been focusing on the demand side, focusing on car demand at individual level or on at product level. Studies such as Berry (1994) and Goldberg (1995) argue that framework of past studies require assumption of functional form that render estimation intuitive. In addition, it is also argued that absence of “outside goods” in the past studies cannot derive a reliable aggregate demand function. Goldberg (1994, p.893) attempts to bridge the gap by applying disaggregate nested logit model and incorporate outside goods using demand side and supply side data.

According to McFadden & Train (2000, p. 448), mixed multinomial logit (MMNL) was introduced in 1980. MMNL is a multinomial logit (MNL) model with random coefficient that measures taste heterogeneity of population of MNL decision makers. The model allows more than two categories of dependent variable while explanatory variables may be binary or continuous. MNL logit model is considered desirable as it does not assume normality, linearity and homoscedasticity.

McFadden & Train (2000) applies MMNL to microeconomic consumer equilibrium analysis accounting for heterogeneity in unobserved variables. The unobserved variables are product characteristics and consumer characteristics that influence their preference. Estimates show that there are correlation structure in unobservable variables and variation in preference for selected car types by types of fuel consumed, size of cars and types of car body. In addition, study suggests that its results reflect the difficulty of determining factor structure of unobserved utility as well as model specification problems associated with omission of observed variables and interactions of the variables.

107

Although study does not show conclusive results of unobserved variables’ influence on car demand, the study shows that MMNL models are flexible and have computational advantage as this approach is able to detect effects of substitutability across choices.

Subsequently, Train (1998) estimate random-parameter logit (RPL) model, a MNL model extended to include vector of both observed and unobserved consumer characteristics in order to capture substitution pattern of fishing sites. Therefore, this approach does not exhibit the independence from irrelevant alternatives (IIA) property and does not restrict substitution patterns of data.

Bhat (2000) applies RPL model to account for heterogeneity in preference for urban transportation choice for work and estimates three models for comparisons of results. These models are multinomial logit model (MNL), deterministic coefficient logit model (DCL), an MNL model in which deterministic function of an observed consumer characteristics is specified, and random coefficient logit (RCL) model. Comparisons of estimation show that RCL model generates larger coefficient of variables, implying consumers’ greater sensitiveness to product characteristics and time values than the other two models. The study concludes that both observable and unobservable consumer characteristics have to be accounted for and that assumption of heterogeneity in preference may have serious implication on the effectiveness of policy.

In light of the implication of observed and unobserved heterogeneity of consumer characteristics, Green, Hensher & Rose (2006) attempt to study heterogeneity of consumer characteristics through decomposition of variance. Although the heteroscedastic mixed logit model used in the study is able to identify factors contributing to heterogeneity in observable consumer characteristics and to heterogeneity in unobservable characteristics. Although it is concluded that the model is able to explain consumer’s decision-making process better, the model generates intuitive estimates. As a result, there remain a gap to bridge in the area of identifying contribution of unobservable variables.

In the meantime, MNL continues to be used for study in other areas of study and development takes place to study phenomena observed in those areas. Hedeker (2003)

108 for example, applies a mixed-effects multinomial logit model to study the effects of homeless adults’ living conditions on their psychiatric conditions. The model is useful to study clustered and longitudinal data. Although Train (1998, p. 230) suggests that logit and nested logit models do not account for “memories” of unobserved factors over time for each households for example, the mixed-effects multinomial logit model can be used for longitudinal categorical data (Hedeker 2003, p. 1444).

Brus et al. (2016) applies mixed MNL logit model which is a variation in logit model, in ecology study. The study investigates relations of habitat types by vegetation with environmental variables such as storm and sea level over a period of about three years. It is concluded that environmental effects expressed as categorical variables, on growth of vegetation can be accounted for by adding random effects to the MNL.

In the study of factors that influence the level of crash severity, Zhen & Wei (2019) classify crash severity into five categories as probability function of eight identified explanatory variables in a MNL model. Results of MNL model generates intuitive results, suggesting that probability of fatal crash for example, is associated with adverse light condition and adverse drivers’ physical condition, higher speed, and road conditions. Study concludes the usefulness of MNL model in the area of study.

4.6 Application of Logit Model - Overview

Discrete choice modelling is an approach to explain or to predict a choice out of a set two or more discrete alternative based on the explanatory variables specified in the model. Although this approach has been more widely used in medical sciences research, its application gained ground since 1970s to study transport mode for policy making. In social science for example, this approach allows analysis of categorical nature of the car features and car buyers’ characteristics.

In the context of study of car market, discrete choice model is built upon a framework of rational choice. It is assumed that when confronted with a discrete set of choices, car buyers choose the option that maximises his/her satisfaction. Since cars are differentiated

109 durable goods, the utility derived from use of the cars is influenced by the car characteristics. Thus, utility of a choice is a function of characteristics of the car buyers and characteristics of the possible choices. The discrete choice model characterizes a function for population of car buyers so that the statistical inference for the functional parameters can be made. For this research, a discrete choice model is mapped on the data collected, containing data of choice of car makes related to the car characteristics and association of choice with characteristics of car buyers.

In discrete choice modelling, binary logit (logistic probability unit) is often preferred to binary probit (probability unit) because of the followings:

(a) although both logit and probit transformation generate sigmoid functions, probit requires integration while logit can be simplified to linear equation;

(b) the coefficients of logistic model, in terms of odds ratios, have simple interpretation;

(c) logistic model is closely related to loglinear model in that the estimates namely: intercepts, coefficient of independent variables, their standard errors and chi- squares, in logistic model are the same as the estimates in the loglinear model;

(d) logistic model has desirable sampling property that is, coefficient estimates are not subject to sampling bias problem (Allison 2012, p. 103); and

(e) logistic distribution approximates a normal distribution.

Page, Whelan & Daly (2000) applies discrete choice model to examine if car buyer characteristics, car characteristics, purposes of car use, sources of information and car buyers’ attachment to cars have influence over the choice of car buyers will buy in the near future. Results of the study shows the association of a set of car buyer characteristics with choice of car make and show how car buyers obtain new car information in the process of survey.

110

Following recognition of unobserved product characteristics’ influence on car demand in Lave & Train (1979), studies begin to take into consideration the unobservable characteristics. Berry (1994) proposes that unobserved product characteristics can be detected at product level as residual. It is stated that presence of unobserved product characteristics contributes to abnormal demand function (Ibid, p. 243).

Berry, Levinsohn & Pakes (1995) introduces a new approach treating cars as discrete goods, specifies utility function of using a car as additively separable into two parts, namely: product observable and unobservable characteristics and consumer characteristics. The study assumes that the distribution of car buyers’ underlying preference is influenced by their income and other social-economic factors. It is also assumed that the buyers’ preference does not vary across markets and time. Using discrete choice models with a vast amount of data and data of second choice, the random coefficient discrete choice approach taken in BLP (1995) allows products’ characteristics to be observed through the substitution patterns found at consumer-level heterogeneity in taste.

Since attributes that contribute to quality of cars may influence car buyers’ decision, study incorporates the cars’ attribute using discrete choice models. This approach allows incorporation of more micro information such as new attributes and thus, allows evaluation of impact of the change on the demand. Other studies allow presence of unobservable characteristics are Berry, Levinsohn & Pakes (1999), Petrin (2002), and Berry, Levinsohn & Pakes (2004).

Petrin (2002, p. 706) applies market level data alongside with micro data in discrete choice model to improve the precision of estimates. Mariuzzo (2002) and Berry, Levinsohn & Pakes (2004) estimate demand function using product level data to replace individual range of cars’ characteristics.

Although a component of residual, termed idiosyncratic error (ξ) (see utility function (1) in section 4.5) is treated as a random variable representing unobserved characteristics in several studies, Haaf et al. (2016, p. 183) argues that the component can be viewed as

111 merely a construct that improves fitness of the model. The study shows that a generalized method of moments that uses instrumental variables (GMM-IV) reduces coefficient bias.

While studies in car market continue to develop, similar approach is applied in studies of slightly different perspective. For example, Allcott & Wozny (2014, p. 794) studies demand for gasoline in relation to car demand using a logit model. Results of the study are consistent with the other literature that study gasoline demand using Bertrand-Nash model (Langer & Miller 2011, p. 2) and applying the approach of estimating reduced- form parameter (Busse, Knittel & Zettelmeyer 2013, p. 224). The study concludes that car buyers weight the cost of using cars in terms of the gasoline consumed over the period of car use and the price of cars.

Antolín, de Lapparent & Bierlaire (2016) attempt to study substitution effects of new cars using a newly proposed cross-nested logit model. In this model, it is assumed that an alternative need not be attached to a nest only but, may be found in other nest that is, an alternative may be one of the alternatives available in is a nest. Based on a relatively large sample of 20,000 observations, price of an alternative is changed in order to observe such change on the market shares of other alternatives in both logit and cross-nested logit model. The results show that when the price of an alternative is increased, the market shares of other alternatives increase, reflecting substitution takes place. Although this study is able to show substitution effects, the results are unable to show that cross-nested logit model is more reliable than logit model in capturing substitution effects.

Østli, Fridstrom, Johansen & Tseng (2017) applies a nested logit model to study car demand in the context of novel fuel demand and technology advancement over a period of 16 years for a very large pool of data. The study is able to show market’s response to price changes due to taxes and price changes due to increases in either operating or manufacturing cost at the supply. This study shows a wider and deeper use of logit model for social sciences.

Another development in logit model is the specification of dynamic component in the model. Dynamic logit modelling began in the 1980s according to Liu & Cirillo (2017, p. 870) and it has since been applied in some studies. When dynamic logit model is applied,

112

Liu & Cirillo (2017 p. 876) applied stated preference survey online to elicit car buyers’ preference in Maryland. Results show that the society is sensitive to fuel price and that there is preference for green vehicles in the future based on a sample of 3598 observations for a hypothetical time period of nine years.

4.7 Models

Since the demand for cars is the demand for a bundle of characteristics, the utility derived from the consumption of a car of range j by individual i is expressed as:

U(ζi, pj, xj, ξj; θ) (1)

where, ζi is a vector of individual i’s characteristics, pj, xj, and ξj are price, observable and unobservable characteristics of cars respectively, and θ denotes parameters that determine the distribution of the consumer characteristics.

The aggregate demand for cars as a function of their prices and characteristics is expressed as:

Aj= {ζ : U(ζi, pj, xj, ξj; θ) ≥ U(ζ, ps, xs, ξs; θ) for s = 0, 1, 2,…J} (2) where s = 0, 1, 2,…, J representing the purchases of competing substitutes.

Berry, Levinsohn & Pakes 2004 (as BLP 2004 thereafter) improves the approach BLP 1995 by introducing additional component in the model that is, the unobservable consumer attributes using second choice data. The unobservable consumer attributes are considered to exist if a similar range of car is chosen by some consumers who do not possess similar relevant observable characteristics.

BLP 2004 is used in this research because it allows culture differences in a multiracial country to be accounted for in the demand for cars. In Malaysia, Chinese who are generally the third and fourth generation in this country, benefit from their parents and

113 grandparents who are frugal. The wealth accumulated by the older generation allows the younger generation to be able to afford expensive properties and luxurious items such as cars that they may not be able to possess if they have to work on their own like their counterparts whose families do not accumulate as much wealth.

Corruption and cronyism in Malaysia also may contribute to the unobservable consumer attributes that may influence the demand for cars. The political economy of the country leads to the emergence of Malay elites who are politically well connected to gain benefits from government’s policies over the years (Gomez & Jomo 1999). Inclusion of the unobservable consumer attributes will also allow the model to capture the effect of corruption. However, the effect will be in aggregate terms.

In contrast with the utility derived from consumption of goods measured by quantity of the goods consumed, utility function of car owners in discrete choice modelling expresses satisfaction derived from the bundles or sets of car characteristics.

In BLP 2004, linear utility enjoyed by consumer i from consumption of car j. The linear utility function is expressed as a function of car’s observable characteristics xjk, cars’ unobservable characteristics ξj, and car owners’ personal preference that is independent of car characteristics εij. The discrete choice model for utility of cars is therefore expressed as:

~ uij= Σxjk β ik + ξj+ εij (3)

~ where β ik denotes car buyer i’s preference and k denotes car characteristics. Car owner’s i’s preference can be expressed as a summation of observable and unobservable car owner attributes as follows:

~ o u β ik = β k + Σzir β kr + β k υik (4)

where zir and υik are vectors of observable and unobservable car owners’ characteristics, r index observable car owners’ characteristics while k indexes the observable car

114 characteristics. Both βo and βu denotes car owners’ observable and unobservable characteristics respectively.

Using this framework, the substitution patterns may be traced from the interaction between consumers’ preference and the car characteristics. This requires that cars be differentiated by characteristics such that the substitution patterns can be observed. Due to the possibility of total large number of differentiated features, BLP 2004 includes only the characteristics that represent the most important sources of differentiation. Similar approach is taken in this research to identify car characteristics that car buyers indicated they considered and compared while making their buying decision. Although the range of car make in this study is smaller than that in BLP 2004, it is expected that not all of the aspects of differentiated features can be taken into account. Some of the differentiated features are not worth included for estimation of preference distribution because they do not contribute significantly to substitution. For example, variances in optional bundle of accessories available for different car makes of similar segment such as for and may not cause substitution because variances in price outweigh the preferences for accessories.

The unobservable car characteristics, ξ, may have a significant effect on the price of cars such as, the quality of the car, and this subsequently may have effect on the price elasticity of demand for the car. The definition of ξ also implies that the unobservable car characteristics will reflect the aggregate impact of characteristics not specified in the models. Therefore, at consumer-level, the model can be expressed as follows:

o u uij= δj + Σ xjk zir β kr + Σ xjk υik β k + εij (5) kr k where j = 0, 1,.., J; and

δj= Σ xjk β ik + ξj (6) k

115

The choice-specific constant δ is a function of car characteristics if car characteristics have a systematic effect on the car demand. To obtain the price elasticity of demand for a specific car make, price’s impact on δ has to be estimated.

The car characteristics included for estimation of β are the important characteristics that constitute to differentiation among the selected car makes. Studies have not proposed the optimum number of car characteristics to be included but suggest that there may be difficulties to estimate preference distribution that is sufficiently large to capture all aspect of production differentiation. Anderson, De Palma & Thisse (1989, p. 32) proposes that characteristics space have the dimension at least (n-1) where n is the number of variants.

At market level, consumer behaviour is the summation of the all of the individual consumer demand implied by their utility function. Therefore, aggregate demand for individual car make is the summation of the individual consumer attributes wi = (zi, vi, εi).

The proportion of spending units Pw that choose car make j is the integration of the fraction of the population that choose car make j corresponding to the market shares of car make j is expressed as follows:

o u sj (δ, β , β ;x, Pw) = ∫ Pw (dw) (7) o u Aj(δ, β , β ; x) where o u o u Aj(δ, β , β ; x) = {w: max [uir(w; δ, β , β ,x)] = uij}

The logit functional form for the choice probabilities conditional on consumers’ observable (z) and unobservable (v) attributes is expressed as follows:

o u exp (δj + Σkrxjkzir β kr + Σkxjkυik β k ) 1 Pr ( yi = j | zi, vi, θ, x) = ̶̶̶̶̶̶̶̶̶̶ ̶ ̶̶̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ ̶ (8) o u 1+ Σqexp(δj + Σkrxjkzir β kr + Σkxjkυik β k )

116

4.7.1 Estimation

Micro data of car buyers or spending units and cars allows estimation of δ, βo and βu for equation (5) or of β , βo and βu if additional restrictions are imposed on the joint distribution of x and ξ in equation (6). This study will not impose restrictions on both the observable and unobservable characteristics of cars because the number of choices of cars available is relatively small compared to the more developed and opened economies. In addition, it is found that the distribution of the car sales in Malaysia is skewed towards the top three car makers namely: Perodua, Proton and Toyota. In 2014, Honda outperforms Toyota by about 6000 units of their popular model sold but, the distribution of car sales remains skewed towards Perodua and Proton. For example, a total of about 65 per cent of the market shares are dominated by 4 car makes and a total of 9 models (KINIBIZ, Sept 29-Oct 12, 2015, p. 6).

Estimation of the vector θ = (δ, βo, βu) however, may not generate true price elasticities because car characteristics may not have systematic effect on car demand that is, β may be non-zero. Therefore, estimation of β is done by estimating equation (6) at product level to investigate if car characteristics have systematic effect on car demand. This requires some assumptions of joint distribution of the products’ both observable and unobservable characteristics (x, ξ). This study follows BLP 2004’s approach to assume that ξj are mean independent of the non-price characteristics of all different car make.

The steps involve in the estimation of variables have been described as below:

(1) Vector θ = (δ, βo, βu) is estimated using the moments estimator methods as it is done in BLP 2004. Using this method, the moments predicted by the model are compared to the sample’s moments. The vector that gives minimum deviation will be chosen;

(2) Three sets of predicted moments are matched to the data analogues set:

(a) the covariance of the first choice observable product characteristics, x, with the car owners’ observable attributes, z, to obtain βo, and 117

(b) the covariance between the first choice’s observable characteristics and second choice’s observable characteristics to identify car owners’ unobservable attributes, βu, that contribute to the buying decision made. If all relevant car owners’ characteristics can be observed, then βu=0 while βo will determine first and second choice car characteristics and correlation between them.

The choice-specific intercept (in equation (5)), demand function for the characteristics of the cars is written as:

K δj= pj β p + Σ xjk β k + ξj (9) k≠p

According to BLP 2004, a firm producing single product will have a mark-up that is the inverse of the semi-elasticity of price elasticity of demand. In Malaysia’s context, foreign car makers that have been operating in the country, including Perodua, Toyota, and Honda are treated as multi-product firms as large range of their products are available in the market. As direct import of other car make leads to limited range of their models available, assumption can be made regarding their price elasticity of demand. Due to excess supply of cars in the world car market, it is reasonable to assume that the supply of foreign cars in Malaysia to be 3 as it is assumed in de Melo & Tarr (1992).

o The coefficient of interaction terms ( β kr , see equation (8)) in nonlinear models is not interpreted similarly as that in linear models. Ai & Norton (2003, p. 123) says that the conditional mean of the dependent variable (y) can be expressed as the standard normal cumulative distribution function (Փ). The interaction terms of nonlinear models (x1x2) are different from that in linear models because of the following:

o (a) There may be non-zero interaction effect while the coefficient ( β kr ) is zero. Due to the nonlinear function, there may be a point where at the turning point of the curve or function, the slope is zero. The coefficient may be positive or negative beyond the turning point;

118

(b) significance and relative importance of the interaction effect is not directly reflected by the coefficient of interaction terms and it cannot be tested using t-test. In short, the coefficient of interaction term is not constant throughout the function;

(c) independent variables influence the interaction effect. Interaction effect is the effect of subsets of independent variables’ effect on dependent variable. For example, interaction of price and horsepower influence the demand for Mercedes Benz. Correlation between price and horsepower arises because the size of engine influences the cost of manufacturing and hence, influence the price of the cars. Since the overall demand for a car make is influenced by the price and characteristics of the car, it may be found that the subset of car characteristics interacts with price, and influence the demand for different car make at different degree; and

(d) signs of interaction effect may change for different covariates. The change of signs reflects nonlinear nature of the function and implies changes in the importance of interaction effect at different level of the car demand in this research.

To estimate price elasticity, income elasticity, and cross price elasticity of demand, the point of estimation is identified by the average price and/or average of car characteristic(s) that is or are found to interact with price. For example, in the case when horsepower is shown to interact with car price for Perodua cars, then average price and average horsepower will be used as the points to identify price elasticity of demand for Perodua.

The estimated elasticity, β , for different car make is then used to estimate the cost of protectionism. As horsepower of cars is one of the important factors that determine road tax and car insurance, samples are classified by their makes and horsepower for the purpose of calculating price elasticity of demand.

The method applied for estimating the logistic model is Maximum Likelihood (ML) method because the consumer-level data is grouped data and the dependent variable car make, is dichotomous.

119

4.7.2 Sample

According to Allison (2012, p. 20) Maximum Likelihood (ML) method in estimating a logistic model is more superior than the other two methods1 used in Discrete Choice Analysis because of the following:

(a) The ML estimators are consistent, asymptotically efficient and asymptotically normal, and

(b) Sampling distribution of the estimates is approximately normal in large samples.

Due to the second feature of the ML estimators, large sample size is preferred to small sample size. However, there is no study that recommends an optimum sample size. The model applied and sample size selected by selected studies applying Discrete Choice Model are summarised in Table 4.1 below.

The table shows that studies such as BLP 1995 and BLP 2004 use very large sample sizes because primary data is readily available. In addition to large car market in the United States, the number of car makes and models are far larger than the numbers choices available in Malaysia. Other studies use small sample for example Petrin 2002 and Berkovec & Rust (1985), to study a segment of car industry.

There are no secondary sources of data related to individual purchases of new cars and demographics of car buyers available in Malaysia. A few studies use large samples (see Table 4.1) probably due to the availability of data. Other studies use relatively very small sample but, there has not been explanation on how the sample size has been determined.

Sample size matters in regression analysis because large sample size may reduce the effect of outliers or improve the accuracy of estimation. However, in logistic regression, sample size does not matter because estimators are expressed in probability as far as outliers are concerned. However, large sample may allow a modelling of car market using disaggregate data. Hence, improves the accuracy of estimates for individual car makes’ various models available in the market.

120

Table 4.1 Summary of selected literature on car market: Logistic Regression Models and sample sizes (in chronological order), and comparison with studies in related areas Studies Models | Countries | Population Sample size Time size 1 Gu et al. 2019 Mixed logit | Austria | Not 203 2017. reported 2 Orlov, A & Multinomial logit | Norway Not 1,093 Kallbekken, S | 2017. reported 2019 3 Bansal, Daziano & Logit-mixed logit | German Not Panel 1: 500 Achtnicht 2018 | 2007-2008. reported Panel 2: 2000 4 †Moon, Shum & (Standard) Logit | U.S. | 10,600,000 37,500 Weidner 2018 1973-1988. 5 Irawan et al. 2018 Ordered-response logit | 12,497,072 336 | 2005-2010. 6 Jørgensen, Logit | Norway | 1985- Not 35,000 Mathisen & 2013. reported Pedersen 2016 7 Train & Winston Mixed logit | U.S. | 1970- 250,000 458 2007 2005. 8 De Borger & Logit | Belgium | 2000. Not Sample Mayeres 2007 reported proportion 9 *Greene, Hensher Multinomial logit & Mixed Not 223 & Rose 2006 logit | Australia | 2003. reported 10 Berry, Levinsohn Logit | U.S. | 1993. 10,600,000a 37,500 (203)b & Pakes 2004 11 Petrin 2002 OLS logit, Instrumental 30,000 337 (4)c logit | U.S. | 1981-1993. 12 Berry, Levinsohn Simple logit, Instrumental Not 2,217 (997)b & Pakes 1995 logit | U.S. | 1971-1990. reported 13 **Imbens & Probit | Netherland | 1977- Not 347 Lancaster 1994 1983. reported 14 Berkovec & Rust Nested logit | U.S. | 1977- Not 237 1985 1978. reported 15 Manski & Herman Multinomial logit | U.S. | Not 875 1980 1967-1976. reported

Notes: † Study uses the dataset used in BLP 2004

121

For comparisons of sample size: * study of socio-economic characteristics and product characteristics using logit ** alternative model, Probit - study using micro level and macro level data a Total vehicles sold in the United State in 1993. b In brackets, total number of models in the car market. c The sample size of four models: , station wagons, sport-utility and full size vans are 120, 63, 131 and 23 respectively.

The benchmark for this research’s sample size is determined using the formula for sample proportion developed by Cochran (1963, p. 75) according to Israel (1992, p.3). The sample for proportion formula expressed as follows:

n = 2 � � � 2 � where n is the sample size, Z is the abscissa of a normal distribution at 93 per cent confidence level, p is the proportion of the individual car makes’ market shares, q is (1- p), and e is precision level.

There are about 45 makes of car listed in the Malaysian Automotive Association’s survey of non-commercial cars sales performance in the country since 2012. A total of five major car makes identified and specified directly in Discrete Choice model by their name. They are Perodua, Proton, Toyota, Nissan, and Honda. Figure 4.2 below shows that these five car makes contribute up to 83 per cent of the total market shares based on 2-year average yearly sales in 2012 and 2014. The other 17 per cent of the market shares are made up of other Asian, American or European car makes.

The sales of high range car for year 2012 are summarized in Table 4.2. Similarly, in the high range segment, two relatively large car makes dominate the market, namely BMW and Mercedes Benz.

122

Figure 4.2 Market shares of major car makes

35.0 30.0 25.0 20.0 15.0 10.0 5.0 0.0

Source: Malaysia Automotive Association 2013.

Table 4.2 High range car makes - Quantity and relative market shares: year 2012

Quantity Relative shares (%)

BMW 6318 32.5 Mercedes Benz 5905 30.4 Chevrolet 2026 10.4 1471 7.6 Audi 1414 7.3 Volvo 937 4.8 643 3.3 Porsche 395 2.0 341 1.8 Total 19450 100.0

Perodua (No 1) 195579

Note: Sales of all high range car makes are relatively very small when compared to Perodua’s sales, the largest sales since 2006.

Source: Malaysia Automotive Association 2013. 123

All of the high range car makes originate from the United States and Europe except Lexus which originate from Japan. Lexus is combined with the American and European category because the population and hence sample sizes of the high range cars are relatively very small. In addition, Lexus is a high range car make targeting the American and European markets. For the purpose of estimating price elasticity of demand for high range car, all the major high range car makes are therefore, aggregated.

Table 4.3 Sample size determination: Low and middle range Market Shares Estimated sample size, n Car makes and model(s) (two-year

percentage e = 0.05 e = 0.07 e = 0.10 average) Perodua 30 316 137 56 Models: Myvi (40%), Alza, Axia Proton 22 260 115 46 Models: Saga (50%), Persona, Exora Toyota 15 196 85 35 Model: Vios (40%) Honda 10 138 60 25 Models: City (45%), Jazz Nissan 6 100 37 15 Model: Almera (50%) Total 83 1010 434 177

Notes: 1. Car makes are arranged by their market shares in descending order. 2. Proportion of sales for best-selling models are in brackets. Sales of other models are not available. 3. Best-selling models’ market shares are based on each individual car makers’ 2014 total sales.

Source: KINIBIZ 2015.

124

Market shares of the high range cars are also highly skewed. For example, about 60 per cent of Mercedes Benz’ total sales in 2018 is contributed by a single premium sedan model produced locally (Lim 2019) while plug-in hybrid models accounted for about 56 per cent of BMW’s sales in Malaysia (Tan 2018).

Statistics show that sales are highly concentrated on a small number of car makes and within the individual dominant car makes, their sales are also highly concentrated on a single model (See Table 4.3).

The estimated sample sizes of the major car makes based on their respective average market share for years 2012 and 2014 at confidence levels of 95, 93, and 90 per cent are shown in Table 4.3 below. Since car sales are highly concentrated by car makes and the model within the car makes, this research chooses significance level of 73 per cent. The targeted sample size of each of the major car makes are proportionate to their respective market shares.

The top five higher range cars are BMW, Mercedes Benz, Chevrolet, Audi, and Lexus with their market shares range from 1.1 per cent to 0.2 per cent. At 93 per cent confidence level, the sample size for BMW and Mercedes Benz that capture about 1.1 per cent of the market share is seven while the sample size for the other three car makes is two. These estimated sample sizes are used as a guide that is, as minimum sizes to obtain.

Using the table in Hsieh (1989, p. 797), at five per cent significance level, (1-β) = 0.7, for correlation coefficient of 0.2 and 0.27, the sample sizes are 400 and 386 respectively. The correlation coefficient used for estimation of sample size is based on the observation of correlation matrix for car characteristics and social-economic factors specified in the models. Estimation of sample size shows that the higher is the correlation coefficient, the smaller is the sample size. Since correlation coefficients for a few variables are smaller than 0.27, larger sample size is chosen. Comparisons of sample size using the approach for linear regression and the approach for logistic regression (see Table 4.3) show that sample size of 434 is acceptable.

125

To ensure that samples are unbiased representatives of population, car service centres are selected from 13 states and the Federal Territory for distribution of questionnaires. The names, addresses and contact numbers of service centres found in the list of car service centres on the internet. Each selected service centres are contacted either via telephone or email to explain the purpose of survey and how surveys are carried out. Upon the approval of their respective manager, hardcopies of 20 to 30 sets of questionnaires together with a self-addressed and pre-paid envelopes are sent to car service centres across the states. Another round of selection had to be carried out because a few selected service centres declined the invitation. In the service centres, car owners who send their cars to the service centres were invited to participate in the survey on the spot.

A few organizations such as government offices, private business organization were also selected. Due to the nature of their economic activities, the number of questionnaires sent were between five to fifteen sets only.

In addition, online questionnaires are sent to institutions after permissions are granted by their respective department heads. Although this approach is cheaper and less time consuming in than the distribution of hardcopy of questionnaires, online survey is less effective than self-administered survey. The response rates of online survey and on-the- spot survey are less than 10 per cent and about 802 per cent respectively.

The primary data is collected is cleaned and coded prior to analysis. To minimize missing data problem, some respondents who left their contact numbers for the purpose of lucky draw are contacted via telephone to obtain the missing data. Respondents who provided less than 70 per cent of the data were excluded from the database for analysis.

4.7.3 Data and variables

A priori studies at aggregate level that use representative consumer models define the demand for car as quantity of cars registered. In time series analysis, the demand for car is found influenced by real income, price, and taste factor (Suit 1958, p. 269; Farrell 1954, p. 188).

126

At disaggregate level, a priori studies define the demand for cars as the value of usage that is, depreciation. The results show that the demand for cars is a function of price and income. Other factors that influence the car demand are the age of spending unit, numbers of dependent, location, and race (Turnovsky 1966, p. 267; Bandeen 1957, p. 248; Bennett 1957, p. 846).

Following Kelvin Lancaster’s proposition that consumers demand goods for their characteristics, Cragg & Uhler (1970, p. 388) incorporate cars’ characteristics such as comfort and aesthetic characteristics in a logit model. However, such characteristics are not directly observable and thus, values are assigned to the cars in the study.

Unlike the previous studies that have to generate reasonable substitution patterns in logit model, Berry, Levinsohn and Pakes (1995, p. 851) relates car characteristics with observed consumer characteristics using both product-level data and aggregate consumer- level data. The results of this study show that car owners are responsive to the oil price changes and efficiency of the cars and that higher income group is less responsive to the oil price change.

