Armenian Journal of Public Policy Volume 3, No. 1 March 2008
LOOKING FORWARD : THE GLOBAL COMPETITIVENESS OF THE ARMENIAN ECONOMY CONTENTS
David Joulfaian Guest Editor’s Note
Mher Baghramyan and Vahram Ghushchyan Implications of Armenian Dram Appreciation for the Competitiveness of Armenian IT, Tourism, and Food Processing Industries
Vahe Heboyan and Lewell F. Gunter Exchange Rate Dynamics in Armenia
Era Dabla-Norris and Holger Floerkemeier Bank Efficiency, Market Structure, and Foreign Ownership: What Determines Banking Spreads in Armenia?
Karen Grigorian and Vahram Stepanyan Productivity and Sources of Enterprise Level Efficiency in Armenia
Manuk Hergnyan, Gagik Gabrielyan, and Anna Makaryan Competitiveness of the Armenian Private Sector: Moving to the Next Stage
Kenneth Fortson, Ester Hakobyan, Anahit Petrosyan, Anu Rangarajan, and Rebecca Tunstall Armenia’s Millennium Challenge Account: Assessing Impacts on Economic Growth and Poverty Reduction in Rural Armenia
A journal of the Armenian International Policy Research Group
Armenian Journal of Public Policy A journal of the Armenian International Policy Research Group
Guest Editor
David Joulfaian US Department of Treasury
Editorial Board
Richard Beilock University of Florida Ara Khanjian Ventura College Arman Grigorian Swiss Federal Institute of Technology Garbis Iradian International Institute of Finance Lucig Danielian American University of Armenia Armine Khachatryan International Monetary Fund
Managing Editor
Heghine Manasyan CRRC Armenia/ Eurasia Partnership Foundation
Assistant to Managing Editor
Artak Galyan Armenian International Policy Research Group Manuscripts, book reviews, and communication with editors should be sent electronically to [email protected]. Published articles will be available online at www.aiprg.net. The views expressed in the Journal are not to be construed as those of the editors or the publisher, the Armenian International Policy Research Group. The financial support of Eurasia Partnership Foundation in Armenia inpublishing this issue is gratefully acknowledged. Copyright © 2008 AIPRG Printed by TOROS Publishing , Yerevan, Armenia.
Armenian Journal of Public Policy Volume 3, No. 1 March 2008
LOOKING FORWARD :
THE GLOBAL COMPETITIVENESS OF THE ARMENIAN ECONOMY
CONTENTS
David Joulfaian Guest Editor’s Note………………………………………………………………………... Mher Baghramyan and Vahram Ghushchyan Implications of Armenian Dram Appreciation for the Competitiveness of Armenian IT, Tourism, and Food Processing Industries………………………………….. Vahe Heboyan and Lewell F. Gunter Exchange Rate Dynamics in Armenia…………………………………………………….. Era Dabla Norris and Holger Floerkemeier Bank Efficiency, Market Structure, and Foreign Ownership: What Determines Banking Spreads in Armenia?...... Karen Grigorian and Vahram Stepanyan Productivity and Sources of Enterprise Level Efficiency in Armenia…………………….. Manuk Hergnyan, Gagik Gabrielyan, and Anna Makaryan Competitiveness of the Armenian Private Sector: Moving to the Next Stage…………….. Kenneth Fortson, Ester Hakobyan, Anahit Petrosyan, Anu Rangarajan, and Rebecca Tunstall Armenia’s Millennium Challenge Account: Assessing Impacts on Economic Growth and Poverty Reduction in Rural Armenia………………………………………………….
Armenian Journal of Public Policy Volume 3, No. 1 March 2008
GUEST EDITOR’S NOTE The Armenian economy experienced rapid growth over the past decade, in particular growing at double digit rates since 2001. At the same time the country experienced a number of shocks that may have implications for the country’s competitiveness and sustainability of continued growth. These include the rapid appreciation of the Dram, low financial intermediation, and stalled reforms among others. This timely volume contains six papers that address various aspects of the competitiveness of the Armenian economy. These include the effects of the appreciation of the exchange rate, competition in the banking sector, productivity at the firm level, the overall competitiveness of the private sector, as well as the agricultural sector and the impact of farmer training. The first paper by Mher Baghramyan and Vahram Ghushchyan addresses the appreciation of the Dram and its implications for the competitiveness of the IT, tourism, and food processing sectors. The authors explore the relationship among exchange rates, technical efficiency, and export and profitability of firms operating in Armenia. They find that exchange rate appreciation has serious implications for firm profitability. The appreciation of the Dram also raises the question of whether the exchange rate is misaligned. The second paper by Vahé Heboyan and Lewell Gunter explores the underlying factors that explain the exchange rate dynamics in Armenia. Their results show that the Armenian currency is misaligned from its long run equilibrium path. However, they also report that the degree of misalignment is sensitive to the set of variables used in the estimation and call for further work. Turning to the banking sector, the third paper, co authored by Era Dabla Norris and Holger Floerkemeier, studies the determinants of banking interest rate spreads and margins in Armenia. They find that a large proportion of the variation in spreads and margins can be explained by bank size, liquidity, and market power, as well as the market structure within which banks operate. The results suggest that there is a large potential to increase cost efficiency and competition in the banking system.
The fourth paper by Karen Grigorian and Vahram Stepanyan examines firm level evidence on the sources of productivity growth in the Armenian economy. They estimate total factor productivity (TFP) using a sample of firms in the manufacturing and service sectors, and investigate its variation across industries, time and region. They find that technical efficiency is highest in the jewelry, furniture and juice sector as well as mining, and lowest in chemicals, textiles and construction; the service sector operates at the mean of technical efficiency across the economy. However, the authors conclude that Armenia is not experiencing any growth in technical efficiency. Moving to the big picture, the next paper by Manuk Hergnyan, Gagik Gabrielyan, and Anna Makaryan examines the competitiveness of the Armenian private sector. The paper reviews Armenia’s economic achievements and studies the underlying factors and causal links to the country’s competitiveness. It also advances a set of recommendations to position Armenia in the region and globally. The last paper, co authored by the team of Kenneth Fortson, Ester Hakobyan, Anahit Petrosyan, Anu Rangarajan, and Rebecca Tunstall, reviews the major features of the Millennium Challenge Account program and lays out the methodology on how to study the impact of farmer training. The results of the evaluation will assess the farmer training program’s success in increasing the adoption of effective agricultural practices, increasing cultivation of higher value crops, improving farm productivity, increasing agricultural profits and household income, and reducing poverty rates. Guest Editor David Joulfaian
IMPLICATIONS OF ARMENIAN DRAM APPRECIATION FOR THE COMPETITIVENESS OF ARMENIAN IT, TOURISM , AND FOOD PROCESSING INDUSTRIES Mher Baghramyan, AIPRG Vahram Ghushchyan, Ph.D., AIPRG ∗ Abstract: The Armenian currency appreciated more than 40 percent during 2003 2006. This sharp change in nominal exchange rate is considered a negative shock for local producers and especially for the exporters. The survey data of fifty eight Armenian companies is used to study how the appreciation has affected the competitiveness of Armenian tourism, IT, and food processing industries. We use the Stochastic Frontier modeling technique to estimate the level of and changes in the technical efficiency of the companies during 2003 2006. The technical efficiency parameters are then included into the regression model in order to reveal the possible impact of the currency appreciation on profits and export levels of the companies. We find systematic and statistically significant impacts of exchange rate changes on the level of technical efficiency of the companies. We also find that work experience is another important determinant of degree of technical efficiency. We study the relationships among exchange rates, technical efficiency, and export and profitability of the companies. We find that a one point appreciation of the nominal exchange rate causes a decrease in the export of an average Armenian IT company by 66 thousand drams (about 200 USD) per year, average food processing company by 12 thousand Drams (about 40 USD) per year, and a loss of profit of an average incoming tour operator and hotel by 112 thousand AMD (or about 340 USD per year). JEL Classification: C1, C3, D2 Keywords: Technical Efficiency, Exchange Rate, Appreciation, Competitiveness
∗ The authors wish to thank Competitive Armenian Private Project (CAPS), USAID, and Norwegian Institute of International Affairs (NUPI) for financial support, AIPRG staff for their dedicated efforts on conducting the survey, and all survey respondents for providing data and making useful comments. The analysis and views presented in this study do not necessarily reflect those of AIPRG or USAID and are those of the authors alone.
I. INDUSTRY OVERVIEW
1.1 IT and Tourism Both Armenian information and communication (ICT) and tourism industries experienced rapid growth during the last decade and are considered to be among the most dynamic and perspective sectors of Armenian economy. According to Enterprise Incubator Foundation (EIF, 2007) and the Ministry of Trade and Economic Development 1 of Armenia, the average annual growth rate of ICT industry was about 30%. 165 companies operating in the industry employ about 5000 people. The industry output in 2006 was about 85 million USD comprising about 2% of GDP of Armenia, and about 63% of the output was exported. The share of tourism in the GDP of Armenia is about 6 7%, being at the same time one of the main export categories. 2 Table 1 .1 shows the dynamics of international tourist arrivals for 2001 2006. The industry maintains high annual growth rates (on average 25%) and has become one of the most important and dynamic sectors of the Armenian economy. It is estimated that, on average, one foreign tourist spends in Armenia about 10 days and about $1,600 USD, not counting international travel expenses. 3
Table 1.1 International Tourist Arrivals to Armenia, 2001-2006 2001 2002 2003 2004 2005 2006 International tourist arrivals, thousand 123,262 162,089 206,094 262,685 318,563 381,136 Annual growth rate of international 31.5 27.2 27.5 21.3 19.6 tourist arrivals, percentage Source: Ministry of Trade and Economic Development of Armenia. Theoretically, domestic currency appreciation can have extremely negative effect both for tourism and IT industries. Along with appreciation, domestic prices (when denominated in foreign currency) become more expensive for foreign visitors and their number may decrease in favor of alternative cheaper destinations. According to the recent study of ECA International, 4 Yerevan is ranked 21 st among most expensive cities for visitors and is ahead of Paris (23 rd ), Vienna (25 th ), Berlin (27 th ), and even Manhattan, NY (28 th ). This position of the Armenian capital is, in part, due to the domestic currency appreciation. Most of Armenian IT companies either export their products or operate as outsourcing contractors. Most of their costs are denominated in Armenian drams, labor being the largest cost category, which makes cost cutting almost impossible. In the competitive market, the entire burden of dram appreciation can be offset only by an increase of dollar price, if that company operates at the possible highest efficiency level. On another side, the real dram wages in Armenian IT sector are driven up by a deficit of properly skilled labor which makes Armenian IT companies even less competitive in the international market.
1 www.minted.am 2 ibid 3 ibid 4 http://www.eca international.com/ASP/ViewArticle2.asp?ArticleID=175
1.2 Food Processing The food processing industry is traditionally one of the important sectors of the Armenian economy. In the mid 80s, the sector accounted for about 18 percent of total industrial output. Armenia was always famous for its brandy and wine, canned fruits and vegetables, traditional meat products, fresh and dried fruits, etc. After the collapse of the FSU and the hard transition process, worsened by the economic blockade and war, the food processing industry experienced a dramatic decline. Many companies stopped operating, and others tried to survive, utilizing just 5 10 percent of their capacity. After the privatization in 1994 1999, the industry started reviving, benefitting from large volumes of foreign investments (about $60 million USD by 2000 (Decay, 2000)) and increased domestic demand driven by import substitution. Table 1.2 summarizes the main indicators of the food industry of Armenia during 1985 2006. Table 1.2 Main Indicators of Food Industry of Armenia 1985 1997 2001 2002 2003 2004 2005 2006 Number of enterprises, 135 178 593 670 769 786 782 n/a units Volume of production, 93.7 114.9 126.7 150.8 161.9 185.4 189.4 bln. Drams Volume of production, n/a 190,913 206,990 220,963 260,539 303,468 405,069 455,288 current USD, 000s Share in total Industrial Output, 18.4 36.8 37.1 37.1 36.6 31.3 29.5 29.4 percentage Number of industrial production personnel, 33.5 15.2 12.1 10.8 11.6 11.6 14.3 n/a persons, 000s* Food Export, current n/a 25,328 50,538 59,212 81,187 82,877 114,112 121,846 USD, 000s Food Export, share in Merchandise Export, n/a 10.9 14.8 11.7 11.8 11.5 11.7 12.4 percentage Food Import, current n/a 277,979 212,405 199,796 223,803 282,659 315,940 343,492 USD, 000s Food Import, shar e in Merchandise Import, n/a 31.2 24.2 20.2 17.5 20.9 17.5 15.7 percentage Food trade balance, 252,651 161,867 140,583 142,616 199,783 201,827 221,646 current USD, 000s Nominal Exchange Rate, (annual average), 490.8 555.1 573.4 578.8 533.5 457.7 416 drams per USD Change in Nominal Exchange Rate, YOY, 3.3% 0.9% 7.8% 14.2% 9.1% percentage *Without small and super small organizations. Source: Statistical Yearbook of Armenia 2006, NSS; “Industry” Statistical Collection, NSS, 1997; Authors’ calculations.
During the 2000s, the share of the food processing industry has further grown, providing about 30 percent of total industrial output in 2005. At the same time, it became the third largest exporting industry, accounting for about 12 percent of total merchandise exports in 2005. The period of 1997 2005 was characterized by substantial import substitution growth. While domestic food production has almost doubled, food imports have increased by just about 14 percent compared to 1997, and the share of foods in total merchandise import decreased from 31 to 18 percent. However, in 2005, the trade balance for foods was still negative representing about 200 mln USD or about 180 percent of the same year food export. Armenia is still highly dependent on imported food products which is about half of the total food consumption. Despite the increase of the number of enterprises operating within the industry, the level of output significantly decreased compared to the pre transition level (see Annex A). In 2006, the volume of production of almost all food products was significantly lower than in the pre transition period, with the only exception being whole milk dairy products. While for some product groups the difference is modest (alcohol free beverages – 86 percent of 1985 level, meat–75, and brandy 77percent), for other products it is striking (grape wine – about 6 percent, sausages 7, and canned products – about 5 percent). II. EXCHANGE RATE : REAL OR IMAGINARY THREAT ? The Armenian national currency – the Dram – was introduced in November 1993 at the rate of 200 Soviet Roubles per Dram. The ensuing few years were characterized by high rates of inflation and currency depreciation. In 1996 the Central Bank of Armenia (CBA) adopted the floating exchange rate regime and announced low inflation as the main target of Bank’s policy. By the end of 1990’s the Government was able to achieve macroeconomic stabilization and the economy started growing at high rates. Table 2.1 AMD/USD Exchange Rate Dynamics, 1997-2007 Oct. 1997 2001 2002 2003 2004 2005 2006 2007* Nominal Exchange Rate, annual 490.8 555.1 573.4 578.8 533.5 457.7 416 331 average, drams per USD Change in Nominal Exchange 3.3 0.9 7.8 14.2 9.1 20.4 Rate, YOY, percentage Source: Statistical Yearbook of Armenia 2006, NSS Note:*Exchange rate as of 19/10/2007 Starting from 2004, the Armenian currency has been experiencing dramatic appreciation (see Table 2.1). The most common official explanation of that phenomenon are high inflows of remittances from abroad, possible undervalued position of the real exchange rate prior 2003, rapid growth in income and productivity, as well as a process of de dollarization of the economy following new banking and legal regulations and the depreciation of dollar with respect to other major world currencies (Euro, Yuan, Yen, etc). Despite high pressure, Central Bank of Armenia continues to follow its policy of prioritizing low inflation rate rather than supporting stable exchange rate. The common reaction of Armenian officials to the complaints of Armenian producers and exporters regarding the negative impact of the appreciation has been advice to increase the productivity. 5 However, few appear concerned with the feasibility of such productivity growth. Moreover, according to a recent study by the World Bank (World Bank, 2007) Armenia has achieved significant improvement in the labor productivity. It is quite possible that many Armenian enterprises that already have modernized their technology and have applied effective management and quality control systems, have already completed the catch up process in productivity, and sustaining further productivity growth might not be feasible for them. This study will attempt to estimate the degree of the technical efficiency of Armenian food processing, Information and Communication Technology companies (ICT), hotels, and incoming tour operators and check how the dram appreciation has affected their performance and competitiveness. 2.1 What do Managers Think? In order to supplement the findings of the empirical analysis and gain better insight into the situation, we conducted a series of brief interviews with CEO’s of a number of companies from all the industries. The answers on 6 questions (Q) formulated below are summarized in the Table 2.2.
Table 2.2 Mean results of the Interview answers Sector Number of companies Q1 Q2 Q3 Q4 Q5 Q6 IT companies 38 35 503 48 23 77 4 Hotels 21 20 523 30 1 34 5 Tour Operators 33 53 460 33 11 76 3 Food Processing companies 32 47 420 26 2 55 5 Weighted average of all companies 124 40 474 35 11 64 4 Q1. What would you expect the percentage difference of Company’s 2006 revenue to be, if the exchange rate remained at the level of 2003, i.e. 580 drams per 1 USD? On average, the managers claimed that their revenues would be about 40 percent higher if the exchange rate were unchanged, the largest impact being for incoming tour operators – more than 50 percent. IT managers mentioned that the appreciation of Dram makes their companies less competitive compared to other outsourcing countries, and at least 5 firms branches of foreign companies, mentioned that their headquarters had been seriously considering the option of shutting down their Armenian office. For hotels and tour operators, the major problem is the continued appreciation of Armenian tourism service which causes Armenia losing the battle in the global tourism market. They argue that it is the flow of tourists of Armenian origin that allows the industry to survive, however this growth is about to slow down. Food companies selling in domestic markets are concerned with the loss of sales because most of the population is surviving with the flow of remittances, and the
5 See for example, Vahagn Grigoryan, “The Future of Export” (in Armenian), ww.cba.am/verluc/2.9.pdf (last visited on May 10, 2008). appreciation of dram is affecting the welfare of remittance recipients and thus the demand for food products. Q2. What AMD/USD exchange rate would be the most favorable for Your Company and would make it competitive? The desirable exchange rate is at the average 474 drams per US dollar which is about 47 percent higher than the rate of 335 drams prevailed during the study. Almost all respondents mentioned that they are concerned not only with the real appreciation of the currency but also with unpredictability of the exchange rates. Sometimes the uncertainty creates even more problems when they need to make price re calculations, sign agreements, print and distribute booklets, catalogues, etc. Q3. and Q4. What is the percentage change of your Company’s export prices (in USD) and domestic prices (in AMD) compared to 2003? Both export and domestic prices have increased during the last 4 years. However, while domestic prices increased at average 11 percent (which is consistent with the inflation in the country for the same period), the export prices have grown at about 35 percent, which in its turn is close to the appreciation rate for the same period. Evidently the companies raise their export prices in order to offset the exchange rate effect, which makes them less competitive in the international market. Q5. What percentage of your Company’s capital assets and human resources is being used (rate of utilization), on average, during the year? The answers to this question reveal serious efficiency problems. The most striking, in this regard, is with respect to hotels, with their average occupancy rate of about 35 percent. While the tourism industry has a seasonal nature and 76 percent of workload can be justified, for the IT industry the rate of 77 percent should be an issue of concern. The food companies operate on average at 50% or their capacity. Q6. Please, evaluate State – Your Company interrelations according to 0 10 point system (0 extremely unfavorable, 10 the most favorable). The average ranking of this answer was 4 points. The main reasons for such low evaluations of the relationship with the State are tax and customs administration problems, corruption, lack of business assistance programs, the State’s inability to improve the quality of education, etc. The predominant opinion among managers is that the process of appreciation is irreversible, regardless the causes of this phenomena. However, they insist that if the CBA follows such a monetary policy, the government should take adequate measures to help domestic companies to survive in this unfavorable environment. Despite the fact that both the IT and tourism industries are priority sector of economy for the Armenian Government, these industries have no privileges or advantages compared to other sectors of economy.
The most common problems identified by the respondents 6 and their suggestions are presented below: Tax and Customs Administration • Currently, all imported equipment is taxed with 20 percent VAT and customs duties are applied to the most of them. Usually, the customs value is determined by customs officers without any justification, regardless of the invoice value, and is based on internal instructions rather than market prices. It is suggested that these industries should be exempt from VAT tax and customs duties on imported equipment and on domestic investments into capital assets. • Decrease profit tax rate for exporters, from the current 20 to 10 percent; • Simplify the procedures and enforce the refund of overpaid VAT tax; • Provide temporary (1 2 year) tax exemption for newly established ICT companies. • Hotel Restaurant taxation . Many hotels complained that the hotel restaurants are taxed identically to other restaurants, tax being calculated on the square meter base. They claim that this approach is not acceptable since hotel restaurants serve only hotel customers and are marginally profitable while regular restaurants earn high profits on hosting different occasions (weddings, birthday parties, etc.), and the two categories cannot be treated identically. Some hotels, especially in the regions, were going to close their restaurants and stop providing breakfast to their customers. Finance • Create a mechanism for providing low interest, long term loans to exporters; • Improve the access to the long term loans for ICT companies. Education and training • A shortage of skilled ICT specialists is observed, and many companies consider this as one of the most important obstacles for further expansion of the industry; • A shortage of specialized managers (hotel managers, IT managers) is another serious problem; • Companies need the Government to provide free training or cover training costs of managers and other key employees; • The Government should cover the costs of participation in various international trade fairs, symposiums, networks, etc. Protection of intellectual property rights. It is well known that the protection of intellectual property rights is one of the most important factors stimulating the investment into R&D since the investors are sure that the potential benefits of new inventions or innovations will belong to them only. Many Armenian IT companies mentioned that as long as the protection of intellectual property right is not enforced,
6 The suggestions and policy recommendations in this chapter are those of respondents and may be different from the views of authors.
there will be no serious development of the ICT industry and little investment into R&D should be expected. De monopolization. This issue is considered especially important by ICT companies. Very high prices and low quality of internet and telephone services, accompanied with the non transparent system of state contracts on IT services, create significant negative spillovers and market distortions. The most important message of the companies’ officials was that the situation is very critical, and if the Government wants to preserve the emerging Armenian food processing, IT, and tourism industries, they should act as quickly as possible, otherwise even in one year it might be too late. III. METHODOLOGY AND MODEL SPECIFICATION 3.1 Stochastic Frontier Model The method used in our analysis is called Stochastic Frontier Model which we will use to estimate the degree of technical efficiency (TE) of Armenian companies. The obtained levels of technical efficiency, then, will be regressed on the exchange rate to see whether the appreciation of the Armenian currency has affected the technical efficiency and thus the competitiveness of the companies. We will assume that the surveyed companies have production functions with inputs X and output Y. In the perfectly efficient world, the i th firm in time t would produce the output Y it (1) Y = f (X ,β ) it ijt , th th and X ijt is the i firm’s j input at time t. However, as Farrell (Farrell, 1957) specifies, in real life two types of efficiencies exist: technical efficiency that allows firms to produce the maximum level of output given the level of inputs and allocative efficiency that requires production of given level of output as cheaply as possible. To understand the level of efficiency of the firm, we need to have the level of output of the absolutely efficient firm, which is known as production frontier, and then compare the output of the firm with the frontier. The stochastic models for estimating the production frontier and level of efficiency were introduced in 1977 by Aigner, Lovell and Schmidt (Aigner, 1977) and Meeusen and Van Den Broeck (Meeusen, 1977). In these models the efficiency is measured as ξi such that (2) Y = f (X ,β )ξ it ijt i , where ξi belongs to the interval within 0 and 1. A firm is perfectly efficient when ξi =1, in which case the firm’s actual production is at the highest possible level and is located on the production frontier (Figure 3.1). If ξi <1, the firm is not producing the maximum
output of the inputs X ijt , given the available technology reflected by the production function f (X ijt , β ) . Figure 3.1 Production Frontier and Technical Efficiency
Production frontier
Outputs . . . . . Actual production ......
Inputs In addition to inefficiency, each firm experiences also some exogenous shocks v it that are introduced into the model as stochastic error term.
(3) Yit = f (X ijt ,β )ξi exp(vit )
Taking logs of both sides and defining ui = −ln(ξi ) gives
(4) lnYit = ln( f (X ijt ,β )) − ui + vit Where vit and ut are independently and identically distributed, truncated at zero, with 2 mean and variance σ , and vit and ut are distributed independently of each other; cov( ui,v i)=0. Following Battese and Coelli (1992), we parameterize ut as (5) u = exp(−η(t −T ))u it i i , th where T i is the last time period in the i panel and η is the decay parameter. When η>0, the degree of inefficiency decreases over time, when η<0 , the degree of inefficiency increases over time. The last period i.e. when t=T i, contains the base level of inefficiency for the given firm. If η>0, the level of inefficiency decreases toward the base level, and if η<0 , the level of inefficiency increases to the base level. In our study, to estimate the production frontier and inefficiency terms of the companies, we will use two specifications of the production function: Translog production function (6) and Cobb Douglas production function (7).
1 1 1 lnY = β + β lnt + β ln K + β ln L + β (t)2 + β (ln K )2 + β (ln L )2 + (6) it 0 t K it L it 2 tt 2 KK it 2 LL it
+ β KL (ln Kit )(ln Lit ) + β Kt (ln Kit )(lnt) + β Lt (ln Lit )(lnt) − uit + vit and (7) lnY = β + β lnt + β lnK + β lnL −u + v it 0 t K it L it it it , where capital (K), labor (L) and time (t) are input factors used to estimate the stochastic frontier model, and Y is the output. We assume that both production functions have constant return to scale: RTS=1, which is tested for both model specifications. Technological Progress (TP) is the derivative of the production function with respect to time:
(8) TP = βt + βtt t)( + βKt (lnKit ) + βLt (lnLit ) If TP is positive (negative), then the production frontier shifts up (down). For the Cobb Douglas production function, TP is constant and is the coefficient of time βt . Change of Technical Efficiency (TE) is the derivative of the negative of the inefficiency measure with respect to time: du (9) TE = − it dt If TE is greater than zero, then the technical inefficiency declines over time and vice versa. IV. DATA The data used for this study was obtained during the survey of Armenian food processing companies, IT companies, tour operators and hotels conducted during June September, 2007. Initially, data of 50 companies from each sector of economy was intended to be studied; however, only 23 food processing companies, 15 incoming tour operators, 7 hotels, and 13 IT companies agreed to provide their firm level data (see Annex B for the summary statistics). The data for Revenue, Capital, Wages and other monetary variables is expressed in Armenian drams; it is adjusted for the inflation by using GDP Deflator of Armenia with base year of 1996 which was obtained from IMF World Economic Outlook Database. The data on domestic inflation and the exchange rate was obtained from the National Statistical Service of Armenia (NSS). We also calculated a measure of foreign inflation, which is the average inflation rate of ten main trade partners of Armenia weighted by their share in total Armenian trade. The data on trade was obtained from NSS, and the data on price levels from IMF World Economic Outlook Database. The data of Real Effective Exchange Rate was obtained from CBA of Armenia. V. EMPIRICAL RESULTS As the number of companies from each industry was insufficient for conducting the stochastic frontier analysis separately for each industry, we decided to group IT and tourism industries together and include an industry dummy variable for accounting for industry specific variation. The number of companies (23) was sufficient for analysis of the food processing industry. We calculated the change in technical efficiency based on two production functions. In Model 1 we used a Translog production function (see equation 6) and in the Model 2 we used a Cobb Douglas production function (7). In both models, the inputs are capital, labor and time. We assume that technical efficiency varies over time (time variant, equation 5) and has a truncated normal distribution. We also measured the technological progress (TP). In the Translog model, e Technological Progress (TP) is calculated using equation 8, and TE is calculated according to equation 9. In the Cobb Douglas model, TP is the coefficient of time and TE is calculated according to equation 9. In both models, the null hypothesis of Constant Return to Scale (CRS) is accepted based on the likelihood ratio test. Technical efficiency and technological progress for each firm, for every year were estimated. The mean TE and TP by year are presented in the Table 5.1 and Table 5.2. Table 5.1 Mean of Estimated Parameters: IT, Tour Operators, and Hotels, 2003-2006
Year te1 te2 tp1 tp2 2003 0.4369728 0.4716637 0.2318835 0.0768452 2004 0.4860817 0.5213609 0.0213009 0.0768452 2005 0.5035055 0.5472735 0.1761738 0.0768452 2006 0.4912892 0.5453566 0.2722712 0.0768452 Total 0.4834649 0.5263201 0.0892652 0.0768452 Note: te – technical efficiency, tp – technological progress ;. 1 and 2 refer to the Model 1 and Model 2 respectively. The positive sign of TE shows that in both cases the degree of technical inefficiency is decreasing over time. An interesting observation is that in both models, the technical efficiency is increasing for 2003 2005 but the rate is slowing down in 2006. Technological Progress (TP) indicates the direction of change of the production frontier. In the first model, starting from 2004, the frontier is shifting up. In the case of second model, the TP is constant and can be interpreted as an average progress during the last 4 years.
Table 5.2 Mean of Estimated Parameters: Food Processing, 2003-2006
year te1 tp1 2003 0.2632050 0.5384401 2004 0.2599808 0.1620835 2005 0.2679276 0.0995709 2006 0.2666353 0.2382905 Total 0.2646437 0.0580394 Note: te – technical efficiency, tp = technological progress. 1 and 2 refer to the Model 1 and Model 2 respectively. For the food processing industry, Model 2 did not provide economically meaningful results, so we drop the results. In Table 5.2 we see that the degree of TE is fluctuating during the specified period. Table 5.3 and Table 5.4 provide summary statistics of the estimated parameters of TE by industry. We can see that TE index for the food processing industry is the smallest, and within the tourism industry, hotels are about 15 20 percent less efficient than tour operators. In the case of IT companies, the mean TE is almost at the same level of about 51 percent in both models, however this industry has the largest spread in terms of technical efficiency (with largest standard deviation) showing that while some IT companies are successful in improving their efficiency, others are lagging far behind. Table 5.3 Summary Statistics of TE1 by Industry, Average 2003-2006 Industry Obs Mean Std. Dev. Min Max Hotel 22 0.4300364 0.2061671 0.2121916 0.8590048 IT 27 0.4987445 0.2722411 0.0623348 0.8645540 Tour Operators 22 0.5181412 0.2335382 0.0999397 0.8691981 Food Processing 73 0.2646437 0.2539802 0. 0324986 0.8200656 Table 5.4 Summary Statistics of TE2 by Industry, Average 2003-2006 Obs Mean Std. Dev. Min Max Hotel 22 0.5058211 0.1912658 0.2499393 0.8819743 IT 27 0.5112230 0.2806895 0.0629765 0.8730741 Tour Operators 22 0.5653473 0.2485631 0.10 85106 0.8742781 We want to estimate how the change in exchange rate affects the technical efficiency of the firms. In regressions, the calculated firm and time specific technical efficiency, TE1 and TE2, are dependant variables. Two measures of exchange rate are alternatively considered: i) nominal AMD/USD exchange rate ( exch ) together with domestic inflation rate ( infa ) and trade weighted foreign inflation rate ( inff ), and ii) Real Effective Exchange Rate 7 (reer ) calculated by CBA. The foreign inflation rate for each year is calculated using average of the inflation rates of the ten largest Armenian trade partners, weighted by share of the trade of each country in the total foreign trade of Armenia. 7 REER is a composite index that incorporates nominal exchange rate and price levels of both Armenia and its trade partners. For additional details on the methodology and calculations of REER refer to website of the Central Bank of Armenia: http://www.cba.am/publications/prog/annex.pdf (last visited on May 10, 2008).
Since we are trying to estimate the changes in the competitiveness of Armenian producers, in addition to macroeconomic variables we also include firm specific variables, such as marketing expenses in thousands of drams (adjusted by GDP Deflator) and average work experience of the employees expressed in years. Unfortunately, the data on trainings was not reliable and consistent to include into the model. The regression results or random effect models are presented in Annex C (IT and Tourism Industries) and Annex D (Food Processing Industry). The first model specification provides the following results for IT and Tourism Industries: (10) te1 = 0.3554323 + 0.0002701*exch*** – 0.0011503*infa – 0.0002147*inff + 0.0005063*exp** + 5.09e 07*marketr + 0.0279733*tour – 0.0501147*hotel Note:* significant at 10%; ** significant at 5%; *** significant at 1%. The regression results using TE2 as a measure of technical efficiency are identical: (11) te2 = 0.4780624 +0.0000438*exch*** – 0.0001404*infa + 0.0002289*inff + 0.0000968*exp** + 1.24e 07*marketr + 0.0578764*tour + 0.0000625*hotel Note:* significant at 10%; ** significant at 5%; *** significant at 1%. The regression results for food processing industry are the following: (12) te1 = 0. 2681958+ 0.0000164*exch*** + 0. 0000148*infa + 0. 0003993*inff + 0.000027*exp – 4.10e 09*marketr Note:* significant at 10%; ** significant at 5%; *** significant at 1%. From both specifications, for all three industries, we can see that the nominal exchange rate has statistically significant impact on the level of technical efficiency. The positive sign shows that the effect of the appreciation of Armenian currency for the technical efficiency and thus competitiveness of the companies is negative. The positive sign is robust to changes in the model specification and independent variable. For example, we run similar regressions by using consumer price index (CPIa and CPIf) instead of inflation rate as a measure of domestic and foreign price levels, and we found similar results. The coefficients for work experience are positive and significant at 5 percent significance level in (10) and (11). This means that work experience is one of the important determinants of technical efficiency in IT and Tourism industries. The coefficients of the domestic and foreign price levels, as well as marketing expenses are not significant and are sensitive to the model specifications. The industry dummies also are highly insignificant, meaning that the determinants of technical efficiency don’t differ between IT and tourism industries.
To better assess the situation, in the next two regressions three of the previously used variables ( exch , infa , and inff ) are substituted for by one variable – Real Effective Exchange Rate ( reer ). (13) te1 = 0.6885364 – 0.0021411*reer*** + 0.0006226*exp* + 7.91e 07*market + 0.0278303*tour – 0.0513877*hotel and (14) te2 = 0. 5328573 – 0.0003383*reer*** + 0.0001125*exp** + 1.65e 07*marketr + 0.0578545*tour – 0.0001084*hotel Note:* significant at 10%; ** significant at 5%; *** significant at 1%. And for Food Processing Industry, we have (15) te1 = 0.2899901 – 0.0001198*reer*** + 0. 0000339*exp – 7.36e 09*marketr The results are comparable with (10) – (12). Again, the exchange rate appreciation has negative and highly significant effects for the degree of technical efficiency, 8 and work experience has positive and significant coefficients for the IT and Tourism industries, but not for food companies. To continue our analysis, we use Tobit model in order to estimate the possible effect of the change in the degree of technical efficiency (te1 and te2) on the exports of IT and food processing companies (Annex E). Unfortunately, the data on number of foreign customers obtained from tour operators and hotels is not consistent and reliable, and we cannot conduct a similar analysis for tourism industry 9. We find that for IT industry (16) export = – 64415.22 + 244478* te1 (17) export = – 66121.35 + 243023.3* te2 and for Food Processing industry (18) export = – 147614.9 + 743663.2* te1 All coefficients of TE are significant at 1% significance level. The results of (16) suggest that, on average, 10 percent improvement in the degree of technical efficiency of a company brings about 24.5 million dram or 73.9 thousand
8 It is important to note that while the sign of reer is opposite to the sign of exch , the effect it similar. It is explained by the methodology of reer calculation: the appreciation means an increase of value of reer but decrease of value of exch . 9 If we had a reliable data on the number of foreign customers for each tour operator, we could use a similar model and regress the number of customers on the degree of technical efficiency to find out indirectly the total number of customers lost due to the exchange rate appreciation.
USD 10 of additional exports of IT products and 74 million (about 225 thousand USD) of additional exports of processed food. Now we can use our estimates for calculating the effect of each point of dram appreciation on exports. According to (10), appreciation of the nominal exchange rate by 1 dram is causing the technical efficiency of an average Armenian IT company to decrease by 0.0002701. On the other hand, from (13) we know that a decrease of TE by 10 percent will decrease the export of an average IT company by 24.5 million AMD. This means that a one point appreciation of the nominal exchange rate will cause the export of an average Armenian IT company to go down by 0.0002701* 244,478,000=66,034 drams which is equal about 200 USD (at the rate of 1USD=331AMD as of 19 October, 2007). If we want to estimate the total impact that the appreciation during a specified period had for the entire industry, we should use the following formula: (19) Loss in Export IT =66,034 * Number of companies * ∆ exchange rate Similarly, a one point appreciation of the nominal exchange rate will cause the export of an average Armenian food processing company to go down by 0.0000164 * 743,663,200=12,196 drams which is equal about 37 USD (at the rate of 1USD=331AMD), and (20) Loss in Export FOOD =12,196 *Number of companies * ∆ exchange rate When we compare IT and food industries, two important observations can be made. The elasticity of technical efficiency of IT companies with respect to nominal exchange rate is more than tenfold higher than in food processing industry which means that IT companies are more sensitive to the exchange rate appreciation. On the other hand, the change in technical efficiency has 3 time larger impact for the export level of food companies implying that the return on TE improvements is larger. Table 5.5 presents the estimated export losses of the Armenian IT industry starting from 2004. We found that starting from 2004 Armenian IT industry has lost export opportunities of about 6 million USD of value. In our survey, the total export of the surveyed IT companies was 1,095 million AMD during 2004 2006. For the same period, the nominal exchange rate has appreciated from 579 AMD/USD in 2004 to 416 AMD/USD in 2006. According to (19), the total exports loss of our 13 IT companies amounts to 140 million AMD or about 13 percent of total exports. Similarly, food industry has lost about 45 million AMD of export opportunities or about 3 percent of actual exports. Next, we estimate how the change in TE due to the dram appreciation has affected the profitability of the tourism and food processing industries. We use a random effect regression model (see Annex F). The results suggest that TE has a significant and positive effect for the profitability of tourism companies. On average, each point of dram appreciation causes an average tour operator and hotel to lose about 112 thousand AMD or about 340 USD and the average food processing company to lose just 14 USD of profit before tax. 10 At the rate of 331 AMD per 1 USD, as of 19 October, 2007
Table 5.5 Estimated Loss of Exports in IT industry due to dram appreciation, 2003-2006 Total, [95% Oct. 2003* 2004 2005 2006* Total Confidence 2007*** Interval] Number of Operating Companies* 110 125 141 160 165 IT Industry Revenue, mln. USD* 37.7 49.3 64.4 84.2 IT Industry average revenue, mln. 0.34 0.39 0.45 0.52 USD* Domestic market, mln. USD* 13.5 17.8 23.5 30.9 Exports, mln. USD* 24.2 31.5 41.0 53.3 69.3 Export loss, mln. AMD** 373 707 441 926 2,446 1,495 3,398 Export loss, mln. USD** 0.699 1.544 1.059 2.798 6.100 3.727 8.473 Ratio of Lost Export to the 2.2 3.8 2.0 4.0 actual Export, %** Nominal Exchange Rate, AMD per 578.8 533.5 457.7 416.0 331.0 USD, annual average, drams Source: * EIF 2007; ** Authors’ calculation. Note: ***Exchange rate as of 19/10/2007 From Table 5.6 we see that only in 2006, an average tourism company has lost about 9.5 million dram of profit before tax (about 29 thousand USD), while the total loss starting from 2004 has amounted to 28 million drams (68 thousand USD). Table 5.6 Estimated Average Loss of Profit per Tour Operator and Hotel Total [95% 2004 2005 2006 Oct. 2007 Total Confidence Interval] Profit loss per company , 5.092 8.521 4.687 9.555 27.855 2.487 53.223 mln. AMD Loss of profit, per 9,545 18,616 11,268 28,867 68,296 6,098 130,493 company, USD, 000s During the period of 2004 2006, the lost profit of all surveyed tour operators and hotels amounted to 401 million AMD or 15 percent of actual profit. For food companies, the profit loss was modest at slightly less than 1 percent. Our analysis strongly suggests that the IT and tourism industries, and to lesser extent food processing industry, have been seriously affected by dram appreciation, and since the appreciation process continues, urgent measures should be undertaken by the government for helping companies to offset this negative pressure and stay competitive in domestic and international markets.