Cross section data of prices and the demand for the selected car makes’ new car registered during the years 2012 to 2015. The demand for car in this research is the demand for the characteristics of the cars. Socio-economic factors, car buyer characteristics and car characteristics cited in the other studies are included in the initial draft of questionnaires. Factors and characteristics that are not selected by respondents in pre-test and pilot are eliminated from the questionnaires.

The observable household characteristics in a priori studies are the gross yearly income, number of dependents, age of dependents, age of car owners, and family size. A few household characteristics not commonly found in many studies but included in this research are gender, race and sector of work. Gender is included because it is observed that mothers often play the roles of driving children to and back from school.

127

Race and sector of work are included for test of their association with car demand because Malaysia has the highest civil servant to population ratio. Table 4.4 below shows that, the ratio of civil servant to population is 1 to 19.4 which is about four times the ratio in Singapore and six times the ratio in Indonesia.

Table 4.4 Ratio of civil servant to population: Selected countries Country Ratio

Malaysia 1 : 19.4 Singapore 1 : 71.4 Indonesia 1 : 110 Japan 1 : 28 1 : 118

Source: The Star Online, 3/2/2017.

Refer to Table 4.5, out of the 1.6 million civil servants in the country, 78.8 per cent are Malays compared to about 11 per cent are other natives while Chinese and Indian make up to about five per cent each group. Therefore, guaranteed loan approval for civil servants and other incentives given to civil servants will likely to be captured by race (RC) or sector of work (SEC) or both. In the case of both variables show statistical significance in influencing the demand for car, one of them will be dropped to reduce collinearity problems.

Table 4.5 Composition of civil servant: Ethnic groups Ethnic group Per cent of civil servant (%)

Malays 78.8 Sabah and Sarawak natives 10.9 Chinese 5.2 Indian 4.1

128

Source: The Star Online, 3/2/2017.

A variable not found in many studies but, tested in this research is number of car in household (NOS). Burns & Golob (1976, pp. 177-178) specifies direct utility functions of households that have different number of cars owned in multinomial logit models. Results show that households consider the choices of destination are influenced by the distant travelled and transport alternatives that is made up of choices of public transport and number of cars owned. When the number of cars owned is accounted for, the results also suggest that there are more social-economic factors and/or car characteristics that influence car demand (Ibid, p. 191) in addition to income, cost of car ownership, and travel time.

According to Lave & Bradley (1980, p. 380), there are empirical evidence that show positive relationship between number of cars owned and the chance of buying a small car and that education influence the choice of car bought. When the demand for cars is segregated by class of domestic and foreign, higher level of education is associated with demand for foreign cars due to attitude differences caused by education.

Ben-Akiva, Manski & Sherman (1981, p. 404) suggests that the ease with which transport mode can be used to reach the desired destination, for work and/or for leisure are influenced by the number of and types of cars owned and availability of public transport. In recognition of the weaknesses of specifying aggregate demand models for car market and weaknesses of econometrics tool to model car buyers’ choice, the study suggests application of discrete choice model.

Malaysia has 211 and 225 cars per 1000 persons in 2002 and 2003 respectively (Econ Stats, viewed 15/10/2018) during which the poverty rate was about 6 per cent. At the average of about five persons to a car in Malaysia and there are many households that live in poverty, mainly rural peasants who have relatively larger family size than households in the cities and urban, it is reasonable to deduce that many households have more than a car. Therefore, this research attempts to find out if the number of cars owned by a household may be influencing the demand for relatively cheaper domestic cars or if the number may influence the demand for certain car characteristics.

129

In addition to income (Y), car buyers’ characteristics included in the tests of correlation with car characteristics are the number of dependent (DEP), gender (GEN), race (RC), the sector (SEC) car buyers are working in, namely: private and public, and number of cars in the household (NOC).

Due to the lack of reliability of the public transport, private transport has become essential, therefore, location measured by urban, suburban and rural, is no longer a variable in this research. Car characteristics measured by payload is eliminated from the questionnaires because it is observed that light trucks particularly twin-cab light trucks are used as passenger cars instead of for transporting goods. A new proxy that is added to this research is car size that is measured by the number of .

Due to the oligopolistic nature of the market, the number of power accessories and number of safety features tend to be similar among the major car makes, car makes may be differentiated by the number of service centres or customer services as a few respondents indicated during the pilot test. Therefore, these two variables are included in the questionnaires as a characteristic that car buyers may consider part of the package that comes with acquisition of new cars.

The common observable car attributes are the average price of the model of the cars, horsepower (HP), number of passengers (NS), number of power accessories (PA) such as power windows and power doors, numbers of safety features (SF) such as antilock brakes and , and dummy variables (UD) for multi-purpose vehicles and sport utility vehicles.

Since the common observable features are based on the studies in developed nations during the period 1990s to early 2000s, this research will include more car features to capture car features that may not be considered in the developed countries and to capture the influence of technological change on car features over the years. To ascertain the relevance of car characteristics in Malaysia today, car characteristics identified in a priori studies are mapped to the car characteristics published in the media or in brochures available in the market. Although there is relatively large amount of information given

130 in any brochures available in the Malaysian market, most of the information is not directly measurable by car buyers such as such as minimum turning radius. Information that car buyers may not directly observe due to insignificant differences between the first choice and the alternative car makes are like kerb weight and dimensions.

Differences are found in the terms used in specifying car characteristics. Some terms are merely technical terms for similar characteristics while other such as fuel economy is different because of the unit used. In many studies, fuel economy is expressed as miles per gallon while in Malaysia, this piece of information is not provided by car makers, hence, not directly used for comparison. However, fuel economy is often measured as kilometre per litre and this information is mostly available in the website of car enthusiasts where car owners share their experience.

Following the mapping process, it is found that payload of large cars such as light trucks and sport utility vehicles, is not published in Malaysia and hence, likely not commonly taken into consideration when car buyers made their buying decision.

A pre-test was conducted on a small group of car owners consisting of three males and three females. The sample size for pilot test is selected in an ad hoc manner. The new features identified are then added to the questionnaires for second round of pilot test. Subsequently, a pilot test is then conducted at a local car service centre so as to include a larger pool of car owners of more diverse background. This is to ensure that if there exist car buyer characteristics and car characteristics, not cited in a priori studies but, are unique to the Malaysian car market, these characteristics can be identified and be taken into account.

There was no additional feature added to the questionnaires following the finding of the second pilot test carried out at a car service centre, implying that the list has all the features that car owners may consider while making their buying decision. Subsequently, the finalized questionnaires are translated to the Malaysian language. English and Malay language questionnaires are made available to car owners to minimize missing data problem and low response rate due to language problems.

131

Data collection approach taken is random across the country by distributing survey forms at car service centres that have agreed to distribute the questionnaires. Packets of 20 to 30 questionnaires were sent to service centres and telephone calls were made to follow up. Respondents are car owners who sent their cars to the service centre and willing to answer the questions while waiting at the service centres. In order to meet the targeted sample size, additional packets were sent out to service centres not selected earlier when three service centres sent their questionnaires back with zero response rate.

Due to the nature of the society that tends to response in groups, 5 to 10 questionnaires were sent to schools, offices, churches and business organizations for samples following approval granted by their heads of administration department. The average response rate of self-administered survey is approximately 65 per cent.

Online questionnaires were sent to universities and offices to minimize cost of data collection. This approach gives response rate of less than 5 per cent. This data collection method is proven to be far less effective than the use of hard copy questionnaires in this research.

Malaysian car industry can be characterised as oligopolistic because the concentration ratios of the top five largest players highly concentrated are 81.1 and 82.7 per cent in year 2012 and 2015 respectively (refer to Table 5.2, p. 165). Figure 4.2 shows that market is dominated by Perodua, Proton, and Toyota in 2012. In 2015, both Honda and Nissan’s market shares improved by 8.6 per cent and 1.3 per cent respectively while Toyota’s market share decreased by 2.7 per cent. Such high level of concentration is attributable to excessive trade barriers to protect the domestic car makers and to the possibility of comparative advantage Toyota and Honda gain from history. Therefore, the targeted sample size is 434 at 93 per cent confidence level.

Lack of clear consumption patterns such as comparisons and buying of cars of the similar segments in Malaysia make it difficult to separate higher range cars from the higher end of the medium range cars in analysis. Berry, Levinsohn & Pakes (1995, p. 852) shows that cars have non-zero cross price elasticity of demand only with other cars within their

132 segment. This implies that car owners of higher end cars are very unlikely to substitute higher end cars with lower and medium range cars when prices changed.

Therefore, in the first stage of analysis, classifying samples into groups involves identifying the dimensions of discrimination (Hair et al. 2006, p. 285) between higher- end cars and non-higher end cars. There is no separation made for the aggregate of medium and lower range cars because they constitute to more than 50 per cent of the car market shares. Separation of these categories of cars may be difficult and meaningless because many households in Malaysia own more than a car, implying that medium income households may buy relatively cheaper cars for a certain use such as basic transportation but, buy more expensive cars for other reasons or for their unique characteristics.

The second stage of the analysis involves identification of car characteristics that differentiate various car makes. Berry, Levinsohn & Pakes (2004, p. 72) suggest that only characteristics that are distinctive to each car makes are included in the model but, has not explained how the selection process is carried out. In this research, the number of times each characteristic is selected relative to individual car makes are calculated. As calculations show that all car characteristics are important, all of the characteristics identified are included in the model.

4.7.3.1 Selection of car characteristics

For the purpose of analysis, only characteristics that differentiate various car makes are included in the model (Berry, Levinsohn & Pakes 2004, p.72). However, a priori studies have not suggested how selection can be done. Due to the time lapse between a priori studies and this research and also due to the differences in the technology in developed nations where a priori studies were carried out and in Malaysia, more car characteristics were added to the list of characteristics in the questionnaires in addition to those identified in some literature.

133

Although survey shows that the car characteristics that buyers considered before making their buying decision have not been very different from the car characteristics identified in the other studies carried out in the United States in general, some characteristics are not specified and measured similarly in Malaysia. For example, comparisons of the number of power accessories and number of safety features across different car makes are difficult to be accounted for specifically in this research.

Due to stiff competition in an environment under the heavy influence of trade barriers, car makers cater to buyers’ wish in domestic market by allowing different combinations of features. For example, two less air bags than the full specification but with additional accessory of other types at slightly lower price. Toyota Fortuner for example, offers up to five combinations of accessories that are optional to individual car buyers. Such large combinations of optional accessories offered by different car makes will require far larger sample size that is not feasible at the time this research was undertaken. Although Berry, Levinsohn & Pakes (2004, p. 72) acquires a sample size of about 37,000 observations, the study does not account for all options to accuracy of results because the mass of data will be unmanageable. Situation in this research is opposite that in the a priori study. Due to the sample size and lack of data, this research will not be able to capture the effect of accessories’ options in car demand.

Among all car characteristics identified in pre-test and included in pilot test, characteristics selected for the models are horsepower (HP), kilometre per litre petrol (KML), car size measured by number of passenger (NS). These characteristics are found statistically significantly influencing the demand for car either directly or through their interaction with other car characteristics.

Prior to performing logistic analysis of car characteristics in this research, each car makes’ characteristics are individually analysed. To identify the differentiating characteristics, the relative importance of the characteristics is calculated by obtaining the ratios of frequencies of characteristics selected over the sample sizes of individual car makes. For example, if the number of power accessories is considered by 30 out of 65 respondents who bought Toyota, the difference between ratio of 0.46 and the ratio for Honda for example, will indicate if the number of power accessories is a differentiating

134 characteristic. The larger is the difference, the greater is the level of differentiation of the car characteristics.

Simple ratio analysis shows that possibly due to their history in the Malaysian car market, Honda and Toyota are often compared for their fuel efficiency, power accessories, prices of parts and safety features. Due to the lack of differences in the number of characteristics across various car makes, “specification” (SPEC) of car is used as a proxy for “package” of characteristics offered. In Malaysia, the variant offered is either “full” or “not full”. The latter is a package that reduces certain features such as the number of air bags or omission of certain accessories targeting car buyers who prefer to pay less for features that are seen of little value. For example, a full specification cars may be slightly more expensive than the similar model but without full specification, in which the latter may not have Bluetooth.

The simple ratio analysis shows that despite technology changes and the time lapse between studies done in the United States and this research, the car characteristics that car owners in Malaysia consider when making their buying decisions have not been significantly different. These car characteristics are horsepower (HP), safety features (SF), power accessories (PA) and number of passengers (NS). Although fuel efficiency measured in kilometre per litre petrol (KML) is found an important characteristic considered in the a priori studies, it has not been a significant characteristic influencing the buyers’ decision in Malaysia based on simple ratio analysis. This may imply that car buyers do not take into consideration petrol price because it is controlled by the Malaysian government.

In Cragg & Uhler (1970, pp. 387-8), services of a car include the level of comfort and aesthetic value that contribute to utility. European cars particularly are often viewed as luxury cars that carry aesthetic and ostentatious value in Malaysia. However, these values are difficult to identify and measure. In this research, a dummy variable is assigned to cars that are generally considered to be a luxury in Malaysia. They are car makes that are targeting the higher income group such as Audi, BMW, and Mercedes Benz.

135

A problem associated with random sampling is the observation of small car makes such as Audi, Volkswagen, Peugeot, Suzuki and Mazda that constitute to less than 1 per cent of the Malaysian car market. The sample size of these car makes is as small as one each reflecting their actual market shares. Omission of small samples is not appropriate although estimation of their demand function is infeasible due to extremely small sample size. Therefore, these car makes that constitute to less than 5 per cent of the market shares are aggregated and classified under two categories: European cars or small Asian cars following Berry, Levinsohn & Pakes (2004, p. 89)’s treatment of small players.

4.7.3.2 Missing data and false data

Prior to keying in data, returned questionnaires are examined for missing data. A few samples are dropped because 30 per cent of the data were missing. Although questionnaires show a dotted line separating questions and respondents’ contact number, and it was stated that respondents are to detach the slip at the end of the questionnaires before returning the questionnaires, some respondents did not detach the slip. To minimise missing data problems, respondents who left their contact numbers for lucky draw purposes and did not detach the slip were contacted to obtain missing data. Respondents who were contacted were informed that they may choose to answer or not to answer.

As a result, the samples excluded from database of this research consists of respondents who omitted up to 30 per cent of the data and did not provide contact numbers for lucky draw or detached the slip containing contact numbers. The action taken was effective as the number of samples provided more than 70 per cent of the data increased by up to ten samples. According to the respondents, the reasons for missing data were the overlooking of the questions while attempting to complete questionnaires in a hurry while a few said they could not recall the information on the spot.

Some missing data like prices and fuel consumption of cars can be found on the internet. Although the information found on the internet is not specific to the individual owners’ car, it helps to reduce missing data problems.

136

Subsequently, every questionnaire returned is checked for false data. A questionnaire is suspected of providing false data when the information given is not consistent with the findings of the a priori studies. For example, a car owner bought a Proton car at the price RM35,000 compared Proton car with Mercedes Benz car that cost RM300,000 as second choice. There are three samples found that shows such inconsistency and are excluded from the database.

A small number of respondents were found providing information that is not consistent with a priori studies but, the samples are included in the database. The inconsistency is found in the correlation between income levels and prices of cars. Due to data collection in small groups of up to five respondents within an organization, it is relatively easy to confirm that such inconsistency is acceptable. For example, a car owner of 26 years old, with income level range from RM2,001 to RM3,000 per month bought a new MINI Cooper car that cost RM199,000. Respondents of such characteristics can be correlated to high end car features because of wealth accumulation of the older generation. Since the wealth of the older generation is not observable, this characteristic may be identified as “unobservable” characteristic.

Due to the highly competitive environment and price distortion, car makers may omit a certain characteristic but, add a new characteristic. For example, both Toyota and Honda offer two air bags but, Honda offers keyless feature in addition to the air bags while Toyota offers multimedia touchscreen head unit. Therefore, using a number of characteristics such as the number of power accessories can be misleading. In this case, the number of power accessories may not be sufficient to proxy preference. However, car buyers’ preference for different power accessories may be captured by the car buyers’ unobservable characteristics.

Other problems encountered were the car owners’ inability to recall the number of features, and other details used for comparisons such as measures for efficiency of the car engines. Most car owners however, are able to recall which characteristics they took into consideration when making their buying decisions. These car characteristics are power accessories, kilometre per litre petrol, safety features, and horsepower.

137

Qualitative characteristics that some respondents took into consideration when making their buying decisions are perceived quality of parts, the access to reliable service centres and perceived customer service quality. Although it is not the objective of this research to measure the qualitative characteristics, their significance may be captured by the unobservable car characteristic component. Therefore, only measurable car characteristics tested in the a priori studies are included in this research.

4.7.3.3 Demographic of data

Federal Territory records the largest number of cars per square kilometre and the largest car population in Malaysia. The number of cars per square kilometre in the Federal Territory and Sarawak are approximately 12,732 cars per square kilometre and 5 cars per square kilometre respectively in 2011. The car population in the Federal Territory five times the population of car in Sarawak (http://www.motortrader.com.my/news/malaysia- s-vehicle-population). Such great differences are due to the location of the country’s capital city in the Federal Territory while Sarawak is an agriculture state. Since the Federal Territory is located within the state of and, most people and respondents who work in the Federal Territory live in Selangor, the sample of cars registered in the Federal Territory and Selangor are aggregated.

Statistics in Table 4.6 shows that the largest and second smallest sample observations are collected from Sarawak and, Selangor and the Federal Territory respectively. Such inconsistency of sample size with car population size is inevitable because samples are collected randomly through car service centres located across the country. Response rates of self-administered and online surveys are as low as zero per cent in the first round in Selangor, the Federal Territory and, Johor. Contrary, response rates for self-administered and online survey are about 90 per cent and 0 per cent in Sarawak. As a second distribution of questionnaires in most the states in West Malaysia gave no response, there was no third attempt to survey the other states while survey continued in Sabah and Sarawak. Consequently, sample size is disproportionately distributed across different states.

138

Table 4.6 Sample size by states

State Sample size Sample proportion Sarawak (SW) 245 54.8 Sabah (SB) 56 12.5 Terengganu (TR) 38 8.5 Pahang (PH) 32 7.2 Penang (PG) 30 6.7 Johor (JH) 24 5.4 Selangor & Federal Territory (SK) 19 4.2 Other states (OS) 3 0.7 Total 447 100.0

Figure 4.3 overleaf shows classification of car makes by buyers’ income group (YGRP). Both Proton and Perodua capture the largest and second largest market shares in the lower income segment respectively. However, Perodua dominates the middle-income segment. Proton’s market share in middle-income segment remains larger than that of Toyota. Toyota contributes to the largest shares of the high-income segment, followed by Honda. Toyota’s largest sales is in the middle-income segment, followed by high-income segment. Honda’s largest sales is however, in the high-income segment, followed by middle-income segment. Both Toyota and Honda’s sales in the high-income segment suggest that competition between the two old foreign car makers is stiff in this segment.

Unlike the other higher range cars, BMW is non-zero in this lower income segment. Market shares of all other higher range cars are non-zero in the middle-income segment. Such unusual observation may be due to the culture of the Asian society in which the older generation accumulate wealth for the younger generations, unreported income or other reasons not known.

In the middle income segment, Perodua and Proton dominate the market, followed by Toyota. Sample statistics show that both Toyota and Honda in descending order, dominate the higher income segment while higher range cars have their largest market

139 share in the higher income segment. Although both Proton and Perodua target the lower income segment, they have non-zero market share in the higher income segment.

Figure 4.3 Car makes by income groups

Notes: 1. “H”, “L”, and “M” represent income groups: “high income”, “low income”, and “middle income”. The sample sizes of three income groups are 81, 150, and 185 respectively. 2. “Missing” shows missing income data of various car makes.

The sample size and basic characteristics of major car makes of lower to medium range and the higher range are summarised in Table 4.7. Statistics show that Proton and Perodua offer cars of lower horsepower at lower price levels. High range cars generally differentiate themselves by their horsepower, with mean horsepower of more than 2. The

140 mean price of the best-selling Perodua cars, Myvi is approximately RM44,000 while the mean price of the best-selling Proton cars, Saga is approximately RM30,000. However, the mean prices of Perodua and Proton reflect otherwise because while large proportion of Perodua sample is made up of Myvi, the best-selling model, while Proton’s sample although made up mostly by their best-selling model, Saga, also have a number of more expensive models such as Iriz and Exora. Thus, the estimated average price of the sample is greater than the average price of a Proton Saga.

Table 4.7 Samples: Mean price, power and sizes Make Mean Price (RM) Mean HP (’000cc) Quantity of car Audi 112,676.1 2.2 1 BMW 230,069.6 2.3 6 Chevrolet 98,532.8 2.4 4 Honda 93,127.9 1.7 35 Lexus 183,921.9 2.7 4 Mercedes Benz 284,597.7 3.5 3 Nissan 97,568.4 1.8 21 Perodua 44,655.9 1.3 154 Proton 51,612.6 1.5 132 Toyota 93,820.0 1.7 65 Other Asian cars 99,112.0 1.9 15

Notes: HP denotes horsepower. Summary statistics do not include car makes that constitute to relatively very small proportion of the lower to medium range and high range car market.

Major characteristics of various car makes are summarised based on their sizes or type in Table 4.8 below. As some buyers use twin-cab light trucks as substitutes for five- passenger sedans, twin-cab light trucks are included in the five-passenger cars category. The “others” of this category of vehicle type are the cross of mini MPV and .

141

Table 4.8 Car types: Characteristics by size Vehicle Type Sample Mean Price Price Range Mean Mean Km (RM) (RM’000) HP (per litres) (’000cc) 4-passenger 32 37,327 27 - 62 0.99 16.9 5-passenger 328 71,237 29 - 540 1.54 15.5 Hatchback 106 48,701 22 - 110 1.33 18.0 Sedan 209 81,946 30 - 160 1.62 14.1 Light trucks 9 100,315 90 - 123 2.59 Not available Others 4 76,176 40 - 127 1.38 Not available 7-passenger 85 80,899 36 - 184 1.70 11.0 Mini MPV 60 64,693 36 - 129 1.55 10.5 MPV 10 117,332 82 - 178 1.95 11.0 SUV 14 138,242 120 - 184 2.04 12.1 Van 1 92,019 3.00 Not available

Notes: HP denotes horsepower. Van’s carrying capacity is 11 passengers.

Statistics of mean price show that price of cars may be positively correlated to the size of cars measured by the number of passengers. However, when the five-passenger category is broken down into segments, statistics show that five-passenger sedan on the average is more expensive than seven-passenger mini MPV. Such inconsistency is due to prices distortion. Proton is able to offer seven-passenger mini MPV at a much lower price than sedans offered by foreign car makers. Outliers in prices found in the sample are reflected by the price range of five-passenger cars that range from RM29,000 to RM540,000. The outliers in prices are a few luxury foreign cars such as Mercedes Benz and BMW.

Statistics of mean horsepower shows that there is overall a positive correlation between size of cars (number of passenger) and power of cars. The outliers in horsepower are the light trucks that are used as passenger cars.

142

It is not a common practice that domestic car makers including the foreign car makers assembling cars in Malaysia provide the information of fuel efficiency in Malaysia. Therefore, information about fuel efficiency is obtained from various websites where new cars’ owners share their experience and information such as paultan.org. The figures of fuel efficiency measured by kilometres per litre petrol are based on car owners’ estimation driving in the cities. Such information for foreign car makes is available as foreign car makers publish the information in foreign newspaper and other forms of media. In this case, the published data is used. Statistics show that generally, larger cars tend to be less fuel efficient than smaller cars.

Table 4.9a Demographic of car owners’ characteristics: number and percentage Car owners’ characteristics Number Percentage Gender (GEN) 398 Female = 0 205 51.5 Male = 1 193 48.5

Race (RC) 374 Non-bumiputra = 0 168 44.9 Bumiputra (native) = 1 206 55.1

Sector (SEC) 391 Public = 0 237 60.6 Private = 1 154 39.4

Number of dependents (DEP) 1 164 47.4 2 107 30.9 3 34 9.8 4 21 6.1 5 11 3.2 6 5 1.4 7 2 0.6 8 1 0.3 9 1 0.3

143

Distribution of car owners by gender is relatively even (see Table 4.9a). Distribution is however, less even by race, with slightly larger percentage for the native and less than 50 per cent for the non-native.

Distribution is also lop-sided when car owners are classified by the sectors they are working in. Statistics show that up to 60 per cent of the respondents work in the public sector and about 40 per cent work in the private sector. Such uneven distribution is highly unlikely to be bias in sampling as surveys are carried out in selected car service centres across the states. Since Proton cars and Perodua cars take up the top two largest market shares, the numbers of car service centres for Proton and Perodua are also larger than the other car makes.

There are 102 car owners which is about 22.8 per cent of the sample, did not indicate the number of dependents. They are treated as “missing data”. Of those who answered this question, there are about 78 per cent of total car owners who indicated that they have one or two dependents. The smallest number of dependents is one but, it is believed that some respondents who did not indicate the number of dependent may have zero dependent. Open-ended question has not been able to capture zero dependent when respondents skipped the question because of zero dependent. As a result, a number of zero dependents might have been treated as “missing data”.

Table 4.9b shows the distribution of income class and group of sample. Income classification is based on the Malaysia Statistics Department’s definition of low, average and high income. It follows that low-income group consist of income class 1 to 2; middle-income group is consisting of income class 3 to 7; and high-income group consist of income class 8 to 17.

The range of each income class is RM999. The constitution of low income, middle income and high-income groups in the sample are 36 per cent, 44.5 per cent and 19.5 per cent respectively. It is believed that the distribution of the income class is influenced by the targeted sample size for the individual car makes particularly the major players in the industry. The largest percentage of car owners’ surveyed fall under the middle-income range.

144

Table 4.9b Demographic of car owners’ income by range and group: number and percentage Income class Number Percentage Income group Number Percentage

1 76 18.3 Low 150 36.1 2 74 17.8 3 72 17.3 Medium 185 44.5 4 44 10.6 5 25 6.0 6 16 3.8 7 28 6.7 8 13 3.1 High 81 19.5 9 13 3.1 10 13 3.1 11 5 1.2 12 0 0.0 13 3 0.7 14 2 0.5 15 4 1.0 16 7 1.7 17 21 5.0 Total 416 100.0 416 100.0

4.8 Inefficiency - Sources and Definition

The cost of protectionism estimated based on price elasticity of demand measures the cost borne by the society because of distorted prices and the cost of implementing protectionist measures. Although the cost can be identified and measured, there may be cost not easily identifiable and hence, measured because the sources are not easily measurable. Since such cost associated with protectionist is not directly measureable it is not addressed and not accounted for in annual financial reporting. Consequently, production units are persistently produced inside the production possibility frontier. This phenomenon is

145 termed as X-inefficiency in Leibenstein (1966). The theory in Leibenstein (1966) says that the excess cost incurred due to monopolist’s failure to minimise cost causes monopolist’s operating cost to be greater than perfectly competitive firms. The cost is the result of X-inefficiency.

A critique to microeconomic theory of production is the theory’s assumption that firms are homogeneous. Consequently, production function observed is overly simplified for firms that employ multiple types of resources that are also not homogenous (Leibenstein 1979, p. 478). The study points out that decision-making process varies across firms of various sizes because of the complexity nature of firms’ objectives for example, profit, growth, sales, and benefits. Individual firms’ decision making process varies within an industry and across industries. Organization structures may become more complicated as firms adopt international business strategies to enter foreign market. As a result, firms’ objective to maximise profit for example, is not a simple production function of capital and labour. Leibenstein (1989, p. 1362) called the missing details and procedures in microeconomic analysis a black box. The study suggests that internal behaviour that is independent of market. As a result, while firms’ production function may be influenced by market, firms’ behaviour may reflect non-market interactions of non-quantifiable factors.

Formby, Keeler & Thistle (1988, p. 119) defines X-inefficiency the cost arising from misallocation of resources that are available but, not used by the firm to the fullest or being “forced” into other uses. Both Leibenstein (1966) and Formby, Keeler & Thistle (1988) are different in their identification of the sources of inefficiency. The latter definition the source of inefficiency as the forgone output due to slack resources. Therefore, the cost of X-inefficiency is the monetary returns not reap.

In the subsequent study of X-inefficiency, Leibenstein & Maital (1994, pp. 253-255) investigates the possible causes of X-inefficiency. The study points out that inefficiency in developing countries is not taken seriously by the government although there is persistent and large gap between the actual output and efficient output.

146

Factors that management consultants believed have been contributing to the gap are as follows (Leibenstein & Maital 1994, p. 253):

(a) Motivation of worker. Leibenstein (1966) says that inefficiency may not be due to lack of motivation of workers. The association between efficiency and motivation of workers may not exist so long as in the presence of pressure, there is labour mobility, and effective and reliable measurement of workers’ performance;

(b) Incomplete contracts that may cause differences in prices of inputs and of outputs;

(c) Imperfect input market that can be due to immobility of inputs, imperfect substitution of inputs, and possibly imperfect information; and

(d) Misspecification of production function due to technical deficiency or lack of information. As the management studies concerned were carried out in developing countries, missing information or unreliable information might have contributed to the gap.

In the study of efficiency in market and in firms, Shen (1984, p. 1363) finds that production in developed economies is characterized by substitution between capital and labour. On the contrary, production in developing countries show labour intensive. The input-output relationship suggests that X-inefficiency contributes to relative high production cost. Leibenstein (1989, p. 1370) concluded that there is a positive correlation between environmental pressure and X-efficiency. While a form of pressure may come from the external environment that is exert by efficient firms, other factors that may contribute to environmental pressures within the firms are as follows:

(a) Factor mobility. In the absence of unions, firms are able to replace inefficient labour with more efficient labour;

(b) Improvement of technology. Firms that adopt the new technology may develop competitive edge hence, able to substitute labour with capital;

147

(c) Reducing fund allocation for related department or division;

(d) Creating a new unit that partially overtakes the roles of the related department or division; and

(e) Harsh performance appraisal.

In conclusion, both Leibenstein (1979, 1989) and Shen (1984) suggest that X-inefficiency is the result of misallocation of resources in the firms. This implies that while price in a free-market drives firms to efficiency, there remain factors that are associated to organization and management of resources within the firms that contribute to efficiency.

X-inefficiency from the perspective of managers is due to mismanagement or mismatch of labour and other inputs (Mefford 2017, pp. 3-4). In this view, labour is considered to be a rather complicated kind of input due to the differences in individual’s skills, needs and expectation all of which in turn, influence their motivation. The difference in the managers’ view and Leibenstein (1979) perhaps arises from the nature of labour during which the studies took place. Leibenstein’s theory was proposed in the 1960s while most of the studies based on management’s view are carried in the 1980s and beyond.

Theory X labour is characterised as management-driven because job security is of primary importance to the former. On the other hand, Theory Y labour can be characterised as self-motivated because of self-motivation and are generally optimistic about achieving their personal goals while achieving organizational objectives. Hence, at firms’ level Leibenstein’s proposition of misallocation of resources is valid when Theory X labour is taken into consideration. As economies become more affluent, technology advances require labour of various skills and of different hierarchy of needs, X-inefficiency of firms may be viewed as mismatch of resources or conflicting interest between labour and firm that leads to production below the efficient level.

Productive efficiency is achieved by a firm if the firm is able to produce at minimum cost. As firms’ historical data is inherently bias, estimation of parameters of efficiency using micro level data can therefore, be misleading. Consequently, comparisons between firms are more meaningful (Kopp 1981). 148

4.9 Data Envelopment Analysis (DEA) - Application

In the analysis of efficiency in this research, published accounting data and statistical data are used to for comparisons among selected car makers. Such comparison implicitly assumes that the unprotected and/or experienced long-time industry player in Malaysia such as Toyota is efficient. Efficiency in this research is therefore, productive efficiency in microeconomics context although most studies in Management do not distinguish between productive efficiency and allocative efficiency.

Due to lack of transparency and government’s protectionist financial and non-financial assistance provided to Proton, there is incomplete data in the market and accounting data may not be reliable due to possible changes made in accounting treatment of fixed assets and expenses. Therefore, this study attempts to take the mathematical programming approach initiated by Charnes, Cooper, and Rhodes (1978, 1981) to identify the cost associated with inefficiency. In Proton’s case, the cost is highly likely to be associated with excess capacity because of the government’s political objectives to provide employment to the natives and to redistribute wealth from the non-natives to the natives.

According to Kopp (1981), this approach is pioneered by Farrell (1957) that focuses on the concept of productive efficiency. Farrell (1957) suggests that there are two components of overall productive efficiency (OPE), namely: physical efficiency of the input-output production transformation (TE), and economic efficiency of optimum allocation of inputs (AE).

At a given input combination represented by point C, the OPE index, a product of TE and AE indexes, is a ratio of the maximum factors combination (0A) on the ray from the point of origin, 0, over the actual output (0C) (refer to Figure 4.4).

When a firm allocated combination of inputs represented by point C, the firm will be able to produce higher level of output reflected by the dotted isoquant qq’ if the firm operated efficiently or operates in the manner that other firm that allocates the same combination of inputs but, produce the output level represented by qq’. Technical inefficiency is measured by the ratio of 0B over 0C that is, the index shows the gap or distant between a

149 selected efficient firm and the less efficient firm. Farrell (1957) assumes that the overall productive efficiency effect can be segregated. The efficient unit isoquant can be seen serves as production possibility frontier that is, efficient standards.

Figure 4.4 Least cost combination and maximum output

Capital q Q C * I q’ F * B A E * Q’

0 F’ I’ Labour

Source: Kopp (1981, p. 480).

The optimum output is the least cost combination of output indicated by the equilibrium point, E, where the isoquant (QQ’) is tangential to the isocost (FF’). Point B is the combination of output that could have been produced given the technology available but, can be attained if more inputs are employed, reflected by greater financial resources used as reflected by the dotted isocost, given the firm’s operation.

This approach estimates the production frontier without assumption of functional forms of relations between factors and output. Therefore, it allows relaxation of homogeneity and homotheticity assumptions in production technology. The frontier is estimated based on a number of efficient firms, thus these firms’ efficiency levels are used as a benchmark for comparisons with other firms. In short, this approach holds inputs constant and measure the output differences between efficient firms that serve as benchmarks and other firms under evaluation.

150

Another approach is the “least cost combination” approach where output is held constant, and comparisons of inputs allocated in production units are made. Models taking this approach assumes functional form and causes of inefficiency are specified by stochastic disturbance structure. In the approach, technical inefficiency is the cause of output deviation. Therefore, technical efficiency is measured by the ratio of actual output over maximum output technically feasible.

Cook, Tone & Zhu (2014, p.1) makes a clear distinction between the use of DEA for generation of a “production frontier” and for generation of a “best-practice frontier” because the purpose of DEA will determine an appropriate model to be used.

The procedures taken in this research are stated in Golany & Roll (1989, p. 238). They are:

(a) selection of decision-making units (DMUs). The DMUs selected in this research are domestic car-makers and foreign car-makers operating in Malaysia. They are: Proton, the domestic car maker, Perodua, a partially Japanese-owned car maker, and UMW Toyota;

(b) determination of inputs and outputs that are relevant for the assessment of the relative efficiency of car-makers. Disaggregate data of DMU’s inputs is not available but, aggregate data is published in the annual financial statement; and

(c) selection of a DEA model and analysis of car-makers’ efficiency. In this research, CCR (Charnes, Cooper & Rhodes 1978) model is selected for evaluation of inefficiencies.

4.9.1 CCR Model

CCR model was initially developed to measure decision making efficiency for the purpose of evaluating non-profit-making entities that is, evaluation of public programs. The model can be applied to firms in private sectors because the model essentially

151 compares decision-making units. To compare efficiency of firms due to protectionism, a protected firm is compared to firms operating locally but, not protected.

Since the CCR model was used for non-profit-making entities, the measurement of both input and output are the actual market price (Charnes, Cooper & Rhodes 1978, p. 429). In the case when the model is used for evaluation of efficiency in this research, prices are distorted by financial assistance given to protected firms and distorted by taxes imposed on output of the foreign firms. Consequently, it is difficult to compare and evaluate inefficiency arises from mismanagement of resources. However, Leibenstein (1966, 1979) and Shen (1984) argue that X-inefficiency arises because of misallocation of resources and since prices distortion is a factor of resource misallocation, this research applies CCR model using the market prices. To investigate the gaps of inefficiency, market prices after removal of taxes will also be accounted for to derive productive index without tax.

The concept of least cost combination or producer equilibrium can be expressed in the following objective function:

Max h0 (u, v) = uryr0 / vixi0 u,v r i Σ Σ where ur and vi are greater than 0, r denotes output type, and i denotes input type. Both u and v are also virtual rate of transformation of outputs and inputs respectively in Charnes, Cooper & Rhodes (1978). The ratio of u over v is the measure of marginal rate of substitution. Both y and x denotes outputs and inputs respectively.

The maximizing output objective can also be expressed as:

h⃰0 = max ∑ ur yr0 – u0 subject to r

∑ vixi0 = 1, i

∑ ur yr0 – ∑ vixi0 – u0 ≥ 0 r i

152

The signs of u indicate the economies of scale experienced by the firm. For example, a positive sign implies increasing returns to scale while a negative sign implies decreasing return to scale.

The weights of inputs in CCR 1978 are the relative weight of input of individual DMUs. It is measured as the ratio of a DMU’s input i over the aggregate of input i of total sample. In Malaysia, the weights of inputs are neither observable nor available. The weight of inputs is the ratio of labour in a unit over the total labour of the sample while the weight of output is measured by the relative weight of average prices.

Since the objective Data Envelopment analysis in this research is to compare efficiency in the environment where there are serious prices distortion and lack of transparency, the efficiency index is estimated as ratio of weighted output to weighted input at aggregate level.

4.9.2 Sample

A ‘rule of thumb’ for sample size of decision making units (DMUs) is double the size of the product of input and output number (Dyson et al. 2001, p. 248; Cook, Tone & Zhu 2014, p. 2). Cook et al. (2014, p. 2) also suggests that sample size may be immaterial when DEA is used as a benchmarking tool. Overly large number of inputs and outputs may render the tool ineffective because the relative efficiency of individual DMU may not be shown (Ibid, p. 3).

The oldest car manufacturer and assembler, non-fully Malaysian-owned, is UMW Toyota. This foreign car maker is chosen because of its well-known production strategy, the ‘Just-in-time’ (JIT) approach. The approach involves management of inputs to minimize carrying cost of material inputs and of outputs, and fully utilised labour’s capability. Study finds that Toyota outperforms car makers in the United States, Sweden, and W. Germany after implementation of the JIT in Toyota for about 20 years (Sugimori et al. 1977, p. 563). Therefore, it is reasonable to assume that Toyota in Malaysia operates in similar model as its Japanese counterpart.

153

Other car manufacturers that are Malaysian-owned are Proton, Perodua, and of which Inokom and Naza are founded about ten years after Proton is established. However, Inokom and Naza are very small. Therefore, for benchmarking purposes, the DMUs chosen are UMW Toyota and Perodua.

DEA essentially assumes homogeneity in inputs, outputs, and operating environment for comparisons of DMUs. Violation of homogeneity assumption in any of the three elements makes it difficult to conclude the source(s) of discrepancies in the efficiency ratios (Dyson et al. 2001, p. 247). Homogeneity in inputs includes categories of labour, raw materials as well as equipment, while homogeneity in outputs may be the similar range of cars and/or services. The assumption that all DMUs operate in similar environment implies that all selected car makers in this research have equal tax treatment and subject to similar non-tax regulations for examples, safety standard, administrative control, and location of the industrial zone approved.

DEA is applied in this research for identification of the source of inefficiency in terms of the relatively lower output of the domestic car maker than what it may produce if similar level of efficiency as its competitor can be achieved. In other words, the cost of inefficiency is the loss of output because the car maker is not producing on the production frontier.

The assumption of homogeneity in operating environment is relaxed in this research because the cost of inefficiency of the firm is possibly the impact of protectionism that makes the operating environment of the domestic car maker different from the operating environment of the other competitors that are not protected by similar protectionist tools.

As such an estimation of a production frontier based on the similar inputs and outputs of UMW Toyota and Perodua will be made and the distant from which Proton is away from the frontier will be measured in terms of gain or loss as a result of Proton outperforming the competitors or the opposite.

154

4.9.3 Data

A requirement for selection of inputs and outputs for analysis is their significance to the management of the DMUs. Disaggregate data of inputs and outputs are considered ‘private and confidential’ hence, not collected and published by statistics department or any local institutions. The data is also not made available for research purposes in the resource centres of the car makers. Therefore, aggregate data at firm level is used. Data of inputs can be found in the DMUs’ yearly financial statement. Annual financial reports are purchased at a price at Companies Commission of Malaysia in Kuching.

In non-profit-making institution, it is suggested that expenses that are consistent with the objectives of the DMUs and their performance measures are selected (Dyson et al. 2001, p. 248). For profit-making organizations like car makers, expenditure of inputs is generally a reflection of production activities if the car makers are competitive organizations. However, when a car maker is protected by the government and when there are non-economic objectives involved where examples include creating job opportunities for the natives and, redistribute wealth from the car buyers to the native shareholders and native employees, the operating costs and true data that reflects production are no longer fair.

Since it is also the objectives of this research to investigate how protectionism influence efficiency of the protected firm, the costs data reflects misallocation of resources. Comparison is made at plant level to find inefficiency arising from excess capacity, which in turn, causes protectionist measures.

Accounting data selected are the cost of goods sold, fixed assets, number of employees, and total revenue. The weights in terms of book values, of fixed assets is analysed separately based on their characteristics because of the following reasons:

(a) Malaysian government has given privileges to Proton as protectionism tools. There is a lack of information related to the privileges given. This research, therefore, overcomes the problem related to lack of transparency by assuming land

155

has been allocated for Proton. Government’s favouritism if exists, can be gauged by the value of land relative to the land used by other DMUs;

(b) Media reported that Proton’s production capacity had increased by 188 per cent, from 80,000 units to 230,000 units in 1997. Therefore, inefficiency arising from excess capacity and privileges can be identified and traced if value of plant and machinery of DMUs can be compared; and

(c) Excess capacity in labour can also be identified and measured if office furniture, fittings and office vehicles can be compared. Although the number of employees can be used as a proxy for excess capacity in labour, fixed assets for employees’ use is a better proxy because fixed assets also reflects privileges given to the protected car maker in the form of non-financial assistance, and to the employees.

As such the fixed assets are segregated into: (i) land, (ii) plant and machinery, and (iii) furniture, fittings and vehicles for preliminary study of resource allocation among the selected car makers. Since disaggregate data of inputs is not available, cost of goods sold is used as a proxy of input in addition to the fixed assets.

Since only the sales of the best-selling models are available, proxies for outputs of DMUs are the total quantity sales in 2015. The prices used for weightage are the market prices of their respective best-selling units.

4.10 Conclusion

The alternative approach taken in this research has not departed from the mainstream microeconomics theories in that satisfaction is derived from consuming goods and the consumers’ willingness and ability to pay for the goods reflect the utility derived from consumption of the goods. The approach taken Discrete Choice Modelling, is an extension of the conventional microeconomic theories that states social-economic factors have influence over the demand for goods. Discrete Choice Modelling is desirable for the study of car demand because cars are durable and differentiated goods.

156

Research methodology adopted for this study is built upon the principle of positivism. Steps taken to ensure relevant information is obtained are face-to-face interview with car buyers of diverse background. Since there is a time gap between the studies carried out mainly in the developed economies and this research, precautionary measures are taken to identified car characteristics no longer taken into consideration by car buyers as some characteristics are already standardized so that these characteristics are excluded in the model. On the other hand, car characteristics that may influence buying decisions are included in the model. Measurable social-economic variables and car characteristics are identified and mapped to the variables identified in other studies to ensure consistency in unit measurement.

Steps are also taken to ensure that primary data collected generate a reasonable representative of the population. In addition to the use of a few open-ended questions, the analytical tools applied do not require specification of functional form. Hence, allowing the data to give an appropriate picture of car demand in Malaysia. The number of outliers is reduced by questioning respondents during the survey as soon as outliers are detected to ensure that outliers are not due to errors.

Logistic regression analysis is applied to study how car buyers characteristics and car characteristics may influence the demand for car without assuming linear relationships between variables. An advantage of logistic regression is that collinearity is allowed because the logistic regression function is a probabilistic function. As such, interaction between car buyers characteristics and car characteristics, and interaction between two pairs of car characteristics are allowed. This can be a more accurate representation of car demand because car characteristics may have influence over each other and car buyers of certain characteristics may have preference over certain car characteristics. Using this approach, market level data for cars and micro data of car buyers can be analyzed together.

Since the introduction of logit function in the early 19th century, variations to simple logit function have been introduced for studies in social sciences. This research applies simple logit model for analysis of Malaysian car demand.

157

Although general equilibrium framework can be used to elicit spillover effect of protectionism, this research applies partial equilibrium framework. Using partial equilibrium framework, this research estimate cost of protectionism in the car industry which is small relative to the size of the manufacturing sector in Malaysia. Demand functions for selected car makes are specified using Discrete Choice models. Price elasticity of demand for each of the five selected major car makes, aggregate of small Asian cars and aggregate of American and European cars are estimated based on the estimation of Logistic Regression analysis.

This approach is relatively new in the discipline of Economics in that Discrete Choice model is used to incorporate car characteristics in specification of demand function for highly differentiated goods like cars. In this approach, consumer data, product-level data and market-level data are used in Logistic Regression model to identify car characteristics that influence car demand. Due to the nature of this research, primary data collected may be restricted by the willingness of respondents to participate in the survey as it is done in random and across all states during the period of the survey.

In the environment where there is lack of reliable data due to both protectionist tools and to lack of transparency, this research has to be contented with the published financial information and pieces of data found from various sources.

Data envelopment analysis is applied to compare a protected firm with firms that are not protected and are competitive in this research. Using Charnes, Cooper & Rhodes (CCR) 1978 model, maximum ratio of weighted outputs to weighted inputs of decision-making units that is, selected car makers are obtained. Gaps between the efficiency ratios of protected and unprotected firms show the inefficiency of a firm relative to the other firm. If the efficiency ratio of a protected firm is less than the ratio of the unprotected firm, it is concluded that the protected firm is less efficient than the unprotected firm. However, the cost of inefficiency cannot be determined from the efficiency index.

158

Notes 1. The alternative two methods Ordinary Least Square (OLS) and Weighted Least Square (WLS) are unable to generate meaningful estimators for grouped data that values either 1 or 0.

159

Chapter 5 Results and Analysis

5.1 Introduction

This research adopts the Discrete Choice Model specified in Berry, Levinsohn & Pakes (2004) in which the model accounts for unobservable car characteristics and unobservable car buyer characteristics in addition to the observable and quantifiable socio-economic factors that are often found influencing demand. These unobservable characteristics have not been accounted for in many a priori studies that adopt the econometric tools although economists have long recognized the importance of product characteristics in influencing product demand. This is mainly due to the lack of knowledge and tool available to study unobservable characteristics that may very well be qualitative in nature.

A weakness of the approach commonly taken in Microeconomics analysis is the assumption of product homogeneity. The proxy for car demand is often depreciation, expressed as a function of measureable car buyer characteristics. Due to technology changes, stiff competition and real GDP growth over the years, car demand may be influenced by the car characteristics that play significant roles in differentiating various car makes.

In this research, car characteristics identified in the survey are added to the social- economic factors that influence the demand for different car makes in the logistic regression models. If car characteristics have significant influence on specific car makes, they will contribute to the goodness of the models. It can also be deduced that if the domestic cars are differentiated that is, car characteristics of domestic car influence the demand for domestic cars, then technology diffusion may have taken place.

This research also attempts to estimate price elasticity of demand for cars. Microeconomics theory states that variety of choice or availability of substitutes influence price elasticity of demand. Therefore, in the presence of trade barriers, price elasticity of demand may be relatively lower than the price elasticity of demand when there is freer trade. Since the marginal change of price is not directly observable from the probabilistic

160 demand function, average prices, averages of measurable car characteristics are used for estimation of price elasticity of demand.

Following the estimation of price elasticity of demand, estimation for cost of protectionism can be made. The cost of protectionism is expected to be larger than 1 per cent of Malaysia’s GDP in contrast to the cost of protectionism estimated in studies carried out in developed economies because of high tariff rates and the use of non-tariff barriers. In this research the cost is also measured as percentage of GDP contribution of manufacturing sector. This is because Malaysia is a developing economy in which manufacturing sector contributes to a small proportion of the GDP compared to manufacturing sector’s GDP contribution in a developed economies.

The hypotheses of this research are as follows:

Hypothesis 1: Social economic factors such as income, price of cars, and some car buyers’ characteristics influence the demand for cars.

Hypothesis 2: Car characteristics do not influence the demand for domestic cars.

Hypothesis 3: Demand for cars is price inelastic due to limited choices when there are trade restrictions.

Hypothesis 4: The cost of protectionism as a percentage of manufacturing sector’s GDP contribution is large.

This chapter presents the results of logistic regression analysis of how car characteristics interact car buyers’ characteristics and their influence on the demand for various major car makes in Malaysia.

Since there is a lack of studies in Malaysian car industry and implicit government safety net to protect the domestic car maker, this research takes a conservative approach to assume these unobservable characteristics exist, then tests the significance of these characteristic in logistic regression analysis.

161

Using Discrete Choice Modelling, the demand function estimated will be used to calculate price elasticity of demand for the selected car makes. Comparisons of estimated price elasticity of demand with price elasticity of demand in the United States show the overall effect of competition on car buyers’ responsiveness to price changes. The cost of protectionism measured as a percentage of real GDP and percentage of manufacturing sector’s real GDP contribution are then estimated based on the price elasticity of demand estimated earlier.

Due to the partial equilibrium framework used in this research, other costs associated with protection of domestic car industry are not measured. However, the spillover cost will be discussed in the subsequent chapters.

5.2 Preliminary Analysis

Table 5.1 shows comparisons of market prices of various models of domestic car makes and selected foreign car makes. Data shows that both the domestic car makers target the low and low-middle income groups, offer small and cheap cars in segment A. Perodua of which about 70 per cent of the interests are owned by Japanese corporations, has advantage in producing cars at lower cost, offers two models in segment A. Both Toyota and Honda do not enter segment A reflecting their marketing strategies that position themselves at a higher market level.

Models and prices in segment B show that both Toyota and Honda position themselves differently within the segment. Honda offers smaller cars such as hatchback models that Toyota does not offer while Toyota offers a wider range of car, that is, offers more characteristics to capture larger segment of the market. Data also shows that since Perodua is partially owned by Toyota, both of them do not compete against each other.

162

Table 5.1 Comparisons of prices of selected car makes - across segments Segments Proton Perodua Toyota Honda Models Prices Models Prices Models Prices Models Prices A Hatchback - - Axia 23-40 - - - - Sedan Saga* 34-42 Bezza 34-48 - - - - B Hatchback Iriz 40-54 MyVi* 43-54 - - Jazz 70-87 ------HRV 109-125 Sedan Persona 42-54 - - Vios* 77-87 City* 74-92 Mini MPV Ertiga 57-62 Alza 51-63 Avanza 81-86 BRV 81-91 - - - - Sienta 102 ------Innova† 115-132 - - SUV - - Aruz 73-78 Rush# 93-98 - - C Hatchback Suprima 65-73 - - C-HR 150 - - Sedan Preve 61-68 - - Altis 120-138 Civic 108-129 Mini MPV Exora 62-69 ------SUV X70⁑ 100-123 - - - - CRV 138-163 D Sedan Perdana 104-127 - - Camry 190 Accord 149-169 MPV ------Odyssey 259 SUV - - - - Harrier 243-266 - - Truck/4x4 Truck - - - - Hilux 90-138 - - 4x4 SUV - - - - Fortuner 170-196 - - Executive MPV - - - - Vellfire 362 ------Alphard 443-510 - -

Note: 1. * best-selling models in 2016 2. All prices are in range of nearest RM’000, 2019 ‘on the road’ (OTR) price in Kuala Lumpur 3. † compact MPV 4. # crossover MPV-SUV 5. ⁑ newly launched in 2019 6. Exchange rate: RM1 to AUD0.35

Sources: https://www.zigwheels.my/new-cars/ https://www.honda.com.my/model/pricing https://www.carbase.my/all-segment Personal conversation with industry players 163

Distribution of models across segments reflects segment B’s constitution to the largest proportion of Malaysian car market. Foreign hatchback and sedan cars are generally 83 per cent to 119 per cent more expensive than domestic car of similar models, reflecting the effect of trade barriers. Prices of mini MPVs in this segment show that is more price competitive than Proton Ertiga while the prices of foreign mini MPVs namely: and Honda BRV, are about 40 per cent more than prices of the domestic mini MPVs.

In segment C, Toyota and Honda are avoiding fierce competition with each other in this smaller segment of the market. Perodua does not compete with Toyota in this segment. Prices of hatchback models and sedan models in this segment show that the average prices of foreign car makes are about double the average prices of domestic cars. In conclusion, tariffs have made foreign car prices about twice the price of domestic cars across segment B, C, and D.

Malaysian car market is unique in its own way. Based on observation, high income households possess lower and medium range cars for several reasons such as thrift, for ferrying children forth and back from schools, or for their elderly parents’ use. There are also young adults whose income level is relatively low but own higher range cars. Face- to-face interviews at service centres revealed that there are parents who earn high income may bought new high range cars for their young adult children for work or for leisure. Such two extreme cases are outliers for the samples selected in this study.

During the process of data collection, selected car centres in a small number of states did not agree to participate in the survey. Since there are very few car service centres in these small states, there is no other service centres available for survey to be carried out. Hence, all of these ten selected car makes are selected in Federal Territory and all states of the country with the exception of two smallest states, namely: Perlis and Kedah.

Comparisons of the market shares and sample proportions are summarised in Table 5.2. The upper frame shows the major lower to medium range car makes: Perodua, Proton, Toyota, Nissan, and Honda. These major car-makes contribute up to 81.1 per cent of the market shares. The lower frame of the table shows that top five higher range car

164 contributes up to 2.8 per cent of the market share. Other car makes not individually identified due to their small market shares are classified as either American/European cars or other Asian cars in analysis.

Table 5.2 Market shares in year 2012, 2015 and sample shares Car Makes Market shares (%) Sample shares (%) (a) Lower- & medium range 2012 2015 Perodua 30.3 32.0 34.5 Proton 22.6 15.3 29.3 Toyotaa 16.8 14.1 14.8 Nissanb 5.8 7.1 4.7 Hondac 5.6 14.2 7.8

(b) Higher range BMW 1.2 1.1 1.3 Mercedes Benz 0.9 1.7 0.7 Chevrolet 0.3 0.1 1.1 Lexus 0.2 0.3 0.9 Audi 0.2 0.2 0.2 Total 83.9 86.1 95.3

Notes: Models targeting higher range segment are aToyota Camry, bNissan Teana, cHonda Accord.

Source: Malaysia Automotive Association 2013.

Statistics show that the sample shares for individual car makes are closed to the actual market shares except for Proton, Chevrolet and Lexus of which the sample shares are relatively large compared to the market shares in 2012 and 2015. Hence, the aggregate

165 sample share of ten selected car makes is about 10 per cent larger than the actual market shares of 83.9 and 86.1 in year 2012 and 2015 respectively.

Based on the sample size of 34 comprise of higher range cars and the high end of Toyota, Nissan and Honda, 71 per cent of the sample shows cross substitution among the higher range car makes. For example, a buyer of Lexus RX considered Honda Accord as the second choice in the process of making his/her buying decision.

About 17 per cent of the sample show car owners chose within the same car makes for models of different segment for example, car owners who bought Toyota Camry considered Toyota Altis which a medium range of Toyota. About 12 per cent shows no second choice considered in the decision-making process. These two sub-samples reflect the possibility of brand loyalty and/or reputation of brand names.

Raw statistics also show that cross substitution among the higher end models of Toyota, Nissan and Honda with the lower end of BMW and Mercedes Benz is very rare that is, less than 2 per cent. This justifies the initial treatment of data to categorise car makes by high range and non-high range. This is consistent with Berry, Levinsohn & Pakes (1995, p. 852) that suggests no cross price substitution between car makes of different segments.

Due to relatively lower tariff imposed on light truck than all other models of foreign cars, the numbers of new light trucks registration increase in recent years. In a few unrecorded discussions with light truck owners, it is found that light trucks are used as passenger cars because twin cab light trucks are able to carry five passengers like foreign five-passenger cars, at the price up to RM50,000 lower than the price of foreign five-passenger cars of similar horsepower.

To understand the market’s response to tariff treatment on light trucks, the sample collected is carefully categorised to find out if there are cross substitutions across segments for examples, if light trucks and multi-purpose vehicles (MPVs) are second choices of car owners who bought sedans. Sample statistics show that 40 per cent of car- owners who bought light truck considered other makes of light trucks as second choice. 60 per cent of car owners either bought light trucks but their second choices were cars of

166 different segments or bought cars of different segments but considered light trucks as their second choices. The summary of sample statistics related to light trucks are as follows:

Table 5.3 Light trucks: First or second choice First choice Second choice Sample proportion (%) Light trucks Light trucks (of different makes) 40 Light trucks Sedans and others* 20 Sedans and others* Light trucks 40 Total 100

Notes: Others* are comprised of hatchback, MPV/SUV and 4x4.

Statistics show that light trucks are used as normal passenger cars instead of used for carrying goods. This is observed because tariffs have distorted prices to a great extent that demand for light truck is influenced by prices’ interaction with certain observable and/or unobservable car characteristics.

Similar situation is observed in the MPV/SUV segment. Out of 126 whose first or second choices were MPV/SUV, only 13 per cent compared MPV or SUV across different makes within the segment and bought MPV or SUV. In other words, 87 per cent of those who bought MPVs/SUVs considered other models such as sedans or hatchback in their decision making process. Consequently, samples are not segregated by their models for example, hatchback, sedans, MPV/SUV, and light trucks for analysis. That is, demand for cars is analysed by their makes instead of by models or by segments. This is not consistent with Berry, Levinsohn & Pakes 2004 in which dummy variables are assigned for , full-size vans, sport cars and four-wheel drive.

The number of dummy variable used in this research is minimised because the sample size used in this research is relatively far smaller than the samples used in most of studies that apply Discrete Choice modelling. Therefore, the number of passengers (NS) is used 167 as a proxy to capture the characteristics of the non-sedans that are used like sedans as in the case of light trucks and SUVs.

Figure 5.1 Higher range cars: Unit sold in year 2012

7000

6000

5000

4000

3000

2000

1000

0

Source: Malaysia Automotive Association 2013.

The market shares of the higher end cars range from 1 per cent for BMW to 0.2 per cent for Audi in year 2012. Figure 5.1 shows that within the category of high range cars, BMW and Mercedes Benz dominate this segment, with each of them sold about 6000 units in year 2012. Due to the extremely skewed market shares and complicated tariff structure imposed on the luxury cars, it is difficult to identify possible unobservable car characteristics that may influence luxury car demand.

Nevo (2000, p. 534) suggests that unobserved product characteristics may have influence over the markup of prices, causing the error terms to be correlated to prices and hence, contribute to biased estimation of price elasticities. Although theoretically, identification of variables that shift cost and variables that are uncorrelated to the demand shock may overcome the problem, published cost data is often insufficiently fine for differentiating car makes (Ibid). In Malaysia, highly fine and specific data is classified private and

168 confidential, making in-depth analysis of the car demand infeasible. Therefore, a dummy variable is assigned to the top high range car makes that are generally considered to be luxuries and may possess ostentatious value for testing of the presence of unobservable car characteristics.

Due to lack of study of the Malaysian car market, a series of preliminary tests of association or correlation between pairs of variables are conducted. For the purpose of determining the presence of correlation between income and car size, a preliminary test on the association between car size measured by number of passengers (NS) and income is performed. The choice of test of association or test of correlation is dependent on the nature of data. In this case, a test of association is preferred to a test of correlation because variation in income measured by range of income, may dissipate the possible relationship between income and car size that is measured by three classes of size, namely: 4-, 5-, and 7-passenger. Therefore, income levels of car buyers are categorised under three classes to match the number of classes of car size. The classes of income are: high (H), middle (M), and low (L).