VI. POLICY RECOMMENDATIONS AND CONCLUSION The survey data of 58 Armenian companies are used to study how the appreciation has affected the competitiveness of Armenian IT companies, hotels, tour operators and food processing industries. We use the Stochastic Frontier Modeling technique to estimate the level and changes in technical efficiency of Armenian companies for the period of 2003 06. The technical efficiency parameters are then included into the regression model in order to reveal the possible impact of currency appreciation on export volumes and profitability of the companies. The model shows that the level of technical efficiency of Armenian companies has been rapidly growing during the last 3 years, but reversed in 2006. We find a systematic and statistically significant negative relationship between dram appreciation and the degree of technical efficiency of the companies. We also find that technical efficiency is an important determinant of export levels. We estimate that starting from 2004 , the Armenian IT industry has lost about 6 million USD of export opportunities. We also found that each point of dram appreciation is causing an average IT company to lose about 66 thousand AMD (about 200 USD) of exports per year and the average food processing company to lose about 12 thousand AMD (37 USD) per year. We study the relationship between the degree of technical efficiency and profitability of Armenian tour operators and hotels. We find that each point of dram appreciation causes an average tour operator and hotel to lose about 112 thousand AMD or about 340 USD of profit before tax. The profit loss of food processing companies is negligible. We also find strong positive correlation between average work experience of the company’s employees and the degree of technical efficiency of that company. In the conditions of continued dram appreciation, the Armenian government and senior company managements should work together in seeking possible ways for overcoming the negative pressure created by the exchange rate appreciation. According to the model (equation 10), one of the company level determinants of technical efficiency is work experience, one year increase of average work experience of the company’s staff offsetting about 2 points of dram appreciation. If we consider trainings as a means of improving skills and adding experience, they can become a powerful tool for improving efficiency and productivity of the company. According to the current legislation, a firm cannot claim more than the equivalent of 1 percent of total revenue as training expense. This restriction should be removed as it will create incentives for the companies to spend more money on staff training. Also, more free training should be organized through state business assistance programs. Improving knowledge and making education better targeted is another challenge especially in IT sector. Many managers complained that new graduates have very poor skills and knowledge, and they have to spend a lot of resources to train and educate them. One of the recommendations is to create a link between educational institutions and employers in the area of curriculum development: before confirming a certain course, the curriculum should be reviewed and discussed with potential employers, and only after their approval the course should be taught in the college. It is well known, that 20 percent VAT tax on investments and on the import of capital assets (such as equipment, electronics, etc) creates an additional tax burden and affects the investment decision of the companies. Of course, it would be ideal if import of capital assets were exempt from VAT tax and customs duties. However, if it not possible at all, the government could consider adding ICT industry into the list of privileged companies that are allowed to pay the VAT by installment , according to the accepted depreciation schedule. For example, if a depreciation period of an imported
server is 3 years, the company could pay the calculated VAT tax during 3 year period, at 3 equal installments. All these and other policy recommendations should be implemented as part of a state private dialogue. The appreciation continues, and there is no doubt that it has created additional (sometimes almost disastrous) challenges for newly emerged Armenian economy, and all available intellectual, financial, and political resources should be mobilized to help Armenian companies overcome this situation.
Appendix Annex A. Output of selected products of Food Industry of Armenia 1985 1997 2001 2002 2003 2004 2005 2006 Meat, ton, 000s 70 32 38 38 41 43 47 53 Sausages, ton 26,200 40 1,108 1,044 998 841 1,053 1,775 Whole milk dairy products (in 177 251 197 207 218 279 299 313 milk equivalent), ton, 000s Cheese, ton 26,000 1,500 4,792 4,819 14,257 14,413 14,403 14,487 Animal butter, ton 390 11 13 29 48 44 105 n/a Vegetable oil, ton 6,792 279 262 1,559 2,204 385 289 2,735 Pasta products (macaroni), ton 14,000 400 67 5 1,085 1,196 2,334 2,634 2,981 Groats, ton n/a n/a 16 12 8 22 1,141 291 Confectionery, ton 40,000 1,500 3,085 3,507 3,969 3,964 4,836 7,454 Flour, ton, 000s 393 143 114 110 132 147 140 152 Bread and bakery products, ton, 312 373 299 294 294 295 295 295 000s Salt, ton, 000s n/a n/a 29 30 32 32 35 37 Canned products, thousands of 494,000 31,000 38,006 52,571 16,955 7,852 12,103 13,890 conditional cans/t * of which Meat n/a n/a 582 525 1,347 n/a Fish n/a n/a 323 266 226 144 87 n/a Vegetable n/a n/a 751 2,471 1,673 705 996 n/a Tomato 185,000 11,000 24,441 43,531 12,945 5,396 5,618 n/a Fruit 216,000 19,000 12,491 6,303 1,529 1,082 4,055 n/a of which Jam, Confiture n/a n/a 406 551 877 826 827 n/a Alcohol free beverages, liter, 44,830 16,360 27,434 26,817 33,183 36,223 31,981 38,409 000s Natural juices, liter, 000s n/a n/a 1,812 2,519 4,248 4,588 4,341 5,971 Cigarettes, 000000s 11,958 815 1,623 2,815 3,222 2,720 3,020 2,825 Mineral water, liter, 000s 147,500 13,000 20,157 18,286 19,542 19,929 24,115 27,240 Alcoholic beverages Beer, liter, 000s 60,370 5,040 9,975 7,078 7,312 8,834 10,751 12,618 Vodka, liter, 000s 15,970 5,920 9,456 10,335 10,122 12,878 13,596 12,801 Brandy (cognac), liter, 000s 11,690 3,920 5,026 6,060 7,217 7,333 9,135 9,060 Grape wine, liter, 000s 66,460 3,370 6,394 4,008 2,046 6,224 6,740 3,826 Champagne, liter, 000s 3,280 1,520 582 622 670 569 519 543 * Since 2003 production of canned products have been calculated in tons. Source: Statistical Yearbook of Armenia 2006, NSS; Social Economic Situation in RA January December, 2006; "Industry" Statistical Collection, NSS, 1997
Annex B. Mean of Key Variables Obtained during the Survey, 2003-2006
Average monthly Revenue, Profit, Capital Labor/ Average monthly wage of administ Year AMD, AMD, assets, employees, wage of productive rative workers, 000s 000s AMD, 000s person workers, AMD, 000s AMD, 000s Food processing companies 2003 413148 54779 163237 43 23 35 2004 494124 39997 189464 51 23 43 2005 466309 48603 202869 63 29 63 2006 559972 64752 173037 71 34 66 Tour Operators 2003 64419 14516 20701 10 45 51 2004 60514 11715 14324 11 52 53 2005 66387 13111 14187 13 63 68 2006 70807 10903 13230 13 72 76 Hotels 2003 316775 107529 413874 58 32 54 2004 402104 120367 514335 61 34 53 2005 316402 89705 644795 68 39 75 2006 316606 126420 1175143 71 46 104 ICT 2003 109591 27133 30523 48 58 128 2004 121564 16629 148931 53 114 106 2005 178755 22780 179223 59 135 148 2006 203816 23906 122773 51 137 158
Annex C. Output of Regression Analysis of TE determinants, IT and Tourism Industries
Model specification 1
Dependent variable: te1 – technical efficiency of IT and Tourism industries obtained from Translog Production Function Independent variable: Nominal exchange rate AMD/USD, domestic inflation rate, weighted foreign inflation rate, average work experience, marketing expenses, dummy variables for tour operators and hotels. Random effects GLS regression Number of obs = 58 Group variable (i): id Number of groups = 17 R sq: within = 0.9679 Obs per group: min = 1 between = 0.0001 avg = 3.4 overall = 0.0143 max = 4 Random effects u_i ~ Gaussian Wald chi2(7) = 1058.48 Corr (u_i, X)= 0 (assumed) Prob > chi2 = 0.0000 te1 Coef. Std. Err. z P>|z| [95% Conf. Interval] exch .0002701 .0000346 7.81 0.000 .0002024 .0003379 infa .0011503 .001325 0.87 0.385 .0037473 .0014466 inff .0002147 .0077871 0.03 0.978 .0154773 .0150478 exp .0005063 .0002398 2.11 0.035 .0000362 .0009763 marketr .000000509 .000000454 1.12 0.263 .000000381 .0000014 tour .0279733 .1471016 0.19 0.849 .2603406 .3162872 hotel .0501147 .1657391 0.30 0.762 .3749573 .2747279 _cons .3554323 .1186831 2.99 0.003 .1228178 .5880469 sigma_u .2609599 sigma_e .00330753 rho| .99983938 (fraction of variance due to u_i) Model specification 2
Dependent variable: te2 – technical efficiency of IT and Tourism industries obtained from Cobb Douglas Production Function. Independent variable: Nominal bilateral exchange rate AMD/USD, domestic inflation rate, weighted foreign inflation rate, average work experience, marketing expenses, dummy variables for tour operators and hotels. Random effects GLS regression Number of obs = 58 Group variable (i): id Number of groups = 17 R sq: within = 0.9534 Obs per group: min = 1 between = 0.0086 Avg = 3.4 overall = 0.0200 Max = 4 Random effects u_i ~ Gaussian Wald chi2(7) = 742.93 corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 te2 Coef. Std. Err. z P>|z| [95% Conf.Interval] exch .0000438 6.53e 06 6.70 0.000 .000031 .0000566
infa .0001404 .0002504 0.56 0.575 .0006312 .0003504 inff .0002289 .0014718 0.16 0.876 .0026557 .0031136 exp .0000968 .0000453 2.14 0.033 .00000795 .0001857 marketr .0000001240 .0000000858 1.44 0.149 .0000000444 .0000002920 tour .0578764 .1562754 0.37 0.711 .2484177 .3641705 Hotel .0000625 .1760605 0.00 1.000 .3450098 .3451349 _cons .4780624 .1068555 4.47 0.000 .2686295 .6874954 sigma_u .28207267 sigma_e .000636 rho .99999492 (fraction of variance due to u_i) Model specification 3
Dependent variable: te1 – technical efficiency of IT and Tourism industries obtained from Translog Production Function. Independent variable: Real Effective Exchange Rate (REER), average work experience, marketing expenses, dummy variables for tour operators and hotels.
Random effects GLS regression Number of obs = 58 Group variable (i): id Number of groups = 17 R sq: within = 0.9346 Obs per group: min = 1 Between = 0.0000 Avg = 3.4 Overall = 0.0151 Max = 4 Random effects u_i ~ Gaussian Wald chi2(5) = 500.10 corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 te1 Coef. Std.Err. z P>|z| [95% Conf. Interval] reer .0021411 .0001045 20.50 0.000 .0023458 .0019363 exp .0006226 .000329 1.89 0.058 .0000222 .0012674 marketr .000000791 .000000636 1.24 0.213 .000000455 .00000204 tour .0278303 .137765 0.20 0.840 .2421841 .2978447 hotel .0513877 .1552333 0.33 0.741 .3556394 .252864 _cons .6885364 .094103 07.32 0.000 .504098 .8729748 sigma_u .23759577 sigma_e .00459649 rho .99962588 (fraction of variance due to u_i)
Model specification 4
Dependent variable: te2 – technical efficiency of IT and Tourism industries obtained from Cobb Douglas Production Function. Independent variable: Real Effective Exchange Rate (REER), average work experience, marketing expenses, dummy variables for tour operators and hotels. Random effects GLS regression Number of obs = 58 Group variable (i): id Number of groups = 17 R sq: within = 0.9220 Obs per group: min = 1
between = 0.0087 Avg = 3.4 overall = 0.0202 Max = 4 Random effects u_i ~ Gaussian Wald chi2(5) = 430.25 corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 te2 Coef. Std. Err. z P>|z| [95% Conf. Interval] reer .0003383 .0000179 18.95 0.000 .0003733 .0003033 exp .0001125 .0000563 2.00 0.046 2.24e 06 .0002228 marketr 1.65e 07 1.09e 07 1.52 0.129 4.81e 08 3.78e 07 tour .0578545 .1462734 0.40 0.692 .2288361 .3445451 hotel .0001084 .1647926 0.00 0.999 .3230959 .3228791 _cons .5328573 .0993873 5.36 0.000 .3380618 .7276527 sigma_u .25720413 sigma_e .00080072 rho .99999031 (fraction of variance due to u_i)
Annex D. Output of Regression Analysis of TE determinants, Food Processing Industry
Model specification 1
Dependent variable: te1 – technical efficiency of Food Processing industry obtained from Translog Production Function. Independent variable: Nominal exchange rate AMD/USD, domestic inflation rate, weighted foreign inflation rate, average work experience, marketing expenses. Random effects GLS regression Number of obs = 61 Group variable (i): id Number of groups = 18 R sq: within = 0.9344 Obs per group: min = 2 between = 0.0434 Avg = 3.4 overall = 0.0044 Max = 4 Random effects u_i ~ Gaussian Wald chi2(5) = 564.24 corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 te1 Coef. Std. Err. z P>|z| [95% Conf. Interval] exch .0000164 .00000244 6.70 0.000 .0000116 .0000211 infa .0000148 .0000937 0.16 0.875 .0001688 .0001984 inff .0003993 .0005416 0.74 0.461 .0006623 .0014609 exp .000027 .0000199 1.36 0.175 .000012 .000066 marketr .0000000041 .0000000052 0.79 0.431 .0000000143 .0000000061 _cons .2681958 .0628235 4.27 0.000 .1450639 .3913276 sigma_u .27152788 sigma_e .00025974 rho .99999908 (fraction of variance due to u_i) Model specification 2
Dependent variable: te1 – technical efficiency of Food Processing industry obtained from Translog Production Function. Independent variable: Real Effective Exchange Rate (REER), average work experience, marketing expenses. Random effects GLS regression Number of obs = 61 Group variable (i): id Number of groups = 18 R sq: within = 0.9121 Obs per group: min = 2 between = 0.0581 Avg = 3.4 overall = 0.0074 Max = 4 Random effects u_i ~ Gaussian Wald chi2(3) = 425.23 corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 te1 Coef. Std. Err. z P>|z| [95% Conf. Interval] reer1 .0001198 .00000705 16.99 0.000 .0001337 .000106 exp .0000339 .0000225 1.51 0.131 .0000101 .0000779 marketr .00000000736 .00000000582 1.27 0.206 .0000000188 .00000000404 _cons .2899901 .0613354 4.73 0.000 .1697749 .4102054 sigma_u .26330417
sigma_e .0002929 rho .99999876 (fraction of variance due to u_i)
Annex E. Output of Tobit models
IT Industry
1. Dependent variable: te1 – technical efficiency of IT and Tourism industries obtained from Translog Production Function Obtaining starting values for full model: Iteration 0: log likelihood = 339.19909 Iteration 1: log likelihoo = 338.09634 Iteration 2: log likelihood = 338.02567 Iteration 3: log likelihood = 338.02453 Iteration 4: log likelihood = 338.02453 Fitting full model: Iteration 0: log likelihood = 209.42266 Iteration 1: log likelihood = 209.15759 Iteration 2: log likelihood = 208.49901 Iteration 3: log likelihood = 208.41576 Iteration 4: log likelihood = 208.41301 Iteration 5: log likelihood = 208.413 Random effects tobit regression Number of obs = 27 Group variable (i): id Number of groups = 8 Random effects u_i ~ Gaussian Obs per group: min = 2 Avg = 3.4 Max = 4 Wald chi2(1) = 25.39 Log likelihood= 208.413 Prob > chi2 = 0.0000 export Coef. Std. Err. z P>|z| [95% Conf. Interval] te1 244478 48519.54 5.04 0.000 149381.4 339574.5 _cons 64415.22 28943.82 2.23 0.026 121144.1 7686.377 /sigma_u 102806.7 9953.772 10.33 0.000 83297.65 122315.7 /sigma_e 47056.11 8959.238 5.25 0.000 29496.33 64615.89 rho .8267862 .0559387 .6961766 .9146014 Observation summary: 16 uncensored observations 11 left censored observations 0 right censored observations
2. Dependent variable: te2 – technical efficiency of IT and Tourism industries obtained from Cobb Douglas Production Function Obtaining starting values for full model: Iteration 0: log likelihood = 339.15012 Iteration 1: log likelihood = 338.03716 Iteration 2: log likelihood = 337.95508 Iteration 3: log likelihood = 337.9536 Iteration 4: log likelihood = 337.9536 Fitting full model: Iteration 0: log likelihood = 209.21367 Iteration 1: log likelihood = 208.77 Iteration 2: log likelihood = 208.39435 Iteration 3: log likelihood = 208.35855 Iteration 4: log likelihood = 208.35791 Iteration 5: log likelihood = 208.35791 Random effects tobit regression Number of obs = 27 Group variable (i): id Number of groups = 8 Random effects u_i ~ Gaussian Obs per group: min = 2 Avg = 3.4 Max = 4 Wald chi2(1) = 26.37 Log likelihood = 208.35791 Prob > chi2 = 0.0000 export Coef. Std. rr. z P>|z| [95% Conf. Interval] te2 243023.3 47327.75 5.13 0.000 150262.6 335784 _cons 66121.35 29029.59 2.28 0.023 123018.3 9224.399 /sigma_u 102511.1 9812.481 10.45 0.000 83278.99 121743.2 /sigma_e 46441.68 8862.952 5.24 0.000 29070.61 63812.75 rho .8297063 .0553546 .7001194 .9164015 Observation summary: 16 uncensored observations 11 left censored observations 0 right censored observations
Food Processing Industry Dependent variable: te1 – technical efficiency of Food Processing industry obtained from Translog Production Function Obtaining starting values for full model: Iteration 0: log likelihood = 924.25177 Iteration 1: log likelihood = 924.19565 Iteration 2: log likelihood = 924.19544 Fitting full model: Iteration 0: log likelihood = 690.76796 Iteration 1: log likelihood = 687.0899 Iteration 2: log likelihood = 686.42755 Iteration 3: log likelihood = 686.37083 Iteration 4: log likelihood = 686.36997 Iteration 5: log likelihood = 686.36997 Random effects tobit regression Number of obs = 69 Group variable (i): id Number of groups = 20 Random effects u_i ~ Gaussian Obs per group: min = 2 Avg = 3.5 Max = 4 Wald chi2(1) = 20.51 Log likelihood = 686.36997 Prob > chi2 = 0.0000 export Coef. Std.Err. z P>|z| [95% Conf. Interval] te1 743663.2 164192.9 4.53 0.000 421851 1065475 _cons 147614.9 68387.89 2.16 0.031 281652.7 13577.11 /sigma_u 156953.4 41315.08 3.80 0.000 75977.33 237929.5 /sigma_e 153072.1 17413.61 8.79 0.000 118942 187202.1 rho .5125175 .14852 .242388 .7767924 Observation summary: 50 uncensored observations 19 left censored observations 0 right censored observations
Annex F. Output of Regression model
Tourism Industry
Dependent variable: prof – reported accounting profit of tour operators and hotels Independent variable: te1 – technical efficiency obtained from Translog Production Function, hotel dummy variables for hotels. Random effects GLS regression Number of obs = 41 Group variable (i): id Number of groups = 11 R sq: within = 0.0030 Obs per group: min = 2 Between = 0.4878 Avg = 3.7 Overall = 0.2743 Max = 4 Random effects u_i ~ Gaussian Wald chi2(2) = 6.41 corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0406 prof Coef. Std. Err. z P>|z| [95% Conf. Interval] te1 416177.7 193381.3 2.15 0.031 37157.25 795198.2 hotel 164544.4 85389.84 1.93 0.054 2816.643 331905.4 _cons 232723.9 129731.3 1.79 0.073 486992.5 21544.74 /sigma_u 128448.33 /sigma_e 80894.634 rho .7160107 (fraction of variance due to u_i) Dependent variable: prof – reported accounting profit of tour operators and hotels Independent variable: te2 – technical efficiency obtained from Cobb Douglas Production Function , hotel dummy variables for hotels. Random effects GLS regression Number of obs = 41 Group variable (i): id Number of groups = 11 R sq: within = 0.0032 Obs per group: min = 2 between = 0.4150 avg = 3.7 overall = 0.2180 max = 4 Random effects u_i ~ Gaussian Wald chi2(2) = 5.26 corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0722 prof Coef. Std.Err. z P>|z| [95% Conf. Interval] te2 426704 223108.4 1.91 0.056 10580.43 863988.5 hotel 158213.3 90090.3 1.76 0.079 18360.4 334787.1 _cons 261552.4 157947.1 1.66 0.098 571123 48018.26 /sigma_u 138338.29 /sigma_e 80884.039 rho .74523748 (fraction of variance due to u_i)
Food Processing Industry
Random effects GLS regression Number of obs = 75 Group variable (i): id Number of groups = 21 R sq: within = 0.0294 Obs per group: min = 2 between = 0.2240 Avg = 3.6
overall = 0.2063 Max = 4 Random effects u_i ~ Gaussian Wald chi2(1) = 5.37 corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0205 prof | Coef. Std.Err. z P>|z| [95% Conf. Interval] te1 | 275796.3 119000.3 2.32 0.020 42559.96 509032.6 _cons 23838 45633.38 0.52 0.601 113277.8 65601.77 sigma_u 138198.63 sigma_e 55290.772 rho .86202027 (fraction of variance due to u_i)
REFERENCES Aigner D.J., C. A. K. Lovell, and P. Schmidt (1977), Formulation and Estimation of Stochastic Frontier Production Function Models , Journal of Econometrics 6, pp. 21 37 Battese, G. E. and T. J. Coelli (1992), Frontier Production Functions, Technical Efficicency and Panel Data: with Applications to Paddy Farmers in India, Journal of Productivity Analysis, 3, 153 169 Enterprise Incubator Foundation (EIF), (2007), Armenian Information Technology Sector, Software and Services, 2006 Industry Report Decaye, J. (2000), Food Processing Industry Profile , AEPLAC (Armenian European Policy and Legal Advice Centre) M. J. Farrell (1957), The Measurement of Productive Efficiency, Journal of the Royal Statistical Society, Series A (General), Vol. 120, No. 3, 253 290 Meeusen, W. and V. D. Broeck (1977), Efficiency Estimation from Cobb Douglas Production Functions with Composed Error, International Economic Review 18, 435 444 STATA Statistical Software: Release 9, 2005. College Station, TX: StataCorp LP The World Bank (2007), Armenia : Labor Market Dynamics , Volume II: Main Report. Report No. 35361 AM. Human Development Sector Unit, Europe and Central Asia Region, The World Bank www.armstat.am, National Statistical Service of Armenia www.cba.am, Central Bank of Armenia www.imf.org, World Economic Outlook Database
EXCHANGE RATE DYNAMICS IN ARMENIA Vahé Heboyan, PhD Candidate, Applied Economics, University of Georgia Lewell F. Gunter, Professor, Department of Agricultural and Applied Economics, University of Georgia Abstract: While experiencing impressive economic growth, Armenia’s exchange rate has significantly appreciated with respect to major currencies in nominal and real terms, and this has triggered concerns in various parts of the society. The Central Bank of Armenia and the government on one side and economists, businessmen, and political leaders on the other side, provided their own, mostly contrary, interpretations and arguments over the causes of the appreciation. This study attempts to understand the underlying economic fundamentals that explain exchange rate dynamics in Armenia. The Behavioral Equilibrium Exchange Rate has been identified as an appropriate approach for modeling long run exchange rate dynamics in Armenia. Results show that Armenian currency has been misaligned from its long run equilibrium path. The estimated degree of misalignment is sensitive to the set of economic fundamental variables used in the estimation. Even though, estimated models presented in this study have strong statistical parameters and yield similar misalignment trends, they differ in the magnitude of the estimated misalignment. Further analysis are necessary to study the sensitivity of the results to the choice of the variables and given the theoretical ambiguity of the signs of majority of variables, additional work is required to make the correct inference about the results. JEL Classification: E50, F31 Keywords: monetary policy, BEER, exchange rate determination
I. INTRODUCTION Economic problems associated with exchange rate dynamics have become one of the central issues dominating macroeconomic scientific research. Determination of the right exchange rate has been one of the key objectives for international investors, multinational corporations, and scientists (Rosenberg, 2003). Equally, the choice/adoption of the right exchange rate regime was on the agenda during 1980s and 1990s transition process from centrally planned to market oriented economies in the countries of the former Soviet Union (FSU) and Central and Eastern Europe (CEE). Several scientists have blamed economic crises in the developing world as being directly or indirectly caused by the inappropriate exchange rate policies in those countries 1. Understanding of exchange rate behavior, underlying determinants, equilibrium path, exchange rate misalignment, and the impact on the overall economic performance and competitiveness have always been of great importance in the exchange rate literature (Edwards, 1989; Égert et al., 2005). Overvaluation of exchange rates has very important macroeconomic consequences. It may alter international competitiveness of the country, may affect inflation, output and foreign direct investments (FDI) significantly, and may also signal a currency crisis (Dibooglu and Kutan, 2001). However, undervaluation of the currency may also have negative economic consequences. For example, it may lead to higher inflation and price instability due to dramatic growth in exports. Thus, adopting the right exchange rate policy and getting the exchange rate right becomes crucial for the overall success of the economy during the transition and in the long run (Égert et al., 2005). This study is part of a larger study that aims to understand the macroeconomic implications of exchange rate dynamics on transition economies with a focus on Armenia. Republic of Armenia is a land locked country situated on the south eastern edge of the Europe. For over 70 years it was part of the Soviet Union. Before the devastating earthquake of 1988, which paralyzed vast parts of the economy and killed more than 25,000 people, Armenia’s scientific institutes were among the best in the entire Soviet bloc, supporting some of the widely recognized Soviet high technology and military developments. Its collective farms, kolkhozes and sovkhozes, vineyards, and factories produced some of the highest quality foods and consumer goods for a market of 286 million people. “Armenia was the California of Soviet high technology, the Italy of Soviet shoe manufacturing, the France of Soviet made cognac” concludes the National Geographic (Viviano, 2004). Fueled by liberal market and economic reforms that promoted investor confidence and boosted exports, Armenia was quick to recover from production decline experienced by all transition countries of the FSU and CEE (World Bank, 2007). Armenia’s economic recovery started in 1994 along with the countries of CEE while the rest of the FSU countries (except Latvia) were still experiencing negative growth. Armenia’s economy has registered double digit economic growth since 2001 averaging 12.3 percent annual growth during 2001 2006 (World Bank, 2007; Roland, 2000; Iradian, 2007). Its exports grew by an average of 20.8 percent annually during 2000 2005 surpassing all countries 1 Edwards (1989) refers to the 1980s debt crisis (Cline, 1983), failed experiments with free market policies in the Southern Cone (Corbo et al., 1986), and the disappointing performance of Africa’s agricultural sector (World Bank, 1984).
in the FSU and CEE. Armenia’s inflation has also been surprisingly low averaging 2.6 percent annually during 2000 2005, which is lower than other countries in FSU and CEE except Lithuania (0.9%), Czech Republic (2.5%), and Poland (2.5%) (World Bank, 2007; Iradian, 2007). Such impressive economic performance in the last years has been accompanied with a rapid appreciation of the national currency, the Dram (ISO code: AMD), which has triggered alarms in various parts of the society. During 2003 2007, the dram appreciated by 46, 27, and 29 percent with respect to US dollar, Euro, and Russian Ruble respectively (IMF, 2008) (see Figure A.1). The Central Bank of Armenia, prominent economists, businessmen, and politicians offered their own interpretation of the causes of the appreciation, in many cases causing further controversies on the matter. The Central Bank of Armenia (CBA) has backed the official government view that the current appreciation is linked to the drastic growth in cash remittances regularly sent home by hundreds of thousands of Armenians working abroad, mainly in Russia, and relatives living in other countries, as well as by the continuous weakening of the US dollar in international currency markets. The CBA estimates that about 40 percent of households in Armenia receive remittances from abroad and insists that the drastic increases in the dollar remittances are the major cause of such changes. According to the CBA, the dollar value of remittances jumped by 50 percent to $760 million in 2004 from a year earlier. Compared to the monetary base of Armenia’s small economy, about $268 million in circulation, the large amounts of remittances may indeed cause such major fluctuations in the currency exchange market. According to Mr. Smbat Nasibian, the chairman of Armenia’s major commercial bank, the Converse Bank, “there are just too many dollars in circulation in Armenia.” To the contrary, critics question the credibility of the official statistics on the remittances. In particular, Eduard Aghajanov, a leading economist and the former head of the National Statistical Service (NSS) has argued that “Armenians living in Russia or the United States could not have gotten 50 percent wealthier within a year” (Danielyan, 2005). Whatever the reason is, the appreciation of the dram has already generated negative reaction from the hundreds of thousands of Armenian families that rely heavily on remittances from abroad. The economic decline that followed the collapse of the Soviet Union has forced nearly one million Armenians, or about 30 percent of Armenia’s current population, to migrate to other countries, primarily to Russia, in search for work. Banaian and Roberts (2007) estimate that remittances constitute 80 percent of the total income in the households that receive them. Despite large numbers, the CBA argues that dependence on remittances is exaggerated as they account for only about 25 percent of income in Armenia. CBA goes further arguing against intervention into the exchange market, as its primary objective is to ensure low inflation, which it has accomplished successfully 2 (IMF, 2007b; World Bank, 2007). The survey of the Armenian European Policy and Legal Advice Center (AEPLAC) conducted in January 2005, found that more than 85 percent of respondents save money in US dollars, that nearly 50 percent of those surveyed claim to have lost from the dram’s appreciation and that only 27.6 percent claimed to have been better off as a
2 The CBA has announced a revision of its 2008 inflation target from 3 to 4 (±1.5) percent (IMFc, 2007). result of appreciation (Yeghiazaryan, 2004). Claims against government manipulation have become stronger due to the fact that virtually no imported product became cheaper due to the appreciation. Mr. Nasibian of the Converse Bank, believes that “the main reason for that is a very small number of importers. Each of them seems to have monopolized a particular field, making disproportionate profits” (Danielyan, 2005). Contrary, the government and the importers claim that the price increases in the world markets offset the potential for decline in the prices for imported goods. The latest International Monetary Fund review (IMF, 2007a) has concluded that further action is needed in Armenia “ to look for ways to reduce monopolistic practices in the import business, with a view to increasing the pass through of exchange rate changes to domestic prices” . Armenian export businesses have also raised alarms regarding continuing appreciation of the dram. Since late 2006, many exporters have articulated for more intervention from the CBA in the currency exchange market. More specifically, the head of the Diamond Company of Armenia, and the President of the International Association of Armenian Jewelers, Mr. Gagik Abrahamian has suggested a return to a fixed exchange rate regime around AMD 400/US$. The Director General of the Shoghakn diamond cutting company, Mr. Sergey Gasparyan, has also spoken with a similar support (Emerging Markets Monitor, 2006). Despite these calls for intervention, the CBA continues to hold to its primary objective of inflation targeting, and the Emerging Markets Monitor (2006) does not believe that there will be substantial changes in Armenia’s monetary policy in the near future. It forecasts that AMD will move towards AMD 300 per US dollar by 2010 and believes that AMD appreciation has also created benefits, such as low inflation and low impact of international price increases on the Armenian market. For example, dram appreciation helped to prevent complete pass through of increasing world oil prices to the domestic market. The CBA also believes that appreciation creates unique opportunity for local businesses to boost their competitiveness through acquiring new foreign technologies as AMD appreciation makes foreign technologies cheaper in local currency (Emerging Markets Monitor, 2006). These contradicting arguments and statements continue to dominate public private discussions in Armenia. Thus, scientifically robust explanations are timely and crucial to avoid further speculation and accusations around monetary developments in Armenia and their use for political manipulations.
II. OBJECTIVES This study is part of a larger research effort to survey, adapt, and extend empirical models from monetary and financial economics to benefit the understanding and practical modeling of exchange rate dynamics and behavior, exchange rate pass through into prices of strategic and important consumer goods in Armenia. A specific objective of this part of the research is to examine the dynamics of the real exchange rate in Armenia and explain its relation to economic fundamentals. The study aims to provide scientifically supported explanations of recent developments in the currency exchange market and the behavior of exchange rates in Armenia, thus offering an alternative to speculative theories and on the causes and effects of exchange rate dynamics in Armenia.
III. THEORETICAL BACKGROUND
The real exchange rate (RER) is one of the most important concepts in international economics and the most important relative price in international finance. It has attracted enormous attention from leading international organizations and researchers who try to understand the behavior of real exchange rates and their determinants. Exchange rate research has been, for the last few decades, one of the most ‘popular’ theoretical and empirical research topics at the International Monetary Fund. Real exchange rate plays a crucial role in determining the competitiveness position of a country in the global market. Real exchange rates directly impact inflation and output in every economy and are more important for the young and fragile economies in transition. In these economies, one of the main macroeconomic discussion topics is the critical role of the real exchange rate in the economic adjustment process. Edwards (1994) believes that there is a general consensus among researchers that sustained real exchange rate misalignment will generate serious macroeconomic imbalances, and to correct the external imbalances, such as the current account deficit, strong demand management policies and real exchange rate devaluation may be required. Additionally, it has been established that much of the economic success in the ‘successful’ developing countries are due to successful exchange rate polices that maintained the real exchange rate at the ‘appropriate’ level. Thus, the behavior of the real exchange rate is a key component in macroeconomic policy evaluation and design. A very important concept related to real exchange rate economics is real exchange rate misalignment . Edwards (1989, p. 8) defines it as sustained deviations of the actual real exchange rate from its long run equilibrium level . When the actual real exchange rate is below the equilibrium value, it is said that the real exchange rate is overvalued , and undervalued if it is above its long run equilibrium mark. Thus, the understanding of the concept of the equilibrium, and equilibrium real exchange rate (ERER) in particular, becomes the critical building block in this entire story. The concept of equilibrium itself has generated heated debates over diverse range of issues, such as its existence, uniqueness, optimality, determination, and evolution over time. Von Neumann and Morgenstern (1944) believe that without imposing a structure by which different models should be judged, one cannot choose between them as the solution to all of them must satisfy the same analytical thinking. Therefore the concept of equilibrium is no less important within the context of the exchange rates than it is for other fields of economics (Driver and Westaway, 2004). Driver and Westaway (2004) discuss the concept of the equilibrium exchange rate in the context of the time horizon, where they identify short , medium , and long run equilibrium exchange rates: (a) S hort run equilibrium is defined as the exchange rate that results when its fundamental determinants are at their current settings after removing the effects of random shocks. (b) Medium run equilibrium is defined as the rate at which the economy achieves internal and external balance. (c) Long run equilibrium is achieved at the point when stock point equilibrium is achieved for all agents of the economy.