Table 5.4a Summary: Number of passengers (NS) and Income group (YGRP)

NS YGRP Total 4 5 7 16 H Frequency 3 60 19 1 83 Per cent 9.4 20.1 24.1 100.0 L Frequency 17 117 15 0 149 Per cent 53.1 39.1 19.0 0.0 M Frequency 12 122 45 0 179 Per cent 37.5 40.80 57.0 0.0 Total Frequency 32 299 79 1 411 Per cent 100.0 100.0 100.0 100.0

Although there is a vehicle size of 16-passenger found in the sample collected, the sample size of 16-passenger van is one only. This sample is not omitted because data in the questionnaire shows that the van is used as a passenger car for a family that has nine

169 children. As a result, there are four classes of car size tested for their association with three income groups.

Statistics in Table 5.4a show that the lower income group dominates small cars category, contributing to 53.1 per cent of four-passenger cars while the high income group contributes to 9.4 per cent of the total small car buyers. In contrast, middle income group dominates the five-passenger and seven-passenger car categories.

Table 5.4b Tests of Association: Income group and car size (NS)

Statistics DF Value Prob Chi-Square 6 20.1324 0.0026 Likelihood Ratio Chi-Square 6 20.5905 0.0022 Mantel-Haenszel Chi-Square 1 0.0000 1.0000 Phi Coefficient 0.2213 Contingency Coefficient 0.2161 Cramer's V 0.1565

The results for test of association are summarised in Table 5.4b. Chi-square statistics suggest that the null hypothesis of no association is rejected. Cramer’s V statistics of near to zero implies that there is a weak association between income group and number of passengers. Phi coefficient 0.2213 and correlation coefficient of 0.2726 at 5 per cent significant level implies a weak correlation between income and car size measured by the number of passengers. The weak association observed possibly due to the outliers: high income group’s acquisition of small cars and low income group’s acquisition of large cars (refer to Table 5.4a).

Subsequently, car size (NS) is tested for its correlation with the numbers of dependent (DEP). The numbers of independent in this test are not classified under high, medium and low to match the number of class for car size because the numbers of dependents range from zero to nine. Results in Table 5.5 show that car size is weakly positively correlated with the number of dependent. This implies that while there may be a need for

170 larger cars for ferrying more dependents, car buyers may buy larger cars for other car characteristics that are unique to larger cars.

The results also suggest that the number of passengers the car owners have to carry may not have significant influence over the decision to acquire Mini MPV, MPV, SUV, double cab light trucks and vans. This is partly due to prices distortion and partly due to differences in car characteristics which in turns, influenced by technology. For example, some respondents chose Toyota Avanza, seven-passenger mini MPVs at prices below RM70,000 instead of a five-passenger sedan, Toyota Vios at prices above RM75,000 although the respondents use their car for themselves only or to carry less than five passengers.

Table 5.5 Test of Correlation - Car size and number of dependents Pearson Correlation Coefficients Prob > |r| under Ho: Rho=0 NS DEP 0.2726 (<.0001)

Notes: p-value in parenthesis. Other related statistics for the variables are available in the Appendix.

A large gap difference in the state of technology between Malaysia and Japan enables Toyota to produce cars in Malaysia using technology that have already reached its maturity in Japan in order to produce cars at competitive price. This may contribute to the observation that some car buyers who prefer foreign cars are willing to pay higher price for larger seven-passenger foreign cars instead of buying cheaper and smaller domestic five-passenger cars.

With respect to the high car ownership per household ratios in Malaysia over the years, number of cars (NOC) is tested for its correlations with socio-economic factors, namely:

171 income level (Y), gender (GEN), race (RC), sector (SEC) and number of dependent (DEP). Correlation test results are shown in Table 5.6a.

Table 5.6a Interaction of number of car in households (NOC) with car buyers characteristics: Correlation, p-value, and covariance Car buyer Y GEN1 SEC2 RC3 DEP characteristics NOC 0.2752 0.0997 -0.1410 -0.1868 -0.0648 <.0001* 0.0903† 0.0148* 0.0018* 0.3034 1.7222 0.0772 -0.1174 -0.1587 -0.1421

Notes: 1. Gender of car buyers is either female (dummy variable 0) or male (dummy variable 1). 2. Sector (SEC) in which car buyers work is either public sector (dummy variable 0) or private sector (dummy variable 1). 3. Race of car buyers is either non-bumiputra (non-natives, dummy variable 0) or bumiputra (natives, dummy variable 1) * significant at 5 per cent level. † significant at 10 per cent level.

Results show consistency of observation with Microeconomic theory that is, NOC is positively correlated with income at 5 per cent confidence level. Statistics imply that higher income group tends to buy more cars. Although NOC is positively correlated with Y, it cannot be concluded that higher income causes larger number of cars owned because of the relatively large covariance between income and number of cars owned.

The correlation matrix also shows that NOC is negatively correlated to race (RC) at 5 per cent significance level. These negative correlations imply that there is a relationship between larger number of car ownership and non-natives households. In other words, it implies that there is a greater chance to find a non-native car owner than a native car owner who own more than a car.

172

The negative relationship between NOC and the sector (SEC) within which car owners are working is statistically significant at 10 per cent level. Statistics imply that there is a greater chance of finding a car owner who own more than a car working in public sector than in private sector.

NOC is also positively correlated to the gender (GEN) at 10 per cent significance level. This statistics imply that there is a greater chance to find a male car owner than a female car owner who own more than a car.

Since the founding of Proton in 1985, Malaysian government label the cars as “national” cars. Malaysians are encouraged to buy national cars while the government promotes national cars by giving incentives to civil servants who are predominantly the Malays who claim to be the natives of the land. Perodua that is founded nearly 10 years later is also labelled as “national” car maker although more than 50 per cent of their shares are foreign owned. Although both “national” car makers receive different treatments in the forms of privileges from government, civil servants are given similar privileges such as guaranteed loan approval and minimum down payment for acquisition of Proton or Perodua car. Therefore, a dummy variable is assigned to both Proton and Perodua cars to account for privileges and other features attached to these cars for promotion.

Tests of correlation are then conducted to investigate the correlations between number of cars (NOC) and car price, and car characteristics that are measureable and non- measureable such as the privileges. The results in Table 5.6b show that there are positive and weak correlation between NOC and car price (LP), and horsepower (HP). It is difficult to interpret the correlation between NOC and price and between NOC and horsepower. It can be interpreted as there is a pattern that shows larger number of cars in a household is related to more expensive cars and cars of larger horsepower.

Positive correlation coefficient for NOC and ND suggests that larger car numbers owned are related to foreign cars. This implies that there is a greater chance that car owners who own more cars bought foreign cars.

173

Table 5.6b Interaction of number of cars in households (NOC) with car characteristics of first and second choice: Correlation, p-value, and covariance First choice LP1 HP1 KML1 SPEC1 ND1 NOC 0.2343 0.2452 0.0295 -0.0043 0.1865 <.0001* <.0001* 0.6233 0.9424 0.0009* 0.0697 0.1410 0.1334 0.0026 0.1014

Second choice LP2 HP2 KML2 SPEC2 NOC 0.2793 -0.0021 0.0309 0.0303 <.0001* 0.9743 0.6616 0.6413 0.0856 -0.0036 0.1516 0.0283

Notes: 1. Dummy variable (ND) is used for differentiating national car (Proton) = 0 and non- national car (others) = 1. * significant at 1 per cent level.

Results also show that NOC is positively correlated to prices of second choice. Hence, it is concluded that car owners with expensive second choices are likely to own a larger number of cars than car owners with relatively cheaper second choices.

Since NOC is not a car buyer characteristic, it is not included in the logistic model for analysis in the subsequent section. However, it is concluded that income attributes to numbers of car owned by households in Malaysia. It can also be concluded that NOC is related to the car buyers’ race and choices.

Correlation and covariance matric in Table 5.7 shows car prices and characteristics of first and are positively correlated to the prices and characteristics of second choice. Although car characteristics of first and second choices are positively correlated, their correlations are very weak relative to their covariance. Positive correlation between characteristics of first choice and prices of second choice suggest that car buyers who buy foreign cars or cars that are perceived luxury, and/or larger car size tend to make comparison with relatively more expensive second choice.

174

Table 5.7 Characteristics of first (1st) and second (2nd) choice: Correlation, p- value, and covariance Pearson Correlation Coefficients Prob > |r| under Ho: Rho=0 Covariances 1st LP1 HP1 KML1 SPEC1 ND1 UD2 NS1 2nd LP2 0.6962 0.4982 -0.1891 0.1352 0.2686 0.5496 0.1740 <.0001* <.0001* <.0015* 0.0215† <.0001* <.0001* 0.0023* 0.0475 0.0475 0.0456 0.0482 0.0475 0.0475 0.0475 HP2 0.1760 0.1869 -0.1018 -0.0432 0.0966 0.1454 0.1711 0.0021* 0.0011* 0.0898 0.4663 0.0933 0.0113† 0.0028* 1.2510 1.2510 -0.4180 -0.0237 0.0469 1.2510 1.2510 KML2 -0.0563 -0.0647 0.0605 -0.0547 0.0701 0.0005 -0.1655 0.3641 0.2965 0.3421 0.3900 0.2585 0.9934 0.0073* -0.0419 -0.0881 0.7586 -0.0960 0.1153 0.0006 13.3156 SPEC2 0.0313 0.0422 -0.0886 0.2252 0.0194 0.0182 -0.0002 0.5881 0.4653 0.1406 0.0001* 0.7373 0.7531 0.9978 0.0042 0.0111 -0.2022 0.4177 0.0053 0.0041 -0.0001 NS2 -0.0290 0.1012 -0.0781 -0.0791 -0.1126 -0.0199 0.3618 0.6187 0.0815 0.1974 0.1864 0.0527 0.7322 <.0001* -0.0059 0.0408 -0.2272 -0.0371 -0.0479 -0.0069 0.9408

Notes: 1. ND (National = 0, Non-national = 1) denotes dummy variable to differentiate national cars (Proton and Perodua) from non-national cars. 2. UD (Non-luxury = 0, luxury = 1) denotes dummy variable to identify possible perceived ostentatious value attached to foreign cars. * significant at 1 per cent level. † significant at 5 per cent level.

For example, when a car buyer considers buying a car of a certain price range, it is highly likely that the price of second choice is also within the same range. Similarly, car buyers compare specification of first choice with specification of second choice. Since in this research, specification of cars is either full specification or non-full specification, the result suggest that buyers who consider full specification of first choice also consider full

175 specification of second choice. In other words, buyers did not consider full specification of first choice and non-full specification of second choice in their decision-making process.

In short, car buyers compare prices, horsepower, specification of cars and size of cars when making their buying decision. However, statistics show that car buyers consider fuel consumption of first choice but, do not consider fuel consumption of second choice. This is probably due government’s control on fuel prices.

The statistics also suggest higher fuel efficiency of first choice is compared to relatively cheaper prices of second choice. For example, there is a great chance to find buyers who consider a more fuel efficient car make like Toyota Vios and compare it to a cheaper model of second choice, either of the same make say, Toyota Avanza which is cheaper than Toyota Vios or other car makes that are cheaper such as Honda Jazz, a hatch back model.

Correlation coefficient for car characteristics of first choice and prices of second choice suggest that decision-making process is not a simple one-to-one process. That is, car buyers may not comparison a car characteristic of first choice with similar characteristic of the second choice such as comparing horsepower of the first choice to horsepower of the second choice or specification of the first choice to specification of the second choice. Instead, comparisons may cross between a characteristic of the first choice and the other characteristic of the second choice because prices of car may be so high that their importance out weights car characteristics of both first and second choice. Therefore, it can be found that a car buyer who wanted a five-passenger foreign car chooses a foreign twin-cab light truck like Mitsubishi Triton that has a capacity for five passengers at a more attractive price than a foreign five-passenger sedan of different make like Honda City.

176

5.3 Estimation/Test Results and Analysis

5.3.1 Estimation and Interpretation

This research uses SAS Enterprise Guide software. Model convergence status is reported to be satisfied for all of the logistic regression procedures suggesting that the results of the tests are certain. Estimated β , is the coefficient measuring the relative weights of car characteristics contributing to the mean utility of a car make. In other words, it is the estimation for coefficients of consumer taste on different car characteristics β . The results of estimation are summarised in Table 5.8 below.

Initial stage of analysis involved specification of individual car makes’ utility as a function of car characteristics identified. The results show that while each of the characteristics are statistically significantly different from zero at P-value less than 0.05, the models’ explanatory power is very low. Hosmer and Lemeshow goodness-of-fit tests show P-value that is less than 0.05 suggesting the null hypothesis is not accepted. In other words, test results do not support the assumption that weighted combination of predictors is linearly related to outcome log odds. It implies that the models are not good fit when demand is expressed as a linear function of selected car characteristics.

However, specification of cross factors that is, the cross of car characteristics improves the explanatory power of the model, generating statistics for Akaike’s Information Criterion (AIC) and -2Log L that are about 50 per cent of those in the models without cross factors or interaction of two factors.

Hosmer and Lemeshow goodness-of-fit test results show P-values are more than 0.05, suggesting that the null hypothesis is not to be rejected implying that the models are good fit. For example, in addition to own price, fuel efficiency and horsepower, the cross of price and horsepower, cross of price and number of passengers, and cross of fuel efficiency and number of passengers explain the demand for Toyota cars well.

Table 5.8 overleaf reports factors and crosses of two factors found statistically significantly influencing the demand for individual car makes. Although studies show 177

Table 5.8 Estimation of β by major car makes Car Prices and First choice First & second makes characteristics choice Proton LP1 -328.1 (<.0001) LP1*LP1 34.9929 (<.0001) Perodua LP1 -37.6318 (0.0105) HP1 -112.6 (0.0051) LP1*HP1 29.5837 (0.0019) LP1*NS 0.9906 (0.0350) HP1*NS -3.9703 (0.0092) NS*KML1 -0.0538 (<.0001) Toyota LP1 -33.8343 (<.0001) KML1 1.1146 (0.0058) HP1 -69.8758 (0.0005) LP1*HP1 14.3488 (0.0004) LP1*NS 0.5711 (0.0038) KML1*NS -0.2154 (0.0046) Honda LP1 -32.7143 -9.7872 (0.0002) (<.0001) HP1 -79.0034 1.3347 (0.0012) (0.0663) LP1*HP1 15.9638 (0.0012) LP2 3.8589 (0.0059) Nissan LP1 -188.0 (0.0141) LP1*LP1 18.5275 (0.0158)

178

European/ LP1 -7.5684 American (0.0002) LP1*HP1 -0.3543 (0.0138) Small LP1 -3.1427 Asian (0.0034)

Notes: Diagnostic test results are attached in the Appendices. p-values in parentheses.

that prices and car characteristics are linearly correlated to the demand for car, this is not observed in Malaysia’s case. This is because the presence of correlation between two variables. For example, there is a correlation between the size of engine and the size of car measured by number of passengers. Both size of engine and size of car also influence car price. Hence, contrary to the studies that show linear correlation between horsepower and the demand for cars, models in this research explain the demand well when crosses of a pair of variables are specified.

Although collinearity is not a violation of the assumption made in linear regression, it is essential to identify and recognize the existence of collinearity in this research. This is because of the presence of collinearity and its implication on the car demand ultimately. To determine the right cross characteristics, each car characteristics are tested for their significance and a correlation matrix is generated. Subsequently, different combination of any two car characteristics based on their significance in the tests performed earlier, are crossed and tested in Logistic Regression models.

Prices and characteristics of the second choices are statistically significantly not different from zero for all car makes except Honda. Models show that prices of second choice influence the demand for Honda, the models generally do not to capture substitution pattern of other car makes. This is possibly due to lack of substitution effect across different car makes or due to relatively small sample size to capture the substitution effect. For example, substitution pattern in Proton’s demand is not captured in this research. This is possibly due to about 26 per cent of the Proton car buyers responded to the survey

179 considered other Proton’s model for their second choice while about 42 per cent of the Proton car buyers did not indicate their second choices. The possible reasons for buyers who did not indicate their second choice are either they did not have a second choice at the time buying decisions were made or they did not want to spend time to answer the questions related to the second choice.

The lack of substitution pattern observed in this research will be discussed further in Chapter 7.

The utility functions of both Proton and Nissan are unique in that they are functions of their own prices while the other car characteristics like horsepower, fuel efficiency and number of passengers are statistically insignificant. Association of predicted probabilities and observed responses statistics, Gamma and c values improve tremendously when their utility functions are quadratic. Hosmer and Lemeshow goodness-of-fit test show p-value greater than 0.05. Therefore, the null hypothesis that weighted combination of predictors is linearly related to outcome log odds is not rejected suggesting that a quadratic function is a better fit for both Proton and Nissan’s demand functions. It is therefore deduced that the demand for the cheapest car makes in the domestic cars category and in foreign cars category may be the demand for basic needs for transport.

Refer to Table 5.8, the demand for the dominating local car make and dominating foreign car make, namely Perodua and Toyota are influenced by the most car characteristics, reflecting more car characteristics the car buyers consider in the process of making their buying decisions than other car makes.

Factors that influence the demand for Perodua cars are prices and horsepower of the cars. The coefficients for LP1 and HP1 are negative suggesting quantity demanded for Perodua cars fall as the price increases but, it is misleading to suggest that demand falls as the horsepower increases. The negative relationship between demand and horsepower suggests that there is greater tendency for Perodua car buyers to buy lower horsepower models than higher power models. This interpretation is consistent with the estimated interaction term for LP1*HP1 that is positive. This interaction term suggests higher prices of Perodua cars tend to be associated with higher horsepower.

180

The interaction term of HP1 and NS for Perodua is -3.9703 although there is a positive interaction between NS and LP1 in the correlation matrix. Such inconsistency may be due to same level of horsepower available in the 5-passenger models and 7-passenger models.

Although fuel efficiency does not have immediate effect on the demand for Perodua cars, there is negative interaction between NS and KML1. Such relationship shows that larger models of Perodua cars tend to be less fuel efficient than the smaller models. This observation is consistent with the findings at aggregate level in Table 4.8 (p. 142).

In short, the mean utility of Perodua cars is contributed by Perodua cars’ own price, horsepower (HP1), fuel efficiency (KML1) and the number of passengers (NS1). Second choices’ characteristics are however statistically not different from zero. Observation of sample data shows approximately 29 per cent of Perodua car buyers considered Perodua’s other models for the second choices, 34 per cent of the buyers did not indicate second choice, and 37 per cent considered other car makes for their second choices.

Estimation of interaction terms show that the demand for Toyota is influenced by price, fuel efficiency and horsepower. The demand for Toyota is consistent with microeconomic theory although over the time, the demand for foreign cars have been showing positive trend over the years since the Asian Financial crisis 1997. Positive coefficient for KML1 shows that fuel efficiency has positive contribution to the demand for Toyota cars. Negative coefficient for HP1 suggests that based on the samples selected, there is a greater tendency for Toyota car buyers to buy smaller models such as Toyota Vios and Toyota Avanza than larger models such as Toyota Camry and Toyota Fortuner. In other words, smaller Toyota models constitute to larger proportion of sales than other larger models.

Positive interaction term for LP1 and HP1 suggests that prices of Toyota cars tend to be higher for the models that have higher horsepower. Positive interaction term for LP1 and NS suggests that larger Toyota cars measured by number of passenger tend to be more expensive. Similar to Perodua cars, interaction of NS and KML1 for Toyota is negative, suggesting that larger Toyota cars tends to be less fuel efficient.

181

Honda is the only car make that shows significant influence of second choice at market level. The demand for Honda cars is negatively related to their price. Similar to Toyota, negative correlation between the demand for Honda and horsepower suggests that smaller Honda models such as Honda City and Honda Jazz contribute to relatively larger proportion of Honda’s sales than larger Honda models such as Honda Accord and Honda CRV. Statistics also show that higher horsepower is correlated to higher prices.

However, when second choice is included in the logistic regression model, the coefficient for horsepower becomes positive. This implies that when second choice is accounted for, car buyers prefer Honda to second choice and there is a preference for Honda’s relatively larger horsepower over characteristics of the second choices. The demand for Honda cars is positively related to the price of second choice, reflecting the substitution effect.

Demand for both the European/American cars are also negatively correlated to their prices although it is generally perceived that European and American cars are luxury cars and that they may have ostentatious value. Estimation of interaction term for the European/American cars shows negative interaction term of price and horsepower. The interaction term suggests that an important characteristic constituting to the demand for European and American cars at aggregate level, is horsepower. Negative sign of the interaction term implies that horsepower is not positively correlated with price like other car makes identified in this research. It is highly likely that there is at least a factor that has influence on the car demand but, it is not specified in the model. Such factor may be a moderating factor that changes the direction of horsepower’s influence on price (Hair Jr. et al. 2014, p. 37).

Results show that demand for other small Asian cars is statistically influenced by price only. It cannot be concluded that price is the only factor influencing other small Asian cars because the average prices of other small Asian cars are greater than Toyota, Honda and Nissan’s cheapest models. Lack of statistical evidence to show the significance of car characteristics on other small Asian cars’ demand may also be due to the very small sample size used in this research. There may be qualitative factors that car buyers take into consideration in their buying decisions. This research is unable to determine other

182 factors that may influence the demand for this category of cars because additional data and different methodology will be required.

5.3.2 Price elasticity of demand and deadweight loss

Price elasticity of demand in logistic regression is estimated using the approach suggested in Ai & Norton (2003). The marginal effect of independent variables cannot be taken directly from the estimated b’s and the sign of the b’s does not indicate the sign of interaction between price and demand. To estimate the price elasticity of demand, first order derivative of the logistic regression is obtained. Estimation of price elasticity of demand for individual car makes is at the point of average price and average of any car characteristics such as, average horsepower and average car size measured obtained from the sample if they are found statistically interacting with price in the demand for a particular car make. Therefore, price elasticity of demand for Toyota for example, is the effect of price change, and horsepower and number of passenger’s contribution to the price. Price elasticity of demand for Proton cars is the first derivative of the logistic regression equation at Proton’s average price because price is the only factor that influence Proton’s demand while car characteristics are found statistically not different from zero.

Due to the approach taken to estimate price elasticity of demand and aggregation of all models of each car makes, the estimated statistics for a car make will be equal across different models of different horsepower and of different car size. For example, Table 5.9 shows price elasticity of demand for Honda to be equal for all Honda’s models namely: City, Civic, Accord and CRV.

Estimates show that prices of the car are negatively correlated to the demand for all car makes. Despite the general perceived quality of the European and American cars, price elasticity of demand does not reflect ostentatious value of the cars that is, the price elasticity of demand is not positive. However, the negative sign can be interpreted as within the group of luxury cars, the demand for luxury cars are negatively correlated to their prices. As such, substitution effect may be observed among the different models of

183 the same car makes or between two different car makes. Answers to an open-ended question reveal that there is a small number of Mercedes Benz buyers consider other Mercedes Benz models for second choice in the decision-making process. However, due to limited sample size and thus limited data available, test results cannot conclude substitution between models of the same car make in this research.

The demand for cars in Malaysia is overall price elastic that is, the demand is sensitive to price changes, reflecting the society’s perception of cars being luxury goods. Refer to Table 5.9, within the range of cars of which the horsepower is below 1800 cc, the demand for Proton cars is far less sensitive to price than the demand for Perodua cars, Proton’s close substitutes and to Japanese car makes. Elasticity of demand suggests that Proton cars are luxury for its niche market, the lower income group although the higher income group is likely to consider other car makes to be normal or luxury goods.

Estimation shows that overall, price elasticity of demand is the highest for Nissan. Demand for foreign cars in general, is more elastic than demand for domestic cars. This may imply that foreign cars are perceived to be more luxurious than domestic cars. Other factors that may contribute to the higher price elasticity of demand for foreign cars are stiffer competition among the foreign brands in Malaysia, and households’ large proportion of income spent on foreign cars.

Although price elasticity of demand for all makes are highly elastic, test results of the model do not show the significance of second choices’ price and/or characteristics except for Honda. Models in this research do not show substitution effect between two different car makes because majority of the sample car buyers considered other models of the same car makes as second choice. Therefore, when logistic regression model is specified for Toyota for example, Toyota cars are assigned ‘1’ and if the second choices are also Toyota models, they are assigned ‘1’. Consequently, substitution effect is not fully reflected in the logistic regression analysis result. As such, estimation for cross price elasticity of demand is not feasible.

184

Table 5.9 Comparison of Price elasticity of demand (h) for selected car makes - by horsepower (cc) Car makes Average Tax Price Quantity Estimated real price rates elasticity of in 2014 Quantity (after tax) (%) demand (h) (Qo) without tax (RM) (Qn) Horsepower below 1800 cc Proton 51,613 85 -1.73 115,783 485,490 Perodua 44,656 85 -4.79 195,579 1,928,485 Toyota 85,615 115 -6.59 68,735 1,042,001 Honda 86,696 115 -6.58 54,690 827,896 Nissan 70,988 115 -8.24 32,500 608,186 Horsepower ranges from 2000 to 2499 cc Toyota 122,668 130 -6.59 5,700 92,041 Honda 129,209 130 -6.58 11,405 183,898 Nissan 156,364 130 -8.24 3,000 33,825 Other Asian 99,112 130 -3.14 65,598 539,754 Horsepower above 2500 cc Toyota 128,393 145 -6.59 27,600 472,939 Honda 145,000 145 -6.58 11,400 195,062 European/American 193,150 145 -6.60 51,024 876,082

Notes: Tax rates for Proton and Perodua are summation of excise duties and sales tax. Tax rates for foreign cars are summation of import duties that vary across four classes of engine capacity, excise duties and sales tax. Price elasticity of demand is not constant price elasticity of demand because cars are highly differentiated products.

Source: Malaysian Automotive Association (2018).

185

At average Honda price of RM101,059, average horsepower of 1.7, and average second choice’s price of RM91,837, the cross price elasticity of Honda is +0.36. Sample statistics show that approximately 69 per cent of Honda car buyers consider other car makes for the second choice while 31 per cent consider other Honda models either more expensive or cheaper models. Although approximately 70 per cent of the Honda car buyers consider other car makes for their second choice, cross price elasticity demand for Honda is inelastic. This may be due to the relatively higher price of Honda cars than their substitutes. The result suggests that it takes greater change in the price of second choice to influence a relatively small change in the demand for Honda. This further suggests that there may be unobservable Honda cars’ characteristics that influence Honda cars’ demand. However, it can also be argued that although the model specified is sufficient to explain Honda cars’ demand, the model may be better specified if additional observable and/or unobservable characteristics may be added to the model.

It is therefore concluded that although the model specified is a good fit for Honda cars, there remains room for in depth study of the demand for Honda cars.

5.3.3 Cost of protectionism

The estimated cost of protectionism is approximately RM169 billion, that is about 16.7 per cent of real GDP in 2014 or approximately 73.6 per cent of the manufacturing sector’s real GDP contribution (see Table 5.10) When tariffs are reduced by 50 per cent, deadweight loss decreases tremendously. At half the current tariff rates, the cost of protectionism is 4.5 per cent of the manufacturing sector’s GDP contribution or 1 per cent of GDP.

Statistics suggest that the cost of protectionism increases faster than the increase in tariff rates. Reduction of tariff rates will lead to more than proportionate decreases in the cost of protectionism.

186

Table 5.10 Cost of protectionism: Current tariff rates and 50 per cent reduction Tariff reduction by Current tariff 50 per cent rates Percentage of manufacturing sector’s 4.5 73.6 GDP contribution (2014) Percentage of GDP (2014) 1.0 16.7

Deadweight Loss in RM 10.4 billion 169 billion

As the number of new cars registered over the years increase, the loss of consumer surplus and deadweight loss borne by the society is likely to increase. In the event that the demand for domestic cars falls while the demand for foreign cars increases over the years, the cost of protectionism is highly likely to increase faster. It is deduced that the cost of protectionism demonstrates positive trend because car buyers are willing to pay more for foreign cars and the numbers of new foreign car makes registered per year have been increasing since after the Asian financial crisis 1998.

The cost of protectionism in this research is underestimated for the following reasons:

(a) Total tariffs imposed on foreign cars of larger horsepower are about 30 to 60 per cent more than the tariffs imposed on domestic cars. This estimation of tariffs is based on horizontal summation of all types of tariffs imposed. In Malaysia, sales tax is based on the price after import duties and excise duties. Hence, the tax effect demonstrate geometric progression instead of arithmetic progression, and

(b) Theoretically, price elasticity of demand is greater at the upper portion of a demand curve. As such, price elasticity of demand is likely to be greater for more expensive cars that is, cars of larger horsepower. Estimating price elasticity at the point of average price and average horsepower therefore, underestimate the price elasticity of expensive cars and hence, underestimating the cost of protectionism for this segment.

187

5.3.4 Socio-Economic Factors that influence the demand for cars in Malaysia

Socio-economic factors that are often found linearly related to demand for goods in many studies do not demonstrate linear relationship with the demand for cars in this research. For example, result shows there is no direct relationship between income and demand for cars when model specifies income as one of the explanatory variable.

Table 5.11 Estimation of interaction terms bo: car characteristics and car buyer attributes Car characteristics and car-buyers attributes bo (First choice only) Price Constant -115.4826 (<.0001) Income 5.6750 (0.0004) Race (Natives1=1) -47.3660 (0.0010) Fuel efficiency Race -0.2242 (<.0001) Horsepower Gender (Male=1) 11.8952 (0.0176) Number of passengers Income 0.0011 (0.0002) Gender 0.0059 (0.0276)

Notes: 1. Natives are called Bumiputra, mainly made up of Malays and other smaller groups of natives. Non-natives are generally made up of other ethnic groups such as Chinese and Indians. p-values in parentheses.

Estimated interaction term (see Table 5.11) shows positive interaction between income and price. This implies that as income level increases, the demand for more expensive

188 cars will increase. This is reflecting the upward trend of new foreign cars registered over the years while the prices of foreign cars increase.

The direct relationship between income level and the demand for foreign cars in a protected car industry suggests that income’s effect on the demand for cars is not easily measured. This is because cars are highly differentiated products. Car characteristics that change as technology advances, may have influence on the demand for cars.

Sample shows that there are car buyers in lower income group buy medium or higher range cars. This is common in Asian countries because it is often found that the older generations accumulate wealth for the younger generations. As a result, there is demand for medium and higher range cars by the lower income younger generation who benefit from the wealth accumulated. Although wealth is a factor influencing car demand, it is not included in the survey and hence, not in the model because data is not available.

Another reason for no statistical evidence of direct correlation between income and demand for car in this study is due to highly distorted car prices. Tariffs and non-tariff barriers have caused foreign cars highly expensive relative to domestic cars. In the domestic car market segment, non-tariff barriers such as guaranteed loan approval for civil servant, domestic car makers’ marketing strategies to capture the lower income group, and government’s control on oil prices make domestic car seemingly cheap. As a result, the linear relationship between demand for cars and income is blur.

However, when income is allowed to interact with price in a discrete choice model, the results show different income levels interact with different car price levels in which the latter is captured by different car makes. This shows an advantage of applying discrete choice model in Microeconomics.

In logistic regression models, related social factors are allowed to interact with car characteristics. Test results show that there are interactions between a social factor and a car characteristic for example, gender and horsepower, both simultaneously influence the demand for car. Gender is also found weakly interacts with number of passengers in their influence on car demand.

189

Statistics also show that race is a social factor influencing car demand in Malaysia. Although both race (RC) and sector of occupation (SEC) are highly correlated, statistics show that race influences the demand for car instead of sector. Negative interaction terms between race and prices implies that the natives in Malaysia are more likely to buy local cars than the non-natives. This is most likely due to the incentives in the form of attractive interest rates and guaranteed loan approval given to the civil servants who are predominantly natives, if they buy Proton cars.

Negative interaction between race and fuel efficiency measured by kilometre per litre petrol implies that there is a weak correlation between the natives and lower fuel efficiency. Race and fuel efficiency do not influence car demand separately but, both the factors combine and exert influence over car demand. This reflects government’s protection for domestic car makers and for natives’ statutory rights for privileges. Due to the incentives given, natives are more incline to buy Proton cars, and since Proton cars are less fuel efficient, then the association between natives and fuel efficiency can be observed. However, this model does not examine the nature of the relationship that is, this model does not differentiate between immediate correlation or indirect correlation. It is not the objective of this research to identify and analyse variables that mediate correlation between two variables. Hence, it can only be suggested that there may be a mediating variable that influence the correlation between race and fuel efficiency.

There is very little statistical evidence of correlation between horsepower and size of cars in this research. This is because at market level, there are mini MPVs such as Toyota Avanza, Perodua Alza, and that have carrying capacity of seven passengers at horsepower of as low as 1600 cc while other models of sedans that have carrying capacity of five passengers at the same level of horsepower. As a result, statistics show that gender is positively and relatively strongly correlated to cars of higher horsepower but, weakly positively correlated to larger cars, measured by the number of passengers. In other words, male car buyers are more likely to buy larger cars and cars of higher horsepower than female buyers.

Since Proton supplies seven-passenger cars at a lower price than five-passenger foreign cars, this may have given rise to outliers for observation of correlation between price and

190 size of cars, and correlation between income and size of cars. Although income remains positively correlated to the size of cars, the weight of outliers may be the reason for the weak correlation between income and size of cars.

Table 5.12 shows the estimation of income elasticity of demand for cars of horsepower below 1800cc. As it is misleading to estimate income elasticity of demand based on the overall sample’s average horsepower, estimation of income elasticity is based on horsepower of the best-selling models of the major car makes. For horsepower below 1800cc category, the average price of domestic and foreign cars are RM48,135 and RM81,100 respectively.

Table 5.12 Income elasticity of demand: Domestic and foreign cars below 1800 cc. Groups of car make Average price1 Race2 Income elasticity of demand3 Domestic 48,135 1 28.7 Foreign 81,100 0 20.3

Notes: 1. Average price of domestic cars is the average price of Proton and Perodua’s best- selling models, not weighted by their sales. Average price for foreign cars is the average of Toyota, Honda, and Nissan cars, not weighted by their sales. 2. Race of car buyers selected for estimation is based on majority of the car buyers. For example, since up to 60 per cent of the domestic car buyers are Natives, estimation of income elasticity of demand takes Race=1. The proportion of non-native foreign car buyers for Toyota, Honda and Nissan are 74, 67, and 70 per cent respectively. 3. Income elasticity of demand is unlikely to be constant elasticity of demand because cars are highly differentiated products.

Since majority of domestic car buyers are native while majority of foreign car buyers are non-natives, the dummy variable’s values assigned to the demand functions are 1 and 0 respectively. Based on the results in Table 5.10, interaction between income and race gives estimation for income elasticity of demand 28.7 and 20.3 for domestic cars and

191 foreign cars respectively. The results suggest that the demand for domestic and foreign cars are highly income elastic.

Table 5.13 Estimation of interaction term bu: unobservable car characteristics by car makes Unobservable car characteristics - by car makes bu Proton 642.0949 (<.0001) Toyota and Honda 75.0273 (<.0001) Nissan 401.6677 (<.0001) Notes: p-values in parentheses.

Following the estimation of bo, interaction terms of unobservable car characteristics, bu are estimated. Table 5.13 shows that there may be unobservable car characteristics attached to Proton, Toyota and Honda as a group, and to Nissan. Contrary to the finding in Table 5.10 that none of the car characteristics identified in this research influence the demand for Proton, interaction term for unobservable Proton car characteristics show that unobservable car characteristics carry relatively large weight in its influence on Proton cars’ demand. Similarly, while none of the car characteristics identified influence Nissan cars’ demand, interaction term for unobservable Nissan car characteristics also carries large weight in its influence on Nissan cars’ demand.

A small number of answers to open-ended question suggest that wait time at service centres or perceived quality of after sales service influence car buyers’ decision. The presence of unobservable characteristics may also implies possible omission of characteristics not specified in the model. In Proton’s case the unobservable characteristics may be government’s incentives that are difficult to measure.

192

The explanatory power of model improves and the interaction term is statistically significant when Toyota and Honda are grouped together. This implies that both Toyota and Honda may have common unobservable car characteristics. Unobservable car characteristics however, are statistically not different from zero for all other car makes.

Despite the general perception that luxury goods possess ostentatious value, there is no statistical evidence for the presence of unobservable car characteristics demand for car makes that are perceived to be luxuries.

Although there are some qualitative factors associated with specific car makes identified by car buyers during survey, more in depth studies of these qualitative variables are needed to measure their relative importance in influencing the demand for different car makes.

For the purpose of identifying qualitative factors that are associated with specific car makes, answers to optional open-ended questions are summarized and manually matched to their respective car makes. The results of tests for unobservable car characteristics and summary of the optional open-ended question in survey questionnaires suggest the followings:

(a) Proton’s unobservable characteristics are factors appealing to the lower and medium-low income groups. These factors may be offers or privileges that have become a bundle of characteristic attach to the Proton cars;

(b) both Toyota and Honda may be enjoying the advantage of their reputation built up since their establishment in the 1960s. Relatively short waiting time for car parts compared to other foreign car makes that do not have manufacturing plants in Malaysia may contribute to the positive influence on the demand for Toyota and Honda cars; and

(c) Nissan may have unobservable characteristics associated with after sales services and value for reliability that are not measurable and observable in the models applied in this research.

193

Table 5.14 Estimation of interaction term bu: unobservable car buyer characteristics Preference influenced by unobservable buyer characteristics bu

Local versus Foreign cars -153.4743 (<.0001) Toyota and Honda versus other car makes 38.7797 (<.0001) Proton -36.4789 (<.0001)

Notes: p-values in parentheses

Finally, residuals are obtained to test for presence of unobservable car buyer characteristics. Refer to Table 5.14, there are unobservable car buyer characteristics that differentiate buyers of local cars namely: Proton and Perodua as a whole, and buyers of foreign cars. There may be buyer characteristics that differentiate Toyota and Honda car buyers as a whole, from buyers of other car makes while there is also a set of unobservable car buyer characteristics that is unique to Proton car buyers. The possible unobservable car buyer characteristics may be revealed if repeating car buyers can be identified. However, this will require different set of questionnaires and different approaches in methodology.

The model has not been able to capture the unobservable buyer characteristics that are associated with acquisition of luxury cars and characteristics that are associated with lower income buyers who buy luxury cars. This is most probably due to the following:

(a) characteristics of buyers who buy luxury cars are captured by their income level or,

(b) there is a very small sample of buyers who had low income but acquired luxury cars.

194

5.3.5 Efficiency

An overview of accounting and statistical data (refer to Figure 5.2) for the year 2015 shows that plant and machinery constitutes to the largest proportion of input for all major car makers followed by land since the car manufacturing industry is a capital-intensive industry.

Figure 5.2 Input (Ringgit) and output (sales in quantity) - Toyota, Perodua, and Proton

Toyota

Perodua

Proton

0 1000000 2000000 3000000 4000000 5000000 6000000

Furniture, fittings, vehicles Plant & machinery Land Sales(Q)

Notes: Berhad’s accouting period is 1st of April to 31st of March after the acquisition of Proton Holdings Berhad by DRB-Hicom in 2012. Therefore, for consistency in accounting data, annual report for the year 2016 is used.

Sources: UMW Toyota Motor Sendirian Berhad (60576 K) Annual Report 2015; Perodua Manufacturing Sendirian Berhad (95999 T) Annual Report 2015; and Proton Holdings Berhad (623177 A) Annual Report 2016, Companies Commission of Malaysia.

195

It also shows that Proton has the largest amount of plant and machinery as a result of expansion in 1997. Proton’s furniture, fittings, and vehicles is a few times greater than that of the other players, reflecting the larger number of employees in Proton than in Perodua and Toyota.

Perodua’s plant and machinery is about 50 per cent that of Proton but, the former’s sales are almost double Proton’s sales. Large amount of inputs relative to the level of sales suggest that Proton is highly likely to be operating in excess capacity in labour, land and capital.

Table 5.15 Accounting data and Statistical data: Major car makes

Plant Furniture, Average Number Cost of & fittings, Price Revenue of goods sold Land machinery vehicles (RM) (RM'000) employees (RM'000) (RM'000) (RM'000) (RM'000) a b c d Proton 51,612.5 4,834,899 12,000 5,355,763 1,921,903 4,918,284 1,275,947 Perodua 44,655.9 5,095,502 10,000 3,811,987 632,024 2,192,311 191,717 Toyota 96,136.6 10,881,306 3,500 10,191,281 833,320 744,573 204,048 Sources: 1. Proton Holdings Bhd (623177 A) Annual Report 2016. 2. Perodua Manufacturing Sdn Bhd (95999 T) Annual Report 2015. 3. UMW Toyota Motor Sdn Bhd (60576 K) Annual Report 2015.

Table 5.16 Ratio Analysis

Land to Plant and Furniture, fittings, Sales to employees machinery to vehicles to employees employees employees Toyota 238.1 212.7 58.3 29 Perodua 63.2 219.2 19.2 20 Proton 160.2 409.9 106.3 10

196

Table 5.16 summarises combinations of variable and fixed inputs of the 3 largest car makers in Malaysia for the year 2015. Due to lack of information for raw and intermediate inputs, cost of goods sold is used as a proxy for all raw and intermediate inputs used. The proxy for output is the revenue. Data shows that Proton has the largest amount of fixed input combined with the largest number of employees but, has the lowest total revenue.

Ratios of revenue over cost of goods sold for Proton, Perodua and Toyota are 0.9, 1.3, and 1.1 respectively. The ratios suggest that Proton is operating at the output level where average price is less than average variable cost that is, below the shut-down point while Perodua and Toyota are operating above the shut-down point. The ratio also suggests that Toyota is operating relatively close to productive efficiency compared to Perodua.

Refer to Table 5.16, Toyota has the largest ratio of land to employees while Proton has the largest ratios of plant and machinery to employees and, furniture, fittings, vehicles to employees. Toyota’s large ratio of land to employees is attributable to a relatively very small number of employees. Perodua has the smallest ratios of land to employees and ratio of furniture, fittings, vehicles to employees. Ratios of sales in quantity to employees in 2015, for Toyota, Perodua and Proton are 29, 20, and 10 respectively. The statistics suggest that on the average, an employee in Toyota sold three times more cars than an employee in Proton but, one and a half times more cars than an employee in Perodua. In short, Toyota is more efficient than both the domestic car makers, and that Perodua is more efficient than Proton.

Proton’s plant and machinery to employee ratio is twice the ratio of Perodua and Toyota while furniture, fittings and vehicles to employee ratio is twice the ratio of Toyota and up to five times the ratio of Perodua. The ratios of fixed input over variable input confirm that there is over-investment of plant and machinery, and furniture, fittings and vehicles. The intensity of plant and machinery use is exceptionally low.

Both the ratios of plant and machinery to employees and of sales to employees suggest that there is excess capacity in plant and machinery in addition to over-staffing problems in Proton.

197

To compare efficiency of the major car makers, efficiency index is constructed using the approach proposed in Charnes, Cooper, & Rhodes (1978, p. 429). Average prices of the all models are used for estimation of weight for output prices. Prices are not weighted by the relative importance of each models in their respective sales because all the best-selling models of each car makes constitute up to 40 per cent of total sales. In addition, detail information of the sales across models is not available. The weighted outputs are therefore, the products of average prices and sales in quantity. The weighted inputs are the products of the numbers of employee and total inputs where the numbers of employee are the weights to inputs.

Efficiency index constructed (see Table 5.17) using accounting data and statistical data shows that both Perodua and Toyota are more efficient than Proton. Using Toyota’s index as a base, Perodua and Toyota’s efficiency index are 0.55 and 1.00 respectively. Since both Proton and Perodua are in the same category labelled as “national cars”, both are imposed the same rates of tariff. Therefore, comparison of efficiency index between Perodua and Proton can be made. Comparison of Perodua and Proton’s index suggests that Perodua is about three times more efficient than Proton.

Table 5.17 Data Envelopment Analysis - Efficiency index

Sales in Weight Total Weight Efficiency Normalised quantity, input (x) Index (u) (v)1 (uy/vx) (y) a+b+c+d Proton2 115,783 0.54 13,471,897 0.47 0.0098 0.16 Perodua 195,579 0.46 6,828,039 0.39 0.0339 0.55 Toyota 102,035 1.00 11,973,222 0.14 0.0621 1.00

Notes: 1. Number of employees is used for estimation of weight for inputs: (a) cost of goods sold, (b) land, (c) plant and machinery, and (d) furniture, fittings and vehicles. 2. Proton’s accounting period is 1st of April to 31st March. Thus, Proton Holding Berhad Annual Report 2016 is used for the required accounting data reflecting 2015’s performance.

198

Toyota on the other hand, is classified as a “foreign make” thus, subjects to much higher rates of tariff. Since prices of Toyota cars are distorted, the weight assigned to sales is also distorted. As a result, comparison of Toyota’s index with Perodua and Proton will violate the assumption of homogeneity in Data Envelopment analysis (Dyson et al. 2001, p. 247).

To find out how tariffs may affect efficiency of the foreign car maker, tariff imposed on foreign car maker is reduced to estimate a fairer efficiency index. Although tariff and other forms of taxes are imposed on foreign car makers, only reduction of import tariff is taken into account for analysis because import tariff is the largest tax component compared to other taxes such as sales tax.

Published information shows that tariff imposed on Toyota is 30 per cent more than the tariff imposed on Proton and Perodua for horsepower below 1800cc. For comparison of efficiency index between Toyota and two domestic car makers, tariffs on Toyota is reduced by 30 per cent. This analysis exclude Toyota cars with engine sizes of 2000 to 2499cc and above 2500cc because all domestic car engine sizes are below 2000cc (see Table 5.9).

Using the estimated price elasticity of demand for Toyota, new quantity after 30 per cent tariff reduction is obtained. The average price after 30 per cent tariff reduction is used as a new weight for output, that is u.

Variable inputs measured by “cost of goods sold” in financial statement is extrapolated to estimate cost of variable inputs for production of the new output level after tariff reduction. It is assumed that the assembly plants operate at constant return to scale because the plants are more specialized in producing the best-selling model, Toyota Vios in Malaysia.

Fixed assets and number of employees are held constant because the other models that have higher horsepower have lower local content than the best-selling model. In addition, certain models with larger engine sizes are assembled in plants within the region for example, in Thailand and in Indonesia.

199

The assumptions made for construction of efficiency index are as follows:

(a) variable inputs, other than labour, increase proportionately as the output increases;

(b) labour does not increase significantly because the industry is capital-intensive; and

(c) capital mobility within the industry that is, Toyota may use the excess capacity of domestic car makers.

Table 5.18 Comparison of Efficiency Index - UMW Toyota: Before and after 30 per cent tariff reduction Sales in Total quantity, Weight input (x) Weight Efficiency (y) (u) a+b+c+d (v) (uy/vx)

Before 102,035 1.00 11,973,222 0.14 0.0061 After 218,752 1.00 23,630,945 0.14 0.0066

Comparison of the efficiency index shows that under a set of assumptions, efficiency index of Toyota improves by approximately 9.4 per cent after 30 per cent reduction in tariffs. The efficiency index is however, deemed underestimated because of the following reasons:

(a) comparison is made for a single firm under two scenarios. If relative efficiency is taken into account, tax reduction will increase sales of Toyota and decrease sales of other car makes, to even a small percentage. As a result, the normalised index is highly likely to show greater improvement, and

(b) estimated increase in sales is 114 per cent after 30 per cent tax reduction. Ceteris paribus, weight of input will decrease when Toyota reap economies of scale. As a result, efficiency index increases.

200

The results of analysis for tariff reduction suggest that Toyota has comparative advantage in producing cars and X-efficiency may contributes to Toyota’s higher efficiency index than the domestic car makers. Study shows that at country level, Japan is more X-efficient than Malaysia in the 1960s. Refer to Table 5.19 below, in the study of X-inefficiency in 18 countries, Shen (1984) finds that capital and labour used in Malaysia for production of a unit of output is about 3 times the capital and labour used in Japan.

Table 5.19 Production input multiples - Capital and labour: Japan and Malaysia Countries Per capita income Production Input Multiples (1970 U.S. dollar) Capital Labour Japan (1966) 941 0.47 0.86 Malaysia (1969) 325 1.08 2.31

Source: Shen (1984, p. 102).

The study also finds that Japan’s per capita income was about 3 times Malaysia’s income in 1970. In 2018, Japan and Malaysia’s GDP per capita are USD39,286.7 and USD11,238.9 respectively that is, Japan’s income is about 3.5 times Malaysia’s income (The World Bank, 2019). Therefore, it can be concluded that Japan’s X-efficiency may have contributed to Malaysia’s car industry through operation of Japanese car maker in Malaysia.

5.4 Conclusion

5.4.1 Car characteristics and car buyer characteristics

Preliminary test results show the presence of the followings:

(a) there is a weak association between income group (YGRP) and car size measured by the number of passengers (NS), implying that higher income group tend to buy larger cars but, this association is weak. This weak association reflects Proton and Perodua’s relatively cheaper 7-passenger models than foreign cars;

201

(b) there is a positive but weak correlation between the number of dependents (DEP) and numbers of passenger (NS), implying that car buyers who have more dependents tend to buy larger cars;

(c) there is a positive but weak correlation between number of cars in a household (NOC) and income (Y). That is, statistics suggest that households that have relatively high income tend to own more than a car; and

(d) there is a positive correlation between NOC and prices of car, and between NOC and horsepower. In short, results show households that own more than a car are associated with relatively more expensive cars, and households that own more than a car tend to own cars that have larger horsepower.

Logistic Regression results show that there is are relationships between car characteristics and demand for car. In other words, the demand for cars is influenced by prices of cars, horsepower, car size and fuel consumption. Car characteristics may influence the demand for cars through the formers’ interaction with car prices or through interaction with other car characteristics.

Results also show that when classified cars into “domestic” and “foreign” categories, the demand for Proton and Nissan, the cheapest in their respective categories, are influenced by their own prices. On the contrary, demand for the most competitive car makes Perodua and Toyota, are influenced by the most numbers of car characteristics. This implies that the most competitive domestic and foreign car makes differentiate themselves from their competitors to add value to the buyers within their segments.

Among all the social-economic factors identified in pilot tests, Logistic Regression analysis results show that income, race and gender have statistically significant influence on car demand through their interaction with car characteristics. Since civil servants are predominantly the natives and government incentives are given to civil servants, sector (SEC) and race (RC) are both tested. Results of logistic regression show that race has negative interaction with car price and fuel efficiency while SEC is statistically highly insignificant. While the preliminary test results support the positive correlation between 202 the numbers of dependent (DEP) and car size (NS), the numbers of dependent generally has not been a significant factor that influences the demand for cars in the Logistic Regression model.

Although fuel-efficiency is found statistically insignificant at market level, it is statistically not different from zero at product level. That is, car buyers take into account fuel efficiency of cars when buyers compared two car makes of their choices. The statistical inconsistency in market level and product level may be due to oil price control in Malaysia. Although the government has cut oil subsidies and allowed oil price to be more flexible since 2014, car demand overall remains insensitive to fuel efficiency of the cars.

Unlike the studies carried out in developed economies, models in this research do not suggest linear correlation between car characteristics’ and car demand for each car makes. This is possibly due to the following reasons:

(a) Inadequacy of models. Excessive tariffs on cars and car parts may have distorted the prices to a great extent that Logistic Regression is unable to capture the correlation between car demand and individual car characteristics;

(b) Effects of very high car prices relative to income level. Excessive tariffs may have given rise to double-taxation and/or tremendously high prices to the extent that car characteristics are weighed far less than the prices of the car; and

(c) Product differentiation. In an environment where there are rapid technology changes, cars are so highly differentiated that car characteristics have become collinearly related.

The interaction term for unobservable car characteristics is significant when Toyota and Honda are grouped together. There may be qualitative factors that are attached to the car makes due to “historical accidents” that allowed both the old car makers in Malaysia to gain dominance (David 1985, p. 335). Other possible unobservable car characteristic may be after sales service and wait time at service centres that are becoming important influence on car demand. 203

5.4.2 Price elasticity of demand and cost of protectionism

Demand for foreign cars is highly elastic. Estimated price elasticity of demand for foreign cars range from -3.1 to -6.6 compared to that of domestic cars that range from -1.7 to - 4.8. Despite excessive tariff imposed on cars, the demand for cars continues to increase every year with the numbers of new foreign cars registered increase faster than the numbers of new domestic cars. This reflects the ineffectiveness of tariffs to stimulate the demand for domestic cars. Excessive tariff rates imposed on foreign cars may have made foreign car luxury goods. Consequently, as the society becomes more affluent, the demand for foreign cars increases.

Excessive trade barriers and government’s lack of commitment to remove or reduce trade barriers may have detrimental effects on the growth of infant industry. The protected infant industry has little motivation to acquire new technology hence, unable to compete with foreign car makers.

The estimated total taxes paid by car buyers in 2014 is approximately RM27.8 billion which is about 11.9 per cent of manufacturing sector’s GDP contribution. The deadweight loss is approximately 73.6 per cent of the manufacturing sector’s GDP or 16.7 per cent of real GDP in 2014. When tariff rates are reduced by 50 per cent, the cost of protectionism decreases faster to RM10.4 billion from 169 billion at current rates. The high cost of protectionism implies that costs arising from trade barriers overwhelms the benefit of trade barriers in the form of tax revenue. As the number of new foreign cars registered increases over the years, the amount of taxes paid and deadweight loss also increase.

The excessive tariff paid by car buyers, not gained by the society as a whole is so large that the amount is more than half the GDP contribution of manufacturing sector, implying the significance of unproductive activities arise as a result of excessive tariff and cost associated with misallocation of resources in terms of labour and capital.

At the estimated price elasticity of demand, cost of protectionism decreases more than the proportionate decrease in tariffs. This implies that cost of protectionism increases faster

204 than the increases in tariff rates. Hence, regardless of how collected tariffs are used, the cost of protectionism is highly likely to be overwhelming and offset the benefits of tariffs.

Finally, it can be concluded that trade barriers have ceased to be effective in protection of domestic economy’s infant industry.

Although the demand for car is highly price elastic, test results do not show significant substitution effect for all selected car makes except for Honda. Cross price elasticity of demand for Honda is inelastic implying that Honda car buyers demand Honda’s horsepower and possibly other characteristics not identified in this research.

Lack of statistical evidence in substitution effect despite high price elasticity of demand is due to the following reasons:

(a) Model specification at car make level. Since analysis is made at car make level instead of car model level substitution effect within the same car makes but different models is not captured by the logistic regression models. Substitution between two models of the same car make may also implies the presence of unobservable car characteristics attached to a particular car make;

(b) Lack of data. Data shows that the first choice and second choice of many buyers are of the same car makes. As such, larger sample size is more desirable to make analysis at models level feasible;

(c) Lack of choices. Due to lack of choices, car buyers may consider two different models of the same car make although the price differences between two models may be large; and

(d) Information of second choice is elicited at the time car buyers are buying new cars. However, substitution may take place when car buyers decided to buy new cars. For example, a car buyer who bought a domestic car ten years ago plans to buy a new car today. Due to a change in his/her preference, the car buyer now consider two foreign cars. As such, survey captures car buyers’ plans today but, not his/her change in preference over the years. 205

At aggregate level, income elasticity of demand for domestic cars and foreign cars are 28.7 and 20.3 respectively. Statistics show that the demand for car in Malaysia is highly income elastic, implying that cars are luxury goods for their respective niche markets.

5.4.3 Efficiency

Efficiency indexes show that Perodua is more efficient than Proton. Although Toyota’s efficiency index is up to six times greater than that of Proton, it is difficult to conclude that Toyota is six times more productive than Proton because of different tax treatment.

The results of analysis show that when tariffs are reduced by 30 per cent, Toyota’s efficiency index improves. Hence, it is concluded at the same tariff level, Toyota is more efficient than the domestic car makers.

Although it may be argued that Toyota may reap economies of scale after tariff reduction, the efficiency index, after adjustment for tariff difference may not be significantly different from the estimated efficiency index before tariff reduction. This is because while tariff removal may reduce Toyota car prices that are used as weight for the sales (u), tariff removal also increases the sales (y). However, as the demand for Toyota car is highly price elastic, tariff removal may cause a greater increase in the sales than falls in the prices.

Inefficiency of Proton is attributable to excessive allocation of fixed factors such as land, and plants and machinery by the government and excessive variable factor, labour. Labour’s contribution to X-inefficiency in Malaysia may be attributable to the protected environment that discourage learning and subsequently, unsuccessful technology diffusion. Other problems that plague the domestic car maker may be saturated market coupled by insignificant repeating buyers or replacement, and excess capacity.

Logistic regression analysis, estimation of price and income elasticity of demand, and analysis of efficiency index conclude the possible significant welfare cost decreases if tariffs are reduced or removed. Tariffs removal allows foreign car makers such as Toyota

206 and Honda, that have been operating in Malaysia to reach potentially far higher production level and achieve greater efficiency. Freer trade will also improves variety of choices for domestic car buyers.

207

Chapter 6 Discussion

6.1 Introduction

The Logistic regression models in this research specify the demand for cars as functions of a set of car buyer characteristics and a set of car characteristics. Car buyer characteristics selected for this study are the characteristics identified in a priori studies and a few characteristics that are relevant in Malaysia but not included in the a priori studies. There are some car characteristics identified in a priori studies are not specified in this research because information is not available in Malaysia and car owners are not given such information for comparison when making their buying decisions. The car characteristics omitted in this study are carbon emission and pickup payload. Fuel consumption is included in this study although this piece of information is not available to car buyers. The statistics of fuel consumption are obtained from some blogs of car owners who share their information and experiences with other owners and from websites managed by organizations for car enthusiasts.

Although selected car characteristics are found statistically significant influencing the demand for car, most of them influence car demand through their interactions with prices and other characteristics. This is contrary to many studies carried out in the developed economies in which car characteristics have linear correlation with the demand for car. Apart from the differences in car characteristics’ influence on demand, this research does not find significant substitution effect between two different car makes except for Honda where statistics show that prices of second choice influence the demand for Honda cars.

The results of logistic regression are used for estimation of own price elasticity of demand, income elasticity of demand, cross price elasticity of demand, and deadweight loss. In addition to logistic regression analysis, this research also adopts Data Envelopment Analysis in the attempt to investigate how the cost of protectionism is imposed on the society through misallocation of resources when missing true data is a major problem.

208

The subsequent section of this chapter revisits the literature and results of both the Logistic Regression models and Data Envelopment Analysis. Due to highly distorted prices and lack of transparency in government policies, unpublished information obtained during face-to-face interviews with industry players is used to complement to the statistical test results.

6.2 Benefits of Trade

The first trade model that is, the Ricardian model of trade and the subsequent models such as Heckscher-Ohlin model and Standard trade model, say that comparative advantage is the underlying reason for trade. In these models, if an economy specialized in producing goods of which it has comparative advantage, more of the goods can be produced at relatively lower opportunity cost and subsequently, able to reap economies of scales. As trade takes place, larger variety of choices will be available hence, trigger competition among economies. Competition in turns, leads to efficiency, technology transfer and innovation.

Malaysian government’s decision to found its own home brand car is not based on comparative advantage to begin with. The domestic car maker has been protected since it was founded in the 1980s. The numbers of car makes available in Malaysia during the period 1980s and 1990s are not available. During the period from 2002 to 2015, the number of passenger car makes found in Malaysia has increased by about 46 per cent, from 24 to 35 car makes as there was pressure to liberalize the car industry (MAA 2002 & 2016). Although liberalization of car industry is not complete, statistics show that variety of choice has improved tremendously since early 2000s.

It is difficult to attain a meaningful measurement of cost associated with efficiency at firm’s level in Malaysia because of missing data. The presence of unfair and untrue published data due to lack of transparency and excessive distortion requires careful selection of published data and building of industry players’ network to elicit qualitative information for better understanding of how the industry operates.

209

Based on market level data, Figure 2.2 (in Chapter 2) shows that beginning 1998, the number of new foreign car makes registered increases steadily while the number of new Proton registered declines and remains fluctuating around 150,000 units in domestic market. Statistics for Proton’s export in the 1980s and 1990s are not available. Limited published statistics show that Proton’s exports were 7 per cent and 14 per cent of Proton’s sales in year 2004 and 2008 respectively (MITI 2009). Although the export statistics have shown increases in growth rates over four years, the increase is not as high as 7 per cent because Proton’s yearly sales have been declining steadily over the years. A recent report says that Proton’s export resumes in August of 2018 (New Straits Times, 9/8/2018), suggesting that Proton has not exported cars to foreign economies possibly during the period of early and mid of 2010s. Proton is operating at about 50 per cent and 70 per cent excess capacity in year 2014 and 2016 respectively at the given production capacity of 230,000 cars per year.