Within the equilibrium setting, it is also important to know which measure of the exchange rate should be used, and why. The choice of the exchange rate measure (nominal versus real), the price deflator, or even if it is bilateral or multilateral measure directly depends on the relevant research agenda. Some choose to use the nominal bilateral exchange rate as determined directly in the financial markets. However, most theories of equilibrium exchange rates refer to the real exchange rate, even though different measures of relevant prices are used. Driver and Westaway (2004) have identified five price measures commonly used to define the real exchange rate: (a) consumer price index, (b) prices of tradable goods or output prices, (c) price of an economy’s exports compared to the price of its exports, (d) relative unit labor costs, and (e) ratio of tradable to non tradable prices. There is a common consensus in the exchange rate literature that there is no single best modeling approach to exchange rate behavior or to identifying a common measure of relevant price. The choice of the approach depends on the question of interest and more importantly on the time horizon of the study. Driver and Westaway (2004) also believe that the approaches may differ in the treatment of dynamics and the time frame they concentrate on. The summarized overview of these different approaches is compiled Driver and Westaway (2004, p.26). The concept of exchange rate misalignment is perhaps the most challenging empirical problem in macroeconomics due to unobservable equilibrium value of the real exchange rate. However, economic theory models ERER as a function of observable fundamental variables of the economy assuming that the actual RER converges to its equilibrium in the long run (Baffes et al., 1999). Since the early 1990s, a growing number of empirical studies have applied the above relationship to studying equilibrium real exchange rates and misalignments associated with them. The behavioral equilibrium exchange rate approach (BEER) has become a standard workhorse in examining real exchange rates and their misalignment from long run equilibrium. The BEER approach has been popularized by Sebastian Edwards (1994; 1989) and MacDonald (1997) and used by many authors in the empirical literature. This approach has also become a standard approach for exchange rate modeling at the International Monetary Fund. The influential works of Edwards (1994; 1989) were the first substantial attempts to understand the behavior of the real exchange rate in a transition economy setting in terms of economic fundamentals. Edwards’ framework has been adapted and extended by many. Khan and Ostry (1991) study the response of the ERER to the real shocks in the developing countries. Using panel data, they estimate ERER elasticities for trade shocks and commercial policies. Elbadawy (1994) applies the simplified version of the model to estimating ERER for Chile, Ghana, and India. Faruqee (1995) and Mongardini (1998) examine the ERER in Egypt, De Broeck and Sløk (2001) extend the model to the transition economies of the CEE and Baltic countries, Lane and Milesi Farretti (2001) for Ireland, and MacDonald and Ricci (2003) for South Africa. Chudik and Mongardini (2007) have applied the Autoregressive Distributed Lag (ARDL) modeling techniques pioneered by Pesaran and Shin (1999) to estimate equilibrium real exchange rates in 39 Sub Saharan African (SSA) countries. BEER has not been successful when applied to a single country due to data limitations (only 26
yearly observations), but their panel data estimates were statistically and economically significant even when different estimation techniques were used. The authors have also developed a user friendly template that automates variable selection and estimation procedures (Chudik, 2006a; Chudik, 2006b). Widespread success of the Edwards’ (1994; 1989) seminal framework and its extended versions in the recent decade or two, and more specifically its use for studying the behavior and determinants of real exchange rates in CEE and FSU countries, provide an appropriate ground to adopt it for the purposes of examining the behavior of the real exchange rate in Armenia and possible deviations from its long run equilibrium path. More specifically, the approach and template used by Chudik and Mongardini (2007) are adopted for this study. Edward’s model is an intertemporal general equilibrium model of a small open economy. The unique ERER is attained when the economy achieves its internal and external balance. The model’s internal balance is achieved when all markets for non tradable goods are cleared (static equilibrium). External balance is achieved when the net present value of the future current accounts is non negative at the given level of exogenous long run capital inflows (dynamic equilibrium). A formal summary of the model is provided by Edwards (1994; 1989) and Mongardini (1998, Appendix II). The basic structure of Edwards’ real exchange rate model is: * * =logetθ( log e t − log e t−1 ) −−+ λ ( ZZ tt ) (1) +φ()()logStt − log S−1 − ψ PMPR t − PMPR t − 1 where, e is the actual real exchange rate, e* is the equilibrium real exchange rate (in turn a function of fundamentals), Zt is an index of macroeconomic policies (i.e. the rate * of growth of domestic credit), Zt is the sustainable level of macroeconomic policies
(i.e. the rate of increase of demand for domestic money), St is the nominal exchange rate, PMPR is the spread in the parallel market for foreign exchange, and θλψ, , ,and φ are positive parameters that capture the most important dynamic aspects of the adjustment process. Equation (1) clearly suggests that the real exchange rate is moving due to three forces. First , the actual real exchange rate will tend to independently correct existing * misalignment through the partial adjustment term θ (loget− log e t −1 ) . The speed of the adjustment is determined by the parameter θ . The larger is the parameter, faster will be the speed at which the real exchange rate misalignment will be corrected. The second term that determines the real exchange rate movements is given by the term for * macroeconomic policies, −λ (Zt − Z t ) . According to Edwards (1994; 1989), if policies are “inconsistent” the real exchange rate will become overvalued, ceteris paribus . However, Chudik and Mongardini (2007) argue that it is not clear whether changes in these policies will have an impact on the ERER in the long run. The third element of the equation is the change in the nominal exchange rate (i.e. nominal devaluation) represented by φ (logSt− log S t −1 ) . Nominal devaluation will in the short run cause the real exchange rate to depreciate. The magnitude will depend on the parameter φ . The forth element refers to the changes in the parallel market premium. An increase in this term will cause a real exchange rate appreciation. The ERER in (1) can be written as: * (2) loget=β0 + β i log ( FUND it) + ε t where, FUND it represents a set of fundamental variables that are assumed to have a determining effect on the ERER. The choice of the fundamentals varies from one country to another. Likely determinants of ERER in a developing country, as discussed for example, in Edwards (1994; 1989), Chudik and Mongardini (2007), and Roudet et al. (2007), are: 1. Terms of trade for goods (TOT). Is defined as the price ratio of the country’s exports over imports. An improvement in the terms of trade implies increase in the international price for exports, which will have a positive impact on the current account and lead to the ERER appreciation. 2. Government spending (GOV). Usually defined as the government consumption of non tradable goods. An increase in the consumption vis à vis tradable goods will improve the current account and lead to appreciation of the ERER. 3. Market openness (OPEN). This is a proxy for trade controls or restrictions. An increase in the openness will lead to an increased trade. The equilibrium response of the ERER will depend on whether this leads to improving or deteriorating current account. 4. Technological progress/productivity (TECHPRO). Allows capturing the famous Balassa Samuelson effect that the productivity improvements will generally be concentrated in the tradable goods sector. Technological progress increases the productivity in the economy and leads to the appreciation of the ERER without hurting its competitiveness. 5. Investment (INV). Usually defined as the ratio of investments to GDP relative to that of foreign partners. Investments in developing countries are usually concentrated in the imports sector, thus have negative impact on the trade balance. But the overall impact on ERER is still ambiguous as it still may capture technological progress. 6. Debt service (DS). Is defined as a share of exports. An increase in the debt service payments leads to the deterioration of the external balance and will require price adjustments to restore the equilibrium. Thus, a ERER deprecation should be expected. 7. Net foreign assets (NFA). Is a proxy for the country’s net external position and is defined as a share of GDP. An improvement in the position (increase in capital inflows) leads to the appreciation of ERER. 8. Aid flows (AID) as a share of exports. It represents a significant share of several developing and low income countries. An increase in aid flows improves the external balance and leads to ERER appreciation. 9. Controls over capital flows (CAPCTRL). Similar to market openness liberalization will have impact on the ERER. The direction will depend on the real interest rate differential and the country’s risk profile.
After substituting (2) into (1), and for convenience and generalization using a single notation of macroeconomic policy variables, say POLICY , and omitting the PMPR term, we obtain a reduced form equation that could be estimated using conventional methods.
(3) logeti=γ log( FUND it) +−−( 1 θλ) log e ti−1 ( POLICY it) + φε NOMDEV tt + where, NOMDEV stands for nominal devaluation and γ ’s are combinations of β ’s and θ . Nominal devaluation ( NOMDEV ) has positive sign and can be quite powerful for reestablishing real exchange rate equilibrium. However, Edwards (1989) believes that “for the nominal devaluation to have a lasting effect, it is necessary that the sources of the original disequilibrium – positive EXCRE and DEH 3 – be eliminated. If this is not the case, soon after the devaluation the RER will again become overvalued ” (p. 141). IV. DATA AND ESTIMATION METHODOLOGY
The study uses quarterly data covering 1996:Q1 2007:Q4. Table A.1 provides a brief description of each variable and its source(s). One of the major problems faced when working with time series data is the issue of stationarity in the series. In the short time series the determination of variables’ order of integration becomes uncertain due to poor performance of unit root tests for small samples (Chudik and Mongardini, 2007). Gregory and Hansen (1996) argue that “ the standard tests for cointegration are not appropriate, since they presume that the cointegrating vector is time invariant under the alternative hypothesis (p. 100) ” and that if there exists a cointegration, the standard ADF test may not reject the null, thus wrongly concluding that there is no long run relationship. Gregory, Nason, and Watt (1996) have found that the power of standard ADF test decreases sharply when a structural break is present. Structural breaks occur with technological progress, economic crises, changes in people’s preferences, policy or regime shifts, and institutional developments, which are very typical to developing and transition countries. The traditional econometric cointegration approaches, such as Johansen’s, require the series to be integrated to the same order, thus introducing a further degree of uncertainty into the analysis of level relationships especially in a transition country setting. Pesaran, Shin, and Smith (2001) developed a new approach for testing for the existence of level relationship between variables irrespective of whether the underlying variables are stationary, integrated to the order of one, or a mixture of the two. This approach has been successful and superior to the traditional Johansen cointegration test in a small sample (Chudik and Mongardini, 2007). 3 EXCRE and DEH are excess supply for domestic credit calculated as the rate of growth of domestic credit minus lagged rate of growth of real GDP and ratio of fiscal deficit to lagged high power money respectively. They measure the role of macroeconomic policies in the RER behavior (Edwards, 1989, p. 137).
The estimation is done using an Autoregressive Distributed Lag (ARDL) Modeling approach proposed by Pesaran and Shin (1999). This approach makes estimation independent of the order of integration in the variables, thus provides statistically better results (Mongardini, 1998). The general ARDL model for a dependent variable yt and independent variables xt with intercept (c) is written as:
py n px (4) yt=∑α iti y− + ∑∑ β jijti xc, − + i=1 j = 1 i = 1 where, py, p x are the orders of lags for dependent and independent variables, respectively, and n refers to the number of regressors. The corresponding error correction will be represented with:
n py −1 n px −1 (5) =yytππ yyt−1 +∑ yxjjt , x , −− 1 + + ∑ θ iti y ∑∑ ψ jijti + xc, − j=1 i = 1 ji == 10 The bounds tests proposed by Pesaran, Shin, and Smith (2001) uses F statistics to tests for the existence of the level relationship. The joint null hypothesis is:
(6) H0:πyy=∩ 0{ π yx , j = 0, j = 1,2,..., n } For the purposes of this research, Chudik’s (2006b; 2006a) econometric template for conducting estimations using Pesaran and Shin's (1999) ARDL approach was slightly modified. It is user friendly and has an MS Excel and EViews based interface. Due to the small sample size, it is impossible to include all potential explanatory variables in the estimation process, thus there is a need for selecting the appropriate set of variables. The empirical literature commonly uses up to four explanatory variables in a small sample setting. The template enables the user to incorporate the variable selection into the estimation process.
V. RESULTS AND CONCLUSIONS A total of 126 models with 3125 possible specifications for each model of 9 potential explanatory variables 4 with up to four lags in the dependent and independent variables are investigated for the best possible combination of four variables, which are then ranked according to the following principles: a) existence of long run cointegration using bounds testing; b) the number of statistically significant variables at 5 percent significance level; and c) the number of correct signs of the parameters as predicted by the economic theory.
4 See Table A.1 for the list and explanation of variables used in this study.
The two best models (Models A and B) with respective long run parameter estimates and t statistics and associated misalignments are illustrated in Figures A.2 A.5. Note that an increase in the exchange rate measures here represents an appreciation of the dram. Model A with (2,0,0,0,0) 5 quarterly lag structure (7) ERER=×0.48 GOV −× 0.80 NFA +× 0.05 INV −× 0.63 EXCRE 3.8− 4.5 2.3− 5.0
Model B with (1,0,4,2,0) quarterly lag structure (8) ERERGOVNFAEXCRENOMDEV=×0.28 −×−× 1.61 0.88 +× 0.03 3.5− 7.3 − 8.5 4.6
All coefficients are statistically and economically significant and all except the macroeconomic policy variable ( EXCRE ), measured as the rate of growth of domestic credit minus lagged rate of growth of real GDP as suggested by Edwards (1989, 1994), have the expected signs. The sign of EXCRE has not been widely discussed in the literature and Edwards (1989, 1994) offers little explanation for the suggested sign. Thus, a further study is necessary for the correct inference. The positive sign on government consumption ( GOV ), as a share of GDP relative to that of foreign trading partners, suggests that government consumption is biased towards non tradables. The negative sign on Net Foreign Assets ( NFA ) as a share of GDP suggests that high NFA increase current account by increasing investments, thus causing depreciation. The sign on direct investments in the country ( INV ) as a share of GDP relative to that of the trading partners suggests that investments are associated with technological progress and productivity increases, thus cause appreciation. The sign on the nominal devaluation is intuitive however the magnitude is much smaller than what has been found in other research.
Real exchange rate misalignment ( mt) is calculated as the deviation of the real exchange rate from its estimated equilibrium value expressed in a percentage term, as expressed in equation 8: * (9) mt=log e t − log e t * where et and et are real and real equilibrium exchange rates respectively. Results of the study show that over the course of the study, the real exchange rate for Armenia was misaligned from its long run equilibrium path. However, the degree of misalignment is sensitive to the set of economic fundamental variables and lags used in the estimation. Due to the small sample size, it is impossible to include all variables that are suggested by the economic theory, thus a subset of those variables needs to be chosen in the study. Even though, both models have strong statistical parameters and yield similar misalignment trends, they differ in the magnitude of the misalignment.
5 Lag structure for the dependent variable first then the explanatory variables.
Further analyses are necessary to study the sensitivity of the results to the choice of the variable set. Additionally, given the theoretical ambiguity of the signs of majority of variables, additional work is required to make better inferences about the results. In the final research report, the feasibility of grouping of several variables and introducing better proxies will be studied to avoid the need for limiting the number of variables in the model. Additionally, the short run volatility of the bilateral nominal exchange rate needs to be studied based on changes in monetary aggregates and expectations. An additional expectations defining variable(s) will be introduced to capture the potential effect of major political and economic events in the country. And finally, the exchange rate pass though into the price of important products will be examined. Understanding the impact of exchange rate fluctuations on prices and output has a critical policy implication in order to determine the appropriate monetary policy response, which may influence the future direction of the economic, social, as well as political developments in the country.
Appendix
Figure A.1 Exchange Rates Indexes for Armenia, 1996-2007 (2000=100)
Source: own calculations, (IMF, 2008) Note: increase means an appreciation. Russia is represented on the right axis.
Table A.1 Data Description and Sources Variable Description Log transformation Data source REER Real equilibrium exchange rate. Is a multilateral CPI based real effective Yes 2 exchange rate. It is defined in terms of Armenian Dram per unit of foreign currency, so that an increase (decrease) in REER r epresents depreciation (appreciation). ToT Terms of trade for Armenia is defined as the ratio of export price index over Yes 1, 3, 4 import price index. GOV Government consumption as a share of GDP relative to that of foreign trading Yes 1, 2, 4 partners. OPEN Openness to trade: exports plus imports as a share of GDP. Yes 1, 2 NFA Net foreign assets as a share of GDP. No 1, 2 INV Direct investments in rep. economy as a share of GDP relative to that of the Yes 1, 2, 5 trading partners. EXCRE Excess supply of domestic credit. Measured as the rate of growth of domestic No 1, 2 credit minus the lagged rate of growth of real GDP. PROD Measure of productivity. Proxied by per capita real GDP relative to that of the Yes 1, 2, 6, 7 trading partners. DS Debt service as a share of exports. No 1, 8, 9 MONDEV Nominal devaluation. Nominal effective exchange rate is used. No 2 Sources: (1) own calculations (9) (CBA, 2006) (2) (IMF, 2008) (3) (NSS, 2003a; NSS, 2006; NSS, 2007a) (4) (Datastream International, May 1, 2008) (5) (Armenian Economic Association, April 29, 2008) (6) (IMF, May 9, 2008) (7) (NSS, 2001; NSS, 2003b; NSS, 2007b) (8) (IMF, 1998; IMF, 1999; IMF, 2001; IMF, 2002)
Figure A.2 REER and ERER for Armenia, Model A (in natural logarithms)
5 4.9 4.8 4.7 4.6 4.5 4.4 4.3 4.2 4.1 1996 1 1997 1 1998 1 1999 1 2000 1 2001 1 2002 1 2003 1 2004 1 2005 1 2006 1 2007
Figure A.3 Real Exchange Rate Misalignment in Armenia, Model A
50%
40%
30%
20%
10%
0%
10%
20%
30%
40%
Figure A.4 REER and ERER for Armenia, Model B (in natural logarithms)
5.4
5.2
5
4.8
4.6
4.4
4.2
4 1996 1 1997 1 1998 1 1999 1 2000 1 2001 1 2002 1 2003 1 2004 1 2005 1 2006 1 2007
Figure A.5 Real Exchange Rate Misalignment in Armenia, Model B
60%
40%
20%
0%
20%
40%
60%
80% 1996 1 1997 1 1998 1 1999 1 2000 1 2001 1 2002 1 2003 1 2004 1 2005 1 2006 1 2007
REFERENCES Armenian Economic Association, (April 29, 2008), Balance of Payments [Online]. Available: Armenian Economic Association/Economic Data (www.aea.am) Baffes, J., I. A. Elbadawi, and S. A. O'Connell (1999), Single Equation Estimation of the Equilibrium Real Exchange Rate , in Hinkle, L. E. and Montiel, P. J. (eds.), Exchange Rate Misalignment, Concepts and Measurement for Developing Countries, New York, New York: Oxford University Press, 405 464 Banaian, K. and B. Roberts (2007), Remittances in Armenia II: The Impact of Remittances on the Economy and Measures to Enhance Their Contribution to Development , Armenian Journal of Public Policy 2, 229 257 CBA (2006), Annual Report 2006, Appendix. Central Bank of Armenia, [Online]. Available at: http://cba.am/publications/anne_2006/appendix.pdf Chudik, A. (2006a), ARDL Approach to Cointegration Econometric Template (available upon request) Chudik, A. (2006b), Documentation for ARDL Approach to Cointegration Econometric Template, Unpublished manual (available upon request) Chudik, A. and J. Mongardini (2007), In Search of Equilibrium: Estimating Equilibrium Real Exchange Rates in Sub Saharan African Coountries , International Monetary Fund Working Paper Danielyan, E. (2005), Armenian Government Under Fire Over Continuing Currency Appreciation, www.EurasiaNet.org. Accessed on 14 Dec 2006. Available at: http://eurasianet.org/departments/business/articles/eav050205_pr.shtml Datastream International, (May 1, 2008), [Online]. Available: Datastream International/Economics De Broeck, M. and T. Slø, (2001), Interpreting Real Exchange Rate Movements in Transition Countries, International Monetary Fund Working Paper Dibooglu, S. and A. M. Kutan (2001), Sources of Real Exchange Rate Flactuations in Transition Economies: The Case of Poland and Hungary, Journal of Comparative Economics 29, 257 275 Driver, R. L. and P. F. Westaway (2004), Concepts of Equilibrium Exchange Rates, Bank of England Working Paper 248 Edwards, S. (1989), Real Exchange Rates, Devaluation, and Adjustment: Exchange Rate Policy in Developing Countries , Massachusetts Institute of Technology Edwards, S. (1994), Real and Monetary Determinants of Real Exchange Rate Behavior: Theory and Evidence from Developing Countries , in Williamson, J. (ed.) Estimating Equilibrium Exchange Rates , Washington, DC: Institute for International Economics
Égert, B., L. Halpern, and R. MacDonald (2005), Equilibrium Exchange Rates in the Transition Economies: Taking Stock of the Issues, William Davidson Institute Working Paper Elbadawy, I. (1994), Estimating Long Run Equilibrium Real Exchange Rates, in Estimating Equilibrium Exchange Rates , in Williamson, J. (ed.) Estimating Equilibrium Exchange Rates, Washington, DC: Institute for International Economics Emerging Markets Monitor (2006), Armenia: Is AMD Strength Starting To Hurt? In Emerging Markets Monitor, Vol. 12. 1 3 Faruqee, H. (1995), Long Run Determinants of the Real Exchange Rate: A Stock Flow Perspective, IMF Staff Papers 42, 80 107 Gregory, A. W. and B. E. Hansen (1996), Residual based tests for cointegration in models with regime shift, Journal of Econometrics 70, 99 126 Gregory, A. W., J. M. Nason, and D. Watt (1996), Testing for structural breaks in cointegrated relationships, Journal of Econometrics 71, 321 341 IMF (1998), Republic of Armenia Recent Economic Developments, IMF Country Report No. 98/22. Washington, D.C.: International Monetary Fund IMF (1999), Armenia: Recent Economic Developments and Selected Issues, IMF Country Report No. 99/128. Washington, D.C.: International Monetary Fund IMF (2001), Republic of Armenia: Recent Economic Developments and Selected Issues, IMF Country Report No. 01/78. Washington, D.C.: International Monetary Fund IMF (2002), Republic of Armenia: Statistical Annex, IMF Country Report No. 02/225. Washington, D.C.: International Monetary Fund IMF (2007a), Statement at the Conclusion of an IMF Mission to Armenia, Press Release No. 07/214. International Monetary Fund IMF (2007b), World Economic Outlook: Globalization and Inequality, Washington, DC: International Monetary Fund IMF (2008), International Finance Statistics, International Monetary Fund IMF (May 9, 2008), Direction of Trade Statistics , [Online]. International Monetary Fund Iradian, G. (2007), Rapid Growth in Transition Economies:Growth Accounting Approach. International Monetary Fund Working Paper Kahn, M. and J. Ostry (1991), Response of the Equilibrium Real Exchange Rate to Real Disturbances in Developing Countries, International Monetary Fund Working Paper
Lane, P. R. and G. M. Milesi Ferretti (2001), The External Wealth of Nations: Measures of Foreign Assets and Liabilities for Industrial and Developing Countries, Journal of International Economics 55 (December), 263 294 MacDonald, R. (1997), What Determines Real Exchange Rates? The Long and Short of It, Working Paper No 97/21. Washington, DC: International Monetary Fund MacDonald, R. and L. Ricci (2003), Estimation of the Equilibrium Real Exchange Rate for South Africa, International Monetary Fund Working Paper Mongardini, J. (1998), Estimating Egypt's Equilibrium Real Exchange Rate, International Monetary Fund Working Paper NSS (2001), Statistical Yearbook of Armenia. Yerevan, Armenia : National Statistical Service NSS (2003a), Prices and Price Indexes in The Republic of Armenia 1992 2002 (in Armenian). Yerevan, Armenia : National Statistical Service NSS (2003b), Statistical Yearbook of Armenia. Yerevan, Armenia: National Statistical Service NSS (2006), Prices and Price Indexes in The Republic of Armenia 2001 2005 (in Armenian), Yerevan, Armenia: National Statistical Service NSS (2007a), Prices and Price Indexes in The Republic of Armenia 2002 2006 (in Armenian), Yerevan, Armenia: National Statistical Service NSS (2007b), Statistical Yearbook of Armenia. Yerevan, Armenia: National Statistical Service Pesaran, M. H. and Y. Shin (1999), An Autoregressive Distributed Lag Modeling Approach to Cointegration Analysis . in Strom, S. (ed.), Econometrics and Economic Theory in the 20th Century: The Ragnar Frisch Centennial Symposium, Cambridge, England: Cambridge University Press Pesaran, M. H., Y. Shin, and R. J. Smith (2001), Bounds Testing Approaches to the Analysis of Level Relationships, Journal of Applied Econometrics 16, 289 326 Roland, G. (2000), Transition and Economics: politics, markets, and firms , Cambridge, Massachusetts: Massachusetts Institute of Technology Rosenberg, M. R. (2003), Exchange rate determination: models and strategies for exchange rate forecasting, New York: McGraw Hill Roudet, S., M. Saxegaard, and C. G. Tsangarides (2007), Estimation of Equilibrium Exchange Rates in the WAEMU: A Robustness Analysis, International Monetary Fund Working Paper
Viviano, F. (2004), The Rebirth of Armenia, in The National Geographic, Vol. March 2004 . 28 49 Von Neumann, J. and O. Morgenstern (1944), Theory of Games and Economic Behavior, Princeton University Press World Bank (2007), World Development Indicators 2007 , CD ROM. Washington, DC: World Bank Yeghiazaryan, A. (2004), The Latest Developments in the Foreign Exchange Market: An Attempt to Make a Quantitative Analysis, Armenian Trends Q3, 14 16
BANK EFFICIENCY , MARKET STRUCTURE , AND FOREIGN OWNERSHIP : WHAT DETERMINES BANKING SPREADS IN ARMENIA ? Era Dabla Norris, International Monetary Fund Holger Floerkemeier, International Monetary Fund* Abstract: Despite far reaching banking sector reforms and a prolonged period of macroeconomic stability and strong economic growth, financial intermediation in Armenia has lagged behind other transition countries, and interest rate spreads have remained higher than in most Central and Eastern European transition countries. This paper examines the role of bank characteristics, market structure, foreign ownership, and macroeconomic factors in determining interest rate spreads and margins in Armenia. We find that bank specific factors, such as bank size, liquidity, and market power, as well as the market structure within which banks operate, explain a large proportion of cross bank, cross time variation in spreads and margins. In contrast to the experience of other transition countries, foreign bank ownership is not directly associated with lower spreads and margins. However, we find that foreign bank origin matters for banking efficiency. JEL Classification Numbers: E43, G21, C13 Keywords: Banking system, interest rate spreads, foreign ownership, Armenia
* The views of this paper represent those of the authors and do not reflect the views of the International Monetary Fund.
I. INTRODUCTION As in other transition countries, the structure of Armenia’s banking system has undergone significant changes. Bank restructuring and privatization has been accompanied by consolidation, market entry of new banks, an overhaul of the legal framework, and a strengthening of prudential regulation and supervision. However, despite these reforms and a prolonged period of macroeconomic stability and economic growth, financial intermediation remains low by regional standards (Table 1.1). At the same time, interest rate spreads have remained persistently high and well above those in most transition countries, averaging over 12 percent since 2003 (Figure 1.1). Table 1.1 Financial Intermediation in Selected Transition Countries, 2005
Broad Money Private Credit Bank Assets Bank Deposits /GDP /GDP /GDP /GDP Armenia 16.4 8.2 20.2 10.7 Azerbaijan 15.4 9.7 25.5 11.7 Estonia 50.1 60.0 108.2 44.8 Georgia 16.6 14.8 26.3 12.7 Kazakhstan 26.6 26.7 63.3 25.6 Latvia 45.4 59.9 115.9 22.4 Lithuania 40.7 34.9 59.9 32.1 Russia 33.4 23.9 44.8 25.0 Ukraine 45.5 33.4 52.3 31.6 Source: WEO and IFS databases
Figure 1.1 Regional Comparison of Interest Rate Spreads
Source: IMF IFS. The spread between lending and deposit rates is widely regarded as an indicator of the efficiency of financial intermediation. High interest rate spreads are an impediment to financial intermediation, as they discourage potential savers with low returns on deposits and increase financing costs for borrowers, thus reducing investment and growth opportunities. This is of particular concern for the less advanced transition countries where financial systems are largely bank based and tend to exhibit high and persistent spreads. Understanding the determinants of high spreads in Armenia is, therefore, important to inform policy for improving banking efficiency and achieving financial deepening. High intermediation spreads in many countries are attributed to a variety of factors. Bank specific factors, such as overhead costs, bank size, risk assessment capacity, and investment management, influence banking efficiency (Demirgüç Kunt and Huizinga, 1999; Dell’Arriccia and Márquez, 2004). The legal and institutional environment can contribute to the efficiency of intermediation through its impact on loan recovery rates, enforceability of foreclosures, and collateral collection, as well as market transparency and information sharing on borrowers (Demirgüç Kunt et al., 2004, Haselmann and Wachtel, 2006). At the same time, the market structure within which banks operate (ownership structure, market concentration, competition) can have important implications for the incentives of banks to overcome market frictions and efficiently intermediate the economy’s savings to borrowers. In particular, foreign bank entry, by imposing competitive pressure and bringing in expertise in risk management, can play an important role in determining the overall efficiency and profitability of the banking sector (Martinez Peria and Mody, 2004; Claessens et al. , 2001, Van Horen, 2007). This paper has two objectives. First we analyze the determinants of banking sector efficiency, as measured by spreads and net interest margins, in the Armenian banking
system using a bank level panel dataset over the period 2002 2006. 1 In particular, we examine whether bank market shares and market power, concentration, and operational efficiency play an important role in determining spreads and margins. The dataset also includes information on banks’ loan portfolio composition which allows us to assess whether varying risk premiums across sectors and variations in market segments account for differences in spreads and margins across banks. Second, in light of the financial sector reforms and changes in market structure and ownership that have taken place in Armenia, we investigate the impact of foreign bank entry in determining interest rate spreads and margins. In particular, we examine whether foreign banks operate with lower spreads and margins relative to domestically owned banks and whether foreign bank origin matters for banking efficiency. Previous studies on transition countries have typically focused on the importance of financial reforms, foreign ownership, and competition in determining banking sector efficiency in the more advanced transition countries (see Fries et al., 2002; Fries and Taci, 2005 for an overview). Claeys and Vander Vennet (2004) find that bank concentration, operational efficiency, and capital adequacy are important determinants of bank interest margins in Central and Eastern European (CEE) countries. Using data for banks in eleven transition countries, Bonin et al. (2005) show that majority foreign owned banks tend to be more efficient and have a positive impact on bank competition in transition countries. 2 Very few studies have examined the effect of financial sector reforms and changes in market structure on banking system efficiency in the CIS countries of the Caucasus and Central Asia. This paper contributes to the literature by investigating the determinants of banking system efficiency in relatively financially underdeveloped CIS country. Our results show the importance of bank specific characteristics in explaining the variation of interest rate spreads and margins across banks and across time in Armenia. We find that larger and more liquid banks with less exposure to agricultural and consumer loans are associated with lower spreads. Banks exhibiting higher market power, as measured by individual market shares, are associated with higher spreads but with lower margins. More profitable banks are associated with higher margins, while higher capital adequacy and liquidity ratios are associated with lower margins. Higher concentration in loan and deposit markets has a positive and economically significant effect on both spreads and margins. However, macroeconomic variables were found to have only a minimal impact on both spreads and margins. In contrast to the experience of other transition countries, the presence of foreign banks does not directly seem to have contributed to lower spreads and margins, reflecting the limited presence of first tier international banks in the Armenian banking sector. However, we find that foreign bank origin matters for banking efficiency, with first tier
1 Ex post interest rate spreads are the difference between the implicit interest rates charged to borrowers and paid to depositors, while net interest margins capture the difference between bank’s total interest income and total interest expenses. 2 Several single country studies have investigated the relationship between foreign ownership and bank efficiency in transition countries. Jemric and Vujcic (2002, Croatia), Hasan and Marton (2003) for Hungary, and Weill (2003) for the Czech Republic and Poland, find that foreign owned banks are more efficient than domestically owned banks.
(western) foreign banks having a spill over effect on interest rate spreads. The presence of banks from other countries, however, is associated with higher spreads. The remainder of the paper is organized as follows. Section II gives an overview of the Armenian banking sector and provides a regional comparison. Section III discusses methodology and data. Section IV presents the main results and discusses robustness tests. Section VI concludes and discusses policy implications. II. THE ARMENIAN BANKING SYSTEM The Armenian banking system has experienced substantial consolidation over the past decade. As in other transition countries, liberal licensing and regulation policies led to the creation of a large number of banks during the early 1990s. The gradual tightening of prudential regulations since the mid 1990s and a number of bank failures between 2000 and 2002 reduced the number of banks from 74 in 1994 to 21 currently. Strengthening of the regulatory and supervisory framework also resulted in improvements in system wide asset quality, capitalization, and profitability, which are high in comparison with other transition countries (Floerkemeier, 2006). However, most banks are still very small, exhibiting low indicators of banking productivity, such as the amount of deposits and assets per employee. The Armenian banking system has also undergone significant changes in its market structure. Since 2001, all banks have been privately owned. Foreign participation in the banking system increased following the removal of limits on foreign ownership in the mid 1990s. Foreign banks now account for 60 percent and 70 percent of system wide loans and deposits, respectively. However, with one key exception, foreign investors were mainly Armenian Diaspora individuals without the benefit of international banking expertise and reputation. Some banks from other CIS countries have also entered the Armenian market by taking over local banks, but until mid 2006, only one first tier international bank operated in Armenia. Market concentration, as measured by the Herfindahl Hirshman Index (HHI) and concentration ratios (share of the top three banks) in loans and deposits, has been low or moderate. 3 The top three banks account for about 35 40 percent of the system’s assets and loans. Market concentration as measured by the HHI for deposits is higher than that for loans (Figure 2.1) mainly due to the dominant position of the only first tier international bank in demand deposits, with a market share in this segment of more than 50 percent. However, as a result of market segmentation, concentration of credit on individual economic sectors and on demand deposits is significantly higher than the overall concentration measures suggest. Market concentration is high in agricultural and mortgage loans and moderate in lending to the transport and communication, and construction sectors. Most banks are only active in two to four of the nine main economic sectors, resulting in generally high HHI credit portfolio specialization indices.
Figure 2.1 Market Concentration in the Armenian Banking Sector
3 The U.S. Department of Justice considers markets in which the HHI is between 1000 and 1800 points to be moderately concentrated, and those in which the HHI is in excess of 1800 points concentrated.
Sources: Central Bank of Armenia and ARKA As mentioned earlier, interest rate spreads in Armenia are significantly higher than in other transition countries. One reason for this is the small size of the financial sector, suggesting unrealized economies of scale. Other factors explaining high spreads include low intermediation activity (as measured by loan to deposit ratios). Competitive pressures appear to be limited, which possibly allows banks to charge above market spreads, resulting in internationally high returns on assets. Problems with asset quality, on the other hand, are less likely to play an important role, as nonperforming loan ratios are low by regional comparison (Figure 2.2). Figure 2.2 Regional Comparison of Determinants of Interest Rate Spreads, 2005
Source: IMF Focusing solely on the aggregate banking sector indicators may, however, be misleading as bank activities, market shares, banking productivity, profitability, and financial soundness indicators vary greatly among individual banks. Banks differ greatly in size, and customer base. Around half of all Armenian banks are very small with limited market shares in deposits and loans . A number of these banks concentrate their activities in trade financing, money transfers, and investment banking, rather than in deposit taking and loan making. Anecdotal evidence suggests that several small banks serve as “pocket banks” of enterprise groups or wealthy individuals, which use them for treasury operations, or as sources of cheap liquidity, and equity investment. Other banks serve a very large number of small scale depositors and borrowers. While they may have comparatively large shares of the overall deposit and credit markets, the size of the average financial service per customer is extremely small, with average loans and deposits of US$200 per customer.
III. METHODOLOGY AND DATA We utilize a panel dataset of 20 commercial banks’ spreads and interest margins to empirically investigate which bank specific, market structure, and macroeconomic characteristics are the main determinants of banking spreads in Armenia. 4 Following Demirgüç Kunt and Huizinga (1999), Demirgüç Kunt et al., (2004), and others, we estimate a general class of regressions for spreads and net interest margin of the form: (1) Ii,t = α + β1Bi,t + β2Ct + β3Mt + εi,t Iit is the dependent variable (either spread or margin) for bank i at time t; B i,t is a vector of bank specific variables for bank i and time t; Ct is a vector of time varying, industry specific variables such as measures of concentration and ownership structure in the banking system; Mt is a vector of time varying macroeconomic variables; and εi,t is the residual. We control for time effects by including quarterly and yearly time dummies. The spread and margin equations are estimated with both pooled OLS and fixed effects
4 We omit banks that were under central bank administration and/or went out of business during the sample period. Of the five banks that are excluded from the sample, two went out of business, one was acquired by another bank, one became a credit organization, and one emerged from temporary central bank administration under new ownership.
regression. The latter controls for time invariant bank specific effects. Table 3.1 and 3.2 provide summary statistics and correlations for the employed variables. Table 3.1 Summary Statistics
Variable Observations Mean Median Std Dev. Min. Max. Spread 307 3.63 3.7 1.10 0.10 7.70 Margin 307 1.48 1.4 0.69 0.70 4.00 Spread2 279 12.58 12.7 4.18 4.20 27.50 Overhead 307 6.71 5.5 4.89 0.3 36.48 Non interest income 307 0.90 0.54 1.27 0.18 9.80 Bank size 307 7.11 7.07 0.38 6.03 7.88 Market share deposits 307 5.04 3.2 6.02 0.1 32.20 Capital Adequacy 309 58.28 32.4 51.24 11.2 326.50 ROA 310 2.02 2.45 5.41 38.6 24.20 Current liquidity 305 157.02 115 109.20 75.5 804.90 Liquid assets ratio 308 42.95 39.8 14.94 17 90.60 Portfolio industry 310 23.66 18.96 17.84 0 98.53 Portfolio agriculture 310 4.81 0.31 11.03 0 58.11 Portfolio consumer 310 23.55 21.14 17.18 0 84.20 Market share industry 310 5.13 4.10 5.47 0 32.65 Market share agriculture 310 5.16 0.13 14.78 0 73.65 Market share consumer 310 5.16 3.85 5.23 0 31.94 Foreign bank share loans 310 56.80 62.05 10.63 38.3 68.50 Foreign bank share deposits 310 69.87 71.45 6.95 58.4 78.30 Herfindahl loans 16 798.75 797.10 54.70 696.7 864.20 Herfindahl deposits 16 1182.61 1154.10 104.97 1055.8 1417.30 Concentration ratio (3) loans 16 0.35 0.35 0.02 0.33 0.38 Concentration ratio (3) deposits 16 0.48 0.49 0.03 0.43 0.52 Exchange rate 16 2.54 1.65 3.07 8.9 1.30 Real GDP growth 16 12.46 11.7 3.44 7.5 18.40 Inflation 16 3.57 3.19 3.22 1.05 8.76 Real money market rate 16 1.57 2.27 3.76 4.57 11.20
Table 3.2 Correlation Matrix Non share share share ROA share Spread Market Capital Market Market interest income Margin dummy Current Foreign Market industry deposits industry Spread2 liquidity Portfolio Portfolio Portfolio adequacy consumer Overhead Bank size agriculture agriculture consumer
Spread 1 Margin 0.527 1 Spread2 0.444 0.361 1 Overhead 0.217 0.189 0.122 1 Non interest 0.177 0.097 0.097 0.814 1 income Bank size 0.062 0.064 0.068 0.166 0.008 1 Market share 0.101 0.176 0.167 0.171 0.036 0.777 1 deposits Capital 0.325 0.226 0.209 0.088 0.207 0.478 0.273 1 adequacy ROA 0.247 0.384 0.123 0.116 0.221 0.108 0.075 0.036 1 Current 0.058 0.149 0.089 0.126 0.062 0.435 0.346 0.392 0.001 1 liquidity Foreign 0.247 0.407 0.205 0.208 0.266 0.037 0.225 0.297 0.231 0.148 1 dummy Portfolio 0.307 0.196 0.069 0.085 0.021 0.247 0.271 0.200 0.006 0.011 0.203 1 industry Portfolio 0.208 0.324 0.253 0.070 0.137 0.158 0.102 0.201 0.029 0.101 0.227 0.209 1 agriculture Portfolio 0.286 0.036 0.104 0.087 0.157 0.024 0.002 0.443 0.047 0.267 0.173 0.494 0.116 1 consumer Market share 0.174 0.082 0.029 0.050 0.031 0.616 0.487 0.281 0.019 0.249 0.215 0.628 0.027 0.177 1 industry Market share 0.231 0.358 0.260 0.028 0.118 0.208 0.076 0.164 0.053 0.149 0.246 0.191 0.953 0.121 0.010 1 agriculture Market share 0.195 0.132 0.135 0.056 0.159 0.361 0.234 0.538 0.044 0.248 0.014 0.183 0.072 0.641 0.324 0.117 consumer 1 Observations: 303 Foreign Foreign Real Herfindahl Herfindahl CR3 Exchange Real GDP bank share bank share CR3 loans Inflation money loans deposits deposits rate growth loans deposits market rate Foreign bank share 1 loans Foreign bank share 0.965 1 deposits Herfindahl loans 0.908 0.960 1 Herfindahl deposits 0.220 0.418 0.461 1 Top 3 banks loans 0.764 0.855 0.901 0.448 1 Top 3 banks deposits 0.760 0.830 0.911 0.561 0.816 1 Exchange rate 0.405 0.343 0.365 0.031 0.217 0.270 1 Real GDP growth 0.507 0.552 0.604 0.428 0.582 0.450 0.243 1 Inflation 0.209 0.108 0.097 0.547 0.061 0.108 0.241 0.156 1 Real money market 0.387 0.459 0.493 0.632 0.394 0.412 0.440 0.537 0.743 1 rate Observations: 16
Two measures of intermediation efficiency used in the study are spreads and net interest margins. 5 Spread is the difference between the ex post implicit interest rate charged on loans and the implicit interest paid on deposits. Net interest margin is the difference between total interest income and total interest expense divided by average assets. 6 This variable measures the gap between the amount a bank pays the providers of funds and what it receives from users of bank credit. The average ex post interest spread in our sample is 3.6 percent, while the net interest margin is 1.5 percent, compared to an average ex ante interest rate spread of 12.7 percent. The variation of spreads and margins across banks is about twice as large as the variation over time. Although the net interest margin can be interpreted as an indicator of bank efficiency, changes in margins could also be due to unrelated factors. A reduction in the net interest margin can, for example, reflect a reduction in bank taxation or, alternatively, a higher loan default rate. In addition, cross bank differences in net interest margins may reflect differences in bank activity, asset allocations, and risk tastes, rather than differences in efficiency. These issues emphasize the need to control for bank specific characteristics, to conduct robustness tests, and to use alternative measures of bank efficiency and performance. We use a variety of control variables and sensitivity checks to mitigate problems with interpreting the net interest margin variable. We use a variety of bank specific variables to explain the variation in interest margins and spreads. Overhead costs refer to the ratio of administrative expenses (including payroll and fixed assets) to total bank assets. We use this variable to capture cross bank differences in the organization and operation of the bank. If banks incur high administrative costs in the process of providing their services as intermediaries, they are likely to increase their interest spreads. Bank size is the logarithm of total bank assets. Size may be an important determinant of net interest margins and spreads if there any scale related cost or revenue advantages. Non interest income equals non interest operating income divided by total assets. Typically banks have different product mixes, which may influence the pricing of loan products. Therefore, banks with well developed non interest income sources may have lower interest margins due to cross subsidization of bank activities. Capital adequacy is defined as the ratio of regulatory capital to risk weighted assets. This variable captures banks’ overall financial soundness. Banks with higher capital adequacy ratios may be able to withstand shocks to their balance sheets, but they also give up financial leverage which may lead to lower interest margins and returns on equity. Return on assets (ROA) is defined as profits over total assets. More profitable banks may be able to charge lower interest rate spreads, or the higher spreads and margins could themselves constitute a mechanism through which banks generate profits. Current liquidity is the ratio of highly liquid assets to demand liabilities and measures the extent to which deposit takers could meet a withdrawal of funds. Banks
5 In the robustness section, we consider an alternative measure of ex ante interest spreads – the difference between the weighted average lending and deposit rates for each bank, where the weights are the relative amounts of deposits or loans contracted at the specific interest rate. However, this data series is incomplete as it only includes information on deposits and loans made during the last week of each quarter. 6Due to data unavailability, we use average total assets instead of earning assets in the denominator. with high levels of liquid assets (e.g. cash and government securities) may receive lower interest income than banks with less liquid assets (e.g. loans). If the market for deposits is reasonably competitive, greater liquidity will tend to be negatively associated with spreads and net interest margins. Deposit market share is the ratio of individual banks’ deposits to total banking system deposits. To the extent that market shares get translated into market power, banks with higher market shares may be able to charge higher spreads. Therefore, a bank that dominates the banking system may enjoy higher spreads and net interest margins than a smaller bank, after controlling for bank size related economies of scale. Following Beck and Hesse (2006), we also distinguish between bank portfolio shares of industry , agriculture, and consumer loans . Lending rates and thus spreads could reflect risk premiums that may vary across sectors, while interest margins are affected by loan losses which could again vary across sectors. In addition, we include bank market shares in industry, agriculture, and consumer loans. These variables account for market segmentation and the potential for market power in specific sectors of the economy. The dummy variable foreign indicates foreign ownership, where a bank is characterized as foreign if at least 50 percent of its capital is held by foreigners. By introducing this variable, we can test whether the average spreads and margins for foreign banks are significantly different from the average spread for domestically owned banks. We include several measures of market structure, some of which do not vary significantly over time. Foreign bank participation is the share of loans or deposits in the hands of foreign banks. As in Martinez Peria and Mody (2004), this variable is included to test whether there is a spillover effect arising from the presence of foreign banks in the system. 7 We also study the impact of bank concentration on spreads and margins by including several measures of market concentration in our estimations. Specifically, we include the Herfindahl Hirshman index for both the deposit and the lending market as well as the share of loans and deposits held by the top 3 largest banks. We interpret a positive association between concentration measures and the margin and spread variables as an indication of greater market power and less competition in the banking system. Given that the level of bank spreads and margins can be affected by the macro environment in which banks operate, we control for variables such as real GDP growth , inflation , the real money market rate and the change in the nominal exchange rate . Inflation can affect spreads if monetary shocks are not passed through to the same extent to deposit and lending rates or if the adjustment occurs at different speeds. The short term money market rate proxies the marginal cost of funds faced by banks. Finally, exchange rate risk, as measured by changes in the exchange rate, resulting from currency mismatches between dollar denominated assets and liabilities could be important, as Armenia is a highly dollarized economy. 7 For instance, it could be argued that foreign bank entry would force domestic banks to lower spreads, either because increased competition drives them to become more efficient or to relinquish some of the margins they were able to charge before. Alternatively, Dell’Ariccia and Marquez (2004) argue that when faced with foreign bank competition, domestic banks may redirect their lending to segments that are more opaque, where they have greater market power and an information advantage vis à vis foreign market entrants, allowing them to charge higher spreads.
We use quarterly bank balance sheet and income statement data compiled by ARKA News Agency and the Central Bank of Armenia for the period between the last quarter of 2002 and the third quarter of 2006. Data on inflation, output growth, and the real short term interest rate are from the IMF’s IFS database. Appendix contains a detailed description of the variables used in this paper.
IV. EMPIRICAL RESULTS The first section presents the main results for bank specific variables using pooled OLS and fixed effects regressions. The next section discusses the impact of market structure and macroeconomic variables; robustness tests are presented in the last section. In all tables, the standard errors shown are clustered by banks to allow for potential unobserved factors that cause a correlation of error terms for individual banks across time. We also include but do not report quarterly and annual time dummies in most regressions. 4.1 Basic Results Table 4.1 shows the relationship between bank level characteristics and interest spreads and margins. Columns 1–3 report regressions of interest rate spreads, while columns 4– 6 report regressions of net interest margins. Column 1 presents a pooled OLS regression, while in columns 2 and 3 bank level fixed effects are included. The contribution of bank specific fixed effects is large, accounting for roughly 50 percent and 40 percent of the total explained variation in the spread and margin regressions, respectively. Table 4.1 Interest Rate Spreads and Margins and Bank Characteristics Ex post Interest Spread Net Interest Margin (1) (2) (3) (1) (2) (3) Overhead 0.079 0.000 0.009 0.058 0.008 0.008 (0.037)** (0.030) (0.025) (0.022)** (0.014) (0.014) Non interest income 0.199 0.010 0.058 0.236 0.068 0.079 (0.148) (0.097) (0.093) (0.084)** (0.056) (0.057) Bank size 1.657 2.444 1.909 0.365 0.042 0.201 (0.853)* (0.715)*** (0.676)** (0.333) (0.395) (0.427) Market share 0.070 0.100 0.058 0.025 0.078 0.070 (0.031)** (0.042)** (0.040) (0.011)** (0.017)*** (0.016)*** Capital adequacy 0.011 0.007 0.005 0.004 0.003 0.003 (0.003)*** (0.005) (0.004) (0.001)*** (0.001)* (0.001)** ROA 0.047 0.025 0.028 0.064 0.025 0.024 (0.019)** (0.026) (0.026) (0.021)*** (0.012)** (0.012)* Liquidity 0.000 0.003 0.003 0.001 0.0005 0.0004 (0.001) (0.001)*** (0.001)*** (0.000)*** (0.000)** (0.000)** Foreign bank 0.163 0.335 0.356 0.364 0.011 0.005 (0.306) (0.293) (0.276) (0.163)** (0.226) (0.280)
Agriculture 0.043 0.004 (0.006)*** (0.003) Industry 0.007 0.005 (0.012) (0.004) Consumer 0.014 0.003 (0.005)** (0.006) Constant 15.296 21.837 17.068 0.820 3.037 4.492 (5.905)** (5.128)*** (4.834)*** (2.413) (2.830) (3.131) Fixed effects NO YES YES NO YES YES Observations 303 303 303 303 303 303 R squared 0.284 0.605 0.648 0.481 0.777 0.781 Quarterly and annual time dummies are incl uded but not reported. Clustered standard errors in parenthesis ***, **, and * denotes significance at the 1%, 5%, and 10% levels, respectively. 1/ Time dummies not included in all equations because of multicollinearity problems. Bank size and deposit market share are the main bank level characteristics explaining variation in ex post interest spreads . Larger banks generally have lower interest rate spreads than smaller banks, reflecting economies of scale. At the same time, bank spreads increase with their market shares. 8 This suggests that while larger banks enjoy scale economies, to the extent that market shares get translated into market power, banks with larger market shares charge higher spreads. The impact of bank size and market shares on spreads is not only statistically significant, but also economically important. On average, a one standard deviation increase in bank size results in a reduction of spreads by about 0.8 percentage points, while a one standard deviation increase in the market share variable increases spreads by 0.5 percentage points. The current liquidity ratio is found to be negatively and significantly associated with ex post spreads, reflecting the need of banks with a volatile funding base to hold more liquid assets. A one standard deviation increase in liquidity reduces spreads by 0.3 percentage points. Overhead costs, non interest income, capital adequacy, and ROA are not significant once bank specific fixed effects are included in the regressions. The foreign ownership dummy does not have a statistically significant impact on spreads, suggesting that foreign owned banks do not charge lower spreads than domestic banks. This result contradicts the usual findings in the literature regarding the importance of foreign bank presence in lowering spreads. Column 3 indicates that a higher share of agricultural and consumer loans in a bank’s loan portfolio is associated with higher spreads, suggesting that these loans are perceived as being riskier. In terms of economic magnitudes, banks with a 10 percent higher loan portfolio share in agriculture and consumer loans charge 0.4 and 0.1 percentage points higher spreads relative to the average, respectively. The main determinant of net interest margins is the deposit market share. In contrast to the spread regressions, the market share of banks enters negatively in the margin
8 The market share of loans (results not reported), on the other hand, does not have a significant effect on spreads and margins. Results are available from the authors upon request.
regressions. This result could possibly reflect individual banks’ aggressive pricing strategies to gain market share in the deposit market. The effect of this variable is economically important, as a one standard deviation increase in market share reduces margins by 0.4 percentage points. In contrast to the spread regressions, the impact of bank size on margins is not significant. As in the case of the spread regressions, overhead costs and non interest income are insignificant once bank specific fixed effects are included. Higher capital adequacy ratios lead to a significant reduction of margins, reflecting the trade off between keeping greater safety cushions and reaching higher margins by increasing financial leverage and investing in longer term assets. The association between margins and ROA is significant and positive. However, we also acknowledge the possibility of reverse causation, that is, banks that realize higher margins are likely to be more profitable. The liquidity variable is negatively and significantly associated with higher margins in the fixed effects regressions. The impact of capital adequacy and ROA is modest: one standard deviation increase in capital adequacy reduces spreads by about 0.2 percentage points, while variations in ROA of the same magnitude change margins only by 0.1 percentage point. In contrast to the spread regressions, the loan portfolio variables do not have a significant effect on margins. Finally, the foreign ownership dummy is insignificant with the inclusion of bank specific fixed effects. 4.2 Market Structure and Macroeconomic Characteristics Tables 4.2 and 4.3 report results for interest spreads and margins, respectively, including market structure and macroeconomic characteristics. Columns 1–7 in tables 4.2 and 4.3 examine the effect of bank concentration and foreign participation in the loan and deposit markets. Again, we employ clustered standard errors and fixed effects but quarterly and yearly time dummy variables are not included in all regressions because of multicollinearity issues. We continue to find that bank spreads decline with size and liquidity and increase with bank market shares, while margins decrease with bank market shares, liquidity, and capital adequacy, and increase with ROA.
Table 4.2 Ex post spreads, banking structure and macro variables
(1) (2) (3) (4) (5) (6) (7) (8) (9) Overhead 0.001 0.006 0.001 0.012 0.021 0.013 0.028 0.019 0.024 (0.030) (0.0 28) (0.029) (0.028) (0.770) (0.029) (0.039) (0.029) (0.024) Non interest income 0.005 0.000 0.005 0.005 0.017 0.003 0.004 0.043 0.047 (0.100) (0.108) (0.102) (0.101) (0.063) (0.101) (0.100) (0.105) (0.078) Bank size 2.452 2.562 2.615 1.742 1.313 1.732 1.737 1.597 1.391 (0.711) 3 (0.676) 3 (0.691) 3 (0.673) 2 (0.673) 2 (0.677) 2 (0.678) 2 (0.688) 2 (0.539) 2 Market share 0.099 0.078 0.092 0.072 0.043 0.073 0.069 0.074 0.052 deposits (0.041) 2 (0.031) 2 (0.038) 2 (0.044) (0.052) (0.045) (0.045) (0.032) 2 (0.045) Capital adequacy 0.007 0.010 0.009 0.007 0.005 0.007 0.006 0.008 0.004 (0.005) 1 (0.004) 2 (0.004) 2 (0.005) (0.005) (0.005) (0.005) (0.004) 1 (0.005) ROA 0.025 0.026 0.025 0.027 0.026 0.027 0.027 0.033 0.026 (0.027) (0.025) (0.026) (0.027) (0.025) (0.028) (0.028) (0.029) (0.025) Liquidity 0.003 0.003 0.003 0.003 0.003 0.003 0.003 0.003 0.003
(0.001) 3 (0.001) 3 (0.001) 3 (0.001) 3 (0.001) 3 (0.001) 3 (0.001) 3 (0.001) 2 (0.001) 3 Foreign bank 0.348 0.332 0.313 0.339 0.321 0.523 0.299 (0.283) (0.285) (0.262) (0.281) (0.297) (0.340) (0.314) Foreign participation 0.039 0.041 loans (0.025) (0.026) Foreign participation 0.033 deposits (0.023) Origin 1 0.061 0.087 (0.009) 3 (0.015) 3 Origin 2 0.011 0.007 (0.006) 1 (0.003) 2 HHI (loans) 0.003 (0.002) 2 HHI (deposits) 0.001 (0.000) 2 CR3 (loans) 7.016 (3.069) 2 CR3 (deposits) 5.547 (2.487) 2 Market share 0.064 (industry) (0.027) 2 Market share 0.009 (agriculture) (0.034) Market share 0.010 (consumer) (0.028) Exchange rate 0.004 (0.022) RGDP 0.000 (0.011) Inflation 0.018 (0.029) Money market rate 0.0004 (0.000) 2 Constant 21.520 22.354 22.339 15.755 17.154 15.733 15.615 19.193 15.253 (5.335) 3 (5.145) 3 (5.380) 3 (5.221) 3 (5.058) 3 (5.094) 3 (5.078) 3 (4.482) 3 (3.921) 3 Fixed effects YES YES YES YES YES YES YES YES YES Time dummies 1/ YES YES YES NO NO NO NO YES NO Observations 303 303 303 303 303 303 303 303 303
R squared 0.608 0.632 0.625 0.593 0.567 0.591 0.594 0.602 0.557 Clustered standard errors in parenthesis. 3, 2, and 1 denote significance at the 1 , 5 , and 1 0 percent levels, respectively. 1/ Time dummies not included in all equations because of multicollinearity problems. Table 4.3 Net Interest Margin, Banking Structure and Macro Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) Overhead 0.007 0.006 0.005 0.011 0.010 0.011 0.001 0.003 0.009 (0.012) (0.012) (0.011) (0.013) (0.015) (0.250) (0.011) (0.014) (0.014) Non interest 0.058 0.061 0.044 0.072 0.064 0.037 0.025 0.049 0.051 income (0.049) (0.054) (0.045) (0.051) (0.055) (0.042) (0.043) (0.056) (0.048) Bank size 0.154 0.170 0.340 0.080 0.069 0.468 0.504 0.219 0.209 (0.377) (0.384) (0.398) (0.368) (0.366) (0.386) (0.393) (0.448) (0.375) Market share 0.080 0.077 0.072 0.083 0.084 0.068 0.067 0.094 0.076 de posits (0.016) 3 (0.017) 3 (0.017) 3 (0.017) 3 (0.017) 3 (0.016) 3 (0.017) 3 (0.022) 3 (0.015) 3 Capital adequacy 0.003 0.003 0.003 0.003 0.003 0.003 0.004 0.003 0.003 (0.002) 1 (0.002) (0.002) (0.002) 1 (0.002) 1 (0.002) 2 (0.002) 2 (0.002) 1 (0.002) 2 ROA 0.022 0.022 0.021 0.025 0.025 0.021 0.021 0.022 0.024 (0.011) 1 (0.011) 1 (0.011) 1 (0.012) 2 (0.012) 2 (0.011) 1 (0.011) 1 (0.010) 2 (0.012) 1 Liquidity 0.001 0.001 0.0005 0.0004 0.0003 0.0004 0.0004 0.0004 0.000 (0.000) 3 (0.000) 3 (0.000) 2 (0.000) 2 (0.000) 2 (0.005) 1 (0.005) 1 (0.000) 2 (0.000) Foreign bank 0.027 0.011 0.013 0.060 0.067 0.009 0.055 (0.249) (0.227) (0.232) (0.231) (0.201) (0.169) (0.207) Foreign 0.009 0.010 participation (0.003) 3 (0.003) 2 Foreign 0.009 participation (0.004) 2 Origin 1 0.006 0.006 (0.004) (0.005) Origin 2 0.000 0.001 (0.003) (0.003) HHI (loans) 0.000 (0.001) HHI (deposits) 0.0004 (0.000) 1 CR3 (loans) 3.462 (1.331) 2 CR3 (deposits) 0.624 (0.802) Market share 0.019 (industry) (0.006) 3 Market share 0.002 (agriculture) (0.018) Market share 0.032 (consumer) (0.018) 1 Exchange rate 0.017 (0.207) RGDP 0.017
(0.007) 2 Inflation 0.021 (0.015) Money market 0.011 rate (0.013) Constant 4.639 4.623 6.096 2.513 2.167 7.635 2.135 4.623 4.243 (2.735) (3.774) (2.906) 2 (2.960) (2.710) (2.572) 2 (2.771) (3.774) (2.727) Fixed effects YES YES YES YES YES YES YES YES YES Time dummies YES YES YES NO NO NO NO YES NO Observations 1/ 303 303 303 303 303 303 303 303 303 R squared 0.756 0.775 0.749 0.786 0.777 0.749 0.741 0.786 0.760 Clustered standard errors in parenthesis. 3, 2, and 1 denote significance at the 1%, 5%, and 10% levels, 1/ Time dummies not included in all equations because of multicollinearity problems. We first examine whether changes in the market structure of the banking system have had an impact on bank spreads and margins. As in the previous regressions, we do not find that foreign banks, on average, have lower spreads or margins. However, we find that foreign banks have an indirect “spillover effect” on the overall level of spreads and margins. Column 1 in Table 4.2 and 4.3 suggests that foreign bank participation in the loan market (i.e., the share of loans held by banks that are at least 50 percent foreign owned) does not have a statistically significant negative effect on interest spreads but results in lower margins. To examine this further, we include two interaction terms between the origin of the bank and foreign market share in loans in column 2.9 The origin 1 variable is the interaction between the first tier international bank and the foreign market share variable, while origin 2 is the interaction between other types of foreign banks and foreign market share. We find that the interaction term between the first tier international bank and the foreign market share variable is negatively and significantly associated with interest spreads, while the interaction term between other foreign banks and foreign market share is positively and significantly associated with spreads. Consistent with Van Horen (2007), this result suggests that bank origin could matter for banking efficiency, with only the western foreign bank having a spill over effect in lowering interest rate spreads. However, the presence of other foreign banks is associated with higher spreads. In the margin regressions, this effect is found to be insignificant (column2, Table 4.3). A similar result is obtained when we consider the foreign market share in deposits (column 3 of Tables 4.2 and 4.3). Foreign bank participation in the deposit market does not have a statistically significant effect on interest spreads but results in lower margins, again providing evidence of the spill over effect. However, as before, the interaction terms between the bank origin and foreign market share variable are again significant in the spread regressions, suggesting that the presence of first tier international banks is important for banking efficiency.
9 We do not include the foreign bank dummy in the regressions in columns 2 and 3 due to potential multicollinearity with the origin variables.
Higher bank concentration, as measured by the Herfindahl Hirshman index (HHI) (columns 4 and 5 in Tables 4.2 and 4.3) in both the loan and deposit markets, raises spreads significantly, while a higher HHI in the deposit market is positively and significantly associated with margins. 10 A one standard deviation increase in the HHI for loans results in an 0.2 percentage point increase in spreads, while one standard deviation increase in the HHI for deposits results in a 0.1 percentage point increase in spreads. Similar results are obtained when we consider the shares of loans and deposits held by the top 3 banks in the case of spreads, and the share of loans held by the top 3 banks in the case of margins. Bank market shares in loans to the industrial sector are found to be significantly and negatively associated with lower spreads, consistent with lower risks of loans to a sector with “easy” collateral. At the same time, industrial loan market shares are positively associated with margins (column 8 in Tables 4.2 and 4.3), possibly reflecting lack of competition due to market segmentation within the industrial loan market. Finally, market shares in consumer loans are significantly and positively associated with higher margins. This could be an indication of market segmentation and high bank switching costs in a credit market that is characterized by relationship banking. Next, we extend the model specification with four key macroeconomic variables. The results in column 9 of Tables 4.2 and 4.2 indicate that higher real GDP growth is positively and significantly associated with higher margins, reflecting better investment opportunities and lower default risks in a high growth environment, but this effect is insignificant in the spread regressions. Likewise, the inflation rate has a small but significant effect on margins. The exchange rate enters positively and significantly in the margin regression, suggesting a narrowing of margins in periods of the exchange rate appreciation. The coefficient of the inflation adjusted money market rate is significant and weakly positive in the spread regression, indicating some effect of funding costs. Overall, the economic effects of the macroeconomic characteristics on spreads and margins are very small.
4.3 Robustness In this section, we briefly discuss whether the reported findings are robust to alternative definition of spreads, the treatment of outliers, and the inclusion of other variables. As discussed in footnote 9, spread2 is the difference between ex ante contracted loan and deposit interest rates, and is computed as the difference between the weighted average lending and weighted average deposit rates. As can be seen from column 1 in Table 4.4 the results are consistent with those found for the ex post spread variable. We continue to find that larger banks have lower spreads, while greater market power, as measured by a higher market share of deposits, is positively and significantly associated with higher spreads. In contrast to the regressions for spreads and margins, the foreign ownership dummy is now significant and enters the equation with a negative sign, that is, majority foreign owned banks appear to have lower ex ante spreads. However, this result is not robust in alternative specifications.
10 We do not include the HHI for loans and deposits variables with the relevant foreign participation variables simultaneously in the same regression as they are highly correlated with each other, with correlation coefficients of 0.9 for the loan market.
Table 4.4 Robustness Test
(1) (2) (3) (4) (5) (6) Dependent variable Spread2 Spread Margin Spread2 Spread Margin Overhead 0.059 0.029 0.010 0.029 0.002 0.007 (0.088) (0.070) (0.020) (0.156) (0.031) (0.012) Non interest income 0.089 0.022 0.027 0.162 0.012 0.065 (0.279) (0.183) (0.051) (0.435) (0.130) (0.056) Bank size 7.568 2.214 0.472 9.417 2.303 0.052 (2.349) 3 (1.115) 1 (0.420) (2.120) 3 (0.735) 3 (0.386) Market share deposits 0.351 0.085 0.059 0.419 0.103 0.073 (0.114) 3 (0.052) (0.017) 3 (0.115) 3 (0.041) 2 (0.017) 3 Capital adequacy 0.002 0.008 0.003 0.004 0.004 0.001 (0.018) (0.005) (0..001) 1 (0.015) (0.005) (0.001) ROA 0.024 0.059 0.035 0.051 0.009 0.017 (0.042) (0.015) 3 (0.014) 2 (0.029) 1 (0.031) (0.012) 1 Current liquidity (LR2) 0.004 0.006 0.001 0.011 (0.004) (0.003) 2 (0.001) (0.013) Liquid assets ratio (LR1) 0.005 0.015 (0.009) (0.004) 3 Foreign bank 1.269 0.435 0.023 0.522 0.269 0.126 (0.666) 1 (0.291) (0.218) (0.695) (0.253) (0.114) Constant 71.375 22.443 6.622 88.108 21.351 3.754 (17.877) 3 (8.232) 2 (3.060) 1 (17.227) 3 (5.352) 3 (2.747) Fixed effects YES YES YES YES YES YES Time dummies YES YES YES YES YES YES Observations 274 269 269 249 305 305 R squared 0.462 0.637 0.812 0.452 0.564 0.794 Clustered standard errors in parenthesis. 3, 2, and 1 denote significance at the 1%, 5%, and 10% levels, respectively. To test the sensitivity of our results to outliers, we estimate the basic fixed effects regressions reported in Table 4.1 by excluding 3 of the smallest banks in our original sample. These institutions are among the smallest banks in the sector. In contrast to other banks, these banks focus on fee generating investment and private banking activities rather than deposit taking and loan making. Columns 2, 3, and 4, report the results for all three independent variables. 11 The basic thrust of our results remains unchanged. In columns 5 and 6, we replace the current liquidity ratio variable with the liquid asset ratio (ratio of liquid assets to total assets). In both equations, the coefficients have the expected negative sign, indicating that more liquid banks receive lower returns on holding cash or securities than banks that invest a higher share of their funds in loans. However, as opposed to the current liquidity variable, the liquid assets ratio has only a significant impact on margins, but not on spreads.
11 We also ran regressions with the share of nonperforming loans as well as loan loss provisions in some specifications. However, these variables were found to have no impact on spreads and margins. The results are not reported here, but are available upon request.
V. CONCLUSIONS In this paper, we use a bank level panel dataset to examine the determinants of ex post interest rate spreads and net interest margins in Armenia over the period 2002–06. Our results provide evidence of the important role that bank specific characteristics and market structure play in explaining the variation of interest rate spreads and margins. In contrast to findings from other transition and developing countries, the presence of foreign banks was not found to directly contribute to lower spreads and margins in Armenia. However, we find evidence of a spill over effect of foreign bank presence on interest spreads and margins. Specifically, we find that foreign bank origin matters for banking efficiency, with the presence of the first tier western foreign bank having a spill over effect in lowering spreads. Banks from developing countries or other types of foreign banks tend to charge higher spreads. The results suggest that there is a large potential to increase cost efficiency and competition in the banking system, including by encouraging further western bank entry. The large explanatory power of time invariant bank level fixed effects in the panel analysis may be an indication of high market segmentation and low competition in the banking sector. Improving information on both sides of the market, that is, information about borrowers for banks and information about banks for depositors, is likely to contribute to a more competitive banking market and declining market segmentation. The introduction of a credit registry in 2003 and the establishment of a private credit bureau, which became operational in 2007, have been important steps towards improving information sharing on the creditworthiness of borrowers in Armenia. These will serve to increase borrower discipline by reducing moral hazard, and at the same time mitigate the information monopoly that banks have over their existing clients, thus reducing bank switching costs. At the same time, increased public information about banks as well as bank ratings by international rating companies in Armenia could contribute to improving market transparency, the functioning of market forces, and depositor confidence.
Appendix Variable Definitions
Variable Name Description Difference between total interest income received in quarter divided by average loans for the quarter and total interest Spread expenses in quarter divided by average deposits for the quarter. Difference between total interest income and expenses over Margin average assets Difference between the weighted average lending rate and the weighted average deposit rate for each bank and each quarter Spread2 where the weights are the relative amounts of deposits or loans contracted at specific interest rates. Overhead Administrative costs over total assets Non interest income Non interest operating income divided by total assets Bank size Log of total bank assets Bank deposits divided by total commercial bank deposits in Market share deposits economy Capital adequacy Ratio of regulatory capital to risk weighted assets ROA Profits over total assets Current liquidity Liquid assets over demand liabilities Portfolio (i ndustry, agriculture, Bank loans to sector over its total loan portfolio consumer) Foreign Equals 1 for foreign owned bank; 0 otherwise Foreign bank participation Share of loans/deposits in the hands of foreign banks loans/deposits Herfindahl index loans/deposits Sum of squared loan/deposit market shares of banks Concentration ratio (3) loans/deposits Share of three largest banks in loans/deposits Market share (industry, agriculture, Bank loans to sector divided by total commercial bank loans consumer) to sector GDP growth Quarterly real GDP growth Inflation Quarterly change of the CPI index Real money market rate Quarterly money market rate adjusted for inflation Change in the exchange rate Quarterly change in Dram per US$
REFERENCES Beck, T., and H. Hesse (2006), Bank Efficiency, Ownership and Market Structure. Why are Interest Rate Spreads so High in Uganda? , World Bank Working Paper No. 4027, Washington: World Bank Bonin, J., I. Hasan and P. Wachtel (2005), Bank Performance, Efficiency and Ownership in Transition Countries , Journal of Banking and Finance, 29, 31–53 Claessens, S., A. Demirgüç Kunt H. and Huizinga (2001), How Does Foreign Entry Affect Domestic Banking Markets? Journal of Banking and Finance, 25, 891–911 Claeys, S. and R. Vander Vennet (2005), Determinants of Bank Interest Margins in Central and Eastern Europe: A Comparison with the West , Ghent University Working Paper No. 316, Belgium: Ghent University Dell’Ariccia, G. and R. Marquez (2004), Information and Bank Credit Allocation , Journal of Financial Economics, 72, 185–214 Demirgüç Kunt, A. and H. Huizinga (1999), Determinants of Commercial Bank Interest Margins and Profitability: Some International Evidence , World Bank Economic Review, 13, 379–408 Demirgüç Kunt, A., L. Laeven and R. Levine (2004), Regulations, Market Structure, Institutions, and the Cost of Financial Intermediation , Journal of Money, Credit and Banking, 36, 593–622 Floerkemeier, H. (2006), Armenia’s Financial System: Why is it Small, and What Reforms are Needed? , IMF Selected Issues Papers, SM/06/434, Washington: International Monetary Fund Fries, S., D. Naeven and P. Seabright (2002), Bank Performance in Transition Countries , EBRD, London: Economic Bank for Reconstruction and Development Fries, S., and A. Taci (2005), Cost Efficiency of Banks in Transition: Evidence from 289 Banks in 15 Post communist Countries , Journal of Banking and Finance, 29, 55 – 81 Hasan, I. and K. Marton (2003), Development and Efficiency of the Banking Sector in a Transitional Economy: Hungarian Experience , Journal of Banking and Finance, 27, 2249 2271 Martinez Peria, M. and A. Mody (2004), How Foreign Participation and Market Concentration Impact Bank Spreads: Evidence from Latin America , Journal of Money, Credit, and Banking, 36, 511–537 Van Horen, N. (2007), Foreign Banking in Developing Countries: Origin Matters , Emerging Markets Review, 8, 81 105 PRODUCTIVITY AND SOURCES OF ENTERPRISE LEVEL EFFICIENCY IN ARMENIA
Karen Grigorian, Ph.D, Economist, World Bank Vahram Stepanyan, Ph. D, Economist, World Bank *
Abstract: This study reviews in detail the sources of productivity growth in the Armenian economy and suggests policies to encourage sustainable growth in the future. It summarizes the theoretical frameworks and empirical findings on the determinants of enterprise productivity in developed and transition economies, highlighted by a number of key factors including competition, foreign direct investment and the institutional environment. The central part of the study comprises the empirical analysis of a sample survey of 300 Armenian firms in the manufacturing and service sectors during 2003 2005 to provide estimates of total factor productivity (TFP). The study suggests that Armenia is not experiencing significant growth in technical efficiency, and possibly in TFP, contrary to the arguments of macroeconomic papers. Technical progress at the level of firms seems to be playing no major part in the Armenian growth process. The results suggest no significant difference in TFP on average between the industrial and service industries. The study finds virtually no difference in average levels of TFP between firms that export and those that do not, indicating that competitive pressures in export markets have not acted to raise productivity in Armenian firms. This indicates the urgent need for policy to ensure faster dissemination of new technologies and know how and the improvement of labor skills. The mean estimated values of TFP are virtually identical in foreign owned and domestic firms, suggesting serious weaknesses in the institutional environment which prevent foreign owned firms from transferring their technology and know how to Armenian companies. There is at best mixed evidence that competition is having a positive effect on TFP because competitive pressures are not yet strong enough in Armenia to influence managerial decisions. The strong macro economic performance of the Armenian economy is probably based primarily on the successful implementation of reforms at the start of transition, notably price liberalization, opening the economy to trade and reliance on the private sector. However, the time has come for a second phase of reforms to enhance the institutional environment, increase competition in the economy and encourage deeper penetration of foreign direct investment.
JEL Classification: D24, O47, C3, O43, Keywords: Armenia, productivity, stochastic frontier analysis * The authors are grateful to Prof. S. Estrin (LSE) for his valuable guidance and useful suggestions.
I. INTRODUCTION
Armenian firms have, for some years, been operating in an environment where widespread economic reforms have been implemented, and this has laid the foundations for a strong macro economic performance. As a result, the economy has maintained strong economic growth for more than a decade, though employment (in the formal sector) has not been rising and investment rates have been moderate. This suggests that efficiency gains may have been one of the important engines of economic growth (e.g., The World Bank, 2007, henceforth The Caucasian Tiger) via both the reallocation of labor from low to high productivity uses within and between sectors and from technical advance which may have increased technical efficiency of production over time. In this study, we try to understand in detail the sources of productivity growth in the Armenian economy, in order to be able to develop policies which encourage sustainable growth in the future. In this paper we explore in detail the sources of productivity growth in the Armenian economy. First, we review the growing literature, both theoretical and empirical, on the potential determinants of firm level productivity. We survey this literature focusing on findings of potential relevance to Armenia. We then move to consider the economic situation in Armenia at the aggregate level with respect to these determinants, in particular, competition, ownership and foreign direct investment. Next we examine a sample survey of approximately 300 Armenian firms in the manufacturing and service sectors in the years 2003 to 2005 to provide estimates of total factor productivity (TFP). We will use these estimates to explore the determinants of productivity and to see how these differ from other economies. We also investigate the pattern of variation in TFP in firms across industries, time and region over this period. The findings can be used to identify the areas of greatest potential in terms of productivity, and those where greater efforts will be required to bring the firms to the productive frontier. We base our formulations on studies of productivity in other transition economies, including those of Central and Eastern Europe (CEE) and the former Soviet Union. This will allow us to compare and contrast our findings concerning the factors influencing productivity in Armenia with the paths followed in comparable economies elsewhere. The paper has five further sections. In the following section we survey the large literature concerning the definition and the determinants of productivity growth. The former helps us to identify appropriate estimation methods for calculating productivity, namely estimation of TFP via stochastic frontiers. The latter provides us with a number of hypotheses concerning the possible determinants of productivity differentials in Armenia, in particular, with respect to competition and the impact of foreign direct investment. In the third section we present information on the movements of the key determinants of productivity in Armenia, before outlining in the fourth the estimation methods used in this study and the data employed in our empirical work. The results of our estimations are reported in the fifth section, which provides estimates of TFP by year, industry and region as well as explores the principal determinants of company level total factor productivity in Armenia. We bring together the findings and draw policy conclusions in the sixth section.
The strengths of this study are its ability to derive analytically advanced and robust measures of company efficiency across different industries and time, as well as the ability to explore, based on the theoretical literature and the plethora of studies for other transition economies, carefully developed hypotheses concerning the determinants of differential productivity performance in Armenian firms. This work can provide a sound basis for the development of policy to encourage the enhancement of productivity in Armenian firms. However, it is also useful to note the limitations of this Paper. Our research is based on a relatively short sample (three years) and hence is not well suited to the evaluation of long term trends. Moreover, the strength of the Paper – its ability to identify firm level productivity and its determinants is also its weakness. Because it is based on a survey of existing enterprises, it is not designed to explore some of the issues which have been given stress in the productivity literature, for example, the role of so called “creative destruction” as proposed by Joseph Schumpeter (1942) whereby productivity within a sector is increased by the entry of new and relatively more efficient firms. Because our sample is constructed to cover a group of firms that survive throughout the sample period, we exclude all entrants and exiters and hence cannot explore the impact of “creative destruction” on productivity growth. II. MEASURING COMPANY PRODUCTIVITY AND ITS DETERMINANTS
2.1 Estimating Total Factor Productivity
There are two approaches to estimating TFP in the literature, and we employ both in this Paper. The first is based on the Solow production function and develops Tornquist indexes by constructing the residual in a production function calculated at the mean of the sample. Hence one directly estimates TFP by estimating a production function. The TFP is the residual in this function, calculated for example for the Cobb Douglas specification as the difference between actual output (value added) and predicted output using the factor shares of labor and capital as the weights for the two inputs. An early example of this approach is by Jorgenson, Gollop and Fraumeni (1987) for the US. For developing countries, the work was extended by Chenery, Robinson and Syrquin, in their path breaking 1986 book, "Industrialization and Growth". For transition economies, the approach has been used especially to analyze the impact of privatization and there are well over 100 studies using some sort of production function method (e.g., Djankov and Murrell, 2002). The second approach to identifying TFP is via the estimation of frontier production functions. The idea here is to compare the efficiency of firms, on average, with a constructed composite representing the most efficient producer. Thus one estimates the relative distance for a firm between its actual combination of inputs and the predicted input combination for a given level of output that would be required by the estimated “most efficient” firm. The approach is to calculate, using information about the input output combinations of the most efficient firms, the technology frontier and then measuring TFP in other firms as a distance from this, controlling for capital intensity. Estimation is usually undertaken using non parametric estimation methods (e.g., Schmidt, 1985; Cornwall, Schmidt, and Sickles, 1990). This approach has been used less frequently in the analysis of developing and emerging markets, perhaps because of data limitations, but it provides a clear picture of the dispersion of efficiency within each sector.