Low efficiency index based on accounting data in 2015 and 2016 (see Table 5.16) suggests that Proton has not been able to improve efficiency after about 30 years of protection. Accounting data (see Table 5.14) shows that relatively high fixed costs and high salaries caused by overstaff problem might have contributed to low profitability of Proton. Misallocation of fixed and variable resources in Proton is highly likely to be the reason for diseconomies of scales. Statistics suggest that Proton is unable to or has no incentive to achieve efficiency, to acquire technology despite the prolonged protection from foreign car makers. Lack of evidence of growth is mainly the results of government’s disincentive to lift protection.

Although Proton experience output growth in the 1980s and early 1990, its growth does not persist. Proton’s initial growth was due to penetration into low and low-middle income segments in which foreign car makers find unprofitable to tap into. Proton’s output began to take a turn after the Asian Financial crisis in 1998 and Proton’s sales statistics have not shown significant improvement since. Proton’s low output growth rates implies the followings:

(a) the low and low-middle income segments of the car market may be saturated and income of these segments have not increased as fast as the higher income segment;

210

(b) entrance of car makers from the emerging economies like China and India, that make smaller and affordable cars is eroding Proton’s niche market away; and

(c) Proton has not been able to compete in the market segments where car buyers have stronger buying power because Proton cars are not significantly differentiated.

Further investigation is required before conclusive remarks can be make about the effect of Proton’s specialization in small and affordable cars. However, it may be concluded that Proton’s inability to compete with foreign cars within domestic market as well as international market suggests that there is insufficient technology diffusion taken place over the years. As a result, as income level increases, car buyers substitute Proton cars with foreign cars that possess different sets of characteristics possess higher market value.

6.3 Costs of Protectionism

Panagariya (2002, p. 2) quotes that Krugman (1990) states the cost of protectionism is as low as 0.25 per cent of United States’ GNP although there are major protectionist measures used in the car, steel and textile industries. A summary of estimated protection costs in a small economy and a few large economies during the period in 1960s to early 1970s shows that the cost is relatively larger in Chile, a small economy than in larger economies such as in the United States, Germany and a few industrial countries (Panagariya 2002, p. 5). Studies show that the cost of protectionism in percentage of GDP, is less than 2.5 per cent of Chilean economy but less than 1 per cent of the United States and a few industrial countries.

Cost of protectionism in the Malaysian car industry is about 6.7 per cent of real GDP in 2014. It is more than five times the relative size of the cost estimated in major protected industries in the United States. The cost in Malaysia is far greater than the cost estimated in the United States mainly because the tariff rates are as high as 145 per cent imposed on foreign cars while tariff equivalent estimated range from 4 per cent to 85 per cent across major protected industries in the United States. There are two possible reasons for the observed pattern of protection cost size across large and small countries. They are: 211

(a) more diverse economic activities in the developed countries. Hence, their GDP is not heavily dependent on a small number of industries, and

(b) relatively much larger real GDP in developed countries than smaller developing nations.

A few studies suggest that the low estimated cost is due to omission of spillover costs. These costs are allocative inefficiency, disappearance of products, unproductive profit- seeking activities, and X-inefficiency. Each of these costs are discussed accordingly from Malaysia’s perspective as follows:

6.3.1 Allocative inefficiency

DEA is developed to measure relative efficiency of decision-making units that is, the selection of decision-making units influences the degree of inefficiency of specific decision-making units a study concerns. Efficiency index in this research does not show the productivity trend of car makers selected but, provides a snapshot of their efficiency. Although the index is in relative terms, it provides a reliable measure for comparison when a set of decision-making units chosen as benchmarks are proven highly efficient. In this research there is no attempt to decompose allocative efficiency from other forms of efficiency. Inefficiency reflected by the gap between Proton’s and other two selected car makers’ efficiency index is analyzed at aggregate level. That is, the gap is contributed by physical efficiency of input-output production transformation and the economic efficiency of optimal factor allocation as defined in Kopp (1981, p. 479)

Efficient index constructed shows that protected decision-making unit, Proton is highly inefficient relative to Perodua, and an experienced industry player Toyota of which the mother plant located in Japan has been proven efficient because of the strategies adopted.

Competition is often cited to be a factor that contributes to efficiency. It is found that efficiency is low in developing countries because of government intervention that gives rise to monopoly power, excessive and unproductive red tapes, and trade barriers

212

(Lagarde 2017). Other factors are inequality in income distribution and wealth distribution that impede equal opportunities to education and subsequently to employment (Isaksson, Ng & Robyn 2005, p. 38).

Although productivity slowdown in developed countries may affect productivity in Malaysia, it is highly likely that inefficiency in Proton arises in the microenvironment. Although technology advancement, joint-venture projects and freer trade have made knowledge available, Proton’s efficiency index is 0.16 compared to Perodua’s efficiency index of 0.55 based on 2015-2016’s accounting data.

Since the objectives of protecting Proton have not been published, discussion in this subsection review some commonly cited justifications for protectionism. These justifications are as follows:

(a) To close the technology gap between advanced foreign competitors and domestic producer (Ederington & McCalman 2011, p. 37). Study at countries level have shown that capital and labour productivity in Malaysia are relatively low when compared to developed nations (Shen 1984, p. 102). This is supported by a review of low productivity trends in developing countries in Isaksson, Ng & Robyn (2005).

Logistic regression analysis show that Proton has not differentiated their cars but rely heavily on pricing strategy to maintain its position as a large player in Malaysian car market. In contrast, Perodua differentiates themselves as the second domestic car maker through price and other car characteristics such as size of car and horsepower, reflecting a relatively small technology gap between Perodua and an experienced industry player such as Toyota. The technology gap is however, large between Proton and Toyota as Proton cars are not differentiated by their characteristics.

It is therefore, concluded that protection has not acquire new technology and hence, unable to close technology gap between Proton and foreign competitors.

213

In Malaysia’s case, the results support Sauré (2007, p. 115)’s proposition that protectionism may hinder growth.

(b) “Narrow moving band” (Krugman 1990, p. 107) argues that protectionism allows creation and preservation of specialization pattern. Hence, it is justifiable if Proton is able to reap dynamic economies of scale and competes in the domestic economy when protectionist tools are lifted after a designated period. However, this argument does not hold in Malaysia’s case because Malaysian government is lack the commitment to lift the protectionist measures and Proton’s declining market shares over the years since the Asian Financial crisis 1998 shows that Proton is unable to compete in the domestic market. This implies that Proton has not achieved the technical component of efficiency during more than 35 years of protection period. The objective to create a specialization pattern in domestic car industry is futile.

(c) Trade policy. Malaysia’s trade policy (Ministry of International Trade and Industry 2019) focuses on regional trade and gears towards more liberalized and fair trade. Specialization of foreign car makers’ plants within the region may leads to allocative efficiency at plants level. This may have contributed to Toyota’s relatively high efficiency index. Despite the access to markets in the region, Proton’s export to foreign markets in the region has not been significant.

Table 5.19 shows that Toyota’s efficiency index improves when there is 30 per cent tariff cut. The results suggest that without trade barriers, experienced foreign car makers operating in Malaysia may achieves allocative efficiency and contributes to trade and GDP growth of the manufacturing sector. Thus, the argument for protectionism is not justifiable in this context.

(d) Employment. Statistics show that Proton has the largest number of employees and has the lowest sales to employee ratio, reflecting the lowest labour productivity among the three selected car makers. This objective is achieved on the expense of efficiency and high cost borne by the society in the form of taxes paid for private car ownership and job destroyed in the public transport industry, mainly public bus and school bus operators. 214

Domestic car maker Proton is unable reap economies of scale and unable to achieve allocative efficiency because of the followings:

(a) low labour productivity that can be an effect of moral hazard problems associated with government’s implicit safety net and privileges given to the natives, and

(b) under provision of education and government’s deliberate action to manipulate the education system for the purpose to increase the natives’ enrolment rates in tertiary education. This has repercussion effects on the domestic labour market of which the most commonly cited problem in the labour market is the mismatch of labour and skill required by employers.

6.3.2 Disappearance of products

This argument against trade barriers was contributed by Dupuit (1969) Subsequently Romer (1994) reviews the argument and suggests that in the presence of economies of scale, low level of protection may leads to high protectionism costs. The costs in the context is the disappearance of products of which the producers find it unprofitable to produce because of high fixed cost.

In the case of car industry, production is capital-intensive and technology is readily available, fixed cost may not be high because different stages of product life-cycle in foreign countries and in the domestic country. As technology in Malaysia is lagging behind the stage of technology in the exporting economies such as Japan and the United States, it is reasonable to deduce that capital becomes relatively cheaper in Malaysia than in Japan when the technology becomes available in Malaysia. Since foreign producers have set up the plants about two decades before the founding of Proton, transportation cost has greatly reduced. Therefore, tariff in this case, need not cause disappearance of products.

The cost of product disappearance has become a lot less direct when the demand for car is not merely for the service derived from the cars but, derived from a set of car

215 characteristics. For example, in Malaysia, car accessories such as navigation system, blind spot monitor, panoramic view monitor are available for a Toyota Fortuner of 2800cc that is priced RM182,000 or approximately A$58,000 at the exchange rate of 3.15. Car buyers may choose to install those accessories at a higher car price. In Australia, a Toyota Fortuner of similar horsepower is priced A$45,000, is equipped with all the accessories not available in Malaysia’s version. In addition to the differences in accessories, there is also variant in the horsepower in Malaysia. The horsepower available in Malaysia is either 2400cc or 2800cc while the horsepower of a Fortuner is 2800cc in Australia. Foreign car makers that entered Malaysian car industry before the founding of domestic car have the advantage of altering the variants of their cars and percentage of local content to compete with other imported foreign car makes.

Foreign car makers operating in Malaysia adopt innovative strategies to create more variants in Malaysia to increase the variety of choices within a model. This allows the car makers to price their cars differently to compete with each other while maximizing their sales. While trade barriers in Malaysia’s case may cause disappearance of some imported products the former also cause creation of some products with minor differences. The products disappeared may be certain car makes or car makes available in domestic market but, disappearance of certain models because of the high price attached to the latest technology.

While trade barriers protect domestic cars, the former also indirectly protect the foreign car makers that successfully entered the industry before other foreign car makers do. Although history allows foreign car makers operating in Malaysia to take advantage of being “inside” the industry that is, low transport cost and low risk associated with production in domestic economy, the cost of protection in the form of high tariffs may still erodes the gains.

6.3.3 Unproductive profit-seeking activities

There are two main purposes of introducing import licenses in Malaysia. They are to restrict imports of foreign cars as well as to redistribute wealth to the natives, mainly the

216

Malays who are politically connected to the government. While the cost associated with import licenses may be reflected by the higher price of foreign cars, the cost associated with unproductive profit-seeking activities is not easily traced and measured.

In Panagariya, (2002, p. 24), Bhagwati (1989) suggests that measurement of protectionism cost is to include the downstream directly unproductive profit-seeking activities because activities at the downstream are the effects of unproductive profit- seeking behaviour while upstream unproductive profit-seeking activities are the causes of such activities. Bhagwati (1989) also suggests that cost of downstream unproductive profit-seeking activities may be smaller than the total rent from quota because in the presence of distortion, the shadow prices of resources are likely to be lower than the market prices.

Although it is not formally documented in literature, casual conversation with industry players reveals that Malaysian native license holders often obtain the license through relationship instead of through competitive bids. Consequently, the cost of obtaining the license may be much smaller than the benefit of re-selling the license. The significant gap between the cost of obtaining a license and the anticipated short term benefit of selling the license subsequently turns import licenses into commodities traded in a non-existence market where aggressive non-native buyers are willing to pay higher prices than the actual prices of the licenses. In this case, measurement of protectionism cost is made complicated when the licenses are traded as goods for immediate gains.

This research is unable to capture the cost arising from directly unproductive profit- seeking activities mainly due to lack of transparency. Both the total number of licenses and the license prices are not publicly available. The part of the cost associated with import licenses accounted for in this research is the cost borne by car buyers in the form of higher imported car prices.

Additional cost associated with unproductive profit-seeking activities is the cost of obsolete technology. Due to tariffs and non-tariff barriers, imported new cars are excessively expensive. As a result, the market for imported new car is relatively very small. This gives rise to market for used and re-conditioned foreign cars. Re-conditioned

217 foreign cars are imported at relatively low price to offset the cost of license so that imported cars are more affordable to a large segment of the car buyers. As a result, car buyers who demand foreign cars may pay high price for obsolete technology.

6.3.4 X-efficiency

X-efficiency is a term Leibenstein (1966) uses for efficiency specific to firms and industry instead of an economy as a whole. Two sources of X-inefficiency identified are:

(a) protected domestic firms that cannot compete with imports, and

(b) protected domestic firms that can compete with imports at non-competitive cost.

The cost associated with X-efficiency is therefore, aggregate of avoidable high cost and monopoly return. Bergsman (1974, p. 410) cites Leibenstein’s suggestion that cost associated with X-inefficiency may be 50 to 100 per cent larger than cost of allocative inefficiency associated with misallocation of resources. This implies that measurement of cost at firms’ level allows for more specific sources of cost to be identified and be accounted for.

Bergsman (1974, p. 421) shows that in Malaya (Malaysia prior to the joining of East Malaysia to the 11 states in Peninsula Malaysia), the cost of protection associated with X-inefficiency and monopoly returns is 0.8 per cent of Malaya’s GNP. In Malaya’s case desirable monopoly returns gained by the protected firms are less than the cost of allocative inefficiency. In other words, the benefit of protecting the industry is less than the cost that arises because of inefficiency. The result suggests that in a small and open economy, the cost of protection is overwhelming because of misallocation of resources.

Despite protection, Proton constantly makes loss since the late 1990s. Although the car industry is oligopolistic, non-tariff privileges such as guaranteed loan approval for civil servants and government’s acquisition of Proton cars have not been able to turn Proton’s

218 returns to positive. Statistics in Table 5.14 show that Proton is operating below shut- down point, implying that the source of X-inefficiency cost is incompetency of Proton. Foreign car makers operating in Malaysia achieve efficiency through specialization of their plants located in different economies within the region. For example, the largest foreign car maker in Malaysia, Toyota has an assembly plant in specializes in Camry, Vios, Fortuner, Innova, Hiace and Hilux. It is able to export up to 200 units of light truck to Thailand (Khoo 2014). Toyota in Malaysia imports other models of Toyota cars from their plants in Thailand, Indonesia and Japan.

6.3.5 Other costs borne by car buyers and society as a whole

Face-to-face interviews of a few industry players and a number of Proton car owners revealed exorbitant Proton parts prices makes ownership of Proton car expensive to the low and low-middle income groups. Car owners further suggest that large proportion of their income is spent on maintenance if owners chose to maintain their Proton cars. Due to expensive parts coupled with lax in road safety regulations, many cars particularly in the rural areas are not maintained and not insured. Hence, unfit cars impose social cost to road users.

In addition to the possible dissipation of trade benefits, excessive protectionism in Malaysia’s case also leads to the followings:

(a) Lack of incentive for protected industry to acquire knowledge and growth. Cost associated with lack of incentive to growth is the cost of low quality domestic cars that lower income group acquire at high price. This cost is very likely to be high relative to aggregate income of the low income segment because private transport owner takes up a large proportion of the household income and have negative impact on the income spent on other goods;

(b) High Proton car part prices relative to durability of the parts. Due to expensive car parts, private transport ownership is more costly than uses of public transport. However, the costs of car parts are often not accounted for in low income group’s

219

decision making process due to lack of information related to durability of the cars and the parts;

(c) Misallocation of public funds. Large amount of public funds had been used for bailing out of Proton despite opposition of the general public. Additional cost associated with misallocation of public funds is the cost associated with moral hazard problem; and

(d) Cost of excess capacity imposed on public transport providers because of low demand for public transport following artificially lower price of domestic cars and subsidised oil price.

Although most of these costs are direct and quantifiable, estimation of these costs involves different approaches, models and sets of data. Hence, estimation of these costs are not feasible in this research.

6.4 Elasticities and Car Characteristics

Comparisons of price elasticity of demand for cars show that demand for domestic car is more elastic in Malaysia than most of the studies carried out in the United States using the Econometric tools. However, demand for both domestic cars and foreign cars are highly elastic in Berry, Levinsohn & Pakes (2004). At this stage, it cannot be concluded that estimation of price elasticity of demand without incorporation of car characteristics generate lower estimation of price elastic of demand than the estimation generated if Discrete Choice approach is taken.

Comparisons of elasticities of demand across studies are summarized in Table 6.1 below. Table 6.1 shows that income elasticity of demand is far lower in the studies carried out in the 1950s and 1960s. Comparisons also show that generally, price elasticity of demand for domestic cars and for foreign cars are higher in the more recent studies.

220

Table 6.1 Summary of elasticity findings in a priori studies Price Literature Income Domestic Foreign Cross* Bandeen (1957) 0.9 Bennett (1957) 1.4 to 1.7 Suits (1958) 3.8 to 4.5 -0.4 to -0.6 Turnovsky (1966) Below 5 -0.6 to -9.8 Hickok (1985) -0.2 Tarr & Morkre (1984) -1.1 -3.4 Levinsohn (1988) -0.8 to -2.0 Hufbauer & Elliot -1.2 -1.5 (1994) Berry, Levinsohn & -3.9 to - -3.1 to - Pakes (2004) 24.3 27.5 This research Domestic cars: 28.7 -1.7 to -4.8 -3.1 to -8.2 0.36 Foreign cars: 20.3

Notes: * Cross price elasticity is based on the demand for Honda cars and the price of the second choice.

Comparison of results between Berry, Levinsohn & Pakes (2004) and this research shows that the range of price elasticity of demand is very large for both domestic and foreign cars in the United States. The range is however, very small in Malaysia. These results suggest that in both domestic and foreign cars in the United States are highly differentiated and high values are attached to luxury cars of each category. On the other hand, both domestic and foreign cars are relatively less differentiated in Malaysia than in the United States. This implies that when prices are very high relatively to per capita income, the major factor that influence car demand will be the price. From the car makers’ perspective, there is a need to balance car differentiation with production cost in order to produce cars affordable to their target markets while offering unique car characteristics to capture new buyers and to retain repeating buyers.

221

Estimation in Table 5.8 and Table 5.12 suggest that demand for Proton cars, the cheaper “national car” is mainly influenced by their own prices and unobservable car characteristics. This implies that the lower income segment demands private transport for its basic service and there are some unique features attached to Proton which may not be specific characteristics of the car but, features attached to the brand such as guaranteed loan approval or privileges of government loans. Demand for Proton is created through manipulation of price signals and creation of features not on Proton cars but as a package attached to the cars. Government’s policy to achieve political objectives is implemented on the expense of the low income segment.

Comparisons also show that the demand for imported cars is consistently more elastic than the demand for domestic cars. This implies that car buyers tend to be more sensitive to price changes of imported cars than in domestic cars. The difference may be due to the following reasons:

(a) Imported cars are relatively more expensive than domestic cars due to transportation cost and tariffs. Hence, acquisition of foreign cars takes up a larger proportion of household income than acquisition of domestic cars, and

(b) Presence of unobservable car characteristics unique to imported cars.

Within the category of foreign cars, estimation of interaction terms in Table 5.10 suggests that Nissan do not differentiate their best-selling model that is the cheapest within the segment. On the other hand, Toyota differentiate their cars mainly the best-selling model through car characteristics. Literature in marketing suggest that product differentiation has been long recognized as a significant factor contributes to a firm’s success (Phongpetra & Johri 2011; Train & Winston 2007; Dickson & Ginter 1987; Smith 1956). The rationale underlying this argument is that as the national income of the economy increases, demand for car is no longer the derived demand for basic transport but, utility derived from car characteristics. Test results support literature in Marketing in that the most competitive car makes of domestic and foreign categories present the most number of car characteristics for car buyers’ to consider during the process of making their buying decision.

222

6.5 Other Welfare Effects of Trade Barriers

Malaysia records high ratio of car to person at 1:3 compared to Australia that records 1:1.5 while Malaysia and Australia’s real GDP per capita are USD7,278 and USD42,404 respectively (World Bank data, accessed 25/10/2012). At aggregate level, through the establishments of Proton and subsequently Perodua, the National Automotive Policy has contributed to high car ratio to person in Malaysia. The increases in the numbers of car on the roads over the years contribute to traffic congestion and pollution.

The environmental problems worsen when trade barriers coupled by the lax of road safety regulations lead to the increasing numbers of old and inefficient cars on the roads over the years.

Other welfare costs are the cost of traffic congestion when old cars broke down on the roads, and the default risk that banks have to take when loans are extended to domestic car buyers. Although interest rates are about 2 per cent higher on loans for domestic car buyers than for foreign cars, there are still implicit costs borne by banks. These costs are monitoring cost, impounding cost and transaction costs when car buyers default their payments. The default risk is relatively high in the domestic car segment because of low re-sale value of the cars, short life expectancy of the cars, and high maintenance cost. Therefore, the default risk is likely to remain high if quality of domestic cars is not improved.

Government’s implicit protectionist measures such as guaranteed loan approval, other promotional strategies, and controlled petrol prices distort the true cost of private transport ownership. Other cost of private car ownership is the expensive but low quality car parts prices. Private transport becomes very expensive when government reduced subsidies on oil. Due to the low quality of Proton cars in particular, their short life expectancy and high maintenance cost, the cost borne by this segment of car buyers is enormous by their income level.

The 2010 J.D Power Asia Pacific1 reports that in the segment, Perodua Myvi achieved the highest ranking for 4 consecutive years with its score of 157 PP100, followed

223 by that score 178 PP100. Proton Savvy scored 199 PP100. In the MPV/van segment, Toyota Avanza assembled in Malaysia and Proton Exora score 85 PP100 and 143 PP100 respectively. Report suggests Perodua has outperformed Proton because Perodua cars have better quality than Proton cars in compact car segment and Toyota outperformed Proton in the MPV/van segment.

In 2011, J.D. Power Asia Pacific reports Proton Saga scores 124 PP100 compared to 62 PP100 for Honda City and 64 PP100 for Toyota Avanza (mini MPV) that are assembled in Malaysia. In the midsize car segment, Nissan Sylphy, Toyota Altis and Honda Civic score 39 PP100, 43 PP100 and 55 PP100 respectively. These results show that foreign car makes assembled in Malaysia possess higher quality than Proton cars.

In 2016, Proton reports that their cars meet the ASEAN NCAP (New Car Assessment Program) standard. The NCAP results are misleading because the models assessment were the more expensive models such as Persona and Iriz while the best-selling model is the cheapest model Proton Saga.

Comparisons of reports from different sources show the followings:

(a) Proton, a fully domestic car, heavily protected, has not been able to compete with Perodua that is partially Japanese owned and not as heavily protected as Proton;

(b) Proton has not been able to compete with foreign cars assembled in Malaysia in terms of their quality, in different segment;

(c) Higher range models of foreign cars possess higher quality standard than their lower range models. This may be due to the lower percentage of local content in the higher range models than the lower range models;

(d) It follows that the locally made car parts are likely to have lower quality than the parts made in car makers’ home countries; and

224

(e) Large gap in quality standard between Proton and foreign car makers enable foreign car makers to alter their product characteristics, hence reducing production cost in order to compete in the presence of trade barriers.

6.6 The Benefits of Protectionism

An argument for founding and protection of domestic car is employment. Table 5.15 shows that both Proton and Perodua record low ratio of sales to employees based on the 2015 accounting data compared to foreign car maker, Toyota. Ratios suggest that there is over staff problem that leads to inefficiency in both Proton and Perodua. Consequently, the argument for protectionism on the ground of employment is not justifiable.

The argument that car industry contribute to GDP growth is based on the assumption that the industry’s linkages to other industries are strong. However, there have been lack of study to support this argument. If it is indeed true that multiplier size of car industry is large and growth in the car industry may leads to significant GDP growth, then freer trade will allows the multiplier to work to its maximum.

Ratio analysis (Table 5.15) and Data envelopment analysis (Table 5.16) suggest that the protected domestic car maker Proton, is both unproductive and inefficient. Evidence also suggests that foreign car maker has comparative advantage in producing cars. Therefore, trade barriers removal or reduction will lead to faster GDP growth.

Although at product level, substitution effect is not detected in most of the selected car makes except for Honda in logistic regression model (see Table 5.8), there may be substitution between domestic cars and foreign cars. Figure 2.4 shows that when Proton and Perodua are aggregated and categorised as “domestic”, the number of new domestic cars registered is a mirror image of the number of new foreign cars registered from year 2001 to 2017. The graph also shows that foreign cars outperform domestic cars since 2014. These observations suggest the followings:

(a) excessive trade barriers are not effective anymore;

225

(b) car buyers are willing to pay up to 3 times the price of similar foreign car models in overseas for their characteristics; and

(c) as income level increases, coupled with effective pricing strategies, foreign car makers are gaining ground in the competitive market.

It is therefore concluded that the existing foreign car makers in Malaysia may contribute significantly to GDP growth in the absence of domestic cars and absence of trade barriers. However, more substantial study is required to make a definite conclusion on car industry’s contribution to GDP growth.

6.7 Tariffs and Deadweight Loss

Although tariffs are introduced to protect Proton specifically and to lesser extent, to protect Perodua, domestic car makers are also subject to about 85 per cent tariff rate. In comparison, tariffs imposed on foreign cars is at high as 145 per cent. The tariff rates imposed on Malaysian car industry are 34 times to 58 times greater than that in the United States in which tariff and tariff equivalent voluntary restraint arrangement in car industry are 2.5 per cent and 4.1 per cent respectively (Hufbauer & Elliot 1994, p.100).

Due to excessive tariffs imposed on both domestic and foreign cars, the social cost of protectionism is extremely high at 16.7 per cent of real GDP and more than half the size of manufacturing sector’s GDP contribution. Such massive amount of tariffs imposed has the following impact on Malaysian economy:

(a) Redistribution of income from the lower income segment to the a small number of Malay elites who are politically connected to the government, who have interest in both Proton and Perodua or owns domestic car parts manufacturing and distributing firms, and to employees of Proton;

(b) Redistribution income from tax payers to employees and shareholders of Proton through taxes and huge amount of public funds injection to bailout Proton;

226

(c) Redistribution of wealth from tax payers to import license holders who are mostly government’s cronies;

(d) Misappropriation of public funds leads to under provision of public goods; and

(e) Misallocation of resources that leads to excess capacity in manufacturing plants and in labour.

Tariffs are significant sources of government revenue if the numbers of new cars registered over the years increase as income increases. However, Laffer curve suggests that if tariffs are set too high, beyond an optimum tariff rate, tariffs collection will fall, causing government revenue to fall. Laffer curve’s implications on car industry in Malaysia are:

(a) Government revenue from tariffs on cars may be higher if the rates were lower. This is because high price elasticity of demand implies that demand increases faster than the price falls if tariffs are reduced or removed;

(b) There may be disincentives for foreign car makers to bring new technology in to Malaysia, causing lack of technology diffusion because high tariffs rates erode profitability of the foreign car makers;

(c) High cost of administration in the form of transaction cost and monitoring cost, is borne by the society; and

(d) In Malaysia’s case, large tariff collection is offset by the large cost incurred to keep employment and privileges to the natives. Consequently, the benefits of tariff collection on research and development for the car industry is not reap.

6.8 Conclusion

Trade barriers erected to protect domestic car maker Proton is proven to be extremely costly to the economy. The estimated cost of protectionism associated with tariffs alone 227 is as large as 70 per cent of manufacturing sector’s GDP contribution. The cost is expected to be larger if other costs are identified, measured and taken into account. The costs not accounted for in this research are as follows:

(a) The cost of lower product quality in the form of high price paid by car buyers and in the form of accidents caused by defects in cars;

(b) Environmental cost;

(c) Cost of unproductive profit-seeking activities;

(d) Economies of scales reap by the foreign car makers operating in Malaysia;

(e) Misallocation of capital and labour; and

(f) Under-provision of public goods.

The cost identified in this research are not identified in studies carried out in other countries because of the differences in the stage of economic development in Malaysia and in other countries. These costs are not accounted for in this research because of missing data and different methodology required.

Notes

1. J.D. Power, an organization that surveys customer satisfaction and product quality, and ranks new cars based on the number of defects or problems identified during the first 2 to 6 months of ownership. The measures include more than 200 problem symptoms covering the exterior and interior of the cars, driving experience, car characteristics and engine. The problems are summarized and measured as number of problems reported per 100 cars (PP100). Therefore, a lower score is preferred to a higher score.

228

Chapter 7 Conclusion

7.1 Introduction

This research applies partial equilibrium framework to study the car industry of a small economy. Although general equilibrium framework is able to account for spillover effect on other related industry, partial equilibrium framework is chosen because of imperfect substitution of Proton with all other range of cars available in Malaysia and because car industry contributes to a very small proportion of GDP in Malaysia.

This research essentially answers the questions following observation of the protected domestic car industry. These questions are as follows:

(a) Do car characteristics that differentiate various car makes have significant influence on car demand in a developing economy such that car buyers are willing to buy foreign cars despite high prices?

(b) Has technology diffusion taken place in the infant car industry during the period of protection?

(c) What is the cost of protectionism, as percentage of manufacturing sector’s contribution to real GDP, borne by the domestic car and foreign car owners when car features are accounted for?

(d) Does competition in the industry lead to greater price elasticity of demand?

This research is different from most a priori studies in cost of protectionism as Discrete Choice analysis is adopted to specify the demand function of cars in Malaysia. Using this approach, car characteristics are accounted for their influence over car demand in addition to the socio-economic factors that have often been cited in conventional econometrics models. Through the estimates of Logistic Regression models, price elasticity of demand is estimated for each of the major car makes in the categories of lower, middle, and higher range cars.

229

Primary data is collected across all states in Malaysia using self-administrated survey and online survey. As data collection is based on random sampling and voluntariness of car owners at service centres, the sample size distribution reflects the friendliness of car owners across different states of the country instead of reflecting proportion of new car registered.

Results of Logistic Regression Analysis generally suggest that social-economic factors as well as car characteristics influence demand for cars. However, statistics suggest car characteristics’ contribution to demand for Proton is insignificant.

Comparisons of price elasticity of demand for car with the statistics estimated in studies carried out in developed and freer economies show that price elasticity of demand is greater in developed and freer economies than in Malaysia. The statistics support the argument that competition leads to greater price elasticity of demand.

Due to lack of true and fair data, marginal cost and supply curve of Proton are not observable. Data Envelopment analysis is therefore, adopted to complement the investigation on cost of protectionism through investigation on productivity and excess capacity. However, cost of excess capacity is not estimated.

The cost of protectionism measured by deadweight loss as percentage of Malaysia’s GDP is very large. Although a series of study over a period of time provides a fuller and fairer picture of technology changes and their effect on cost of protectionism, this study provides a snapshot of total cost imposed on the society at a given state of technology due to data constraint. In addition, due to lack of transparency and complexity of the trade barriers introduced, it is difficult to separate the effect of tariff from non-tariff barriers. Therefore, welfare effects of the tariff and non-tariff barriers are embodied in prices of cars. The cost estimated in this study is the combined effects of tariffs and non-tariffs barriers that reinforce each other in their influence over the car industry.

230

7.2 Welfare Effects of Trade Barriers

It is difficult to separate the effects of tariff barriers from non-tariff barriers on car industry in Malaysia because there are non-tariff barriers that are not directly measurable but have impact on prices. Cost of protectionism in this research is relatively very large when compared to that of larger economies that is, at about 16.7 per cent of Malaysia GDP compared to a priori studies that are often found less than 1 per cent of the United States’ GDP. This implies that the negative effects of trade barriers are significantly larger in Malaysia than those of larger economies.

The overall measurable component of welfare effects is captured by the deadweight loss estimated based on price elasticity of demand for the selected car makes. Deadweight loss is the cost borne by the society as a whole in the form of tariffs paid by car buyers but not benefitting any party in the society. As tariffs imposed are based on the categorisation of cars namely, domestic or foreign, and the engine size, the cost of protectionism will increase as the demand for foreign cars increases. It follows that the cost of protectionism increases as the income level of the economy increases.

Other welfare effects either not captured by partial equilibrium framework or not easily traced because of market failure are as follows:

(a) Allocative inefficiency. One of the domestic car makers Proton is unable to achieve allocative efficiency because the lack of technology diffusion in a highly protected industry. Although Proton captures the second largest market share is highly protected, does not operating at supernormal profit level but, operate below the shut-down point in certain years. Its 15.3 per cent market share in 2015 for example is achieved at the expense of economies of scale that can be reap by more efficient foreign car makers that have been operating in Malaysia. In addition, Proton’s market share is also obtained on the expense of the public transport such as bus operators. Although the second domestic car maker Perodua is also protected to a lesser extent, Perodua competes with and outperforms Proton. Perodua’s performance reflects comparative advantage of foreign car makers because Perodua is partially Japan-owned;

231

(b) X-inefficiency. Protected infant industry is unable to compete with foreign car makers operating in Malaysia and the foreign car exporters. This is highly likely to be associated with complacency resulting from government’s lack of will to lift the protectionist measures. Consequently, there is lack of technology diffusion which in turns, leads to lack of competitiveness of the domestic cars despite their low price;

(c) Economies of scale. Without trade barriers, foreign car makers operating in Malaysia are able to reap economies of scale due to comparative advantage and specialization of their plants within Southeast Asian region. Therefore, the exports of foreign cars assembled in Malaysia is highly likely to achieve success as the point in case of Thailand. Public transports such as public buses, operators are unable to reap economies of scale because of the falls in the demand for public transport. Consequently, lack of profitability leads to less re-investment in the sector. This further reduces the demand for public transports when buses are not upgraded to accommodate public’s expectation of service quality;

(d) Unproductive profit-seeking activities. The cost associated with unproductive profit-seeking activities are relatively high prices of car parts, indirect transaction cost, and cost associated with trading of import licences in the black market;

(e) Misallocation of resources. Excess capacity and over-staffing in Proton and Perodua impose high cost of resource misallocation on the society. Job creation is often cited as the reason for the founding of Proton hence, the reason for not shutting down Proton despite making subnormal profit made over the years. The reason for not closing down Proton is not justifiable. Labour employed are proven unproductive and highly likely to be the main reason for lack of technology diffusion. Excess capital and labour in the industry could have been allocated to other industries where the marginal returns are higher. The cost associated with misallocation of resources is not only confined to the cost of hiring unproductive labour and the cost of maintaining plants but, the higher production cost in other industries in which the cost could have been lower if unproductive labour and capital are channelled to these industries;

232

(f) Redistribution of income and wealth. Income is redistributed from tax payers to employees and shareholders of Proton in the form of financial assistance while Proton is making losses. Wealth is also redistributed from general public to shareholders of Proton in the form of benefits for holding the assets. Automotive Development Fund founded in 2006 aims to improve the competitiveness of the car industry by channelling funds into Proton to keep the firm solvent. Following RM400 million and RM200 million allocated in 2006 and 2009 respectively, information is unavailable to the public in regard to the use of funds and if there is any allocation of the funds in the other years. There was no further explanation from the government on how the objectives of the fund can be achieved and how the funds are used. Funds have also been given to assist Proton’s distributors but, no further information is available regarding the amount and how the funds are used. Other source of funds includes Malaysian Industrial Development Finance Berhad (MIDF). Through MIDF, funds continued to be injected into selected firms in the industry;

(g) Other social costs. There are cost associated with the lower income group who buys Proton because of the seemingly cheap private transport. Cost of maintaining private transport is hidden because of unexpected short life expectancy of the cars as well as their parts. The cost of private transport ownership increases further when there is unexpected reduction in subsidies on petrol reducing the proportion of income spent on other goods. Consequently, unexpected high private transport cost borne by the lower income group is not to be taken literally in monetary terms but, in relative terms. Other social costs not directly measurable are the cost associated with increasing numbers of unfit cars on the road that cause traffic congestion when broke down, and road accidents occurred due to defects of new cars or unfit cars;

(h) Environmental cost. Traffic congestion is posing a serious problem in many cities in Malaysia and poses travel cost to the society in terms of longer travel time. This problem leads to greater demand for flyovers and highways, and subsequently cost of maintaining these public goods financed either by taxes or by tolls. Air pollution on the other hand, contributes to health problems associated with respiratory system. Statistics show that major cause of death in Malaysia is 233

Ischaemic heart diseases followed by pneumonia in 2016 (Department of Statistics Malaysia, 2017). Although studies are required to confirm the significance of the relationship between air pollution and the principal cause of death, statistics nevertheless, show the need to abate air pollution that is associated with high ratio of car to population in Malaysia; and

(i) Under provision of public goods. Misallocation of funds and subsidies on oil may lead to under provision of public goods. Subsidies on oil distort the actual cost of transport hence, lead to high demand for oil that in turns, leads to higher amount of subsidies on oil. Consequently, less funds available for public goods such as healthcare, education and infrastructure in the rural areas.

7.3 Strengths and Weaknesses of Research

The major strength of this research is the application of discrete choice model to incorporate car characteristics in specification of car demand function. There are studies suggesting that the demand for product is influenced by product characteristics when product is differentiated. However, studies begin to analyse demand for differentiated product when knowledge improved and statistical tools are available.

Another strength of the research comes from the strength of logistic regression. That is, logistic regression allows use of micro data at product level and market level for more insightful analysis of highly differentiated product. Logistic regression allows collinearity among explanatory variables. Therefore, analysis provides more realistic picture of how factors or variables interact with each other in their influence on the demand for cars. Results of logistic regression show the significance of different car characteristic set on the demand for different car makes.

The results also show that there are unobservable car characteristics that influence different selected car makes, implying that further research can be carried out to identify the unobserved characteristics. Identification of these characteristics may be useful for marketing strategies. However, more in-depth study of Malaysian car industry is required

234 to investigate and to understand the industry and finally, to identify the unobservable car characteristics.

Studies also suggest that logistic regression is not influenced by outliers like linear regression analysis because logistic regression analysis reports probability of outcome. As such, the estimators generated are not influenced by values of outliers. This feature is important for this research because of cultural differences observed in Malaysia and other Asian countries. For example, the accumulation of wealth by the old generation for younger generation may have effect on the demand for cars but this is not be taken into account in other studies.

The major weakness of this research is the lack of data. There is time lapse between the dates of new car purchases and the time primary data is collected. Although data collection is carried out mainly in car service centres where new cars are serviced and only cars bought within 3 years are considered, there may be difficulties for car buyers to recall information and hence, difficulties to provide accurate information.

Another weakness of this research is derivation of demand logistic function for individual car makes instead of individual models of each car makes. Car makers target different market segments by differentiating their cars through variations made to car characteristics hence, creating values to different groups of car buyers. This research is unable to study the market by analysing the market at a deeper level on individual models of each major car makes. Since there are variations to car characteristics are made to individual models of different car makes, calculation of price elasticity of demand at disaggregate level that is, on micro level, is therefore, more appropriate.

In addition to variants available for each model from individual car make, there are also optional accessories available for car buyers to select. These options generate different numbers of combination and hence, different prices for the same model. This study does not take into consideration differences in accessories within models of each car makes. Therefore, logistic regression result does not show association of different groups of car buyers to different combinations of features that also contribute to price differences. For example, Toyota Fortuner (model) offers 3 variants in Malaysia namely: 2.4L VRZ AT

235 , 2.4L AT diesel engine, and 2.7L SRZ petrol engine. In contrast, the same model offers only 2.8L GX diesel engine and 2.8L GLX diesel engine in Australia. In Malaysia, the difference between 2.4L VRZ AT and 2.4L AT is the different combinations of accessories and different in some features such as paddle shift, rear differential lock, power back door, cruise control, material and reverse camera and other optional accessories. In other words, Malaysian car buyers have the options to install desired accessories at additional costs. In this research, variations at model level is not taken into account. Therefore, this research is unable to investigate car buyer characteristics’ interaction with different bundles of accessories. Substitution of a car with a another car of the same make and same model but different bundles of accessories at different price is not captured by the model. Hence, logistic regression results do not show substitution effect.

This study assumes that the coefficient of variables in the logistic regression models is similar for all car buyers for each car make. It implies that all car buyers of a car make such as Perodua equally value a car characteristic such as size of car but, in reality, some car buyers from the same cohort may place greater emphasis on horsepower more than car size.

The analytical tool used in the research assumes independence from irrelevant alternative (IIA). It is assumed that a change in the attributes of an alternative changes the probabilities of other alternatives proportionately. Consequently, this research lacks the predictive power to forecast the effect of changes in car characteristics of a car makes on the demand for other car makes. However, this may not post significant problem as logistic regression results show that there is lack of substitution across different car makes.

It is assumed that the unobservable characteristics of car and car buyers are independent over time for each car buyers. This is not realistic due to changes in technology changes and social-economic factors over time.

236

While the unobservable variables in the logistic regression model may be quantitative or qualitative variables, the unobservable variables may also imply under-specification of models.

Preliminary graphical analysis (Figure 2.4) and statistical analysis of Discrete Choice model in this research show contradicting results. The former show movements of yearly domestic new car registered are mirror image of the movements of yearly foreign new car registered while results of logistic regression analysis suggest that there is insignificant substitution effect. These contradicting results are due to market level data used in preliminary graphical analysis while product level data is used for statistical analysis.

7.4 Conclusion

This research has contributed to the discipline of Economics in that, product characteristics can be accounted for in the estimation of demand function using Discrete Choice modelling. Subsequently, through this framework, estimation of price elasticity of demand and cost of protectionism are made. This research also highlights that when market fails, Discrete Choice modelling is an alternative to allow more factors to be included for more accuracy of demand function specification.

In this research, it is found that trade barriers erected to protect the Malaysian car industry is extremely high at 16.7 per cent of the country’s GDP and almost as large as the manufacturing sector’s GDP contribution based on 2014 statistics. As the number of new cars registered increases after 2014, the cost of protectionism is anticipated to grow. Since Proton has been protected for nearly 30 years, the cost of protectionism is likely to be enormous. The results of this research also warrant shutting down of Proton to reduce cost imposed on the society. Argument for employment and GDP growth are not justifiable because of low productivity of production factors and excess capacity in Proton.

Various sources of spillover cost associated with trade barriers are identified but, not accounted for. Estimation of these costs requires extensive studies using different sets of

237 primary data and different framework, hence not feasible in this research. Hence, it is anticipated that the actual cost of protectionism is likely to be far greater than 16.7 per cent of Malaysia’s GDP if total spillover cost is taken into consideration. It is therefore, concluded that it is not worth to develop Malaysia’s own car brand like South Korea. The overwhelming cost of protectionism suggests that like Thailand, Malaysia may be better off opening the industry for experienced foreign car makers.

All selected car makes in this research have elastic demand that is, the demand for cars in Malaysia is sensitive to price changes. Demand for the cheapest car make Proton is the least elastic compared to other car makes. This reflects the lack of choices for lower income group or lack of product differentiation in Proton cars. The estimates are based on average of significant car characteristics. Therefore, measure of price sensitiveness is more accurate for the best-selling models of each car makes.

7.5 Future Research

Future research in international economics requires extensive studies on spillover cost that comprises environmental cost and other social costs because of the effect of international trade on domestic economy. Discrete choice analysis is essentially a tool for better analysis of market of highly differentiated products particularly those that are heavily influenced by technology changes.

Since the demand for Proton car is heavily influenced by government’s policies, econometric tools may not be sufficient to capture the qualitative factors both known and unknown currently. In addition to different tool to use, there is a need to re-look at the demand function for domestic products that are protected because of prices distortion. The law of demand although still applies, the relationship between price and quantity demanded may not be easily measured. A research tool that allows distorting factors to be removed may be essential for a better study of buyers’ response to car price changes and hence, estimation of elasticities and consumer welfare changes. In addition, more in- depth studies of new car defects and unfit car’s correlation with road accidents are desirable for more research of Malaysian car industry.

238

Future study may also include switch-over cost of substituting public transport with domestic cars.

239

REFERENCES

Agarwal, MK & Ratchford, BT 1980, ‘Estimating demand functions for product characteristics: The case of automobiles’, Journal of Consumer Research, vol. 7, pp. 249- 262.

Ahmad, A 2018, The key to Proton’s success is exporting, New Strait Times, 9 August 2018, viewed 9 August 2018, .

Ai, C & Norton, EC 2003, ‘Interaction terms in logit and probit model’, Economic Letters, 80, pp. 123-129.

Aitchison, J & Silvey, S 1957, ‘The generalization of probit analysis to the case of multiple responses,’ Biometrika, 44, pp. 131-140.

‘All segment’ 2019, CarBase.my, viewed 1 April 2019, < https://www.carbase.my/all- segment>.

Allcott, H & Wozny, N 2014, ‘Gasoline prices, fuel economy, and the energy paradox, Review of Economics and Statistics, vol. 96, no. 5, pp. 779-795.

Allison, PD 2012, Logistic Regression Using SAS® - Theory and application, 2nd edn, SAS.

Alvarez, F & Lucas, RE Jr, 2007, ‘General equilibrium analysis of the Eaton-Kortum Model of international trade,’ Journal of Monetary Economics, vol. 54, pp. 1726-68.

Amiti, M & Konings, J 2007, ‘Trade liberalization, intermediate inputs, and productivity: Evidence from Indonesia’, The American Economic Review, vol. 97, no. 5, pp. 1611- 1638.

Anderson, JE & van Wincoop, E 2004, ‘Trade costs’, Journal of Economic Literature, vol. XLII (September 2004), pp. 691-751.

Anderson, SP, De Palma, A & Thisse, JF 1989, Demand for differentiated products, discrete choice models, and the characteristics approach, Review of Economic Studies, vol. 56, pp. 21-35.

Antolín, AF, de Lapparent, M & Bierlaire, M 2016, ‘Uncovering substitution patterns in new car sales using a cross nested logit model,’ proceedings of the 16th Swiss Transport Research Conference (STRC) 18-20 May 2016, viewed 22 February 2019, .

ASEANNCAP 2016, ‘5-star ASEAN NCAP rating for ’, viewed 22 August 2018, .

240

Athukorala, PC 2014, Industrialisation through state-MNC partnership: Lessons from the Malaysia’s national car project, Working Paper No. 2014/06, Working Papers in Trade and Development, ANU College of Asia and the Pacific.

Automotive Industry Portal 2018, Marklines Information Platform - Thailand, viewed 14 September 2018, .

Autoworld 2011, ‘MAA car sales statistics for 2011’, viewed 7 May 2012, .

Autoworld 2009, ‘MAA car sales statistics for 2009’, viewed 2 May 2012,

Autoworld 2004, ‘MAA car sales statistics for 2003’, viewed 2 May 2012, .

Azhar, K 2012, “Vendors come under scrutiny,” Malaysia, 23 January 2012, p. 50.

Azhar, S 2006, “Car dealers hit by glut”, The Star, 16/07/2006, p. 1&3.

Baldwin, RE 1969, ‘The case against infant-industry tariff protection,’ Journal of Political Economy, vol. 77, no. 3 (May - Jun 1969), pp. 295-305, viewed 24 March 2014, .

Baharom, H 2013, Cover story: Rocky road for Proton, Focus Malaysia, March 23-29, 2013, pp. 6-8, 10.

Balassa, B 1971, ‘Evaluation of the system of protection’, in Balassa, B & Associates, The structure of Protection in Developing Countries, Baltimore: Johns Hopkins Press.

Bansal, P, Daziano, RA & Achtnicht, M 2018, ‘Comparison of parametric and semiparametric representations of unobserved preference heterogeneity in logit models’, Journal of Choice Modelling, 27, pp. 97-113.

Bandeen, RB (1957). ‘Automobile consumption, 1940-1950’, Econometrica, 25, pp. 239-248.

Ben-Akiva, M, Manski, CF & Sherman, L 1981, ‘A behavioural approach to modelling household motor vehicle ownership and applications to aggregate policy analysis,’ Environment and Planning A, vol. 13, pp. 399-411.

Bennett, WB (1967). “Cross-section studies of consumption of automobiles in the United State”, American Economic Review, 57, pp. 841-50.

Bergsmen, J 1974, ‘Commercial policy, allocative efficiency and ‘X-efficiency’’, Quarterly Journal of Economics, vol . 56, no. 3 (August 1974), pp. 409-433.

Berkovec, J & Rust, J 1985. ‘A nested logit model of automobile holdings for one vehicle households’, Transport Research Part B: Methodological, vol. 19B, no. 4, pp. 275-285. 241

Berkson J 1951, ‘Why I prefer Logits to Probits,’ Biometrics 7, pp. 327-339.

Bernard, AB, Eaton, J, Jensen, BJ, & Kortum, S 2003, ‘Plants and productivity in international trade’, The American Economic Review, vol. 93, no. 4, pp. 1268-1290.

Berry, ST 1994, ‘Estimating discrete-choice models of product differentiation’, RAND Journal of Economics, vol. 25, no. 2, pp. 242-262.

Berry, S, Levinsohn, J, & Pakes, A 1995, ‘Automobile prices in market equilibrium’, Econometrica, vol. 63, no. 4, pp. 841-890.

Berry, S, Levinsohn, J, & Pakes, A 2004, ‘Differentiated products demand systems from a combination of micro and macro data: The new car market’, Journal of Political Economy, vol. 112, no. 1, pp. 68-105.

Bhagwati, J 1982, ‘Directly unproductive profit-seeking (DUP) activities’, Journal of Political Economy, October 1982, vol. 90, no. 5, pp. 988-1002.

Bhat, CR 2000, ‘Incorporating observed and unobserved heterogeneity in urban work travel model choice modeling’, Transport Science, vol. 34, no. 2, May 2000, pp. 228-238, viewed 12 February 2019, .

Bombardini, M, Kurz, CJ, & Morrow, PM 2012, ‘ Ricardian trade and the impact of domestic competition on export performance’, Canadian Journal of Economics, vol. 45, no. 2, pp. 585-612.

Bresnahan, TF 1987, ‘Competition and collusion in the American automobile industry: The 1955 price war’, The Journal of Industrial Economics, vol. 35, no. 4, The Empirical Renaissance in Industrial Economics (Jun. 1987), pp. 457-482, viewed 18 March 2013, .

Brus, DJ, Slim, PA, Gort, G, Heidema, AH, & van Dobben, H 2016, ‘Monitoring habitat types by the mixed multinomial logit model using panel data’, Ecological Indicators, vol. 67, pp. 108-116.

Burns, LD & Golob, TF 1976, ‘The role of accessibility in basic transportation choice behavior’, Transportation, vol. 5, pp. 175-198.

Busse, MR, Knittel, CR & Zettelmeyer, F 2013, ‘Are consumers myopic? Evidence from new and used car purchases’, American Economic Review, vol. 103, no. 1, pp. 220-256, viewed 5 March 2019, .

Canto, VA, Eastin, RV, & Laffer, AB 1982, ‘Failure of protectionism: A study of the steel industry’, Columbia Journal of World Business, Winter 1982, pp. 43-57.

‘Car Direct’ 2011, Online Car Portal, viewed 31 January 2012, .

Charnes, A, Cooper, WW, & Rhodes, E 1978, ‘Measuring the efficiency of decision making units’, European Journal of Operational Research, vol. 2, pp. 429-444. 242

‘Chinese carmaker Geely to acquire Proton’ 2017, The Star online, viewed 6 September 2018, .

‘China’s 2012 vehicle sales up 4.3% miss forecast’ 2013, Market Watch, The Wall Street Journal, 11 January 2013, viewed 25 February 2014, .

Choong, EH 2012, “Proton bids farewell to Bursa after 9-year listing”, The Star, 4 May 2012, viewed 8 May 2012,

Corden, WM 1997, Trade policy and economic welfare, 2ndedn, Clarendon press, Oxford.

Cook, WD, Tone, K, & Zhu, J 2014. ‘Data envelopment analysis: Prior to choosing a model’, Omega, vol. 44, pp. 1-4.

Costinot, A 2009, ‘Jobs, jobs, jobs: A “new” perspective on protectionism’, Journal of the European Economic Association, vol. 7, no. 5, pp. 1011-1041, viewed 4 April 2014, .

Coughlin, CC, Chrystal, AK, & Wood GE 2000, ‘Protectionist trade policies: A survey of theory, evidence, and rationale’, Federal Reserve Bank of St Louise, January/February 1988, viewed 15 May 2014, .

Craft, ED 2002. ‘The demand for vanity (plates): Elasticities, net revenue maximization, and deadweight loss’, Contemporary Economic Policy, vol. 20, no. 2, April 2002, pp. 133-144.

Cragg, JG and Uhler, RS 1970, ‘The demand for automobiles’, Canadian Journal of Economics, 3, pp. 386-406.

Cramer JS 2004, ‘The early origin of the logit model’, Studies in History and Philosophy of Biological and Biomedical Sciences, 35, pp. 613-626.

David, PA 1985, ‘Clio and the Economics of QWERTY’, The American Economic Review, vol. 75, no. 2, Papers and Proceedings of the Nineth-seventh annual meeting of the American Economic Association (May 1985), pp. 332-337.

Das, S 1995, ‘Size, age and firm growth in an infant industry: The computer hardware industry in India’, International Journal of Industrial Organization, vol. 13, pp. 111-126.

Dasgupta, P & Stiglitz, J 1988, ‘Learning-by-doing, market structure and industrial and trade policies’, Oxford Economic Papers, New Series, vol. 40, no. 2 (Jun 1988), pp. 246- 268, viewed 25 April 2013, .

243

De Borger, B & Mayeres, I 2007, ‘Optimal taxation of car ownership, car use and public transport: Insights derived from a discrete choice numerical optimization model’, European Economic Review, 51, pp. 1177-1204.

De Melo, J & Tarr, D 1992, A general equilibrium analysis of US foreign trade policy, MIT Press.

De Pelsmacker, P 1990, ‘A structural model of the demand for new cars in Belgium’, Applied Economics, vol. 22, pp. 669-686.

Deardorff, A 2006, ‘The Heckscher-Ohlin model: features, flaws, and fixes’, GEP newsletter, Leverhulme Centre for Research on Globalisation and Economic Policy, issue 16 winter, 2006.

Deardorff, A 2005, ‘How robust is comparative advantage?’ Research Seminar in International Economics, Discussion Paper No. 537, viewed 28 March 2014, .

Deardorff, A 2004, ‘Local comparative advantage: Trade costs and the pattern of trade’, Research Seminar in International Economics, Discussion Paper No. 500, viewed 28 March 2014, .

Department of Statistics Malaysia 2013, ‘Monthly external trade statistics’ , viewed 29 August 2014, .

Dickson, PR & Ginter, JL 1987, ‘Market segmentation, product differentiation, and marketing strategy’, Journal of Marketing, vol. 51, no. 2 (Apr. 1987), pp. 1-10

Diewert, EW & Lawrence, DA 1996, ‘The deadweight costs of taxation in New Zealand,’ Canadian Journal of Economics, vol. XXIX, Special Issue, pp. 658-673.

‘Drive’ 2011, The Age, Holiday edition, 30-31 December, 2011, p. 14, viewed 22 May 2012,

‘Drive’ 2011, The Age, October - December, 2011.

‘Drive’ 2011, The Saturday Age, 12 November 2011.

Dupuit, J 1969, On the measurement of the utility of public works, in KJ Arrow & T Scitovsky, ed., reprinted in Readings in welfare economics, Irwin, Homewood, Illinois.

Dyson, RG et al. 2001. ‘Pitfalls and protocols in DEA’, European Journal of Operational Research, vol. 132, pp. 245-259.

Earl, PE 1995, Microeconomics for Business and Marketing - Lectures, cases and worked essays, Edward Elgar.

244

Ebert, RR & Montoney, M 2007, Performance of the South Korean Automobile Industry in the domestic and United States markets, The Baldwin-Wallace College Journal of Research and Creative Studies, Fall 2007, vol. 1, no. 1, pp. 12-24.

Economic and Social Commission for Asia and the Pacific (ESCAP) 2002, ‘The development of the automotive sector in selected countries of the ESCAP region’ 2002, , proceedings and country papers presented at the regional consultative meeting on Promotion of intraregional trade and economic cooperation in the automotive sector, 2002, United Nation: New York, viewed 2 February 2012,

Economist, The 2012, ‘Car industry: Danger ahead-The car industry’s crisis is over. Its long-term problems are not’, 13 January 2012, viewed 20 June 2012, .

Ederington, J & McCalman, P 2011, ‘Infant industry protection and industrial dynamics’, Journal of International Economics, vol. 84, pp. 37-47.

Farrell, MJ 1954, ‘The demand for motor-cars in the United States’, Journal of the Royal Statistical Society, 117A, pp. 171-201.

Federal Reserve Bank of New York 1985, The consumer cost of U.S. trade restraints, FRBNY Quarterly Review, Summer 1985, pp. 1-12.

Feenstra, RC 1992, ‘How costly is protectionism?’, Journal of Economic Perspectives, vol. 6, no. 3, summer 1992, pp. 159-178.

Feenstra, RC 1985, ‘Automobile prices and protection: The U.S.-Japan trade restraint’, Journal of Policy Modeling, vol. 7, no. 1, pp. 49-68.

Feenstra, RC & Levinsohn, JA 1995, ‘Estimating mark-ups and market conduct with multidimensional product attributes’, The Review of Economic Studies, vol. 62, no. 1, pp. 19-52, viewed 15 July 2013, .

Formby, JP, Keller, JP & Thistle, PD 1988, ‘X-efficiency, rent-seeking and social cost’, Public Choice, vol. 57, pp. 115-126.

Fountas, G & Anastasopoulos, P Ch. 2018, ‘Analysis of accident injury-severity outcomes: The zero-inflated hierarchical ordered probit model with correlated disturbances’, Analytic Methods in Accident Research, 20, pp. 30-45.

Fuss, M & Waverman, L 1986, ‘The Canada-U.S. auto pact of 1965: An experiment in selective trade liberalization’, NBER Working Paper Series, Working Paper No. 1953, National Bureau of Economic Research, Massachusetts.

German Chamber Network 2012, ‘Market Watch 2012: The Malaysian automotive and supplier industry’, viewed 29 August 2014, .

245

Golany, B & Roll, Y 1989, ‘An application procedure for DEA,’ OMEGA International Journal of Management Science, vol. 17, no. 3, pp. 237-250.

Goldberg, PK 1995, ‘Product differentiation and oligopoly in international markets: The case of the U.S. automobile industry’, Econometrica, vol. 63, no. 4 (July, 1995), pp. 891- 951, viewed 21 February 2019,

Gomez, ET and Jomo, KS 1999, Malaysia’s political economy - Politics, patronage and profit, 2nd edn, Cambridge University Press, UK.

Greenaway, D & Kneller, R 2007, ‘Firm heterogeneity, exporting and foreign direct investment’, The Economic Journal, vol. 117, pp. 134-161.

Greenbaum, A 2002, The globalization of the Korean automotive industry, Economic Strategy Institute, Washington DC.

Greene, WH, Hensher, DA & Rose, J 2006, ‘Accounting for heterogeneity in the variance of unobserved effects in mixed logit models’, Transportation Research Part B, 40, pp. 75-92.

Grossman, GM & Horn, H 1987, ‘Infant-industry protection reconsidered: The case of informational barriers to entry’, National Bureau of Economic Research, Working paper no. 2159, February 1987, viewed 20 March 2014, .

Gu, G, Yang, D, Feng, T & Timmermans, H 2019, ‘Household vehicle holding decisions in response to life cycle events’, Transportation Research Procedia, 37, 21st EURO Working Group on Transportation Meeting, EWGT 2018, 17th - 19th September, 2018, pp. 171-178.

Haaf, GC, Morrow, RW, Azevedo, IML, Feit, EM, & Michalek, JJ 2016, ‘Forecasting light-duty vehicle demand using alternative-specific constants for endogeneity correction versus calibration,’ Transportation Research Part B, vol. 84, pp. 182-210.

Hair, Jr, JF, Hult, GT, Ringle, CM, & Sarstedt, M 2014, A Primer on partial least squares structural equation modelling (PLS-SEM), Singapore: SAGE.

Hair, Jr, JF et al. 2006, Multivariate Data Analysis, 6th edn, New Jersey: Pearson Prentice Hall.

Harris, R 1984. “Applied general equilibrium analysis of small open economies with scale economies and imperfect competition,” American Economic Review, vol. 74, no. 5, pp. 1016-1032.

Hausman, JA & Wise, DA 1978, “A conditional probit model for qualitative choice: Discrete decisions recognizing interdependence and heterogeneous preference.” Econometrica, vol. 46, pp. 403-426.

246

Havemen, J & Hummels, D 2004, ‘Alternative hypotheses and the volume of trade: the gravity equation and the extent of specialization’, Canadian Journal of Economics, vol. 37, no. 1, pp. 199-218.

Head, K 1994, ‘Infant industry protection in the steel rail industry’, Journal of International Economics, vol. 37, pp. 141-165.

Hedeker, D 2003, ‘A mixed-effects multinomial logistic regression model’, Statistics in Medicine, vol. 22, pp. 1433-1446.

Helpman, E & Krugman, PR 1985, Market structure and foreign trade - Increasing returns, imperfect competition, and the international economy, The MIT Press, Massachusetts.

Hensher, DA & Le Plastrier, V 1985, ‘Towards a dynamic discrete-choice model of household automobile fleet size and composition’, Transportation Research, vol. 19B, pp. 481-495.

Hickok, S 1985, ‘The consumer cost of U.S. trade restraints’, Federal Reserve Bank of New York - Quarterly Review (Summer 1985), pp. 1-12, viewed 10 August 2013, .

Honda 2019, ‘Pricing’, viewed 1 April 2019, .

Hsieh, FY 1989, ‘Sample size tables for logistic regression,’ Statistics in Medicine, vol. 8, pp. 795-802.

Hufbauer, GF & Elliot, KA 1994. Measuring the cost of protection in the United State, Institute for International Economics, Washington DC.

Imben, GW & Lancaster, T (1994). Combining micro and macro data in microeconometric models, Review of Economic Studies, vol. 61, pp. 655-680.

Irawan MZ, et al. 2018, ‘A market share analysis for hybrid cars in Indonesia’, Case Studies on Transport Policy, vol. 6, pp. 336-341. ir.ChartNexus 2013, UMW Toyota, viewed 12 May 2014, .

Irwin, DA 2009, Free trade under fire, 3rd edn, Princeton University Press, Princeton, USA.

Irwin, DA 2000, ‘Could the U.S. iron industry have survived free trade after the civil war?’, National Bureau of Economic Research, Working paper no. 7640, viewed 20 March 2014, .

247

Irwin, DA 1998, ‘Did late nineteenth century U.S. tariffs promote infant industries? Evidence from the tinplate industry’, National Bureau of Economic Research, Working paper no. 6835, December 1998, viewed 20 March 2014, .

Isaksson, A, Ng, TH, & Robyn, G 2005, ‘Productivity in developing countries: Trends and Policies’, UNIDO Research Programme, United Nations Industrial Development Organization, viewed 9 April 2019, .

Israel, GD 1992. Sampling the evidence of extension program impact. Program Evaluation and Organizational Development, IFAS, University of Florida, PEOD-6.

Jabatan Pengangkutandan Jalan Raya (JPJ) (Road and Transport Department, Malaysia), ‘Statistics of the numbers of Proton cars registered for the period 2001 to 2010’, unpublished data.

Jabatan Perdana Menteri (Prime Minister Department, Malaysia) 2005, ‘National automotive policy framework’, viewed 08 February 2012, .

J D Power Asia Pacific 2011, ‘J D Power 2011 Malaysia Initial Quality Study’, viewed 16 November 2012, .

J D Power Asia Pacific 2010, ‘J D Power Asia Pacific Reports: Overall New-Vehicle Initial Quality in Malaysia declines slightly’ 2010, J D Power Asia Pacific Press Release 2010, The McGraw Hill Companies.

Jaipragas, B 2018, ‘Is Chinese carmaker Geely being anti-Malay in cost-cutting drive at Malaysia’s Proton?’, South China Morning Post, viewed 6 August 2018, .

Jomo, KS 2004, ‘The New Economic Policy and interethnic relations in Malaysia, Identities, Conflict, and Cohesion,’ Programme Paper Number 7, September 2004, United Nations Research Institute for Social Development, viewed 15 May 2014, .

Jørgensen, F, Mathisen, TA & Pedersen, H 2016, ‘Brand loyalty among Norwegian car owners,’ Journal of Retailing and Consumer Services, 31, pp. 256-264.

Katayama, J, Lu, S, and Tybout, J 2003, ‘Why plant-level productivity studies are often misleading, and an alternative approach to inference,’ NBER Working Paper Series, Working paper 9617, viewed 25 July 2014, .

Kana, G 2017, ‘Zhejiang Geely buys Proton, Lotus Stakes for RM1bil’, The Start online, 23 Jun 2017, viewed 7 September 2018,

248

.

Keller, W 2004, ‘International technology diffusion,’ Journal of Economic Literature, vol. XLII (September 2004), pp. 752-782.

Keller, W 2001, ‘The geography and channels of diffusion at the world’s technology frontier,’ NBER Working Paper Series, Working Paper 8150, viewed 23 July 2014, .

Keller, W 1996, ‘Absorptive capacity: on the creation and acquisition of technology in development’, Journal of Development Economics, vol. 49, pp. 199-227.

Kemal, AR 1979, ‘Infant industry argument, protection and manufacturing industries of Pakistan,’ The Pakistan Development Review, vol. XVIII, no. 1 (Spring 1979), pp. 1-19.

Khoo, D 2014. Auto sales continue to drive UMW, The Star Online, Business News, viewed 17 July 2017, < http://www.thestar.com.my/business/business- news/2014/09/27/auto-sales-continue-to-drive-umw-car-business-remains-a-mainstay- of-the-company-automotiverelated-b/>.

Kon 2010, ‘Proton Inspira - First driving impressions,’ Autoworld.com.my - Test drives & Reviews 27 Oct 2010, viewed 16 April 2019, .

Kopp, RJ 1981. ‘The measurement of productive efficiency: A reconsideration’, The Quarterly Journal of Economics, pp. 474-503.

Krueger, AO 1974, ‘The political economy of the rent-seeking society’, The American Economic Review, vol. 64, no. 3 (June 1974), pp. 291-303.

Krueger, AO & Tuncer, B 1982, ‘An empirical test of the infant industry argument,’ American Economic Review, December 1982, vol. 72, pp. 1142-52.

Krugman, P 1990. Rethinking International Trade, The MIT Press, Massachusetts.

Krugman, P & Obstfeld, M 1994, International Economics - Theory and Policy, 3rd edn, Harper Collins College Publishers

Krugman, J, Obstfeld, M, & Melitz, MJ 2012, International Economics - Theory & Policy, Global Edition, 9ed, Pearson, England.

Kunst, R & Marin, D 1989, ‘On exports and productivity: A causal analysis,’ The Review of Economics and Statistics, vol. 71, no. 4, pp. 699-703.

Lagarde, C 2017, ‘Reinvigorating productivity growth,’ Speech, IMF Communication Department, viewed 9 April 2019, . 249

Lancaster, KJ 1966, ‘A new approach to consumer theory,’ The Journal of Political Economy, vol. 74, no. 2 (April 1966), pp. 132-157.

Langer, A & Miller, NH 2011, ‘Automakers’ short-run responses to changing gasoline prices and the implications for energy policy,’ Review of Economics and Statistics, vol. 95, 1198-1211.

Lave, CA & Train, K 1979, ‘A disaggregate model of auto-type choice,’ Transport Research, vol. 13A, pp. 1-9.

Lave, CA & Bradley, J 1980, ‘Market share of imported cars: A model of geographic and demographic determinants,’ Transport Research A, vol. 14A, pp. 379-387.

Leahy, D & Neary, JP 1999, ‘Learning by doing, pre-commitment and infant-industry promotion,’ Review of Economic Studies, vol. 66, pp. 447-474.

Lee, C 2011, ‘Trade, productivity, and innovation: Firm-level evidence from Malaysian manufacturing’, Journal of Asian Economics, vol. 22, pp. 284-294.

Lee, J 1997, ‘The maturation and growth of infant industries: The case of Korea,’ World Development, vol. 25, no. 8, pp. 1271-1281.

Lee, L 2014, ‘Longer turnaround seen for Proton’, StarBiz, 4 March, 2014, p. 4.

Lee, SI 2013, ‘Malaysia - Dr Mahathir says ending APs means ending Proton and Perodua’, 28 November 2013, viewed 15 May 2014, .

Leibenstein, H 1966, ‘Allocative efficiency vs. ‘X-efficiency’, The American Economic Review, vol. 56, no. 3 (June, 1966), pp. 392-415.

Leibenstein, H 1979, ‘A branch of economics is missing: Micro-micro theory,’ Journal of Economic Literature, vol. XVII (June 1979), pp. 477-502.

Leibenstein, H 1989, ‘Organizational economics and institutions as missing elements in economic development analysis,’ World Development, vol. 17, no. 9, pp. 1361-1373.

Leibenstein, H & Maital, S 1994, ‘The organizational foundations of X-inefficiency – a game-theoreti interpretation of Argyris’ model of organizational learning,’ Journal of Economic Behavior and Organization, vol. 23, pp. 251-268.

Levinsohn, J 1988, ‘Empirics of taxes on differentiated products: The case of tariffs in the U.S. automobile industry,’ in R Baldwin (ed.), Trade Policy Issues and Empirical Analysis, Chicago: NBER.

Lim, A 2019, ‘Another record year for Mercedes-Benz Malaysia - 13,079 vehicles delivered in 2018, up by 8.6%,’ viewed 7 March 2019,

250

.

Liu, Y & Cirillo, C 2017, ‘A generalized dynamic discrete choice model for green vehicle adoption,’ 22nd International Symposium on Transportation and Traffic Theory, Transport Research Procedia 23, pp. 868-886.

Lo, M 2007, High amount of taxpayers money is used to pay salaries and pensions of govt staff, The Star Online, 03/02/2017, viewed 29 March 2017, .

Luzio, E & Greenstein, S 1995, ‘Measuring the performance of a protected infant industry: The case of Brazilian microcomputers’, Review of Economics and Statistics (February 1995), pp. 622-633.

Lye, G 2019, ‘Vehicle sales performance in Malaysia, 2018 vs 2017 – a look at last year’s biggest winners and losers’, 23/01/2019, viewed 9 August 2019, .

Mahalingam, E 2011, ‘Many market players don’t think APs will be abolished’, Starbizweek, 19 March, p. SBW22.

Malaysia Automotive Association (MAA) 2018, ‘Malaysia: Duties & taxes on motor vehicles’, viewed 28 March 2019, < http://www.maa.org.my/info_duty.htm>.

Malaysia Automotive Association (MAA) 2013, ‘Market Review for 2012’, MAA press conference 23 January 2013, viewed 7 November 2013, .

Malaysia Automotive Association (MAA) 2012, ‘Press conference on Market review for 2012 and outlook for 2013’, viewed 2 April 2013, .

Malaysia Automotive Association (MAA) 2008, ‘Malaysia automotive info - duty structure’ 2008, viewed 31 January 2012, .

Malaysia Automotive Association (MAA) 2006, ‘National Automotive Framework’, viewed 25 July 2019, .

Malaysia Automotive Association (MAA) 2006, ‘Malaysia automotive info - Summary of sales and production data’ 2006, viewed 31 January 2012, .

Mannering, F & Train, K 1985, ‘Recent directions in automobile demand modelling,’ Transportation Research, vol. 19B, pp. 265-274.

251

Manski, CF & Sherman, L 1980, ‘An empirical analysis of household choice among motor vehicles,’ Transportation Research, vol. 14A, pp. 349-366.

Maps of World 2013, ‘Top 10 car manufacturing companies’, viewed 6 March 2013, .

Mariuzzo, F 2002. ‘Automobile equilibrium prices: An empirical study on the Italian market,’ Working paper 2002.06, Univerita’ degli Studi di Venezia, viewed 9/7/2013, .

Markusen, JR 1992, Trade and the gains from trade with imperfect competition. In Imperfect competition and international trade, ed. Gene M Grossman, The MIT press, Massachusetts.

McFadden, D 1974, ‘Conditional logit analysis of qualitative choice behaviour’. In Frontiers in Econometrics, edited by Paul Zarembka, New York: Academic Press, pp. 105-142.

McFadden, D & Train, K 2000, ‘Mixed MNL models for discrete response’, Journal of Applied Econometrics, 15, pp. 447-470.

Mefford, R 2017, ‘X-efficiency: economists and managers view it differently,’ Journal of Behavioral Economics for Policy, vol. 1, no. 2, pp. 25-30.

Melitz, MJ 2003, ‘The impact of trade on intra-industry reallocations and aggregate industry productivity,’ Econometrica, vol. 71, no. 6, pp. 1695-1725.

Melitz, MJ 2005, ‘When and how should infant industries be protected?’ Journal of International Economics, vol. 66, pp. 177-196.

Melitz, MJ & Ottaviano, GI 2008, ‘Market size, trade, and productivity’, Review of Economic Studies, vol. 75, pp. 295-316.

Ministry of Information, Communication and Culture, Malaysia 2012, ‘Achievement: 1980 Heavy Industry Policy’, viewed 26 March 2012, .

Ministry of International Trade and Industry 2019, ‘Free Trade Agreement 2019,’ Official portal of the Ministry of International Trade and Industry, viewed 12 April 2019, .

Ministry of International Trade and Industry (MITI) 2010, MITI Weekly Bulletin, vol. 103, 27 July 2010.

Ministry of International Trade and Industry (MITI) 2009, MITI Weekly Bulletin, vol. 33, 25 February 2009.

252

Minondo, A 2010, ‘Exports’ quality-adjusted productivity and economic growth’, The Journal of International Trade & Economic Development, vol. 19, no. 2, pp. 257-287.

Moon, HR, Shum, M & Weidner, M 2018, ‘Estimation of random coefficients logit demand models with interactive fixed effects’, Journal of Econometrics, 206, pp. 613- 644.

Morkre, ME & Tarr, DG 1980, ‘Effects of restrictions on United States imports: Five case studies and theory, Bureau of Economics Staff Report to the Federal Trade Commission, USGPO.

Motor Trader 2012, ‘And Proton’s new owner will be…’, viewed 8 May 2012, .

Munger, MC 1984, ‘The costs of protectionism’, Challenge, January-February 1984, pp. 54-58.

National Roads and Motorists’ Association Limited 2010, ‘Proton S16 2010 on sedan ANCAP Review and Crash Test’, viewed 22 May 2012, .

Nevo, A 2000. ‘A practitioner’s guide to estimation of random-coefficients logit models of demand,’ Journal of Economics & Management Strategy, vol. 9, no. 4, Winter 2000, pp. 513-548.

Nishiwaki, M 2006, ‘Measuring the effect of the infant industry protection - The Japanese automobile industry in 1955-1965,’ Discussion paper #2006-23, Graduate school of Economics, Hitotsubashi University.

Noorzoy, MS 1979, ‘Tariff reductions and gains in efficiency - Some evidence from Canadian Data,’ Economics Letters, vol. 2, pp. 51-57.

Obstfeld, M & Rogoff, K 2000, ‘The six major puzzles in international macroeconomics: Is there a common cause?,’ NBER Working Paper Series, Working Paper 7777, viewed 23 July 2014, .

Okamoto, Y & Sjӧholm, F 2000, ‘Productivity in the Indonesian automotive industry’, ASEAN Economic Bulletin, vol. 17, no. 1, pp. 60-73.

Orlov, A & Kallbekken, S 2019, ‘The impact of consumer attitudes towards energy efficiency on car choice: Survey results from Norway’, Journal of Cleaner Production, 214, pp. 816-822.

Østli, V, Fridstrom, L, Johansen, KV & Tseng, YY 2017, ‘A generic discrete choice model of automobile purchase,’ European Transport Research Review, vol. 9, no. 16, viewed 1 March 2019, < https://link.springer.com/article/10.1007/s12544-017-0232-1>.

Page, M, Whelan, G & Daly, A 2000, ‘A discrete choice model of car type choice,’ paper presented at the European Transport Conference 2000. 253

Panagariya, A 2002, ‘Alternative approaches to measuring the cost of protection,’ Mimeo, University of Maryland, January 2002, viewed 20/04/2012, .

‘Passenger cars per 1000 people’, Econ Stats, viewed 5 October 2018, . Pavcnik, N 2002, ‘Trade liberalization, exit, and productivity improvements: Evidence from Chilean plants’, Review of Economic Studies, vol. 69, pp. 245-276.

Perodua Manufacturing Sendirian Berhad (95999 T) 2015, Annual Report 2015, Companies Commission of Malaysia.

‘Perodua pushes Proton to the brink’ 2015, KINIBIZ Cover, pp. 53-68.

Petrin, A 2002, ‘Quantifying the benefits of new products: The case of the minivan,’ Journal of Political Economy, vol. 110, no. 4. pp. 705-729.

Phongpetra, V & Johri, LM 2011, ‘Impact of business strategies of automobile manufacturers in Thailand’, International Journal of Emerging Markets, vol. 6, no. 1, pp. 17-37.

Proton Holdings Berhad (623177 A), Annual Report 2016, Companies Commission of Malaysia.

Ravenhill, J 2001, ‘From national champions to global partnerships: The Korean auto industry, financial crisis and globalization’, MIT Japan Program, Working Paper 01.04, pp. 1-22.

Romer, P 1994, ‘New goods, old theory, and the welfare costs of trade restrictions’, Journal of Development Economics, 43, pp. 5-38.

Santerre, RE & Vernon, JA 2006, ‘Assessing consumer gains from a drug price control policy in the United States’, Southern Economic Journal, vol. 73, no. 1, pp. 233-245.

Saunders, M, Lewis, P & Thornhill, A 2009, Research methods for business students, 5th edn, Prentice Hall Financial Times, Pearson Education.

Sauré, P 2007, ‘Revisiting the infant industry argument,’ Journal of Development Economics, vol. 84, pp. 104-117.

Shen, TY 1984, ‘The estimation of X-inefficiency in eighteen countries,’ Review of Economics and Statistics, vol. 66, no. 1, pp. 1361-1373.

Sidhu, JS 2014, ‘Proton seeks more funds - up to RM3billion needed to develop new models, including electric car’, StarBiz, 5 March 2014, p. 1.

‘Slow take-up for Automotive Development Fund’ 2007, Bernama.com, viewed 14 June 2012, .

254

Smith, W 1956, ‘Product differentiation and market segmentation as alternative marketing strategies’, Journal of Marketing, vol. 21, pp. 3-8.

Statista 2018, ‘Leading car manufacturing countries worldwide in 2016 ranked by production volume’, viewed 20 September 2018, .

Statista 2018, ‘Leading motor vehicle manufacturer worldwide in 2017, based on global sales (in million units)’, viewed 24 September 2018, .

Stopher, PR 1969, ‘A probability model of travel mode choice for the work journey,’ Highway Research Record, Issue 283, pp. 57-65.

Sugimori, Y, Kusunoki, K, Cho, F & Uchikawa, S 1977, ‘Toyota production system and Kanban system materialization of just-in-time and respect-for-human system,’ The International Journal of Production Research, vol. 15, no. 6, pp. 553-564.

Suits, D 1958, ‘The demand for new automobiles in the United States 1929-1956,’ Review of Economics and Statistics, vol. 40, pp. 273-80.

Tan, D 2018, ‘BMW Malaysia’s H1 2018 sales up 11%, MINI up 18%,’ viewed 7 March 2019,

Tan, P 2017, ‘Vehicle sales performance in Malaysia - 2016 vs 2015 - a look at last year’s biggest winners and losers’, viewed 22 August 2017, .

Tarr, DG & Morkre, ME 1984. Aggregate costs to the United States of tariffs and quotas on imports: General tariff cuts and removal of quotas on automobiles, steel, sugar, and textiles - An economic policy analysis, Bureau of Economic Staff Report to the Federal Trade Commission, December 1984.

Thailand Board of Investment 2017, ‘Thailand’s automotive industry - The next generation’, viewed 11 September 2018, .

Thailand Board of Investment, 2012 ‘Thailand: Automotive hub of Asia’, 21 September 2012, viewed 10 October 2013, .

Thailand Board of Investment 2012, ‘Thailand’s automotive industry’, viewed 20 August 2014, .

255

Theil, H 1969, ‘A multinomial extension of the linear logit model,’ International Economic Review, vol. 10, no. 3 (October 1969), pp. 251-259, viewed 1 February 2019, .

Tractus Asia Ltd 2014,‘Overview of automotive industry sector and route to market’, viewed 26 August 2014, .

Train, KE 1998, ‘Recreation demand models with taste differences over people,’ Land Economics, vol. 74, no. 2 (May 1998), pp. 230-239.

Train, KE & Winston, C 2007, ‘Vehicle choice behaviour and the declining market share of U.S. automakers,’ International Economic Review, vol. 48, no. 4, pp. 1469-1496.

Turnovsky, SJ 1966, ‘The New Zealand automobile market, 1948-63: An econometric case-study of disequilibrium,’ Economic Record, vol. 42, no. 98, pp. 256-273.

UMW Toyota Motor Sendirian Berhad (60576 K) 2015, Annual Report 2015, Companies Commission of Malaysia.

U.S. Department of State 2012, ‘Background note: Malaysia’, viewed 2 April 2012, .

Wad, P & Govindaraju, VGRC 2011, ‘Automotive industry in Malaysia: an assessment of its development’, International Journal of Automotive and Management, vol. 11, no. 2, pp. 152-171.

Wagner, J 2008, ‘Export entry, export exit and productivity in German manufacturing industries,’ International Journal of the Economics of Business, vol. 15, no. 2, pp. 169- 180.

Weidenbaum, ML 1983, ‘The high cost of protectionism,’ Cato Journal, vol. 3, no. 3 (Winter 1983/84), pp. 777-791.

Wikipedia 2012, Proton, viewed 8 May 2012, .

Wrightreport 2008, Proton Holdings Berhad, The Winthrop Corporation, viewed 24 February 2011, .

Wonnacott, P & Wonnacott, RJ 1980, ‘Free trade between the United States and Canada: Fifteen years later,’ Working papers, Paper 23, viewed 25 April 2014, .

Workman, D 2018, Car exports by country, viewed 21 September 2018, .

256

World Bank 2019, ‘GDP growth (annual %),’ The World Bank - Data’, viewed 17 April 2019, .

World Bank 2019, ‘GDP per capita: Japan (current USD)’, viewed 20 August 2019, .

Yusof, A 2018, ‘Proton resume cars export to the Middle East, via Jordan’, New Strait Time online, viewed 7 September 2018, .

Zhen, C & Wei, (D) F 2019, ‘A multinomial logit model of pedestrian-vehicle crash severity in North Carolina,’ International Journal of Transportation Science and Technology, vol. 8, pp. 43-52.

Zigwheels 2019, ‘Perodua cars Malaysia’, viewed 1 April 2019, .

Zigwheels 2019, ‘Proton cars Malaysia’, viewed 1 April 2019, .

Zigwheels 2019, ‘Toyota cars Malaysia’, viewed 1 April 2019, .

257

APPENDICES

A. Statistics

Test of Correlation - Car size and number of dependents Variance and Covariance Covariance / DEP Variance / NS Variance / DF NS 0.3815 1.8283 DEP 1.0710 349

Variable N Mean Std Dev Sum Minimum Maximum Dep 350 1.9257 1.3521 674 0 9 NS 442 5.3326 0.9990 2357 4 16

258

B. SAS® Generated Output

1. Proton

Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 86.0765 2 <.0001 Score 33.0283 2 <.0001

Wald 31.6228 2 <.0001 Analysis of Maximum Likelihood Estimates Standard Wald Parameter DF Estimate Error Chi-Square Pr > ChiSq Intercept 1 769.2 146.9 27.4038 <.0001 LP1 1 -328.1 62.3583 27.6869 <.0001

LP1*LP1 1 34.9929 6.6133 27.9979 <.0001 Association of Predicted Probabilities and Observed Responses Percent Concordant 73.0 Somers' D 0.460 Percent Discordant 26.9 Gamma 0.461 Percent Tied 0.1 Tau-a 0.192 Pairs 41580 c 0.730

Hosmer and Lemeshow Goodness-of-Fit Test Chi-Square DF Pr > ChiSq 7.3735 8 0.4969

259

260

2. Perodua

Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 244.3172 6 <.0001 Score 160.8270 6 <.0001

Wald 84.3541 6 <.0001 Analysis of Maximum Likelihood Estimates Standard Wald Parameter DF Estimate Error Chi-Square Pr > ChiSq Intercept 1 148.9 61.6547 5.8319 0.0157 LP1 1 -37.6318 14.7087 6.5458 0.0105 HP1 1 -112.6 40.2338 7.8297 0.0051 LP1*HP1 1 29.5837 9.5310 9.6344 0.0019 LP1*NS 1 0.9906 0.4699 4.4428 0.0350 HP1*NS 1 -3.9703 1.5233 6.7928 0.0092

NS*kml1 1 -0.0538 0.0122 19.5139 <.0001 Association of Predicted Probabilities and Observed Responses Percent Concordant 90.8 Somers' D 0.816 Percent Discordant 9.2 Gamma 0.816 Percent Tied 0.0 Tau-a 0.379

Pairs 39713 c 0.908

Hosmer and Lemeshow Goodness-of-Fit Test Chi-Square DF Pr > ChiSq 10.7969 8 0.2135

261

262

263

3. Toyota

Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 83.3210 6 <.0001 Score 67.7286 6 <.0001

Wald 46.6092 6 <.0001 Analysis of Maximum Likelihood Estimates Standard Wald Parameter DF Estimate Error Chi-Square Pr > ChiSq Intercept 1 151.3 34.7097 19.0030 <.0001 LP1 1 -33.8343 7.3378 21.2607 <.0001 HP1 1 -69.8758 19.9897 12.2192 0.0005 kml1 1 1.1146 0.4041 7.6079 0.0058 LP1*HP1 1 14.3488 4.0271 12.6953 0.0004 LP1*NS 1 0.5711 0.1973 8.3760 0.0038

kml1*NS 1 -0.2154 0.0760 8.0400 0.0046 Association of Predicted Probabilities and Observed Responses Percent Concordant 84.9 Somers' D 0.699 Percent Discordant 15.1 Gamma 0.699 Percent Tied 0.0 Tau-a 0.166

Pairs 20349 c 0.849

Hosmer and Lemeshow Goodness-of-Fit Test Chi-Square DF Pr > ChiSq 8.6662 8 0.3712

264

265

266

4. Honda

Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 50.1622 3 <.0001 Score 37.4768 3 <.0001

Wald 26.5226 3 <.0001 Analysis of Maximum Likelihood Estimates Standard Wald Parameter DF Estimate Error Chi-Square Pr > ChiSq Intercept 1 163.3 43.1539 14.3240 0.0002 LP1 1 -32.7143 8.7273 14.0514 0.0002 HP1 1 -79.0034 24.4109 10.4743 0.0012

LP1*HP1 1 15.9638 4.9138 10.5546 0.0012 Association of Predicted Probabilities and Observed Responses Percent Concordant 83.6 Somers' D 0.673 Percent Discordant 16.4 Gamma 0.673 Percent Tied 0.0 Tau-a 0.098

Pairs 14385 c 0.836

Hosmer and Lemeshow Goodness-of-Fit Test Chi-Square DF Pr > ChiSq 3.0713 8 0.9298

267

268

5. Nissan

Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 25.7000 2 <.0001 Score 18.4790 2 <.0001

Wald 12.6598 2 0.0018 Analysis of Maximum Likelihood Estimates Standard Wald Parameter DF Estimate Error Chi-Square Pr > ChiSq Intercept 1 478.8 190.2 6.3360 0.0118 LP1 1 -188.0 76.6218 6.0217 0.0141

LP1*LP1 1 18.5275 7.7127 5.7706 0.0163 Association of Predicted Probabilities and Observed Responses Percent Concordant 79.6 Somers' D 0.593 Percent Discordant 20.3 Gamma 0.593 Percent Tied 0.1 Tau-a 0.053

Pairs 8946 c 0.796

Hosmer and Lemeshow Goodness-of-Fit Test Chi-Square DF Pr > ChiSq 7.8525 8 0.4480

269

270

6. American/European

Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 86.3725 2 <.0001 Score 119.8515 2 <.0001

Wald 31.6018 2 <.0001 Analysis of Maximum Likelihood Estimates Standard Wald Parameter DF Estimate Error Chi-Square Pr > ChiSq Intercept 1 44.1517 9.7776 20.3906 <.0001 LP1 1 -7.5684 2.0051 14.2481 0.0002

LP1*HP1 1 -0.3543 0.1438 6.0682 0.0138 Association of Predicted Probabilities and Observed Responses Percent Concordant 95.6 Somers' D 0.912 Percent Discordant 4.4 Gamma 0.912 Percent Tied 0.0 Tau-a 0.082

Pairs 8925 c 0.956

Hosmer and Lemeshow Goodness-of-Fit Test Chi-Square DF Pr > ChiSq 4.3520 8 0.8241

271

272

7. Other Asian

Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 7.4823 1 0.0062 Score 9.4976 1 0.0021

Wald 8.5978 1 0.0034 Analysis of Maximum Likelihood Estimates Standard Wald Parameter DF Estimate Error Chi-Square Pr > ChiSq Intercept 1 18.9860 5.3301 12.6880 0.0004

LP1 1 -3.1427 1.0718 8.5978 0.0034 Association of Predicted Probabilities and Observed Responses Percent Concordant 80.4 Somers' D 0.609 Percent Discordant 19.6 Gamma 0.609 Percent Tied 0.0 Tau-a 0.029 Pairs 4796 c 0.804

Hosmer and Lemeshow Goodness-of-Fit Test Chi-Square DF Pr > ChiSq

3.9726 8 0.8596

273

274

8. Honda (with second choice)

Testing Global Null Hypothesis: BETA=0 Test Chi-Square DF Pr > ChiSq Likelihood Ratio 36.1048 3 <.0001 Score 36.8158 3 <.0001

Wald 25.9694 3 <.0001 Analysis of Maximum Likelihood Estimates Standard Wald Parameter DF Estimate Error Chi-Square Pr > ChiSq Intercept 1 29.0188 6.8278 18.0633 <.0001 LP1 1 -9.7872 2.0611 22.5481 <.0001 HP1 1 1.3347 0.7269 3.3716 0.0663

LP2 1 3.8589 1.4017 7.5790 0.0059 Association of Predicted Probabilities and Observed Responses Percent Concordant 80.7 Somers' D 0.613 Percent Discordant 19.3 Gamma 0.613 Percent Tied 0.0 Tau-a 0.107

Pairs 7917 c 0.807

Hosmer and Lemeshow Goodness-of-Fit Test DF Pr > ChiSq 8 0.8938

275

276

C. Application for Ethics Approval of a Research Protocol

277

278

279

280

281

282

283

284

285

286

287

288

289

290

D. Application for Ethics Approval of a Research Protocol - Additional Notes

291

292

293

294

295

296

E. Human Research - Modifications/Additions to Approved Projects

297

298

299