Using either approach, but more commonly with the Solow method, one can identify the factors that influence TFP. Using the Solow approach, one can estimate simultaneously a second round regression which includes potential determinants on the right hand side. In the frontier method, one can either compare the estimates of mean (median) TFP in subsamples defined across determinants or include these as independent variables in the “augmented” frontier production function estimates. We used both approaches in this study, but have tended to report the findings from the former method because given the sample size these are probably more reliable for inference. 2.2 Determinants of TFP The Western literature on the determinants of enterprise productivity has tended to focus on the role of competition, organizational factors (notably ownership and the effects of foreign direct investment) and institutional environment, including government regulations. These factors have also proved to be important in transition, so it is upon these and the findings of the huge literature on the determinants of TFP in transition economies that this brief survey will concentrate. One can therefore identify three factors in the transition literature that are expected to influence the level and rate of change of total factor productivity: competition, ownership (notably foreign ownership), and characteristics of the firm (often proxied by the age of the firm).
A. Competition It is usually hypothesized that company efficiency will be positively affected by the competitiveness of the market in which the firm operates. The argument is that monopoly power provides firms with a margin of comfort which they exploit by failing to undertake the organizational changes necessary to maintain and increase TFP. Competition puts pressure on management because they have to work harder to maintain their profits against the encroachments of other firms and potential entrants. This leads them to undertake more efficiency enhancing activities, for example, innovation introducing new production methods and new products, and rearranging organizational structures so as to keep performance ahead of other firms in the market. These pressures are necessarily reduced in a less competitive environment, where managers, because they do not face the same “struggle for survival”, may become complacent or indolent. In a closed economy, the forces of competition derive principally from the domestic market. In this study, we proxy these buy industry dummy variables. They may also derive from international competition. Thus firms which operate primarily in overseas markets, exporting the bulk of their output, may face much greater competitive pressure and this may force them to enhance their TFP relative to firms facing competitive pressures only in their domestic markets. Finally, firms that are innovating new products may also be more productive, in that their lower cost base allows them to successfully bring to market new products. Alternatively, high rates of product innovation may instead indicate firms that are struggling to survive given their current levels of TFP, and represent more a signal of desperation than efficiency.
B. Private Ownership The privatization policies of transition economies are based on the hypothesis that privately owned firms will have higher TFP than state owned ones, because of their superior corporate governance. There is very considerable evidence for this argument (e.g., Djankov and Murrell, 2002). The higher levels of efficiency arise because it is easier in a privately owned firm to align the interests of owners and managers and to reduce the problems of asymmetric information between owners and managers which lie at the heart of the corporate governance problem (e.g., Estrin, 2002). With private ownership, monitoring can be efficient and the flows of information transparent either via direct owner representation on boards or through the competitive pressure of the stock market. Moreover, bankruptcy laws impact on private but not state firms, and the absence of a bankruptcy threat can lead to “soft budget constraints” that distort managerial incentives. The structure of ownership is in principle a very important issue in Armenia as in all transition economies, but is not of direct empirical relevance for our study because all the firms in our survey are privately owned. C. Foreign Direct Ownership
In the transition context foreign ownership has proved to be the most significant source of productivity gains. One can consider the impact at three levels: • On the macro economy • On the firms which have been bought by new foreign owners • On other firms in the economy, through either horizontal or vertical spillovers. Since this is such an important area, we will consider each of these impacts in turn. The impact of FDI on host economies is complex as foreign investors interact with, and thus influence, many local individuals, firms and institutions. However, on average, the net effect may well be less than many observers expect. Figure 2.1 outlines various channels of impact. Any FDI project closely interacts with local businesses; most of the impact on the host economy is transmitted through this interaction. Beyond this, FDI also impacts on other aspects, including macroeconomic variables, the host economy’s institutional framework as well as the natural and social environment. Most of these interactions are bilateral. On the one hand, foreign investors adapt to the local institutional, social and natural environment in designing their strategies. On the other hand, they would – intentionally or not – influence the environment through, for instance, political lobbying, setting good examples of labor standards, or polluting the environment. The FDI project in turn is designed by a Multi National Enterprise (MNE) located outside the country. The structure and strategies of this MNE thus shape the project and its interactions with the local environment. Macro Effects
FDI is argued to influence the main macro economic variables of concern to policy makers: balance of payments, employment, and gross domestic investment. FDI is commonly believed to have a positive effect on each of these variables. At the country level, scholars have attempted to relate the inflow of FDI to macroeconomic growth in terms of GDP on the basis of endogenous growth models (e.g. Borensztein, Gregorio and Lee 1998). They find a complementary effect of
countries’ absorptive capacity, measured by proxies for human capital, which positively moderates the relationship between FDI inflows and GDP growth. In particular, a minimum threshold level of human capital is required to benefit from inward FDI. Balasubramanyan, Salisu and Sapsford (1996) differentiate countries by their trade openness, and find that FDI has a more positive effect on economic growth of countries with export oriented trade regime compared to countries with import substitution type trade regimes. Figure 2.1 Channels of Impact of FDI Parent MNE FDI Project Knowledge Local Firms = country of origin = subsidiary role Linkage = intra industry spillovers = industry = mode of entry = inter industry spillovers = organizational = centralization Competition = absorptive capacity centralization = knowledge = entrepreneurship = clusters = size & experience management
Natural Social Issues Institutions Macro economy environment = ’ethical’ business = policy framework = balance of = pollution havens practices = FDI laws payment = global standards = labour standards = competition laws = capital stock ... = wages = educational system = employment ... ... ... Source: Meyer (2004)
More recently, Li and Liu (2005) examine the macroeconomic relationship taking account of the endogeneity of FDI, that is foreign investors are likely to seek locations with higher economic growth as well as contribute to this growth. Their empirical study shows that such an endogenous relationship has increased over the time period of their study from 1970 to 1999. This endogeneity reinforces the complementary relationship between FDI and local human capital in promoting economic growth. Horizontal spillover effects to local firms Hirschman (1958) argues that poor countries would benefit from pursuing unbalanced industrial growth promoting, in particular, the developing of industries with strong backward and forward linkages. The benefits that local firms may attain arise through several channels: • Demonstration effects work through the direct contact between local agents and MNEs operating at different levels of technology. After observing an innovation adapted to local conditions, local entrepreneurs may recognize their feasibility, and thus strive to imitate them. As local businesses observe existing users,
information about new technologies and business practices is diffused, uncertainty is reduced, and imitation increases. • Foreign investors affect local businesses not only through productivity effects, but in a variety of other ways. For example, scholars have investigated market access benefits generated by foreign investors (e.g., Aitken, Hanson and Harrison, 1997; Greenaway, Sousa and Wakelin, 2004). The rationale of this literature is that MNEs would directly or indirectly share knowledge on how to operate in international markets, by building trade channels, and by enhancing the country of origin reputation. • FDI contributes to human capital formation, especially through training and labor mobility. Trained local employees may move to locally owned firms or set up own entrepreneurial businesses. MNEs typically pay salaries above local standards to discourage highly trained employees from leaving, yet they may not oppose such movements if the new firms become business partners. Many successful local firms trace their origins to entrepreneurs or top managers that had prior links to MNEs (e.g., Altenburg, 2000). Even where few employees move, those that move may make a substantive contribution to local business. • FDI may help local firms to access export markets. MNEs are more likely to share general trade knowledge, as it is less industry specific and not part of their core capabilities and its diffusion to local businesses does not endanger their own competitive advantage. Moreover, foreign investors may help building trade channels and a country of origin reputation that local followers may use for their exports (e.g., Altenburg, 2000). • Foreign investors may support local supplier industries and markets for specialized inputs, such as labor and materials. Beyond the quality of physical products this may enhance in particular the quality of services provided by suppliers, such as just in time delivery and low default rates. With these improved inputs, local firms in turn may enhance their productivity. Negative spillovers on local firms are also possible, notably through crowding out effects. Foreign investors may gain market share at the expense of local firms. This would leave the local firms, at least in the short run, with excess production capacity and thus low productivity and low profitability. Moreover, foreign investment may source internationally and thus weaken the local industry’s domestic supplier base. These seem less likely to be relevant than the positive effects in the case of Armenia which is a relatively small and open economy with low levels of FDI. Perez (1997) offers an evolutionary model of technology spillovers that depend on the absorptive capacity of local firms and are inversely related to the technological gap; yet received spillovers influence the market share dynamics between local and foreign competitors. He thus suggests that strong local firms would benefit from competition from foreign investors, while weak firms are likely to be crowded out completely. Markusen and Venables (1999) analyse the relationship between the number of foreign owned firms and the number of domestic firms under a range of assumptions and scenarios. They argue that, under certain conditions, entry of foreign investors would trigger entry of new domestic firms in vertically related industries. The research question that has probably attracted most empirical research is horizontal spillovers, in particular the productivity benefits that local firms attract from foreign investment in their industry. This literature bypasses the fact that knowledge flows are
not measurable directly by estimating local firms’ productivity as a function of, among other factors, the presence of foreign investors in the industry. This stream of research was initiated by Caves (1974), and in 2007 we counted over 60 studies using this approach (e.g., Meyer and Sinani, 2007). This literature has evolved in several stages, notably to employ more complex datasets and more sophisticated analytical techniques, and to incorporate moderating variables that may influence this relationship. Important references include Caves (1974), Blomström and Persson (1983), Haddad and Harrison, (1993) and Aitken and Harrison (1999). Görg and Strobl (2001) review this literature using a Meta analysis of 21 studies and find that these methodological issues substantially affect the results, such that early cross sectional studies may have overstated the actual effects of FDI. Moreover, they point to important variations of spillovers across countries.
Vertical spillover effects to local firms Local firms may benefit from vertical linkages in a supply chain, benefiting from knowledge transfers to suppliers and customers. MNEs may make a deliberate effort to improve the quality of local suppliers, especially for components that cannot be cost efficiently imported due to high transportation costs or where the local industry has a natural cost advantage (e.g. for labor intensive components). These effects benefit also firms in other industries, for instance providers of business services, such as accounting or legal services. Similarly they may support their customers, for instance by providing training in sales and marketing. We have less empirical evidence on this matter, mainly because the datasets required to analyze vertical interactions along the supply chain are fairly complex and hard to obtain. Lall (1980) provides the first major study on vertical spillovers. Building on Hirschman (1958), Lall develops the theoretical arguments on why backward linkages would emerge, and he provides probably the first systematic empirical evidence. Javorcik (2004) employs industry level input output data from Lithuania, and finds higher productivity in industries which are suppliers to industries with high foreign presence. This productivity effect is larger when the foreign investors are domestic market oriented rather than export oriented. At the same time, she finds no evidence of spillovers within the same industry. Sources of Variation in FDI Impact
Multinationals vary in their internal operations, including for instance the centralization of decision making, organizational cultures, and human resource management practices. Consequently, subsidiaries in emerging economies would vary in their interactions with other business units of the parent’s network. This in turn affects interactions with local businesses, for instance, the development of local supply networks, investment in human capital, employee mobility, and the stages of the value chain located in the host economy.
Some of these variations are due to industry specific features (e.g., Grosse 2005). Infrastructure FDI for instance in transport or telecommunication can greatly enhance productivity in other sectors of the economy, yet at the risk of foreign control – possibly even monopoly – if the sector is not appropriately regulated. Similar benefits and risk arise from financial sector investment. Services such as information technology operate in more competitive markets and may benefit a wide range of other business. In manufacturing, major variations arise from the need or opportunity to produce close to the market due to high transportation costs or low scale economies. An aspect of particular relevance for MNE spillovers is intra firm knowledge transfer. Knowledge sharing within the MNE is a precondition for knowledge spillovers. Typically investors would transfer ‘know how’ to their affiliates to enhance efficiency and productivity. Yet they would keep tighter control over their ‘know why’, because such knowledge could – if diffused to other firms – threaten the international market position of the firm. Knowledge spillovers would also rise with higher value added activities, such as complex manufacturing processes like making customized machinery, rather than mass assembly of, for example, garments or shoes. In particular, research and development (R&D) is commonly believed to generate positive spillovers. Traditionally, MNE would keep their R&D activities close to their home base, or locate it in leading edge clusters such as Silicon Valley. However, recent data show that R&D is increasingly located in countries such as China, India, Singapore and Brazil (UN 2005). This potentially boosts the technology flows between MNEs and local suppliers or local institutions, such as universities. These variations influence the effectiveness of government designing policies aimed to attract FDI. The literature shows that policies ought to consider explicitly what type of FDI would benefit the host economy, rather than focusing on quantitative targets for FDI. Moreover, evaluation of policies should analyze what types of investors, and with what type of projects would consider the local environment (incl. political institutions) attractive. D. Age of the Firm A large literature (e.g., Djankov and Murrell, 2002) attests to the argument that in transition economies, de novo entrants may be more efficient than existing firms. This is because new firms do not carry the heritage from the Soviet era of hoarded labor, antiquated capital and weak management. As we have noted above, privatization may improve corporate governance of former state owned firms, but former state owned enterprises (SOEs) may, for some years, face problems in their attempts to enhance performance. This is because SOEs may take considerable time to reduce employment to appropriate levels and to invest in new equipment. New firms will not face the problems of restructuring to the new market environment that is at the heart of the transition problem for SOEs. Being created from scratch, they do not inherit the structural problems over manning, underinvestment, poor quality control, weak marketing and financial control and all the other difficulties which beset SOEs and former SOEs. “Firms” under socialism did not have many of the functions of independent Western enterprises; sales, marketing, distribution, supply, finance or investment. In many cases, the inherited structures, attitudes and organisational cultures of the old state owned firms are so strong that such radical restructuring is
impossible or at least very slow. This implies that it may be easier and more successful to satisfy the demands of the market economy in entirely new organisations (e.g., Estrin, Meyer and Bytchkova, 2005). Moreover the selection process that determines the foundation of new firms may also ensure that more market focused and entrepreneurial people will lead them from the outset. We are not able to explore this issue using our Armenia sample.
2.3 Findings from the Literature on Productivity in Transition Economies The literature on the determinants of productivity in transition economies is huge, and there are a number of major surveys including Djankov and Murrell (2002) and Estrin et al, 2007. These papers each cite more than one hundred works respectively. Most papers in the literature focus on the effects of TFP of ownership and especially foreign ownership. Thus the Djankov and Murrell concluded that the effect of private ownership was positive in Central and Eastern Europe (CEE) but insignificant in the Commonwealth of Independent States (CIS). There are around 20 more recent studies that analyze the impact of ownership on TFP or rate of change of TFP in the transition economies, using value added, total product or sales revenues as the dependent variable and either dummy variables or percent share ownership as measures of different types of ownership (e.g., Estrin et al ., 2007). A number of these studies have simply examined the differential effect of state versus private ownership, while others examine the effects of other sub categories of ownership. The studies cover both the CEE and CIS regions. With the possible exception of Russia, studies usually find the effect of private ownership on TFP to be positive or non negative. Moreover, studies that break private ownership into several categories show that the overall private v. state ownership dichotomy includes different private ownership effects. Except for two of the three studies of Slovenia, all studies uniformly suggest that privatization to foreign owners’ increases efficiency. The effect of foreign ownership is strong and robust across regions. This is a very important finding for policy makers in Armenia to which we return in the conclusions. The effect of domestic private ownership on TFP is by and large also found positive in the CEE region and in Ukraine, but it is quantitatively much smaller than that of foreign ownership. This is probably because the countries of CEE are more advanced on the transition path (see e.g. EBRD Transition Report Indicators and World Bank Doing Business Indicators). Russia appears to be different from Ukraine in that Sabirianova, Svejnar and Terrell (2005) and Brown, Earle and Telegdy (2006) find with large data sets the effect of domestic private and mixed ownership to be negative or insignificant. Similarly, Commander and Svejnar (2007) use a large firm level data set from 26 transition economies and find an insignificant average (across countries) effect of domestic private ownership relative to that of the state ownership. In general, the effect of domestic private ownership appears to be more positive in the CEE region than in the CIS and it seems likely that this is a consequence of the greater progress made in institutional development. This confirms the results from findings for other emerging markets. Once again this is an important conclusion for Armenian policy makers.
It is also important to understand the factors driving technical progress; the change in company TFP. Studies that examine the change in productive efficiency show that foreign owned firms improved efficiency faster than domestic private and state owned firms in the 1990s and early 2000s. This differential effect is not detectable, however, in Commander and Svejnar’s (2007) study of the 2002 2005 panel data from the 26 transition economies. It is hence possible that foreign owners brought about a sizable increase in efficiency in the period immediately after acquiring the local firms in the 1990s, but that later on the rate of change in efficiency has been on average similar in all the principal types of ownership of firms. It is also interesting to consider the experience of China, where growth has proceeded at high levels for several decades, arguably on the basis of both increases in factor inputs and rises in TFP. There are a number of important studies of TFP in China. Probably because privatization is a relatively recent phenomenon in China, a number of studies, including Jefferson and Rawski (1996), address TFP issues with firm level data but do not examine differences in TFP related to privatization or ownership. Studies that address these issues find diverse results, with the effect of non state ownership being mostly positive but sometimes statistically insignificant and sometimes negative. Thus Jefferson and Su (2006) use a large sample of firms (N > 20,000) and show that the effect of private joint stock ownership on the level of TFP is positive. Hu, Song, and Zhang (2004) in turn use a much smaller sample of firms in selected regions (N > 700) and find the effects of cooperative as well as domestic and foreign private ownership to have a positive effect on the level of productivity. Yusuf, Nabeshima, and Perkins (2006) use a relatively large sample of firms (N > 4,000) and find the effect of domestic private, collective and complete foreign ownership on the level of productivity to be statistically insignificant, the effect of foreign joint ventures to be positive, unreformed state ownership negative, and reformed state ownership positive. Finally, Dong, Putterman, and Unel (2006) use firm level data from Nanjing (N = 165) to examine the effect on the rate of change of TFP, and they find the effect of state urban ownership to be positive, while the effect of state rural and both private urban and private rural ownership is found to be insignificant. The TFP studies of CEE, CIS and China hence generate a fairly clear overall picture which is highly relevant to Armenia. There is clear evidence that foreign ownership raises productivity. In China and CEE, the estimates suggest that private domestic ownership also raises TFP relative to state ownership but the effect is quantitatively smaller than that of foreign ownership. Russia and elsewhere in the CIS is different in that the performance effect of privatization to domestic owners is found to have a negative or insignificant effect on TFP, and that is probably because of weaknesses in the institutional framework. In addition, concentrated (especially foreign) private ownership has a stronger positive effect on performance than dispersed ownership in CEE and CIS. Data from CEE and CIS suggest that new firms are equally or more productively efficient than firms privatized to domestic owners. In view of the above results, the question naturally arises as to why the TFP effect of privatization to domestic owners has been much smaller than the TFP effect of privatization to foreign investors. Discussions with managers, policy makers and analysts suggest three leading possible explanations. The finding may reflect in part the limited skills and access to world markets on the part of the local managers. Domestically owned privatized firms are also the ones where performance reducing
activities such as looting and defrauding of minority shareholders have been most frequent. Finally, in a number of countries the nature of the privatization process initially prevented large domestic private owners from obtaining 100% ownership stakes and insiders or the state often owned sizeable holdings. It often took these large shareholders several years to squeeze out minority shareholders and in the process the large shareholders sometimes artificially decreased the performance of their newly acquired firms in order to squeeze out the minority shareholders at low share prices.
III. CORPORATE ECONOMIC BEHAVIOR IN ARMENIA The previous discussion indicates that it will be important to focus on the following determinants of TFP in Armenia: • Competition • Foreign direct investment • Institutional development In this section we briefly review the situation in Armenia, where possible against potential comparators among transition economies. 3.1 Competition There is little direct information about market structure in Armenia. As noted, in the empirical work which follows we control for market power effects by the use of industry dummy variables. However, there is some encouraging evidence on this subject. As can be seen in Table 3.1, Armenia has made considerable progress in the areas of price and trade liberalization, relative to other countries in the FSU and the Balkans. Table 3.1 Price and Trade Liberalization in Armenia and Comparators, EBRD Transition Indicators in 2007
Private Sector Price Liberalization Trade System Competition Policy Share of GDP Armenia 75 4+ 4+ 2+ Bulgaria 75 4+ 4+ 3 Estonia 80 4+ 4+ 4 Georgia 80 4+ 4+ 2
Russia 65 4 3+ 2+ Ukraine 65 4 4 2+ Source: EBRD, Transition Report.
Thus the private sector share is amongst the highest in the sample, and for both price and trade liberalization, Armenia joins the more advanced transition economies in achieving the highest possible ranking. However, the possibility of serious monopoly problems is indicated by the low score with respect to competition policy, though the levels are not out of line with other countries in the sample. We return to this issue below. 3.2 Foreign Direct Investment and Ownership
It can be seen in Table 3.2 that FDI inflows into Armenia have been relatively low in comparison with the sample of FSU economies. It should be noted that in fact the levels of FDI have been even higher for the transition economies which have acceded to the EU, for example Czech Republic, Hungary and Poland.
Table 3.2 FDI Inflows into Armenia and other economies (ratio to GDP, percent) 2007 2001 2002 2003 2004 2005 2006 Estimate Geo rgia 2.5 3.6 8.4 8.1 8.3 14.3 15.5 Estonia 5.5 2.1 7.8 5.9 15.9 3.4 4.5 Armenia 3.3 4.7 4.3 6.1 5.1 5.3 5.2 Ukraine 2.0 1.6 2.8 2.6 8.7 5.4 6.6 Russia 0.1 0.0 0.4 0.3 0.0 1.1 0.8 Source: EBRD, Transition Report . Table A.1 shows the rate of FDI inflows and GDP growth in Russia, Ukraine, Poland and Bulgaria from 1992 to 2007. It can be seen that there has been an upswing in FDI levels in the FSU economies and Bulgaria from a low base, and this is associated with an acceleration of economic growth (one should be careful not to infer causality here – the two phenomena are correlated but the causality is not clear). However, it can be seen that the increase in growth in Armenia since 1999 is not correlated with changes in FDI, indeed, FDI levels remain very low throughout. Table A.2 highlights that FDI flows remain low in Armenia, even by the standards of many economies of the FSU. 3.3 Institutional Development According to the EBRD Transition Reports, Armenia has made considerable progress in terms of institutional development. The average score across the EBRD indices is around 3.1, which is the highest in the CIS though below Lithuania (3.7), Latvia (3.64) and Estonia (3.75). It is above possible comparator states like Georgia (3.05) and also Russia and Ukraine (3.0). A richer framework can be developed from the Heritage Foundation that provides indications of “economic freedom” on a scale from zero to 100 (best). According to this assessment, Armenia’s situation in the key measures with respect to FDI and productivity look strong with respect to business and investment freedom and the overall indicator also suggests that the policy environment is rather benign. However, the assessment also indicates some deep rooted problems in Armenia with respect to the enforcement of property rights and corruption. As indicated in Section 2, these are likely to have had a negative effect on the flows of FDI. 3.4 Summary The literature on productivity in developed and transition economies places emphasis on three determinants: competition, foreign direct investment and institutional development. In this section, we have considered Armenia’s performance in these three areas, in comparison with other transition economies. We find that reforms with respect to price and trade liberalization have been very effective but Armenia lags with respect to Competition Policy. Since the Western literature stresses the role of competition in
TFP determination, this may be a source of difficulty. However, the most significant issue is the low levels of FDI in Armenia. These have remained at a low level for a number of years, despite fast growth and a modestly good institutional environment among transition economies. It seems likely that this will hinder the evolution of TFP in Armenia.
IV. ESTIMATION METHODS AND DATA 4.1 Estimation Methods
A stochastic frontier framework is employed to measure the technical inefficiency of the economic agents, either firms or industries, which are the objects of the assessment. This methodology allows us to evaluate the performance of each agent, relative to a common estimated best practice frontier of production. The estimates of this frontier are based on the performances of the other agents in the economy. In fact, according to the classical microeconomic theory, functioning markets do not tolerate inefficiency. But this axiom is easily contradicted by any empirical analysis. In the empirical context therefore, efficient producers will be those who produce as much as possible with the employed inputs. Since the analysis concerns a single output production, we can think of technical efficiency in terms of TFP. That is the ratio between actual output and the optimal value as specified by a theoretical production function. The production function is the mechanism through which the firm transforms inputs into outputs. There can be different specifications of the production function that reflect different restrictions on its properties. It is important to point out that, given this definition of technical inefficiency, the correct specification of the theoretical production function and the list of inputs are decisive in avoiding errors in its measurement. The output of the fully efficient firm can be expressed as: (1) yi = f(X i,β)exp(v i) where f(X i,β) is the theoretical production function, β is the vector of parameters of the production function to be estimated. i indexes the i th firm in the sample of N firms and exp(v i) is an unrestricted idiosyncratic and stochastic component of the model. This last component embodies measurement errors and any statistical noise and random shock of the frontier. The output of the less than efficient agent is: (2) yi = f(X i,β)ξ iexp(v i) where 0 < ξ i < 1 is the actual technical efficiency of the firm. The empirical analysis carried out with stochastic frontier models was originally proposed by Aigner, Lovell and Shmidt (1977) who made distributional assumptions on the composed error ε i = v i ui defining it as the sum of a symmetric normally distributed variable (the idiosyncrasy) and the absolute of a normally distributed variable (the inefficiency): (3) vi ~ N[0, σ v] (4) ui = |U i| where U ~ N [0, σ u] This is the half normal specification of the model. Both components are assumed to be independent and identically distributed across observations. The analysis usually focuses on the measurement of ξ i more than on the estimates of the technology parameters and it is carried out in a linear form after the logarithmic transformation of the previous model: (5) lny i = lnf(β‘X i)+ v i ui where u i is –lnξ i and is always non negative and can be interpreted as the percentage variation of observed performance from the firm’s own frontier performance (or efficient performance). The empirical analysis of inefficiency that arises from this framework requires two steps. In the first step the estimates of the technology parameters β, σ v and σ u allow to construct estimates of the composed error ε. In the second step the estimate of ξ i can be calculated using the formula proposed by Jondrow et al . (1982): (6) ξi = E(exp( ui|ε i)]. The assumption of half normality for the inefficiency term constrains the mean of this stochastic component to be equal to zero. However this assumption can be relaxed. Some authors such as Stevenson (1980) extended the model to a truncated normal distribution for ui that allows the mean to be non zero. This new and less restrictive distribution implies: 2 (7) ui = |U i| where U i ~ N [ , σ u ] The mean of the inefficiency component of the model can now be modelled to vary with some factors that the researcher considers decisive such as industry, location and so on. Formally, the inefficiency can be specified as a linear function of these factors: (8) i = 0 + θ’Z i As will be shown later, the parameters of the underlying distribution of u i provide a mechanism to introduce heterogeneity in the production in the model. The available data for our analysis of Armenian firms consist of a sample of N economic agents observed over T years so the model can be now written as: (9) lny it = lnf(β‘X it )+ v it uit where each observation concerns firm i in period t and t = 1,2,…,T.
The greater amount of information contained in the panel structure of the data can be exploited in the estimation of the production function parameters and the efficiency scores. However, while increasing the available information, panel data can exacerbate the presence of heterogeneity. The large variation of characteristics between all the sectors and firms of the Armenian economy produces a large amount of measured and unmeasured heterogeneity. As described earlier, in stochastic frontier analysis the ultimate objective is to obtain estimates of u it . The first step is to estimate the technology parameters β, σ v and σ u. If these estimates are inappropriate or inconsistent then the estimation of ε i and therefore ui is likely to be problematic as well. Heterogeneity in the production function and in the inefficiency distribution can be a cause of inconsistency in the structural parameters estimation. Heterogeneity can be either observable (if we have some measure of it or variables which capture it) or unobservable (such as the individual effects and time effects in panel data). When heterogeneity is observable, it is important to understand how it enters the model: that is whether it affects the production function or the inefficiency distribution, or it scales them in the form of heteroskedasticity. When heterogeneity is unobservable and therefore unmeasured, it is important to use an adequate panel data estimator. In order to address all these heterogeneity issues, two different treatments of stochastic frontiers with panel data have been carried out in order to find the specification of the model that better fits the described data. The first treatment is pooled cross section estimation to be performed within the stochastic frontier framework. Even if this estimation could appear restrictive because it assumes that all observations are independent and it does not exploit the within variation of panel data, it could be an interesting starting point for the analysis. With this specification, when the assumed underlying distribution is the truncated normal, it is possible to introduce heterogeneity into the distribution of efficiency by modelling the mean as: T (10) it = 0 + θ Zit where Z it can include all the factors that are likely to affect the underlying inefficiency distribution, such as binary indicators for employees training and innovative firms. Moreover this specification allows the heterogeneity to be introduced in the model through the heteroskedasticity of both random components v it and u it as, for instance, depending on dimensional variables. Modelling the mean of the inefficiency distribution and the variances of the random components gives more flexibility to the estimation and allows the introduction of full heterogeneity in the model. Moreover the estimates of efficiency are time varying. The second treatment of stochastic frontiers that is feasible with panel data is the so called “true fixed effect” model put forward by Greene (2002): (11) yit = α i + β‘X it + v it uit in which individual effects are introduced simply by placing industry dummies in the classical stochastic frontier model. There are two advantages of this technique: the industry effects are allowed to be correlated with the explanatory variables of the model and the estimated efficiency is time variant. In our empirical work on Armenia, we employed many of these techniques to calculate the measures of technical efficiency. In particular, our analysis provided estimates of technical efficiency (u it ) as well as the technology parameters β, σ v and σ u. We also addressed the heterogeneity issues using the pooled cross section approach. 4.2 Data The empirical work is based on a survey of 300 enterprises in Armenia covering the period 2003 to 2005. It focused on industrial firms in 11 sectors, including chemicals, textiles, jewelry and beverages (including juices) as well as services. It was designed to provide the basis for analysis of trends in company behaviour, including productivity. Thus, it collected information about output, labor input and factors influencing productivity such as ownership, legal status, privatization and competitive pressures (e.g. exports). The data from the questionnaire was supplemented by information from the database of the Annual Administrative and Regulatory Cost Survey of Armenia (on the business environment) and official statistics from the National Statistical Service of Armenia. The latter was used to provide information on capital input. V. ESTIMATES OF TFP AND ITS DETERMINANTS IN ARMENIA
In this section, we discuss the results of a large number of econometric exercises to estimate TFP in Armenia, to analyze the ways that it varies across industries, ownership and time and to understand the forces that drive the determination of levels of productive efficiency in Armenia. 5.1 The Stochastic Frontier Production Function in Armenia
The estimation methods used in this study are discussed in the previous section. We first report the findings based upon the estimates of TFP using stochastic frontier methods. In fact numerous specifications were used but the results of interest did not vary to any significant degree. The particular specification on which this section is based uses a truncated normal model and the results of the estimation are reported in Table 5.1. Table 5.1 Stochastic Frontier Production Function Dependent variable: log Output in 2005
Standard Standard Variables Coefficient Variables Coefficient Error Error ln Capital (k) .13 .03 Equipment 1.02 .36 ln Labor (l) .89 .05 Food .07 .35 Year2 .07 .13 Furniture .63 .44 Year3 .96 .13 Jewelry .25 .45 Cloth ing , Shoes 1.32 .37 Juices, mineral .26 .40 Construction .16 .41 Mining .26 .42 Constr uction Services .06 .36 .49 .36 Material Textile .79 .42
Standard Standard Variables Coefficient Variables Coefficient Error Error _cons 7.58 22.97 sigma_v2 1.18 .47 Mu chi2(1) = 0.13 Any_training .08 .14 Prob > chi2 = 0.71 Any_new_pr .39 .17 chi2( 1) = 0 .00 _cons .20 22.97 Prob > chi2 = 0.99 /lnsigma2 .17 .07 Number of 458 /ilgtgamma 6.25 204.70 observations sigma2 1.18 .08 Wald chi2(15) 860.14 gamma .00 .39 Prob > chi2 0.00 sigma_u2 .00 .46 Log likelihood 687.94 The estimated equation is strongly significant and the coefficients on labor and capital are plausible and highly significant. 1 An important test of whether the results are “sensible” is given by the “Test for CRTS”. The test asks whether the estimated coefficients when added sum to unity, implying constant returns to scale. The estimated coefficients add up to almost precisely unity, and the test confirms that the estimates are not significantly different from unity. This allows one to place some faith in the findings since most other studies for developed and developing economies also tend to find the coefficients summing to around unity. The economic interpretation is that future growth in Armenia deriving from increments of factor inputs labor and capital will not be constrained by the onset of diminishing returns. One should not place excessive emphasis on this finding, however, since more sophisticated specifications of technology have not been tested and would be hard to fit on a dataset of this size. 5.2 Measuring Differential Performance in Armenia
Given the stochastic frontier reported in Table 5.1, one can calculate the variation in efficiency according to a number of criteria. In this sub section, we first consider TFP overall, and then analyze the variation according to the principal drivers discussed in the literature summarized in section 2, as well as the Armenian literature. These include year, activity, legal status, foreign ownership, region and sector. A. Overall TFP Our estimates suggest that productivity in the average firm in Armenia is calculated to be 78.12% of the level found in the most efficient firm in the country. The standard deviation around this estimate is found to be around 20%, which suggests that the least efficient firms are probably operating at approximately 60% of the productivity of the most efficient firms ( if the distribution of the errors were normal). We do not know whether the most efficient firms in Armenia are efficient by international standards. We only know that the forces of competition and openness to
1 These results are preliminary and not entirely comparable with those in the literature. First, TFP estimates are usually undertaken at the sectoral level and with careful control for the quality of labor and capital inputs. The dataset was too small to permit disaggregation at this stage, though it is hoped to undertake finer estimation in the future. There are also some concerns about the quality of the data concerning capital, which was measured at historic cost. international trade, for example, have acted to keep the dispersion efficiency among Armenian firms low by international standards. Perhaps the most important point is that these estimates for Armenia are broadly comparable with those for developed and transition economies, and certainly well within the standard range. This gives us some confidence when considering the results with respect to the determinants of productivity. B. Growth in Productivity The findings with respect to productivity growth over time are reported in Table 5.2 The influential Caucasian Tiger study stressed the potential role of technical advance and increases in TFP as the basis for future Armenian growth. In this context, the mean levels of TFP by year reported in Table 5.2 may seem to be disappointing. They show that while the estimated mean level of TFP in Armenian firms was 0.7812 (78.1 percent) of the most efficient firms, this mean value did not vary to any significant extent across the three years of the study. Indeed, if the mean was changing at all, it was in a downward direction. This implies that we have not been able to identify a positive significant technical progress (increase in TFP) over the period of our study.
Table 5.2 Mean Values of TFP (te1)
By year By Export By Industry By Region Category (Marz) 2003 0,80 0 0,79 Chemic als 0,72 Arag. 0,77 2004 0,79 1 0,78 Clothing, Shoes 0,79 Ararat 0,78 2005 0,76 Constr uction 0,77 Arma. 0,79 By Activity By Foreign Construction 0,78 Gegh. 0,82 Ownership Material Industry 0,78 0 0,78 Equipment 0,77 Kot. 0,79 Services 0,78 1 0,79 Food 0,77 Lori 0,77 Other 0,77 Furniture 0,82 Shirak 0,78 Jewelry 0,82 Syun. 0,82 Juices, mineral 0,80 Vay. Dzor 0,81 Mining 0,81 Yerevan 0,78 Services 0,79 Textile 0,75 Total 0,78 Total 0,78
However, it should be noted that the time period is short and one should not expect great changes in TFP over such a short period. The international companies reported in Caves (1982) also failed to identify dynamic effects in TFP estimates. Even so, we must conclude that in this time period productivity growth as measured by TFP has not been significantly different from zero, which highlights the need for policies to accelerate its growth in the future. 5.3 Determinants of TFP in Armenia We report the variation in TFP by activity in Table 5.2. The results suggest that there is no significant difference in TFP on average between the industrial and service
industries. There is thus nothing in our findings to favor selecting industry over services or vice versa as a basis for a growth strategy founded on productivity growth. There is however, somewhat more useful information when one considers the variation in TFP across twelve industries in Table 5.2, though these must be treated with caution because the standard deviation of the technical efficiency is larger than the estimated differences. We note that, measured at the mean, technical efficiency is found to be highest in the jewelry, furniture and juice sector as well as mining. It is lowest, on the other hand, in chemicals, textiles and construction. The service sector operates more or less exactly at the mean of technical efficiency across the economy. One must take care in interpreting these results. They do not provide information about whether the most efficient firms, at the frontier, are more efficient in the jewelry or furniture sector than in the chemicals or textile sectors. We do not have information on the relative efficiency of the most efficient firms. The results inform us about the dispersion in efficiency within different sectors. They tell us that dispersion is greater in chemical, textiles and construction than in jewelry, furniture and juices. This may be because the latter sectors are more monopolistic, for example, so that inefficient firms emerge and survive in a manner that is harder in more competitive sectors. A. Exporting It might be predicted that firms subject to higher levels of competition, for example because they are selling in overseas markets which are highly competitive, would have higher levels of productivity. However this hypothesis is not confirmed in Table 5.2. In fact, we find virtually no difference in average levels of TFP between firms that export and those that do not. This appears to suggest that competitive pressures in export markets have not acted to raise productivity in Armenian firms. Given the discussion in section 2 about FDI and horizontal spillovers, this is a troubling finding for the Armenian economy. Indeed, the mean level of TFP in exporting firms is very slightly smaller than that found in non exporting firms. However, it is also possible that the appreciation of the Armenian currency is reducing the value of exports relative to domestic goods, and therefore in our frontier functions, which measure output by sales, reducing the productivity measured in drams of exporting firms.
B. Foreign Direct Investment and TFP We noted in section 3 that levels of FDI were very low in Armenia – it is therefore perhaps unsurprising that the impact of FDI on TFP is not discernable. We find in Table 5.2 that the mean estimated values of TFP are virtually identical in foreign owned and domestic firms. This runs strongly counter to the results in the literature for other transition economies, and indicates there may remain serious weaknesses in the institutional environment which prevent foreign owned firms from transferring their technology and know how to Armenian companies. 1 This negative result is probably our most important finding, and we return to the policy implications in the conclusions. 1 However the sample contains very few foreign owned firms, only 4 out of more than 300. The fact that the sampling method identified so few foreign firms is indicative that foreign ownership has not yet penetrated very deeply into the Armenian economy, and that this should be high as an objective of a second phase reform agenda.
C. Regional Effects
The analysis reveals considerable variation in technical efficiency on average across regions, though once again the differences in mean TFP are not great in comparison with the standard errors. We observe that TFP is highest on average in firms operating in Gegharqunik, Syunik and Vayots Dzor, and the levels are some 3 or 4 percent on average higher than those found in the least productive regions; Aragatsotn, Lori, Shirak and Yerevan. 5.4 Solow Production Function Results
As an alternative to the above, we next estimate a Solow type augmented production function in order to understand the determinants of the residual, technical efficiency. These results are reported because they allow a direct comparison with the transition literature. However, because the estimation method is completely different, the results cannot be compared directly with those of the previous sub section. We estimate a production function using firm fixed effects in which log output is the dependent variable and the independent variables are log labor (l), log capital (k), year dummies and a number of variables that are often identified in the literature with higher levels of TFP. These include export category (as an indicator of competitive pressure in product markets), foreign ownership and expenditure on training of labor. We report the results of the fixed effects estimation in Table 5.3. The estimated function is largely consistent with our results using stochastic frontier methods. We find that the sum of coefficients on labor and capital, is not significantly different from unity, which provides a “common sense” check on the estimation. We also confirm in a single equation most of the findings with respect to variation in TFP discussed above. Thus, the dummy variables for time (year 2 and 3 respectively) are not statistically significant, indicating that TFP did not increase over our sample period. We cannot test for activity or industry effects using this method because all such variation is removed by the fixed effects estimation. However, in this approach we find a very weak but positive effect of competition, via export category, on TFP, though it is only significant at the 10% level. Once again, we are not able to identify any positive impact of foreign ownership on TFP in these equations.
Table 5.3 Augmented Production Function Estimate (Fixed-Effects) Dependent variable: log Output in 2005
Variables Coefficient Standard Error ln Capital (k) .22 .09 ln Labor (l) .65 .11 export_category | .27 .16 fdi_category .11 .33 any_training .04 .23 year2 .03 .08 year3 .08 .09 _cons 6.70 .88 sigma_u 1.10 Sigma_e .60
Rho* .77 F test that all u_i=0: F(127, 222) = 6.56; Prob > F = 0.0 0 Number of observations = 359 Number of groups = 128 Observations per group: minimum = 1 (max=3; average = 2.8) F(9,222) = 6.94; Prob > F = 0.00 Group v ariable (i): id R sq: within = 0.22 between = 0.57 overall = 0.55 corr(u_i, Xb) = 0.08 * fraction of variance due to u_i VI. POLICY CONCLUSIONS
In this paper, we have summarized the theoretical frameworks and empirical findings in the literature on the determinants of enterprise productivity in developed and transition economies. In the empirical section, we employ a new dataset on Armenian firms to measure total factor productivity and explore its determinants as well as examine its variation across sectors. Thus, the strong macro economic performance of the Armenian economy is probably based primarily on the successful implementation of reforms at the start of transition, notably price liberalization, opening the economy to trade and reliance on the private sector. However, the time has probably come for a second phase of reforms to enhance the institutional environment, increase competition in the economy and encourage deeper penetration of foreign direct investment. This view is strengthened by the fact that in our empirical analysis, we were unable to identify a significant relationship between productivity and the quality of the labor force, as indicated for example by levels of education. This indicates that productivity growth was probably largely determined by demand rather than supply side factors. Our results obtained from a survey of 300 manufacturing and service firms provide plausible and robust estimates of production technology in Armenian enterprises. Our principal findings are: 1. The average firm in Armenia operates at some 78% of the level of productivity (TFP) of the most efficient firms. These levels are comparable to those found in both developed and transition economies. The study suggests that Armenia is not experiencing any growth in technical efficiency, and possibly in TFP, contrary to the arguments of macroeconomic papers. Thus technical progress at the level of firms seems to be playing no part in the Armenian growth process. This indicates the urgent need for policy to ensure faster dissemination of new technologies and know how and the improvement of labor skills. 2. There is at best mixed evidence that competition is having a positive effect on TFP. This is probably because competitive pressures are not yet strong enough in Armenia to influence managerial decisions, for example forcing firms to improve their efficiency to the levels of benchmark firms.
There is no evidence that FDI acts to increase productivity in Armenian firms. Foreign ownership in Armenian firms remains low by the standard of the region, and perhaps the economic and institutional environment is not adequate to ensure either adequate FDI inflows or that the benefits of FDI are reaped by recipient firms. The empirical analysis therefore reveals some divergence between the situation with respect to productivity in Armenian firms and international experience in transition and developed economies. Taken together, the comparison of the situation in Armenia with international experience suggests that the Armenian economy may benefit from a second stage of reforms so that economic development can become more firmly based on productivity growth at the enterprise level. A comparison of Armenia with the international experience highlights three inter related areas where policy development could improve the environment in a manner consistent with attaining this objective: (i) policies to improve domestic competition; (ii) policies to improve further the institutional environment, especially with respect to the enforcement of property rights and the reduction of investor risk; and (iii) policies to encourage foreign direct investment, targeted to sectors where the horizontal and vertical linkages could be maximized. It is very important that a sound Competition Policy is put in place, that it is operated by an agency which is independent of the government, and that it is quickly seen to be effective in the implementation of the Law. The first issue can be addressed quite easily by the adoption of legislation, for example, based on the regulations of the European Union. This was the method adopted by the EU Accession economies of Central and Eastern Europe. The latter two requirements can be more difficult, because they require governments foregoing considerable discretionary power in the development of industrial policy. Nonetheless, the Armenian authorities would be well advised to look to the experience of other transition economies, perhaps notably the Baltic States which have also emerged from the legal structures of the former Soviet Union. Turning to the institutional environment, we noted earlier that Armenia has performed rather well in terms of the EBRD’s Transition Indicators of institutional development, but that when one considered the deeper questions raised for example by international investors and considered in the Heritage Foundation Index, there was still considerable progress to be made. The critical issues seem to be around protection of property rights via the legal system, and these concern less the nature of the laws that have been enacted than the consistency of their enforcement. Once again, there is very considerable experience, notably in the multilateral agencies, on how to make the legal system operate in such a way as to give confidence to investors, especially foreign investors. This brings us to the third and most important area for policy development; encouraging foreign direct investment. The most striking difference between the experience of Armenia and the other transition economies studied is that, in the latter, FDI has been pivotal in productivity growth. In Armenia, productivity growth has been very slow at the enterprise level and FDI inflows have been small. It seems likely that these two facts are correlated. In terms of policy, one should distinguish between activities to improve institutional and environmental factors likely to encourage FDI,
factors that will encourage the best sort of FDI (in terms of productivity) and policies to encourage FDI.
Appendix
Table A.1 GDP real growth rates and Net FDI inflows as percent of GDP, selected countries
Armenia Russia Ukraine Poland Bulgaria Year GDP FDI GDP FDI GDP FDI GDP FDI GDP FDI 1992 41.8 0.0 14.8 NA 9.7 0.8 2.6 0.3 7.3 0.5 1993 8.8 0.2 8.7 NA 14.2 1.4 3.8 0.7 1.5 0.4 1994 5.4 1.2 12.7 0.1 22.9 0.4 5.2 2.0 1.8 1.1 1995 6.9 2.0 4.0 0.5 12.2 0.7 7.0 2.6 2.9 0.8 1996 5.9 1.1 3.6 0.4 10.0 1.2 6.2 2.8 9.4 1.4 1997 3.3 3.2 1.4 0.4 3.0 1.2 7.1 3.1 5.6 4.9 1998 7.3 11.7 5.3 0.6 1.9 1.8 5.0 3.5 4.0 4.2 1999 3.3 6.6 6.4 0.6 0.2 1.5 4.5 4.3 2.3 6.2 2000 5.9 5.5 10.0 0.2 5.9 1.9 4.3 5.4 5.4 7.9 2001 9.6 3.3 5.1 0.1 9.2 2.0 1.2 3.0 4.1 5.9 2002 13.2 4.7 4.7 0.0 5.2 1.6 1.4 2.0 4.5 5.6 2003 13.9 4.3 7.3 0.4 9.6 2.8 3.9 2.0 5.0 10.4 2004 10.1 6.1 7.1 0.3 12.1 2.6 5.3 4.8 6.6 11.7 2005 14.0 5.1 6.4 0.0 2.7 8.7 3.6 2.3 6.2 14.7 2006 13.4 5.3 7.4 1.1 7.3 5.4 6.2 2.9 6.3 23.2 2007 Est. 13.7 5.2 8.1 0.8 7.3 6.6 6.5 3.8 6.2 20.6 Source: EBRD, Transition Report.
Table A.2 Net FDI inflows as percent of GDP in CIS countries
2001 2002 2003 2004 2005 2006 2007 Est. Armenia 3.3 4.7 4.3 6.1 5.1 5.3 5.2 Ukraine 2.0 1.6 2.8 2.6 8.7 5.4 6.6 Estonia 5.5 2.1 7.8 5.9 15.9 3.4 4.5 Tajikistan 0.9 3.0 2.0 13.1 2.4 2.3 2.0 Latvia 1.4 2.7 2.3 3.8 3.6 7.5 7.9 Georgia 2.5 3.6 8.4 8.1 8.3 14.3 15.5 Turkmenistan 4.6 6.2 3.8 5.0 5.0 7.2 6.4 Lithuania 3.6 5.1 0.8 2.3 2.6 5.2 3.5 Russia 0.1 0.0 0.4 0.3 0.0 1.1 0.8 Kazakhstan 12.9 8.8 7.2 12.6 3.7 8.2 4.9 Kyrgyz Republic 0.1 0.3 2.4 5.9 1.7 6.4 6.0 Source: EBRD, Transition Report.
REFERENCES
Aigner, D., K. Lovell and P. Schmidt (1977), Formulation and estimation of stochastic frontier production function models, Journal of Econometrics 6, 21 37 Aitken, B.J. and A.E. Harrison (1999), Do domestic firms benefit from direct foreign investment? Evidence from Venezuela , American Economic Review 89(3), 605 618 Aitken, B., G. Hanson, and A. Harrison (1997), Spillovers, Foreign Investment, and Export Behavior, Journal of International Economics, 43, 103 32 [30] Altenburg, T. (2000), Linkages and spillovers between transnational corporations and small and medium sized enterprises in developing countries: Opportunities and policies, Proceedings of the UNCTAD X Special Round Table, TNC SME Linkages for Development, Bangkok, February Balasubramanyan, V. N., M. Salisu, and D. Sapsford (1996), Foreign direct investment and growth in EP and IS countries , Economic Journal 106(434), 92 105 [14] Bartels, F. L. (2005), UNIDO’s contribution to technological development: enhancing developing countries’ ability to absorb and master technology , Technology Paper Series, No. 4/05 (Vienna: UNIDO) Battese, G. and T. Coelli (1995), A Model for Technical Inefficiency Effects in a Stochastic Frontier Production for Panel Data, Empirical Economics 20 Blomström, M. and H. Persson (1983), Foreign investment and spillover efficiency in an underdeveloped economy: Evidence from Mexican manufacturing industry, World Development 11, 493 501 Borensztein, E., J. Gregorio, and J. Lee (1998), How does foreign direct investment affect economic growth?, Journal of International Economics 45, 115 135 Brown J. D., J. S. Earle, and A. Telegdy (2006), The Productivity Effects of Privatization: Longitudinal Estimates from Hungary, Romania, Russia, and Ukraine, Journal of Political Economy, 114(1), 61 99 Caves, R.E. (1974), Multinational firms, competition and productivity in host country markets, Economica 41, 176 193 [18] Caves, D.W., L. R. Christensen, and W. E. Diewert (1982a), Multilateral Comparisons of Output, Input, and Productivity Using Superlative Index Numbers, Economic Journal, 92, 73 86 Chenery, H., Sh. Robinson and M. Syrquin (1986), Industrialization and Growth , Oxford University Press Commander, S., and J. Svejnar (2007), Do Institutions, Ownership, Exporting and Competition Explain Firm Performance? Evidence from 26 Transition Countries,
Cornwall, Ch., P. Schmidt and R. C. Sickles (1990), Production frontiers with cross sectional and time series variation in efficiency levels , Journal of Econometrics, 46 (1 2), 185 200 Djankov, S., and B. Hoekman (2000), Foreign Investment and Productivity Growth in Czech Enterprises, The World Bank Economic Review, 14 (1), 49 64 Djankov, S., and P. Murrell (2002) Enterprise Restructuring in Transition: A Quantitative Survey, Journal of Economic Literature, 40(3), 739 92 Dong, Xiao Yuan, L. Putterman, and B. Unel (2006), Enterprise Restructuring and Firm Performance: A Comparison of Rural and Urban Enterprises in Jiangsu Province, Journal of Comparative Economics, 34(3), 608 633 Estrin, S., K. E. Meyer, and M. Bytchkova (2005), Entrepreneurship in Transition Economies, in Oxford Handbook of Entrepreneurship, ed. Mark Casson et al . Oxfrod, Oxford University Press Estrin, S. (2002), Competition and Corporate Governance in Transition , Journal of Economic Perspectives 16(1), 101 124 Estrin, S., J. Hanousek, E. Kocenda, and J. Svejnar (2007), Privatization in Central East Europe and the CIS, Mimeo, University of Michigan LSE CERGE EI Good, D., I. Nadiri, L. Roller, and R. Sickles (1993), Efficiency and Productivity Growth Comparisons of European and U.S. Air Carriers: A First Look at the Data, The Journal of Productivity Analysis 4 Görg, H. and Strobl, E. (2001), Multinational Companies and Indigenous Development: An Empirical Analysis, European Economic Review 46, 1305 1322 [18] Greenaway, D., N. Sousa and K. Wakelin (2004), Do Firms Learn to Export from Multinationals? , European Journal of Political Economy 20, 1027 1043 [17] Greene, W. (2002), Fixed and Random Effects in Stochastic Frontier Models, Working Papers 02 16, New York University, Leonard N. Stern School of Business, Department of Economics Grosse, R., ed. (2005), International Business and Government Relations in the 21 st Century, Cambridge: Cambridge University Press Haddad, M. and A. Harrison (1993), Are there positive spillovers from direct foreign investment? Evidence from panel data for Morocco, Journal of Development Economics 42, 51 74 [24] Hirschman, A. (1958), Interdependence and Industrialization, Chapter 6 of The Strategy of Economic Development , New Haven, Yale University Press, 98 119 [20]
Yifan, H., F. Song, and J. Zhang (2004), Competition, Ownership, Corporate Governance and Enterprise Performance: Evidence from China, Hong Kong Institute of Economics and Business Strategy Working Paper 1111 Javorcik, B.S. (2004) Does foreign direct investment increase the productivity of domestic firms? In search of spillovers through backward linkages, The American Economic Review, 94, 605 627 Jefferson, Gary H., Thomas G. Rawski (1996), Chinese Industrial Productivity: Trends, Measurement Issues, and Recent Developments, Journal of Comparative Economics, 23(xxx), 146–180 Jefferson, G. H. and S. Jian (2006), Privatization and restructuring in China: Evidence from shareholding ownership, 1995–2001, Journal of Comparative Economics, 34(1), 146–166 Jondrow, J., C.A. Knox Lovell, I. S. Materov, and P. Schmidt (1982), On the estimation of technical inefficiency in the stochastic frontier production function model, Journal of Econometrics 19, 233 238 Jorgenson, D. W., Gollop F.M and B. Fraumeni 1987, Productivity and US Economic Growth, Harvard University Press Schumpeter, J. A. (1942), Capitalism , Socialism and Democracy , London, Unwin Lall, S. (1980), Vertical Inter firm Linkages in LDCs: An Empirical Study , Oxford Bulletin of Economics and Statistics 42, 203 226 [24] Li, X. and Liu, X. (2005), Foreign Direct Investment and Economic Growth: An Increasingly Endogenous Relationship , World Development 33, 393 407 [15] Markusen, J.R. and A. Venables (1999), Foreign direct investment as a catalyst for industrial development, European Economic Review, 43, 335 356 [22] Meyer, K. E. (2004), Perspectives on Multinational Enterprises in Emerging Economies, Journal of International Business Studies 34, no. 4, 259 277 Meyer, K.E. and E. Sinani (2007), When and where does foreign direct investment generate positive spillovers? A meta analysis, mimeo , University of Reading and Copenhagen Business School, July Perez, T. (1997), Multinational enterprises and technological spillovers: an evolutionary model, Journal of Evolutionary Economics 7, 169 192 [24] Sabirianova K., J. Svejnar, and K. Terrell (2005), Foreign Investment, Corporate Ownership, and Development: Are Firms in Emerging Markets Catching Up to the World Standard?, William Davidson Institute Working Paper 734 Schmidt, P. (1985 86), Frontier Production Functions, Econometric Reviews 4, 289 328
Shahid Yusuf Kaoru Nabeshima Dwight H. Perkins (2006), Under New Ownership Privatizing China’s State Owned Enterprises, World Bank Publication Stevenson, R. (1980). Likelihood Functions for Generalized Stochastic Frontier Estimation. Journal of Econometrics, 13(1), 58 66 The World Bank (2007), The Caucasian Tiger; Sustaining Economic growth in Armenia, The international Bank for Reconstruction and Development
COMPETITIVENESS OF THE ARMENIAN PRIVATE SECTOR : MOVING TO THE NEXT STAGE *
Manuk Hergnyan, Economy and Values Research Center Gagik Gabrielyan, Economy and Values Research Center Anna Makaryan, Economy and Values Research Center
Abstract: The paper examines Armenia’s economic achievements, and explores the anatomy of Armenia’s competitiveness by studying its underlying factors and causal links. It contrasts Armenia’s economic performance and microeconomic foundations of competitiveness with peer or comparator countries. The “synthesis” part of the paper advances a set of recommendations and thoughts on the future development of Armenia and enhancement of national competitiveness, and covers such issues as the context of future policies, frameworks and mechanisms for setting priorities and designing short to long term development strategies, approaches for positioning Armenia in the region and globally. JEL Classification: O11, O12, O19, O29, O30 Keywords: Armenia, business environment, competitiveness and regional policies, international linkages and internationalization
* This paper is based on the content of the first National Competitiveness Report of Armenia, EV, 2008.
I. INTRODUCTION
Currently, competitiveness is a major challenge facing Armenia’s decision makers in both the private and public sectors. While Armenia’s economic growth rate is exceptionally high, it has only recently recovered to GDP levels seen in 1990 and wage levels are still less than half of levels seen in 1990. Economic performance has been largely dependent on external factors (e.g. remittances, assistance from international financial and donor organizations). Large and increasing regional disparities and continued poverty among some segments of the Armenian population create a sense of urgency to bring economic dynamism to all geographic areas and to all segments of the population. Armenia’s disadvantage as a landlocked country increases the need for repositioning Armenia towards high value products and services that are less subject to transportation cost disadvantages. Attaining higher levels of competitiveness will determine whether Armenia can achieve sustainable and harmonious economic growth beyond that bolstered by remittances, foreign assistance and resource exploitation. Competitiveness, which we define here as high and rising levels of productivity, is determined more by created factors than its resource endowments. It is a function of the nation’s ability to develop an environment enabling firms and individuals to utilize the nation’s resources and factors effectively and efficiently. It is about the ability of the nation, its firms and citizens to design, adopt, and implement sophisticated strategies and operations which allow efficient utilization of available resources and factors. In other words, competitiveness is about the ability to be productive. Various benchmarks of competitiveness exist. Those generally considered most authoritative are provided by the World Economic Forum (WEF), International Institute for Management Development (IMD), and the Institute of Industrial Policy Studies (IPS). However, only WEF currently includes Armenia in its reports. Therefore, the paper we will heavily (but not exclusively) rely on WEF’s indexes for assessing relative competitiveness. II. ARMENIA ’S RECENT ECONOMIC ACHIEVEMENTS
Prosperity is ultimately a choice. A nation’s welfare or prosperity depends on its choices regarding how efficiently and effectively the nation utilizes and allocates its resources, or in other words, on how productive or competitive the nation is. We start by looking at what the economic system of Armenia has yielded to date. First, we look at the achievement measures prosperity to assess the performance of the economic system and its results. Then we look at its key driver – productivity measures, and, after, two major enablers that stimulate the productivity growth, namely – internationalization and knowledge, skills and creativity measures. These are four closely linked indicators of economic achievement of a country. Prosperity depends on the productivity of a nation and on the way wealth is created and distributed in society. Innovative capacity is an important indication of economic achievement, and also an engine for productivity increase. Knowledge, skills and creative potential constitute the innovative capacity of a nation, a cluster, a firm and an individual. Internationalization illustrates the degree to which a country is successfully competing in foreign markets for goods, services and capital. Usually, high productivity is accompanied by increased
levels of internationalization, and they together are strong supporters of knowledge and skills creation and upgrade. Figure 2.1 Economic Yields and Competitiveness
Prosperity
Knowledge, Skills and Productivity Internationalization Creativity Source: Adapted from Michael Porter (1998); EV, 2007 Armenia’s economic performance is viewed in the international context, i.e. in comparison with regional and international competitor and/or peer countries and benchmark countries. For this purpose, peer and competitor countries are selected from the following regions: (a) Eastern Europe (EE), which includes Central and Eastern European (CEE) and Baltic countries (Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovak Republic, Slovenia) and Southeastern European (SEE) countries (Albania, Bulgaria, Bosnia Herzegovina, Croatia, FYR Macedonia, Montenegro, Serbia, and Romania); (b) Commonwealth of Independent States (CIS), which includes Russia, Belarus, Kazakhstan, Kyrgyzstan, Moldova, Tajikistan, Turkmenistan, Ukraine, Uzbekistan); and (d) the regions which we refer to as Eurasian Crossroad Region between the Black, Mediterranean and Caspian Seas, of which Armenia is a part and which also includes Azerbaijan, Georgia, Jordan, Iran, Israel, Lebanon, Syria, and Turkey. This regional coverage allows a fairly comprehensive comparison and a broad view on Armenia’s position based on geographical and historical as well as social and economic factors. Countries of Eastern Europe and the CIS are all transition countries with a wide range of similarities inherited from communist regimes. The recent history of Armenia is closely linked to CIS countries, and there are still close economic and trade relations with many CIS countries. Comparison with European countries is also in line with current priorities in Armenia’s foreign policy to achieve closer integration with the EU, and with the fact that the EU is now a major trade partner for Armenia. At the same time, Armenia has much longer historical, cultural and economic ties with countries of the Middle East and the Mediterranean, although they were interrupted during Soviet times. We refer to this region as the Eurasian Crossroad region. After the collapse of the Soviet Union, Eurasian Crossroad countries became important economic and trade partners for Armenia, and this region provides broad, but still unexploited market opportunities. 2.1 Prosperity
After more than 50 percent decline in GDP between 1991 and 1993, Armenia recorded 5.4 percent growth in 1994, and since has grown at an annual compound rate of 8.2 percent. The growth rate has been remarkably high during the last six years, at two digit levels. In 2004, after a decade of economic growth, the GDP level surpassed the 1990 level (Figure 2.2). Armenia’s GDP growth rate outpaced almost all transition countries in EE, the CIS and the Eurasian Crossroad regions (Figures 2.3) in 2006. An exception was Azerbaijan, where a 34.5 percent GDP growth rate was registered versus 13.2 percent in Armenia for the year 2006 (and 13.8 percent in 2007).1 Azerbaijan’s high growth rate was strongly driven by the energy sector which benefited from increasing oil and gas prices worldwide. Figure 2.2 Armenia GDP and Wage Growth Rates (1990-2006)
Source: NSS, various publications There has been a similar trend with regard to growth rates of GDP per capita. Between 2000 and 2006, Armenia’s compound annual growth rate (CAGR) of GDP per capita was 12.9 percent, outpacing neighboring countries, as well as its regional peer/comparator countries, although lagging behind Azerbaijan (15.8 percent). With this growth rate, Armenia’s GDP per capita in PPP terms surpassed USD 5000 in 2006, placing Armenia in the group of lower middle income countries according to World Bank criteria. Economic growth has brought about considerable improvements in the population’s welfare in terms of real wage increases, poverty reduction, and increased spending on social services and transfers. However, a closer look shows that in absolute terms the level of Armenia’s GDP, based on purchasing power parity (PPP) as well as GDP per capita, still remains low and lags behind most of its competitor and peer countries. In 2006, Armenia’s GDP per capita (PPP) was higher than that of Georgia, Moldova, Kyrgyzstan, Uzbekistan, Tajikistan, Syria, Egypt, Lebanon and Albania, but was significantly lower than other EE, CIS and Eurasian Crossroad countries. Although the country’s economy has been growing annually by two digits, this growth should be viewed against a very low GDP base.
1 National Statistical Service of Armenia (NSS), 2008
Figure 2.3 Comparative Economic Performance (2000-2006)
Source: WB, WDI Online, 2007 (last accessed October 2007); Authors’ own calculations. Note: The data for Israel is for 2005. Yellow bubbles indicate countries of Eastern Europe, dark blue – CIS countries, pink – Eurasian Crossroad countries. Additionally, economic growth in Armenia has not resulted in corresponding wage growth; wage levels reached only around 40 percent of 1990 levels (Figure 2.2). Real unemployment levels (over 30 percent) and poverty (circa 26 percent in 2006) still remain quite high; the distribution of wealth in society is highly inequitable; and disparities between regions in terms of economic and social development are significant. 2.2 Internationalization
A. Export Performance
We look at Armenia’s export performance as an important indicator for measuring a country’s comparative productivity and ability to compete in world markets. For a country like Armenia (with a small internal market) export performance is a crucial factor for economic development, as it permits the achievement of economies of scale and concentration on those sectors where the country is more competitive. The Government of Armenia has indicated that export promotion is one of its main priorities. Since 2000 Armenia’s share of world exports has more than doubled, although in absolute terms it remains very small and lags behind most of the EE, CIS and Eurasian Crossroad countries (Figure 2.4). Armenia’s total share of world exports is slightly less than its share of world GDP. This is an indication that economic growth has not been export driven, but has been attained primarily through import substitution, growth of domestic demand, and growth of non traded sectors of the economy. This is a warning sign for Armenia’s international competitiveness given the small domestic market and, hence, the need for export led growth as a long term sustainable strategy. The difference between Armenia’s share of world exports and share of world GDP is better than that of Turkey, Lebanon, and Albania, but is worse than that of other countries in EE, the CIS and Eurasian Crossroad regions.
Figure 2.4 Export Share in World Export (2000-2005, percent)
= difference between shares in world exports and world GDP; white collor- negative size
Source: WB, WDI Online, 2007 (last accessed October 2007); Authors’ own calculations. Note: Bubble size indicates the difference of shares of world exports and world GDP. White color indicates that the share of nation’s export in the world exports is less than the nation GDP share in the World GDP. These are generally considered underperformers. Data for Syria is for 2004. Armenia’s exports are largely resource intensive. The latter represented about 64 percent of the country’s total merchandise exports in 2005 which increased to 69.5 percent in 2006; this is quite high compared with indicators of most counties in EE, the CIS and Eurasian Crossroad regions (Figure 2.5). It should be noted that in 2002 2004, the indicator of resource intensity of Armenia’s exports was higher still at over 70 percent (2002 at 75, 2003 at 77, and 2004 at 71 percent)2. The drop of the indicator in 2005 was due to the drop in the export of diamonds (in 2002 the share of diamonds among all merchandise exports was 48 percent compared with only 30.9 percent in 2005). High resource intensity also means low value added of Armenian exports. Higher value added products and services are required to attain higher levels of productivity and competitiveness.
2 Resource intensive product groups were defined using the Lall (2000) classification of export by their technological intensity. Two groups of products were treated as resource based products: primary products (fresh fruit, meat, rice, cocoa, tea, coffee, wood, coal, crude, petroleum, gas) and resource based manufactures agro/forest based products (prepared meats/fruits, beverages, wood products, vegetable oils)) and other resource based products (ore concentrates, petroleum/rubber products, cement, cut gems, glass). Source: Chami Batista, 2004; UNCTAD, 2004; Mahmood, 2004
Figure 2.5 Resource Intensity of Merchandise Exports (2006)
Source: WB, WDI Online (last accessed: October 2007); UNCOMTRADE online database (last accessed: October 2007); Authors’ own calculations. Note: PPP adjusted GDP per capita of Israel is for 2005.
B. Investment Performance Foreign direct investment (FDI) is an important factor that can contribute to economic performance and to productivity in several ways, including: (a) introduction of new technologies, managerial competences, and knowledge; (b) enhancing the business climate; (c) improving domestic competition; (d) job creation; and (e) increasing the quality of human resources. FDI driven growth has been a model for economic growth in many countries (e.g. Hungary). In most such countries, the bulk of investment has been domestic, but FDI creates a leading edge and triggers new domestic investments. It is logical to expect, and the recent history of FDI has demonstrated, that high levels of FDI are recorded in countries with primary extractive industries, with a superior cost position, with large domestic markets, or in countries that can supply highly skilled scientists and engineers for high value added processing and R&D activities. Given the constraints in natural resource endowments, size and geopolitical situation, in order to attract large and high quality FDI it is vital for Armenia to enhance knowledge intensive industries. This would boost the competitiveness of the nation, and over time would allow establishing or bringing high value processing elements of the global value chains of multinationals to Armenia. Between 2003 and 2007, the CAGR of inward FDI to Armenia was 52.9 percent, which, however, was driven mainly by reinvestments (Figure 2.6). The largest inflow of equity investments was observed in 2006, peaking at 216.7 million. The tremendous growth of FDI was due to the rise of reinvested profits and other investments. In 1998 through 2007, the major FDI recipients were electricity, gas and water supply (25.2 percentage share), post and telecommunications (19.8), mining and quarrying (14.6), and food products and beverages (7.7) sectors 3. If the current trends continue, Armenia will soon exhaust its potential to attract even resource driven FDI into new sectors.
Figure 2.6 Inward FDI, Armenia (2001 - 2006)
Sour ce: NSS of Armenia, Balance of Payments, various publications
2.3 Knowledge, Skills and Creativity Since independence, the R&D and innovation performance of Armenia has been rather disappointing. Once an important center for scientific research and high tech production in the former Soviet Union 4, Armenia has lost most of its scientific and technological (S&T) resources and has yet to exploit its S&T potential. Until recently, the promotion and development of innovation were, practically, out of the Government’s policy agenda. Armenia has a higher tertiary education expenditure/GDP per capita ratio than most other countries, but in absolute terms tertiary student expenditure is lower than that of most countries. Government expenditure on R&D has been extremely low and negligible and below that of most competitor and peer countries in EE, the CIS and Eurasian Crossroad regions (both as a percentage of GDP and, more dramatically, in absolute terms). In 2006, R&D expenditures accounted for 0.22% of GDP. In 2005, Government R&D expenditure was only 0.12% of GDP, and there has been a similar ratio in the previous five years. Armenia’s private sector businesses also had very low expenditure for R&D and for employee training (less than Government expenditure). In 2005, total R&D expenditure in the country was only 0.23% of GDP.5 3 National Statistical Service of Armenia, 2008. 4 Armenia had 25,000 scientists and engineers in 1990 vs. 6,700 in 2005 (Arzumanyan, 2006). 5 NSS; UNESCO online database, 2007.
2.4 Productivity Armenia’s prosperity measured in terms of GDP per capita reflects both the level of labor force utilization and the productivity at which labor is employed. Labor productivity is an important measure of economic performance. However, ensuring higher labor participation in the economy also is a crucial economic and social achievement. Labor productivity can be increased by cutting the labor force, however, this will not bring about real increase in prosperity of the society or the competitiveness of the country. Labor force utilization depends on labor force as a share of the working age population, employment rate (share of working age population) and hours worked (per employee). Armenia has a low labor force participation rate (59 percent) lagging behind most of the EE, CIS and Eurasian Crossroad countries (Figure 2.7). This reflects the effects of massive emigration of mostly young people since the early 1990s. Official unemployment has been decreasing since 2000 from 11.7 percent in 2000 to 7.5 percent in 2006. But, official figures for unemployment (or registered unemployment) differ significantly from real unemployment in the country. Real unemployment still remains quite high – in 2006 the unemployment rate was about 30 percent.6 Moreover, there is a quite high level of underemployment in the economy, particularly in the agricultural sector, where more than half of the labor force was underemployed and expressed readiness to work more hours. In the industrial and services sector, the levels of underemployment in 2004 were 37.9% and 36%, respectively. 7 The level of underemployment in terms of working less than 40 hours per week was 30.2%, 39.4% and 73.4% in the industrial, services and agricultural sectors, respectively. Figure 2.7 Labor Force Participation Rate and Labor Productivity per Employee (2006)
6 NSS Survey data (ILO methodology); ILO. 7 NSS, Social Snapshot, 2004.
Source: WB, WDI Online (last accessed October, 2007), ILO Laborsta (last accessed October 2007); IMF, International Financial Statistics May 2007; Authors’ own calculations. Note: Data for Egypt, Georgia, Iran, Israel and Kyrgyz Republic are for 2005 Based on official data, the most productive sector of the Armenian economy was construction in 2006 (2.7% of the total employment, and 26% of gross value added). 8 The mining and quarrying sectors also had relatively high productivity (increased due to the rising base metal prices worldwide, which suggests that this may not be a sustainable source of productivity). The productivity level is particularly low in the agricultural sector which employed 46% of the total labor force in 2006, but represented only 20% of gross value added in the country (Figure 2.8). While development of the construction sector can have an important multiplier effect for economic growth in the short run, it is not exportable and, given the small local market, in the long run is unlikely to maintain its leadership position. It is essential therefore, that Armenia in the medium to long run relies on the promotion and development of export oriented, high value added sectors, if it is to enhance its international competitiveness.
Figure 2.8 Shares of Main Sectors - Employment and Value Added (2006)
8 Although it is obvious that the construction sector is the most productive sector in Armenia, it may be argued that the reported level of employment and, thus, the real productivity in the construction sector is much less that the real level due to non registered employees.
Source: NSS, 2007. III. ARMENIA ’S COMPETITIVENESS POSITION AND ITS CAUSES In the GCR 2007 2008, Armenia ranked 93 rd out of 131 countries in the Global Competitiveness Index (GCI), i.e. Armenia was in the 8 th decile of countries a relatively low ranking, indicating that there is still much room for improvement if Armenia is to become a globally competitive country. 9 The GCI provides a holistic overview of around 110 factors that influence productivity and competitiveness of countries, grouping these factors into “twelve pillars” that are arranged into three broader categories, namely, basic requirements, efficiency enhancers, and innovation and sophistication factors , as presented in Figure 3.1:
Figure 3.1 GCR’s Twelve Pillars of Competitiveness of a Nation
9 The rankings and scores in this section are those of the Global Competitiveness Report 2007 2008 (GCR) if not otherwise noted. The comparisons are with GCR 2006 2007 rankings and scores, if not otherwise mentioned.
COMPETITIVENESS 95 99 110 104 104 40 111 57 87 96 – – 111 94 Institutions ial market efficiency Infrastructure Macroeconomy Goods market efficiency Health and Primary educ. High. education & training Labor market efficiency Technological readiness Market size Business sophistication Innovation Financ
Basic Requirements Efficiency Enhancers Innovation & Sophistication Factors 91, down by 14 points 101, down by 10 points 103, down by 18 points
Source: World Economic Forum, 2007. Note: Rankings are that of GCI 2007 08 compared to GCI 2006 07.
Compared to the previous year, Armenia improved its scores modestly in several areas, but saw its rankings fall because many other countries achieved faster progress. This highlights the fact that we are living in a highly competitive and integrated world, where trivial improvements will be inadequate to stand the competition and achieve sustainable prosperity. Armenia’s ranking in the GCR acknowledges that Armenia did well in the areas of macroeconomic stability, labor market efficiency and security for doing business. The Government of Armenia should be commended for its notable efforts aimed at: ensuring macroeconomic stability, including good government debt management and low inflation; facilitating procedures for starting a business; establishing a flexible labor market; and ensuring security for businesses. However, Armenia’s solid macroeconomic performance is not backed by adequate progress in other aspects or pillars of competitiveness, which are necessary for ensuring the sustainability of economic growth. There are serious shortcomings or weaknesses in a number of areas, especially at the micro level, which significantly lower Armenia’s competitiveness in the global context and endanger the sustainability of economic development. Areas of concern are so wide that they point out to the need for a comprehensive reform of public and private sectors. They span from general quality of the business environment, especially the financial sector, the lack of judicial independence and the prevalence of favoritism in government decisions to weak performance in higher education and training and low level of innovation. The sustainability of economic growth and the ability to transform it into a sound basis for economic development in the future depends greatly on the ability of both the public and private sectors to reshape and build their policies on a truly pro development
context. While much emphasis is rightly placed on the need for Government reforms, there is also a pressing need for the private sector to improve its operational and strategic performance. The current situation with high rates of economic growth, but low and regressing competitiveness is not balanced and sustainable. It’s a specific situation that can be characterized as a “growth competitiveness paradox”. This is another impetus for comprehensively understanding the architecture of Armenia’s economic growth, its current and future drivers as well as possible sources of improving its competitive position.
3.1 Micro-level Foundations of Competitiveness We will evaluate the micro level foundations of Armenia’s competitiveness through the lens of the Business Competitiveness Index (BCI) of the GCR, and use the “Diamond Model” (Figure 3.2) developed by M. Porter as an intellectual framework for analyzing the business environment. The BCI is based largely on Executive Opinion Survey data and, in broad terms, measures two broad determinants of the microeconomic environment: (a) quality of national business environment, and (b) sophistication of company strategies and operations. The “Diamond Model” depicts the business environment as a system of four interrelated areas: (a) the quality of factor (input) conditions, (b) the quality of local demand conditions, (c) the context of firm strategy and rivalry, and (d) the presence of related and supporting industries. It has proved to be a very powerful tool in analyzing industry and country competitiveness. Overall, Armenia has fallen 17 spots in the BCI in 2007 (i.e. BCI index in GCR 2007 2008) relative to 2006 (i.e. BCI index in GCR 2006 2007), falling from 91 st rank to 108 th among 127 countries.
Table 3.1 Armenia’s Business Competitiveness Rankings in the GCR
Business Competitiveness Index 2007 Rank 108, Lost 17 points 17 points Quality of National Business Environment, Sophistication of Company Str ategies and Lost 17 Points Operations, Lost 23 Points 2006 2007 2006 2007 89 106 92 115 Source: World Economic Forum, 2007. The quality of the national microeconomic business environment is an important determinant of the level of sophistication an individual company can reach. No matter how capable firms are, in order to employ sophisticated operations and strategies, they need developed administrative and physical infrastructure, access to financial resources, and a highly skilled labor force. They need also suppliers of necessary services and materials to produce sophisticated goods; demanding customers urging them to innovate and upgrade to build competitive edge over time; and stringent local competition to enhance productivity. However, a firm’s ability to reach an increased level of sophistication, upon which the nation’s productivity depends, is also partly a function of its own strategy and operations. No matter how favorable or developed a nation’s microeconomic business environment is, a nation needs companies that are able to design and employ sophisticated operations and strategies. Ultimately wealth is created by businesses and as a general rule, productive companies tend to have a high degree of sophistication, with more efficient operational processes and strategies. Competitive companies have professional management who are able to design winning strategies and drive the company; employ efficient and innovative ways to produce and market unique and sophisticated goods; collaborate and make effective use of services and “unusual efforts” of skilled engineers and scientists; invest in R&D adopt sophisticated technologies to upgrade their current competitive position; and hire and train high quality professionals. The assessment of the “national diamond” of the business environment is provided below in Figure 3.2:
Figure 3.2 National Business Environment (Diamond Model)
Context for Firm Strategy and Rivalry
A local context and rules that determine the extent of investments and sustained upgrading (e.g., IPR protection), the types of corporate strategies, Factor and the intensity of competition Demand Conditions (Input) Conditions among locally based rivals.
Sophistication of local demand and the The quality and specialization of inputs pressure of local customers to upgrade available to firms: products and services. • Human resources • Capital resources • Physical infrastructure • Administrative infrastructure • Information infrastructure • Scientific and technological Related and Supporting infrastructure Industries • Natural resources
The availability and quality of local suppliers and related industries, and the state of development of clusters. Source: Michael Porter, 1990. Since 2001, recognizing the need for improving the business environment in the country, the Government initiated a number of reforms including the consolidation and reduction of business inspections, simplification of administrative procedures, reduction of time for business registration, and streamlining of the licensing regime. The government’s consultation mechanisms with the private sector were strengthened and a high level Business Council, chaired by the Prime Minister, was established. In 2007 the government also created a National Competitiveness Council to boost the competitiveness of the country.
According to the most recent World Bank’s Doing Business Report 2008, Armenia was ranked 39 th among 179 countries, and was ahead of most countries of the EE, CIS and Eurasian Crossroad regions, but lagging behind Estonia (17 th ), Georgia (18 th ), Latvia (22 nd ), Lithuania (26 th ), Isprael (29 th ) and the Slovak Republic (32 nd ). In Doing Business 2008, there was a significant positive change related to the improvement of legislation aimed at easing access to credit. However, under the GCR (which measures actual practice rather than a formal regulatory framework), Armenia had a disappointing showing in terms of the difficulty in obtaining a bank loan (123 rd out of 131 countries). Table 3.2 Ease of Doing Business in Armenia, 2007-2008 (among 179 countries)
Ease of... Doing Business 2008 Doing Business 2007 Change in Rank Rank Rank Doing Business 39 46 +7 Starting a Business 47 44 3 Dealing with Licenses 73 72 1 Employing Workers 48 48 0 Re gistering Property 2 2 0 Getting Credit 36 62 +26 Protecting Investors 83 81 2 Paying Taxes 143 137 6 Trading Across Borders 118 133 +15 Enforcing Contracts 64 64 0 Closing a Business 42 43 +1 Source: World Bank, 2007; http://www.doingbusiness.org/ExploreEconomies/?economyid=10 (last accessed October, 2007) A. Management The lack of professionally trained and globally experienced management and the unwillingness to delegate authority to professionals are significant impediments for Armenian businesses in gaining the ability to choose the most efficient operations and effective strategies and, hence, to increase the productivity and competitiveness of companies. This situation may in particular be caused by (a) a lack of awareness of the importance of professionally trained managers, which, in turn, could be a result of the lack of knowledge, or of imperfections in competitive environment due and absence of strong pressures towards increasing efficiency and management capacities (b) lack of trust; (c) a “non merit based” culture of hiring managers and other key professionals, i.e. managerial and other key positions in companies are most frequently held by relatives or friends independently of their professional qualifications. Poor management and a non merit based system of hiring employees are priority problems that have to be tackled if Armenian companies want to become productive and competitive in global context. IV. TOWARDS A COMPETITIVE ECONOMY We believe that Armenia has yet to adopt a development driven policy context. Today Armenia faces a serious challenge of rethinking and reshaping the context of its policies in order to introduce a truly development oriented agenda, which would enable the country to become a competitive player in the global marketplace. Immediately following independence, Armenia undertook comprehensive reform efforts toward establishing a market economy and democratic society. This was a period characterized by dramatic economic decline, disruption of trade, the Nagorno Karabakh conflict and transportation blockade, shortage of energy, food, and other consumer products, hyperinflation, and high levels of unemployment and poverty. Under these circumstances, the government grew accustomed to governing in “crisis mode”, forced to find quick solutions to urgent, short term problems. Armenia’s policy context became a “survival context”, which resulted in a short sighted view of policy and a general lack of policy coordination and of an overall strategy toward policy development. This has been exacerbated by very weak public and private institutional capacities. After achieving economic stability and reaching a ceasefire in the Nagorno Karabakh conflict in the mid 1990s, the nature of Armenia’s policies became more redistributive. The underlying context of economic reconstruction was the redistribution of the country’s wealth among its citizens. Mass privatization of small and medium entities was the major direction of reforms. The privatization has been highly inequitable and resulted in the concentration of the country’s productive assets in the hands of a few groups. The context continued to be characterized by a “fire fighting” approach, with remaining elements of the “survival context”. The “redistribution context” implied better, but still weak coordination. At the beginning of this decade, Armenia adopted a poverty reduction approach (“social context”). In 2003, for instance, Armenia adopted the Poverty Reduction Strategy Paper (PRSP), which became an overarching strategic document based upon which the Government developed its policies and programs, and designed its medium term budgetary expenditures. The PRSP has become a single comprehensive framework that established long term measurable objectives and established important linkages between different aspects of social, economic and institutional life. In contrast with previous approaches to policy development, policies during this period have been characterized by the introduction of elements of strategic thinking, longer term planning, better coordination of policies and operations, improved institutional capacities, and an attempt to set out priorities. However, policies have remained primarily “inward looking”, without regard to the international context in which they are being made. Additionally, the government’s policies have not made economic development a centerpiece of their approach. While the PRSP contains objectives related to economic development, it remains unclear what the drivers of Armenia’s competitiveness and economic development are. It fails to deal with fundamental questions, such as the sectors in which Armenia can successfully compete globally, what its value proposition is, or what its competitive strengths and weaknesses are. We believe that these can be best addressed if Armenia adopts an economic development focused strategy. Policies and projects should be viewed and measured against their implications for economic development. This will require a higher level of
coordination of the country’s policies and projects and clear linkages and concerted interaction between numerous sectors and aspects of economy. Such a strategy should be based on a realistic and honest assessment of the nation’s capabilities and its opportunities in the global marketplace, which would permit Armenia to identify the unique strengths upon which the country can compete internationally, and which would set out a road map to development, enabling effective coordination of activities and establishing priority areas or sectors. This would allow for more efficient use of scarce financial and other resources. We propose a structured approach to enhancing Armenia’s competitiveness distinguishing two layers of action, namely “Quick Wins” and “Strategic Breakthrough”. The Quick Wins are targets that are achievable in a short time period within the limits of current resources and competencies, while actions in the Strategic Breakthrough sections define factors that may move Armenia to the next level of competitive position vis à vis other countries and direct competitors. 4.1 Strategic Breakthrough Armenia is in need of new growth drivers based on sustainable sources of competitive advantage. Given current trends, this calls for real strategic breakthrough which leaves the decision makers with two fundamental questions: 1.What should be the role of Armenia in the regional and global economy? 2.What are the tools to achieve it? A. Strategic Positioning of the Armenian Economy The experience of many resource scarce but highly successful economies that managed to transform their early acceleration into sustainable economic growth suggests that clear positioning of a country is a key ingredient of success. Such positioning implies defining a unique and sustainable set of local conditions, skills, products and services that will lead to the creation of competitive advantages. At the heart of the positioning is the definition of a country’s value proposition . A core value proposition points to the specific role the country plays in the world or regional economy. These can include its value as a business location, the range of businesses, functions and competences for which the country can become a base for globally competitive companies. Value proposition inherently incorporates the notion of competitive arena – the region, range of countries, competitors or the entire world. Given the interplay of the factors mentioned earlier in and around Armenia, it has a narrow menu of possible choices for its positioning and value proposition. Unique human capital represents one of the key ingredients for such a value proposition. At the same time, human capital can hardly represent a competitive advantage unless it has unique characteristics. For a country like Armenia, a few unique characteristics of human capital deployed in areas producing high value added products and services may become an economy wide source of advantage and key differentiating feature. For that, the targeted areas need to be narrow enough to enable focused efforts, greater leverage and larger impact.
Armenia may strive to become an R&D center at the regional level and use it to become a leader in a few areas. Those areas could be either technology intensive (such as selected segments in software programming or “green” mining and metallurgy), network and knowledge intensive (for example selected niches in financial and educational services), or creativity intensive (including selected areas of culture related commercial activities (cultural tourism centered around key established cultural events/artifacts of global or regional significance). Becoming an R&D center requires development of key competencies and skills that may have business appeal for regional and global players to attract investments and technologies into Armenia. The effort should capitalize on past tradition, selected functioning R&D and scientific institutes and emerging trends in some areas (such as IT). Such positioning will require re definition of a notion of region as Armenia’s perceived positioning arena. Re defining “competitors” should become a core theme for the country’s branding and image building. That theme should dominate all key communication efforts and be shared by the government, private sector, academia and society at large. Currently, it is most common to consider Armenia as a part of the South Caucasus, CIS or Eastern Europe. However, none of these groups of countries alone provide favorable positioning settings for Armenia. An unfavorable location and the conflict with Azerbaijan in the case of South Caucasus, dominance of Russia and lack of perception as a region in case of CIS, and impossibility to attain leadership roles in case of Eastern Europe make these regions not well suitable for positioning purposes. Armenia may consider building its communication strategy utilizing the notion of “Eurasian Crossroad”. While the latter is not a well defined region in the political, economic or geographic sense, it has multi layered and meaningful connotations which are exploitable for positioning purposes. The concept of “Eurasian Crossroad” has deep cultural, historic, economic and political roots. It appeals to the idea of intersection of civilizations, cultures, religions and political systems. While the interpretation as to its geographic frontiers may be voluntary, it provides countries with greater flexibility for positioning and communication purposes. The region will provide more effective positioning platforms if it includes countries such as Turkey, Georgia, Azerbaijan, Armenia, Iran, Iraq, Syria, Jordan, Lebanon and Israel (Central Asian or Gulf countries can be considered a part of the region in a much broader sense). The strategic positioning based on the concept of a crossroad will also help overcome the perception of Armenia as a country in a landlocked location without significant natural resources. The region defined in this way provides opportunities for Armenia to capture leading roles in a few selected areas that demand highly developed human capital, access to global networks and a tradition of science and technology. This option will imply positioning Armenia as an R&D center in skill intensive areas in the Eurasian crossroad region. As such a claim or vision is highly ambitious given the competitive positions and claims by direct competitors in the region, this will require an organized effort with an integrated strategy framework and identification and deployment of key levers. B. Key Levers for Breakthrough
Strategic breakthrough refers to the creation, nurture and development of internationally competitive industries that will make Armenia’s value proposition to the world well shaped and visible. As resources are scarce, this calls for identification of a few selected strategic levers that could be deployed to attain high impact. The latter does not eliminate the need for broad based public and private sector reforms, improvement of general business environment, provision of general pro development economic policy, etc. Instead these two streams need to reinforce each other. As in mechanics, levers are deployed using a base or platform. Base in this case may be an effective sector specific, micro level policy platform. Such a platform will direct the levers’ energy and impact to specific sectors, technologies and locations. The combination of levers and the bases can be called a leverage system. The leverage system that has been identified for the purposes mentioned above will have a three tier hierarchy as described in Figure 4.1 The Leverage Platform will consist of three sets of policies and initiatives – cluster policy, innovation policy (which is related to and mutually reinforces cluster policy) and regional development policies. The Key Levers are targeted foreign direct investment (FDI), formation and development of Diaspora entrepreneurial networks and strategic initiatives in education. Figure 4.1 System of Levers The central role in shaping the leverage platform wSystem of Leversill belong to cluster initiatives (or policies). This will be called the policy nexus . The cluster initiatives in potentially internationally Leverage competitive Platform clusters will be enhanced by supportive Key Levers innovation and regional development policies which can be called policy add ins . The combined and integrated application of this platform is intendedInnovation Policy to create an efficient and conducive FDI context for applying the levers.
Diaspora Networks Cluster Initiatives
Regional Education Development Policy Leverage Platform
Policy Nexus - Cluster Policies. The concept of clusters has become very popular in many countries during recent years. As its practical manifestation, a large number of cluster initiatives at national and regional levels have been initiated with massive public support and resource infusions. The cluster policies can create an overarching policy context where different functional or sector specific policies can be incorporated and implemented. At the current stage the most urgent competitiveness policy ingredients with strong synergies are seen to be innovation and regional development. The regional disparity impedes competitiveness by skewing the benefits of economic growth. Fortunately, regional policy is an emerging policy priority for the Armenian government; however, it should be incorporated into the right mix of measures and broader economic development policy context. Another emerging policy domain is innovation policy that is supposed to energize Armenia’s potential to become a location for producing high value added products. The effectiveness of these two policy priorities is dependent on their optimal combination and coordination as well as their targeted deployment. The cluster policies can provide an appropriate context to assure these two aspects, as clusters are the ultimate development targets incorporating both geographical as well as technological dimensions. Generally, well designed frameworks of collaboration can be worked out by the government so that different efforts don’t overlap and result in a waste of resources. However, it should be noted that different clusters require different strategies and different sets of interventions; therefore, those frameworks should be flexible enough to capture these differences. They must also take into account location based specifics of the clusters which can be part of regional development policies. In addition to this, a very focused and intensive cluster development effort is required in selected areas to raise Armenia’s international competitiveness. Currently, from the perspective of competitiveness, one of Armenia’s key challenges is the development of a few internationally competitive industries that will set the economy on the path to innovation. No country can be competitive in all industries. Specialization is at the heart of competitiveness. Only a handful of industries can become drivers of Armenia’s international competitiveness. Given also the limited public and private resources that can be mobilized, the efforts should be highly focused and deliberately designed. While there is a need for thorough research for short listing the promising clusters there is already a high degree of consensus around some clusters among key stakeholders as to their potential as priority areas. The Armenian government has already prioritized IT and tourism; however, despite some strategic approach is emerging, this still requires a political will and competence to commit necessary resources.
Innovation Policies. As Armenia strives to build a knowledge based economy, the creation and development of innovation supporting infrastructure becomes a critical challenge for the next 5 10 years. The international competitiveness of the Armenian economy depends greatly on its companies’ abilities to create and market innovative products and services in foreign markets. While still sporadic and inconsistent, there is an emerging debate about the creation of elements of a national innovation system in Armenia. The government has adopted a program on the formation of an innovation system for 2005 2010; however, the committed funds are extremely small, the framework is not yet developed and there is no adequate understanding of its role and place in the overall policy context. The general cluster approach to organizing and stimulating innovative and productive entrepreneurship can become a coherent platform for innovation policy in Armenia. Under this approach the cluster priorities will become priorities for innovation policy based on the need for support of innovative process by specific clusters. The selection of tools may include: (a) critical infrastructure development, (b) knowledge and information centers, (c) incentive schemes, (d) procurement policies.
Regional Development Policies. Despite the fact that the government has already prioritized this issue, there is still a vague understanding of how to combat the problem. The key pitfall here is that it may be dealt with as if it is a purely social problem. Sustainable development is possible only if it is driven by commercially viable projects based on key specializations of each region. Specializations should be based on the right mix of comparative advantages of regions that can be gradually developed into competitive advantages. Specialization naturally caters to the concept of clusters. Thus, instead of current infrastructure building projects implemented in isolation without complex development plans, this approach calls for highly coordinated, deliberately crafted public private collaborative efforts to create regional clusters. On the part of the government this may still include infrastructure development; however, in this case well tailored to the needs of emerging business sector in the region and enhanced by private investments. The approach will require development of strategic profiles of each region. The strategic profiles and action plans of each region should be synthesized and integrated into the general cluster based economic development policy. This approach will help introduce elements of a bottom up and top down, business oriented process to the design of economic policies. It will also ensure a shift from a macroeconomic focus toward a microeconomic focus, which is an absolute imperative for public policy in the economic area in Armenia.
Key Levers The three component leverage platform will create a pro business economic development policy context, within which different reform initiatives can be implemented. However, there are three key factors that can be deployed to bring a true breakthrough. In the context of Armenia these are: (1) technological FDI, (2) Diaspora networks and (3) superior education. These factors were determined taking into account three criteria: (1) Armenia’s possible value proposition alternatives suggested in this paper, (2) Armenia’s comparative advantages that can be developed into unique competitive advantages, and (3) the need for mobilization and prioritization of primary tools for strategic breakthrough FDI. Despite the fact that Armenia declared an open door policy to FDI, this has not become a major driver of technological upgrade and specialization in the Armenian economy. In order to achieve this, Armenia should pursue a clearly targeted FDI attraction strategy. The building blocks of such an effort should be: • well defined target markets – right mix of target multinationals; • clear priorities deriving from policy context (fit with cluster priorities, applying innovation incentives); • initial “seed” public investments in creating fundamentals/basic infrastructure in targeted clusters (in many cases specially tailored to certain multinationals’ needs); • aggressive marketing and targeted promotion plans supported by an adequate budget; • involvement of a few world renowned top executives (preferably non Armenians) as leaders in promotion and communications; • utilization of the Diaspora executives holding key positions at targeted companies; • upgrade of institutions involved in FDI attraction.
Diaspora. The Diaspora may become a unique source of competitive advantage for Armenia that can be matched only by a few other nations. An effective partnership can be built on a “hub and spoke” model. Within that model Armenia should be viewed as a center (hub) for global Armenian business and other partnership flows. The gradual move of HQs, coordinating units, information centers and other resources of pan Armenian organizations and networks to Armenia is only one of many ways to create such a model. Connections and competences are the most critical resources that Diaspora can invest into Armenia’s economy. It is a unique source for creating training and innovation centers, bringing in world class expertise, and accessing the most sought after corporate leaders of global companies. This calls for a set of coordinated activities as well as highly efficient institutional forms of engagement. Education. Education must be made a priority if Armenia’s competitiveness is to be based on knowledge and skill intensive characteristics and if its regional value proposition will be based on unique human capital. This leads to several preconditions (principles) that need to be satisfied for a successful deployment of education as a competitiveness lever: 1. Very high standards of primary education matching the best international practices. Basic education should not be biased towards any specialization, but should provide comprehensive knowledge, creative thinking and high ethical values. However, mathematics (and some natural sciences) should be a high priority as a basic discipline for all natural sciences. 2. Basic education should become a key communication channel of national ideals and major aspects of national identity. Such values will constitute the fundamentals of competitiveness thinking as they will define common national goals and aspirations. 3. Stress on specialization in special and higher education congruent with cluster development preferences. Those areas shall receive most of the resources. 4. Leadership aspirations in 2 3 areas of specialization in the region. Armenia should strive to become a leader in the provision of special educational services in a few areas linked with overall cluster and specialization preferences. 5. Creation and development of a few “centers of excellence”, scaling them up, leveraging them and replicating them on a larger scale to achieve spillovers throughout the entire system. Quick Wins We developed a special analytical tool to identify short term priorities and select the quick wins. The tool has been entitled the “Prioritization Filter”. The competitive disadvantages or areas where Armenia trails most nations are assessed from the viewpoint of feasibility of rapid improvements given the current resource and competence constraints as well as its overall impact on the entire economy. The Prioritization Filter incorporates four key criteria applied to assess and prioritize different competitive areas. Thus, the mentioned four criteria are: - Resource Restrictions - Competence Gap - Time Span
- Spillover Effect. The factor of “Resource Restrictions” assesses the availability of all types of resources (excluding human) necessary to address a certain competitive disadvantage in the short term. The “Competence Gap” factor assesses the existence or lack of required knowledge, expertise, experience, and will to mitigate a specific disadvantage. The “Time Span” is the most important factor as it assesses the feasibility of improvement of a certain issue in 2 3 years, while the “Spillover Effect” provides an assessment of the potential positive effects such improvements might have on the entire economy at the current stage, i.e. positive externalities of specific actions. Each criterion is assigned a weight and assessed using a 1 5 scale. Figure 4.2 The Logic of the Prioritization Filter to Identify Quick Wins
Priority Filters
Resource Restrictions
Weight – 0.2 Competitive Quick Wins Disadvantages Competence Gap Weight – 0.2
Time Span Weight – 0.35
Spillover Effect Weight – 0.25 Using this methodology, 10 factors have been identified as possible quick wins, grouped into three distinct areas: – Energizing financial sector – Improving selected elements of business environment – Promoting technology usage (see Table 4.1).
Table 4.1 Identified Quick Wins
Energizing Financial 1. Easing access to loans and expanding credit activity by banks Sector 2. Reduction of interest rate spread 3. Raising effectiveness of antitrust policy Improving Selected 4. Improving the effect of taxation (creating tax incentives, total Elements of Business tax rate, burden of customs procedures, non wage labor costs) Environment 5. Introducing incentives for FDI in prioritized areas 6. Spreading the use of cellular telephones 7. Encouraging the use of personal computers Promoting Technology 8. Encouraging Internet usage Usage 9. Regulatory framework encouraging use of ICT 10. Government procurement of technology products
V. CONCLUDING REMARKS
Armenia still has a long way to go in order to create a highly competitive economy. It has already recorded notable achievements in ensuring basic conditions for economic development. However, the next stage requires more focused efforts, greater skills and higher aspirations. Success will depend on the nation’s firm choice, its ability to set clear goals, achieve consensus, mobilize resources and work diligently towards those goals. The energizing force will have to be the country’s leadership since “if a man knows not what harbor he seeks, any wind is the right wind” (Seneca).
REFERENCES Arzumanyan, T. (2006), Current Issues of Research, Development and Innovation in Armenia , International Journal of Foresight and Innovation Policy, Vol. 2 No. 2, pp. 133 145 Chami Batista, J. (2004), Latin American Export Specialization in Resource Based Products: Implications for Growth , The Developing Economies, Vol. 42 No. 3, p337 370 International Labor Organization (2007), Laborsta Internet, Geneva International Monetary Fund (2007), International Financial Statistics, Washington D.C. Lall, S. (2000), The Technological Structure and Performance of Developing Country Manufactured Exports, 1985–1998 , Queen Elizabeth House Working Paper Number 49, Oxford University Mahmood, A. (2004), Export Competitiveness and Comparative Advantage of Pakistan’s Non agricultural Production Sectors: Trends and Analysis , the Pakistan Development Review, Vol. 43 No 4, p541 561 National Statistical Service of Armenia (2008), Yerevan, Armenia (www.armstat.am) Porter, M. E. (1990), The Competitive Advantage of Nations, New York: The Free Press. Porter, M. E. (1998), On Competition , the Harvard Business Review Book Series, Boston, MA, United States UNESCO Institute for Statistics (2007), Online Database, Paris United Nations Conference on Trade and Development (2007), UNCTAD Handbook of Statistics On line, New York and Geneva, United Nations United Nations Conference on Trade and Development (2004), Development and Globalization: Facts and Figures, New York and Geneva, United Nations World Bank (2007a), Doing Business 2008 , Washington DC World Bank (2007b) World Development Indicators On line, Washington D.C. World Economic Forum (2006), The Global Competitiveness Report 2006 2007: Creating an Improved Business Environment , Hampshire: Palgrave Macmillan World Economic Forum (2007), The Global Competitiveness Report 2007 2008 , Hampshire: Palgrave Macmillan, New York, United States
ARMENIA ’S MILLENNIUM CHALLENGE ACCOUNT : ASSESSING IMPACTS ON ECONOMIC GROWTH AND POVERTY REDUCTION IN RURAL ARMENIA Kenneth Fortson, Mathematica Ester Hakobyan, MCA Armenia Anahit Petrosyan, MCA Armenia Anu Rangarajan, Mathematica Rebecca Tunstall, MCC * Abstract: The Millennium Challenge Corporation’s $236 million Compact with Armenia aims to reduce rural poverty through investments in irrigation infrastructure, rural road rehabilitation, and farmer training. This paper discusses the design for a rigorous impact evaluation of the farmer training project, which focuses on water management and cultivation of high value crops. The evaluation uses a random assignment design, whereby rural communities are randomly assigned to a treatment group, who are offered agricultural training early in the Compact, or a control group, who are not. We will then compare farmers’ outcomes in treatment group villages to farmers in control group villages. The results of the evaluation will assess the farmer training program’s success in accomplishing its key objectives: Increasing adoption of effective agricultural practices, increasing cultivation of higher value crops, improving farm productivity, increasing agricultural profits and household income, and reducing poverty rates. Moreover, the lessons learned in this context can be applied to other countries with similar economic climates so that future investments can be efficiently allocated to interventions with proven economic impacts. JEL Classifications: O12; O13; Q12 Keywords: Poverty; Agriculture; Training; Random Assignment; Armenia
* This project has benefited greatly from the hard work of numerous colleagues at MCC, MCA Armenia, and Mathematica. We are especially grateful to Emily Andrews, Sergey Balasanyan, Randall Blair, Barry Deren, Katherine Farley, Melik Gasparyan, Mary Grider, Ara Hovsepyan, Julie Ingels, Lusine Kharatyan, Sergey Meloyan, Daniel Player, Stephanie Roueche, Peter Schochet, and Sonya Vartivarian for their input throughout the project. We also thank VISTAA and ACDI/VOCA for facilitating a rigorous evaluation as they implement their programs, and AREG for implementing the baseline survey of farmer households that serves as the key data source for the evaluation.
I. INTRODUCTION In March of 2006 Armenia signed a $235.65 million agreement (the “Compact”) with the Millennium Challenge Corporation (MCC) with the goal of reducing rural poverty through improved performance of the country’s agricultural sector. Armenia plans to achieve this goal through a five year program of investments in rural roads, irrigation infrastructure, and technical and financial assistance to support farmers and agribusinesses. This paper will focus on the technical training in agricultural practices provided to Armenian farmers through the Compact’s Water to Market Activity (WtM), and particularly, the accompanying impact evaluation. The MCC program aims to train sixty thousand farmers in region specific water management techniques. These methods will help farmers to use water more efficiently, which can promote both cost savings (through water conservation) and increased quantity and quality of crops. Many farmers who will be trained in water management technologies will also receive training in higher value agricultural (HVA) methods. The combination of these activities is expected to increase beneficiaries’ average net income by about 25 percent. The Compact provides funding for an innovative impact evaluation to test the assumptions used in the pre Compact economic analysis by assessing the impact of the agricultural training on farmer productivity and income. The key research questions guiding the design of the evaluation for the agricultural training are: • Did the program affect the irrigation and agricultural practices of Armenian farmers? • Did the program affect agricultural productivity? • Did the program improve household well being for the targeted farmers, including income and poverty? The backbone of our research design is the use of random assignment to create two statistically equivalent groups of farmers, with the only difference between them being that one group can participate in training while the other cannot. This research design is considered the most rigorous methodological approach for estimating program impacts. Random assignment has been implemented to estimate the effects of programs in many contexts, and is especially widespread in developed countries (Michalopoulous, 2005; Kling, 2007). Recent research has also expanded the use of random assignment into studies of programs in developing countries (Duflo and Kremer, 2004). To our knowledge, this evaluation of the Compact’s WtM training program is the first large scale random assignment evaluation in Armenia. The next section provides a summary of the recent history of rural Armenia as context (Section II), followed by an overview of the Compact programs (Section III). Section IV provides a detailed description of the WtM activity. With this as background, we describe the evaluation design, beginning with the random assignment design in Section V. We then discuss the main source of data, the Farming Practices Survey (Section VI), followed by the Irrigation PIU data that could be used as a supplemental data source (Section VII). Lastly, we discuss in detail our econometric approach for estimating program impacts in Section VIII before concluding with a summary of the next steps in Section IX. II. RECENT DEVELOPMENTS IN ARMENIA In January 1991, Armenia began implementing a comprehensive land reform program. Rapid privatization in 1991 1992, when one third of all agricultural land and 70 percent of arable land were transferred to family farms and the Soviet style collective farms were abolished, created a base for a stable recovery of the sector. The decision to move rapidly reflected the fact that implementation of the land reform had the potential to boost a rapid supply of agricultural products, thus increasing the speed of moving to a market economy. In parallel with the land privatization, a set of reform measures designed to liberalize agriculture was implemented. Price controls were removed and food subsidies were abolished. The major support measures to local agricultural producers included VAT and land tax exemptions as well as subsidies for irrigation water. After the sharp decline of agricultural output at the beginning of the 1990s, the economy of Armenia started to recover beginning in 1994. The agricultural sector has continued to register steady growth since that time. However, agriculture has lagged behind other sectors of the economy such as the industrial, service, and construction sectors; as noted in Table 2.1, agriculture represents a shrinking share of GDP in spite of increased rates of growth. Table 2.1 GDP and Agricultural Product
Average Average 1994 1999 2000 2007 GDP Growth rate, percent 5.4 11.8 Agriculture Product Growth rate, percent 3.2 6.8 Share of Agriculture Product in GDP, percent 35.5 22.5 Source: Armenia in Figures, National Statistical Service of the Republic of Armenia Another feature of Armenian agriculture had been the unfavorable changes in relative prices. Average annual growth of agricultural prices in 1995 1999 was much slower than increases in industrial and consumer prices (Table 2.2). A similar pattern, though dampened, was registered during 2000 2007. In fact, lagging agricultural prices reflect both weaknesses of demand, which steadily improved during recent years, and weak market power of farmers compared to wholesales and food processors. The disproportionate price development indicates that the growth in agriculture was mostly beneficial to consumers and the food processing industry. Table 2.2 Changes in Price Indices Annual average Annual average for for 1995 1999 2000 2007 Consumer Price Index 143.6 103.2 Industrial Price Index 166.5 105.3 Agricultural Price Index 127.3 102.4 Source: Armenia in Figures, National Statistical Service of the Republic of Armenia
The lower growth rate of the agricultural sector compared with other sectors reflects the relatively lower productivity of the agriculture that is explained by different factors: • Small, fragmented farms (average land holding per family is 1.4 ha) limit the use of agricultural machinery. As a result production is labor intensive, and labor productivity is low relative to capital intensive production; • The poor condition of irrigation systems combined with climatic conditions of Armenia decreases efficiency of agricultural production; • Because a large share of the rural population, previously working in the non agricultural sector, became subsistence farmers in the 1990s, many farmers are inexperienced and lack basic knowledge about farming practices; • Inadequate use and availability of improved inputs such as seed, pesticides and appropriate fertilizer applications reduces the productivity of the sector; • Limited opportunities for storage, grading, packaging, and even processing of agricultural production reduces the value added and limits the supply of fresh produce throughout a year, and also makes agricultural prices volatile during a year; • Farmers have limited knowledge of the markets or access to information on demand; • Inadequate financial services are available for the rural sector and farmers have limited access to the financial resources; and • Collaboration among farmers in production and marketing of their products is limited. III. MCA-ARMENIA PROGRAM The Compact is designed to address many of the above issues and to increase agricultural productivity and reduce rural poverty in Armenia. MCA Armenia Program is addressing these obstacles through two projects: Rural Road Rehabilitation and Irrigated Agriculture Projects. The first project implemented under the MCA Armenia Program is the Rural Roads Rehabilitation Project. The implementation of the Project will expand the access of rural communities to agricultural markets and social infrastructure, as well as increase non farm income opportunities by improving the condition of rural roads. The Compact was designed to include MCC funding to rehabilitate up to 943 km of the rural roads from ‘very poor’, or ‘poor’, to ‘good’ condition. The project is expected to benefit hundreds of thousands of rural Armenians. The second project, the Irrigated Agriculture Project, will increase the profitability of the agricultural sector by extending and improving the irrigation network, strengthening irrigation entities to better manage the network, and supporting farmers to commercialize their production. For these purposes two types of activities will be implemented under the Irrigated Agriculture Project. The first activity, Irrigation Infrastructure Rehabilitation, will improve existing infrastructure in order to increase irrigated area and to improve the efficiency and sustainability of sourcing and delivery of water. It is expected that the Project implementation will result in expansion of irrigated areas, conversion of selected irrigation areas from pump to gravity irrigation to reduce energy consumption and make water more affordable, as well as reduce water losses in the irrigation network. The primary project beneficiaries are expected to be more than 115,000 farming households (or about 40 percent of all rural households) who will be able to increase the productivity of their irrigated land. The second activity in the Irrigated Agriculture Project, the Water to Market Activity (WtM), will accelerate the transition to more profitable, commercially orientated agricultural production by introducing and encouraging best practices in irrigated agriculture, fostering the adoption of improved water management techniques, shifting or expanding to higher value crops, strengthening the post harvest and processing enterprises linking producers to their markets, both domestic and international, and strengthening the capacity of credit providers to fund viable proposals in production and post harvest activities. IV. WATER -TO -MARKET TRAINING The focus of the paper is on the Water to Market Activity, specifically, how this activity impacts agricultural productivity and how the impact of the activity will be evaluated. The WtM Activity is designed to ensure that MCC Funding to the Irrigation Infrastructure Rehabilitation will contribute to a sustainable increase of agricultural productivity and incomes from farming. The Water to Market Activity includes four types of sub activities covering the whole chain from irrigation to the delivery of agricultural products to the consumers. MCA has contracted with ACDI/VOCA and their partners, VISTAA and Euroconsult (hereafter referred to collectively as ACDI) to implement the WtM activities that include training farmers in water management and high value agriculture, as well as credit and post production services.
Introduction of New On-Farm Water Management Technologies. The objective of the sub activity is to improve farmers’ skills in on farm water management techniques and their access to farm equipment to enhance the efficiency of water use in irrigation. The objective will be reached through implementation of the integrated program including in class training and on field demonstrations of the advanced and location appropriate on farm irrigation practices. Training and demonstrations will be provided to 60,000 farmers, of whom approximately 65 percent are expected to adopt water saving and productivity innovations that will increase the net benefit of their farming operations.
The demonstration farm is a critical part of the initial training and also can help reinforce the lessons even after training is completed. Farms are selected to serve as demonstration sites based on their proximity to other farmers in the village and the demonstration farmer’s willingness to adopt new technologies and facilitate other farmers’ understanding. ACDI provides equipment for the demonstration farms; in exchange, the demonstration farms are used as the site of the practical training, and also a resource where farmers in the village can go to see the technologies in practice, beyond the official training session. Demonstration farms will serve anywhere from one to five villages, depending on the number of eligible farmers in those villages and their proximity. Some demonstration farms exist from previous activities, and these will be supplemented with new demonstration farms. Most demonstration farms will be new. The trainers will be agricultural experts who come from the same region so as to ensure that they are familiar with the local climatic and agricultural conditions, and so that they are
available for technical assistance beyond the training sessions. Once a demonstration farm is established, ACDI will provide several rounds of training at that demonstration farm to saturate the associated villages as much as possible, because high participation rates will maximize the use of a single demonstration farm.
Coordinators will target farmers who are members of Water User Associations (WUAs). They will use posters to advertise at village centers and work with mayors to mobilize farmers to participate. In some villages, mayors may also be able to identify farmers who are particularly likely to participate, and these farmers could be targeted for additional recruitment efforts.
Transition to Higher Value Agriculture. The objective of this sub activity is to support the farmers’ transition to more profitable agricultural production through a combination of crop substitution, increased cropping intensity, and the use of higher yield generating planting materials and related inputs. Like the on farm water management training, the objective of this sub activity will also be reached through implementation of the integrated program including in class training and on field demonstrations of advanced, location appropriate farming practices. 30,000 farmers will be trained in HVA to increase the commercial value of their farms’ output. By the end of the Program, it is expected that approximately 7,800 hectares of land will be converted into higher value agricultural cultivation as a result of increased access to water combined with an effective training program.
Post-Harvest, Processing and Marketing. The objective of this sub activity is to introduce and expand post harvest operations that best preserve the quality of agriculture products and add value to production. For this purpose new strategies and technologies for storage, sorting and packaging, transportation and processing of agricultural products will be introduced. Improved supply relationships between post harvest enterprises and their farm suppliers will be another important aspect of this sub activity. The assistance to 12 firms includes diagnostic analysis of enterprises and providing recommendations for improvements, conducting food safety and hygiene training, providing weekly market information on fresh fruits and vegetables as well as other related activities. Reliable information on market conditions and opportunities will contribute to the better positioning of fresh and processed food products in the domestic, regional and international markets and create trade relationships. Food safety and quality assurance issues such as Hazard Analysis and Critical Control Point (HACCP) and International Standards Organization (ISO) certification will also be addressed through this sub activity. By the end of the Program, it is expected that these activities will impact 300 agribusinesses and 15,000 farmers.
Improved Access to Credit. The objectives of this sub activity are two fold. Improved irrigation and rural roads will present new opportunities for farmers and agricultural related businesses. However, improved irrigation and rural roads will create new opportunities for financial institutions as well. Therefore, the objective of this sub activity is to increase the availability of longer term, affordable credit to beneficiaries of the Project. To reach this objective, the sub activity will provide USD 8.5 million on lending resources to banks and credit organizations. The sub activity will also develop the capacity of banks and credit organizations to lend efficiently and prudently in the agriculture sector. The sub activity will provide assistance in applying for loans to the potential loan applicants as well. The assistance will improve the ability of the farmers to access financial resources and increase their awareness and understanding of credit. These actions will facilitate the intermediation process and reduce transaction costs and risk for credit providers by developing better informed and better prepared borrowers. Table 4.1 summarizes Water to Market Activity’s training and adoption targets for these four sub activities. Through these activities, the direct impact of the WtM Activity will be to generate new employment opportunities and increase income among farms and rural businesses. The synergy between rural infrastructure development and agribusiness development is designed to result in a significant, sustainable reduction in rural poverty. Table 4.1 WtM Adoption Targets
On Farm Water Management Training 60,000 farmers Higher Value Agriculture Training 30,000 farmers Post harvest Enterprises Technical Assistance 300 enterprises Bank Loans provided to project beneficiaries an d related businesses $8.5 million in loans Adoption of improved on farm water management 38,350 farmers Hectares Converted to High Value Agriculture 7,845 hectares The implementation of the program started in February of 2007. The implementation process includes two major stages. The first stage, the pilot stage, covered February September 2007 and it had a very specific purpose to provide lessons learned for full scale program implementation. The first stage was mostly focused on on farm water management training as a strategic tool to improve farmers’ skills in the on farm water management and irrigation techniques, as well as their access to new irrigation technology to enhance the efficiency of water use. Through demonstration and training, farmers are encouraged to adopt new or improved irrigation methodologies and methods which are effective and affordable for different categories of farmers. Therefore, the activities in demonstration sites have to suit all categories of farmers in terms of size and show a wide range of system improvements from the simple and inexpensive to the more costly ones. The general emphasis is on saving water and labor through proper irrigation scheduling, more uniform distribution of water within the farm and more effective delivery to the field. Development of skills and improved farmers’ knowledge are viewed as the most crucial factors for achieving a satisfactory level of adoption and sustainability of the expected changes. Full scale implementation of the Water to Market program started in October 2007. By the end of March 2008 the number of participants completing the On Farm Water Management training reached 8,186 farmers. Besides the training on on farm water management techniques, other activities were implemented as well. The number of participants completing training in High Value Agriculture, which helps farmers identify more profitable crops for their specific regions and best practices for cultivating those crops, totaled 495 farmers. V. USING RANDOM ASSIGNMENT TO ESTIMATE IMPACTS
The ideal method for separating program influences from other factors is to compare outcomes for the group who are provided the intervention with the outcomes for the same group if they were not offered the intervention. However, once persons or communities are offered the intervention, it is not possible to know what their outcomes would have been if they were not given the opportunity to participate. It can only be approximated by comparing their outcomes to the outcomes of some other group. Random assignment, the most rigorous way to measure program impacts, is frequently referred to as the gold standard of evaluation designs. Essentially, when implemented carefully, random assignment leads to the creation of two virtually identical groups at baseline, with the only difference being that only one group (the treatment group) is exposed to the intervention, while the other group (the control group) is not. As a result, any changes observed between the two groups over time can be attributed to the effects of the intervention with a known degree of statistical precision. (Michalopoulous, 2005; Kling, 2007; Duflo and Kremer, 2004) Unit of Random Assignment. Ideally, we would randomly assign individual farmers to receive training or not, and compare outcomes for the two groups. However, because these training sessions are community level interventions, making it difficult to exclude individual farmers, such an approach is not feasible in this context. Our basic approach is to randomly assign villages to the treatment group of farmers, who are eligible for on farm water management training and HVA training, or the control group of farmers, who are not. ACDI has grouped villages into clusters. Most clusters include only one village, but some include as many as five villages that are in close geographic proximity. All farmers who are WUA members and live in a cluster of villages selected for the treatment group will be permitted to participate in water management training. Farmers who are in the control group of village clusters would not be offered water management training until several years later. All villages will ultimately be provided training, and random assignment is used to determine when they are offered training. Randomly assigning entire village clusters in this way, rather than individual farmers or villages, guards against contamination of the control group the possibility that control group members get the same services as the treatment group. There are two types of contamination. The first type of contamination is if farmers in control group villages nonetheless participate in training. This could be problematic if control group members hear about the training activities and show up to training themselves. A different type of contamination could occur if farmers who participate in training teach farmers in the control group about the techniques they learned. Either of these types of contamination would be problematic for the evaluation because we would be unable to compare those who were offered training to those who were not offered training; with contamination, both the treatment and control group have access to or benefit from training. Generally, ACDI has chosen village clusters that are sufficiently far apart geographically to ensure that there is little chance that farmers in a control group village cluster would either participate in the training or learn about the water management techniques through other means. However, in some areas particularly the Ararat Valley region many villages are located in close proximity. While we cannot completely eliminate the possibility of contamination here, it will be important for the planned implementation to strive to avoid such contamination problems by, for example, ensuring that recruiting techniques for the training attract treatment group farmers without influencing control group farmers. The WtM training program exit questionnaire will also inform us about where farmers reside, which will help us assess the extent to which control group farmers are “crossing over” and receiving training in spite of being randomly assigned to not be eligible. Implementing Random Assignment. Random assignment was conducted for the subset of villages that have adequate water and could potentially be served early in the Compact. We randomly assigned villages to one of three groups: those who would be served in the second year of the Compact; those who would be served in either year 3 or year 4 of the Compact; and those who can be served in the final year of the Compact. The earliest group constitutes our treatment group, and the latest group our control group impacts will be measured after the treatment group has been provided training but before the control group has. The middle group, those who are served in the third or fourth year, will not be included in the impact evaluation. Only villages that were considered ready for WtM training were included in the randomization; some villages currently have poor sources of water, and thus, would not benefit from training until their irrigation systems are rehabilitated. Such villages may receive training in the future, but they will not be included in the impact evaluation. We also excluded from the random assignment all villages that were included in the pilot phase of the WtM training or where ACDI has already developed demonstration farms. Random assignment was conducted within strata defined by WUAs to preserve regional balance, which created balanced treatment and control groups along this dimension. The distribution of villages by treatment status for each agricultural zone is reported in Table 5.1 The probability of a village being assigned to the treatment group was approximately the same for all WUAs, with most of the deviations occurring due to rounding. An exception, however, is the Mountainous Zone, where a smaller proportion of villages were selected for the research groups (years 2 and 5), while most villages were assigned to the non research group. This zone was undersampled largely because MCC anticipates very low impacts, so the evaluation will focus more on the other zones where MCC is more optimistic about the prospects for improvement. A total of 120 clusters were assigned to the treatment group and 80 to the control group, with these 200 clusters containing 223 villages in total. (For simplicity of exposition, we hereafter refer to village clusters as “villages.”)
Table 5.1 Distribution of Village Clusters by Year of Training and Agricultural Zone
Yearly Ararat Valley Pre Mountainous Mountainous Sub Tropical Total Year 2: Treatment 44 58 12 6 120 Years 3 and 4: Nonresearch 18 19 38 2 77 Year 5: Control 28 38 10 4 80 Total 90 115 60 12 277 To ensure the process was transparent to all villages that will be served in the coming years, random assignment was conducted in public using a specially designed computer program that randomly sorted villages into the three groups.
VI. FARMING PRACTICES SURVEY The Farming Practices Survey (FPS) will serve as the primary data source for the impact evaluation. Approximately 25 interviews will be completed in each of the villages in the treatment and control groups, with fewer in the smallest villages and more in the largest. MCA Armenia has contracted with AREG to field the FPS in the first, baseline year, and the FPS will subsequently be conducted each year of the follow up period, at the end of 2008, 2009, and 2010. Sample Frame. With the help of village mayors, the FPS targets the households of farmers who are most likely to benefit from the training programs: those who are actively engaged in farming and have been tied to the community for several years. These farmers are identified through an iterative process. First, MCA Armenia requested that the WUAs work with village mayors to compile a list of farmers meeting our specific criteria in each village. The number of farmers requested depended on the size of the village, but averaged about 60. The sample was then drawn from these lists. Pretesting revealed that these lists were of mixed quality, however, often because the WUAs had not consulted with the mayors in compiling them. Thus, the sample was updated with the assistance of village mayors and marz officials, either at the marz offices or in the village itself. The mayors reviewed the lists to determine whether the farmers indeed met our criteria. If an insufficient number of farmers from the lists were eligible that is, in cases where the WUA had failed to consult with the mayor then the mayor helped AREG update the list in accordance with our survey eligibility criteria. Relying on mayors to identify eligible farmers in each village is not ideal, as the lists they provide may not be representative of the village as a whole. Unfortunately, a reliable sample frame was not available, and the cost of conducting a census in project communities is prohibitive. However, as the method AREG used mirrors the manner in which ACDI identifies farmers to participate in training, the sample will still provide good coverage of the population most likely to benefit from training. Moreover, because of the random assignment design, treatment and control villages should still be comparable (on average), preserving the internal validity of the evaluation. Intermediate Outcomes. While most of the outcomes of primary interest are longer term outcomes, such as economic improvements, these outcomes may not be immediately observable. Consequently, we will closely examine intermediate outcomes through which the training programs are intended to improve household income. We would expect an impact on households’ income only if we observe that a substantial proportion of the targeted farmers are actually participating in training, and perhaps most importantly, are then applying the techniques they learn. Examining the intermediate outcomes also establishes the counterfactual what services the villages would have received and what practices they would have adopted even in the absence of the WtM programs. Table 6.1 summarizes the key intermediate that can be examined using the FPS data. Table 6.1 Intermediate Outcome Measures
Intermediate Outcome Measures Time Frame
Participation in Agricultural Training. Whether attended any irrigation or agriculture training (including training sponsored by other sources); type Last Year of training attended (e.g., classroom, video, or practical); whether received a certificate indicating the full training was attended. Adoption of HVA and Irrigation Practices. Which irrigation practices were used, focusing on those taught in training sessions; whether those Last Agricultural Season practices had perceived time or labor savings. Investment in Agricultural Technology or Equipment. Ownership of personal reservoir or water pump; ownership or rental of trucks, tractors, Last Agricultural Season combines, seed planters, and sprayers. Cropping Patterns. Specific crops grown, especially high value crops; amount of land devoted to cultivation of each crop; total hectares of land Last Agricultural Season devoted to crops; whether household cultivates a kitchen plot; reason(s) for changes in cropping patterns. Final Outcomes. The ultimate goal of the MCA Armenia programs is to increase household income in rural Armenia, and hence, these outcomes are an important focus of the FPS instrument. Because a full accounting of all sources of household income would require far longer to administer than the allotted time for each interview, the survey concentrates on sources of income that are most directly affected by the training programs, specifically, income from agricultural production and from employment by the farmer and his or her immediate family. We can also use the average sale price of specific crops for other farmers in the village to monetize crops that are consumed by the household or bartered. Additionally, the FPS asks for estimates of expenditures on key categories of consumption, and for income from other sources. Table 6.2 summarizes the key final outcomes that can be examined using the FPS data.
Survey Nonresponse. All interviews are conducted in person, and a limited span of time is available for interviews in the majority of villages. Therefore, survey nonresponse is a concern. Substantial survey nonresponse can damage the validity of impact estimates. Nonresponse weights can account for some of the differential nonresponse, but only to the extent that nonresponse is explained by household characteristics that are known for both respondents and nonrespondents. In the worst case, survey nonresponse might be different for treatment and control villages, contradicting the core assumptions of a random assignment design. More commonly, however, survey nonresponse affects both the treatment and control groups equally. In this scenario, the impact estimates remain internally valid, but may not generalize beyond the select group of survey respondents. As a salient example, if farmers’ absences are due to trips to the markets to sell their produce, then the respondents may have a disproportionate share of the less engaged farmers, for whom program impacts could be minimal. We have instituted several safeguards against survey nonresponse. Working with mayors to clean the lists in advance can help in this regard. Whenever possible, village mayors would also contact the sampled households in advance to ensure they would be available for interviews on the day AREG visited their village. In instances where a household is not available on the first attempt, interviewers would return to the household throughout their time in that village. AREG also has reserve lists of farmers which they can draw on to help them reach their targets for completed interviews within each village.
Table 6.2 Final Outcome Measures Final Outcome Measures Time Frame Continuing Use of HVA and Irrigation Practices. Same as above, but focusing on changes in these practices relative to the initial follow up Last Agricultural Season years. Agricultural Production . Total amount of specific crops grown; amount of Last Agricultural Season crops grown per square meter; total value of all crops cultivated. Livestock. Number of cows, pigs, and sheep owned. As of Survey Date Revenue from Agricultural Production. Value of crops sold; total value of Last Agricultural Season all crops (including those sold, bartered, or consumed). Agricultural Costs. Expenditures on fertilizers, pesticides, irrigation Last Agricultural Season water, hired labor, rented equipment and taxes (individually and in total). Profit from Agricultural Production. Revenues minus c osts —the income Last Agricultural Season from agricultural activities. Income from Employment. Whether household head, spouse, and any grown children were employed (besides work on the family farm); total Last Month earnings from employment. Income from Pensions, Remittances, or Social Programs. Can also be added to profits and employment income to construct a rough measure of Last Month total income. Household Consumption. Expenditure on purchased food, health care, housing products, utilities, and transportation; cost of purchased goods Last Month/Last Year (converted from monthly to annual) plus value of crops consumed by the household.
Follow-up Surveys. Ideally, each round of the FPS would interview the same set of households, yielding a longitudinal data set. Analytically, longitudinal data allow for the cleanest estimation of program impacts, and also provide the most statistical precision, because changes from the baseline to the follow up period are not confounded with sampling variability. As a practical matter, however, it may not be as easy to track specific households from year to year. Our plan is to survey the same set of households in subsequent rounds of the FPS to the extent possible, but given the nonresponse issues described previously, we anticipate that these will need to be supplemented with additional households, yielding a mixed longitudinal repeated cross sectional data set. The sample frame will remain consistent; so as to avoid having the samples for treatment and control villages diverge over time. VII. IRRIGATION PIU DATA The Irrigation PIU has an impressive database for a subset of the WUAs, and by the end of the Compact, it is planned that information on members of all WUAs will be included in their databases. These data could be used to supplement the survey data in two important ways. First, they provide some outcome measures that would not be obtainable from farmers, such as energy use and water distribution. Second, they will provide data on some outcome measures for the entire population of registered WUA members in Armenia. These data items are defined in more general terms than the survey data for example, the amount of land the WUA member plans to grow wheat on, but not actual production but they can still be used to inform us about broad national and regional trends.
VIII. ESTIMATING PROGRAM IMPACTS Random assignment ensures that, on average, treatment group villages and control group villages are the same, with the exception that treatment group villages are offered WtM training. Hence, the difference between the mean of the outcome of interest for the treatment group and the mean for the control group yields an unbiased estimate of the WtM program’s impact. The precision of the impact estimates can be improved, however, by controlling for other covariates in a regression model. Regression adjustment can also help alleviate any differences between the treatment and control groups in baseline characteristics that arose by chance. Core Specification. The survey data will be cross sectional, with a new cross section of respondents drawn each year. 1 Given this data structure, our econometric specification is designed to compare how treatment group villages changed over time to how control group villages changed over time, controlling for idiosyncratic differences in the two groups. The basic model can be expressed as follows:
(1) yivt=β′ x iv + λ TFF v ×+ t θηε t ++ v ivt , where yivt is the outcome of interest for household i in village v at time t (where Ft = 0 in the baseline year and 1 in the follow up year); xiv is a vector of time invariant characteristics of household i in village v; t accounts for any time trends between the base and follow up years; Tv is an indicator equal to one if village v is in the treatment group and zero if it is in the control group; ηv is a village specific error term (a village “random effect”); and εivt is a random error term for household i in village v observed at time t. The parameter estimate for λ is the estimated impact of the program. The vector of baseline characteristics xiv will include both household and village characteristics. At a minimum, we will control for village characteristics such as the geographic region, WUA, and the baseline water conditions. We will also control for household size and composition, and characteristics of the household head, namely, education level, gender, age, and number of years farming. In the framework of a repeated cross sectional model, however, the characteristics that are included must be restricted to those that are unaffected by the WtM programs. We must be careful with land holdings, for example, as the WtM program could conceivably induce some farmers to cultivate more land, and controlling for it would therefore understate the full program impact. The model in equation (1) is designed to answer the general research question, “How have villages in treatment group changed from the baseline year to the follow up year, relative to villages in the control group?” This core model can be tweaked in a variety of ways to explore alternative specifications. A simple example would be to allow the 1 As described previously, there will be substantial overlap in the household samples from the baseline year and subsequent years, but the samples will likely not be identical. If, however, the survey in subsequent years uses the same sample, we will be able to employ panel (longitudinal) data models. The intuitive interpretation of panel data models is similar to models of repeated cross sectional data, but the estimation techniques differ somewhat from those described here.
time trends to vary across regions. The specification also (implicitly) weights all respondents equally, which could be modified to either give all villages equal weight, or weights equal to the village populations. Such explorations would not change the general interpretation of the impact estimate, but they can provide insights on two important issues. First, and of most direct interest, we can explore how robust the impact estimates are to these alternative specifications. Beyond this, however, the other regression covariates may be of independent interest, and may also provide context for interpreting the impact estimates. Pooled Model. Instead of using data from only the base year and one follow up year, we can also pool data from multiple waves of follow up year surveys. The econometric specification would be very similar to (1), but with a separate impact estimate for each of the n follow up years:
(2) yivt=βλ′ xTFTF iv +×+1122 v t λ v ×++ t... λ n TFFF v ×+ nt θθ 1122 t + t ++ ... θηε n F nt ++ v ivt , where Fnt = 1 if t = n and 0 otherwise. These impact estimates can then be compared to one another to see how program impacts changed over time, and could be particularly important to see whether any impacts on farming practices that are observed early on persist, and also whether impacts on longer term outcomes, such as agricultural productivity, grow after farmers have had more time to implement new techniques and benefit from their innovations. Clustering. The estimation techniques must take into account the correlation of outcomes for households in the same village, as they may be exposed to similar idiosyncratic influences that are not otherwise captured in the regression model, and therefore, the individual households cannot be considered statistically independent. As an example, a particular village might have abnormally good or bad weather, or could experience other economic “shocks” that are unrelated to the training program but nonetheless affect the entire village. The econometric models will account for this clustering with methods that allow flexibility in the correlation structure of the error terms. (Deaton, 1997).
Impact on Participants Only. Randomly assigning communities to be eligible for WtM training programs provides an unbiased estimate of the impact of offering this training in the villages selected for training the “intent to treat” (ITT) effect. The ITT effect combines the effect of the intervention on both participants and non participants in treatment villages. In many contexts, people who are offered program services but opt out of participating are unaffected by the program, while in other situations the program may nonetheless have within village spillover effects on the outcomes of non participants. By including questions about both participation in WtM training and adoption of WtM techniques, we will be able to determine whether there are sizable within village spillover effects present, and how best to account for them. When spillover effects are known to be minimal, a simple but powerful adjustment can be made to calculate the effect of the training program on participants the effect of “treatment on the treated” (TOT). This adjustment known colloquially as the Bloom adjustment calculates the effect of the training program on participants by dividing the estimated impact (the ITT) by the participation rate. The intuition for this elegant result is that, if the effect of the program on non participants is known to be zero, the estimated impact can be attributed entirely to the proportion of the treatment group that actually participated in training. Importantly, however, while the Bloom adjustment can potentially be used to account for non participation in the impact estimate, it cannot alleviate the problem non participation introduces for the variance of the impact estimate. If participation rates are low, we will not be able to detect impacts that are statistically reliable. (Bloom, 1984; Angrist, Imbens and Rubin, 1996) Subgroup Analysis. For many of the outcome measures, it is conceivable that the effects of the interventions will vary by observable characteristics. Estimating differential impacts on female headed households, for example, is of particular interest to MCC. We will examine whether the interventions’ effects differ for key subgroups defined by the characteristics of the households such as gender, age, and level of education of the household head; size of the household; or size of farm holdings operated by the household. Similarly, we will also examine how effects vary by subgroups defined by village characteristics. It is straightforward to embed subgroup estimates into the framework of equation (1). To do so, we include an interaction term that distinguishes treatment group members in subgroup S from those who are not in the subgroup:
(3) yivt=βλ′ x iv + S=1 TFS v ××=+ t( iv 1) λ S = 0 TFS v ××=+++ t ( iv 0) θηε F t v ivt In equation (3), the estimate of λS=1 represents the estimated impact for members of subgroup S, and we can test whether the impacts differ for members of that subgroup compared to everyone else by statistically testing whether λS=1 and λS=0 are equal. Distributional Effects. The implicit focus of the analysis plan outlined above is on examining differences on the mean household. In conducting the analysis, it is also important to examine whether the interventions’ effects vary at different levels of the outcome distribution. For example, the impact on agricultural real income for households with very low or very high income may differ from the impact on households at the mean. Specifically, the training programs may be such that only the higher income households will benefit, if, for example, implementing the techniques taught in training requires investment in equipment that lower income households cannot afford. Conversely, the poor in the community might benefit more than the wealthy if the training focuses primarily on techniques that are useful only to smaller scale farms. As Armenia has among the highest levels of income inequality in Europe, this distinction is not a trivial one. We will use quantile regression analysis to determine whether the intervention effects vary at different points in the distribution. Quantile regressions are analytically appealing because, similar to standard regression analysis, the quantile regression coefficients have direct and simple interpretation, thereby making it very appropriate for communicating impact estimates with policymakers. Estimating impacts for specified quantiles starts with the same regression model as a standard model. The difference is in the methodology for estimating the parameters, which in turn, affects the interpretation of those impact estimates. While a standard
regression model compares the impact for mean households, a quantile regression instead compares the impact of the interventions for a specified percentile, such as the 25th or the 75th percentile. Quantile regressions at the 50th percentile, the median, are also more robust to the influence of extreme outliers in the data, and thus can serve to validate the findings from standard regression analysis. (Koenker and Hallock, 2001; Deaton, 1997). IX. NEXT STEPS The analyses described above will be employed in a series of reports. The first report will cover the baseline FPS, and will be a short report focusing on the current state of the villages in the evaluation. The second report will cover the second round of the FPS, after the training programs have begun at least one round of training in most villages. This report will focus on the intermediate outcomes, to gauge participation rates and preliminary adoption rates for the new technology and practices. The final report will follow the fourth year of the Compact, the last year before the control group villages will become eligible for WtM training. This report will focus on the longer term outcomes, but as discussed, it will also examine intermediate outcomes such that we can assess not only whether there have been tangible impacts on poverty and household income, but also whether there is evidence from the intermediate outcomes that the full economic impact of the WtM may not have been fully manifested yet. The results from these studies will directly inform MCC and other international agencies on the efficacy of this approach for training in agricultural practices, and whether it is an effective pathway for reducing poverty in rural Armenia. Moreover, the lessons learned in this context can be applied to other countries with similar economic climates and will help shape future MCC Compacts with other countries.
REFERENCES Angrist, J., G. Imbens, and D. Rubin (1996), Identification of Causal Effects Using Instrumental Variables, Journal of the American Statistical Association, vol. 91, no. 434 Bloom, H. (1984), Accounting for No Shows in Experimental Evaluation Designs, Evaluation Review , vol. 8 Deaton, A. (1997), The Analysis of Household Surveys: A Microeconomic Approach to Development Policy, Johns Hopkins University Press: Baltimore, MD Duflo, E. and M. Kremer (2004), Use of Randomization in the Evaluation of Development Effectiveness , in O. Feinstein, G. Ingram, and G. Pitman (ed.) Evaluating Development Effectiveness, Transaction Publishers: New Brunswick, NJ Kling, J. (2007), Methodological Frontiers of Public Finance Field Experiments , National Tax Journal, vol. 60 Koenker, R. and K. Hallock (2001), Quantile Regression , Journal of Economic Perspectives, vol. 15, no. 4 Michalopoulos, C. (2005), Precedents and Prospects for Randomized Experiments, in Bloom, H. (ed.), Learning More from Social Experiments, Russell Sage Foundation: New York, NY National Statistical Service of the Republic of Armenia (Multiple Years). Armenia in Figures series, available at www.armstat.am
ՀՀՀոդվածների ամփոփագրերՀոդվածների ամփոփագրերոդվածների ամփոփագրեր Հայկական դրամի արժևորման ազդեցությունը տեղեկատվական տեխնոլոգիաներիտեխնոլոգիաների, , , , զբոսաշրջության և սննդամթերքի վերամշակման ձեռնարկությունների մրցունակության վրաձեռնարկությունների մրցունակության վրա Մհեր Բաղրամյան, Հայկական միջազգային տնտեսական հետազոտությունների խումբ Վահրամ Ղուշչյան, Հայկական միջազգային տնտեսական հետազոտությունների խումբ* Հայկական դրամը 2003 2006 թթ. արժևորվել է ավելի քան 40 տոկոսով: Լայնորեն տարածված տեսակետի համաձայն՝ անվանական փոխարժեքի այս կտրուկ փոփոխությունը բացասական ազդեցություն է ունեցել տեղական արտադրողների և հատկապես արտահանողների վրա: Հետազոտությունում օգտագործվել են 58 հայկական ընկերությունների ընտրանքային հետազոտության արդյունքները` պարզելու, թե արժևորումը ինչ ազդեցություն է ունեցել Հայաստանում զբոսաշրջության, տեղեկատվական տեխնոլոգիաների և սննդամթերքի վերամշակման ձեռնարկությունների մրցունակության վրա: 2003 2006թթ ին ընկերությունների տեխնիկական արդյունավետության գործակիցները և դրանց փոփոխությունը գնահատելու համար օգտագործել ենք ստոխաստիկ սահմանային մոդելավորման մեթոդը: Արժույթի արժևորման հնարավոր ազդեցությունը ընկերությունների եկամուտների և արտահանման ծավալների վրա պարզելու նպատակով տեխնիկական արդյունավետության գործակիցները օգտագործվել են ռեգրեսիայի հավասարումը գնահատելու համար: Հետազոտությամբ պարզվել է, որ փոխարժեքի փոփոխությունը հետևողական և վիճակագրորեն նշանակալի ազդեցություն է ունեցել ընկերությունների տեխնիկական արդյունավետության վրա: Պարզվել է նաև, որ փորձի կուտակումը տեխնիկական արդյունավետությունը պայմանավորող էական գործոն է հանդիսացել: Հետազոտությունում ուսումնասիրվել են փոխարժեքի և ընկերությունների տեխնիկական արդյունավետության, արտահանման ու եկամտաբերության միջև փոխկապակցությունները: Հետազոտությամբ պարզվել է, որ անվանական փոխարժեքի մեկ կետով արժևորման դեպքում տեղեկատվական տեխնոլոգիաների ոլորտի ձեռնարկությունների արտահանումը կրճատվում է տարեկան միջինը 66 հզ. դրամով (կամ շուրջ 200 ԱՄՆ դոլարով), սննդամթերքի վերամշակման ձեռնարկություններինը` 12 հզ. դրամով (կամ շուրջ 40 ԱՄՆ դոլարով), իսկ հյուրանոցների ու ներգնա զբոսաշրջությամբ զբաղվող ընկերությունների միջին տարեկան հասույթը՝ 112 հզ. դրամով (կամ շուրջ 340 ԱՄՆ դոլարով):
Փոխարժեքի դինամիկան ՀայաստանումՓոխարժեքի դինամիկան Հայաստանում Վահե Հեբոյան , Ջորջիայի Համալսարան Լեվել Գանթեր , Ջորջիայի Համալսարան Տպավորիչ տնտեսական աճին զուգընթաց՝ հայկական դրամի փոխարժեքը և՛ անվանական, և՛ իրական արտահայտությամբ զգալիորեն արժևորվել է, ինչը հասարակության տարբեր հատվածների մոտ անհանգստությունների տեղիք է տվել: Հայաստանի կենտրոնական բանկը և կառավարությունը մի կողմից, տնտեսագետները, գործարարները ու քաղաքական գործիչները մյուս կողմից տարբեր, գերազանցաբար հակադիր մեկնաբանություններ են տալիս արժևորման պատճառների վերաբերյալ: Սույն հետազոտության նպատակն է պարզել Հայաստանում փոխարժեքի դինամիկան պայմանավորող տնտեսական գործոնները: Փոխարժեքի երկարաժամկետ դինամիկան մոդելավորելու համար կիրառվել է վարքագծային հավասարակշիռ փոխարժեքի մոտեցումը: Հետազոտության արդյունքները ցույց են տալիս, որ հայկական արժույթը շեղվել է իր երկարաժամկետ հավասարակշիռ ուղղուց: Շեղման չափի գնահատականը զգայուն է դիտարկվող փոփոխականների համախմբի նկատմամբ: Թեպետ գնահատված մոդելներում փոփոխականները վիճակագրորեն նշանակալի են, իսկ շեղման ուղղությունը միանման է, սակայն շեղման մեծության գնահատականները տարբեր են: Լրացուցիչ վերլուծություն է անհրաժեշտ արդյունքների զգայունությունը փոփոխականների ընտրությունից պարզելու, ինչպես նաև, հաշվի առնելով փոփոխականների մեծամասնության նշանների տեսական ոչ միարժեքությունը, ստացված արդյունքներից ճշգրից եզրահանգումներ կատարելու համար: Բանկային համակարգի արդյունավետությունը , շուկայի կառուցվկառուցվածքըածքը ևևև օտարերկրյա սեփականությունը . ինչով էէէ պայմանավորված ավանդների ներգրավման ևևև վարկերի տրամադրման տոկոսադրույքների տարբերությունը ՀայաստանումՀայաստանում Էռա Դաբլա Նորրիս , Արժույթի միջազգային հիմնադրամ Հոլգեր Ֆլյորկեմայեր, Արժույթի միջազգային հիմնադրամ * Չնայած բանկային համակարգի լայնածավալ բարեփոխումներին, մակրոտնտեսական երկարատև կայունությանը և տնտեսական աճի բարձր տեմպին` ֆինանսական միջնորդության մակարդակով Հայաստանը դեռևս ետ է մնում այլ անցումային տնտեսությամբ երկրներից, իսկ ավանդային և վարկային տոկոսադրույքների տարբերութունը ավելի մեծ է, քան` Կենտրոնական ու Արևելյան Եվրոպայի անցումային տնտեսությամբ երկրների մեծ մասում: Սույն աշխատությունը ուսումնասիրում է բանկերի բնութագրիչների, շուկայի կառուցվածքի, օտարերկրյա սեփականության և մակրոտնտեսական գործոնների դերը Հայաստանում ավանդային և վարկային տոկոսադրույքների տարբերության, ինչպես նաև զուտ տոկոսային եկամտի որոշման գործում: Հետազոտությամբ
պարզվել է, որ բանկային բնութագրիչները, մասնավորապես` բանկի չափը, իրացվելիությունը, դիրքը շուկայում, ինչպես նաև շուկայի կառուցվածքը բացատրում են ավանդային և վարկային տոկոսադրույքների տարբերության և զուտ տոկոսային եկամտի միջբանկային և միջժամանակային տատանումների մեծ մասը: Ի տարբերություն այլ անցումային տնտեսությամբ երկրների`ուղղակի կապ գոյություն չունի օտարերկրյա սեփականության ու ավանդային և վարկային տոկոսադրույքների տարբերության, ինչպես նաև զուտ տոկոսային եկամտի միջև: Սակայն, բանկերի արդյունավետության տեսանկյունից, նշանակություն ունի՝ արդյոք բանկը օտարերկրյա սեփականություն է, թե՝ ոչ։ Արտադրողականությունը և ձեռնարկությունների աաարդյունավետության գործոնները Հայաստանումարդյունավետության գործոնները Հայաստանումրդյունավետության գործոնները Հայաստանում
Կարեն Գրիգորյան , Համաշխարհային բանկ Վահրամ Ստեփանյան, Համաշխարհային բանկ *
Հոդվածում վերլուծվում են արտադրողականության աճի գործոնները Հայաստանում և առաջարկվում են ապագա կայուն աճի խրախուսմանը միտված միջոցառումներ: Ամփոփված են զարգացած և անցումային երկրներում ձեռնարկությունների արտադրողականության տեսական գործոնները և փաստական արդյունքները, առանձնակի ուշադրություն է դարձված այնպիսի առանցքային գործոնների, ինչպիսիք են մրցակցությունը, օտարերկրյա ուղղակի ներդրումները և ինստիտուցիոնալ միջավայրը: Հետազոտության առանցքը 2003 2005թթ–ին արդյունաբերության և ծառայության ոլորտների 300 հայկական ձեռնարկությունների ընտրանքային հետազոտության արդյունքների վերլուծությունն է` ընդհանուր գործոնային արտադրողականությունը գնահատելու նպատակով: Հետազոտությամբ պարզվել է, որ Հայաստանում տեխնիկական արդյունավետության և նաև ընդհանուր գործոնային արտադրողականության նշանակալի աճ տեղի չի ունենում` ի հակադրություն որոշ հետազոտությունների արդյունքների: Ձեռնարկություններում տեխնիկական առաջընթացը, ըստ երևույթին, Հայաստանի տնտեսական աճի էական գործոն չի հանդիսացել: Հետազոտության արդյունքները վկայում են, որ ընդհանուր գործոնային արտադրողականությունը արդյունաբերության և ծառայության ոլորտների ճյուղերում էապես չի տարբերվում: Գրեթե որևէ տարբերություն չի բացահայտվել նաև արտահանող և արտահանում չիրականացնող ձեռնարկությունների ընդհանուր գործոնային արտադրողականությունում, ինչը վկայում է, որ արտահանման շուկաներում մրցակցային ճնշումը չի նպաստել հայկական ձեռնարկությունների արտադրողականության աճին: Վերջինը ցույց է տալիս, որ նոր տեխնոլոգիաների և գիտելիքի տարածմանը ու աշխատուժի հմտությունների բարելավմանը միտված քաղաքականության հրատապ անհրաժեշտություն կա:
Ընդհանուր գործոնային արտադրողականության միջին գնահատված արժեքները գործնականում նույնն են նաև տեղական և օտարերկրյա սեփականությամբ ընկերություններում, ինչը ցույց է տալիս, որ ինստիտուցիոնալ միջավայրում առկա են լուրջ թերություններ, որոնք արգելակում են օտարերկրյա սեփականությամբ ձեռնարկություններից տեխնոլոգիաների և գիտելիքի փոխանցումը տեղական ձեռնարկություններին: Անորոշ է նաև մրցակցության դրական ազդեցությունը ընդհանուր գործոնային արտադրողականության վրա, քանի որ Հայաստանում մրցակցային ճնշումները դեռևս բավականաչափ ուժեղ չեն` կառավարչական որոշումների վրա ազդելու համար: Հայաստանում մակրոտնտեսական նշանակալի առաջընթացը թերևս պայմանավորված է անցման սկզբնական փուլում իրականացված հաջող բարեփոխումներով , հատկապես ` գների և արտաքին առևտրի ազատականաց ٳٵ և մասնավորեց ٳٵ : Սակայն « եկել է երկրորդ սերնդի բարեփոխումների ժամանակը ` ինստիտուցիոնալ և մրցակցային միջավայրը բարելավելու և օտարերկրյա ուղղակի ներդրումները խթանելու համար : ՀայաստանիՀայաստան ի իի մ մասնավոր հատվածի մրցունակությունըմմասնավոր հատվածի մրցունակությունը. ասնավոր հատվածի մրցունակությունը . անցում հաջորդ փուլ. . անցում հաջորդ փուլ անցում հաջորդ փուլ Մանուկ Հերգնյան, “Տնտեսություն և արժեքներ” հետազոտական կենտրոն Գագիկ Գաբրիելյան, “Տնտեսություն և արժեքներ” հետազոտական կենտրոն Աննա Մակարյան, “Տնտեսություն և արժեքներ” հետազոտական կենտրոն Հոդվածում ուսումնասիրվում են Հայաստանի տնտեսական ձեռքբերումները և վերլուծվում է Հայաստանի մրցունակությունը` այն պայմանավորող գործոնների ու պատճառահետևանքային կապերի հետազոտության միջոցով: Հայաստանի տնտեսական նկարագիրը և մրցունակության միկրոտնտեսական հիմքերը համադրվում են նման կամ մրցակից երկրների հետ: Հոդվածում ներկայացված են առաջարկություններ` Հայաստանի հետագա զարգացման և ազգային մրցունակության բարելավման վերաբերյալ։ Առաջարկությունները առնչվում են ապագա քաղաքականությունների համատեքստին, նախապատվությունների սահմանման ու կարճաժամկետ և երկարաժամկետ զարգացման ռազմավարությունների մշակման մեխանիզմներին, տարածաշրջանում և գլոբալ տնտեսությունում Հայաստանի դիրքավորման մոտեցումներին:
Հայաստանի Հազարամյակի մարտահրավեր հիմնադրամհիմնադրամ. . . . գյուղական բնակավայրերում տնտեսական աճի և աղքատության կրճատման ազդեցության գնահատազդեցության գնահատումումումում Կեննեթ Ֆորստոն, «Մաթեմատիկա – քաղաքականության հետազոտում» ընկերություն Էսթեր Հակոբյան, Հազարամյակի մարտահրավեր հիմնադրամ – Հայաստան Անահիտ Պետրոսյան, Հազարամյակի մարտահրավեր հիմնադրամ – Հայաստան Անու Ռանգարաջան, «Մաթեմատիկա – քաղաքականության հետազոտում» ընկերություն Ռեբեկա Թանսթալլ, Հազարամյակի մարտահրավերներ կորպորացիա*
Հազարամյակի մարտահրավերներ կորպորացիայի 236մլն. ԱՄՆ դոլար արժողությամբ ծրագիրը Հայաստանում նպատակ ունի կրճատել աղքատությունը գյուղական բնակավայրերում` ոռոգման ենթակառուցվածքում ներդրումների, գյուղական ճանապարհների վերանորոգման և ֆերմերների վերապատրաստման միջոցով: Հոդվածում քննարկվում է ֆերմերների վերապատրաստման ծրագրի ազդեցության գնահատումը` ջրի կառավարման և բարձրարժեք մշակաբույսերի մշակման բնագավառում: Գնահատման համար կիրառվում է պատահական ընտրության սկզբունքը, որի համաձայն գյուղական համայնքները պատահականորեն ընտրվում են որպես ազդեցության խումբ, որոնց առաջարկվում է համապատասխան վերապատրաստում, կամ որպես ստուգիչ խումբ, որոնց նման վերապատրաստում չի առաջարկվում: Այնուհետև համեմատվում են ազդեցության և ստուգիչ խմբերում ֆերմերների ցուցաբերած արդյունքները: Գնահատման արդյունքները կօգտագործվեն վերապատրաստման ծրագրի արդյունավետությունը, իր հիմնական նպատակների տեսանկյունից, գնահատելու համար: Այդ նպատակներից են արդյունավետ գյուղատնտեսական արտադրության ներդնումը, բարձրարժեք մշակաբույսերի ցանքատարածությունների ընդլայնումը, ֆերմերային տնտեսությունների արտադրողականության և եկամուտների բարձրացումը, աղքատության կրճատումը: Ստացված արդյունքները կարող են օգտագործվել նաև նման տնտեսական միջավայր ունեցող այլ երկրներում, այնպես, որ ներդրումները արդյունավետորեն տեղաբաշխվեն այն ոլորտներում, որոնց տնտեսական ազդեցությունը հայտնի է: