Food Insecurity in Developing Economies

FOOD INSECURITY IN DEVELOPING ECONOMIES:

CAMBODIAN AND INTERNATIONAL EVIDENCE

KIMLONG CHHENG

A THESIS SUBMITTED FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY

OF THE AUSTRALIAN NATIONAL UNIVERSITY

DECLARATION

I declare that this thesis is my own work, unless otherwise indicated.

Kimlong Chheng February 2018

ACKNOWLEDGEMENTS

I am indebted to the panel of my academic supervisors for their advice, guidance, and support leading to the successful completion of this thesis. Professor Budy P. Resosudarmo, chair of the supervisory panel, has been instrumental throughout my PhD program. I am grateful for his excellent supervision and patience since the early stage of my consultations with him in developing and streamlining the research focus, to checking the progress of my research work, to reviewing my many drafts, and to sharing his knowledge and expertise. I am fortunate to have Dr Paul J. Burke, my advisor and assistant supervisor, to supervise my dissertation. Dr Burke has been exceptional in providing both technical and moral support. I thank him for encouraging me to participate in academic and scholarly research events and professional development. I am grateful to Dr Sarah Xue Dong for her support and attention to the detail of analytical parts of my thesis. Dr Dong has been very resourceful and very kind to share her experience and knowledge to improve the writing of this thesis. I acknowledge the financial support from the Government of Australia through the Endeavour Awards postgraduate scholarship, which has made my PhD program and research at the Australian National University possible. My thanks also go to Dr Sommarat Chantarat, Professor Yasuyuki Sawada, and Dr Sothea Oum for inviting me to participate in the household survey in 2014. I also thank the Economic Research Institute for ASEAN and East Asia for a research grant. I am grateful to them for allowing me to use the survey data in this thesis. Additionally, I thank the survey team members, namely Dr Vathana Sann from the Council for Agriculture and Rural Development, Minea Kim, Kakada Kuy, Vanna Meas, and all the research assistants for their assistance and support during the household survey. My thanks also go to Dr Ryan B. Edwards, Yessi Vadila, Moh Widodo, Umbu Raya, Dr Michael Cabalfin, Dr Marcel Schroder, Dr Matthew McKay, Dr Wee Koh, Dr Omer Majeed, Arndt- Corden Department of Economics (ACDE) faculty members, and Crawford School PhD colleagues for suggestions to improve the chapters in this thesis. Any errors are my own. There are many people whom I am thankful for their various kinds of support. I sincerely thank Melissa Sweeney, Gina Lopez, Professor Hal Hill, Professor Prema-chandra Athukorala, Professor Bruce Chapman, Dr Robert Sparrow, Dr Blane Lewis, Dr Ross McLeod, Dr Arianto Patunru, Dr Heeok Kyung, Sandra Zec, and other faculty members of

iii

the Arndt-Corden Department of Economics and the Crawford School of Public Policy at the Australian National University for their excellent support. I am grateful for the comments received from the audience during my presentations at the Arndt-Corden Department of Economics PhD Seminar, at the Crawford PhD Conference, and at the University House Student Conference. Finally, I thank the audience members at an Australian Agricultural and Resources Economics seminar, held in Canberra, for their suggestions and feedback to improve the analysis and writing of my chapters in this thesis. Thanks also go to Karin Hosking, who copyedited this thesis. I dedicate this thesis to the following people. In the loving memory of my mother, Chheng Kuoy, whose love, care, and parental virtues she provided to me and the family have been eternal and unsurpassed. My father, Chheng Im, has taught me moral and ethical values since my young age, and most importantly his firm belief in the value of education and freedom of choice. My great aunt, Kuong Kuoy, has taken the place of my mother to take care of the family to ensure that everyone in the family was properly fed and healthy. Finally, thanks are extended to my brother and sisters, nieces and nephews for their unlimited reservoirs of support and encouragement. I especially thank and dedicate this thesis to my wife, Shoko Kudo, whose love, care, support, and encouragement are unsurpassed. During my studies, she has been a mentor, advisor, and companion who always ensured that I made continuous progress leading to the successful completion of the thesis. I thank my in-laws who have been supportive during my PhD program. I also dedicate this research work to my ancestors and all relatives. Finally, I thank the Cambodian-Australian communities, including the Cambodian Student Association in Canberra (CSAC), for their moral support throughout my PhD program in Australia.

ABSTRACT

The seriousness of food insecurity in many developing economies has prompted this research into its key potential drivers. The thesis assesses primary data on Cambodia as a case study to examine potential impacts of (i) agricultural land property rights on household food insecurity of rice farmers in rural Cambodia, (ii) of excessive flooding and irrigation on rice productivity and rice revenue, and (iii) of rice productivity and rice revenue on household food insecurity. The primary data are taken from a household survey conducted between March and May 2014, administered to 256 households in 32 rural villages in rural Cambodia. The second part of the thesis examines (iv) the international experience of private property rights impacts on food insecurity, using data from 57 developing economies over 1990 to 2011. This cross-country examination is motivated by the Cambodian evidence to investigate whether the international evidence on the linkage exists. The plot-level evidence from Cambodia indicates that a one-unit increase in security in agricultural land property rights could reduce household food insecurity by about 1 day per annum on average. Security in agricultural land property rights could improve credit access, collateralisation, and farmers’ revenue-cost ratios. For rural rice farmers in Cambodia, simply holding ‘land documents’ of any type does not appear to have a strong impact on their food insecurity. The international evidence provides similar results to the Cambodian evidence: countries with greater private property rights experienced less food insecurity. A one-percent increase in property rights security potentially reduces prevalence of undernourishment and prevalence of food inadequacy by 0.85 percent and 0.64 percent on average, respectively. The plot-level evidence from Cambodia shows that providing irrigation for the currently unirrigated plots could raise per-harvest rice yield by about 0.7 tonnes and per- harvest rice revenue by about USD150 on average. Expanding access to formal irrigation, i.e., from reservoirs, dykes, or canals, could raise rice yield and rice revenue by about USD200 per harvest, relative to other irrigation types, such as river or groundwater irrigation. The household-level evidence from Cambodia shows that rice productivity and rice revenue are significantly, negatively associated with household food insecurity. Plots affected by excessive flooding had lower rice yield by about 0.7 tonnes per hectare, lower per-harvest

rice revenue by about USD150, or lower per-hectare rice revenue by about USD140 on average, relative to those plots unaffected by excessive flooding. The thesis has identified four policy options for tackling food insecurity in Cambodia and developing economies. First, strengthening security in private agricultural property rights is an option for reducing household food insecurity in rural Cambodia. Relatedly, greater security in agricultural land property rights would improve credit access and land-based collateral use. Second, enhancing security in private property rights in developing economies would be key for lowering their national food insecurity. Third, expanding formal irrigation access and is another option for improving rice production and rice revenue. Fourth, strengthening mechanisms to cope with excessive flooding in rice-producing areas in rural Cambodia is key for improving rice production and rice revenue.

TABLE OF CONTENTS

DECLARATION ...... ii ACKNOWLEDGEMENTS ...... iii ABSTRACT ...... v TABLE OF CONTENTS ...... vii LIST OF TABLES ...... x LIST OF FIGURES ...... xi LIST OF ABBREVIATIONS ...... xii

CHAPTER 1 INTRODUCTION ...... 1 1.1 Motivation of this thesis ...... 1

1.2 Scope and research questions ...... 4

1.3 Goals of this thesis ...... 5

1.4 Contributions of the thesis ...... 5

1.5 Food (in)security ...... 7

1.6 Land property rights ...... 7

1.7 Methods ...... 8

1.8 Thesis organisation ...... 10

CHAPTER 2 IMPACT OF LAND PROPERTY RIGHTS ON FOOD INSECURITY IN RURAL CAMBODIA ...... 11 2.1 Introduction ...... 11

2.2 Literature review ...... 12

2.3 Land property rights in Cambodia ...... 15

2.4 Basic model ...... 17

2.5 Data collection ...... 19

2.6 Estimation strategy ...... 33

2.7 Results ...... 34

2.8 Possible channels ...... 41

vii

2.9 Conclusions ...... 44

CHAPTER 3 FLOOD RISK, RICE PRODUCTIVITY, AND FOOD INSECURITY IN RURAL CAMBODIA ...... 46 3.1 Introduction ...... 46

3.2 Literature review ...... 48

3.3 Flood risk and irrigation systems in Cambodia ...... 50

3.4 Basic model ...... 53

3.5 Data collection ...... 56

3.6 Estimation strategy ...... 67

3.7 Results: Flood risk, irrigation, and rice production ...... 67

3.8 Results: Rice production and food insecurity ...... 75

3.9 Conclusions ...... 78

CHAPTER 4 PROPERTY RIGHTS AND FOOD INSECURITY IN DEVELOPING ECONOMIES ...... 80 4.1 Introduction ...... 80

4.2 Literature review ...... 83

4.3 Basic model ...... 86

4.4 Measurement of property rights ...... 88

4.5 Measuring food (in)security indicators ...... 89

4.6 Other data ...... 91

4.7 Empirical strategy ...... 97

4.8 Results ...... 98

4.9 Robustness tests ...... 103

4.10 Conclusions ...... 108

CHAPTER 5 CONCLUDING REMARKS ...... 110 5.1 Introduction ...... 110

viii

5.2 Key findings ...... 112

5.3 Policy implications ...... 113

5.4 Suggestions for future studies ...... 115

REFERENCES ...... 116

APPENDICES ...... 134 Appendix 1: Questionnaire for the household survey in Cambodia ...... 135

Appendix 2: Test results excluding China and random and fixed effects results in Chapter 4 .... 160

Appendix 3: Descriptive statistics for indicator of security in land property rights by village and

province ...... 168

Appendix 4: Summary results for the three empirical chapters ...... 169

LIST OF TABLES

Table 2.1 List of the survey fields ...... 21 Table 2.2 Descriptive statistics at household level ...... 26 Table 2.3 Correlations of variables ...... 29 Table 2.4 Basic data on food insecurity, land property rights, and credit access ...... 30 Table 2.5 Descriptions of variables ...... 32 Table 2.6 Impact of agricultural land property rights on household food insecurity ...... 37 Table 2.7 Impact of agricultural land property rights on length of household food insecurity ...... 39 Table 2.8 Average marginal effects of agricultural land property rights on household food insecurity ...... 40 Table 2.9 Possible channels ...... 42 Table 3.1 Plot-level and household-level basic survey data ...... 58 Table 3.2 Descriptive statistics of plot-level data ...... 59 Table 3.3 Descriptive statistics of household-level data ...... 60 Table 3.4 Descriptions of variables ...... 56 Table 3.5 Flooding and irrigation impacts on total rice yield of the latest harvest at plot level ...... 65 Table 3.6 Flooding and irrigation impacts on per-hectare rice yield of the latest harvest at plot level ...... 69 Table 3.7 Flooding and irrigation impacts on total rice revenue of the latest harvest at plot level ...... 71 Table 3.8 Flooding and irrigation impacts on per-hectare rice revenue of the latest harvest at plot level ...... 72 Table 3.9 Impact of rice productivity and rice revenues on household food insecurity ...... 76 Table 3.10 Impact of rice productivity and rice revenues on length of food insecurity ...... 77 Table 4.1 List of variables and data sources for developing economies ...... 93 Table 4.2 Descriptive statistics ...... 94 Table 4.3 Correlation of variables ...... 95 Table 4.4 List of 57 developing economies in the sample ...... 96 Table 4.5 OLS results over 1990‒2011 ...... 98 x

Table 4.6 Between estimator panel results (1990‒2011) ...... 101 Table 4.7 Robustness OLS results: Prevalence of undernourishment and property rights measures ...... 105 Table A4.1 OLS results over 1990‒2011 (Excluding China) ...... 160 Table A4.2 Between estimator panel results (1990‒2011) (Excluding China) ...... 161 Table A4.3 Robustness OLS results: Prevalence of undernourishment and property rights measures ...... 162 Table A4.4 Random effects panel results (1990‒2011) ...... 164 Table A4.5 Fixed effects panel results (1990‒2011) ...... 165 Table A4.4 Random effects panel results (1990‒2011) (Excluding China) ...... 166 Table A4.5 Fixed effects panel results (1990‒2011) (Excluding China)...... 167 Table A5.1 List of estimation results ...... 169

LIST OF FIGURES

Figure 2.1: Map of the household survey fields ...... 20 Figure 3.1 Irrigation intensity in Cambodia vs. Thailand and Vietnam ...... 53 Figure 4.1 Log of property rights protection against log of prevalence of undernourishment in 57 developing economies, 2011 ...... 82 Figure 4.2 Conceptual framework for factors influencing food security ...... 86

xii

LIST OF ABBREVIATIONS

AAPR Annual average precipitation rate ACDE Arndt-Corden Department of Economics ADB Asian Development Bank ADESA Average dietary energy supply adequacy AGCOM Agriculture share of Gross Domestic Product BE Between estimator CA Credit access CF Annual crop frequency COL Collateral CPI Consumer price index DFD Depth of food deficit EFW Economic Freedom of the World ELC Economic land concession ERIA Economic Research Institute for ASEAN and East Asia ETHNO Ethnolinguistic fractionalisation FAO Food and Agriculture Organization of the United Nations FE Fixed effects FI Food insecurity GCR Global Competitiveness Report GDP Gross Domestic Product GNI Gross National Income HC Human capital index HHFCEPC Per capita household final consumption expenditure INFL Inflation rates IPRI International Property Rights Index LATIT Latitude LFI Length of food insecurity LPM Linear probability model LPR Land property rights xiii

MAFF Ministry of Agriculture, Forestry and Fisheries ML Maximum likelihood MLE Maximum likelihood estimator MLM Maximum likelihood model MOWRAM Ministry of Water Resources and Meteorology OLS Ordinary least squares OPEN Openness OVB Omitted variable bias PFI Prevalence of food inadequacy POP Population PRA Property Rights Alliance PPRS Physical property rights score PRRG Property rights and rule-based governance PR Property rights PRP Property rights protection PU Prevalence of undernourishment PWT Penn World Table RCR Revenue-cost ratio RE Random effects REGPR Registering property

REVp Rice revenue at plot level

REVh Rice revenue at household level RRRP Regulatory restrictions on sale of real property SP Soil problem Unicef United Nations Children’s Fund USAID United States Aid for International Development USDA United States Department of Agriculture

Yp Rice yield at plot level

Yh Rice yield at household level

xiv

CHAPTER 1

INTRODUCTION

1.1 Motivation for this thesis Food insecurity in many developing economies remains a serious development issue. About one in nine of the global population is facing persistent and chronic food insecurity (FAO, 2013, 2014). About 780 million out of almost 800 million global undernourished in 2014‒16 were in developing countries (FAO, 2015). The number of undernourished people in the world has fallen by 216 million since 1990‒92, down 21.4 percent. The share of the undernourished in developing countries fell from 23.3 percent in 1990‒92 to 12.9 percent in 2014‒16. However, in 2015 only 72 out of 129 developing countries reached the hunger target of the Millennium Development Goals. Between 1990‒92 and 2014‒16, the number of the undernourished in Sub-Saharan Africa rose from 176 million to 220 million. Its regional share of the undernourished rose from 17.4 to 27.7 percent. Over the same period,

the level of hunger prevalence in South Asia declined from 291 million to 281 million. But its regional share of hunger rose from 28.8 to 35.4 percent. Both SA and Africa have suffered serious deficiencies in micronutrients, such as essential vitamins and minerals for healthy physical growth and development (FAO, 2013). Food insecurity occurs when a person or household has an insufficient intake of nutritious food, i.e., below a minimum daily dietary energy requirement of caloric intake to remain healthy and active (FAO, 2015; USAID, 2016). Some people might have experienced food insecurity more persistently than others. Insufficient intakes of nutrition or poor diets in developing countries may have resulted from their restricted access to food. Chronic or persistent food insecurity could cause many adverse consequences, including (i) worsening human capital, (ii) lost labour productivity, and (iii) lost agricultural productivity. At a micro level, it could affect almost every social facet (Jones et al., 2013), such as health, ability of children to stay in school and earn income in the long run, and child mortality. Every year, undernutrition causes almost 45 percent of deaths among those under five years of age globally (Unicef, 2014). That amounts to around 3.1 million deaths, of which 1.1 million deaths are caused by deficiency in micronutrients, essential minerals and vitamins required 1

for body growth and cognitive health (IFPRI, 2014). Similarly, about 2.8 million children and 300,000 women in developing countries die every year because of malnutrition (Guha- Khasnobis et al., 2007). Multiple causes of food insecurity have been debated. For example, insecurity in land property rights has been highlighted as a driver of food insecurity in developing economies (Maxwell & Wiebe, 1999; World Bank, 2003). Most rural poor in many developing countries engage in agriculture for food and income. Attenuation of their agricultural land property rights could possibly have prevented them from productive activities, making them worse off. Existing evidence tends to point to positive impacts of strong land property rights on land productivity and investments in land-related crop production (Newman et al., 2015) and income gains (Lawry et al., 2016). In contrast, providing secure access to agricultural land for farmers for their productive and creative use could be key to enhancing their livelihoods (Lawry et al., 2016). Natural disaster could be another driver of food insecurity in developing economies (FAO, 2000). While annual floods provide irrigation in agriculture, bring nutrients, and stabilise soil conditions, disastrous or catastrophic flooding can cause a lot of damage, particularly to rural agriculture. For example, when severe flooding becomes more frequent and worsens, production of rice and other food crops could face higher risks. Rice farmers who rely on rice cropping for income and food might then be unable to enhance crop productivity and secure food sources for consumption. Other potential drivers of food insecurity include, but are not restricted to, poor crop productivity (Markussen, 2008), high prices (Chan, 2011; CDRI, 2008; Maltsoglou et al., 2010; Warr, 2014), poverty or low income (Timmer, 2000 & 2005; Banerjee & Duflo, 2007), population growth (FAO, 2012), and trade restrictions (Gillson & Fouad, 2015). These factors may explain why food is insecure for large pockets of poor people in the developing world. In fact, a country can import food to overcome food insecurity (Gillson & Fouad, 2015), without having to produce food. However, many rural people in developing economies may not be able to purchase food imports because of a limited income from agricultural activities. In the case of Cambodia, food insecurity is an important issue, especially among the rural population. Between 25 and 28 percent of rural Cambodians, or almost four million people, were food insecure for about two to three months in 2015 (USAID, 2016). For around

70 to 80 percent of rural Cambodians, rice cropping is their main source of food and income (USAID, 2016). Furthermore, rice consumption constitutes two-thirds of their total caloric intake (Maltsoglou et al., 2010). Insecurity in private land property rights, shortage of irrigation, and annual flooding have been debated as drivers of food insecurity in Cambodia. Post-conflict Cambodia is rife with weak land property rights. Private property rights, including land rights, were abruptly abolished during the genocidal regime of the Khmer Rouge (1975‒1978). After about two decades of civil wars that ended in 1998, land property rights reforms were reignited. Only an estimated 20 to 30 percent of rice plots in Cambodia have received legal land titles (USAID, 2016). Conversely, between 70 and 80 percent of rural agricultural land is at risk of land dispossession in land grabs or land acquisitions, including under the government’s economic land concession (ELC) schemes. Landlessness due to land losses or stressed land sales has risen about two percent annually since around 1990 (So et al., 2001). The current rate of landlessness stands at 29 percent among rural households (Phann et al., 2015). An infringement on agricultural land rights may threaten rural farmers’ crop production, income, and food security. Cambodia suffers floods and droughts, which might have influenced food insecurity of many rural Cambodians. While the country is prone to these disasters, its irrigation infrastructure and flood-control systems are limited. Only about 24 percent of agricultural land in Cambodia is irrigated (FAO, 2012; USAID, 2010). Low crop productivity in Cambodia has been driven by a lack of agricultural infrastructure and limited access to high- yielding seeds and fertilisers (USAID, 2013). The lack of agricultural infrastructure, coupled with extreme climate conditions, may be among the major causes that destabilise food systems and affect crop production capabilities of rural farmers. Instability in food systems and agricultural production systems may threaten food availability (supply), food access (price), and food utilisation (consumption). Moreover, a loss of rice income due to crop losses or undersupplied irrigation may worsen livelihoods. The poorest 40 percent of Cambodians, most of whom live in rural areas and earn income from agriculture, spend around 70 percent of their income on food (CDRI, 2008; Chan, 2011). Quantitative evidence on the links between land property rights, excessive flooding, and irrigation on rice productivity and food insecurity is lacking (see Dininger & Ali, 2008; Maxwell & Wiebe, 1999). None of the existing studies highlighted in Chapter 2 quantitatively examined the direction of causality between agricultural land property rights

and food insecurity. Additionally, there is no quantitative study on the impact of excessive flooding and irrigation on rural food insecurity that uses primary survey data at the plot and household levels. Only a few studies, including Maxwell and Wiebe (1999) and World Bank (2003), assessed the link between land tenure security and food insecurity. However, none of these studies examined these links quantitatively. Therefore, it is important that this thesis conducts quantitative examinations to contribute to evidence-based policymaking in tackling food insecurity in the developing world.

1.2 Scope and research questions This thesis presents three empirical chapters on factors that could influence food insecurity in developing countries. To explore the links between agricultural land property rights and household food insecurity (Chapter 2) and potential impacts of excessive flooding and irrigation on rice productivity and food insecurity (Chapter 3), the thesis uses Cambodia – my home country – as a case study (Chapters 2 and 3). Chapter 2 is the first study that examines primary household survey data on land property rights and rural household food insecurity in Cambodia. Chapter 3 is the first study that uses primary plot- and household- level data. To explore the potential impact of security in private property rights on food insecurity, the thesis conducts cross-country observations among 57 developing economies (Chapter 4) for the period 1990 to 2011. Chapters 2 and 3 use two indicators of household food insecurity. The first indicator captures the reported experience of hunger per surveyed household over the past 12 months (March 2013 to February 2014). It is a binary indicator (1 and 0), which reflects food availability (supply) and food utilisation (consumption of diets and nutrition) or food access. The second indicator measures reported length of household food insecurity. Chapter 4 uses three indicators of food insecurity (prevalence of undernourishment, prevalence of food inadequacy, and depth of food deficit) and one indicator of food security (average dietary energy supply adequacy). Further details on indicators are provided in Section 1.5. Chapters 2 to 4 conduct econometric tests of key empirical questions. Chapter 2 asks the question: do agricultural land property rights affect household food insecurity among rice farmers in rural Cambodia? If so, to what extent? Through what mechanisms do agricultural land property rights have an impact on rural household food insecurity? Chapter 3 asks: do

(i) excessive flooding and (ii) the availability of irrigation affect rice productivity and household food insecurity among rural rice farmers in Cambodia? If so, to what extent? Chapter 4 asks: does security in private property rights matter for food insecurity in developing economies? And to what extent?

1.3 Goals of this thesis The thesis examines whether there is any evidence on impacts of agricultural land property rights, excessive flooding, and irrigation on rice productivity and rural food insecurity in Cambodia and the potential impact of property rights security on food insecurity in developing economies. The purpose is to provide quantitative evidence that is currently lacking in the literature and to contribute to food policymaking that addresses food insecurity in developing countries. The thesis is divided into two analytical parts, and it approaches the analysis as follows. The first part focuses on Cambodia as a case study, using primary household survey data. It first examines potential impacts of agricultural land property rights on food insecurity among rural rice farmers in Cambodia, using survey data at household level. It then examines potential impacts of irrigation and excessive flooding on rice productivity and rice revenues among rural rice farmers in Cambodia, using the survey data at plot and household levels. It tests the household-level survey data to examine potential impacts of rice productivity and rice revenues on household food insecurity. Additionally, it investigates potential channels through which private property rights security, especially in agricultural land, impacts food insecurity, such as (i) credit access, (ii) land-based collateralisation, (iii) revenue-cost ratio, (iv) rice productivity, and (v) rice revenues. The second part presents an analysis of 57 developing countries. It examines whether and to what extent private property rights protection impacts food insecurity in developing economies, using data at the national level.

1.4 Contributions of the thesis One of the main strengths of this thesis is the primary household survey in which I participated to interview rice farmers among 32 villages in eight districts of four provinces in rural Cambodia (Chapters 2 and 3). The survey was conducted between March and May

2014. The other strength comes from examining data for a large cross-section of developing economies (Chapter 4). Chapter 2 makes three contributions. First, it provides new empirical evidence from Cambodia, a post-conflict developing economy. The results demonstrate strong evidence of a negative link between agricultural land property rights and household food insecurity. Second, it confirms that land property rights are significantly, positively associated with credit access, land-based collateralisation, and farms’ rice revenue-cost ratios. Third, rice productivity and rice revenues are each significantly, negatively associated with food insecurity. Chapter 3 makes two contributions. First, it provides new evidence on the links between excessive flooding risk, irrigation, rice productivity, and household food insecurity. Second, it identifies at least two potential channels through which rural food insecurity can be reduced. Investing in flood-coping mechanisms to control disastrous flooding risks or adapting to seasonal, annual flooding and the ecosystems of the rice-cropping floodplains can be an option for improving rice productivity. In addition, investing in irrigation infrastructure is another option for enhancing rice productivity and reducing rural food insecurity in Cambodia. Third, higher rice productivity and rice revenues could potentially lower rural household food insecurity. Chapter 4 makes two contributions. First, it provides new quantitative evidence on links between property rights and food insecurity in a large cross-section of developing countries, which has never previously been done. Second, it examines the effects of different measures of property rights on food insecurity to assess data validity and robustness. The evidence that emerges out of this thesis informs policymaking regarding food insecurity in Cambodia and developing Agricultural policy that incorporates agricultural land rights and mechanisms that ensure full protection of the rights can prove beneficial. economies. Securing land property rights for farmers could help secure access to land, expand credit access, and enable land-based collateralisation. In sum, the quantitative evidence sheds light on local solutions to address food insecurity in developing economies through (i) enhancing private property rights (including agricultural land), (ii) controlling excessive flooding, and (iii) building irrigation networks.

1.5 Food (in)security Food security has been defined to exist “when all people have unobstructed physical, social and economic access to sufficient, safe and nutritious food to meet their dietary needs and food preferences for an active and healthy life” (1996 World Food Summit). Food insecurity occurs when hunger and undernutrition prevail because people have less than the minimum dietary energy requirement for a healthy and nutritious caloric intake. Food security, according to the Food and Agriculture Organization (FAO) of the United Nations, consists of (i) food access or affordability, (ii) food availability, (iii) food utilisation, and (iv) stability of food systems. Food insecurity can easily be confused with food poverty. However, food insecurity encompasses the four dimensions rather than a lack of food as in the case of food poverty. Chapter 4 of this thesis tests the following four food (in)security indicators, compiled and determined by FAO. The measurement methods are explained in FAO’s 2011 Food Balance Sheets handbook and FAO’s 2014 Food Security Indicators. The first indicator of food insecurity is prevalence of undernourishment. This is the share of the population with an insufficient intake of food calories (IFRI, 2016). The second indicator of food insecurity is prevalence of food inadequacy. It measures the percentage of the population unable to cover adequate food requirements to perform normal physical activity, including those who are likely being conditioned in their economic activity by insufficient food, even though they are not chronically undernourished. The third indicator of food insecurity is depth of food deficit. It measures kilocalories per capita per day or how many calories are needed for a person to be no longer undernourished. The depth of food deficit measures the average intensity of food deprivation of the undernourished. The fourth indicator, average dietary energy supply adequacy, is a measure of food security. It expresses the average dietary energy supply as a percentage of the dietary energy requirement. The other measures of food insecurity and of agricultural productivity used in Chapters 2 and 3 come from the Cambodia household survey in which I participated in 2014.

1.6 Land property rights Property rights have been studied by many researchers and pioneers in development and institutional economics, including Demsetz (1967), North (1990), de Soto (2000), and De Janvry et al. (2015). The social, cultural, political, and economic implications of property

rights were discussed in Adam Smith’s (1776) The Wealth of Nations. In Smith’s own terms, “no commerce and manufactures could flourish sufficiently long enough unless their private properties receive full protection”. Property rights cover multiple entitlements or a bundle of land rights which allow land owners or authorised land holders to capitalise on the rights. In broader terms, land property rights include ownership rights, use rights, transfer rights, and mortgage rights. They refer to a set of rights to make change, develop, and exclude others from accessing a property of a holder. An important aspect of property rights to highlight is agricultural land property rights. Complete land property rights comprise (i) transferability, (ii) inheritability, (iii) exclusivity, and (iv) enforcement (Alchian & Demsetz, 1973; Feder & Feeney, 1991; Lawry et al., 2016). These land property rights are collectively a measure of land property rights (in)security. These features are not inseparable, albeit complete, rather than partial, land rights are desirable (Fenske, 2011; Deininger & Ali, 2008). The attenuation of these rights may harm investment in land and productivity of land use and might affect a multitude of socio-economic outcomes of many people. Strengthening insecurity in private property rights requires mechanisms that decode and register property rights and legal institutions that enforce contracts and property rights (de Soto, 2000). As de Soto suggests, clear articulation and configuration of land property rights that fit the social, political, and economic conditions of an economy is essential. However, under secure private land property rights, social gains may prevail and the benefits of enforcing the land property rights should outweigh the costs (La Porta et al., 1999).

1.7 Methods This thesis follows the “research papers” style. Chapters 2 and 3 use primary household-level cross-sectional survey data collected in 2014. Chapter 4 uses pooled cross- sectional and pooled panel data for 57 developing countries for the period 1990 to 2011. Each of the three empirical chapters use econometric methods. The theoretical framework behind the empirical method utilized is the Agricultural Household Model (AHM) (Sadoulet and de Janvry, 1995). The modification of the model is needed to include some relevant assumptions in the context of Cambodia. In the Cambodian context, there have been some observations of imperfect markets for credit (Yagura, 2005), labour and product markets (Han, Fukui & Miwa, 2008),

particularly in developing economies. Imperfection of labour, credit, capital, and product markets in rural Cambodia are observed and realising this potential issue, I attempted to account for variation in cross-sectional data by including other factors that might affect food production and food security as a result. In this connection, I tested three potential impacts of agricultural land rights security, namely credit access, collateralisation, and revenue-cost ratio (known as input efficiency). Most rural farmers have faced difficulties with expansion of land size and finding off-farm work to earn extra income to support welfare (Han et al., 2008), therefore younger adults in rural Cambodia tend to migrate to other regions, including the city and neighboring countries. Two key points differentiate households with more productive assets, such as land, but less family labour from those with less of these assets but more household farm labour, leading to divergence in credit access from land-based collateralisation and access to farm technology. Where capital market is missing or limited in rural villages, households with larger landholding tends to have better access to lower-cost loans (Han et al., 2008), which would encourage their capital investment in farming. Imperfections in labour market and lack of regulatory support for labour and social insurance has appeared to hinder employment creation (Cho et al., 2012), leading farm and nonfarm labour to seek jobs elsewhere. The labour-rich, land-poor households would tend to seek farm and off-farm labour to compensate for little landholding for crop cultivation. While land- rich, labour-poor households would continue operating on their farmland by investing in more productive inputs and hiring extra labour, they could reap more agricultural revenues than their counterparts. Based on these assumptions of the imperfect labour, credit, and capital markets, and in fact they are factual observations, I employed this variation in cross- sectional data to decompose potential impacts of differing insecurity in land property rights on rice and land productivity and food insecurity of rural rice farmers in Cambodia. In sum, my model specification has responded well to the agricultural household model and imperfect market assumptions postulated by Sadoulet and de Janvry (1995) and other researchers. The econometric methods for Chapters 2 to 3 are probit, logit, and Tobit or Poisson regressions, as suited to situations where the dependent variables are binary and count data (Cameron & Trevidi, 2005; Wooldridge, 2010; Greene, 2012; Amemiya, 1973). Chapter 4 uses the ordinary least squares (OLS) method and the between estimator (BE) approach. The estimation techniques for the pooled cross-country data rely on the BE method. The use of the BE method is suitable because of the nature of the cross-country data on private property

rights and food (in)security, where only a little temporal variation within a country over the past years is observed. The property rights and food insecurity within a country do not change rapidly or their changes can be observed over the long run and across countries. Moreover, year-to-year measures of these data might not be well measured. I use the multi-year data (1990‒2011) that allow a country-level panel to exploit between variation. The coefficients for the BE will provide estimates of long-run relationships and interpretation (Burke & Yang; 2016; Stern, 2010; Pesaran & Smith, 1995). The long-run effects captured by the BE method are important for identifying potential impacts of cross-country variations in private property rights on food insecurity. The BE is not affected by time-series issues related to unit roots (Stern, 2010) and is superior to alternative panel estimators when there is measurement error in explanatory-variables (Hauk & Wacziarg, 2009). The thesis conducts robustness checks for the cross-country observations, using different data settings and different measures of property rights and of food insecurity. The thesis approaches the econometric tests of whether agricultural land property rights impact household food insecurity, using primary household survey data. It examines three channels, namely credit access, land-based collateral usage, and revenue-cost ratio, known as input efficiency. In testing whether excessive flooding influences food insecurity, the thesis first estimates (i) the effect of excessive flooding and irrigation on rice productivity and rice revenues, using plot-level and hectare-level household data and then (ii) the effect of rice productivity and rice revenues on household food insecurity, using household-level data.

1.8 Thesis organisation This thesis consists of five chapters. Chapters 2 to 4 present the main research content. Chapter 2 examines potential impacts of agricultural land property rights on food insecurity. It also assesses three potential channels: access to credit, land-based collateralisation, and revenue-cost ratio. Chapter 3 discusses whether excessive flooding and irrigation interventions affect rice productivity and rural household food insecurity. Chapter 4 presents a quantitative analysis using cross-country secondary data on 57 developing countries for the period 1990 to 2011. Chapter 5 concludes with recommendations for future research.

CHAPTER 2

IMPACT OF LAND PROPERTY RIGHTS ON FOOD INSECURITY IN RURAL CAMBODIA

2.1 Introduction

Private property rights in Cambodia have undergone critical transitions under unstable political regimes. They were abruptly abolished by the genocidal Khmer Rouge regime that took power in 1975 (Oldenburg & Neef, 2014). The current property rights are largely undeveloped, fragmented, and fragile. An estimated 80 percent of agricultural land held by rural households was not titled by 2002 (Sik, 2000). By 2015, this number had reduced to about 70 percent (USAID, 2016). Having no land title presents a high risk of land dispossession or expropriation by the state or private elites. However, holding registered or titled farmland in rural Cambodia may still face risk as well (Sekiguchi & Hatsukano, 2013). The attenuation of agricultural land property rights means that farmers stand the risk of losing their cropland or losing rightful access to engage in agricultural production when their cropland is withheld in a land dispute. The risk of losing cropland among rural rice farmers in Cambodia is not trivial. But what does insecurity in land property rights entail? Agricultural land is a productive input in an economy (Feder, 1987). For farmers, large landholding embodies possession of a large asset. Farmers often seek a land title to prevent expropriation if expected long-run benefits of securing the land are greater than the costs (Jacoby & Minten, 2007). Insecurity of agricultural land rights may threaten the agrarian sectors, including rural food security and the livelihoods of farmers. Existing evidence appears to suggest that amount and quality of land which rural households have access to and control over could determine their food insecurity (Ballad in Torhonen & Groppo, 2006). Despite insecurity in their agricultural land, farmers continue cultivating plots and making investments to maintain or improve crop productivity (Hayes et al., 1997) because they have no choice.

Between 70 and 80 percent of rural Cambodians engage in agriculture, mostly in rice sector, for food and income (USAID, 2016; Oldenburg & Neef, 2014). Rice consumption constitutes two-thirds of total caloric intake in Cambodia (Maltsoglou et al., 2010), and agriculture contributes to 33 percent of the gross domestic product (GDP). By 2015, about 25 to 28 percent of rural Cambodians, roughly 3.7 to 4 million, were food insecure for about two to three months (USAID, 2016). Some households have experienced more prolonged food insecurity. Facing food insecurity means having less than the minimum required daily dietary caloric intake (USAID, 2016) to consume to remain healthy. This chapter examines whether and to what extent agricultural land property rights affect household food insecurity of rural rice farmers in Cambodia. The food insecurity variable as an outcome is binary. It is equal to 1 if a household reports experiencing food insecurity during the past 52 weeks (April 2013 to March 2014) and 0 otherwise. The main variable of interest is agricultural land property rights. The chapter found that security in agricultural land property rights is significantly, negatively associated with household food insecurity. This chapter also tests self-reported land documents as an alternative measure of land property rights. Land documents are not significantly linked to household food insecurity. Additionally, the chapter examines the potential impact of land property rights on the duration of food insecurity, measured in weeks per year. The chapter then tests the link between agricultural land property rights and three potential channels, namely (1) collateral, (2) credit access, and (3) revenue-cost ratio.1 The chapter proceeds as follows. Section 2.2 provides a literature review. Section 2.3 discusses land property rights in Cambodia. Section 2.4 describes the basic model. Section 2.5 explains the data collection. Section 2.6 discusses the estimation strategy. Section 2.7 explains the key findings. Section 2.8 discusses channels. Section 2.9 concludes with policy recommendations and a discussion of future research directions.

2.2 Literature review Land property rights have been discussed by many scholars, including Hernando de Soto (2000), Fenske (2010, 2011), and Newman et al. (2015), as having social and economic impacts on income and poverty (Besley & Burgress, 2000), agricultural investment

1 Revenue-cost ratio is also known as input efficiency. 12

(Markussen, 2008), and food production (Besley, 2005).2 Only a few studies, including Maxwell and Wiebe (1999) and World Bank (2003), Newman et al. (2015), Lawry et al. (2016), and Golay and Biglino (2013) have assessed the link between land tenure security and food insecurity. However, to the best of my knowledge, none of these existing studies have analysed the empirical link quantitatively. Land property rights cover land access, which includes the rights to use, transfer, and mortgage. Complete land property rights comprise exclusivity, inheritability, transferability, and enforcement mechanisms (Alchian & Demsetz, 1973; Feder & Feeney, 1991; Lawry et al., 2016; Besley & Ghatak, 2009). Where weak or unprotected private property rights prevail, owners or holders of land are at risk of losing that land due to land grabbing or land dispossession by the government or power private individuals. Therefore, in the presence of weak or insecure property rights, rural farmers tend to engage in land disputes over overlapping claims. This insecurity in land property rights is common in many of the developing countries (de Soto, 2000). A study on a land certification program in Vietnam only examined the land program impact on consumption levels for rural households (Kemper et al., 2013). It has been observed that securing land tenure may better protect tropical forest outcomes, which affects food sources (Naughton-Treves & Wendland, 2014).3

Investment impact of land property rights: Existing evidence tends to report that security in agricultural land property rights has encouraged investment (Markussen, 2008) in inputs or land improvement (Deininger & Jin, 2006; Hayes et al., 1997; Newman et al., 2015) and in innovation in agricultural production (Besley, 1995), production behaviours of farmers (Li, Rozelle, & Brandt, 1998), and farmers’ decision to invest in land improvement and conservation measures (Abdulai et al., 2011). It seems to suggest that investment incentives could depend on expectations of land tenure rights over returns to an investment and nature of the land property rights (Goldtein & Udry, 2008). Increased tenure security was found to have led to more investment in northeast China (Jacoby, Li, & Rozelle, 2002). A similar evidence shows that land titling positively affected agricultural investment and capital intensity in two of three provinces in Thailand (Feder & Onchan, 1987), and crop

2 Other leading scholars who have studied these links include Field (2005), Field and Torero (2006), Goldstein and Udry (2008), Place (2009), Galiani and Schargrodsky (2010), among many others. 3 In their explanation, incentive-based strategies, for example payments for ecosystem services or Reducing Emissions from Deforestation and Degradation (REDD+) help protect tropical forest outcomes. 13

diversification in Vietnam (Do & Iyer, 2008). In Ghana, insecure land tenure reduced investment in land fertility (Goldstein & Udry, 2008).4 Land tenure differences could affect farmers' decisions to invest in land-improving and conservation measures, and land tenure security could affect farm productivity (Abdulai et al., 2011). However, impact of land property rights on investment remains ambiguous. The impact appears to depend on land tenure regime or land property rights practices (Place & Otsuka, 2002). While there is evidence that informal property rights positively impacted investment in the Wassa region of Ghana, there was no such evidence for the Anloga region (Besley, 1995). In Hayes et al. (1997), customary land tenure security encourages farmers to make investments in input use. However, such evidence does not seem to exist in Burkina Faso (Brasselle et al., 2002) or in Sub-Saharan Africa (Place & Hazell, 1993).

Productivity impact of land property rights: Some existing evidence shows that property rights security increases productivity and land values (Besley, 1995; Markussen, 2008; Babu et al., 2014). It was found to have positively affected agricultural productivity in peri-urban areas of the Gambia (Hayes et al., 1997) and a southern province of Zambia (Smith, 2004). Another piece of evidence suggests that land tenure security releases resources spent on conflicts to fund agricultural investment (Deininger & Castagnini, 2006). However, there appears that overall evidence remains mixed. While land registration positively affected yields and productivity in rural in Kenya (Place & Migot-Adholla, 1998), there was no such evidence in Uganda (Deininger & Ali, 2008). Place and Hazell (1993) and Migot-Adholla et al. (1991) found that indigenous land rights did not influence land productivity in Sub- Saharan Africa. Gavian and Ehui (1999) also found no differences in total factor productivity impacts of land tenure.

Credit access impact of land property rights: Existing evidence seems to have shown that secure land property rights facilitate credit access to finance investment (Brasselle et al., 2002) or turn unused assets into capital (Piza et al., 2016; Fenske, 2011; Macours et al., 2010; Besley, 1995; de Soto, 2000). Weak property rights, in contrast, hinder access to investment capital (Atwood, 1990; Markussen, 2008; Place & Hazell, 1993). However, there is still

4 Complete land property rights can boost investment (Carter & Olinto, 2003; Fenske, 2011), and government- issued land documents also increased investment in agriculture (Markussen, 2008). 14

mixed evidence on credit impact of property rights. While there is evidence that land titling stimulated collateral-backed credit use, investment, and land values in Indonesia (SMERU, 2002), indigenous land rights were not found to influence credit access in Sub-Saharan Africa (Place & Hazell, 1993). Moreover, land tenure was found to have no or only very weak impact on credit access in other studies (Feder & Nishio, 1999; Deininger & Feder, 2009; Lawry et al., 2016; Markussen, 2008). Lastly, weak or non-existent credit markets did not improve capital access in Thailand (Feder & Onchan, 1987) and in Africa (Gavian & Fafchamps, 1996), despite improved land rights. In such cases, titled and untitled plots make no difference in access to credit (Jacoby & Minten, 2007; Lawry et al., 2016).

2.3 Land property rights in Cambodia Post-conflict Cambodia faces challenges of land property rights insecurity. The abrupt abolishment of private land rights during the Khmer Rouge regime (De Lopez, 2002; Rudi et al., 2014) was one of the most tragic changes in governance in human history. After the regime collapsed in 1979, a new socialist state began a collective land property rights system. Private land property rights were reintroduced by the state in 1989 (So et al., 2001), and were officially reinstated later in the 1992 Land Law. In early 2000, systematic land registration was planned to resume. By 2002, only 10 percent of rural households possessed land titles (World Bank, 2002). Currently, about 20 to 30 percent of agricultural plots have received title (USAID, 2016). Agricultural landlessness resulting mostly from land grabbing is rising.5 The number of agricultural landless and near-landless households in Cambodia increases by about 2 percent annually (Torhonen & Groppo, 2006). In 2011, the agricultural landlessness rate among rural households was 29 percent. The World Bank (2002) reported that rate of landlessness, both agricultural and residential, could have been as high as 20 percent of rural Cambodians by 2002. The total number of landless households in Cambodia increased from 16.3 percent in 2001 to 23.2 percent in 2014 (Phann et al., 2015). Several factors have contributed to weak land property rights in Cambodia. For example, the adoption of an open economic system by the mid-1990s has seen a jump in land prices and eruption of land disputes and landlessness. Complexity in administering land property rights has resulted from the existence of modern land rights alongside customary

5 “Agricultural landlessness” means lacking agricultural land while holding residential land. 15

land rights. Customary practices remain prevalent in rural Cambodia (Sekiguchi & Hatsukano, 2013), and has created controversies over land claims. The prevailing 2001 Land Law aimed to end the customary land rights regime to unify and consolidate land property rights. Under customary law, farmers can obtain a property rights over the land if they cultivate the land for three consecutive years (Sekiguchi & Hatsukano, 2013).6 In contrast, the 2001 Land Law recognises legal ownership even if the property rights holders do not cultivate the land. Armed conflict, which continued until 1998, weakened the state capacities to regulate land use and settle land disputes (Cooper, 2002). Land registration processes often involve unofficial charges (Markussen, 2008), restricting access to land titling services. Farmers sometimes pay between US$300 to $400, more than the official fee of US$3 to $4 (So et al., 2001). Furthermore, sporadic and tedious land titling processes might have threatened private land rights. In 2003, the Cambodian government began a new comprehensive land administration program (Markussen, 2008), aiming to issue one million titles between 2003 and 2007 (World Bank, 2002).7 There are over 6 million rice plots in Cambodia, but only 38,481 had received titles by the end of 2004 (Deutsch, 2006). The number of registered plots has increased a few years later. By end of 2009, over 1.7 million plots or approximately 20 to 30 percent of the total plots were registered and about 1.25 million land titles had been distributed (GTZ, 2009). The economic land concession (ELC), granted by the government to private investors for use in industrial and agricultural development projects, could have contributed to rising land disputes. By end of 2003, 2.7 million hectares of land were given as ELCs (Leuprecht, 2004). Forbes reported that in 2014 land acquisitions under ELC schemes had affected over 770,000 Cambodians, who ended up in land disputes and with land losses. The absence of map and satellite use in boundary demarcations when ELCs are granted is one of the causes of land disputes over overlapping claims (Neef et al., 2013). Violent dispossessions have affected communal land titling efforts in eastern provinces of Cambodia (Milne, 2013). By

6 The 2001 Land Law states that any regime of ownership of immovable property prior to 1979 (before the collapse of the Khmer rouge regime) is not recognised. It only recognises possession of immovable property from 1989. Under the Land Law, any land occupation for less than five years before August 2001 cannot receive formal land title. The law also considers any land occupation after August 2001 without a title illegal. 7 Cambodia has benefited from World Bank land projects introduced in 2004 (Conning & Deb, 2007). However, in recent years the land titling projects have seen little progress. The World Bank stopped funding the land-titling projects in 2009 due to frustration over irregularities in project implementations, and the German government ended its land rights project in early 2016 (Cambodia Daily, 2016). 16

2007, indigenous minorities had lost 30 percent of their traditional forestland and by 2012 over 53 percent of Cambodia’s arable land had been given as ELCs (Neef et al., 29013).

2.4 Basic model

This chapter tests the following specifications:

퐹퐼ℎ } = α +  . LPRℎ + Ω. 퐗풉 + ℰ풉 (2.1) 퐿퐹퐼ℎ where:

- 퐹퐼ℎ is food insecurity. 퐹퐼ℎ is binary, which takes the value of 1 when a household reported food insecurity over the past 52 weeks, and 0 otherwise. This indicator reflects food availability (supply) and food utilisation (nutrition) or food access.

- 퐿퐹퐼ℎ is length of food insecurity, expressed as number of weeks of reported food insecurity. - The term food insecurity was briefly explained to farmers when asking them about their rice production, food consumption, food availability and sufficiency, and experience and duration of not having food to eat. The interviewed farmers appeared to understand the questions and the purpose of the interview well.

- LPRℎ measures the degree of perceived security in agricultural land property rights reported by household head for their rice plot(s). The household land property rights are derived from summing up the per-plot scores. Total cultivated land size per household is used to weight it.

- 퐗풉 is a set of controls, including HH size (SH), cultivated land size (CLS), HH head’s years of rice-growing experience (YRCE), annual crop frequency (CF), short-run shocks (SRS), rice quality (RQ), soil problems (SP), land documents as ownership proof (LD), village-member trust (VMT), government trust (GT), village-town closest distance (VTD), and annual average precipitation rate (AAPR). - 훼 is a constant. Ω is a vector of parameters, and  is the parameter of interest to

estimate. ℰℎ is an error term.

Land documents that households reported as evidence or proof of land ownership are also included in a separate estimation. Rice quality is controlled for to capture the type of seeds or the variety of rice that are suitable for the soil quality, seasonality, and availability of irrigation which rice growth depends.8 Additionally, I intended to take into account of remittances and off-farm income in the specifications. However, the data for these two variables are not available as there are a lot of missing values in the data, which become unusable in the regressions. The chapter also tests the following specification to assess potential impacts of land property rights on three potential channels: productivity, collateral, and credit access.

퐶퐻ℎ = α +  . LPRℎ + Ω. 퐗풉 + ℰℎ (2.2)

where CHh stands for channels or transmission mechanisms. CHh includes collateral (COL), credit access (CA), and the revenue-cost ratio (RCR). Credit access, collateral, and revenue-cost ratio can be possible channels to food insecurity for the following reasons. First, when land property rights are well recognized and protected in rural credit markets to obtain loans to fund agricultural production and expansion of agribusiness or to engage in land instance. Second, collateralisability or collateralisation could be closely related with augmented access to credit, mostly ‘good’ credit, which in general carry cheaper interest rates from credit suppliers. Third, improved revenue cost ratio could mean higher revenue from rice production, given unchanged cost of production. Revenue-cost ratio is known as ‘input efficiency’ in agricultural production. Revenue-cost ratio is ratio of rice revenues to cost of rice production, where rice revenue is total rice harvest times household average price of rice derived from diving sum of prices of rice by number of crops per household. Cost of rice production excludes other costs, such as chemical input, irrigation cost, and depreciation of farm equipment. I calculated the household revenue-cost ratio for each household for household-level data analysis. Improved input efficiency may mean an improvement in allocative efficiency in input use to reap the same harvest or higher input use has grown less than gains from harvest. However, as revenue-cost ratio is more

8 Rice quality is proxied by the price obtained for the cultivated rice of each household. The Cambodian farmers normally grow different varieties of rice that suit their land and soil characteristics and availability of water for irrigation. 18

commonly used, a similar, but related, interpretation can be useful. Higher income from production that result from higher yield or higher prices would allow farmers to acquire additional food other than rice production from their family farming. These potential gains can be attributed to security in agricultural land property rights which, as the evidence has shown, tend to positively affect farmers’ access to credit and ability to collateralise their cropland and improved revenue-cost ratio. Therefore, as agricultural land property rights would lower food insecurity, they could possibly do so through improved credit access, land- based collateral use, or higher revenue-cost ratio.

2.5 Data collection

The data used in this chapter come from a household survey conducted between March and May 2014, after the wet-season rice harvest, administered to 256 rice-growing households in 32 rural villages across four provinces. The survey was funded by the Economic Research Institute for ASEAN and East Asia (ERIA), with support from Cambodia’s Council for Agriculture and Rural Development. It was part of an ERIA research project on natural disasters in agriculture and household risk behaviours among ASEAN countries. The data collection received help and support from local authorities, including village chiefs and commune councils. In about one-third of the survey fields, the village chiefs were present during the interviews. However, the survey team attempted to make sure that no pressure from the local authority influenced the responses from households. The sampling strategy involved stratifying the survey sites to ensure a degree of homogeneity in average household characteristics in each village and variation in village characteristics in each surveyed commune. The survey team also aimed to ensure heterogeneity in social and economic aspects. The four surveyed provinces are Kampong Thom in the central region, Banteay Meanchey and Battambang in the northwestern region, and Prey Veng in the southern region. The pink shaded area in Figure 2.1 represents areas which were flooded in either 2011 or 2013 or in both years. I do not have access to ArcMap services to have a location map created. The current map of Cambodia shows the sites of the interviews.

Figure 2.1 Map of the survey fields

In selecting households, the survey team chose 256 rice farmer households randomly in the 32 villages (marked by red boxes in Figure 2.1), namely eight households were selected from each village and two villages from each of the 16 communes. Two districts were chosen from each of the four provinces. Table 2.1 lists the names of villages, communes, districts, and provinces. Note that the star signs represent the locations of the household survey. 16 of the 32 surveyed villages are in lowland floodplains, and they have been affected by annual flooding at least once over the past 10 years. The other 16 villages are close to these flood-afflicted villages, but they have never affected by severe annual flooding. However, some of the rice plots of the households in these villages could have been affected to some extent by the flooding that occurred in other villages because their rice plots are located near to or in the flood areas of other villages that were flooded. Likewise, not all rice plots of the households in the flooded villages got flooded because those rice plots are located elsewhere. The survey data show that 286 plots of the 251 households were not flooded and the other 284 plots were flooded during 2013–2014. Of note, although 256 households were interviewed, only the data of 251 households are used. The 5 households were excluded from the analysis because there were missing data on many variables for the 5 households. The missing data appear to have happened during data collection. 20

Table 2.1 List of the survey fields Province District Commune Village Kampong Thom Krong Steung Saen Srayov Bramatdei Srayov Thboung Aur Kanthor Prek Sbov Aur Kanthor Thbong Kampong Svay Sankor Kra Sang Balang Kdei Doung Peam Kreng Kdei Doung Banteay Preach Netr Preah Preach Netr Preah Sresh Lech Meanchey Tapen Chab Vari Kouk Lorn Brasat Mongkul Borey Banteay Neang Kouk Kduoch Kouk Trolerb Russei Kroak Aur Takol Chamkeav Battambong Thmor Koul Boeng Pring Snoul Koang Boeng Pring Tapung Ta Pung Kouk Kdouch Moung Russei Robos Mongkol Prey Prom I Prey Prom II Prey Touch Prey Touch Kon Khlong Prey Veng Preah Sdach Rom Jek Chongros Tropoeng Chhuk Bateay Chakrey Brobos Rolauy Rorka Jour II Kampong Trabek Peam Montea Takeo Krocham Luer Cham Cham Sdach 4 provinces 8 districts 16 communes 32 villages Notes: Kampong Thom and Battambang provinces are in Tonle Sap floodplains that receive floodwater from the Mekong River during June‒September. Banteay Meanchey is in the northwest part of Cambodia and it mostly receives water from overflows of the Tonle Sap, major upstream catchment areas, and natural river streams. Prey Veng Province is in the southern part of Cambodia. It receives floodwater from the Mekong River, natural rivers, and river streams.

The questionnaire for the household survey is provided in Appendix 1. The five main questions asked are as follows. 1) Land property rights security question:

‘Please assign the coins to the following land (in)security events based on your opinion about the likelihood that they will occur to your plots in the next 10 years’. ‘What is the likelihood that this plot j will not be taken away, or lost to others?’

The household head was asked to assign i coins (i = 0‒10) on plot j (j = 0‒6) for perceived land property rights security over the next 10 years. Plot j is the plots owned and operated by households. The number of coins assigned to each of the plots suggests the number of coins assigned to each of the plots suggests as expected probability of land property rights security. If no coin was placed on plot j, it suggests the lowest land property rights security or highest risk of land loss for that plot j. In contrast, if 10 coins were placed, it suggests the highest security or lowest risk. The score of land property rights for each plot is weighted by its plot size. The sum of the scores for land property rights per household is the relative measure of land property rights.9 This perception-based question on land documents could be correlated with level of land property rights security of each of their plots.

2) Land documents question: ‘Could you tell about all agricultural land owned or operated by your household? If you own plot j, (j= 1‒6), do you have and what type of paper or documents to certify your ownership or land use rights?’

(0 = Do not have; 1 = Land investigation paper; 2 = Certificate (title) from government; 3 = Paper from local authority; 4 = Application receipt; 5 = Other (specify); 6 = Don't know/not sure)

There are six responses. The household head can choose any of the six responses for each of their plots. Responses #0 and #6 will turn into 0. The rest, namely that a household holds or possesses a land document or land paper to prove or claim land rights over that plot, will be given the value of 1. I then weighted it by the

9 The household head made a judgement based on their own experience of either having lost some land or having some of their land formerly or currently in dispute. The fear of land loss comes from the fact that their farmland was unregistered or untitled at the time of the survey. Hence, the estimated score of land property rights is the perceived likelihood or perceived risk of land property rights security. 22

plot size. The sum of estimates for all plots provides the household estimated score. This measure differs from the perception-based land property rights question in that it does not capture the perceived land loss risk or land rights security.10 However, the perception of land property rights is determined by not only whether households hold a general document of land property rights but also by whether the land has been registered and granted a hard title. In fact, there is difference between holding a land document, in which case if it is not a legal hard land title, and farmers’ perceived security in land property rights. The four different measures of land rights being asked in the land documents question, include (i) land title, (ii) land registration paper from local authority, (iii) land investigation paper, and (iv) land application receipt. Only (i) is the formal land title recognised as an established form of legal land ownership. The rest are either land use certificate or a land measurement or registration receipt issued by local authorities before (i) is granted. In practice, most private banks and financial institutions would accept only (i), otherwise they would charger higher interest rates and with more conditions attached.

3) Food insecurity questions (1) ‘In the past 12 months, were there any days and weeks that your household had very little, not enough or no food (“was hungry”)?’ (2) ‘If YES, how many weeks of the past 12 months did the household have so little food or no food at all that the household was hungry?’11

4) Collateral question: ‘Can you use this plot as collateral for a loan?’ The response is binary (1 = YES; 0 = NO). I then calculated the measure of collateral score, weighted by the total land size owned by each household.

5) Credit or loan access question:

10 It only reports what land documents each household holds, and it tends to contain measurement error or error by household head in distinguishing an official land title from a land investigation paper, for instance. 11 Number of weeks of food insecurity is denoted in this thesis as length or duration of experience of food insecurity. 23

‘Would you like to borrow more from banking/credit institutions? If YES, can you borrow?’ The response is binary (1 = YES; 0 = NO).

Before the actual survey, the questionnaire was pretested, followed by minor modifications to improve the quality of the questionnaire and data collection. The questionnaire modules incorporate aspects of food insecurity and of land property rights, in addition to other modules, such as rice production, household socioeconomic characteristics, credit access, trust, food insecurity experience, natural disaster risks, and many more. The questionnaires cover the history of land rights security issues and land documents. However, the main measure of land property rights is based on farmers’ subjective perception of land property rights security (Question 1), not the reported land documents (Question 2). The sampling strategy was designed to reflect the local features of rural Cambodia, where rice fields can be prone to or affected by either flooding or droughts. The purpose is to ensure that not all households selected for the sample would fall into floodplains only or drought areas only. However, some of the rice-cropping might have been affected by both weather shocks. Additionally, the selection of households to measure land property rights was not based on an a priori criterion that households had been affected by past land insecurity risks. While the selection of households into the category of flooding is not purely random, the selection of households within and across villages into the survey sample is random. The 256 households were selected without reference to household land property types and history of household characteristics. The collected data appear to display sufficient variation in land property rights and landholding sizes, between villages and regions. The little variation in private agricultural land property rights is mostly observed within the same village. In contrast, more variation of those land property rights is mostly observed across villages. It is not surprising that the mean scores of property rights across provinces vary by very small margins. The descriptive characteristics of farmers’ agricultural land property rights across villages and provinces are given in the Appendix 3. Each of the survey households hold more than one plots or parcels of land. On average, a household holds between two to three plots, up to five or six plots. Although in few of villages, the agricultural land property rights of rural rice farmers appear to show little variation within the same villages, the security in land property rights is widely dispersed

between the 32 villages and varies across the districts and provinces. That could be because the process of land registration and titling is relatively homogenous within the same village and because of their diverse development potentials or political interests of the state. The food insecurity variable contains two separate indicators. FI is a binary variable of 0 and 1 and LFI is a non-negative response, measured in number of weeks of reported food insecurity over the past 52 weeks. The measured score of land property rights per plot ranges between 0 and 10, so does the overall land property rights score per household (0 = highest land property rights security or lowest land property rights insecurity or no risk of land loss; 10 = lowest land property rights security or highest land property rights insecurity or highest risk of land loss). Breaking down the analysis by income/expenditure quantiles can be worthwhile. However, there are important challenges in decomposing and calculating the food consumption (or expenditure) into quantiles, particularly conversion of the different units of measurements, such as kilograms, bottles, cups, cans, packs, etc., liters, and monetary values, into food caloric intakes. Any measurement error resulting from decomposing these different units into caloric values that correspond to human physical and physiological needs can happen. FAO, among others, have recognised this challenge (FAO, 2012 & 2014). Therefore, I chose to use subjective indicator of food insecurity to test the data. While income can be a predictor for food insecurity, existing evidence has highlighted some potential for reverse causality between income and welfare indicators, for example health (Hoffman, Kröger & Geyer, 2018; Ettner, 1996), education expenditure (Sylwester, 2000), and food security (Hoden & Ghebru, 2016). Given this, controlling income may cause further complexity to address causality effects between income and food insecurity. Therefore, I did not control income in the model. I would have controlled for nonagricultural income; however, the data collected could not be used as there are a lot of missing values. There is possibility that land property rights could be endogenous by some factors already explained in the section on empirical strategy (section 2.6). If any, wealth status appears to be correlated with farmers’ ability to get land registration. On reverse causality between credit access, collateral, and revenue-cost ratio, higher crop revenues may encourage farmers to secure their farmland land because of the many positive benefits farmers have acquired from the land. Relatedly, agricultural land with higher soil quality would provide higher yields, and as a result yielding higher revenue, and farmers would tend to find ways

to secure these plots. It is less likely, however, that the direction drives from credit access to land rights security because to obtain a loan, it requires that farmers have a legally accepted form of guarantee, which is mostly land-based collateral, where land title is the most demanded in both markets. A reverse causality between land tenure security and food security could potentially exist, and if any it could be attributed to land rights restrictions and obligations (Hoden & Ghebru, 2016). However, all these concerns cannot be addressed in one single study, whose focus is on food security outcome. It would be best to explore these potential links and address these concerns in future research. Table 2.2 provides summary statistics of the variables, measured at the household level. The average score of land property rights among the surveyed households is 8.22 out of 10. It suggests that more than half of the surveyed households perceived that their plots are secure against expropriation risks over the next 10 years.

Table 2.2 Descriptive statistics at household level

Variable Mean Standard Minimum Maximum Deviation Food insecurity (FI) (binary 1 or 0) 0.25 0.43 0 1 Length of food insecurity (LFI) 1.04 3.25 0 26 (in number of weeks) Land property rights (LPR) 8.22 3.32 0 10 (probability rank: 0 = lowest, 10 = highest) Land documents (LD) 0.77 0.40 0 1 Collateral (COL) 0.72 0.43 0 1 Credit or loan access (CA) 0.61 0.49 0 1 Revenue-cost ratio (RCR) 5.21 7.97 0.008 63.88 (in million Cambodian riel) Short-run shocks (SRS) 0.71 0.45 0 1 Rice quality (RQ) (in thousand) 0.80 0.13 0.5 1.2 Cultivated land size (CLS) (in hectares) 2.76 2.95 0.1 20 Soil problem (SP) (binary 1 or 0) 0.61 0.46 0 1 HH head’s years of rice-cultivating 26.95 11.94 1 61 experience (YRCE) Annual crop frequency (CF) 1.34 0.48 1 3 Size of household (SH) 5.18 1.99 2 16 Village-member trust (VMT) 0.87 0.34 0 1 Government trust (GT) 0.40 0.49 0 1 Village-town closest distance (VTD) 11.10 6.02 1.97 24.21 (in kilometres) Annual average precipitation rate (AAPR) 54.72 0.56 54.1 55.6 (in inches) Notes: The number of households in the sample, used in the estimations, is 251. Source: Author’s calculations

Land property rights are correlated with collateral and access to loan or credit at 1 and 5 percent significance levels, respectively (Table 2.3). In terms of land property rights, 24 households, or 9.55 percent of total households, believed that all their plots were insecure. About 77 percent of households hold at least a land document or land record, which can be a land registration paper, land transfer certificate, land investigation record, or land title. 47 households did not hold or possess a land document of any kind. 182 households, or 72 percent of total households, reported that they could use some of their plots as collateral. The other 69 households, or 27 percent, were not able to collateralise any parcel of their land. 152 households, or 61 percent, had access to loans or financial services during 2013‒2014. The remaining 99 households, or about 39 percent, could not borrow from credit markets.13 Agricultural land property rights in rural Cambodia can vary across districts or regions and among villagers. For example, farmers who have closer relationships or stay in closer proximity to the local government, they might have been able to have their plots registered or titled more quickly and at lower cost. This would allow them to secure their rice plots better than those who live far or do not have a close relationship with the local government. Similarly, wealthier households or those which are willing to pay extra money for land registration could have their land certificate issued more quickly. Having this privilege may explain why those farmers appear to be able to claim legal ownership rights to land better than those who do not have extra money to pay the fee for a quicker land registration. It can appear that households can take actions to improve land rights. However, official land rights status could be determined by holding an official “hard” land title, which is issued under a land titling process undertaken by a local government and administered by the central government. According to the prevailing Land Law of 2001, farmers who hold this kind of land documents can only use or make minor changes, but they do not have complete land rights to ownership and transfer. Regional characteristics, such as prosperity of the region, i.e., wealthier regions or regions with more development potential, might have better land property rights. Although variation in agricultural land property rights can be

13 The household survey data I collected show that only six percent of the survey households had invested in land improvements, such as tree planting or dike construction on the plots. 27

more noticeable across villages or regions and less so within the same village, the province- average security in land property rights is not much different across the four provinces. One may expect that rice farmers that could have stronger agricultural land property rights than others would possibly cultivate their plots without interruption or would not be involved in a land dispute. Consequently, households with greater security in agricultural land property rights could possibly be able to produce more food and income from rice production.

Table 2.3 Correlation of variables

FI LFI LPR LD COL CA RCR SRS PRi CLS SP YRCE CF SH VMT GT VTD AAPR

FI 1 LFI 0.505*** 1 LPR -0.128* -0.0387 1 LD - 0.0323 - 0.0487 0.172** 1 COL - 0.0496 - 0.0109 0.304*** 0.528*** 1 CA 0.0892 - 0.0520 0.168** 0.00101 0.0576 1 RCR - 0.151* - 0.125* 0.124* 0.0847 0.0657 0.123 1 SRS - 0.0989 - 0.0594 - 0.0472 0.0322 - 0.0106 0.0289 0.0157 1

PRi 0.0369 - 0.0586 0.0236 0.148* 0.0003 0.105 -0.0112 -0.0366 1 CLS - 0.218* ** - 0.145* 0.00615 - 0.0893 - 0.122 0.0785 0.181** 0.0645 -0.106 1

2 -0.0223 0.0535 -0.106 0.0712 -0.0611 0.0370 -0.0467 0.0726 -0.0899 -0.0715 1 9 SP

YRCE - 0.0471 0.0608 - 0.0764 0.0761 0.00105 -0.119 0.00165 0.0943 0.0372 0.00871 0.0414 1 CF - 0.0476 - 0.0745 0.127* 0.0702 0.128* 0.0828 0.0751 -0.104 0.100 0.0184 0.0247 -0.087 1 SH -0.0837 0.0149 0.0189 -0.0315 -0.0505 0.0771 0.0731 -0.0137 -0.0750 0.141* -0.0266 0.087 0.0177 1 VMT -0.0860 0.0453 0.215*** -0.0290 0.157* 0.0479 0.0677 0.0400 -0.0881 0.134* 0.0119 0.103 0.164** 0.0530 1 GT -0.117 0.0516 0.0713 -0.0818 -0.0916 -0.0360 -0.0505 0. 125* -0.147* -0.0100 0.0836 0.042 -0.0348 0.0282 0.199** 1 VTD -0.080 -0.097 -0.242*** -0.142* -0.052 -0.031 0.079 0.045 -0.140* 0.180** 0.00426 0.076 -0.0605 0.139* 0.0151 0.036 1 AAPR -0.098 -0.048 0.026 -0.097 -0.020 0.055 -0.078 0.101 -0.164** 0.261*** 0.0183 -0.010 0.0892 0.0952 0.174** 0.006 0.004 1

Notes: *, **, *** indicates significance level at 10 percent, 5 percent, and 1 percent, respectively.

Table 2.4 Basic data on food insecurity, land property rights, and credit access (1) (2) (3) (4)

Variable Description # of HHs (of 251) Percent (of 251)

Food insecurity (FI) Binary (1 = food insecurity; 0 65 HHs = 1 25.89 = no food insecurity) 186 HHs = 0 74.1 Mean = 0.24 Length of food insecurity # of weeks ranging from 1 to 32 HHs = 1 week 12.75 (LFI) 26 weeks 17 HHs = 2‒4 weeks 19.52 Mean = 1.04 weeks 8 HHs = 5‒8 weeks 3.19 8 HHs = 8‒26 weeks 3.19 Land property rights Score ranges from 0 (highly 24 HHs = 0 9.56 (LPR) insecure) to 10 (highly secure) 227 HHs > 0 90.44 Mean = 8.22 141 HHs ≤ 1.68 56 Collateral (COL) Binary (1 = YES; 0 = NO) 182 HHs = 1 72.50 Mean = 0.6 69 HHs = 0 27.50 Credit or loan access (CA) Binary (1 = YES; 0 = NO) 152 HHs = 1 60.56 Mean = 0.61 99 HHs = 0 39.44

Source: Author’s calculations from the household survey data. The data show number of HHs and percentage of the total 251 HHs in the sample.

Table 2.4 presents the basic statistics used in estimating the three specifications. Column (1) shows the names of the variables and column (2) describes the variables. Columns (3) provide the number of households that reported facing food insecurity and how many weeks they had faced it over the past 12 months. Column (4) shows the percentage of households that were facing the corresponding number of weeks of food insecurity. As the table shows, 65 households, or about 25 percent, reported suffering food insecurity. The overall average duration of food insecurity among the 65 households which experienced food insecurity was for 1.04 weeks during the year (Table 2.2). 32 households, representing 12.75 percent of the sample, had faced food insecurity for one week in the year. 49 households or 19.52 percent of total households faced food insecurity for between one to four weeks of the year. Among the 251 surveyed households, 16 of them or 6.37 percent faced prolonged food insecurity over 4 to 26 weeks over the past 12 months.

Other controls are explained as follows. Total cultivated land size (CLS) measures actual cultivation, including plots which households leased. The interviews with household heads revealed that some households had cultivated less than the land they ‘own’. Some households reported cultivating more than their land, due to either additional leased-in land or cleared land or both. However, the survey data show that the amount of leased-in land was minimal. Village member trust (VMT) is binary, and it measures social trust among households within the same village or community. Government trust (GT) measures trust in public institutions and their services. Revenue-cost ratio is the total amount of rice harvests times the average price of rice (Cambodian riels) to total cost of rice production, excluding irrigation costs and cost of chemical inputs in rice production. Annual crop frequency (CF) measures the frequency of crops each household cultivated per year. For example, household j cultivated two crops, i.e., twice per year on the same plot i, so the number of crops is 2. The soil problem (SP) variable is controlled for its possible association with the type and quantity of fertilisers used in rice cultivation. For example, poor quality soil or soil that has fertility problems, such as salinity, requires more use of fertiliser to maintain yields. Otherwise, it needs more irrigation and a higher amount of investment in high-yielding seeds. The short- run shocks variable measures shocks that households faced during 2013‒2014, including floods and drought. Other agricultural shocks include pests, insects, and livestock mortality. Non-agricultural shocks include such shocks as death of family members and loss of assets. I have consulted the work by Delavande and Kohler (2009) and Delavande, Giné and McKenzie (2011) regarding validation tests of perceived risk and actual risk experience. Although it could be useful to test the validation of actual risk experience versus risk expectation, it is not possible to conduct the test because the perceived land security question asks the respondents about the expected risk of land loss that could happen over the next 10 years, while there is no meaningful data on actual land losses reported by farmers. Therefore, we could not conduct a validation test. Similarly, perceived flood question asked the respondents about the perceived risk of flooding that could potentially happen over the next 10 years, while the flooding experience question asked the respondents about flooding instances that occurred over the past 12 months (2013˗2014). Therefore, because the two questions on flooding, namely actual versus expected, cover different time periods, it could not be used to test their validity.

Table 2.5 provides descriptions of the variables and how they are measured.

Table 2.5 Descriptions of variables

Variables Descriptions and how each variable is measured A. Household-level Food insecurity (FI) Equals 1 if household i experienced food insecurity over the past 52 weeks, and 0 otherwise. Length of food insecurity (LFI) Measured in number of weeks of reported food insecurity over the past 52 weeks. Land property rights (LPR) Weighted average of land property rights score for all plots. Cultivated land size per plot was used for the weighting. Normally there are multiple plots per household and sizes vary across plots and households. Land document as ownership proof Binary variable, which equals 1 if a household reported owing (LD) at least a plot and was holding a proof of land ownership. These documents can be in the form of a land certificate, land title, or merely a land registration paper. Collateral (COL) Binary variable. It equals 1 if household had use or was able to use any of their plots as collateral in obtaining a loan, capital, or credit or in gaining other access. Credit or loan access (CA) Binary variable. It equals 1 if household reported that they could borrow or obtain credit from a credit institution. Revenue-cost ratio (RCR) Ratio of rice revenues to cost of rice production. Rice revenue (known also as input efficiency) is total rice harvest times the average price of rice per household. Average price of rice is derived from diving sum of prices of rice by number of crops per household. Cost of rice production excludes other costs, such as chemical input, irrigation cost, and depreciation of farm equipment. Short-run shocks (SRS) If household experienced any shocks, such as droughts, agricultural shocks and non-agricultural shocks during 2010‒ 2014. It equals 1 if household faced any of these shocks and is 0 otherwise. Rice quality (RQ) Is proxied by different prices of rice. Cambodian farming households cultivate different types of rice variety which are suitable to their land type, natural ecosystems and irrigation. Cultivated land size (CLS) Measured in hectares of all cultivated plots per household Soil problem (SP) Is a dummy variable. It equals 1 if household has had any soil quality problems, such as salinity, erosion, infertility, with any of the plots. It is 0 otherwise. HH head’s years of rice-growing Measured in number of years the head of household has been experience (YRCE) cultivating rice. Annual crop frequency (CF) Frequency of rice cultivation, for example 1‒3 crops, per year Size of household (SH) Number of members of household. Village-member trust (VMT) Binary variable, which equals 1 if household has trust in members of the same village. It is 0 otherwise. Government trust (GT) Binary variable, which equals 1 if household has trust in the national and local governments and law enforcement bodies, (local police and court). It is 0 otherwise. B. Village-level Village-town closest distance (VTD) Measured in kilometres to a major town area or district. C. Province-level Annual average precipitation rate Measured in inches as yearly average rate of precipitation for (AAPR) each of the four provinces.

2.6 Estimation strategy

This chapter uses a linear probability model (LPM) and maximum likelihood model (MLM) to test specification (2.1). The ordinary least squares (OLS) method is for the LPM, while logit, probit, and Tobit methods are for the MLM. The use of logit and probit techniques is because the dependent variable is a binary choice, which takes the value of either 1 or 0. The use of Tobit techniques is for when the dependent variable, the length or duration of food insecurity, is measured in number of week(s) of reported experience of food insecurity, censored at zero. The use of the Poisson technique is because the length of food insecurity variable has a characteristic of count data. For specification (2.2), this chapter uses the OLS method. Unlike the OLS method, which minimises the sum of the squared residuals, the logit, probit, Tobit and Poisson methods maximise the log-likelihood function. The use of Tobit and Poisson regressions can also serve as a test of robustness. Factors that have driven variation in private agricultural land property rights across regions in rural Cambodia include the strategic location of the land, its development potential, and wealth status of households. Farmers or landholders would seek legal recognition of land near major towns or district areas suitable for agriculture or industrial development. The land in these areas is in strong demand. The land markets normally require a legal land certificate or land title in a land transfer process, which also stimulates a higher level of property rights registration and titling than if otherwise. Wealth class of households is another factor, i.e., richer households could afford to pay extra money or higher fees to have their land registered and titled. Finally, differences in land management capacities of local governments across districts and provinces might also cause variation in land property rights, which are generally observed in different regions, such as districts and provinces. Therefore, in the estimations I exploit the variation in farmers’ agricultural land property rights. The data used in the estimations are based on subjective choices of survey respondents. Inaccuracy in measuring data based on subjective response method is noted in empirical surveys (Migotto et al., 2005). In my survey, household heads reported their probabilistic expectation of per-plot land rights insecurity based on individual experience. Housewives gave accounts of food insecurity and recounted the quantity of food they consumed. It is likely that farmers could not judge their hunger or length of food shortage precisely and the same way within and across households. Webb et al. (2002) suggested that experience-based measures of food security rely on information that reflects cultural or

personal values of deprivation that may not correspond to more objective measurement. Therefore, the constructed food insecurity indicators derived from subjective, self-reported experiences are not precisely measured. Using the data from this method may involve inaccuracy in estimation results. However, research using subjective questions on experience and expectations in welfare and social studies, has been advanced (McFadden et al., 2004, 2005; Matzkin, 2007; Manski, 2004; Migotto et al., 2005; Jones & Samman, 2016; Hicks, 2011; Tinkler & Hicks, 2011; Beegle, Himelein, & Ravallion, 2012; Ralph et al. (2012), Bover, 2015). Such household surveys have been widely used, such as by Deaton (1997), Pradhan and Ravallion (1998), OECD (2013, 2014), and World Bank (2016) because subjective measurement of expectations can help enrich choice data and provides some usefulness (Manski, 2004). Data on local infrastructure and government support, such as agricultural extension services, are missing because of the difficulty in measuring and obtaining them. Despite these limitations, relevant variables for household characteristics, village characteristics, and province characteristics are accounted for. Indicators, such as village member trust (VMT) to account for social capital or social institutions as a support network, and government trust (GT) to account for trust in public services, are also considered. Village member trust and government trust account for informal networks that different groups of households might have, such as informal financial and seed loans. The closest distance to a major district or town area from each of the 32 villages is controlled for. The distance control captures local infrastructure and market access. The closer the distance, the easier the access and the less costly the transportation and business transactions are expected. In addition, annual precipitation rates for each of the four provinces are also controlled for. Province or district effects are not controlled for. The reason is that this would remove the cross-province variation and leave only the within-province variation. The within-province variation might be relatively more prone to issues arising from the subjective data, due to measurement error. To include the variation across provinces, therefore, I do not include the province or district effects in the estimations.

2.7 Results The estimation results are reported in Tables 2.6 to 2.9. The results on the average marginal effect (AME) are reported in Table 2.8. The results on the tested channels are given 34

in Table 2.9. The logit regressions provide similar results to the probit regressions. For simplicity, the chapter bases the analysis on the probit and the Tobit results. The logit regressions and the Poisson regressions are a test of robustness for the probit and the Tobit regressions, respectively. The OLS results in Table 2.6 (Columns 1‒7) show that there is a significant, negative association between agricultural land property rights and household food insecurity. The logit and probit results (Columns 8‒9) also indicate that land property rights are significantly, negatively associated with food insecurity. The findings point to the significance of private property rights in influencing food insecurity in rural Cambodia. As the evidence indicates, providing more protection of agricultural property rights to farmers, through land registration and recognition, could be an effective tool the government can use to potentially enable rural rice farmers to secure their food access and food availability. Mechanisms that strengthen and uphold their land rights would be significant for farmers to engage in crop production for food security. As for unequal land property rights are concerned, factors that may have caused the variation in obtaining legal recognition of land rights can be many (refer to Section 2.6). They include ability and willingness to pay land registration and titling fees, location of agricultural land and its development potential, and proximity to local land registrar office. As the results indicate, households with weaker or lower land property rights could be expected to have higher degree and longer duration of food insecurity among household members. The main variable of land property rights is perception-based. Despite its subjectivity, it could capture land rights security better than reported land documents. The responses of farmers to land documents question can appear problematic. Most of the surveyed households hesitated to show the documents as proof because they did not hold one or because they were not able to distinguish “hard” land title from other land use rights papers or land registration certificates. Most households consider a land paper they hold an official land title, which is not the case. Column 6 of Table 2.6 shows that when land property rights variable is replaced by self-reported land documents variable, all other coefficients are unchanged, except for government trust which then becomes significant. The negative correlation between government trust and food insecurity would signify that higher trust would not encourage farmers to engage fully in crop production as they would expect that the government would provide support for them if their crops fail. The

question in the questionnaire was intended to solicit information from farmers that could be interpreted as follows. Government trust variable was controlled to account for factors that might be influencing likelihood that farmers would expect the government to support them during time of food crisis or in time of crop loss due to natural disasters, for instance, the effects of El Nino that occurred in 2015 on farmers’ food production capacity. Poor farmers tend to get food relief in such a hard time. The evidence shows no significant association between self-reported land documents and food insecurity. Despite its insignificance, the land documents variable is negatively associated with food insecurity. A possible explanation for the insignificance is that the estimate of land documents is not precise. Table 2.6 shows that the standard error for the land documents variable (0.040) is quite larger than those of the perception-based land property rights variable (0.009). The measurement error might be because of inaccurate reporting of land documents by the surveyed households. One may need to treat land documents variable with care when estimating its potential effects on an outcome. Thus, I used the perception- based land property rights variable in the rest of the estimations. The OLS results for the control variables are also reported. The evidence shows that cultivated land size has a significant, negative link with food insecurity. Village-town closest distance is significantly, negatively associated with food insecurity, suggesting that food insecurity is lower in villages with a shorter distance to major towns and districts. However, it is not significant in the logit and probit estimates. The average rate of precipitation covers the rainfalls over 12-month period (that cover the survey period corresponding to rice cropping seasons) for each of the surveyed province. Using averages of the precipitation rates would, in fact, cancel out some cyclical, seasonal climatic patterns and make the data more correspond to seasonal and annual crop production in Cambodia. Note also that the R2s for the OLS estimates are small, suggesting that some factors that explain or affect household food insecurity are missing from the specification. Despite its low R2s, the OLS results show that land property rights explain differences in food insecurity across households in all estimations. However, the logit and probit methods normally estimate the data better for the binary dependent variable models. The probit and logit estimation results show that land property rights are robust to the inclusion of different sets of control variables and to alternative specifications.

Table 2.6 Impact of agricultural land property rights on household food insecurity Dependent variable Food insecurity (FI) Independent variables OLS OLS OLS OLS OLS OLS OLS Logit Probit (1) (2) (3) (4) (5) (6) (7) (8) (9) Land property rights (LPR) -0.020** -0.019** -0.019** -0.018* -0.017* -0.018** -0.098** -0.060** (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) (0.048) (0.028) Cultivated land size (CLS) -0.028*** -0.028*** -0.028*** -0.028*** -0.028*** -0.028*** -0.351*** -0.194*** (in hectares) (0.007) (0.007) (0.007) (0.007) (0.007) (0.007) (0.117) (0.062) Soil problem (SP) -0.044 -0.045 -0.037 -0.030 -0.015 -0.033 -0.162 -0.091 (0.058) (0.058) (0.057) (0.058) (0.058) (0.058) (0.345) (0.204) HH head’s years of rice- -0.002 -0.002 -0.002 -0.001 -0.001 -0.001 -0.006 -0.005 cultivating experience (YRCE) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.013) (0.008) Annual crop frequency (CF) -0.026 -0.025 -0.029 -0.034 -0.042 -0.036 -0.197 -0.124 (0.054) (0.055) (0.055) (0.055) (0.055) (0.055) (0.347) (0.202) Size of household (SH) -0.008 -0.008 -0.008 -0.009 -0.008 -0.039 -0.023 (0.013) (0.012) (0.013) (0.012) (0.013) (0.086) (0.050) Village-member trust (VMT) 0.013 0.012 -0.023 0.014 0.128 0.064 (0.100) (0.099) (0.098) (0.099) (0.460) (0.269) Government trust (GT) -0.091 -0.087 -0.095* -0.084 -0.583* -0.345*

7 (0.056) (0.057) (0.056) (0.056) (0.347) (0.200) Short-run shocks (SRS) -0.068 -0.063 -0.070 -0.387 -0.218 (0.062) (0.063) (0.062) (0.343) (0.203) Land documents (LD) -0.040 -0.064 (0.070) (0.068) Rice quality (RQ) 0.009 0.004 -0.042 -0.026 -0.019 -0.038 -0.446 -0.233 (0.232) (0.233) (0.243) (0.244) (0.246) (0.243) (1.229) (0.710) Village-town closest distance -0.008* -0.006 -0.005 -0.005 -0.005 -0.003 -0.005 -0.024 -0.015 (VTD) (in kilometres) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) (0.029) (0.016) Annual average precipitation -0.072 -0.031 -0.029 -0.031 -0.027 -0.027 -0.025 -0.095 -0.061 rate (AAPR) (in inches) (0.050) (0.052) (0.052) (0.052) (0.051) (0.052) (0.051) (0.308) (0.179) N 251 251 251 251 251 251 251 251 251 R2 0.038 0.077 0.078 0.088 0.095 0.079 0.094 n.a. n.a. Notes: Robust standard errors are in parentheses for OLS regressions. Standard errors are in parentheses for logit and probit regressions. *, **, *** indicates significance level at 10 percent, 5 percent, and 1 percent, respectively. Land documents that households reported are not statistically significant. It is likely that the self-reported land documents do not capture formal land ownership. Only the perception-based land rights security (PRP) is tested in the rest of the estimations.

The length of food insecurity could be influenced or determined by the degree to which private property rights are observed and the extent to which private property rights, for instance farmers’ agricultural land rights, are guaranteed in the land and financial markets, for instance. The attenuation of private property rights may have the potential to discourage private investment in making economic use of their private property rights, such as agricultural, to produce food and industrial crops, for instance. Furthermore, the test of length of food insecurity on land tenure security serves as a robustness check for the binary food insecurity variable. In fact, the length of food insecurity is a subcomponent of the binary food insecurity indicator. The negative association indicates that as land tenure security improves, there is a higher likelihood that household would face shorter period of food insecurity. The evidence suggests that households with more security in land property rights were having lower probability of food insecurity relative to those households with lower insecurity in land property rights, at least among the rural rice farmers surveyed. In Table 2.7, the OLS estimates in Column (1) are not significant. Because the data for the dependent variable are censored from zero, the Tobit results provide more reliable estimates than OLS results. However, the Tobit and Poisson results presented in Columns (2) and (3) of Table 2.7 indicate that agricultural land property rights have a negative, significant link with the length (in weeks) of reported food insecurity. The R2s for the OLS result in Column (1) of Table 2.7 is very low. A plausible explanation for the OLS results being incorrectly estimated may be because some factors that might affect food insecurity have not been accounted for although efforts were made to account for as many relevant controls as possible. For example, data such as agricultural extension services are not available, and therefore, they have been missed out from the estimated equations. It is interesting to note that trust among the surveyed farm households should be considered good and important. However, it appears that too much trust among themselves may have spoiled their motivation to invest more in rice production because so much trust among them is likely to force them to instead borrow some food from their members when they lack food and when their crops fail. As the result indicates, it appears that those who have too much trust could be those who lack food to consume.

Table 2.7 Impact of agricultural land property rights on length of food

household insecurity (1) (2) (3) Dependent variable Length of food insecurity (LFI) Independent variables OLS Tobit Poisson Land property rights (LPR) -0.067 -0.349* -0.056*** (0.070) (0.201) (0.019) Cultivated land size (CLS) -0.146*** -1.526*** -0.394*** (in hectares) (0.040) (0.458) (0.058) Soil problem (SP) 0.235 -0.109 0.245* (0.457) (1.451) (0.145) HH head’s years of rice-cultivating 0.015 0.017 0.016*** experience (YRCE) (0.020) (0.055) (0.005) Annual crop frequency (CF) -0.514 -1.761 -0.538*** (0.368) (1.456) (0.156) Size of household (SH) 0.064 0.228 0.066** (0.076) (0.337) (0.030) Village-member trust (VMT) 0.753* 1.721 0.773*** (0.428) (1.948) (0.237) Government trust (GT) 0.222 -1.029 0.061 (0.480) (1.412) (0.133) Short-run shocks (SRS) -0.510 -1.922 -0.448*** (0.515) (1.429) (0.137) Rice quality (RQ) -1.770 -3.287 -1.784*** (2.360) (5.075) (0.524) Village-town closest distance (VTD) -0.061* -0.163 -0.056*** (in kilometres) (0.036) (0.117) (0.013) Annual average precipitation rate (AAPR) -0.152 -0.097 -0.065 (in inches) (0.454) (1.237) (0.121) N 251 251 251 R2 0.059 n.a. n.a. /sigma n.a. 7.655*** n.a. (0.753) Notes: Robust standard errors are in parentheses for OLS regressions. Standard errors are in parentheses for logit and probit regressions. *, **, *** indicates significance level at 10 percent, 5 percent, and 1 percent respectively.

Although using the lagged rainfall can help validate the flood risk expectation, the data on rainfall is already average annual precipitation rate per province. Note also that the estimated coefficients of the Tobit and Poisson results reported in Table 2.7 only indicate the possible causal relationship or direction of effects, rather than the size effects. The potential average marginal effects are reported in Table 2.8.

Table 2.8 Average marginal effects of agricultural land property rights on household food insecurity AME AME AME AME logit probit Tobit Poisson Dependent variable Food insecurity (FI) Length of food insecurity (LFI) Independent variables Land property rights (LPR) -0.016** -0.017** -0.082* -0.059***

Cultivated land size (CLS) -0.057*** -0.054*** -0.358*** -0.411*** (in hectares) Soil problem (SP) -0.026 -0.025 -0.025 0.256*

HH head’s years of rice-cultivating -0.001 -0.001 0.004 0.016*** experience (YRCE) Annual crop frequency (CF) -0.034 -0.034 -0.413 -0.562***

Size of household (SH) -0.006 -0.006 0.054 0.069**

Village-member trust (VMT) 0.021 0.018 0.403 0.807***

Government trust (GT) -0.094* -0.095* -0.241 0.063

Short-run shocks (SRS) -0.063 -0.060 -0.451 -0.467***

Rice quality (RQ) -0.072 -0.064 -0.771 -1.862***

Village-town closest distance (VTD) -0.004 -0.004 -0.038 -0.058*** (in kilometres) Annual average precipitation rate -0.015 -0.017 -0.023 -0.068 (AAPR) (in inches) Notes: *, **, *** indicates significance level at 10 percent, 5 percent, and 1 percent respectively. The regressions for the average marginal effects of the logit, probit, Tobit and Poisson equations are based on delta-method standard error (not reported) and are evaluated at the means values of the right-hand-size variables. AME stands for average marginal effect.

Columns (1) and (2) show the average marginal effect (AME) for the logit and probit regressions. The evidence indicates that a one-unit increase in land property rights from the average of 8.22 is associated with a 1.7 percentage point lower probability of household food insecurity per annum. The average marginal effects of Tobit and Poisson regressions are provided in Columns (3) and (4). The Tobit result (Column 3) shows that when there is one- unit increase in land property rights from the average of 8.22, there is a higher probability that households could shorten the length of household food insecurity by about 1 day per household annually, on average. The Poisson result (Column 4) provides similar evidence to the Tobit result. However, the Poisson estimate shows a slightly lower average marginal effect than that of the Tobit estimate. Both Tobit and Poisson estimates show that the average

marginal effects of security in agricultural land property rights on household food insecurity are small. However, strengthening security in agricultural land property rights for rural farmers can still be worthwhile. The results on the control variables are explained as follows. First, the household cultivated land size is significantly, negatively associated with food insecurity (Table 2.6: Columns 2‒9 and Table 2.7: Columns 1‒3). The average marginal effects (AMEs) are significantly large. For example, a one-hectare increase in cultivated land size (CLS) from the average of 2.76 hectares is associated with a 5.5 percent higher probability of reducing household food insecurity (Column 2) or a higher probability that households could reduce household food insecurity by about 2.5 weeks annually (Column 3), on average. Village- member trust appears to be significantly, positively linked with food insecurity (Table 2.7: Columns 1 & 3). The OLS or Poisson results show that rice quality (RQ), short-run shocks (SRS), village-member trust (VMT), and village-town closest distance (VTD) are significantly correlated with food insecurity. Unlike the OLS and Poisson results, the Tobit results show no significant relationship of these variables with household food insecurity.

2.8 Possible channels The chapter also examines whether agricultural land property rights affect credit access, collateral use, and revenue-cost ratio (or input efficiency). The results in Table 2.9 demonstrate that agricultural land property rights are significantly, positively associated with access to rural credit and usage of land as collateral. On average, households that had an increase in agricultural land property rights security by one unit tend to obtain better credit access by about 2.4 times more, could use land as collateral by about 3.8 times more, and potentially increase revenue-cost ratio by about 31.8 times more respectively, relative to households which did not have any improvement in agricultural land property rights security. The positive association between illustrates that an improvement in enhanced security of farmers’ agricultural land property rights will tend to allow farmers to use of their secured farmland to access credit, or to collateral their land in more productive land use initiatives or raise input use efficiency (known as revenue-cost ratio). This evidence highlights the significance of security of private property rights in agricultural land in not only food security but also in boosting farmers’ crop revenue and capital investment in farming.

Table 2.9 Possible channels (1) (2) (3) Dependent variable Credit access Collateral Revenue-cost (CA) (COL) ratio (RCR) OLS regressions

Independent variables Land property rights (LPR) 0.024** 0.038*** 0.318** (0.010) (0.009) (0.127) Cultivated land size (CLS) 0.011 -0.022** 0.510*** (in hectares) (0.011) (0.010) (0.166) Soil problem (SP) 0.080 -0.038 -0.281 (0.068) (0.056) (1.013) HH head’s years of rice-cultivating -0.005* 0.001 0.002 experience (YRCE) (0.003) (0.002) (0.039) Annual crop frequency (CF) 0.033 0.076 1.154 (0.062) (0.058) (1.033) Size of household (SH) 0.020 -0.011 0.183 (0.014) (0.015) (0.276) Village-member trust (VMT) 0.026 0.162* 0.869 (0.095) (0.092) (1.258) Government trust (GT) -0.040 -0.127** -1.174 (0.065) (0.057) (1.077) Short-run shocks (SRS) 0.052 0.031 0.666 (0.070) (0.058) (1.108) Rice quality (RQ) 0.483** -0.150 -1.266 (0.239) (0.199) (2.690) Village-town closest distance (VTD) 0.001 0.004 0.096 (in kilometres) (0.005) (0.005) (0.088) Annual average precipitation rate (AAPR) 0.028 -0.017 -2.197*** (in inches) (0.055) (0.052) (0.836) N 251 251 251 R2 0.077 0.154 0.085 Notes: Robust standard errors are in parentheses. *, **, *** indicates significance level at 10 percent, 5 percent, and 1 percent respectively. Village-town distance is in kilometres. Annual average precipitation is measured in inches. Cultivated land size is measured in hectares. Some important descriptive statistics of households by provinces are provided in the Appendix 3 of the thesis.

With a somewhat major improvement in agricultural land property rights security, there may be growth in credit markets to facilitate loans to rural farmers. Greater security in land property rights could induce not only an expansion of rural credit markets to benefit rice farmers but would also encourage producers of other food and nonfood crops. Similarly, stronger agricultural land property rights would facilitate collateralisation of land, either to access ‘better’ credit or to engage in other productive purposes. Better credit is loans that charge affordable, market-based interest and the one that is without undue risk to borrowers. Normally, banks and credit institutions in Cambodia charge lower interest rates on loans when borrowers have a formal land title to pledge as collateral. But when farmers are not 42

able to collateralise their land or their other fixed property, they are more likely to be restricted from low-cost loans in formal credit markets. Therefore, greater security in land property rights is important when farmland is made liquid in credit markets and is protected in land markets. Rural farmers can possibly use their agricultural land in economic activities more securely, which ultimately would stimulate rural economy. One might also expect a growth in land market alongside credit markets to support farmers’ demand for loans or other financial services. The evidence from this study signifies some significant roles of land rights security in bolstering rural agrarian economy. The local Cambodian rural economy and its disaggregated sectors would potentially internalise the benefits of enhanced land rights security and would in turn facilitate enhanced credit access for rural farmers in the financial and land markets. The evidence is supported by some existing studies, including Brasselle et al. (2002) found evidence of strong, positive impact of land rights security on credit. Other studies, such as Piza et al. (2016), Fenske (2011) Macours et al. (2010), Besley (1995), and de Soto (2000) found that land rights security could enhance not only credit, but it could also turn unused assets into capital. The result in Column (3) shows that land property rights are significantly, positively associated with the revenue-cost ratio. This potential positive impact of land property rights may encourage farmers to deepen their investment in rice agriculture. The revenue-cost ratio can potentially be a mechanism through which an enhancement of agricultural property rights affects rural household food insecurity. The negative coefficient sign of size of cultivated area (CLS) to household food insecurity indicates that as household cultivated area increases, household food insecurity would tend to decline. This evidence simply signifies that households with larger harvested or cultivated land size appears to have had lower food insecurity. This negative correlation between government trust (GT) and food insecurity indicates that trust in the government by local villagers is key to securing food production and food security. It can be a validation that households that had higher trust in the government, including police, judiciary body, and other local authorities, tend to have experienced less food insecurity. It is possible that farm households in the past might have received support from the government in the form of food relief and public services to smooth an instance of food consumption shock, to some extent. The test results of potential impact of (i) credit access and (ii) collateral usage on agricultural land property rights and the results (Table 2.9) show positive and significant link

between security in agricultural land property rights and farmers’ access to credit and land- based collateralization. This positive association illustrates that improved security in land property rights, as explained in the Chapter 2, tends to allow land with proper legal title to be easily accepted in the land and credit markets. This acceptance and recognition of land title has in fact made farmers’ agricultural land more marketable and tradeable. The evidence demonstrates that agricultural land tenure security and cultivated land size each would aid farmers in obtaining higher revenue-cost ratio. In the case that cost of production is not changed or merely changed, higher revenue would render higher revenue-cost ratio or positive revenue-cost ratio. Potentially, farmers who could achieve higher revenue-cost ratio tend to experience less food insecurity. As security in land property rights would lower food insecurity, it can potentially do so in raising farmers’ crop revenue-cost ratio by yielding higher revenue gains per unchanged production cost. Although the data used in the estimation are cross-sectional, if an achievement of higher revenue-cost ratio is sustained over an extended period, farmers could see reduced household food insecurity over time.

2.9 Conclusions This chapter presents evidence on the potential impact of agricultural land property rights on household food insecurity of rural rice farmers, using primary data from a household survey administered to 256 households in 32 rural villages of 8 districts in 4 provinces in Cambodia. The two food insecurity indicators used reflect food availability (supply) and food utilisation (consumption). The first is binary (1 and 0) and the second expresses the number of weeks of reported food insecurity over the past 12 months. The land property rights indicator reflects a binary expectation of security in agricultural land property rights. The alternative land rights measure used self-reported land documents. The results indicate that agricultural land property rights are significantly, negatively associated with household food insecurity. This evidence suggests that security in agricultural land property rights may influence household food insecurity among rice farmers in rural Cambodia. Security in agricultural land property rights could potentially lift rural farmers, most of whom are smallholders, gradually out of hunger in the long run. It may prevent most of the rice farmers from falling further into deeper food insecurity or at least help them smooth food consumption in the short run. The self-reported land documents

variable is insignificantly, negatively associated with the two indicators of food insecurity, but it may not be estimated precisely. The tests on the possible channels provide evidence that land property rights are significantly, positively associated with credit access, collateral, and farmers’ revenue-cost ratio. Securing private property rights in agricultural land can be an option for tackling food insecurity in developing economies, including Cambodia. There can be many drivers and mechanisms of securing food, such as through market participation by smallholder farmers in increasing farm productivity and household income (Barrett, 2008; Rios et al., 2008; Matz, 2015). Other potential mechanisms of securing food for farm household may include farmers’ access to rural credit and collateralisation of farmland. These two indicators could have important effects on rural farmers’ ability to acquire investment capital to fund their agricultural production, such as using high-yielding seeds and cost-effective farm technology to boost farm productivity and to enable farmers to commercialize their farm produce in wider markets. These findings validate the findings and theoretical propositions in some of the existing studies: stronger land property rights enable holders to collateralise their land and gain access to credit. Revenue gains from rice production will encourage farmers to deepen agricultural investment and production. Potentially through augmented credit access, collateralisation, and revenue gains, security in agricultural land property rights will tend to lower rural food insecurity among rice-farming households in Cambodia. Strengthening agricultural land property rights for rural rice farmers, especially for those households that do not have or have least security in land property rights, can be an option for reducing rural food insecurity. It can also be a development tool for improving socioeconomic livelihoods of rural rice farmers, including those households that have never received any land certification and titling services. Strategy for improving security in agricultural land property rights can be an effective option even it is undertaken under governance and institutional transitions. It potentially transcends positive impacts on reducing rural food insecurity and enabling rural credit market and capital formation. The evidence points to a possibility that farmers can obtain loans or financial credit from gaining trust by credit institutions in using their plots as collateral. It is possible that rural credit markets and other agrarian sectors in Cambodia will take advantage of enhanced land property rights when farmers and private credit markets capitalise on secure land rights. Future research could use a larger dataset from a large collection of households.

CHAPTER 3

FLOOD RISK, IRRIGATION, RICE PRODUCTIVITY, AND FOOD INSECURITY IN RURAL CAMBODIA

3.1 Introduction Prevalence of rural food insecurity in Cambodia is as high as 25 to 28 percent (CSES 2011; USAID, 2016). Insufficient intake of food and nutrition for an extended time affects health, labour productivity, and child mortality.14 In Cambodia, rice consumption constitutes about two-thirds of daily caloric intake (Maltsoglou, Dawe & Tasciotti, 2010).15 The poorest 40 percent, most of whom live in rural areas, spend 70 percent of their income on food (CDRI, 2008; Chan, 2011). Between 70 and 80 percent of rural Cambodians rely on rice cropping for food and income (USAID, 2016). Several factors, including extreme flooding phenomena and prolonged droughts, appear to have affected Cambodian agriculture (USDA, 2013; ADB, 2012) and rural livelihood sources (FAO, 2012). While annual floods provide water for agriculture, bring nutrients, and stabilise soil conditions, some phenomena of annual floods have negatively affected rural sectors as well. Annual floods are vital in Cambodian agriculture (MRC, 2016).16 The MRC estimated that the annual floods contribute around USD8‒10 billion annually to the Lower Mekong Basin (LMB) region, much greater than annual cost of flood damage of USD60‒70 million. However, Cambodian agriculture has suffered wet-season flooding (FAO, 2011; USAID, 2013). Existing evidence has shown that excessive flooding undermines systems of food

14 Food (in)security is measured on a yearly basis. It means having unobstructed social and economic access to sufficient intakes of affordable, nutritious food to ensure a healthy and productive life (FAO, 2014; Naylor, 2014; Wegren et al., 2017; Wegren, 2013). Refer to data section in Chapter 1 for further detail of the definition and how indicators of food insecurity are measured and calculated by the FAO. 15 But rice consumption can constitute between 55 and 80 percent of total caloric source in many developing countries (Kiple & Ornelas, 2000, p.132). 16 Annual floods in many parts of the world benefit ecosystems, biodiversity, and economy by recharging groundwater resources, filling wetlands, relocating sediment and nutrients (Queensland Government, 2011). 46

production and reduces crop production, food availability, and rural income (Malla, 2008;

Rosenzweig et al., 2001; Rockström, 2003; Douglas, 2009; Wassman et al., 2009a). Because Cambodia lacks flood-control mechanisms, excessive floods might have undermined its food production capabilities (Johnston, Try, & de Silva, 2013; Wokker et al., 2014; FAO, 2012). Droughts, another recurrent natural phenomenon, could possibly have affected Cambodian agriculture. Only about 24 percent of 4.65 million hectares of cultivable land in Cambodia was irrigated in 2011 (MOWRAM, 2012). An annual increase in acreage under irrigation has remained slow. Covering around 3.1 million hectares or 85 percent of total cultivated area (USDA, 2013), rice cultivation in Cambodia is susceptible to severe flooding, droughts, and climatic abnormalities. Additionally, most irrigation systems in Cambodia are not efficient (Levidow et al., 2014). Among slightly over 2,000 irrigation schemes, only 7 percent are fully functional, and 34 percent are partly functional, while 59 percent are unusable because of inadequate maintenance (Yang et al., 2011). Existing studies have not conducted quantitative investigation into impact of irrigation and extreme flooding on household food insecurity in Cambodia. This chapter examines the links quantitatively, using primary data from the household survey (as in Chapter 2). This chapter first tests whether excessive flooding impacts rice productivity and rice revenue at plot level. The plot-level results indicate that excessive flooding tends to lower rice yields by about 0.7 tonnes per hectare or between 0.3 to 0.4 tonnes per hectare on average, respectively. It tends to reduce rice revenue by about 0.6 million riels (≈USD150) per harvest or 0.3 million riels (≈USD75) per hectare on average, respectively. Second, the chapter tests three irrigation indicators: (i) reservoir, dyke, canal (RDC) irrigation, (ii) river, lake, pond (RLP) irrigation, and (iii) underground or pumping (UP) irrigation, for their impact on rice productivity and rice revenue measured at plot level.17 Formal irrigation (RDC) is found to be associated with higher total rice yield by about 0.98 tonnes per harvest or 1.2 tonnes per hectare and total rice revenue by about 0.81 million riels (≈USD200) per harvest or 0.93 million riels (≈USD225) per hectare. RLP irrigation is associated with higher per-hectare rice yield and per-hectare rice revenue by about 0.38 tonnes and 0.34 million riels, respectively. However, there is no significant link between RLP irrigation and total rice

17 Underground irrigation is limited to small-scale vegetable gardens or fruit farms in the dry season (MOWRAM, 2012). 47

yield per harvest and per-hectare rice revenue. UP irrigation has no significant link with rice productivity and rice revenue. Next, the chapter tests a binary irrigation variable, whether a plot is irrigated regardless of sources, for its impact on rice productivity and rice revenue at plot level. The results show that plots with irrigation tend to have higher rice yield and higher rice revenue per harvest or per hectare. On potential impact, supplying irrigation for the currently unirrigated plots has the potential to improve rice yield by about 0.7 tonnes per harvest, and raise rice revenue by about 0.6 million riels (≈USD150) per harvest, or raise rice revenue by about 0.57 million riels (≈USD140) per hectare, respectively. It is possible that wealthier households or more successful rice farmers would seek ways to irrigate their plots, either by paying water user fees or by purchasing irrigation equipment if their rice revenues are greater than the cost of irrigation. Finally, the chapter tests household food insecurity data on rice yields and revenues measured at household level. The findings show that household-level rice productivity and rice revenue are each negatively associated with household food insecurity. They are the potential channels through which irrigation and excessive flooding could affect household food insecurity in rural Cambodia. The chapter proceeds as follows. Section 3.2 provides key literature. Section 3.3 explains flood risks and irrigation systems in Cambodia. Section 3.4 discusses the model. Section 3.5 describes the data collection. Section 3.6 discusses the estimation strategy. Sections 3.7 and 3.8 discuss the results. Section 3.9 concludes.

3.2 Literature review

Flooding in food production and food insecurity: Many developing countries often face big shortfalls in food supply induced by droughts and floods (Kumar et al., 2013).18 Although annual floods contribute to ecosystems, agriculture, and the economy (Queensland Government, 2011; MRC, 2016), flooding could generate huge impacts on food production outcomes for many Asian countries (Wassman et al., 2009a). It was found to affect food supply and rural income (Malla, 2008; Rosenzweig et al., 2001; Rockström, 2003; Douglas, 2009). The Mekong River Commission (MRC, 2016) highlights that severe flooding worsens

18 In FAO (2000), flooding caused by climatic anomalies has been common in the Horn of Africa. 48

food insecurity and damages crops, property, and infrastructure and disrupts social and economic activities throughout the river basin. Floods, droughts and land degradation deteriorate rice and other food crop ecosystems (Malla, 2008)19, which worsens food insecurity (Rockström, 2003; Wassman et al., 2009b Douglas, 2009; Campbell et al., 2016; Rosenzweig et al., 2001;). Floods and droughts have negatively affected crop production and yields (Piao et al., 2010) and rural livelihood systems (Toufique & Islam, 2014). Toufique and Islam found that households whose access to agricultural land is affected by prolonged or intense flooding produce fewer crops and suffer intensified food scarcity. Floods compromise crop productivity and bring pests and diseases that jeopardise food ecosystems (Malla, 2008). Flooding damages wet-season rice crops, causing food insecurity for poor households (Mirza, 2011). In Bangladesh, vulnerability to floods and droughts has caused shortfalls in food production and fluctuation in food availability despite rice production growth augmented by small-scale irrigation expansion (Hossain et al., 2005).20 However, whether agricultural production affects food insecurity in developing countries directly (Mârza et al., 2015), is unclear.

Irrigation in food production and food insecurity: Existing studies have argued for the importance of irrigation in crop production and rural livelihoods and in reducing food insecurity (Mueller et al., 2012; Pradhan et al., 2015; Rosegrant et al., 2009; Wassman et al.,

2009a; Kang et al., 2009). Irrigation contributes to greater availability of food, more nutritional intake, diversified and balanced diets, and income (von Braun et al., 1989l; Sampath, 1992) and improved nutritional and health outcomes (Rockström et al., 2010; Namara et al., 2010; Domènech, 2015). Lall (2013) and Rosegrant et al (2002) argue that water, used in irrigation, has special roles in agriculture and food security. It provides about 40 percent of global food and fibre supply (Evans & Sadler, 2008). However, irrigation for unirrigated plots can produce between 24 and 80 percent more crop calories than the 2000 levels (Pradhan et al., 2015). Enhanced crop productivity would then lower food insecurity (Hossain & Fischer, 1995). Nguyo et al. (2002) (in Namara et al., 2010) show the percentage of food insecure households living in and outside an irrigation scheme in Kenya: 13 percent versus 33 percent. However, a wide distribution of small irrigated plots across farming areas

19 Floods, droughts and land degradation which are caused by extreme climate or erratic weather 20 It appears that sustainable rice agriculture requires sufficient interventions to cope with weather shocks and irrigation shortage (Bandara & Cai, 2014). 49

is vital to counter adverse effects of seasonal irregularities on food and nutrition consumption (von Braun et al., 1989). Additionally, support systems are needed to boost food production (Carruthers et al., 1997). On a global scale, trade has apparently enabled national food and water security, but increasing food prices and land grabbing might have threatened national food and water security (Allouche, 2011). Insufficient water for irrigation could threaten food production (Ali & Talukder, 2008; Namara et al., 2010; Rockström et al., 2010; Cabangon et al., 2002; Ghosh et al., 2014).21 Some existing studies tend to show that the amount and application of irrigation interrupts crop growth and productivity (Kang et al., 2009; Bandara & Cai, 2014). Other evidence appears to show that irrigation would affect food insecurity (Ragab & Prudhomme, 2000).

3.3 Flood risk and irrigation systems in Cambodia Most rice production in Cambodia is in riverine floodplains that receive floodwaters from the Mekong River and Tonle Sap Lake (FAO, 2012). Among the four surveyed provinces, Kampong Thom and Battambang in Tonle Sap floodplains receive floodwaters from the Mekong River. Banteay Meanchey in the northwest part of Cambodia receives water from overflows of Tonle Sap Lake, upstream catchments, and natural river streams. Prey Veng in the southern part of Cambodia receives floodwaters from the Mekong River, natural rivers, and subsidiary river streams. Despite benefiting from the annual floods of the Mekong River, there are downsides to catastrophic flooding, such as reduced crop productivity (USDA, 2013).22 The flash floods in 2011 caused by excessive rainfall damaged about 10 percent of the wet-season crop (ADB, 2012). The severe flooding in 2013 affected 12 percent of total rice area, slightly less than the affected area of 415,000 hectares in 2011. It damaged around 113,260 hectares or 4 percent of total wet-season rice. The shortage of irrigation infrastructure and mechanisms to control floods has been found to constrain rice productivity (USDA, 2010). The irrigation sector in Cambodia faces some challenges. The irrigation systems in Cambodia date back to the 9th‒10th century (FAO,

21 However, deficit irrigation reduces wasteful consumption of water (Panda et al., 2003) and deficit irrigation practices improve crop water productivity significantly and sometimes by 200 percent (Zwart et al., 2004). 22 The MRC (2016) reported that among the countries in the Lower Mekong Basin, Cambodia and Vietnam pay two-thirds of the cost of annual flood damage. 50

2012; Fletcher et al., 2008). Most irrigation systems were built during 1950‒1953, most of which functioned until the beginning of the Khmer Rouge regime (1975‒1979), which abolished all previous governance systems. The Khmer Rouge undertook a massive project of constructing irrigation canals and dykes (Himel in FAO, 2012). The 12th century Angkorian ‘Water Policy’ developed four reservoirs, storing about 100‒150 million cubic metres of water to irrigate 14,000 hectares (FAO, 2012). By the 13th century, a vast network of reservoirs, canals, and embankments were built in the Northern provinces, covering over 1,000 square kilometres, for use in flood control, agriculture, and rituals (Fletcher et al., 2008). A system of overflows and bypasses were built to carry surplus water into the Tonle Sap Lake. However, they ceased being used in about the 14th‒15th centuries. First, there is unequal distribution of water in agriculture as most irrigation facilities are concentrated only in major rice-growing regions. Second, increasing human settlements on the irrigation structures have affected the functionality of canals and subsidiary networks. Third, insufficient investment and maintenance of the irrigation systems have left major irrigation schemes in limbo.23 The prevention and control of flood in Cambodia is the work of Ministry of Water Resources Management and Meteorology (RGC, 2007); however, there is no established legal and implementation frameworks (NCDM, 2013). Lacking these frameworks might have inhibited the development of flood control systems and disaster management mechanisms in most of rural Cambodia. Ministry of Agriculture, Forestry and Fisheries (MAFF, 2013) has devised a plan of action to tackle disasters in agriculture; however, the plan lacks a long-term strategy for building flood control systems in rice-growing areas to counter shocks from flooding and to prevent and manage floodwaters. To improve rice yields and food security or self-sufficiency and to reduce vulnerability to drought, the Cambodian government has expanded irrigation systems (USDA, 2010; FAO, 2012). In 2010, the Cambodian government committed USD1 billion for irrigation sector development based on its ‘Hegemonisation of Irrigation System Strategy’ (USDA, 2010; FAO, 2012). Over ten dams and irrigation systems have been planned in four northwestern provinces to irrigate wet- and dry-season rice production (FAO, 2012). It was expanded to cover 650,000 hectares of rice area during 1996‒2007 and is expected to irrigate additional 800,000 hectares during 2008‒2017 (USDA, 2013).

23 Rasmussen & Bradford (1977) maintain that large-scale irrigation from wells was minimal and unlikely. 51

The Royal Government of Cambodia (RGC) has prioritised irrigation infrastructure development to increase paddy production and rice productivity as laid out in its ‘Rectangular Strategy’ and ‘Policy Document for Paddy Rice Production and Milled Rice Export’ (FAO, 2012). The RGC’s Document on Promotion of Paddy Rice Production and Export of Milled Rice which aims to boost rice production and productivity to address domestic food shortages and food insecurity and (ii) RGC’s Rectangular Strategy Phase IV, which aims to reduce poverty and improve rice productivity. The RGC’s National Strategy for Agriculture and Water aims to develop and rehabilitate large-scale irrigation infrastructure (Johnston et al., 2013), while National Strategy on Food Security and Nutrition 2014‒2108 (CARD, 2014) emphasises smallholder agricultural development and strategies for reducing rural food insecurity. For example, coping with natural disasters and strengthening land rights of rural farmers is laid out in the RGC’s Rectangular Strategy Phase III (and the updated Rectangular Strategy Phase IV). The National Strategic Development Plan 2014‒2018 and the Rectangular Strategy also outline investment promotion in agriculture as a strategy for ensuring food security and poverty reduction. The proposed agricultural insurance scheme, including rice mortgage program, has never actually been implemented, nonetheless. Although the progress in improving rice productivity and irrigation development has been made over the past 20 years, there are limits as to what impacts could have had on rural food insecurity. Only around 24 percent of all agricultural land was irrigated in Cambodia in 2007, and it has not improved much since (MOWRAM, 2012; FAO, 2012). Irrigation in Cambodia is mainly used for dry-season rice and to irrigate wet-season rice in times of poor rainfall

(Smith & Hornbuckle, 2013) and is still undeveloped (World Bank, 2015a). It still plays a minor role in food production (Tully, 2005) because of inadequate irrigation infrastructure and a lack of secure access to water resources. Compared to its neighbours, Cambodia has about 50 percent lower physical and water productivity (Smith & Hornbuckle, 2013; USAID, 2016) and a higher prevalence of undernutrition and hunger (FAO, 2015).

Myanmar Myanmar Laos

Thailand

Cambodia

Vietnam

Figure 3.1 Irrigation intensity in Cambodia vs. Thailand and Vietnam

Source: FAO Global Map of Irrigation Areas, 2013

Figure 3.1 illustrates that Cambodia has much lower irrigated acreage (namely, 24 percent) than its neighbours. For example, between 50 and 75 percent of the lowlands of south-central Thailand and of the Mekong delta of southern Vietnam are irrigated annually. Some areas in southern Vietnam have between 75 and 100 percent irrigation capacity.

3.4 Basic model

The chapter tests the following specifications:

퐹푅 푌푝 = α + . 퐷푝 + Ω. 퐗풑/퐗풉 + ℰ풑 (3.1) 퐹푅 푌ℎ푎 = α + . 퐷푝 + Ω. 퐗풑/퐗풉 + ℰ풑 (3.2)

푌푝 is a per-harvest measure of rice productivity and rice revenue: (i) total rice yield per harvest and (ii) rice revenue per harvest. Yha is a per-hectare measure of rice productivity 퐹푅 and rice revenue: (iii) per-hectare rice revenue, and (iv) per-hectare rice revenue. 퐷푝 is a dummy for flooding risk, which equals 1 if flooding occurred to plot i during 2010‒2014 and 퐹푅 0 otherwise. 퐷ℎ is a dummy for flooding risk, which equals 1 if flooding occurred to the household j during 2010‒2014 and is 0 otherwise. α is a constant and  is the parameter of interest. Ω is a vector of parameters. ℰ풑 is the error term.

퐗푝 and 퐗ℎ each consist of a similar set of control variables to those used in Chapter 2. However, the controls used in this Chapter are measured at plot and household level. Some household characteristics are included with plot characteristics in the estimation equations. X include land property rights (LPR), size of household (SH), household head’s years of rice- growing experience (YRCE), cultivated land size (CLS), annual crop frequency (CF), soil problem (SP), short-run shocks (SRS), rice quality (RQ), cost of per-hectare chemical input in rice production (CCIRP), cost of other inputs in rice production (COIRP) per hectare, closest distance between village to major district or town (VTD), and annual average precipitation rate (AAPR). Flood risk (FR) variable measures the farmer’s vulnerability to annual flooding, possibly associated with weather shocks, that occurred over 2010‒2014. The other short-run shocks (SRS) variable measures both agricultural and non-agricultural shocks, such as loss of assets and death of cattle or family members, pests, insects, which occurred to individual households during 2010‒2014.

This chapter also tests the following specifications using plot-level data:

퐹푅 푅퐷퐶 푅퐿푃 푈푃 푌푝 = α + . 퐷푝 + . 퐷푝 + θ. 퐷푝 + μ. 퐷푝 + Ω. 퐗풑/퐗풉 + ℰ푝 (3.3) 퐹푅 퐼푃 푌푝 = α + . 퐷푝 + β. 퐷푝 + Ω. 퐗풑/퐗ℎ + ℰ푝 (3.4)

퐼푃 퐷푝 is an irrigation dummy for plots irrigated with any sources of water. Specifically, regardless of irrigation source, if any of the plots are supplied with any of the 퐼푃 푅퐷퐶 three types of irrigation, 퐷푝 equals 1 if a plot is irrigated, otherwise it is 0. 퐷푝 is the dummy 푅퐿푃 for reservoir, dyke, or canal (RDC) irrigation. 퐷푝 is the dummy for river, lake, or pond 푈푃 irrigation. 퐷푝 is the dummy for underground or piping irrigation. α, β, , , θ, and μ are the

parameters. Ω is a vector of parameters. ℰ푝 is the error term. 퐗푝 and 퐗ℎ contain the same set of control variables as in specifications (3.1) and (3.2). I then test the following specifications:

퐹푅 푅퐷퐶 푅퐿푃 푈푃 푌ℎ푎 = α + . 퐷푝 + . 퐷푝 + θ. 퐷푝 + μ. 퐷푝 + Ω. 퐗풑/퐗ℎ + ℰ푝 (3.5) 퐹푅 퐼푃 푌ℎ푎 = α + . 퐷푝 + β. 퐷푝 + Ω. 퐗풑/퐗ℎ + ℰ푝 (3.6)

The specification (3.7) below examines the linkage between household food insecurity and measures of household rice yields and rice revenue Yℎ:

퐹퐼ℎ } = α + ϑ . 푌ℎ + Ω. 퐗ℎ + ℰ풉 (3.7) 퐿퐹퐼ℎ where:

- 퐹퐼ℎ is food insecurity and is a binary response outcome variable. 퐹퐼ℎ takes the value of 1 if household j reported food insecurity and is 0 otherwise. This indicator reflects food availability and food consumption or food utilisation of household. The concept of food security cover four elements. The first food insecurity variable focused in Chapters 2 and 3 reflects food availability or access to food. It somehow captures an aspect of food utilisation, which reflects degree of consumption of diets and nutrient. These explanations are based on the FOA (2012, 2014).

- 퐿퐹퐼ℎ is a non-negative outcome variable, expressed as number of weeks of reported food insecurity that occurred over the past 52 weeks per household. This second indicator captures the length or duration of food insecurity or hunger experience.

- 푌ℎ variables are measured at household level. 푌ℎ includes household rice yield per harvest, household rice yield per hectare, household rice revenue per harvest, and household rice revenue per hectare.

- 퐗ℎ is a set of controls, measured at the household level and the same as those variables in specifications 3.1‒3.6. However, 퐗ℎ excludes short-run shocks, cost of chemical input per hectare, cost of other input in rice production per hectare, and irrigation variables. - ϑ in specifications is the parameter of interest to estimate.

- Ω is a vector of parameters, and ℰ풉is an error term.

3.5 Data collection The sampling strategy, selection of households into the sample, and some data issues are explained in the data section of Chapter 2.24 The following four key questions are taken from the survey questionnaire.

1. The flood risk question: ‘Please assign the coins to the flood events based on your opinion about the likelihood that the flood events will occur in the next 10 years from now?’ This question asks the household head about the probabilistic expectation or likelihood of flooding that would occur over the next 10 years based on their experience.

2. Questions on other shocks (i) ‘What types of serious shocks occurred to your household over the past ten years, i.e., from 2005‒2014, which affected your household’s rice production?’ Shock number = 1–8. Examples of shocks are flood, drought, insect, pest, disease, others (specify). (ii) ‘When (month & year) did it/they happen? How did it/they affect your rice production/rice income?’ I calculated the short-run shocks (agricultural and non-agricultural) by restricting to shocks other than flooding that occurred during 2010‒2014. Short-run shocks are assumed to directly impact rice yields in the year that the flooding occurred and rice- growing capability of rice farmers in the following 2‒3 years or so.

3. Irrigation questions: (i) ‘Among all the plots your household cultivated, is plot i irrigated (i = 1‒6)?’ This is a binary question (1 = Yes; 0 = No) (ii) ‘What are the sources of water your household use for rice cultivation?’

24 The survey also incorporates village characteristics considering their vulnerability to natural disaster risks, such as floods and droughts. The purpose is to avoid sampling households that are in floodplains only or drought areas only. This sampling strategy reflects the local features of rural Cambodia, where rice fields can be prone to either flooding or major droughts. Lastly, the selection of households into the sample is random, not based on a criterion that households had been affected by past flooding or food insecurity.

Multiple choices (rainwater, underground or pumped water, natural ponds or lakes, and water from reservoirs, dams, or canals) are supplied to the question.

4. The food insecurity questions are the same as those used in Chapter 2: (i) ‘In the past 12 months, were there any days and weeks that your household had very little, not enough or no food (“was hungry”)?’ (ii) ‘If YES, how many weeks of the past 52 weeks did the household have so little food or no food at all that the household was hungry?’

Two dimensions of food insecurity, namely food access (supply) and food utilisation (nutrition), are examined. Heterogeneity in plot sizes and household characteristics were incorporated in the survey. The data display variation in plot characteristics and sources of irrigation each household had access to. In selecting households, there was no prior information about household irrigation sources, access, and choices. Farmers use three other sources of irrigation: (i) RDC, (ii) RLP, and (iii) UP irrigation, in addition to rainwater. These sources present variation in quality and access to irrigation. The first category is probably the least expensive. The second category is managed by individual households. The third category is a common pool resource. Where rivers, lakes, and ponds are distant from farmers’ rice fields, their own investment to pump water into their rice fields can be costly. The variation in irrigation access and vulnerability to flooding among households are explained later in this section. However, in data collection it is possible that data collected are not precisely measured. If any, measurement error could possibly occur when farmers recalled their experiences about flooding and food insecurity from memory. They gave an account of their experiences subjectively. This method, as discussed in Chapter 2, may not ensure data accuracy. Despite the shortcomings, the recall method has been widely used in household surveys and social studies. Plots that rely solely on rainfall were regarded as not having access to modern irrigation supply. No matter where plots are located, at least they are irrigated by rain, and because the distribution of rain is a natural phenomenon, I treated rainfed farming as not having irrigation access. Comparing different types of irrigation types and their access with rainfed farming can be useful a policy analysis. Importantly, different irrigation types reflect access, availability, and associated costs in obtaining irrigation for their rice fields. Distance

from sources to rice fields, equipment used to pump water into rice fields, and availability and access fees (free vs. paid) affect the ways and to the extent to which farmers choose to irrigate that would affect their crop production. Because rainfall can be unpredictable and vary by climatic conditions, yields can be much affected by rainfall variation.

Table 3.1 Plot-level and household-level basic survey data # Variable Description # of HHs or plots Percent 1 Food insecurity 1 = food insecurity; 0 = no 65 HHs = 1 25.89 food insecurity 186 HHs = 0 74.1 2 Length of food insecurity 1 to 26 weeks; mean weeks of 32 HHs = 1 week 12.75 food insecurity = 1.04 17 HHs = 2‒4 weeks 6.77 8 HHs = 5‒8 weeks 3.19 8 HHs > 8‒26 weeks 3.19 3 Irrigation type RDC irrigation 58 plots 10.18 RLP irrigation 134 plots 23.51 UP irrigation 37 plots 6.49 4 Irrigated plots Plots irrigated regardless of 238 plots 41.77 irrigation types Source: Author’s calculations. # of HHs is 251 and total plots is 570. The data on (1) and (2) are measured at household level. The data on (3) and (4) are measured at plot level.

Table 3.1 shows that about 25 percent of 251 households, or 65 households, reported food insecurity, from 1 week to 26 weeks over the past 12-month period (March 2013 to May 2014). 12.75 percent, or 32 households of total households, did not have enough food to eat for about one week. 6.77 percent, or 17 households of all households, did not have food to eat between 2 and 4 weeks. 6.38 percent, or 16 other households, did not have food from 5 to 26 weeks. 74.1 percent, or 186 households, of the total households reported having no food insecurity over the same period. Only 58 plots or 10.18 percent of 570 plots were irrigated with water from RDCs. 134 plots or 23.51 percent and 37 plots or 6.49 percent of 570 plots were irrigated with RLP irrigation and UP irrigation, respectively. About 42 percent of the plots were irrigated with one of the three irrigation sources. Rainfall is not tested because it is the general source of water that is available to all farmers despite its some variation across locations. Self-supplied irrigation from UP and RLP sources by means of water pumping tools is more expensive and farmers tend to suffer inefficiency loss and sunk costs. Farmers can have all types of irrigation, but in general they would weigh the costs of using an alternative irrigation sources over an existing one(s) in relation to availability of irrigation. I proxied a dummy for whether a plot was irrigated with at least one of the three

irrigation types, which was roughly 42 percent of 570 plots. In so doing, the concern about whether risk of food insecurity would be spread out or not has been addressed by estimating impact of the irrigated dummy. I controlled flood risk in the data in both plot-level and household-level estimations.

Table 3.2 Descriptive statistics of plot-level data

Variable Mean Standard Minimum Maximum Deviation Rice yield from latest harvest at plot level 2.78 5.26 0.02 72 (in tonnes) Rice productivity per hectare 2.43 1.99 0.05 9 (in tonnes) Rice revenue from latest harvest at plot level 2.20 4.23 0.02 57.6 (in million Cambodian riel) Rice revenue per hectare 1.94 1.64 0.04 9 (in million Cambodian riel) Land property rights 8.41 3.26 0 10 (0 = lowest, 10 = highest) Short-run shocks (Binary 0 & 1) 0.69 0.46 0 1 Flood risk (Binary 0 & 1) 0.67 0.47 0 1 Rice quality (in thousand) 0.80 0.13 0.5 1.2 Cultivated plot size (in hectares) 1.22 1.41 0.05 12 Soil problem (Binary 0 & 1) 0.58 0.49 0 1 HH head’s years of rice-cultivating 27.23 12.15 1 61 experience (# of years) Annual crop frequency 1.33 0.49 1 3 (Number of crop cultivation per year) Cost of chemical inputs in rice production per 0.01 0.08 0 1.25 hectare (in billion Cambodian riel) Cost of other inputs in rice production per 0.03 0.37 0.00005 6.25 hectare (in billion Cambodian riel) RDC irrigation 0.10 0.30 0 1 (binary: 1 if RDC irrigated, 0 otherwise) RLP irrigation 0.23 0.42 0 1 (binary: 1 if RLP irrigated, 0 otherwise) UP irrigation 0.06 0.25 0 1 (binary: 1 if UP irrigated, 0 otherwise) Irrigated plots 0.42 0.49 0 1 (binary: 1 if irrigated, 0 otherwise) Household size 5.19 2.01 2 16 (absolute number) Village-town closest distance 11.58 6.16 1.97 24.21 (in kilometres) Annual average precipitation rate 54.67 0.53 54.1 55.6 (in inches) Notes: Number of plots in this sample is 570. The data on precipitation rates are taken from https://weather-and-climate.com/average-monthly-Rainfall-Temperature-Sunshine-in-Cambodja

The household-level data presented in Table 3.3 are used to examine the potential impact of rice productivity and rice revenues on household food insecurity.

Table 3.3 Descriptive statistics of household-level data

Variable Mean Standard Minimum Maximum deviation Food insecurity 0.24 0.43 0 1 (binary: 1 = yes, 0 otherwise) Length of food insecurity (# of weeks) 1.04 3.25 0 26 Total household rice yield from latest 6.32 9.32 0.03 84 harvest (tonnes) Household rice productivity per hectare 2.38 1.87 0.06 9 (tonnes per hectare) Total household rice revenue from latest 5.01 7.52 0.024 67.20 harvest (in million Cambodian riels) Rice revenue per hectare 1.91 1.55 0.048 8 (in million Cambodian riels) Land property rights 8.22 3.32 0 10 (0 = lowest, 10 = highest) Rice quality (in thousand) 0.79 0.13 0.50 1.20 Total cultivated land size (in hectares) 2.76 2.95 0.10 20 Soil problem (1 = yes, 0 otherwise) 0.61 0.46 0 1 HH head’s year of rice-cultivating 26.95 11.94 1 61 experience (# of years) Annual crop frequency 1.34 0.48 1 3 (Number of crop cultivation per year) Household size (Absolute number) 5.18 1.99 2 16 Village-town closest distance 11.10 6.02 1.97 24.21 (in kilometres) Annual average precipitation rate 54.72 0.56 54.10 55.60 (in inches) Notes: Number of households in this sample is 251. On average, each household cultivated 2 to 3 plots. The data on precipitation rates are taken from https://weather-and-climate.com/average-monthly-Rainfall-Temperature-Sunshine-in-Cambodja

The summary statistics of the village-town closest distance are given in Tables 3.2 and 3.3. The descriptive statistics on this variable and the data that show variation in access to different types of irrigation are already provided (Tables 3.2 and 3.3). Additional statistics are also provided. For example, out of the 32 surveyed villages, 16 were not flooded during that time. Despite this, some of the rice plots of farmers in the unflooded villages could in fact have been flooded because they are located away from the villages or are in the flooded areas of other villages. Likewise, not all rice plots of each household in the flooded villages were flooded because some of their rice plots are located elsewhere away from the flooded areas. The survey data show that 286 plots of the 251 households were not flooded and the other 60

284 plots were flooded during 2013–2014. In the household data, about 23.9 percent were identified as having an ID-poor card given by the government. In Cambodia, although an irrigation may exist in an area, it might not function properly due to lack of maintenance and rehabilitation. Moreover, there has been an issue with water use management and maintenance. Thus, most irrigation systems in Cambodia are not operational. However, there data on year of construction, conditions, and irrigation capacities for each irrigation scheme are not available. The measures of per-hectare rice productivity and rice revenue are computed at the plot and household levels. The length of experience in rice farming captures agricultural techniques and technical capacities of the household head accumulated over the years. The subjective flood risk measures flooding that lasts more than 10 days and is higher than knee depth. Longer periods and deeper depths than these are considered major flooding of alarming magnitude. In the survey, each of the households was given 10 coins and the household head was asked to assign them to each of their plots based on their experience or the perceived likelihood that flooding occurs over a period of 10 years. Suppose that a household head placed three coins on plot i. This suggests that the farmer expected the flooding to occur for 3 years on that plot i over a period of 10 years. However, the severe flooding variable used in the estimation is calculated to measure whether plot i or household j experienced severe flooding during 2010‒2014. I restricted it to this period, the most recent flooding experience, so that farmers could easily recall it and importantly that accounts for recent crop loss caused by flooding. If plot i or household j experienced flooding during this period, the flood risk variable is 1 and 0 otherwise. The flooding indicator being estimated in this chapter features a seasonal or annual phenomenon that occurs from early stage of rice planting to late stage of rice growth to harvest stage. The flooding indicator captures flooding experiences that occurred during 2010‒2014. Although it happened to rice plots or households prior to rice production in 2013‒2014, the crop damage reported by households reflects the estimated loss for the preceding years that correspond to their rice cultivation experiences over 2010‒2014. The extent and intensity of flooding can vary from year to year. I controlled for village-district distance to capture village effects. The flood risk variable has been controlled in all estimations. The village-town distance would capture access to markets, public services, and other information, as well as the issue of flooding.

Other short-run shocks are shocks other than flooding that occurred during 2010‒ 2014. Those other shocks include droughts, agricultural, and non-agricultural shocks other than flood risk, which includes pests and insects, livestock mortality, death of family members, and loss of assets. It is reasonable to argue that shocks that occurred in the past 1 to 5 years have discouraged farmers or at least they continue to influence farmer’s decisions and ability to invest in agriculture. The restriction of shocks that occurred over this time span is sensible. Annual average precipitation rates are in inches, measured at the province level, and are used because data at the village, commune, and district levels are not available. The precipitation rate relates to whether supplemental irrigation is required and whether observed climatic factors influence annual cropping frequency or intensity. In sum, the specifications manifest a correction for some unobserved effects by the inclusion of various household characteristics, village characteristics, and province characteristics. Average prices of rice for different rice varieties farmers were planting is different from rice revenue per household. They are somehow related but are different. Rice quality is based on different types of rice varieties farmers were planting to suit their plot characteristics, such as soil quality, availability of irrigation, and length of planting season and expected harvest. It is derived from averaging prices of rice. In contrast, rice revenue is derived from price of rice per plot times total production per plot. To derive household rice revenue, all revenues from cultivated plots per household are summed. Only the data on annual average precipitation were available for the analysis. Even though plots are located far from each other, precipitation rate would be the same for households within the same province. However, the precipitation rates only differ across provinces, rather than within province. This, in fact, is the same as controlling for province fixed effects for households within the same province. There is possibility that availability or access to formal irrigation in Cambodia is endogenous. That might happen when wealthier households, households with easier access to low-cost loans, or more successful rice farmers invest in purchasing own irrigation equipment where access to formal irrigation is limited. Therefore, wealthier households, those households willing to pay water user fees, or more successful farmers could be able to obtain access to irrigation water, which is generally managed and operated by a local farmer water user community (FWUC). Most formal irrigation infrastructure in Cambodia are built

by the government, and so there is very slight possibility that households could have influenced the government decision. Although households do not appear to have no control over government investment in the construction and expansion of reservoirs, dykes, or canals, the government might have responded to some of their needs to retain their loyalty in national and subnational elections. It is difficult to identify the endogenous process precisely. However, somewhat subtle variations in irrigation availability and access across the sample could be traced to two other factors. First, the geographical locations of the rice fields and the availability of water resources could be a primary reason why an expensive irrigation infrastructure could be developed. For an area to receive an irrigation system developed is because such factors such as the natural topography of the areas next to or around the Tonle Sap Lake and other natural river streams. Areas with better development potential could have an irrigation infrastructure developed in their community. However, it may not be the primary reason the government has endorsed irrigation sector development strategy. It is also important to note that the differences in availability of irrigation sources can be noticeably observed within a commune or across districts but less within a village. Therefore, I specify the estimation equations to exploit the variations in irrigation access and availability. One might argue that wealthier households would seek to acquire plots away from areas easily affected by seasonal or flash floods. Furthermore, if their plots are in a flood- prone area, those richer households could possibly build drainage canals or embankments to avert some flooding risk. However, there is little information that only poorer households live in or have their plots in flooded areas. When an excessive flooding occurs, partial to total flood drainage systems are damaged because they are not built of fortified materials resistant to severe floods. It has been observed that rice fields of both poor and richer households have been flooded by seasonal or flash floods with similar or varying magnitudes. In fact, there is a somewhat exogenous variation in flooding risk among poor and rich households in rural Cambodia. Hence, I exploit this variation in flooding risk in the estimations. The four selected provinces are in the low floodplains and they are among the main rice-producing areas. They are not very different in terms of agroecological conditions from other rice-producing areas. However, what make them different is supply or availability of irrigation or access to different modes of access to irrigation. In addition, their altitudes or climatic conditions can differ among them. I exploit this variation in these factors by

controlling for average annual precipitation to capture variation in topographical conditions and climates that would affect crop production. The list of variables and the detailed descriptions of how each variable is measured or calculated are given in Table 3.4.

Table 3.4 Descriptions of variables

Variables Descriptions and how variables are measured D. Plot-level data

Reservoir, dyke, or canal The formal system of irrigation is from reservoirs, or dykes, irrigation (RDC) or canals. A farmer water user community (FWUC) runs and operates an irrigation scheme at commune, and in some areas at district, level. A user fee is generally charged for use of irrigation, and water usage is regulated. River, lake, or pond irrigation It is a common pool resource. The irrigation from natural (RLP) rivers, lakes, or ponds can be used by individual households against competing demands. The use of it is free-of-charge, except for own investment in irrigation equipment, such as machinery and water pipes. Underground or pumping It is operated privately by an individual household that can irrigation (UP) afford to invest in water pumping equipment. This option is used mainly by those that have little or no access to formal irrigation or who choose it as complementary irrigation. Irrigated plots (IP) If plot i is irrigated with any of the three irrigation types, it equals 1. If it is rain-fed, meaning it is not irrigated with the any of the three sources, it equals 0. Rice yields from latest harvest Yields from each plot from the latest rice harvest. (Yh) Rice revenue from latest It is measured in tonnes per plot times price of rice for each harvest at plot level (REV) plot. Cost of chemical inputs in rice Include costs for chemical fertiliser, weeding, insecticide, production per hectare (CCIRP) and chemical pesticide (in billion Cambodian riels). It is measured as ratio of total cost of chemical inputs to total cultivated land size per household. Cost of other inputs in rice Cost of other inputs in rice production (in billion Cambodian production per hectare riels), which excludes cost of irrigation and fertiliser, over (COIRP) total cultivated land size per household. Land property rights (LPR) Whether plot i is secure equals 1, and is 0 otherwise Soil problem (SP) If plot i has had any soil quality problem it equals 1, and 0 otherwise. Flood risk (FR) It equals 1 if plot i was flooded over the past five years, namely during 2010‒2014, and is 0 otherwise. Rice quality (RQ) Is captures different types of rice seeds or variety of rice that farmers cultivated. If farmers adopted more expensive seeds, it means they cultivate higher rice quality rice. Annual crop frequency (CF) Frequency of rice cultivation, ranging from 1‒3 crops, per year per plot.

Variables Descriptions and how variables are measured E. Household-level Food insecurity (FI) It equals 1 if household j experienced food insecurity over the past 52 weeks and is 0 otherwise. Length of food insecurity (LFI) Measured in number of weeks of reported food insecurity over the past 52 weeks. Flood risk (FRh) It equals 1 if household j experienced flooding that occurred to any of the plots over 2010‒2014 and is 0 otherwise. Short-run shocks (SRS) If household experienced any of shocks: other agricultural shocks, drought and non-agricultural shocks during 2010‒ 2014. These shocks include death of family members, loss of assets or cattle, pest or insects, and long-term illness or disability. It equals 1 if household faced any of these shocks and is 0 otherwise. Household rice yields per Yield or harvest from the latest crop season per each harvest (Yh) household and is measured in tonnes. Household per-hectare rice Total yield or harvest from all plots per household divided yield (Yh) by total cultivated land size. Household rice revenue per Yield per plot times the average price of rice varieties harvest (REVh) cultivated by each household. Household per-hectare rice Rice productivity per hectare times the average price of rice revenue (REVh) varieties cultivated by each household. Annual crop frequency Frequency of rice cultivation, ranging from 1‒3 crops, per (CF) year per household. Land property rights Ratio of sum of land property rights for all plots to total (LPR) cultivated plot size per household. Normally there are multiple plots per household and sizes vary across plots and households. Cultivated plot size is used to weight household LPR. HH head’s years of rice-growing Measured in number of years the head of household has been experience (YRCE) cultivating rice. Soil problem Is a dummy variable. It equals 1 if household has had any of (SP) the following problems: salinity, erosion, infertility, with any of the plots per and it is 0 otherwise. Cultivated land size (CLS) Measured in hectares of all cultivated plots per household. Size of household (SH) Number of members of household. Rice quality (RQ) It aims to capture different quality of rice or rice variety which each household cultivated. F. Village-level Village-town closest distance Measured in kilometres to a major town area or district. (VTD) G. Province-level Annual average precipitation rate Measured in inches as yearly average rate of precipitation for (AAPR) each of the four provinces.

3.6 Estimation strategy This chapter assesses the household survey data using the ordinary least squares (OLS) method and maximum likelihood estimations (MLE) of probit and Poisson techniques. The chapter relies on the MLEs that use probit regressions for the binary dependent variable and Poisson regressions for the count dependent variable. This is because the OLS method tends to estimate data outside the boundary when the dependent variable is binary or is count data. As in Chapter 2, this chapter uses two indicators of food insecurity: (i) binary 0 and 1 and (ii) number of weeks of food insecurity. The chapter uses two levels of data: (i) plot and (ii) household. The chapter first assesses the plot-level data when examining the potential impact of irrigation and flooding on rice productivity and revenues. It then examines household-level data when analysing the potential impact of rice productivity and revenues on household food insecurity. In so doing, the chapter takes advantage of data availability at different levels to examine the linkages.

3.7 Results: Flood risk, irrigation and rice production The OLS results reported in Tables 3.6 to 3.8 show that severe, excessive, or extreme flooding is significantly, negatively associated with per-hectare and total rice yields and per- hectare and total rice revenues, measured at plot level. 25 The plot-level evidence in Table 3.5 shows that severe flooding tends to reduce total rice yield by about 0.7 tonnes per harvest on average. This potential crop loss caused by an extreme flooding phenomenon per rice- cropping season is somewhat noticeable. The evidence in Table 3.6 shows that severe flooding reduces rice yields by about 0.3 to 0.4 tonnes per hectare on average, relative to plots that were not affected by severe flooding. The evidence in Table 3.7 shows that severe flooding tends to lower total rice revenue by about 0.6 million Cambodian riels per hectare (≈USD150) on average. The evidence in Table 3.8 indicates that severe flooding tends to reduce rice revenue by about 0.3 million Cambodian riels (≈USD75) per hectare on average. The estimated impact is an average impact of a severe flooding phenomenon that occurred during 2010‒2014. The estimated crop damage of extreme flooding is the crop loss calculated after rice harvest per cropping season. In a few cases, farmers’ rice paddies were completely

25 This evidence is consistent with Karunasagar and Karunasagar (2016), Gregory et al. (2005), and Malla (2010). 67

ruined by extreme flooding, and therefore the estimated loss is the amount of rice that household would expect to be able to harvest in a lean season without the extreme flooding.26

Table 3.5 Flooding and irrigation impacts on total rice yield from latest harvest at plot level Dependent variable Total rice yield (in tonnes) (1) (2) (3) (4) (5) Independent variables OLS Flood risk_p -0.718** -0.734** -0.683** -0.663** -0.689** (2010‒2014) (0.300) (0.301) (0.301) (0.296) (0.299) Reservoir/dyke/canal irrigation_p 0.982* (RDC) (0.516) River/lake/pond irrigation_p 0.366 (RLP) (0.555) Underground/pumping irrigation_p 0.863 (UP) (0.602) Irrigated plots_p 0.643 (IP) (0.439) Short-run shocks_h 0.398 0.349 0.335 0.274 0.294 (SRS) (agri & nonagri) (0.293) (0.291) (0.290) (0.284) (0.284) Rice quality_p 0.623 0.595 0.617 0.663 0.629 (RQ) (0.830) (0.821) (0.806) (0.808) (0.792) Cultivated land size_p 2.947*** 2.958*** 2.966*** 2.973*** 2.967*** (CLS) (0.484) (0.487) (0.487) (0.495) (0.491) Soil problem_p 0.381 0.372 0.426 0.475 0.443 (SP) (0.317) (0.315) (0.320) (0.322) (0.318) HH head’s years of rice-cultivating -0.032*** -0.031*** -0.029*** -0.026** -0.026** experience_h (YRCE) (0.011) (0.011) (0.011) (0.012) (0.012) Annual crop frequency_p 1.682*** 1.690*** 1.645*** 1.433*** 1.358*** (CF) (0.357) (0.362) (0.358) (0.508) (0.480) Cost of chemical inputs in rice 4.495*** 4.619*** 4.366** 5.012*** 4.993*** production per ha_p (CCIRP) (1.731) (1.741) (1.733) (1.576) (1.540) Cost of other inputs in rice -0.534 -0.562* -0.530 -0.628** -0.623** production per ha_p (COIRP) (0.341) (0.341) (0.341) (0.312) (0.305) Village-town closest distance_v -0.012 -0.009 -0.002 -0.001 0.001 (VTD) (0.023) (0.022) (0.022) (0.022) (0.022) Annual average precipitation -1.457** -1.443** -1.451** -1.423** -1.421** rate_pv (AAPR) (0.580) (0.574) (0.573) (0.605) (0.588) Size of household_h -0.073 -0.084 -0.102 -0.103 (SH) (0.071) (0.071) (0.067) (0.066) Land property rights_p 0.077* 0.089** 0.084** (LPR) (0.041) (0.041) (0.040) N 570 570 570 570 570 R2 0.608 0.609 0.611 0.615 0.614 Notes: *, **, **each denotes significance at 10%, 5%, 1% level, respectively. The robust standard errors for OLS estimations are in parentheses. N is total number of plots of all 251 households in the sample. p, h, v, and pv denote data at plot, household, village, and province level, respectively.

26 Note that not all plots were flooded in the same year. Plots that were flooded are close to the major river streams in the flood-prone areas. Some plots have never been flooded. Each household normally has more than one plot in different locations. Refer to the sampling strategy on household selection in the data section. 68

The results indicate that excessive flooding is bad for total rice productivity and rice revenues and per-hectare rice productivity and rice revenues, which could be translated to reduced food production and therefor reduced food availability. These results are consistent even when household size and agricultural land property rights are included (Columns 2‒5).

Table 3.6 Flooding and irrigation impacts on per-hectare rice yield at plot level Dependent variable Rice yield (in tonnes per hectare) (1) (2) (3) (4) (5) Independent variables OLS Flood risk_p -0.375** -0.370** -0.344* -0.303* -0.350** (2010‒2014) (0.177) (0.178) (0.177) (0.172) (0.174) Reservoir/dyke/canal irrigation_p 1.201*** (RDC) (0.286) River/lake/pond irrigation_p 0.378* (RLP) (0.223) Underground/pumping irrigation_p 0.503 (UP) (0.365) Irrigated plots_p 0.676*** (IP) (0.194) Short-run shocks_h 0.010 0.027 0.020 -0.028 -0.023 (SRS) (agri & nonagri) (0.178) (0.180) (0.179) (0.171) (0.175) Rice quality_p -0.010 -0.000 0.011 0.049 0.024 (RQ) (0.565) (0.566) (0.561) (0.562) (0.545) Cultivated land size_p -0.081 -0.085 -0.081 -0.078 -0.080 (CLS) (0.064) (0.065) (0.064) (0.066) (0.066) Soil problem_p -0.064 -0.062 -0.034 0.014 -0.016 (SP) (0.161) (0.161) (0.162) (0.162) (0.163) HH head’s years of rice-cultivating -0.021*** -0.022*** -0.021*** -0.018*** -0.017*** experience_h (YRCE) (0.006) (0.006) (0.006) (0.006) (0.006) Annual crop frequency_p 1.377*** 1.374*** 1.351*** 1.172*** 1.049*** (CF) (0.173) (0.174) (0.175) (0.202) (0.197) Cost of chemical inputs in rice 1.894 1.852 1.724 2.404* 2.383** production per ha_p (CCIRP) (1.441) (1.443) (1.447) (1.256) (1.167) Cost of other inputs in rice 0.743*** 0.753*** 0.769*** 0.666*** 0.671*** production per ha_p (COIRP) (0.287) (0.287) (0.287) (0.248) (0.229) Village-town closest distance_v 0.006 0.006 0.009 0.012 0.013 (VTD) (0.011) (0.011) (0.011) (0.011) (0.011) Annual average precipitation -0.087 -0.092 -0.097 -0.068 -0.065 rate_pv (AAPR) (0.150) (0.151) (0.151) (0.152) (0.152) Size of household_h 0.025 0.020 0.001 -0.001 (SH) (0.037) (0.037) (0.036) (0.036) Land property rights_h 0.039* 0.049** 0.046** (LPR) (0.022) (0.021) (0.021) N 570 570 570 570 570 R2 0.192 0.193 0.197 0.228 0.218 Notes: *, **, **each denotes significance at 10%, 5%, 1% level, respectively. The robust standard errors for OLS estimations are in parentheses. N is total number of plots of all 251 households in the sample. p, h, v, and pv denote data at plot, household, village, and province level, respectively.

The evidence appears to correspond to the crop losses reported by the Cambodian government and numerous international organisations. In Cambodia, extreme flooding in 2011 damaged about 10 percent of total cultivation (USDA, 2013). In 2013, severe flooding ruined about 8 percent of farmers’ rice crops and their other assets. The plot-level evidence reported in Tables 3.5‒3.8 show that providing formal irrigation, i.e., irrigation supplied from reservoirs, dykes, or canals (RDC) for the currently unirrigated plots, is linked with raising total rice yield (Table 3.5), per-hectare rice yield (Table 3.6) by about 0.98 tonnes, 1.2 tonnes, 0.81 million riels, and 0.93 million riels, respectively. The evidence also shows that access to formal irrigation could raise total rice revenue (Table 3.7) and per-hectare rice revenue (Table 3.8) by about 0.81 million riels, and 0.93 million riels, respectively. RLP irrigation is linked with higher per-hectare rice yield (Table 3.6) and per-hectare rice revenue (Table 3.8) by about 0.38 tonnes and 0.34 million riels, respectively. However, there is no significant link between RLP irrigation and rice yield per harvest (Table 3.5) and rice revenue per harvest (Table 3.7). Lastly, UP irrigation has no significant relationship with any of the measures of rice productivity and rice revenue. The results on irrigated plots, a binary irrigation variable representing whether individual households had access to RDC, UP, or RLP irrigation, are provided in Tables 3.6 to 3.8. The results indicate that having irrigation access is significantly, positively associated with rice productivity and rice revenues.27 Plots with irrigation (or simply irrigated plots) tend to have higher rice yield and higher rice revenue. As the evidence indicates, providing irrigation for the currently unirrigated plots could potentially raise per-hectare rice yield by about 0.7 tonnes (Table 3.6) or higher rice revenue per harvest by about 0.6 million Cambodian riels (≈USD150) on average (Table 3.7). Plots with irrigation also tend to have higher per-hectare rice revenue by 0.56 million Cambodian riels (≈USD140) on average (Table 3.8). Higher revenues for farmers when their plots become irrigated, having previously been unirrigated, mean that rice production becomes more profitable. The potential gains in rice productivity and rice revenues are important for rural farmers, most of whom do not have access to sufficient water for rice farming and could benefit from an expansion in irrigation systems and access.

27 The results are consistent with Wokker et al. (2014) and Rosegrant et al. (2009): irrigation affects rice yields and productivity. 70

Table 3.7 Flooding and irrigation impacts on total rice revenue from latest harvest at plot level (1) (2) (3) (4) (5) Dependent variable Total rice revenue (in million Riels) Independent variables OLS Flood risk_p -0.586** -0.601** -0.562** -0.546** -0.567** (2010‒2014) (0.241) (0.242) (0.242) (0.238) (0.240) Reservoir/dyke/canal irrigation_p 0.806* (RDC) (0.416) River/lake/pond irrigation_p 0.416 (RLP) (0.452) Underground/pumping irrigation_p 0.730 (UP) (0.479) Irrigated plots_p 0.595* (IP) (0.355) Short-run shocks_h 0.289 0.244 0.234 0.184 0.195 (SRS) (agri & nonagri) (0.239) (0.239) (0.239) (0.234) (0.234) Rice quality_p 1.780** 1.754** 1.771** 1.811** 1.782** (RQ) (0.790) (0.780) (0.771) (0.768) (0.755) Cultivated plot size_p 2.366*** 2.376*** 2.382*** 2.386*** 2.383*** (CLS) (0.387) (0.389) (0.389) (0.396) (0.393) Soil problem_p 0.376 0.369 0.409 0.446* 0.426* (SP) (0.256) (0.254) (0.258) (0.260) (0.257) HH head’s years of rice-cultivating -0.024*** -0.023** -0.021** -0.019** -0.018* experience_h (YRCE) (0.009) (0.009) (0.009) (0.010) (0.009) Annual crop frequency_p 1.388*** 1.396*** 1.361*** 1.145*** 1.095*** (CF) (0.288) (0.292) (0.289) (0.407) (0.385) Cost of chemical inputs in rice 2.837* 2.950* 2.758* 3.346** 3.339** production per hectare_p (CCIRP) (1.656) (1.662) (1.657) (1.460) (1.430) Cost of other inputs in rice -0.247 -0.273 -0.248 -0.338 -0.335 production per hectare_p (COIRP) (0.329) (0.328) (0.328) (0.291) (0.285) Village-town closest distance_p -0.010 -0.008 -0.002 -0.000 0.001 (VTD) (0.018) (0.018) (0.017) (0.018) (0.018) Annual average precipitation rate_pv -1.165** -1.152** -1.158** -1.124** -1.130** (AAPR) (0.467) (0.462) (0.461) (0.488) (0.474) Size of household_h -0.067 -0.075 -0.091* -0.093* (SH) (0.057) (0.058) (0.055) (0.054) Land property rights_p 0.059* 0.068** 0.065** (LPR) (0.033) (0.033) (0.032) N 570 570 570 570 570 R2 0.605 0.606 0.608 0.612 0.612 Notes: *, **, **each denotes significance at 10%, 5%, 1% level, respectively. The robust standard errors for OLS estimations are in parentheses. N is total number of plots of all 251 households in the sample. p, h, v, and pv denote data at plot, household, village, and province level, respectively.

The point about increasing irrigation facilities is relevant but the cost of irrigation development does not rest with farmers themselves. As explained in the thesis, the national government plans and construct networks of canals and irrigation systems. However, given rural farmers rely on farming, investment in irrigation sectors would be a sensible, meaningful option. Another point might be about which irrigation type suits farmers most in 71

terms of location of their plots, soil quality, or quality of access. Quality of access refers how much control over water use and how easy can a household have, for instance, over a community water reservoir where households belonging to that water user group must pay water use fee and maintenance fee.

Table 3.8 Flooding and irrigation impacts on per-hectare rice revenue at plot level Dependent variable Rice revenue (in million Riels per hectare) (1) (2) (3) (4) (5) Independent variables OLS Flood risk_p -0.283* -0.280* -0.260* -0.228 -0.265* (2010‒2014) (0.144) (0.145) (0.144) (0.140) (0.142) Reservoir, dyke, canal irrigation_p 0.933*** (RDC) (0.234) River, lake, pond irrigation_p 0.342* (RLP) (0.184) Underground, pumping irrigation_p 0.380 (UP) (0.288) Irrigated plots_p 0.563*** (IP) (0.158) Short-run shocks_h 0.013 0.022 0.017 -0.019 -0.019 (SRS) (agri & nonagri) (0.145) (0.148) (0.147) (0.141) (0.144) Rice quality_p 1.585*** 1.590*** 1.599*** 1.629*** 1.609*** (RQ) (0.555) (0.555) (0.551) (0.551) (0.536) Cultivated land size_p -0.060 -0.062 -0.059 -0.058 -0.058 (CLS) (0.051) (0.051) (0.051) (0.052) (0.052) Soil problem_p -0.036 -0.035 -0.014 0.022 0.002 (SP) (0.133) (0.133) (0.134) (0.135) (0.135) HH head’s years of rice-cultivating -0.016*** -0.016*** -0.015*** -0.013*** -0.013** experience_h (YRCE) (0.005) (0.005) (0.005) (0.005) (0.005) Annual crop frequency_p 1.119*** 1.118*** 1.100*** 0.946*** 0.849*** (CF) (0.143) (0.143) (0.144) (0.162) (0.158) Cost of chemical inputs in rice 1.036 1.014 0.915 1.464 1.464 production per hectare_p (CCIRP) (1.184) (1.184) (1.187) (1.020) (0.947) Cost of other inputs in rice 0.515** 0.521** 0.533** 0.450** 0.451** production per hectare_p (COIRP) (0.237) (0.237) (0.237) (0.202) (0.187) Village-town closest distance_v 0.005 0.004 0.007 0.009 0.010 (VTD) (0.008) (0.008) (0.009) (0.009) (0.009) Annual average precipitation rate_pv -0.077 -0.080 -0.083 -0.057 -0.057 (AAPR) (0.120) (0.122) (0.121) (0.123) (0.122) Size of household_h 0.013 0.009 -0.005 -0.008 (SH) (0.029) (0.029) (0.028) (0.028) Land property rights_h 0.030* 0.037** 0.036** (LPR) (0.018) (0.017) (0.017) N 570 570 570 570 570 R2 0.188 0.188 0.191 0.220 0.213 Notes: *, **, **each denotes significance at 10%, 5%, 1% level, respectively. The robust standard errors for OLS estimations are in parentheses. N is total number of plots of all 251 households in the sample. p, h, v, and pv denote data at plot, household, village, and province level, respectively.

Attempts to minimise an issue of unobserved factors and reverse causality were made. For example, I controlled for soil quality as a proxy for level of wealth and investment in land improvement. In addition, I used an irrigation dummy for plots that were irrigated with any or all the types of irrigation. Using this dummy is the same as averaging the irrigation access per plot, which in fact could minimise a potential issue of simultaneous effects, if any. Importantly, the tests that use irrigation dummy are a robustness check against the estimations that tested individual types of irrigation. A past study by Vollrath (2007) found similar evidence on positive impact of cultivated land size on agricultural productivity. However, high inequality in land holding size could have determined large disparity in agricultural performance of farm households. One might argue for the importance of total cultivate land per household as it could translate into food production and household income. The plot-level findings in Tables 3.5 to 3.8 for the control variables are given below. Land property rights (LPR) have a significant, positive relationship with rice yield per harvest (Table 3.5), per-hectare rice yield (Table 3.6), rice revenue per harvest (Table 3.7), and per- hectare rice revenue (Table 3.8). Cultivated land size (CLS) and annual crop frequency (CF), each have a significant, positive link with rice yield per harvest, rice yield per hectare, rice revenue per harvest, and rice revenue per hectare. The plot-level evidence provided in Tables 3.5 to 3.8 reveals that annual average precipitation rate (AAPR) and HH head’s years of rice-growing experience (YRCE) each appears to be significantly, negatively associated with total rice yield, per-hectare rice yield, total rice revenues, and per-hectare rice revenues. It might be possible that rainfall rates appear to have invariably affected crop yields across the provinces studied and in bad years rainfalls could have been too little or too much such that it could have affected rice production and revenue from rice production. The evidence on YRCE is somewhat counterintuitive in that the longer the years of rice experience would aid famers in raising yields and productivity. However, the inverse relationship can be possible because farmers who have longer experience might have lost motivation in rice cropping or their aging years would force them to work less on the farms. Household heads who are now aging may not be able to work on the farms as productively as before in taking care of their farms, including fertilising, weeding, and irrigating water into their farms, for instance. Therefore, their lower labour productivity could be explaining their lower crop yields and crop revenue.

The evidence also shows that rice quality (RQ) is significantly, positively linked with per-hectare rice revenue at plot level (Table 3.7) and per-hectare rice revenue at plot level (Table 3.8). However, RQ does not statistically significantly affect total latest rice yield at plot level (Table 3.5) and per-hectare rice yield at plot level (Table 3.6). In Tables 3.5 to 3.7, the cost of chemical inputs in rice production (CCIRP) is significantly, positively associated with total rice yield, per-hectare rice yield, and total rice revenue. However, CCIRP does not have a significant relationship with per-hectare rice revenue (Table 3.8). The cost of other inputs in rice production (COIRP) is significantly, negatively linked with total rice yield (Table 3.5), but it is significantly, positively linked with per-hectare rice yield (Table 3.6) and per-hectare rice revenues (Table 3.8). However, COIRP has no significant link with total rice yield (Table 3.7). The rest of the control variables do not have any significant relationship with the tested measures of rice productivity and revenue. The inverse relationship can be possible because farmers who have longer experience might have lost motivation in rice cropping or their aging years would force them to work less on the farms. Household heads who are now aging may not be able to work on the farms as productively as before in taking care of their farms, including fertilising and irrigating water into their farms, for instance. Therefore, their lower labour productivity could be explaining their lower crop yield and crop revenue. Households reported some of the cases of short-run shocks. It might be possible that any instance of short-run shocks was very short and trivial that households were able to overcome relatively easily. Other support from village members or government might influence the way households were copying any shocks and therefore they could manage to overcome without any serious impact on their production capability. In Tables 3.5 to 3.8, agricultural land property rights (LPR) appears to be significantly, positively associated with total rice yield, per-hectare rice yield, total rice revenue, and per-hectare rice revenue. This evidence demonstrates that improvement security of agricultural land property rights could be beneficial for farmers to improve their crop yields and revenue.

3.8 Results: Rice production and food insecurity Tables 3.9 to 3.10 report the household-level estimation results. The OLS results (Columns 1‒4) and the probit results (Columns 5‒8) in Table 3.9 indicate that household rice productivity and household rice revenues are each significantly, negatively associated with household food insecurity. The evidence shows that an improvement in household rice yields from latest harvest (in tonnes), per-hectare household rice yield, total rice revenue from latest harvest, and per-hectare household rice revenue will each have a potential to reduce household food insecurity. These indicators of rice productivity and rice revenues could be key potential channels through which extreme flooding and irrigation affect household food insecurity. Higher rice productivity and greater rice revenues could be achieved through an expanded access to irrigation and a greater ability to control excessive flooding risk. Having a greater access to irrigation and being able to avoid crop damage from extreme flooding in rice production, rice farmers in rural Cambodia could reap the benefits of higher income and enhanced food access to lower the duration, and possibly the severity, of household food insecurity. The findings in this study highlight that the potential gains in rice production and rice revenue would help them improve their food access and possibly nutritional intakes through an increase in food availability. The OLS and probit estimates in Table 3.9 show that among the control variables, only household cultivated land size (CLS) and agricultural land property rights (LPR) are significantly, negatively associated with household food insecurity (FI). The negative association between CLS and FI (in Table 3.9) signifies that as cultivated farm size per household increases, household food insecurity would tend to decline. Further evidence reveals a significant, positive relationship between CLS and total rice yield (Tables 3.5) and between CLS with total rice revenue ((Table 3.8). However, the CLS does not tend to be significantly associated with per-hectare rice yield and per-hectare rice revenue, despite being positive. The negative association between LPR and FI (Table 3.9) reveals that security in land property rights can help lower household food insecurity. The R2 for the OLS results (Table 3.9) are small. It is possible that some factors that might have affected food insecurity are not accounted for because the data on those characteristics are not available.

Table 3.9 Impact of rice productivity and rice revenues on household food insecurity Dependent variable Household food insecurity (FI)

(1) (2) (3) (4) (5) (6) (7) (8) Independent variables OLS probit

Household rice yields from latest -0.007** -0.099*** harvest_h (measured in tonnes) (0.003) (0.033) Total household rice yield_h -0.053*** -0.227*** (measured in tons per hectare) (0.013) (0.065) Household rice revenue from latest -0.009** -0.114*** harvest_h (in million Cambodian riel) (0.004) (0.040) Total household rice revenue_h -0.059*** -0.254*** (in million Cambodian riel) (0.016) (0.079) Land property rights_h -0.018* -0.015 -0.018* -0.016* -0.055** -0.053* -0.057** -0.054** (LPR) (0.009) (0.010) (0.009) (0.010) (0.028) (0.028) (0.028) (0.028) Rice quality_h 0.015 0.016 0.038 0.118 -0.138 -0.109 0.115 0.261 (RQ) (0.232) (0.227) (0.234) (0.235) (0.706) (0.711) (0.709) (0.718) Total cultivated land size_h -0.011 -0.030*** -0.012 -0.029*** -0.051 -0.204*** -0.057 -0.203*** 76 (CLS) (0.009) (0.007) (0.009) (0.007) (0.066) (0.062) (0.066) (0.062) Soil problem_h -0.041 -0.054 -0.040 -0.051 -0.150 -0.208 -0.144 -0.196 (SP) (0.058) (0.057) (0.058) (0.058) (0.205) (0.207) (0.204) (0.206) HH head’s years of rice-cultivating -0.002 -0.002 -0.002 -0.002 -0.008 -0.006 -0.008 -0.006 experience_h (YRCE) (0.002) (0.002) (0.002) (0.002) (0.008) (0.008) (0.008) (0.008) Annual crop frequency_h 0.003 0.045 0.001 0.041 0.131 0.223 0.123 0.205 (CF) (0.057) (0.057) (0.057) (0.057) (0.207) (0.215) (0.207) (0.214) Size of household_h -0.009 -0.007 -0.009 -0.007 -0.029 -0.026 -0.029 -0.027 (SH) (0.012) (0.012) (0.012) (0.012) (0.050) (0.051) (0.050) (0.051) Village-town closest distance_v -0.005 -0.004 -0.005 -0.004 -0.012 -0.010 -0.013 -0.011 (VTD) (0.005) (0.005) (0.005) (0.005) (0.016) (0.017) (0.016) (0.017) Annual average precipitation rate_pv -0.042 -0.037 -0.041 -0.034 -0.098 -0.086 -0.104 -0.086 (AAPR) (0.053) (0.051) (0.053) (0.052) (0.174) (0.176) (0.174) (0.175) N 251 251 251 251 251 251 251 251

R2 0.090 0.123 0.088 0.115 n.a. n.a. n.a. n.a. Notes: *, **, **each denotes significance at 10%, 5%, 1% level, respectively. The robust standard errors for OLS estimations and standard errors for probit estimations are in parentheses, respectively. N is total number of households in the sample. h, v, and pv denote data at household, village, and province level, respectively.

Table 3.10 Impact of rice productivity and rice revenues on length of food insecurity Dependent variable Length of household food insecurity (LFI) (1) (2) (3) (4) Independent variables Poisson

Household rice yield from latest -0.225*** harvest_h (measured in tonnes) (0.036) Total household rice yield_h -0.325*** (measured in tonnes per hectare) (0.052) Household rice revenue from latest -0.270*** harvest_h (in million Cambodian riel) (0.045) Total household rice revenue_h -0.317*** (in million Cambodian riel) (0.061) Land property rights_h -0.013 -0.014 -0.016 -0.019 (LPR) (0.019) (0.018) (0.019) (0.018) Rice quality_h -1.888*** -1.752*** -1.524*** -1.430*** (RQ) (0.477) (0.475) (0.481) (0.491) Total cultivated land size_h -0.158*** -0.423*** -0.173*** -0.422*** (CLS) (0.057) (0.057) (0.057) (0.057) Soil problem_h 0.222 0.145 0.235 0.170 (SP) (0.144) (0.145) (0.144) (0.145) HH head’s years of rice-cultivating 0.015*** 0.017*** 0.015*** 0.017*** experience_h (YRCE) (0.005) (0.005) (0.005) (0.005) Annual crop frequency_h -0.087 -0.061 -0.092 -0.123 (CF) (0.156) (0.157) (0.156) (0.159) Size of household_h 0.075** 0.080*** 0.076*** 0.077** (SH) (0.029) (0.030) (0.030) (0.030) Village-town closest distance_v -0.040*** -0.039*** -0.041*** -0.041*** (VTD) (0.012) (0.012) (0.012) (0.012) Annual average precipitation rate_pv -0.053 -0.035 -0.060 -0.056 (AAPR) (0.114) (0.114) (0.114) (0.114) N 251 251 251 251 Notes: *, **, **each denotes significance at 10%, 5%, 1% level, respectively. The standard errors for Poisson estimations are in parentheses. h, v, and pv denote data at household, village, and province level, respectively.

The Poisson results in Table 3.10 show that all the household-level measures of rice productivity and rice revenues (i.e., household rice yield from latest harvest, total household rice yield, household rice revenue from latest harvest, and total household rice revenue) are significantly, negatively associated with the length of food insecurity. The coefficients are statistically significant at 1 percent level. This household-level evidence reveals that household rice productivity and household rice revenues could possibly be the channels that alleviate rural household food insecurity. Any policy options that would enhance household rice production and household rice revenues would potentially be conducive for lowering the duration of hunger and malnutrition for at least some of the rice farmers in the surveyed villages. Note the results in Tables 3.9 and 3.10 do not show size effects, they only show the potential causal relationship or the direction of effect.

3.9 Conclusions In this chapter, the quantification of impacts of flooding and irrigation on rice productivity and food insecurity in rural Cambodia provides some useful evidence from this understudied, developing economy. The main results show that plots or households affected by severe flooding had lower rice productivity and rice revenues. Severe flooding potentially reduces rice yields per-hectare estimated at plot and household levels and rice revenues per-hectare measured at both plot and household levels. The results are consistent with some existing studies, including Karunasagar and Karunasagar (2016), Gregory et al. (2005), Malla (2010), among others, in that floods decrease crop production directly. Severe flooding can potentially reduce food availability, which affects nutrition consumption and worsens rural food insecurity. This chapter also finds that irrigation, regardless of source, is significantly, positively associated with rice yields and rice revenues. Formal irrigation, whether from reservoir, dyke, or canal (RDC), is significantly, positively associated with all the tested measures of rice productivity and rice revenues.28 River, lake, or pond (RLP) irrigation is significantly, positively linked with per-hectare rice yields at plot level and per-hectare rice revenues at household level, but not with per-hectare rice revenues at plot level and per-hectare rice yields at household level. Underground or piping (UP) irrigation is not significantly linked with the productivity and revenue measures. Although UP irrigation does not appear to be statistically significantly related with rice productivity and rice revenues, it has a positive association. The data show that only a very small fraction of the surveyed households were using this irrigation because it tends to be more costly that all other types. The differences between access to, supply, and availability of the irrigation types are explained in Table 3.4. Finally, all measures of household rice productivity and rice revenues have negative relationships with household food insecurity.29 This chapter has identified two possible channels. First, expanding formal irrigation that ensures irrigation availability and access to improve rice production and revenues in rice agriculture. Developing formal irrigation facilities or expanding irrigation access can provide several benefits, including supplying water to irrigate rice fields and addressing unexpected water shortages during droughts. Irrigation availability and access may enable farmers to improve agricultural land use and rice productivity. It

28 The findings are in line with Rosegrant et al. (2009), Wokker et al. (2014), Kirby et al. (2016). 29 Turner et al. (2004) highlight that improving rice productivity can prevent price rises associated with reduced production and lowered productivity. This may explain food and nutrition insecurity because price is a factor that influences access to food. 78

could enable farmers to innovate or further optimise their rice production and boost food availability. Therefore, irrigation development can have important implications for improving rice production and revenues from rice production, which could lower rural food insecurity. Second, because excessive flooding in rice-growing regions tends to damage rice production and lower rice revenues, managing or controlling excessive flooding risk could minimise the damage. Potentially through these channels could rural food insecurity in Cambodia be gradually reduced. While the chapter did not assess cost-benefit analysis of irrigation interventions, it investigates effectiveness of irrigation access in rice production. Future research that looks at cost-effectiveness could be valuable.30 Additionally, future work that uses panel data or a larger cross-section of households, when resources and data are available, could be beneficial.

30 Cost-effectiveness in irrigation sector is required for irrigation to contribute positively to food production (Evans & Sadler, 2008) and maybe to rural food security. Smith and Hornbuckle (2013) stress that improving efficiency in irrigation in rice farming would improve rural food and nutrition security. But irrigation efficiency is not explored in this thesis. 79

CHAPTER 4

PROPERTY RIGHTS AND FOOD INSECURITY IN DEVELOPING ECONOMIES

4.1 Introduction Food insecurity in developing countries is critically high and challenging (West et al., 2014). Currently, about 800 million people in the world are undernourished, of whom 780 million were in developing countries during 2014‒16 (FAO, 2015). About one in nine people is persistently and chronically food insecure (FAO, 2014). The global number of undernourished people has fallen by 216 million since 1990‒92, a decline of 21.4 percent. Over this period, the share of undernourished in developing countries fell from 23.3 percent to 12.9 percent. The number of regional undernourished in Sub- Saharan Africa, on the contrary, rose from 176 million in 1990‒92 to 220 million in 2014‒ 16, and the regional share of undernourished people jumped from 17.4 to 27.7 percent. In South Asia, the level of hunger prevalence declined only slightly from 291 million to 281 million over the same period. Its regional share increased from 28.8 to 35.4 percent. Both South Asia and Africa remained strongly exposed to micronutrient deficiency, which affects two billion people globally every year (FAO, 2013).31 The food price increases during the 2007−2008 was a major phenomenon of the response to staple food trade restrictions, which has led to food price crisis across the global (Anderson et al., 2013; Barrett, 2013). The sustained food price crisis could adversely exacerbate global food insecurity enormously, affecting particularly the low-income, net food importing countries. Being food insecure means having an insufficient intake of nutrition to remain healthy. Severity of food insecurity imposes a multitude of social and economic consequences. Insufficient consumption of nutrition or poor diets over a prolonged period can possibly have adverse effects on health, life expectancy, labour productivity, and child ability to stay in school and income-earning capabilities. Existing evidence shows that hunger and malnutrition have led to high death rates in developing countries. Every

31 According to the Global Hunger Index (2016), only 72 out of 129 developing countries reached the Millennium Development Goal (MDG) hunger target in 2015. 80

year global undernutrition causes almost 45 percent of child under-five deaths (Unicef, 2014). This amounts to 3.1 million deaths, of which 1.1 million deaths were caused by micronutrient deficiency (IFPRI, 2014). Similarly, every year about 2.8 million children and 300,000 women in developing countries die of malnutrition (Guha-Khasnobis et al., 2007). Weak, poor, or undeveloped property rights mean the rights to own, use, make change, transfer, exclude, develop or trade properties, including land, are limited or restricted by insufficient legal recognition and protection of the rights. It appears that property rights in developing countries, in which large pockets of the world food insecure reside, have not improved adequately. Existing evidence suggests that weak or insecure property rights could possibly exist in many developing countries, including both land- poor and land-rich areas of the world, such as Southern and Eastern Africa, South and Southeast Asia, and Latin America (Barrett, 2013). Where weak or restricted land property rights prevail, those countries or regions tend to be exposed to a high amount of land grabbing. In Ethiopia, the state had once nationalised all land in the country. Currently, the state owns the land and only grants leaseholds to farmers. There is evidence that shows that agricultural productivity and farm investment in Ethiopia have been limited by land tenure insecurity (Deininger & Jin, 2006; Barrett & Upton, 2013, p.326). Low agricultural productivity and low agricultural investment may have contributed to low quality of life, including undernourishment, for many people in Ethiopia. Figure 4.1 shows that property rights protection (in logarithm) is negatively correlated with prevalence of undernourishment (in logarithm), using cross-sectional data for 57 developing countries in 2011. In Guha-Khasnobis et al. (2007), undernourished people normally suffer one or more forms of malnutrition due to taking insufficient caloric intake, consuming poor-quality diets, having protein deficiency, and consuming insufficient concentration of proteins and micronutrients. For simplicity, only some names of the 57 countries are shown. On the Figure, rightward on the horizontal axis indicates a higher level of property rights protection. Upward on the vertical axis indicates a higher level of undernourishment prevalence. The negative relationship implies that in general the economies with lower or weaker property rights appear to have higher prevalence of undernourishment, at least in the year 2011 among the developing countries in the sample.

Prevalence of Undernourishment and Property Rights

Burundi

Mozambique Swaziland Tanzania Uganda Namibia SierraGuatemala Leone Rwanda Kenya Botswana Bolivia Mongolia Sri Lanka Paraguay MalawiSenegal

3 EcuadorPakistanNepal CambodiaPhilippinesIndia Lesotho Bangladesh Gambia Peru ElColombia Salvador China Honduras Indonesia Panama Mauritania

2 ThailandBelize CostaBrazil Rica GabonUruguay Morocco EgyptMexico Turkey JordanMalaysiaSouth Africa Prevalence of Undernourishment (log) of Undernourishment Prevalence Chile linear fit 95% CI

1 1.5 2 Property rights protection (log)

Figure 4.1 Log of property rights protection against log of prevalence of undernourishment in 57 developing countries, 2011

Source: Adapted from FAO’s 2014 Food Security dataset and Fraser Institute’s 2014 Economic Freedom of the World report

Impacts of property rights, including agricultural land property rights, on food insecurity have not been previously analysed quantitatively.32 This chapter examines this linkage, using data from a large cross-section of developing economies. In testing specification (4.1), the chapter uses (i) the ordinary least squares (OLS) method and (ii) the between estimator (BE) method for pooled panel data over 1990 to 2011. The chapter first regresses the prevalence of undernourishment (PU), prevalence of food inadequacy (PFI), depth of food deficit (DFD), and average dietary energy supply adequacy (ADESA) on a measure of private property rights security (PRP). The key estimation

32 Existing studies on property rights have examined other impacts rather than food security. For example, Markussen (2008) and Deininger and Jin (2006) analysed impacts of property rights on agricultural production and productivity; Carter and Olinto (2003), Belsey (1995), Deininger and Jin (2006): property rights on investment; Collier and Hoeffler (1998, 2002 & 2004) and Lawson-Remer (2014): property rights on onsets of civil wars; World Bank (2003): land property rights on poverty and growth. Existing studies on food security have examined different drivers. For example, Timmer (2000), Myers (2006), and Brinkman et al. (2010): food security and economic growth; Timmer (2005): food security-poverty linkages; Timmer (2000) and Myers (2006): food security and food prices; and Parry et al. (1999): climate change and food security; Smith, Obeid and Jensen (2000) and Timmer (2005): geography and causes of food insecurity; Anderson (in CB Barrett 2013) and Warr (2014): prices and food insecurity. 82

results from the OLS and BE methods indicate that security in private property rights (PRP) is significantly associated with reduced food insecurity and with improved food security, respectively. The between estimator (BE) results show that a one-percent increase in security in private property rights (PRP) potentially reduce prevalence of undernourishment (PU) and prevalence of food inadequacy (PFI) by about 0.85 percent and 0.64 percent on average, respectively. Next, the chapter uses the OLS method to test six other different measures of property rights in different time periods, for when the data on those property rights measures were available, with the four measures of food (in)security. However, since the tests provide similar results, only the results on PU are reported. The results from this robustness check show that strong property rights, such as property rights (PR), the property rights and rule-based governance (PRRG) rating, and the international property rights index (IPRI) are significantly, negatively associated with PU. The physical property rights score (PPRS) is insignificantly, positively associated with PU (Table 4.8: Column 3). In contrast, the two measures of property rights, namely (i) regulatory costs or restrictions of sale of real property and (ii) property registration, are significantly, positively associated with PU. This suggests that longer steps, complicated procedures, and higher costs discourage formal land registration, which could prompt weak property rights and worsen food insecurity outcomes in developing economies. The chapter proceeds as follows. Section 4.2 reviews the key literature. Section 4.3 describes the basic models. Section 4.4 explains the empirical specifications, methods, and empirical strategy. Section 4.5 explains the data, providing key definitions of the variables and how they are calculated. Section 4.6 discusses key results. Section 4.7 concludes with suggestions for future research.

4.2 Literature review Different systems of land property rights are practised in developing countries based on their varying legal traditions (La Porta et al., 2008; Levine, 2005; Posner, 1998).33 However, allocation and enforcement of property rights are generally influenced by the political regime that administers the property rights (Levi, 1988; Libecap, 1989; La Croix & Roumasset, 1990; Ensminger, 1992; Knight, 1992; Firmin-Sellers, 1996; Alston et al., 2009). Unstable property rights systems have been affected by rent-seeking

33 These include indigenous, customary, and modern land property rights systems. 83

elites or the military (Besteman, 1990 in Lawry et al., 2016). Weak property rights mean the rights associated with ownership, use, and transfer of property are inadequate and fragile. Their weak land property rights have resulted from weak legal and judicial systems and complicated land regulations and practices (La Porta et al., 2004).34 In general, only a small fraction of land being held or controlled by people in the world has received official, legal recognition. For example, de Soto (2000) estimates that by 2000 about 80 percent of the poor in developing economies were not able to collateralise their land to obtain credit because it has not been officially registered. The Rights and Resources Initiative (2015) reported that by 2015 only 18 percent of the land area in 64 developed and developing countries studied is officially recognised as legally owned or controlled by local communities and indigenous people. The most recent estimate shows that by 2017 about 70 percent of land rights are registered or titled worldwide (World Bank, 2017). Undesired consequences of insecure property rights include land acquisitions by the state or powerful private individuals and forced evictions (Allen, 1982). Literature appears to show that most large-scale land acquisitions that mostly involved involuntary dispossessions have occurred in areas or regions where property rights are weak.35 They tend to involve expropriations of cropland owned or occupied by farmers and the poor. It has been observed that insecurity in private property rights in most developing countries prevails where property rights institutions are dysfunctional or underdeveloped. Land property rights could alter economic and social outcomes (Lipton, 2009). While undue risks of property expropriation might have prevented economies from performing efficiently (Acemoglu, 2009; Acemoglu et al., 2004), high security in land property rights appears to have had positive impact on investment and economic growth (Carter & Olinto, 2003; Rodrik et al., 2004; Smith, 1976; North, 1990). This possibly happens when investors perceive reduced likelihood of losing their land and when there is growth in collateral-based credit markets (de Soto, 2000; Markussen, 2008). Existing studies tend to suggest that insecurity in formal and informal property rights could have affected agricultural production, natural resource, food security, and the economic

34 A disruptive political regime or a radical change of political system can be causes of weak property rights (Seguchi & Hatsukano, 2013; Unruh & Williams, 2013). 35 That include countries in Africa, Latin America, South and Southeast Asia. Barrett (2013) reported that land acquisitions in Africa by foreign investors totalled 40 million hectares in 2009, and about 1.3 million hectares of land were acquired by a foreign firm in 2008 in Madagascar, which affected almost one-third of the country’s arable land (Barrett, 2013). There must be reason to believe that some acquisitions could have affected the agricultural land of the farmers in the region. As it has been documented that many rapid, large-scale land acquisitions in China have displaced many rural Chinese farmers, 84

conditions of the poor in developing counties (Atwood, 1990; Lawry et al., 2016; Deininger & Feder, 2009). Evidence on impacts of land property rights on investment, productivity and credit access remains mixed and contested. In some studies, land property rights were found to positively affect productivity (Markussen, 2008; Goldstein & Udry, 2008; Feder, 1987; Holden et al., 2009; Hayes et al., 1997; Deininger & Jin, 2006; Ali et al., 2011). In other studies, land rights security did not impact rice productivity (Deininger & Ali, 2008; Chankrajang, 2015; Gavian & Ehui, 1999; Place & Otsuka, 2002; Brasselle et al., 2002). Other existing evidence seems to suggest that land tenure insecurity does not negatively affect farm input use (Holden & Yohannes, 2002), or it does not positively affect investment (Sjaastad & Bromley, 1997). Restricted land rights were not found to affect income but households with restricted rights had to use more labour to produce the same output (Markusssen et al., 2011). Studies by Atwood (1990), Besley (1995), Field and Torero (2006), de Soto (2000), Piza et al. (2016), Holden et al. (2009), and Deininger and Castagnini (2006) have either found or argued that secure property rights encourage credit access. However, Lawry et al. (2016) and Markussen (2008) did not find evidence of a positive impact of land property rights on credit access. Other evidence tends to show positive credit access effects of land rights security were conditional on an existence of a well-established, properly-functioning rural credit market (Carto & Olinto, 2003; Deininger & Feder, 2009; Brasselle et al., 2002). 36 Insecurity in property rights has been reported as driving food insecurity in developing economies. It can possibly be a driving force that prompts social conflicts or civil wars (Collier & Hoeffler, 1998, 2002, 2004), potentially disrupting food production, food distribution capabilities, and the functioning of markets (Barrett, 2013, p.12). From observing the literature on land property rights, it can be concluded that few studies have focused their analysis on the relations between land property right and food insecurity. Maxwell and Wiebe (1999), Newman et al. (2015), Lawry et al. (2016), and World Bank (2003) are exceptions. However, none of these studies examined the link quantitatively.

36 For example, Markussen (2008) found that formal land titling in Cambodia had positively affected agricultural production and productivity among farm households. 85

4.3 Basic model Figure 4.2 depicts some key factors that are assumed to influence food production and food insecurity. Among these factors, property rights is proposed as a potential driver which may influence food production in agriculture and food security. Other factors, such as prices and inflation rates, can also influence food security (Timmer, 2005; Brinkman et al., 2010). However, the focus of this diagram is to explain how property rights can transcend potential impacts on national food production and food security in developing countries.

Institutions Property Rights: Protection of property rights & enforcement of Human capital rule of laws and contracts Trade Educational attainment Openness and & return to education access to international Population markets

Geography Agricultural land & Climate National food production Latitude, a measure associated Ethnolinguistic with fractionalisation climate and rainfall Prices & Agriculture Inflation share of GDP Purchasing power Per capita household consumption Food insecurity expenditure (i) Utilisation or consumption of food, (ii) food access, (iii) availability, (iv) stability

- Prevalence of undernourishment - Prevalence of food inadequacy - Depth of food deficit - Average dietary energy supply adequacy

Figure 4.2 Conceptual framework for factors influencing food insecurity

Source: Author

This chapter tests the following empirical specification:

퐹퐼푖푡 푌푒푎푟 푅푒𝑔푖표푛 } = ∝ + 훽. 푃푅푖푡 + 훾 . 푿푖푡 + 훿. 퐷푡 ∗ 퐷푗 + 휀푖푡 (4.1) 퐹푆푖푡

where:

- 퐹퐼푖푡 is log of food insecurity. Three indicators of FI are used (prevalence of undernourishment, prevalence of food inadequacy, and depth of food deficit). FS stands for food security. The indicator for food security is average dietary energy supply adequacy (ADESA). Because of suspected multicollinearity and reverse causality between economic growth and food security, economic growth is not controlled for. Gross domestic product per capita was not controlled either for the same reason. To overcome this, I instead used household final consumption expenditure per capita (HHFCEPC). This would be a better control to capture its potential relational effect on food insecurity outcome.

- 푃푅푖푡 is log of private property rights. 푃푅푖푡 consists of seven measures of property rights. The main variable of interest is property rights protection (PRP). The other six measures of property rights are property rights (PR), physical property rights score (PPRS), property rights and rule-based governance (PRRG), regulatory restrictions on sale of real property (RRRP), registering property (REGPR), and international property rights index (IPRI), each of which is tested separately to check robustness.

- 푿푖푡 is the set of control variables. Subscripts i, j and t refer to country, region, and time, respectively. The control variables include log human capital index (HC), log of consumer price index as a measure of inflation (INFL), log of population (POP), log of per capita household final consumption expenditure (HHFCEPC), ethnolinguistic fractionalisation (ETHNO), latitude (LATIT), trade openness (OPEN), percent of agricultural land to total land area (AGRILD), and agriculture share of GDP (AGCOM). The index of human capital (HC) measures the average years of schooling. Latitude is used as a proxy for geography. It represents factors such as climate and rainfall that could affect agriculture and crop production. The agricultural land (AGRILD) can capture factor endowments specific to an agrarian economy. The consumer price index (CPI) is used as a proxy for the inflation rate. The data used in this chapter, including the CPI, are estimated based on 2005 price levels.

푅푒𝑔푖표푛 푌푒푎푟 - 퐷푗 is a regional dummy, and 퐷푡 is a time or year dummy. - ∝ is a constant. 훾 and 훿 are parameters, and 훽 is the parameter of interest to estimate.

- 휀푖푡 is the error term.

To reduce possible skewness in residuals to obtain residuals that are approximately symmetrically distributed, I use the logarithms for the dependent and for some of the independent variables. In so doing, it could help linearise the relationships for the main variables of interest. Additionally, logging those variables may provide convenience in interpreting the results. The independent variables in logarithm include human capital, inflation, population, per capital household final consumption expenditure. Other independent variables, including ethnolinguistic fractionalisation, latitude, percent of agricultural land to total land area, and agriculture share of GDP are not logged, however. The inclusion of regional dummies and time dummies can partially capture some specific effects, while the rest is in the residual. The chapter attempts to minimise omitted variable bias (OVB) by including relevant controls.

4.4 Measurement of property rights The chapter uses protection of property rights (PRP) indicator as the main variable of interest. For ease of interpretation for its significant or insignificant relationship with food (in)security, PRP is referred to in this chapter as security in private property rights in the discussion of the results. The chapter testes six other measures of property rights to gauge if a relationship between property rights and food (in)security exists when property rights indicators are measured differently. The purpose of testing these different indicators of property rights is to assess whether and how different property rights measures may influence food (in)security in the developing economies. Note also that the data on the seven indicators of property rights are not available for all countries and some of them are only available from 1996, for instance. Therefore, the number of observations varies depending on the availability of data (Table 4.2). The definitions and the differences between the property rights measures are explained as follows. First, according to the 2014 Economic Freedom of the World (EFW) Report, property rights protection (PRP) was adapted from the Global Competitiveness Report (GCR) which originally used a 1‒7 scale. In the 2014 EFW Report, the PRP was converted from the original scale to a zero (lowest) to 10 (highest) scale. The revised formula becomes: EFWi = ((GCRi − 1) ÷ 6) × 10. The original GCR statement to solicit information on property rights is: ‘Property rights, including over financial assets, are poorly defined and not protected by law (=1) or are clearly defined and well protected by law (=7).’

Second, the property rights (PR) measure is like the protection of property rights measure. The PR data are scaled from 0 (lowest) to 100 (highest). However, the PR data are only available from 1995 or 1996 onwards. Third, the physical property rights score (PPRS) is a measure which encompasses three elements: (i) the protection of physical property rights, (ii) property registration, and (iii) access to loans. The score is ranked on a scale from 0 (lowest) to 10 (highest). Fourth, the property rights and rule-based governance (PRRG) measure assesses the extent to which private economic activity is facilitated by an effective legal system and rule-based governance structure in which property and contract rights are reliably complied with and enforced. The data on property rights and rule-based governance range from 1 (lowest) to 6 (highest). Fifth, regulatory restrictions on sale of real property (RRRP) data measures the freedom in property rights transactions in land markets. According to the World Bank’s Doing Business report (2015) and the Fraser Institute (2014), RRRP data are measured on a zero (worst) to 10 (best) scale based on (1) the time cost, measured in terms of the number of days required to transfer ownership over land and a warehouse, and (2) the monetary cost of transferring the ownership, measured as a percentage of the property value. The two subcomponents are then averaged to derive the RRRP score. Note that countries high on PPR can be low on the RRRP ranking, and vice versa. Sixth, the registering property (REGPR) indicator measures the time, cost, steps, and procedures required to completely register a property, or the ease in transferring the property title and using it as collateral to access credit or to expand a business (World Bank, 2015). The REGPR data are on a scale from zero (worst) to 100 (best) scale. Seventh, the international property rights index (IPRI) is a composite variable that consists of physical property rights, intellectual property rights, and the legal and political environment. A higher index score signifies better property rights.

4.5 Measuring food (in)security indicators Food insecurity is measured at four different levels: global, national, household, and individual. The four indicators used in this chapter are measured at the national level, and they are taken from the Food and Agriculture Organization (FAO) of the United Nations. Food security ‘exists when all people, at all times, have physical, social and economic access to sufficient, safe and nutritious food to meet their dietary needs and 89

food preferences for an active and healthy life’ (World Food Summit: 1996; FAO, 2014). Food insecurity applies when this criterion is unmet, implying hunger and undernutrition prevail where people have less than the daily dietary energy requirement to consume to remain healthy and economically active. Food is necessary for life, as food does not have direct substitutes (Warr, 2014). Persistent or chronic food insecurity means lacking access to sufficient and healthy food and diets or consumption of insufficient nutrition and unhealthy diets over a prolonged period. Transient food insecurity, in contrast, could happen over multiple short periods of time within a course of one calendar year. The main food insecurity variable used in this chapter is prevalence of undernourishment (PU). Two other measures of food insecurity, namely prevalence of food inadequacy (PFI) and depth of food deficit (DFD), are also tested. The chapter tests one measure of food security, namely average dietary energy supply adequacy (ADESA). In the proceeding section, there are explanations of food (in)security variables as measured by the Food and Agriculture Organization of the United Nations (FAO). The data are taken from FAO’s 2014 Food Security Index. The FAO uses the following methods of calculating food insecurity indicators based on the FAO’s 2011 Food Balance Sheets handbook. The purpose of testing these four food insecurity indicators is to check whether different measures of food insecurity are robust in the same specification as with the baseline results. The first indicator of food insecurity is prevalence of undernourishment (PU). It measures the number of undernourished or people at risk of undernourishment. This measure reflects the share of the population with insufficient caloric intake (IFRI, 2016). The estimated PU multiplied by the size of population gives the measure of PU. In its simple form, the FAO explains that PU is expressed in the probability (percentage) an individual selected from the population consumes an insufficient quantity of calories to meet her or his energy requirement to live an active and healthy life.37 The second indicator of food insecurity is prevalence of food inadequacy (PFI). It measures the percentage of population that is unable to cover sufficient food requirements to perform normal physical activity. Therefore, it also tends to include those who are likely being conditioned in their economic activity by insufficient food, although they

37 The prevalence of undernourishment (PU) is calculated by comparison of a probability distribution of habitual daily Dietary Energy Consumption (DEC), f(x), with a threshold level, called the ‘Minimum Dietary Energy Requirement’ (MDER), to capture the estimate for an average individual in the reference population. FAO uses the following formula to estimate PU. ( ) pu = ∫x

cannot be considered chronically undernourished. According to the FAO (2011), a higher PFI constitutes a higher intensity of food insecurity. The estimated PFI is derived from aggregating the food inadequate in all regions of a country and then dividing that by population to have country-level PFI data. It is calculated on a three-year average.38 The third indicator of food insecurity is depth of food deficit (DFD). It measures kilocalories per capita per day and estimates how many calories are required to remove the undernourished from the state of undernourishment. Specifically, it measures the average intensity of food deprivation of the undernourished and is expressed as the gap between average dietary energy requirement (ADER) and average dietary energy consumption of the undernourished. The estimated DEF is then multiplied by the number of undernourished to derive an estimate of the total food deficit for each country. It is an estimate for a three-year average, weighted by population. The fourth, average dietary energy supply adequacy (ADESA), is the indicator of food security. ADESA is expressed as a percentage of ADER. The average supply of calories for food consumption is normalised by ADER, and then weighted by population average to provide an estimate of calorie-based adequacy of the food supply. The indicator is calculated as a three-year average for each year from 1990‒92 to 2014‒16, measuring structural food supply adequacy. According to the FAO (2011), this method reduces some possible errors in estimated DES, which may arise because of the difficulty in accounting for stock variations in major food items. The FAO food insecurity data are based on the FAO Food Balance Sheet (2011), constructing three-year averages using data from national household surveys. The FAO acknowledges that there are advantages and disadvantages in constructing the food (in)security data. Refer to the data measurement sections in FAO Food Balance Sheets for further explanations on the measurement methods, advantages and disadvantages.39

4.6 Other data

38 The prevalence of food inadequacy (PFI) is like PU, but its estimate is derived from using a higher physical activity level (PAL) coefficient: 1.75 versus 1.55 for PU (FAO, 2014). PU measures chronic food deprivation or hunger, but PFI is a less conservative measure of food inadequacy. 39 Past per capita caloric consumption statistics by household income are classified in the FAO’s 2011 Food Balance Sheets Handbook. FAO estimates an individual’s CV of food consumption, then calculates variation in average caloric consumption between income classes. The CV of an individual’s habitual food consumption is CV(x) = √(CV|y)2 + (CV|r)2 , where (CV|r) captures some variation induced by factors that cause variability in food consumption and are not correlated to income (y). 91

The other control variables used in the estimated equations are explained as follows. - AGCOM measures agriculture share of gross domestic product (GDP). - Consumer price index used to capture inflation (INFL) measures fluctuations in prices of consumer goods. - Per capita household final consumption expenditure (HHFCEPC) measures final expenditure on consumption made by households at a per capita level. - Population (POP) is used to control for its impact on food insecurity. - Percent of agricultural land (AGRILD) measures the percentage of arable agricultural land over total land area. It represents factor endowments specific to agriculture. According to the World Development Indicators 2014, agricultural land includes all arable land area under permanent crops and pastures. - Openness (OPEN) index is ratio of volume of international trade to GDP. - Human capital (HC) index per person measures years of schooling taken from the Penn World Table (PWT) 8.0. See detail in Feenstra et al. (2013 & 2015). - The latitude variable (LATIT) is the absolute value of the latitude of each country in the sample, measured on a scale between zero and one. It represents geography, which is normally associated with soil quality or land use and crop production patterns. For instance, temperate zones have more conducive climates, better rainfall patterns, and therefore higher crop productivity.40 Engerman and Sokoloff (1997) use latitude to control for factors that are assumed to contribute to better institutions, performance, and socioeconomic outcomes for the people. - Ethnolinguistic fractionalisation represents friction and social conflicts caused by expropriation of properties by the ethnic majority that holds power or those who have access to individuals with power.

40 The data on latitude is taken from La Porta et al. (1999), which follows the calculation method used n World Factbook by the Central Intelligence Agency (CIA 1996). 92

Table 4.1 describes data sources and lists the dependent and explanatory variables used in the tested equations.

Table 4.1 List of variables and data sources for developing economies Acronym Variable name Data source ADESA Average dietary energy supply FAO Food Security Index adequacy 2015 AGCOM Agriculture share of GDP World Bank World Development Indicators 2014 AGRILD Percent of agricultural land World Bank World Development Indicators 2014 DFD Depth of food deficit FAO Food Security Index 2015 ETHNO Ethnolinguistic fractionalisation La Porta et al. (1999) HC Human capital index Penn World Table PTW8.0

HHFCEPC Household final consumption World Bank World expenditure per capita Development Indicators 2014 INFL Inflation World Bank World Development Indicators 2014 IPRI International property rights index International Property Rights Alliance (2015) LATIT Latitude La Porta et al. (1999)

OPEN Openness Penn World Table (PWT) 8.0

PFI Prevalence of undernourishment FAO Food Security Index 2015 POP Population World Bank World Development Indicators 2014 PR Property rights Heritage Foundation

PRP Property rights protection The Fraser Institute’s 2014 Economic Freedom of the World Report PPRS Physical property rights score Property Rights Alliance

PRRG Property rights and rule-based World Bank World governance Development Indicators 2014 PU Prevalence of undernourishment FAO Food Security Index 2015 REGPR Registering property Doing Business Reports 2015 (World Bank, 2015c) RRRP Regulatory costs or restrictions on sale The Fraser Institute’s 2014 of real property Economic Freedom of the World Report

Table 4.2 provides summary statistics of the variables used in the estimation equations. The number of countries in observations and the time periods vary depending on the availability of data for each of the seven tested variables of property rights. Some data of the variables cover only from 1996, rather from 1991 and they do not cover all countries in observations. However, for robustness check against the main variable of interest, I used the data as much as they became available.

Table 4.2 Descriptive statistics

Variable name Unit N Mean St. Dev. Min. Max. Property rights protection Range 0‒10 545 4.70 1.44 1.17 8.37 (PRP) Property rights (PR) Range 0‒100 469 42.86 16.28 10 90 Physical property rights score Range 0‒10 178 5.61 0.96 2.50 7.50 (PPRS) Property rights and rule-based Range 1‒6 136 3.07 0.39 2.50 3.50 governance (PRRG) Regulatory costs or restrictions on Range 0‒10 370 7.05 1.65 1.35 9.60 sale of real property (RRRP) Registering property (REGPR) Range 0‒100 299 62.53 13.54 22.77 94.54 International property rights index Range 0‒10 178 4.76 0.90 2.20 6.90 (IPRI) Prevalence of As % of population 545 17.99 11.70 3.90 69.70 undernourishment (PU) Prevalence of food inadequacy As % of population 545 25.38 13.97 3.70 78.30 (PFI) Depth of food deficit (DFD) Kilocalories/person/ 545 122.36 95.01 1 605 day Average dietary energy supply As % of Average 545 2569.16 391 1630 3720 adequacy (ADESA) Dietary Energy Requirement Per capita HH final consumption In USD 545 1638 1380.50 120.6 1026 expenditure (HHFCEPC) 0 6.19 Population (POP) In million 545 104 267 0.274 1340 Percent of agricultural land In % of land area 545 37.76 23.36 0.45 84.90 (AGRILD) per country Agriculture share of GDP In % of GDP 545 0.14 0.09 0.02 0.52 (AGCOM) Openness (OPEN) Trade as a % of 545 74.10 38.56 14.93 220.4 GDP 1 Human capital (HC) Average schooling 545 2.23 0.43 1.13 3.16 years Inflation (INFL) CPI-based index. 545 100.26 36.51 0.001 205.3 2005 = 100 4 Latitude (LATIT) Range 0‒1 545 0.19 0.12 0.01 0.51 Ethnolinguistic fractionalisation Range 0‒1 545 0.39 0.29 0 0.89 (ETHNO) Notes: Data for the six additional property rights measures other than property rights protection (PRP) and all other control variables are not available from as early as 1990. Data for some variables are only available from after 2005. I use the latitude data in La Porta et al. (1999), which followed the calculation method in the CIA World Factbook, i.e., “latitude of capital of a country divided by 90”. La Portal then scaled it to take values between zero and one.

Table 4.3 provides the correlations of the variables and their significance level.

Table 4.3 Correlation of variables

PRP PR PPRS PRRG RRRP REGPR IPRI HHFCEPC INFL POP AGRILD AGCOM OPEN HC LATIT ETHNO PU PFI DFD ADESA PRP 1

PR 0.684a 1

PPRS 0.453a 0.577a 1

PRRG 0.678a 0.542a 0.154 1

RRRP -0.020 0.117 0.396a -0.019 1

REGPR 0.003 0.123 0.492a -0.123 0.856a 1

IPRI 0.602a 0.624a 0.780a 0.545a 0.365b 0.397a 1

HHFCEPC 0.149 -0.001 0.217 0.036 0.097 0.205 0.112 1

INFL -0.078 -0.014 0.388a -0.168 0.311b 0.163 0.169 -0.070 1

POP 0.401a 0.53a 0.447a 0.289c -0.039 -0.054 0.416a 0.028 -0.058 1

95 AGRILD -0.014 0.006 0.145 -0.109 0.039 0.175 -0.048 0.154 -0.164 0.431a 1

AGCOM -0.101 -0.060 -0.045 -0.130 0.108 0.221 0.039 -0.720a 0.009 -0.275c -0.194 1

OPEN -0.113 -0.175 -0.054 0.073 0.295c 0.319b 0.125 0.509a -0.188 -0.258c 0.022 -0.169 1

HC 0.004 -0.048 0.018 -0.028 0.076 0.014 -0.076 0.687a 0.089 -0.085 -0.171 -0.757a 0.113 1

LATIT -0.148 -0.077 0.100 -0.213 -0.212 0.084 -0.169 0.132 -0.174 0.259c 0.672a -0.086 -0.065 -0.148 1

ETHNO 0.172 0.277c 0.194 0.225 0.188 -0.001 0.373a -0.284c 0.037 0.188 -0.223 0.012 -0.181 -0.095 -0.204 1

PU -0.390a -0.127 -0.221 -0.173 0.139 -0.121 -0.213 -0.492a 0.267c -0.163 -0.305c 0.313b -0.249c -0.228 -0.502a 0.028 1

PFI -0.392a -0.110 -0.209 -0.196 0.125 -0.143 -0.223 -0.486a 0.276c -0.111 -0.296c 0.257c -0.287c -0.176 -0.496a 0.053 0.993a 1

DFD -0.384b -0.118 -0.186 -0.209 0.094 -0.124 -0.228 -0.431a 0.281c -0.134 -0.280c 0.303c -0.273c -0.208 -0.439a -0.088 0.985a 0.975a 1

ADESA 0.348b 0.046 0.206 0.190 -0.062 0.212 0.226 0.594a -0.243c 0.0262 0.260c -0.239c 0.389a 0.189 0.459a -0.209 -0.930a -0.958a -0.885a 1

Notes: a significant at 1% level; b significant at 5% level; c significant at 10% level. The full list of variable names is explained in Table 4.2.

Table 4.4 lists the 57 developing economies in the sample. 18 countries are from Latin America and the Caribbean, 5 from North Africa and the Middle East, 22 from Sub- Saharan Africa, 7 from East and Southeast Asia, and 5 from South Asia. The classification of these countries is based on the country classification method of the United Nations World Economic Situation and Prospects 2015.

Table 4.4 List of 57 developing countries in the sample Latin America & the North Africa & the South Asia

Caribbean Middle East 1 Barbados 19 Egypt 36 Mozambique 53 Bangladesh 2 Belize 20 Jordan 37 Namibia 54 India 3 Bolivia 21 Morocco 38 Rwanda 55 Nepal 4 Brazil 22 Tunisia 39 Senegal 56 Pakistan 5 Chile 23 Turkey 40 Sierra Leone 57 Sri Lanka 6 Colombia Sub-Saharan Africa 41 South Africa 7 Costa Rica 24 Benin 42 Swaziland 8 Dominican Republic 25 Botswana 43 Tanzania 9 Ecuador 26 Burundi 44 Togo 10 El Salvador 27 Cameroon 45 Uganda 11 Guatemala 28 Republic of Congo East & Southeast Asia 12 Honduras 29 Gabon 46 Cambodia 13 Jamaica 30 Gambia 47 China 14 Mexico 31 Kenya 58 Indonesia

15 Panama 32 Lesotho 59 Malaysia 16 Paraguay 33 Malawi 50 Mongolia 17 Peru 34 Mali 51 Philippines 18 Uruguay 35 Mauritania 52 Thailand

The pooling of the developing countries in the sample is to benefit from the law of the large number. The pooling of cross-countries observations into multiple years also benefits the cross-country observations to make use of variation in country’s specific characteristics to capture potential impact of private property rights on food insecurity outcome. The analysis of data by regions would add additional useful insight. However, instead of separating the data by regions and running regressions by regions, I used region fixed effects in the pooled panel data setting. I grouped the developing countries by regions according to their inherent geographical locations and developmental level. This grouping method has been used widely (UNTAD, UNDP and World bank; for instance, http://unctadstat.unctad.org/EN/Classifications.html). I calibrated the classification into regional dummies, and thus each group of regional countries are similar in terms of both geography and GDP level.

A comparison between ‘South’ and ‘East and Southeast’ Asia would be interesting. However, since I already grouped the countries and used their respective dummies, their estimation results also provide interesting results from which we could compare and draw some useful insights.

4.7 Estimation strategy This chapter uses ordinary least squares (OLS) and the between estimator (BE) methods. The OLS regression method is used as the baseline. The BE approach is used for the pooled panel data to examine long-term effects of variation in property rights and food security between developing countries. Using the BE method provides longer-run effects and possibly corrects some short-run measurement error that may be present in the data on property rights and food insecurity because the BE approach uses data averaging over the estimated period or the mean of each series for each of the countries in the sample to exploit between variation (Stern, 2010; Burke & Yang, 2016; Hauk & Wacziarg, 2009; Pesaran & Smith, 1995; Baltagi, 2005). The between estimator uses average data for each country and provides estimates of long-run effects (Baltagi 2005; Burke & Yang, 2016; Stern 2010). The BE performs best despite the extent of bias on each of the estimated coefficients (Hauk & Wacziarg, 2009). In the presence of measurement error because the year-to-year variations in the measures of property rights are not accurately measured, the use of BE method appears to be more suitable for the data characteristics (Hauk & Wacziarg, 2009). Because the property rights data do not vary much over time within countries but somewhat noticeably across countries, the use of the BE method becomes suitable. The chapter also tests the data using random effects (RE) and fixed effects (FE) methods; the tables of RE and FE estimates are given in the Appendix. Both RE and FE methods do not appear to suit the characteristics of the property rights data and food (in)security indicators because there is not much accurate time-series variation in either the dependent or independent variable. The chapter, in addition to testing the 6 other variables of property rights, checks robustness by excluding China from the sample to examine whether China could be an outlier. In so doing, it helps address a concern about a possibility that the test results could be driven by “China effect”, I have conducted separate tests by removing China from the baseline sample. I attached the China-excluded test results at the Appendix section of the thesis. The China-excluded results show even stronger correlations between property rights indicators with the four tested variables of food (in)security. The tests could serve as a robustness check and could provide further evidence that with or without China in 97

the sample, the results remain consistently significant across the estimations and data settings.

4.8 Results The OLS results provided in Table 4.5 show that developing countries with greater private property rights experienced less food insecurity.

Table 4.5 OLS results over 1990‒2011 Dependent variable Ln PU Ln PFI Ln DFD Ln ADESA Independent variables Ln property rights protection (PRP) -0.365*** -0.223*** -0.131 0.045*** (0.095) (0.081) (0.132) (0.017) Ln household final consumption -0.580*** -0.577*** -1.033*** 0.113*** expenditure per capita (HHFCEPC) (0.055) (0.051) (0.093) (0.009) Ln population (POP) -0.151*** -0.162*** -0.323*** 0.0438*** (0.028) (0.025) (0.049) (0.005) Agriculture share of GDP -3.366*** -3.682*** -6.569*** 0.689*** (AGCOM) (0.600) (0.553) (0.953) (0.107) Percent of agricultural land -0.0003 -0.001 -0.005*** -0.0001 (AGRILD) (0.001) (0.001) (0.001) (0.0002)

Latitude (LATIT) -0.623*** -0.437** -0.653** 0.188*** (0.236) (0.190) (0.272) (0.037) Ln openness (OPEN) -0.269*** -0.250*** -0.486*** 0.042*** (0.081) (0.070) (0.113) (0.014) Ln human capital (HC) -0.425** -0.517*** -0.966*** 0.099*** (0.174) (0.141) (0.228) (0.029) Ln inflation (INFL) -0.029 0.002 0.031 -0.007** (0.027) (0.018) (0.025) (0.003) Ethnolinguistic fractionalisation 0.139 0.140 0.255* -0.050** (ETHNO) (0.103) (0.090) (0.136) (0.024) Year & Region Effects YES YES YES YES N 545 545 545 545 Countries 57 57 57 57 R2 0.658 0.708 0.703 0.764 Notes: ***, **, * denotes a significance level at 1%, 5% & 10%, respectively. Robust standard errors in parentheses. PU is prevalence of undernourishment; PFI is prevalence of food inadequacy; DFD is depth of food deficit; ADESA is average dietary energy supply adequacy.

As the estimation results indicate, security in private property rights (PRP) is significantly, negatively associated with prevalence of undernourishment (PU), prevalence of food inadequacy (PFI), and depth of food deficit (DFD). In addition, PRP is significantly, positively associated with average dietary energy supply adequacy

(ADESA). The empirical insight provided by this evidence may explain that strong institutional infrastructure that protects private property rights could be essential for alleviating food insecurity in developing economies. The potential impact is that a one- percent increase in security in private property rights (PRP) is associated with lowering prevalence of undernourishment (PU) and prevalence of food inadequacy (PFI) by about 0.37 percent and 0.22 percent, on average, respectively. The results also show that a one- percent increase in security in private property rights (PRP) is associated with improving ADESA by about 0.05 percent, on average. PRP, however, has no significant association with depth of food deficit (DFD). Tables 4.5 to 4.7 provide the results when all control variables are included. For the control variables, the results in Table 4.5 show that per capita household final consumption (HHFCEPC) is significantly, negatively associated with prevalence of undernourishment (PU), prevalence of food inadequacy (PFI), and depth of food deficit (DFD). However, per capita household final consumption (HHFCEPC) is significantly, positively associated with average dietary energy supply adequacy (ADESA). Agriculture share of GDP (AGCOM), human capital index (HC), and agricultural land (AGRILD) are significantly, negatively linked with prevalence of undernourishment (PU), prevalence of

food inadequacy (PFI), and depth of food deficit (DFD), and are significantly positively linked with average dietary energy supply adequacy (ADESA). Openness (OPEN) is significantly associated with food insecurity at 1 percent significance level, but OPEN is not significantly linked to ADESA (Table 4.6, Column 4). The population measure (POP) is statistically significant but is negatively correlated with food insecurity. Latitude (LATIT) has no important relationship with food insecurity measures (namely PU, PFI, and DFD), but it is significantly, positively associated with the food security indicator (namely ADESA).

The findings indicate that inflation (INFL) is not significantly associated with prevalence of undernourishment (PU), prevalence of food inadequacy (PFI) and depth of food deficit (DFD). However, inflation (INFL) appears to have a significant, yet negative association with average dietary energy supply adequacy (ADESA). Finally, ethnolinguistic fractionalisation (ETHNO) is significantly, positively associated with food insecurity, but is significantly, negatively associated with ADESA. Security in private property rights, denoted by property rights protection variable (PRP), appears to have a strong and significant relationship with the measures of food (in)security across the estimations (Tables 4.5‒4.6). As the results demonstrate, security in private property rights (PRP) remains statistically significantly correlated with food insecurity in most specifications even with the inclusion of these control variables. The between estimator (BE) results for the pooled panel data for 57 countries over 1990 to 2011 (Table 4.6) indicate that security in private property rights (PRP) is negatively associated with food insecurity indicators (Columns 1, 2 and 4), all at 5 percent significance level. On impact, a one-percent increase in security in private property rights (PRP) can potentially reduce prevalence of undernourishment (PU) and prevalence of food inadequacy (PFI) by about 0.85 percent and 0.64 percent on average, respectively. The results also show that a one-percent increase in security in private property rights (PRP) potentially improve average dietary energy supply adequacy (ADESA) by about 0.15 percent on average. These potential impacts provided by the BE estimates are larger than those provided by the OLS estimates. That is because the BE estimates provide longer-run effects of security in private property rights (PRP) on food (in)security outcomes (see Section 4.7). In Column (3), however, security in private property rights (PRP) is not significantly correlated with DFD in the pooled panel BE regressions. While security in private property rights (PRP) is not significant to depth of food deficit (DFD), it is negatively correlated with prevalence of undernourishment (PU).

100

Table 4.6 Between Estimator panel results (1990‒2011) Dependent variable Ln PU Ln PFI Ln DFD Ln ADESA Independent variables Ln property rights protection (PRP) -0.853** -0.642* -0.701 0.154** (0.374) (0.330) (0.528) (0.067) Ln household final consumption -0.370** -0.330** -0.633** 0.069** expenditure per capita (HHFCEPC) (0.171) (0.151) (0.241) (0.031) Ln population (POP) -0.072 -0.090 -0.193* 0.023* (0.069) (0.061) (0.098) (0.012) Agriculture share of GDP -1.280 -1.512 -2.405 0.328 (AGCOM) (1.219) (1.076) (1.722) (0.219) Percent of agricultural land -0.001 -0.002 -0.006 0.00002 (AGRILD) (0.004) (0.004) (0.006) (0.001) Latitude (LATIT) -0.510 -0.442 -0.414 0.198 (0.709) (0.625) (1.001) (0.127) Ln openness (OPEN) -0.127 -0.140 -0.227 0.019 (0.271) (0.240) (0.383) (0.049) Ln human capital (HC) -0.069 -0.185 -0.242 0.036 (0.587) (0.518) (0.829) (0.105) Ln inflation (INFL) -0.027 0.045 -0.040 -0.024 (0.232) (0.205) (0.328) (0.042) Ethnolinguistic fractionalisation -0.247 -0.165 -0.166 0.036 (ETHNO) (0.294) (0.260) (0.416) (0.053) North Africa & the Middle East -0.723** -1.048*** -1.590*** 0.165*** dummy (DNAME) (0.325) (0.287) (0.460) (0.059) Sub-Saharan African dummy 0.446 0.238 0.223 -0.077 (DSSA) (0.297) (0.262) (0.419) (0.053) East & SE Asian dummy 0.315 0.269 0.544 -0.030 (DESEA) (0.339) (0.299) (0.478) (0.061) South Asian dummy 0.293 0.221 0.461 -0.063 (DSA) (0.346) (0.306) (0.489) (0.062) N 545 545 545 545 Countries 57 57 57 57 R2 0.675 0.698 0.687 0.737 F 6.230 6.946 6.595 8.413 Notes: ***, **, * denotes a significance level at 1%, 5% & 10%, respectively. Standard errors in parentheses. Region fixed effects are controlled for by adding regional dummies. Latin America & the Caribbean dummy (DLAC) is used as the base, so it is dropped from the model in the estimations. PU is prevalence of undernourishment; PFI is prevalence of food inadequacy; DFD is depth of food deficit; ADESA is average dietary energy supply adequacy.

Among the control variables in Table 4.6, household final consumption expenditure per capita (HHFCEPC) appears to be significantly, negatively associated with prevalence of undernourishment (PU), prevalence of food inadequacy (PFI), and depth of food deficit (DFD). A one-percent increase in HHFCEPC is associated with a reduction in prevalence of undernourishment (PU), prevalence of food inadequacy (PFI),

101

and depth of food deficit (DFD) by about 0.37 percent, 0.33 percent, and 0.63 percent on average, respectively. HHFCEPPC is significantly, positively associated with average dietary energy supply adequacy (ADESA): a one-percent increase in HHFCEPC is associated with improving ADESA by about 0.07 percent. The results show that population (POP) has a significant negative association with DFD and significant positive association with ADESA. A one-percent increase in POP potentially reduces DFD by about 0.19 percent and increase ADESA by about 0.02 percent on average, respectively. If the data are correctly measured, population increase could possibly be compensated by a relatively higher growth in food availability and food access and that smaller developing countries are more exposed to geographic shocks or disadvantages. POP, however, is not significantly linked with PU and PFI. Agricultural land (AGRILD) has no significant relationship with PU, PFI, DFD, and ADESA. None of other control variables appear to have any significant relationship with the four indicators of food (in)security. Among the regional dummies, only the North Africa & the Middle East dummy (DNAME) has a significant, negative association with prevalence of undernourishment (PU), prevalence of food inadequacy (PFI), and depth of food deficit (DFD) and a significant, positive relationship with the average dietary energy supply adequacy (ADESA). Compared to other regions, including Sub-Saharan Africa (SSA), East and Southeast Asia (ESEA), and South Asia (SA), North Africa and the Middle East (or NAME) have done better in reducing food insecurity or in achieving food security. Other things equal, the countries in NAME had substantially less prevalence of undernourishment than the base countries in Latin America and the Caribbean (LAC) by about 72 percent. The countries in NAME appear to have had lower prevalence of food inadequacy (PFI) and depth of food deficit (DFD) by about 100 percent and 150 percent than the LAC countries, other things equal. The countries in NAME appear to have had a higher level of average dietary energy supply adequacy (ADESA) than the LAC countries by about 16 percent. The random effects (RE) and fixed effects (FE) estimates (provided in Tables A4.1‒A4.2 in the Appendix) show that security in private property rights (PRP) is not statistically significant and has an incorrect sign. The presentation of the RE and FE results merely shows that attempts have been made to treat data differently and to test whether the treatment of data to fit RE or FE methods validates or violates assumptions around the characteristics of the property rights data and food insecurity data. In this case, it tends to show that the use of RE and FE methods is not suitable for property rights and

102

food insecurity data that do not have much accurate time-series variations and that are not accurately measured. However, although impreciseness in measuring property rights data would prevail, the main issue would lie with testing method. Because random effects model do not capture within countries variation (see Stern, 2010; Burke & Yang, 2016; Hauk & Wacziarg, 2009; Pesaran & Smith, 1995; Baltagi, 2005), using between estimators approach to make use of this variation across countries appears to be more appropriate.41 In fact, random effect and fixed effects models that attempt to exploit variations in panel data would be useful when the data display sufficient variation within and across observations. However, because the indicators of property rights have been observed to change very little in each country, the use of random effects has provided the results would not be useful as the regressions would only employ impropriate testing methods for this type of data. Therefore, I instead employed between estimators’ approach.

4.9 Robustness tests The chapter used OLS method in addition to BE method to test the data. Although the BE estimation method is used in conjunction with the OLS method to identify longer- term impact of private property rights on food insecurity, the BE method can serve as a test of robustness for the OLS test. The chapter tested six other different measures of property rights. The results in Table 4.7 provides evidence for the alternative testing of robustness against the baseline property rights results. The overall results show that property rights measures potentially affect measures of food (in)security. For instance, security in private property rights (PRP), property rights (PR), property rights and rule-based governance (PRRG), and the international property rights index (IPRI) are significantly, negatively associated with PU, the prevalence of food insecurity. As the evidence indicates, mechanisms and strategies that protect or strengthen private property rights to improve security in private property rights can potentially affect food insecurity outcomes in developing economies. In contrast, the physical property rights score (PPRS) is not significant but it is positively associated with PU (Table 4.7: Column 3). According to the Property Rights Alliance (PRA), a physical property rights score (PPRS) consists of physical property rights protection, registering property, and access to loans. It is not easy to interpret its

41 The references cited in this section are already in the thesis. 103

relationship until we can separate the three subcomponents and test each separately. Furthermore, data on each of these subcomponents from the PRA are not available. One possible argument for the positive relationship between PPRS and PU is that both the registering property rights score and the loan access score, subcomponents of physical property rights, may overshadow measurement of the quality of physical property rights score. It might be possible that some data inaccuracy may exist in measuring the PPRS. Next, the measure of regulatory restrictions on sale of real property (RRRP) is positively correlated with PU (Table 4.7: Column 5). Similarly, the registering property measure (REGPR) is also positively associated with food insecurity (Table 4.9: Column 6). The interpretation of the results on these two indicators can be tricky. If the data on these two indicators are accurately and correctly measured, they can mean that the longer time, more steps, and higher costs of transferring and registering properties can negatively affect property rights quality. This could be because higher costs, longer procedures, and longer times to transfer and register properties will delay or discourage formal property registration and will prompt property rights to remain unregistered or to be registered informally. On the other hand, the nonexistence of clear steps, procedures, dissemination of land-related information, documentation, and support to access publicly available documents also tend to make land transfer and registration difficult. Hence, when transparent and reliable mechanisms are not clearly written and are not available to the public, property rights registration often involves unofficial fee payments, which ultimately discourages formal property registration. One possible drawback for the data on RRRP and REGPR is that differences in price levels are not adjusted when measuring costs of property rights transfer and registration across countries. In this respect, RRRP and REGPR indicators are quite different to PRP and PR indicators. This chapter also conducted tests by excluding China to check robustness of the results over suspicion that China could be an outlier that would overestimate the coefficients. However, the OLS results for the cross-sectional data and BE results for the pooled panel data are consistent across different data settings and test techniques. The results when China is excluded from the sample is provided in Appendix 2. Although this chapter does not test the channels, protecting property rights can influence food insecurity outcomes through several possible mechanisms. First, enhanced security in private property rights has potential to reduce food insecurity in developing economies. Second, better protected property rights potentially improve access to capital resources, such as through better collateralisation of land in credit markets to fund investment in agriculture and production of other food energy sources to feed populations 104

in developing countries. For many of the developing economies, surpluses from crop production because of improved security in private property rights, can be supplied to other economies to earn export incomes or to purchase other food supplies they cannot produce locally. In this way, the potential gains are essential to overcome overall food insecurity. 42

42 The findings in Chapter 2 in this thesis, which are in line with Roth (2010) and de Soto (2000), show evidence of important relationships between property rights in land and food insecurity as enabling of investment in inputs to food and crop production through improved credit access or revenue. 105

Table 4.7 Robustness OLS results: Prevalence of undernourishment and property rights measures Dependent variable Ln prevalence of undernourishment (PU)

Independent variables Period 1990‒ 2011 1995‒2011 2007‒2011 2005‒2011 2003‒2011 2005‒2011 2007‒2011

Ln property rights protection (PRP) -0.365***

(0.095) Ln property rights (PR) -0.198*** (0.048) Ln physical property rights score 0.234 (PPRS) (0.213) Ln property rights & rule-based -0.421** governance (PRRG) (0.189)

Ln regulatory restrictions on sale of 0.119** property (RRRP) (0.055)

106 Ln registering property (REGPR) 0.141 (0.103)

Ln international property rights -0.838*** index (IPRI) (0.273)

………………….....

(table continues below) N 545 780 180 172 425 354 180 Countries 57 61 42 29 61 61 42 R2 0.658 0.644 0.734 0.549 0.648 0.650 0.751 F n.a. 20.29 18.53 5.322 23.98 30.08 21.58 Notes: ***, **, * denotes a significance level at 1%, 5% & 10%, respectively. Robust standard errors in parentheses. The F-statistics in column (1) are not available and so are not reported. The number of countries in observations and the time periods vary depending on the availability of data for respective indicators of property rights.

Table 4.7 Robustness OLS results: Prevalence of undernourishment and property rights measures (continued) Dependent variable Ln prevalence of undernourishment (PU)

Independent variables Period 1990 ‒ 2011 1995‒2011 2007‒2011 2005‒2011 2003‒2011 2005‒2011 2007‒2011 (continued from table above)

Ln household final consumption -0.580*** -0.647*** -0.807*** -0.604*** -0.626*** -0.588*** -0.615*** expenditure per capita (HHFCEPC) (0.055) (0.043) (0.095) (0.088) (0.053) (0.060) (0.084) Ln population (POP) -0.151*** -0.071*** -0.111*** -0.070** -0.107*** -0.064*** -0.078* (0.028) (0.020) (0.043) (0.031) (0.026) (0.023) (0.047) Agriculture share of GDP -3.366*** -2.304*** -4.702*** -1.641*** -2.249*** -1.905*** -4.446*** (AGCOM) (0.600) (0.333) (1.172) (0.453) (0.430) (0.467) (1.105) Percent of agricultural land -0.0003 -0.003*** 0.003 -0.0005 0.0004 -0.0001 0.002 (AGRILDPC) (0.001) (0.001) (0.003) (0.001) (0.001) (0.001) (0.003) Latitude (LATIT) -0.623*** -0.287 -0.776* -0.667** -0.408* -0.230 -0.803** (0.236) (0.194) (0.417) (0.289) (0.241) (0.259) (0.396) Ln openness (OPEN) -0.269*** -0.108* -0.244** -0.205* -0.280*** -0.150* -0.140

107 (0.081) (0.064) (0.112) (0.110) (0.087) (0.084) (0.115) Ln human capital (HC) -0.425** -0.193 -0.304 0.287 -0.275 -0.156 -0.333 (0.174) (0.133) (0.337) (0.243) (0.203) (0.219) (0.316) Ln inflation (INFL) -0.029 -0.162*** 0.692* 1.668*** 1.221*** 1.682*** 0.726** (0.028) (0.045) (0.374) (0.368) (0.305) (0.293) (0.343) Ethnolinguistic fractionalisation 0.139 -0.176*** 0.256* 0.121 0.028 -0.028 0.284* (ETHNO) (0.103) (0.066) (0.152) (0.116) (0.098) (0.108) (0.163) Year & Region Effects YES YES YES YES YES YES YES N 545 780 180 172 425 354 180 Countries 57 61 42 29 61 61 42 R2 0.658 0.644 0.734 0.549 0.648 0.650 0.751 F n.a. 20.29 18.53 5.322 23.98 30.08 21.58 Notes: ***, **, * denotes a significance level at 1%, 5% & 10%, respectively. Robust standard errors in parentheses. The F-statistics in Column (1) are not available and so are not reported. The number of countries in observations and the time periods vary depending on the availability of data for respective indicators of property rights.

4.10 Conclusions The quantitative examination conducted in this chapter provides evidence on international experience of potential impacts of private property rights on food insecurity in developing countries. The chapter uses three measures of food insecurity (i.e., PU, PFI, and DFD) and one measure of food security (i.e., ADESA) as the left-hand-side variables. It uses seven indicators of property rights, including the main variable of security in private property rights (PRP) as right-hand-side variables. Pooled panel regressions using a between estimator (BE) method reveal that stronger security in private property rights plays a significant role in reducing food insecurity. The difference between the OLS results and the BE results is that the BE method provides longer-run effects of security in private property rights (namely property rights protection, PRP) on food insecurity outcomes across the developing economies. Therefore, the potential effects provided by the BE estimates appear to be larger in size than those in the OLS estimates. The BE results demonstrate that a one-percent increase in security in private property rights (PRP) potentially leads to a reduction in prevalence of undernourishment (PU) and prevalence of food inadequacy (PFI) by about 0.85 percent and 0.64 percent on average, respectively. The BE results also show that a one-percent increase in property rights protection (PRP) potentially improve average dietary energy supply adequacy (ADESA) by about 0.15 percent on average. Although security in private property rights (PRP) is not significantly associated with depth of food deficit (DFD), it has the expected positive relationship with depth of food deficit (DFD). The OLS estimates provide similar evidence to the BE results. As the results show, security in private property (PRP) is significantly associated with a reduction in prevalence and depth of food insecurity. The results of the tests on the other six indicators of property rights indicate that measures that uphold property rights or mechanisms that facilitate fast registration or transfer of property rights are positively linked to reducing food insecurity. Among the six property rights measures tested, property rights (PR), property rights and rule-based governance (PRRG), and the international property rights index (IPRI) are significantly correlated with reduced prevalence of undernourishment (PU). The overall evidence suggests the significance of securing private property rights, such as against expropriation and toward observing inclusive private property rights. Because a big majority of the poor in the developing countries engage in subsistence farming or small-scale agriculture, securing their agricultural land rights could influence their food availability, food access,

108

and food consumption. Collectively, strengthening private property rights, including agricultural and non-agricultural and, can impart positive implications for a gradual reduction in their food insecurity. Arguably, that can have positive impact on overall socioeconomic livelihoods of those segments of population who rely on it for income and food. In sum, the findings in this chapter emphasise the significance of security in private property rights in reducing food insecurity in developing countries.

109

CHAPTER 5

CONCLUDING REMARKS

5.1 Introduction The seriousness of food insecurity, including persistent hunger and chronic undernutrition, in many developing economies has prompted this research into key drivers of food insecurity. The thesis began with quantifying (i) potential impacts of agricultural land property rights on household food insecurity in rural Cambodia. It then examined (ii) potential impacts of excessive flooding and irrigation access on rice production and household food insecurity in rural Cambodia. In assessing (i) and (ii), it used primary data from a household survey administered to 256 households in 32 rural villages across four major rice-growing provinces in Cambodia in 2014. Lastly, the thesis assessed (iii) potential impacts of private property rights on food insecurity in 57 developing countries, using secondary pooled panel data over 1990 to 2011. The examinations in this thesis showcase the first empirical research that not only assessed the empirical links quantitatively but also used the data from a large cross-section of developing countries and the primary data from the household survey in rural Cambodia. The findings discussed in the next section provide new empirical insights that could contribute to an effective food policy for tackling food insecurity in the developing economies, including Cambodia. The estimation results in this chapter indicate that there seems to be strong association between increased security in agricultural land property rights and farmers’ access to credit and their ability to collateralize their titled rice plots. Several studies have assessed land tenure security impacts on rural credits and have found their positive relationships (Markusssen et al., 2011; Atwood, 1990; Besley, 1995; Field & Torero, 2006; de Soto, 2000; Piza et al., 2016, Holden et al., 2009; Deininger & Castagnini, 2006) have either found or argued that secure property rights encourage credit access. It has been documented that rural farmers tend to sell their farms as well since around the mid of 1990s (So et al., 2001). However, a large proportion of rural farmers rely on farmland for food and income and they could sell, if any, some part of the land but not all of it. This similar pattern has been observed not only in Cambodia but also in many other developing countries, for instance China, during and at the wake of the global financial crisis in 2008

110

when they lost their jobs in the industrial and service sectors and had to return to their rural farmland as an effective buffer. While some farmers in rural Cambodia, and possibly in Thailand and other developing countries, would seek to acquire more land through encroaching state land, that process has been reduced dramatically because most of the state land has been placed as protected land and forest. If they could encroach, the type of land encroached is not suitable for rice farming because until today only forest land can still be encroached and that type of land is in general suitable for industrial crops, such as cashew, coffee, cassava, rubber, and so on.

5.2 Key findings The main results and statistical inferences are summarised as follows. Note that all the estimation results are collated and given in Appendix 4: Table A5.1. The Cambodian evidence indicates that households with greater agricultural land property rights had less food insecurity. It shows that an increase in household agricultural land property rights potentially leads to lower probability of facing household food insecurity. An increase in security in agricultural land property rights of household tends to lead to higher probability of shortening the length of household food insecurity among rural farmers. Strengthening agricultural land property rights potentially benefits rural farmers when they could capitalise on secure land rights in the production of food crops to tackle their food insecurity. Additional evidence from Cambodia shows positive impacts of agricultural land property rights on the following possible channels. Greater agricultural land property rights appear to have enabled rice farmers to obtain greater access to credit from credit institutions. Similarly, greater security in agricultural land property rights tends to enable them to collateralise their plots, for instance, in land and rural financial markets and possibly in other creative use for their economic gains.43 When land property rights and credit markets complement each other, growth in the latter is essential. As land property rights contribute to improving farmers’ revenue-cost ratios, it could help farmers optimise

43 This evidence is in line with some existing studies, including Atwood (1990), Besley (1995), Field and Torero (2006), Piza et al. (2016), and de Soto (2000). However, the evidence contrasts with Lawry et al. (2016), Carto and Olinto (2003), Deininger and Feder (2009), and Brasselle et al. (2002). In Carter and Olinto (2003), Deininger and Feder (2009), and Brasselle et al. (2002), land property rights have positive impacts on facilitation of credit access if properly functioning rural credit markets exist. 111

gains from rice production. The Cambodian evidence also reveals that greater agricultural land rights tend to improve rice productivity and raise rice revenue.44 The cross-country examination provides similar evidence: developing countries with greater private property rights appear to have experienced less food insecurity. The between estimator (BE) estimation results indicate that a one-percent increase in property rights security (namely PRP) would potentially reduce prevalence of undernourishment (PU) and prevalence of food inadequacy (PFI) by about 0.85 percent and 0.64 percent on average, respectively. The evidence also shows that a one-percent increase in security in private property rights (PRP) can potentially improve dietary energy supply adequacy (ADESA) by about 0.15 percent on average. However, property rights security is not significant but is negatively associated with depth of food deficit (DFD). In tandem, the robustness test results of the ordinary least squares (OLS) regressions show that a one- percent increase in PRP is associated with a reduction in prevalence of undernourishment (PU) by about 0.37 percent and in prevalence of food inadequacy (PFI) by about 0.22 percent on average. Additionally, a one-percent increase in private property rights security (PRP) can potentially improve average dietary energy supply adequacy (ADESA) by about 0.05 percent, on average. The plot-level evidence from Cambodia shows that plots having access to irrigation seemed to have higher rice productivity and rice revenue. On potential impact, providing irrigation for the currently unirrigated plots could improve rice yield by about 0.7 tonnes per harvest or 0.6 million riels per hectare on average, respectively. The plot- level evidence shows that plots with access to formal irrigation, i.e., the one which is sourced from rivers, dikes, or canals (RDC), tends to have higher rice yield and rice revenue, at per harvest and per hectare levels, relative to access to river, lake, or pond (RLP) irrigation and underground or piping (UP) irrigation. The plot-level evidence also shows that plots with access to formal irrigation (RDC) is associated with higher total rice yield by about 0.98 tonnes per harvest and 1.2 tonnes per hectare, relative to plots having no irrigation. The plot-level evidence also demonstrates that having formal irrigation supply tends to raise rice revenue by about 0.81 million riels (≈USD200) per harvest and 0.93 million riels (≈USD225) per hectare on average respectively, relative to plots without access to irrigation. Having access to RLP irrigation appears to provide lower rice production and rice revenue than formal irrigation access. RLP could improve

44 This evidence contrasts with Deininger and Ali (2008), Chankrajang (2015), Gavian and Ehui (1999), Place and Otsuka (2002), and Brasselle et al. (2002). 112

rice yield and rice revenue by about 0.38 tonnes and 0.34 million riels per harvest per hectare, respectively, relative to no irrigation. RLP irrigation has no significant link with per-hectare rice yield and per-hectare rice revenue. Lastly, UP irrigation does not have any significant link with any of the tested measures of rice productivity and rice revenue. The Cambodian results indicate that plots affected by extreme flooding had lower rice production and rice revenues, relative to those plots unaffected by extreme flooding.45 The extreme flooding would reduce per-hectare rice yield by about 0.7 tonnes per harvest or 0.3 to 0.4 tonnes per hectare on average, respectively. Extreme flooding tends to reduce rice revenue by about 0.6 million riels (≈USD150) per harvest or 0.3 million riels (≈USD75) per hectare on average, respectively. Another piece of plot-level evidence indicates that formal irrigation (RDC) tends to be associated with raising rice productivity and rice revenues. Providing irrigation to the currently unirrigated plots could potentially improve rice yield and rice revenue by about 0.7 tonnes per harvest and 0.57 million riels per hectare, respectively. This potential impact could help uplift the rice farmers, at least most of those surveyed households, in rural Cambodia not only out of food insecurity situations in the medium term but also out of abject poverty totally in the long run.

5.3 Policy implications The empirical findings in this thesis provide an insight on food policy implications that could be important to Cambodia and the developing economies as follows. First, securing private property rights in agricultural land can be an option for tackling food insecurity in developing economies, including Cambodia. Activities that promote private land property rights in developing economies should continue and intensify. In the case of Cambodia, speeding up the current land certification and titling efforts to allow rural farmers, including those in the survey sites who currently do not have a land title, would enable them to multiply their income and further access to loans. For rural farmers in Cambodia, being able to collateralise their agricultural land could enable many of them to obtain low-interest loans or to use the land for other purposes that would make them economically better off. Activities that guarantee or secure private property rights over agricultural land will tend to make rice farmers less food insecure gradually and probably into the long run. Measures or mechanisms that protect private

45 This evidence is in line with Karunasagar and Karunasagar (2016), Gregory et al. (2005), and Malla (2010). 113

property rights should proceed under a legal and judicial system that enforces and recognises necessary land rights to enable or facilitate productive land use, for example, in crop production and land markets. The governance and conduct of agricultural land property rights should be articulated to suit social, cultural, political, and economic contexts of the country. Second, providing formal irrigation is another key option for improving rice yields and rice revenues for rice farmers in rural Cambodia, especially for those surveyed households who lack irrigation supply. The current shortage of irrigation supply in the country, i.e., only about one-fourth of the total agricultural land is irrigated, and particularly the lack of irrigation among the rice farmers, may have been one of the obstacles to improve agricultural production and productivity. Despite a growing budget expenditure in the past 10 years, the state budget allocated for agriculture, particularly irrigation infrastructure to irrigate wet-season and dry-season rice, remains significantly low, much less than levels in countries with similar GDP per capita (World Bank, 2015b). Expanding the acreage under irrigation, especially to provide irrigation for the currently unirrigated plots, would support government’s efforts to improve agricultural production, tackling food insecurity, and fighting poverty, as articulated in the Policy Paper on the Promotion of Paddy Production and Rice Export (RGC, 2010), the National Strategic Development Plan 2014‒2018 (RGC, 2014) and the Rectangular Strategy Phase III (RGC, 2013). On immediate impact, expanding irrigation availability and access would encourage farmers to increase rice cropping frequency, diversify their cropping to respond to seasonality, water availability, and climatic challenges. In so doing, rice farmers could potentially enhance their agricultural productivity and revenue, including rice and other food and non-food crops. The enhanced rice productivity and revenue will tend to reduce the length of household food insecurity. Improvement in irrigation supply would possibly allow the production of other food crops and energy crops farmers as well. Third, developing a strategy to minimise damage caused by annual excessive flooding within the rice-growing areas in rural Cambodia is another key solution. Although developing flood-control systems can be costly, investing in it for long-term positive impacts on rural agriculture and rural economy could be beneficial. The Cambodian government has long recognised the shortage of flood drainage systems or flood-control networks and the importance of developing these systems (RGC, 2007; MOWRAM, 2012; MAFF, 2013). However, the Cambodian government has yet to adopt a legal framework for natural disaster management (World Bank, 2015b) and long-term

114

development strategy for the systems. This lack of a disaster management law, under which the government has expected to establish the disaster management mechanism, may have stifled efforts to develop rural infrastructure, including flood-control system. To help the Cambodian government achieve its national food security (RGC, 2013), poverty reduction RGC, 2014), and possibly rice export targets (RGC, 2010), expanding investment in flood drainage systems could be essential. That could include building flood-control mechanisms; however, the system development should both maintain the general benefits of annual floods and stabilise the natural ecosystems. In so doing, it would directly benefit rural households whose cropland is in areas easily affected by flooding for them to avoid flood-damaged crop loss and repeated interruptions in crop production activities. In addition, designing flood-control networks to release not only the excess floodwaters but also to harvest the excess floodwaters for use in water-deficit months and underirrigated areas can be beneficial.

5.4. Suggestions for future studies The data constraints and the empirical findings provided in this thesis may warrant further efforts to improve data quality and availability and to undertake a deeper quantitative analysis. Because most of the existing data on property rights consist of a bundle of property rights aspects, such as physical, intellectual, and institutional arrangements of property rights registration and transfer, for instance. Separating individual elements of private property rights, for example, agricultural land rights from the rest would make analytical tasks much easier to delink their respective potential impact on an outcome variable in empirical works. The standardising the measurement of land property rights security and household food insecurity indicators in future household surveys could be useful. Future works that examine impact of individual aspects of property rights on an outcome in question can be beneficial. The future research should use large panel data for an individual country case, in addition to the multi-country analysis when data become available. Similarly, future research on cost-benefit analysis of irrigation and flood- control interventions in developing economies would also be valuable.

115

REFERENCES

Abdulai, A., Owusu, V., and Goetz, R. 2011. Land tenure differences and investment in land improvement measures: Theoretical and empirical analyses. Journal of Development Economics, vol. 98, pp. 66–78.

Acemoglu, D. 2009. Introduction to modern economic growth. Princeton University Press. New Jersey: USA.

Acemoglu, D., Johnson, S. and Robinson, J.A. 2004. Institutions as the fundamental cause of long-run growth, in P. Aghion & S. N. Durlauf (eds). Handbook of Economic Growth, Elsevier, pp. 364–372.

Alchian, A.A and Demsetz, H. 1973. The property rights paradigm. The Journal of Economic History, vol. 33, no. 1, pp. 16‒27.

Ali, D.A., Dercon, S. and Gautam, M. 2011. Property rights in a very poor country: tenure insecurity and investment in Ethiopia. Agricultural Economics, vol. 42, no. 1, pp. 75‒86.

Ali, M.H. and Talukder, M.S.U. 2008. Increasing water productivity in crop production: a synthesis. Agricultural Water Management, vol. 95, no. 14, pp. 1201‒1213.

Allen, R.C. 1982. The efficiency and distributional consequences of eighteenth century enclosures, Economic Journal, vol. 92, pp. 937‒953.

Allouche, J. 2011. The sustainability and resilience of global water and food systems: political analysis of the interplay between security, resource scarcity, political systems and global trade. Food Policy, vol. 36, no. 1, pp. S3‒S8.

Amemiya, T. 1973. Regression analysis when the dependent variable is truncated normal. Econometrica, vol. 41, no. 6, pp. 997‒1016.

Alston, L. J., Harris, E. & Mueller, B. 2009. De facto and de jure property rights: land settlement and land conflict on the Australian, Brazilian, and US frontiers, NBER Working Paper #15264. Anderson, K., Ivanic, M., and Martin, W. 2013. Food price spikes, price insulation, and poverty. The World Bank Policy Research Working Pater #6535.

Asian Development Bank (ADB). 2012. Flood damage emergency reconstruction project: preliminary damage and loss assessment. Accessed at https://www.adb.org/sites/default/files/linked-documents/46009-001-cam-oth- 01.pdf

116

Asian Development Bank (ADB). 2014. Improving rice production and commercialisation in Cambodia: findings from the farm investment climate assessment. Asian Development Bank, the Philippines.

Atwood, D.A. 1990. Land registration in Africa: the impact on agricultural production. World Development, vol. 18, no. 5, pp. 659‒671.

Babu, S.C., Gajana, S.N., and Sanyal, P. 2014. Food Security, Poverty and Nutrition Policy Analysis: Statistical Methods and Applications. 2nd edition, Elsevier Inc., USA.

Baltagi, B.H. 2005. Econometric Analysis of Panel Data. Third Edition, John Wiley & Sons: Chichester: UK.

Bandara, S.J. and Cai, Y. 2014. The impact of climate change on food crop productivity, food prices and food security. Economic Analysis and Policy, vol. 44, pp. 451‒465.

Banerjee, A. and Duflo, E. 2007. The economic lives of the poor. Journal of Economic Perspectives, vol. 21, no. 1, pp. 141‒167. Barrett, C.B. 2008. Smallholder market participation: Concepts and evidence from eastern and southern Africa. Food Policy, vol. 33, no. 4, pp. 299−317.

Barrett, C.B. (ed.). 2013. Food Security and Socio-political Stability. Oxford University Press, Great Britain.

Barrett, C.B. and Upton, J.B. 2013. Food security and social political stability in sub- Saharan Africa, in C.B. Barrett (ed.). 2013. Food security and socio-political stability. Oxford University Press, UK, pp. 323‒356.

Beegle, K., Himlein, K., and Ravaliion, M. Frame-of-reference in subjective welfare regressions. Policy Research Working Paper #4904. The World Bank Research Group, World Bank. Washington D.C., USA.

Besley, T. and Burgess, R. 2000. Land reform, poverty reduction, and growth: evidence from India. The Quarterly Journal of Economics, vol. 115, no. 2, pp. 389‒430.

Besley, T. 1995. Property rights and investment incentives: theory and evidence from Ghana. Journal of Political Economy, vol. 103, no. 5, pp. 903‒937.

Besley, T. and Ghatak, M. 2009. Property rights and economic development, LSE STICERD Research Paper # EOPP 006.

Bover, O. 2015. Measuring expectations from household surveys: new results on subjective probabilities of future house prices. SERIEs Journal of the Spanish Economic Association, vol. 6, pp. 361‒405.

117

Brasselle, A.S., Gaspart, F. and Platteau, J.P. 2002. Land tenure security and investment incentives: puzzling evidence from Burkina Faso. Journal of Development Economics, vol. 67, no. 2, pp. 373‒418.

Brinkman, H-J., de Pee, S., Sanogo, I., Subran, L., and Bloem, M. W. 2010. High food prices and the global financial crisis have reduced access to nutritious food and worsened nutritional status and health, Journal of Nutrition, vol. 140, pp. 153S‒ 161S.

Burkes, P.J and Yang, H. 2016. The price and income elasticities of natural gas demand: International evidence. Energy Economics, vol. 59, pp. 466‒474.

Cabangon, R.J., Tuong, T.P. and Abdullah, N.B. 2002. Comparing water input and water productivity of transplanted and direct-seeded rice production systems. Agricultural Water Management, vol. 57, no. 1, pp. 11‒31.

Cambodia Daily. 2016. ‘In Frustration, Germany Ends Land Rights Work’. Accessed from https://www.cambodiadaily.com/news/in-frustration-germany- ends-land-rights-work-107406/

Cambodia Development Resources Institute (CDRI). 2008. Impact of high food prices in Cambodia. A CDRI survey report. Phnom Penh: Cambodia.

Cambodia Socioeconomic Survey (CSES) 2011. National Institute of Statistics, Ministry of Planning. Phnom Penh: Cambodia (2011).

Cameron, A.C. and Trevidi, P.K. 2005. Microeconometrics: Methods and Applications. New York: USA.

Chan, S. 2011. The impact of high food prices on food security in Cambodia. Development in Practice, vol. 21, no. 4‒5, pp. 718‒731.

Chankrajang, T. 2015. Partial land rights and agricultural outcomes: evidence from Thailand. Land Economics, vol. 91. no. 1, pp. 126‒148.

Campbell, B.M., Vermeulen, S.J., Aggarwal, P.K., Corner-Dolloff, C., Girvetz, E., Loboguerrero, A.M., Ramirez-Villegas, J., Rosenstock, T., Sebastian, L., Thornton, P., and Wollenberg, E. 2016. Reducing risks to food security from climate change. Global Food Security (2016), http://dx.doi.org/10.1016/j.gfs.2016.06.002i

Carruthers, I., Rosegrant, M.W., and Seckler, D. 1997. Irrigation and food security in the 21st century. Irrigation and Drainage Systems, vol. 11, no. 2, pp. 83‒101.

Carter, M.R and Olinto, P. 2003. Getting institutions ‘right’ for whom? Credit constraints and the impacts of property rights on the quantity and composition of investment. American Journal of Agricultural Economics, vol. 85, no. 1, pp. 173‒ 186. 118

Cho, Y., Margolis, D., Newhouse, D., and Robalin, D. 2012. Labor markets in low- and middle-income countries: trends and implication for social protection and labor policies. The World Bank 2012−2022 Social Protection and Labor Strategy’s Discussion Paper # 1207.

Collier, P. and Hoeffler, A. 1998. On economic causes of war, Oxford Economic Papers, vol. 50, pp. 563‒573.

Collier, P. and Hoeffler, A. 2002. On the incidence of civil war in Africa, Journal of Conflict Resolution, vol. 46, no. 1, pp. 13‒28.

Collier, P. and Hoeffler, A. 2004. Greed and grievance in civil war, Oxford Economic Papers, vol. 56. no. 4, pp. 563‒595.

Conning, J. and Deb, P. 2007. Impact evaluation for land property rights reforms. Doing Impact Evaluation Series, no.8, World Bank, Washington, DC: USA.

Cooper, G.W. 2002. Land policy and conflict: the Cambodia portion of an eight-country study by the North-South Institute for the World Bank.

Council for Agricultural and Rural Development (CARD). 2014. National Strategy for Food Security and Nutrition. Phnom Penh: Cambodia.

De Janvry, A., Emerick, K., Gonzalez-Navaro, M., and Sadoulet, E. 2015. Delinking land rights from land use: certification and migration in Mexico. American Economic Review, vol. 105, no. 10, pp. 3125‒3149.

De Lopez, T.T. 2002. Natural resource exploitation in Cambodia: an examination of use, expropriation, and exclusion. Journal of Environment and Development, vol. 11, no. 4, pp. 355‒379.

Demsetz, H. 1967. Toward a theory of property rights. American Economic Review, vol. 57, pp. 347‒369. de Soto, F. 2000. The Mystery of Capital: Why Capitalism Triumphs in the West and Fails Everywhere Else. London: Great Britain.

Deaton, A. 1997. The Analysis of Household Surveys: A Microeconometric Approach. World Bank. Washington, DC: USA.

Deininger, K. and Ali, D.A. 2008. Do overlapping land rights reduce agricultural investment? Evidence from Uganda. American Journal of Agricultural Economics, vol. 90, no. 4, pp. 879‒882.

Deininger, K. and Castagnini, R. 2006. Incidence and impact of land conflict in Uganda. Journal of Economic Behavior and Organization, vol. 60, no. 3, pp. 321‒345.

119

Deininger, K. and Jin, S. 2006. Tenure security and land-related investment: evidence from Ethiopia. European Economic Review, vol. 50, no. 5, pp. 1245‒1277.

Deininger, K., Ali, D.A. and Alemu, T. 2011. Impacts of land certification on tenure security, investment, and land market participation: evidence from Ethiopia. Land Economics, vol. 87, no. 3, pp. 312‒334. Delavande, A. and Kohler, H-P. 2009. Subjective expectations in the context of HIV/AIDS in Malawi. Journal of Demographic Research, vol. 20, no. 31, pp. 817−874. Delavande, A., Giné, X. and McKenzie, D. 2011. Measuring subjective expectations in developing countries: A critical review and new evidence. Journal of Development Economics, vol. 94, no. 2, pp. 151−163.

Deutsch, R. 2006. Beneficiary assessment of land title recipients under the Land Management and Administration Project (LMAP). Prepared for the Ministry of Land Management, Urban Planning and Construction, Phnom Penh: Cambodia.

Do, Q.T. and Iyer, L. 2008. Land titling and rural transition in Vietnam. Economic Development and Cultural Change, vol. 56, no. 3, pp. 531‒579.

Domènech, L. 2015. Improving irrigation access to combat food insecurity and undernutrition: a review. Global Food Security, vol. 6, pp. 24‒33.

Douglas, I. 2009. Climate change, flooding and food security in south Asia. Food Security, vol. 1, pp. 127‒136.

Engerman, S.L. and Sokoloff, K. L. 1997. Factor endowments, institutions, and differential paths of growth among new world economies: a view from economic historians of the United States, in: S. Haber (ed.) How Latin America Fell Behind (Redwood, CA: Stanford University Press), pp. 260‒304. Ettner, S.L. 1996. New evidence on the relationship between income and health. Journal of Health Economics, vol. 15, no. 1, pp. 67-85.

Evans, R.G. and Sadler, E.J. 2008. Methods and technologies to improve efficiency of water use. Water Resources Research, vol. 44, no. 7, W00E04. doi:10.1029/2007WR006200

Feder, G. 1987. Land ownership security and farm productivity: evidence from Thailand. Journal of Development Studies, vol. 24, no. 1, pp. 16‒30.

Feder, G. and Feeny, D. 1991. Land tenure and property rights: theory and implications for development policy. World Bank Economic Review, vol. 5, no. 1, pp. 135‒153.

Feder, G. and Onchan, T. 1987. Land ownership security and farm investment in Thailand. American Journal of Agricultural Economics, vol. 69, no. 2, pp. 311‒ 320. 120

Feder, G. and Nishio, A. 1999. The benefits of land registration and titling: economic and social perspectives. Land Use Policy, vol. 15, no. 1, pp. 25‒43.

Feenstra, R.C., Inklaar, R., and Timmer, M.P. 2013. PWT 8.0: A user guide. Accessed at http://www.rug.nl/research/ggdc/data/pwt/v80/pwt_80_user_guide.pdf Feenstra, R.C., Inklaar, R., and Timmer, M.P. 2015. ‘The Next Generation of the Penn World Table,’ forthcoming in American Economic Review, available for download at www.ggdc.net/pwt

Fenske, J. 2010. L’Étranger: status, property rights, and investment incentives in Côte d’Ivoire. Land Economics, vol. 86, no. 4, pp. 621‒644.

Fenske, J. 2011. Land tenure and investment incentives: evidence from West Africa. Journal of Development Economics, vol. 95, no. 2, pp. 137‒156.

Field, E. 2005. Property rights and investment in urban slums. Journal of the European Economic Association, vol. 3, no. 2/3, pp. 279‒290.

Field, E. and Torero, M. 2006. Do property titles increase credit access among the urban poor? Evidence from a nation-wide titling program, Harvard University. Accessed from http://scholar.harvard.edu/files/field/files/fieldtorerocs.pdf?m=1360040257

Fletcher, R., Penny, D., Evans, D., Pottier, C., Barbetti, M., Kummu, M., Lustig, T. and APSARA Authority. 2008. The water management network of Angkor, Cambodia. ANTIQUITY, vol. 82, pp. 658‒670.

Food and Agriculture Organization of the United Nations (FAO). 2000. The elimination of food insecurity in the Horn of Africa: A strategy for concerted government and UN agency action. FAO Final Report.

Food and Agriculture Organization of the United Nations (FAO). 2012. Irrigation in southern and eastern Asia in figures: a QUASTAT Survey 2011. FAO Water Report #37.

Food and Agriculture Organization of the United Nations (FAO). 2013. FAO Statistical Yearbook 2013: World Food and Agriculture. Rome: Italy.

Forbes Asia. 2014. Cambodians Say Land Grabs Have Led to Crime Against Humanity; Want the ICC to Investigate. Accessed from http://www.forbes.com/sites/meghabahree/2014/10/07/cambodians-say-land- grabs-have-lead-to-crime-against-humanity-want-the-icc-to- investigate/#34b9bece2034

121

Galiani, S. and Schargrodsky, E. 2010. Property rights for the poor: effects of land titling. Journal of Public Economics, vol. 94, no. 9‒10, pp. 700‒729.

Garrido, A. and Ingram, H. (eds). 2011. Water for food in a changing world. New York: USA.

Gavian, S. and Fafchamps, M. 1996. Land tenure and allocative efficiency in Niger. American Journal of Agricultural Economics, vol. 78, no. 2, pp. 460‒471.

Gavian, S. and Ehui, S. 1999. Measuring the production efficiency of alternative land tenure contracts in a mixed crop-livestock system in Ethiopia. Agricultural Economics, vol. 20, no. 1, pp. 37‒49.

Ghosh, S., Srivastava, S.K, Nayak, A.K., Panda, D.K., Nanda, P. and Kumar, A. 2014. Why impacts of irrigation on agrarian dynamism and livelihoods are contrasting: evidence from eastern Indian states. Irrigation and Drainage, vol. 63, no. 5, pp. 573‒583.

Gillson, A. and Fouad, A. (eds). 2015. Trade policy and food security: improving access to food in developing countries in the wake of high world prices. International Bank for Reconstruction and Development/The World Bank.

Golay, C. and Biglino, I. 2013. Human rights responses to land grabbing: a right to food perspective. Third World Quarterly, vol. 34, no. 9, pp. 1630‒1650.

Goldstein, M and Udry, C. 2008. The profits of power: land rights and agricultural investment in Ghana. Journal of Political Economy, vol. 116, no. 6, pp. 981‒ 1022.

Greene, W.H. 2012. Econometric Analysis. Seventh Edition. New Jersey: USA.

Gregory, P.J., Ingram, J. S. I., and Brklacich, M. 2005. Climate change and food security. Philosophical Transition of Royal Society B Biological Sciences, vol. 360, pp. 2139‒2148.

GTZ. 2009. Foreign Direct Investment (FDI) in Land in Cambodia. Eschborn: Germany.

Guha-Khasnobis, B, Acharya, S.S. and Davis, B (eds.). 2007. Food security: indicators, measurement, and the impact of trade openness, Oxford University Press: UNU- WIDER.

Han, P., Fukui, S., and Miwa, K. 2008. Testing the “Wealth Paradox” on the incidence of child labor: a case study in Cambodia. GSICS Working Paper #18, Kobe University.

Hauk Jr., W.R. and Wacziarg, R. 2009. A Monte Carlo study of growth regressions. Journal of Economic Growth, vol. 14, no. 2, pp. 103–147. 122

Hayes, J., Roth, M. and Zepeda, L. 1997. Tenure security, investment and productivity: a generalized probit analysis. American Journal of Agricultural Economics, pp. 369‒382.

Hicks, S. 2011. The measurement of subjective well-being. Paper for Measuring National Well-being Technical Advisory Group, Office of National Statistics.

Hoffmann, R., Kröger, H., and, Geyer, S. 2018. Social causation versus health selection in the life course: does their relative importance differ by dimension of SES? Social indicators Research, pp. 1−27. https://doi.org/10.1007/s11205-018- 1871-x

Holden, S., Deininger, K. and Ghebru, H. 2009. Impacts of low-cost land certification on investment and productivity. American Journal of Agricultural Economics, vol. 91, no. 2, pp. 359‒373.

Holden, S. and Yohannes, H. 2002. Land redistribution, tenure insecurity, and intensity of production: A study of farm households in Southern Ethiopia, Land Economics, vol. 78, no. 4, pp. 573‒590.

Hossain, M. and Fisher, K.S. 1995. Rice research for food security and sustainable agricultural development in Asia: achievements and future challenges. GeoJournal, vol. 35, no. 3, pp. 286‒298.

Hossain, M., Naher, F., and Shahabuddin, Q. 2005. Food security and nutrition in Bangladesh progress and determinants. Journal of Agricultural and Development Economics, vol. 2, no. 2, pp. 103‒132.

International Food Policy Research Institute (IFPRI). 2014 Global Hunger Index: The challenge of hidden hunger. Washington D.C. (2014).

Jacoby, H.G., Li, G. and Rozelle S. 2002. Hazards of expropriation: tenure security and investment in rural China. American Economic Review, vol. 92, no. 5, pp. 1420‒ 1446.

Jacoby, H.G. and B. Minten. 2007. Is land titling in Sub-Saharan Africa cost-effective? Evidence from Madagascar. The World Bank Economic Review, vol. 21, no. 3, pp. 461‒485.

Johnston, R., Try, T., and de Silva, S. 2013. Agricultural water management in Cambodia. In Investing in Water Management to Improve Productivity of Rice- Based Farming Systems in Cambodia Project. Issue Brief #1, International Water Management Institute of the Australian Centre for International Agricultural Research.

Jones, A.D., Ngure, F.M., Pelto, G., and Young, S.L. 2013. What are we measuring when we mean food security? A compendium and review of current metrics. Advances in Nutrition, vol. 4, no. 5, pp. 481‒505. 123

Jones, L. and Samman, E. 2016. Measuring subjective household resilience: insights from Tanzania. Working and Discussion Paper (June 2016), Overseas Development Institute.

Kang, Y., Khan, S., and Ma, X. 2009. Climate change impacts on crop yield, crop water productivity and food security: a review. Progress in Natural Science, vol. 19, pp. 1665‒1674.

Karunasagar I and Karunasagar, I. 2016. Challenges of food security: need for interdisciplinary collaboration. Procedia Food Science, vol. 6, pp. 31‒33.

Kemper, N., Viet Ha, L., and Klump, R. 2013. Property rights and consumption volatility: evidence from a land reform in Vietnam. World Development, vol. 71, pp. 107‒130.

Kiple, K.F. and Ornelas, K.C. (eds). 2000. The Cambridge world history of food, vol. 1, Cambridge University Press, New York: USA.

Kirby, M. and Mainuddin, M. 2009. Water and agricultural productivity in the lower Mekong basin: trends and future prospects. Water International, vol. 34, no. 1, pp. 134‒143.

Kumar, M.D., Savamohan, M.V.K., and Bassi, N. (eds.). 2013. Water management, food security and sustainable agriculture in developing economies. Routledge: Taylor & Francis Group, New York: USA.

Lall, U. 2013. Global freshwater and food security in the face of potential adversity, in C.B. Barrett (ed.). 2013. Food security and sociopolitical stability. Oxford University Press (UK), pp. 120‒141.

La Croix, S.J. and Roumasset, J. 1990. The evolution of private property in nineteenth- century Hawaii. Journal of Economic History, vol. 50, no. 4, p. 829‒852.

La Porta, R., Lopez-de-Silanes, F, Shleifer, A, and Vishny, R. 1999. The quality of government. Journal of Law, Economics, and Organization, vol. 15, no. 1, pp. 222‒279.

La Porta, R., Lopez-de-Silanes, F., Pop-Eleches C., and Shleifer, A. 2004. Judicial checks and balances. Journal of Political Economy, University of Chicago Press, vol. 112, no. 2, pp. 445‒470.

La Porta, R., & Florencio Lopez-de-Silanes & Andrei Shleifer. 2008. The economic consequences of legal origins. Journal of Economic Literature, American Economic Association, vol. 46, no. 2, pp. 285‒332.

Lawry, S., Samii, C., Hall, R., Leopold, A., Hornby, D., and Mtero, F. 2016. The impact of land property rights interventions on investment and agricultural productivity in

124

developing countries: a systematic review. Journal of Development Effectiveness, (online).

Levidow, L., Zaccaria, D., Maia, R., Vivas, E., Todorovic M. & Scardigno, A. 2014. Improving water-efficient irrigation: Prospects and difficulties of innovative practices. Agricultural Water Management, vol. 146, pp. 84‒94.

Levine, Ross. 2005. Law, Endowments and Property Rights, Journal of Economic Perspectives, vol. 19, no. 3, pp. 61‒88.

Li, G., Rozelle, S., and Brandt, L. 1998. Tenure, land rights, and farmer investment incentives in China. Agricultural Economics, vol. 19, pp. 63‒71.

Libecap, G. 1989. Contracting for Property Rights. Cambridge University Press, New York.

Lipton, M. 2009. Land Reform in Developing Countries: Property Rights and Property Wrongs. Routledge Priorities in Development Economics, New York: USA. Macours, K., Janvry, A. and Sadoulet, E. 2010. Insecurity of property rights and social matching in the tenancy market. European Economic Review, vol. 54, no. 7, pp. 880−899.

Malla, G. 2008. Climate change and its impact on Nepalese agriculture. Journal of Agriculture and Environment, vol. 9, pp. 62‒71.

Mainuddin, M., and Kirby, M. 2009. Spatial and temporal trends of water productivity in the lower Mekong River Basin. Agricultural Water Management, vol. 96, no. 11, pp. 1567‒1578.

Maltsoglou, I., Dawe, D. and Tasciotti, L. 2010. Household-level impacts of increasing food prices in Cambodia. Food and Agriculture Organization of the United Nations. Rome: Italy.

Manski, C.F. 2004. Measuring expectations. Econometrica, vol. 72, no. 5, pp. 1329‒ 1376.

Markussen, T. 2008. Property rights, productivity, and common property resources: insights from rural Cambodia, World Development, vol. 36, no. 11, pp. 2277‒ 2296.

Markussen, T., Tarp, F., and Van Den Broeck, K. 2011. The forgotten property rights: evidence on land use rights in Vietnam. World Development, vol. 39, no. 5, pp. 839‒850.

Mârza, B., Angelescu, C., and Tindelche, C. 2015. Agricultural insurances and food security. The new climate change challenges. Procedia Economics and Finance, vol. 27, pp. 594‒599.

125

Matzkin, R.L. 2007. Nonparametric survey response errors. International Economic Review, vol. 48, no. 4, pp. 1411‒1427.

Maxwell, D. and Wiebe, K. 1999. Land tenure and food security exploring dynamic linkages, Development and Change, vol. 30, pp. 825‒849.

McFadden, D.L., Schwarz, N., and Winter, J. 2004. Measuring perceptions and behavior in household surveys. Paper presented at the Annual Meeting of the American Economic Association, San Diego, CA.

McFadden, D.L., Bemmaor, A.C., Caro, F.G. et al. 2005. Statistical analysis of choice experiments and surveys. Marketing Letters, vol. 16, no. 3, pp. 183‒196.

Mekong River Commission for Sustainable Development (MRC). 2016. Reports accessed at http://www.mrcmekong.org/topics/flood-and-drought/

Migot-Adholla, S., Hazell, H., Blarel, B., and Place, F. 1991. Indigenous land rights systems in Sub-Saharan Africa: a constraint on productivity? The World Bank Economic Review, vol. 5, no. 1, pp. 155‒175.

Migotto, M., Davis, B., Carletto, G., and Beegle, K. 2005. Measuring food security using respondents’ perception of food consumption adequacy. ESA Working Paper, no.05‒10, Food and Agriculture Organization of the United Nations.

Milne, S. 2013. Under the leopard's skin: land commodification and the dilemmas of indigenous communal title in upland Cambodia, Asia Pacific Viewpoint, vol. 54. no. 3, pp. 323‒339.

Ministry of Agriculture, Forestry and Fisheries (MAFF). 2013. Plan of Action for Disaster Risk Management in Agriculture 2014‒2018. Phnom Penh: Cambodia.

Ministry of Water Resources and Meteorology (MOWRAM). 2012. Climate change strategic plan for water resources and meteorology: 2013‒2017, Phnom Penh: Cambodia.

Mirza, M.M.Q. 2011. Climate change, flooding in South Asia and implications. Regional Environmental Change, vol. 11, no. 1, pp. 95‒107.

Mueller, N.D., Gerber, J.S., Johnston, M., Ray, D.K., Ramankutty, N., and Foley, J.A. 2012. Closing yield gaps through nutrient and water management. Nature, vol. 490, no. 7419, pp. 254‒257. Muriithi, B.W. and Matz, J.A. 2015. Welfare effects of vegetable commercialization: evidence from smallholder producers in Kenya. Food Policy, vol. 50, pp. 80−91.

Myers, R.J. 2006. On the costs of food price fluctuations in low-income countries. Food Policy, vol. 31, pp. 288‒301.

126

Namara, R.E., Hanjra, M.A., Castillo, G.E., Ravnborg, H.M., Smith, L, and Van Koppen, B. 2010. Agricultural water management and poverty linkages. Agricultural Water Management, vol. 97, no. 4, pp. 520‒527.

National Committee for Disaster Management (NCDM). 2013. Climate Change Strategic Plan for Disaster Management. Phnom Penh: Cambodia.

Naughton-Treves, L. and Wendland, K. 2014. Land tenure and tropical forest carbon management. World Development, vol. 55, pp. 1‒6.

Naylor, R.L. 2014. The Evolving Sphere of Food Security. Oxford University Press. New York: USA.

Neef, A., Touch, S., and Chiengthong, J. 2013. The politics and ethics of land concessions in rural Cambodia. Journal of Agricultural and Environmental Ethics, vol. 26, pp. 1085‒1103.

Newman, C., Tarp, F., and Broeck, K.v.d. 2015. Property rights and productivity: the case of joint land titling in Vietnam. Land Economics, vol. 91, no. 1, pp. 91‒105.

North, D.C. 1990. Institutions, Institutional Change, and Economic Performance. Cambridge University Press: Cambridge, MA.

OECD (2013), OECD guidelines on measuring subjective well-being, OECD Publishing, Paris. DOI: http://dx.doi.org/10.1787/9789264191655-en

Oldenburg, C. and Neef, A. 2014. Reversing land grabs or aggravating tenure security? Competing perspectives on economic land concessions and land titling in Cambodia, The Land and Development Review, vol. 7, no. 1, pp. 49‒77.

Panda, R.K., Behera, S.K. and Kashyap, P.S. 2003. Effective management of irrigation water for wheat under stressed conditions. Agricultural Water Management, vol. 63, no. 1, pp. 37‒56.

Parry, M., Rosenweig, C., Iglesais, A., Fischer, G., and Livermore, M. 1999. Climate change and world food security: a new assessment, Global Environment Change, vol. 9, pp. S51‒S67.

Pesaran, M.H. and Smith R. 1995. Estimating long-run relationships from dynamic heterogeneous panels.” Journal of Econometrics, vol. 68, no. 1, pp. 79–113.

Phann, D., Phay, S., Tong, K., and Pon, D. 2015. Landlessness and child labour in Cambodia. Cambodia Development Resource Institute, Final Report, Phnom Penh: Cambodia.

Piao, S., Ciais, P., Huang, Y., Peng, S., Li, J., Zhou, L., Liu, H., Ma, Y., Ding, Y., Friedlingstein, P., Liu, C., Tan, K., Yu, Y., Zhang, T., and Fang, J. 2010. The

127

impacts of climate change on water resources and agriculture in China. Nature, vol. 467, pp. 43‒51.

Piza, C. and de Moura M.J.S.B. 2016. The effect of a land titling programme on households’ access to credit. Journal of Development Effectiveness, vol. 8, no. 1, pp. 129‒155.

Place, F. 2009. Land tenure and agricultural productivity in Africa: a comparative analysis of the economics literature and recent policy strategies and reforms. World Development, vol. 37, no. 8, pp. 1326‒1336.

Place, F. and Hazell, P. 1993. Productivity effects of indigenous land tenure systems in Sub-Saharan Africa. American Journal of Agricultural Economics, vol. 75, no. 1, pp. 10‒19.

Place, F. and Migot-Adholla, S. E. 1998. The economic effects of land registration on smallholder farms in Kenya: evidence from Nyeri and Kakamega Districts, Land Economics, vol. 74, no. 3, pp. 360‒373.

Place, F. and Otsuka, K. 2002. Land tenure systems and their impacts on agricultural investments and productivity in Uganda. Journal of Development Studies, vol. 38, no. 6, pp. 105‒128.

Posner, R. A. 1998. Creating a legal framework for economic development, The World Bank Research Observer, vol. 13, pp. 1‒11.

Pradhan, P., Fischer, G., van Velthuizen, H., Reusser, D.E., Kropp, J.P. 2015. Closing yield gaps: how sustainable can we be? PLoS ONE, vol. 10, no. 6: e0129487. doi:10.1371/journal.pone.0129487

Pradhan, M. and Ravallion, M. 1998. Measuring poverty using qualitative perceptions of welfare. World Bank Policy Research Working Paper #2011.

Property Rights Alliance. International Property Rights Index (IPRI): 2007‒2011. Washington, D.C.: USA.

Queensland Government. 2011. Understanding Floods: questions & answers. The State of Queensland (2011). Queensland: Australia.

Ragab, R and Prudhomme, C. 2002. Climate change and water resources management in arid and semi-arid regions: prospective and challenges for the 21st century. Biosystems Engineering, vol. 81, no. 2, pp. 3‒34.

Ralph, K., Palmer, K., and Olney, J. 2012. Subjective well-being: a qualitative investigation of subjective well-being questions. Working Paper, the Technical Advisory Group.

128

Rasmussen, W.C. and Bradford, G.M. 1977. Ground-water resources of Cambodia. Contributions to the Hydrology of Asia and Oceania: Ecological Water-Supply Survey Paper 1608-P. USAID, Washington D.C.: USA.

Rights and Resources Initiative (RRI). 2015. Who Owns the World’s Land? A global baseline of formally recognized indigenous and community land rights. Washington, DC: RRI.

Rockström, J. 2003. Water for food and nature in drought-prone tropics: vapour shift in rain-fed agriculture. Philosophical Transitions of Royal Society B Biological Sciences, vol.358, issue#1440, pp. 1997‒2009.

Rockström, J., Karlberg, L., Wani, S.P., Barron, J., Hatibu, N., Oweis, T., Bruggeman, A., Farahani, J., and Qiang, Z. 2010. Managing water in in rainfed agriculture: the need for a paradigm shift. Agricultural Water Management, vol. 97, no. 4, pp. 543‒550.

Rodrik, D., Subramanian, A. and Trebbi, F. 2004. Institutions rule: The primacy of institutions over geography and integration in economic development. Journal of Economic Growth, vol. 9, no. 2, pp. 131‒165.

Rosegrant, M.W., Cai, X., and Cline, S.A. 2002. World water and food 2025: dealing with scarcity. International Food Policy Research Institute, Washington D.C.: USA.

Rosegrant, M.W., Ringler, C., and Zhu, T. 2009. Water for agriculture: maintaining food security under growing scarcity. Annual Review of Environment and Resources, vol. 34, pp. 205‒222.

Rosenzweig, C., Iglesias, A., Yang, X.B., Epstein, R.P., and Chivian, E. 2001. Climate change and extreme weather events: implications for food production, plant diseases, and pests. Global Change and Human Health, vol. 2, no. 2, pp. 90‒104.

Roth, M. 2010. Land tenure, property rights, and food security. USAID Property Rights and Resource Governance Briefing Paper #1.

Royal Government of Cambodia (RGC). 2007. Law on Water Resources Management of the Kingdom of Cambodia. Phnom Penh: Cambodia.

Royal Government of Cambodia (RGC). 2010. Policy Paper on the Promotion of Paddy Production and Milled Rice Export. Phnom Penh: Cambodia.

Royal Government of Cambodia (RGC). 2013. Rectangular Strategy for Growth, Employment, Equity and Efficiency Phase III. Phnom Penh: Cambodia.

Royal Government of Cambodia (RGC). 2014. National Strategic Development Plan 2014‒2018. Phnom Penh: Cambodia.

129

Rudi, L-M, Azadi, H., Witlox, F, and Lebailly, P. 2014. Land rights as an engine of growth? An analysis of Cambodian land grabs in the context of development theory. Land Use Policy, vol. 38, pp. 564‒572.

Sadoulet, E. and de Janvry, A. 1995. Quantitative Development Policy Analysis. John Hopkins University Press, USA.

Sampath, R.K. 1992. Issues in irrigation pricing in developing countries. World Development, vol. 20, no. 7, pp. 967‒977.

Sekiguchi, M. and Hatsunako, N. 2013. Land conflicts and land registration in Cambodia. In Land and Post-Conflict Peacebuilding, J. Unruh and R.C. Williams (ed.), London.

Seng, V., Craswell, E., Fukai, S., and Fischer, K. (eds). 2003. Water in Agriculture. Phnom Penh: Cambodia.

Sik, B. 2000. Land ownership, sales, and concentration in Cambodia: a preliminary review of secondary data and primary data from four recent surveys. Cambodia Development Resource Institute Working Paper #16.

Sjaastad, E. and Bromley, D.W. 1997. Indigenous land rights in sub-Saharan Africa: appropriation, security, and investment demand, World Development, vol. 24, no. 5, pp. 549‒562.

Smith, A. 1776. An inquiry into the nature and causes of the wealth of nations, London: United Kingdom.

Smith, D. and Hornbuckle, J. (eds). 2013. A review on rice productivity in Cambodia and water use measurement using direct and indirect methods on a dry season rice crop. Technical Report to ACIAR, CSIRO Sustainable Agriculture Flagship, Australia.

Smith, L., El Obeid, A., and Jensen, H. 2000. The geography and causes of food insecurity in developing countries. Agricultural Economics, vol. 22, pp. 199‒215.

Smith, R.E., 2004. Land tenure, fixed investment, and farm productivity: evidence from Zambia’s southern province. World Development, vol. 32, no. 10, pp. 1641‒1661.

So, S., Real, S., Uch, U., Sy R., Ballard, B., and Acharya, S. 2001. Social assessment of land in Cambodia: A field study. Working Paper no. 20, Cambodia Development Resource Institute: Phnom Penh, Cambodia.

Stern, D.I. 2010. Between estimates of the emissions-income elasticity. Ecological Economics, vol. 69, no. 11, pp. 2173‒2182. Sylwester, K. 2006. Income inequality, education expenditures, and growth. Journal of Development Economics, vol. 63, no. 2, pp. 379−398.

130

Timmer, P.C. 2000. The macro dimensions of food security: economic growth, equitable distribution, and food price stability. Food Policy, vol. 25, pp. 283‒295.

Timmer, P.C. 2005. Food security and economic growth: an Asian perspective. Asia- Pacific Economic Literature, vol. 19, pp. 1‒17.

Timmer, PC 2013. Food Security and Sociopolitical Instability in East and Southeast Asia, in C.B. Barrett (ed.). 2013. Food Security and Socio-political Stability. Oxford University Press, UK, pp. 452‒470.

Tinkler, L., and Hicks, S. 2011. Measuring subjective well-being. Office for National Statistics of the United Kingdom Statistics Authority. South Wales: UK.

Torhonen, M.P. and Groppo, P. (eds). 2006. Land Reform: Land Settlements and Cooperatives. Rural Development Division, Food and Agricultural Organization.

Toufique, A K. and Islam, A. 2014. Assessing risks from climate variability and change for disaster-prone zones. International Journal of Disaster Risk Reduction, vol. 10, pp. 236‒249.

Tully, J. 2005. A Short History of Cambodia: From Empire to Survival. South Wind Production, Singapore.

Turner, K., Georgiou, S., Clark, R., Brouwer, R., and Burke, J. 2004. Economic valuation of water resources in agriculture: from Sectoral to functional perspective of natural resource management. FAO Water Report #27, Rome: Italy.

United Nations Children’s Fund of the United Nations (Unicef). 2014. United Nations Report 2014: Levels and Trends in Child Mortality. Unicef (2014).

United Nations World Economic Situation and Prospects 2015. United Nations: New York (2015).

United States Aid for International Development (USAID). 2016. Agriculture and food security: Cambodia. Accessed from https://www.usaid.gov/cambodia/agriculture- and-food-security United States Department of Agriculture (USDA). 2010. CAMBODIA: Future Growth Rate of Rice Production Uncertain. Accessed at: http://www.pecad.fas.usda.gov/highlights/2010/01/cambodia/

United States Department of Agriculture (USDA). 2013. CAMBODIA: Seasonal flooding impacts wet-season rice production in 2013. Accessed at http://www.pecad.fas.usda.gov/highlights/2013/11/Cambodia/

United States Department of Agriculture (USDA). 2016. Southeast Asia: 2015/16 Rice Production Outlook at Record Levels.

131

Accessed at: http://www.pecad.fas.usda.gov/highlights/2015/06/Southeast_Asia/Index.htm

Unruh, J. and Williams, R.C. (eds). 2013. Land and Post-Conflict Peacebuilding. Environmental Law Institute and United Nations Environment Programme, London: Earthscan.

Vollrath, D. 2007. Land distribution and international agricultural productivity. American Journal of Agricultural Economics, vol. 89, no. 1, pp. 202‒216. von Braun, J., Puetz, D., and Webb, P., 1989. Irrigation technology and the commercialization of rice in the Gambia: effects on income and nutrition. Research Report 75. International Food Policy Research Institute, Washington, D.C.: USA.

Warr, P. 2014. Food security and its determinants. The Australian Journal of Agricultural and Resources Economics, vol. 58, pp. 519‒537.

Wassman, R., Jagadish, S.V.K. Heuer, S, Ismail, A, Redona, E, Serraj, R., Singh, R.K., Howell, G, Pathak, H, and Sumfleth, K. 2009a. Climate change affecting rice production the physiological and agronomic basis. Advances in Agronomy, vol. 101, pp. 59‒122.

Wassman, R., Jagadish, S.V.K., Sumfleth, Pathak, H., Howell, G., Ismail, A., Serraj, R., E. Redona, E., Singh, R.K., and Heuer, S. 2009b. Regional vulnerability of climate change impacts on Asian rice production and scope for adaptation. Advances in Agronomy, vol. 102, pp. 92‒133.

Webb, P., Coates, J., Frongillo, E.A., Rogers, B.L., Swindale, A. and Paula Bilinsky, P. 2002. Measuring household food insecurity: why it's so important and yet so difficult to do. Journal of Nutrition, vol. 13, no. 5, pp. 1404S‒1408S.

Wegren, S.K. 2013. Food security in the Russian Federation. Eurasian Geography and Economics, vol. 54, no. 1, pp. 22‒41.

Wegren, S.K., Nikulin, A.M., and Trotsuk, I. 2017. The Russian variant of food security. Problems of Post-Communism, vol. 64, no. 1, pp. 47‒62.

West, P.C., Gerber, J.S., Engstrom, P.M., Mueller, N.D., Brauman, K.A., Carlson, K.M., Cassidy, E.S., Johnston, M., MacDonald, G.K., Ray, D.K., and Siebert, S. 2014. Leverage points for improving global food security and the environment. Science, vol. 345, no. 6194, pp. 325‒328.

Wiebe, K. 2003. Linking land quality, agricultural productivity, and food security. Resource Economics Division, Economic Research Service, U.S. Department of Agriculture. Agricultural Economic Report #823.

132

Wokker, C., Santos, P., and Ros, B. 2014. Irrigation water productivity in Cambodian rice systems. Agricultural Economics, vol. 45, no. 4, pp. 421‒430.

Wooldridge, J.M. 2010. Econometric Analysis of Cross-Section and Panel Data. Second Edition, MIT Press, Massachusetts: USA.

World Bank. 2002. Project appraisal document on proposed credit in the amount of SDR 19.3 million to the Kingdom of Cambodia for the land management and administration project. Washington, D.C.: World Bank.

World Bank. 2003. Land policy for growth and poverty reduction. A World Bank Policy Research Report. Washington, D.C.

World Bank. 2014. World Development Indicators Online (WDI) database. Available at: http://data.worldbank.org/data-catalog/world-development-indicators

World Bank. 2015a. Cambodian agriculture in transition: opportunities and risks. Accessed at http://www.worldbank.org/en/news/feature/2015/08/19/cambodian- agriculture-in-transition-opportunities-and-risks

World Bank. 2015b. Cambodia: The agriculture, irrigation, and rural roads sectors -- Public Expenditure Review. Washington, DC: United States.

World Bank. 2015c. Doing Business database (accessed October 2015). Available at: http://www.doingbusiness.org/dat

World Bank. 2016. Living Standards Measurement Study - Integrated Surveys on Agriculture (LSMS-ISA). World Bank Development Research Group. Accessed at http://go.worldbank.org/BCLXW38HY0

World Bank. 2017. Land. World Bank Group’s Brief Report. Accessed at http://www.worldbank.org/en/topic/sustainabledevelopment/brief/land

Yagura, K. 2005. Imperfect markets and emerging landholding inequality in Cambodia. Japan Journal of Rural Economy, vol. 7, pp. 30−48.

Yang, S.K., Khim., S., and Seng, S. 2011. Irrigation development in Cambodia: challenges and opportunities. Accessed at https://trustbuilding.wordpress.com/2011/01/19/irrigation-development-in- cambodia-challenges-and-opportunities/

Zwart, S.J. and Bastiaanssen, W.G.M. 2004. Review of measured crop water productivity values for irrigated wheat, rice, cotton and maize. Agricultural Water Management, vol. 69, no. 2, pp. 115‒133.

133

APPENDICES

134

Appendix 1: Questionnaire for the household survey in Cambodia

HOUSEHOLD RESPONSE TO NATURAL DISASTERS The case of flood 2011 and 2013 in Cambodia Economic Research Institute of the ASEAN and Council for Agricultural and Rural Development

INTERVIEWER

Interviewer’s name Date interviewed Supervisor’s name Date revised

HOUSEHOLD INFORMATION

Household ID number Year experience in rice farming Name of respondent

Name of household head (in case that respondent is not head)

ADDRESS

Village/House No. Commune

District Province

Cell phone GPS locator

KEY HOUSEHOLD CHARACTERISTICS

1. Have your household cultivated rice over the past 5 years? (No=0, Yes=1), in no, stop interviewing (not suitable household) 2. Does your household currently own any plot of agricultural land? (No =0, Yes=1) 3. Was your household affected by the great flood 2011? (No=0, rice income loss=1, asset loss=2, both=3) If rice income affected, from what month to what month did you see flood water on your farm? (mm_mm) If rice income affected, what was the highest water level in your farm? (None=0, very little = 1, knee height=2, chest height= 3, more than chest height=4)

4. Was your household affected by the flood 2013? (No=0, rice income loss=1, asset loss=2, both=3) If rice income affected, from what month to what month did you see flood water on your farm? (mm_mm) If rice income affected, what was the highest water level in your farm? (None=0, very little = 1, knee height=2, chest height= 3, more than chest height=4)

5. Is your household classified as ID Poor? (No=0, ID Poor Class I=1, ID Poor Class II=2)

CONTENT 1. Plot characteristics 7. Asset 2. Rice production system 8. Financial transactions 3. Shocks, coping and risk expectation 9. Land ownership and conflicts 4. Household socioeconomics 10. Social capital and program participation 5. Income 11. Consumption 6. Preferences 12. Food insecurity

135

1. PLOT CHARACTERISTICS AND LAND USE BY HOUSEHOLD

Please provide land use information about all agricultural plots operated by your household. These can include plots you own and rent in Lan Lan Plot Soil Irrigat Main source Primary land use Flood 2011 Flood 2013 d d location problem ion of water size type in Crop Total Numb Average Aver Was When When How How Actual Actual Was When When How How Actual Actual cultivate er of yield per age the did did many many yield price the did did many many yield price 1=Rice crop price plot water plot 1 =Piped d ha per water days days harvest per kg water water days days harve per water 2=Maize areas seaso cropping per kg floode start flow out was the was the in 2011 sold in floode start flow out was was the st in kg n d? rising d? 1=Protec 0=No 2 =Undergro 3=Vegetab season of the crop crop 2011 rising at of the the crop 2013 sold 1=salty grown at the ted zone und le plot complet under the plot? plot crop under in 2=acid per plot? 2=Buffer 3 =Natural 4=Sugarc complet ely knee- (MM/DD complet compl knee- 2013 zone 3=lack ane year (MM/DD 0=No river/pond ely? submer deep ) ely? etely deep 3=Social nutrient 5=slope ) 4 =Man- (MM/DD ged water? (MM/DD subm water? economi 4=slope 1=Ye 5=other (kg/ha/c 5=other( made (kg/ha/c ) under (days) 0=No ) erged (days) c s (specify) 0=No rop) >> (kg/ha 0=In specify) river/po rop) water? under betwe >> skip skip to /crop) develop nd (days) water en ment 5 = to flood the 136 (ha) 1=lowl 2013 next ? zone Rainfall and (ha) 1=Yes section (days) 2=upl 4=Cleare 6 = 1=Yes and d forest Irrigated 5=Agricu water ltural zone

ERIA AND CARD HOUSEHOLD SURVEY/Page 1

2. RICE PRODUCTION SYSTEM INTRODUCTION: Let us understand rice production system and cropping season in your area. First, let us think of crop year starting from May‒April, that is the crop year 2013 would start from May 2013‒April 2014 (Enumerator: use timeline to explain this) Second, we would like to remind you of the following key events happening during the last three crop years from the current one. Those are mega flood 2011, flood 2013, government's introduction of policies on nationwide rice insurance and rice mortgage program. I hope you are familiar with them. (Enumerator explain) 2010 2011 2012 2013 2014 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jan Feb Mar Apr May Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan FebMar Apr May Jun Calendar Jun Jul Aug Sep Oct Nov Dec Jun Jul Aug Sep Oct Nov Dec Crop calendar Crop year 2010 Crop year 2011 (mega flood year) Crop year 2012 Crop year 2013 (another flood year) Mega floods Nationwide rice insurance program available Rice mortgage program available

Now, I would like to ask you to RECALL part or all your rice cropping seasons during crop year 2010 to the current crop year of 2014. Work from the current one first and back to the past ones

2010 2011 2012 2013 2014

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun 137 Enumerator: please mark crop interval in this table. And put planting month to the table below Now I would like you to RECALL the planting area size, average yield per ha and price per kg in each cropping season grown since 2010.

2010 2011 2012 2013 2014 Crop 1 Crop 2 Crop 1 Crop 2 Crop 1 Crop 2 Crop 1 Crop 2 Crop 1

Planting month (mm) Planting areas (ha) Average rice yield per ha per season (kg/ha/crop season) Average price sold per kg ERIA AND CARD HOUSEHOLD SURVEY/Page 2

3. SHOCKS ON RICE PRODUCTION, COPING AND RISK EXPECTATION

3.1 SHOCKS AND IMPACTS: Please tell me information about shocks that affected household's rice production since 2004 from the most recent to the past. Enumerator: Make sure to ask about flood 2011, 2013 Has your household done anything to Effect of shocks on rice income Effects of shocks on asset and debt Effect on consumption prepare to manage this shock ex ante? Shocks What are the 0=None Crop main assets 1=Have members Harvest change Price change doe income working outside Planting area destroyed/dam due to shock to shock loss due to agriculture aged due to shock 2=Purchase crop shock? insurance Total Affected Actual Expected Actual Expected 3=Saving more money planting areas harvested harvest in price income- How might Reduce 4=Accumulate asset (e.g., area in total total received Expected Actual shock Reduc Reduce gold) food 5=Crop diversification price income affect your e child medical 0=None consum 6=Decrease rice-growing How without loan school expense When in many 1=Housing ption areas shock repayment Time crop Length months 2=Productive ing? s? 7=Change rice planting Num Type of Name of Who were on? occurr cycle did of after non-farm assets Estimated ? patterns ber shock shock affected? ed the shock shock shock (e.g., vehicle, value of (planting date, variety,

occur? can you shops) chemical use) asset lost replant? 3=Productive 8=Construct irrigation 138 due to system farm asset (Rai) shock 9=Construct flood (Rai) (kg/crop) (kg/crop) (equipment, etc.) protection, etc. (Riel/kg) (Riel/kg) (Riel/crop) 4=Livestock 10=Other (specify) 5=Durable assets 1=Only your 0=No debt 1=Drough Please 1=Vegeta household 1=No effect t name the tive 2=The 2=Delay 2=Flood shock 2=Tillerin whole repayment Second 3=Insect (e.g., g village 3=Default (mmyy 0=No 0=No 0=No Most important most 4=Pest pest, 3=Floweri (day) (month) 3=The ) 1=Yes 1=Yes 1=Yes strategy important 5=Diseas insect, ng whole strategy e disease 4=Ripe commune 6=Others name, 5=Harves 4=The (specify) etc.) t whole province 01 02 03 04 05 06 07 08 ERIA AND CARD HOUSEHOLD SURVEY/Page 3

3. SHOCKS ON RICE PRODUCTION, COPING AND RISK EXPECTATION 3.2 COPING STRATEGIES: Please tell me how you prepare your households before shocks occurs and how you cope with shocks After shock, has your household done Has your household done the following to cope with shocks? Has your household received remittances and disaster reliefs after the shocks? Has your household anything to prepare for benefit from debt the future shock? holiday/relief following Increase labour Remittances/gifts Disaster aid from 0=None Increase use of natural Utilise Disaster relief from the shock? Asset sale used inside or Borrowing cash and in kind from NGOs/international 1=Have members resource saving government outside agriculture friend/relatives organisations working outside Clearing Increase Increa agriculture more collection se 2=Purchase crop forest of forest fishin Adequ insurance land? product? g? ate? 3=Saving more money 4=Accumulate asset How Total How 0=No 0=No (e.g., gold) many amou man 5=Crop diversification 1=Yes 1=Yes nt y Estimat Total Did the days 6=Decrease rice Increase Increase Total receiv mor Main type ed total Main source amount givers live since Main Source of growing areas child adult amount ed e of asset valve of loan of loan in this lost did source loan 7=Change rice labour? labour? 0=N received (both days sold borrowed village? you planting patterns o in-kind exte Num (planting date, variety, (mo Receiv and nde ber 0=No 0=No 0=No 0=No e it? 0=No 0=No chemical use) ve cash) d? (move (move (move (move (move (move 8=Construct irrigation 0=No to 1=Yes to the to the to the to the to the to the system the next next next next next next 9=Construct flood next catego catego catego catego catego catego protection, etc. ry) ry) cate ry) ry) ry) ry) 10=Other (specify) 139 1=Yes 1=Yes gory 1=Yes 1=Yes 1=Yes 1=Yes 1=Bank and ) 1=Bank/coo cooperative 1=Housing 1=Y peratives s 2=Productiv (Number es 0=No (Number 2=Relatives (Riel) e nonfarm of 1=Yes 1=Red of adults /friends 2=Relatives/f assets (e.g., children cross (Day Second who 3=Money riends Most vehicle, who 0=No 2=WFP ) most (Riel) become lender (Riel) (Day) 3=Money important shops) become 1=Yes 3=Other (Riel) important employed 4=MFI/infor lender strategy 3=Productiv employed NGOs strategy following mal saving 4=MFI/infor e farm asset following (specify) shock) shock) group m al saving 4=Durable 5=Other group assets (specify) 5=Other (specify)

08 ERIA AND CARD HOUSEHOLD SURVEY/Page 4

3. SHOCKS ON RICE PRODUCTION, COPING AND RISK EXPECTATION

3.3 SUBJECTIVE EXPECTATION ABOUT FLOOD RISK: Now I would like to ask your opinion about the potential of flood events and land insecurity happening in the coming 10 years from now. We will give you 10 coins. You will be asked to assign them to situations to reflect your thought of chances these situations will happen. The situation with larger number of coins reflect the situation that you feel most likely happen over the next 10 years. EXAMPLE: What's the likelihood that different flood events will occur over the next 10 years? Case A: No flood Mind flood with less than 10 days waterlogging Mega flood like that in 2011/2013 with more A: This is someone who foresee mega floods and less than knee high than 10 days waterlogging and higher than knee happening at the frequency of once every 5 years high in the future and foresee the mind flood occurring once every other year.

This person thus thinks his/her farm is quite prone to flooding in the future.

140

ERIA AND CARD HOUSEHOLD SURVEY/Page 5 No flood Mind flood with less than 10 days waterlogging Mega flood like that in 2011/2013 with more B: This is someone who does not foresee any and less than knee high than 10 days waterlogging and higher than mega flooding event in the next 10 years at all knee high and mind flood occurring 3-in 10 years. Thus, this person might not think his/her farm is prone to flood risk.

Enumerator: Test the household by asking them to explain the difference between A and B.

Test household's understanding: Ask household to allocate coin to reflect

C: the case when mega floods tend to occur at very high frequency

D: the case of highland with flood rarely occurs

Make sure household understands this before moving forward!

3.3 SUBJECTIVE EXPECTATION ABOUT FLOOD RISK (continued)

Subjective expectation of flood risk by plot: What's the likelihood that the following flood events will occur in the next 10 years from now in each of your plots currently operate. Please refer to plot numbers in table 1 and for each plot, assign coins into the events based on your expectation of their occurrence.

Please assign the coins to the flood events based on your opinion about the likelihood that the following floods will occur in the next 10 years future

No flood Mind flood with less than 10 days Mega flood like that in 2011/2013 with more than 10 waterlogging and less than knee high days waterlogging and higher than knee high

141

Plot number 1

(coins) (coins) (coins) 2 (coins) (coins) (coins) 3 (coins) (coins) (coins) 4 (coins) (coins) (coins) 5 (coins) (coins) (coins)

ERIA AND CARD HOUSEHOLD SURVEY/Page 6

3.3 SUBJECTIVE EXPECTATION ABOUT FLOOD RISK (continued) Subjective expectation of public, community support and reliance on natural resources for coping against flood events Please assign the coins to the events based on your opinion about the likelihood that the they will occur in the next 10 years future If the mind flood with less than 10 days waterlogging and less than knee high If the mega flood like that in 2011/2013 with more than 10 days waterlogging occurs, and higher than knee high occurs,

What's the likelihood

No loss Partial loss (1‒50% loss) Total loss (100% loss) No loss Partial loss (1-50% loss) Total loss (100% loss) that flood will affect your rice income?

142 (coins) (coins) (coins) (coins) (coins) (coins)

that your household will get Do not get gov. assistance Get gov. assistance Do not get gov. assistance Get gov. assistance disaster relief/assistance from government/NGOs? (coins) (coins) (coins) (coins)

that your household can rely on Do not get others' assistance Get others' assistance Do not get others’ assistance Get others' assistance social network, e.g., relatives, friends in your community for help? (coins) (coins) (coins) (coins)

that your household can rely on Cannot rely on nature Can rely on nature Cannot rely on nature Can rely on nature natural resources (forest product, fishing, forest land) to help smooth consumption? (coins) (coins) (coins) (coins)

None Some debt forgiven (1‒50%) All debt forgiven (100%) None Some debt forgiven (1-50%) All debt forgiven (100%) that your household will get debt relief/forgiveness? (coins) (coins) (coins) (coins) (coins) (coins)

ERIA AND CARD HOUSEHOLD SURVEY/Page 7

4. HOUSEHOLD MEMBERS AND CHARACTERISTICS Please provide the following information on all members. A person is counted as a household member if he/she lives here or has been absent for less than 12 months since January of last year General information Education Health Occupation Migration Impact of flood 2011 Impact of flood 2013 Name/Nickname Relationship to Sex Age Can he/she Is he/she If not in school, If in school, Health status How Does Primary Secondary How Main Did If he/she Did If he/she the head speak other currently what is the highest what is the many he/she occupation occupation many reason for he/she change job he/she change job languages in school? education? current days in work on days absence need to due to flood need to due to flood than Khmer? level of this year rice have be out of 2011, what's be out of 2013, what's education? that farming? he/she school primary school primary 1=Head 1 =Male (Years) 0=No 0 =No he/she 0 =No Occupation code: been 1=Work in due to occupation due to occupation 2=Spouse 2 =Female 1=French 1 =Yes cannot 1 =Full time 0 =Do nothing absent other flood before flood flood before flood 3=Son/Daughter 2=English 2 =Partial 1 =Rice farming 0=No education 1= 1=Healthy work due from district 2011? 2011? 2013? 2013? 4=Parent 3=Chinese time 2 =Other agri farming 1=Primary to home 2=Work in 5=Sibling 4=Vietnamese Primary 2=Occasionally 3 =Livestock 2=Lower 2=Lower sick sickness? during other 6=Grand child 5=Thai 4 =Fishery province 7=Nephew/Niece 6=Lao secondary secondary 3 =Frequently 5 =Wage labour in agri the past 3=Work in 8=In laws 7=Cham 3=Upper 3=Upper sick 6 =Wage labour in non-agri 12 (Days) other 9=Other relatives 8=Other local secondary secondary 4 =Disabled 7 =Commercial business months? 10=Servant language 4 =Vocational 4 =Vocational 8 =Government country 11=Other 5 =Higher 5 =igher 9 =Other salary jobs (Days) 4 =Go to nonrelatives 10=Studying school 11=Other (specify) 5 =Other (specify) 0 =No 0 =No Use Use 0 = No 0 =No occupation occupation 1 =Yes 1 =Yes code if code if change change 143 (1) (2) (6) (3) (3) (11a) (5) (7) (7) (5) (7) (5) (7) (7) (14) (5) (5) (7) (5) (7)

02 03 04 05 06 07 08 09 10 11 12 13 14 15 ERIA AND CARD HOUSEHOLD SURVEY/Page 8

5. CONSUMPTION 5.1 FOOD CONSUMPTION DURING THE LAST 7 DAYS: Please tell me about the household's food consumption during the last 7 days Unit Total value of food item Fraction of Impact of flood 2011 Impact of flood 2013 consumed last week purchased food 1 =Kg Total Price per Total value What's the How much did How much did How much did How much did 2 =Litre quantity unit of food proportion of total household household household consume household 3 =Big cup (lb) consumed consumed food consumed consume before consume after before flood 2013 in consume after 4 =Small cup that household flood 2011 in flood 2011 in proportion to total flood 2013 in (lb) All food Total purchased? proportion to total proportion to total food consumed in proportion to total 5 =Teaspoon consumed quantity × food consumed in food consumed in the past 7 days? food consumed in 6 =Tablespoon price per Food items including the past 7 days? the past 7 days? 1 = 3/4 less the past 7 days? 7 =Bottle ( lb) unit food 0 = None 2 = 2/4 less 8 =Count unit 1 = 3/4 less 1 = 3/4 less purchased, (Riels) 1 = 1/4 3 = 1/4 less 1 = 3/4 less 9 =Set meal 2 = 2/4 less 2 = 2/4 less produced 2 = 2/4 4 = Same 2 = 2/4 less 10=Other 3 = 1/4 less 3 = 1/4 less at home, 3 = 3/4 4 = Same 4 = Same 5 = 1/4 more 3 = 1/4 less (specify) gifts, free 4 = All 5 = 1/4 more 5 = 1/4 more 6 = 2/4 more 4 = Same collections 6 = 2/4 more 6 = 2/4 more 7 = 3/4 more 5 = 1/4 more 7 = 3/4 more 7 = 3/4 more 6 = 2/4 more 7 = 3/4 more Cereals (rice, bread, corn, wheat flour, rice flour, corn meal, rice cakes, noodles, biscuits, etc.) 01 Fish (fresh fish, salted and dried fish, canned fish, shrimp, prawn, crab, etc.) 02 Meat & poultry (beef, buffalo, mutton, lamb, pork, chicken, duck, innards, incl. liver, spleen, dried 03 beef) Eggs (chicken egg, duck egg, quail egg, fermented/salted egg, etc.) 04

144 Dairy products (fresh milk, condensed or powdered milk, ice cream, cheese, other dairy products, 05 etc.) Oil and fats (rice bran oil, vegetable oil, pork fat, butter, margarine, coconut/frying oil, etc.) 06 Fresh vegetables (trakun, onion, shallot, cabbage, spinach, carrot, beans, chili, tomato, etc.) 07 Tuber (cassava, sweet potato, potato, taro, sugar beet, etc.) 08 Pulses and legumes (green gram, dhal, cowpea, bean sprout, other seeds, etc.) 09 Prepared and preserved vegetables (cucumber pickles, other pickles, tomato paste, etc.) 10 Fruit (banana, orange, mango, pineapple, lemon, papaya, durian, water melon, grape, apple, 11 canned and dried fruits, etc.) Dried nuts and edible seeds (coconut, cashew nut, lotus nut, peanut, gourd seed, other nuts) 12 Sugar, salt and spices (sugar, jaggery, salt, chocolate, candy, coriander, red pepper spice, garlic, 13 ginger, soy sauce, fish sauce, monosodium glutamate, etc.) Tea, coffee, cocoa 14 Non-alcoholic beverages (canned or bottled soft drinks, mineral water, fruit juice, fruit syrup, etc.) 15 Alcoholic beverages (beer, wine, whisky, scotch, other distilled spirits) 16 Tobacco products (cigarettes, mild tobacco, strong tobacco, etc.) 17 Other food products (fried insects, peanut preparation, flavoured ice, ice, other food products) 18 Food taken away from home (meals at work, school, restaurants, snacks, coffee, soft drinks 19 purchased outside home) Prepared meals bought outside and eaten at home 20

ERIA AND CARD HOUSEHOLD SURVEY/Page 9

5. CONSUMPTION

5.2 NONFOOD CONSUMPTION: Please tell me about the household's non-food consumption

Total value (Riels) Impact of flood 2011 Impact of flood 2013 How much did How much did How much did How much did household household household household consume before consume after consume before consume after flood 2011 in flood 2011 in flood 2013 in flood 2013 in Total expenditure proportion to total proportion to total proportion to total proportion to total food consumed in food consumed in food consumed in food consumed the past 7 days? the past 7 days? the past 7 days? In-kind in the past 7 No. Items Time period Cash expenditure or days? expenditure gifts given away 1 = 3/4 less 1 = 3/4 less 1 = 3/4 less 1 = 3/4 less 2 = 2/4 less 2 = 2/4 less 2 = 2/4 less 2 = 2/4 less 3 = 1/4 less 3 = 1/4 less 3 = 1/4 less 3 = 1/4 less (in cash + in kind) 4 = Same 4 = Same 4 = Same 4 = Same 5 = 1/4 more 5 = 1/4 more 5 = 1/4 more 5 = 1/4 more 6 = 2/4 more 6 = 2/4 more 6 = 2/4 more 6 = 2/4 more 7 = 3/4 more 7 = 3/4 more 7 = 3/4 more 7 = 3/4 more

Health and medical care (doctors' fees, other medical services, drugs, hospital 01 charges, other medical supplies, etc.) Last 1 month

02 Education (school fees, textbooks, private tutoring charges, etc.) Last 12 months 145 Housing (rental cost per year, furniture and household equipment and operation e.g.,

03 curtain, household appliances, cooking utensils, light bulbs, soap and detergents etc.) Last 12 months Clothing and footwear (tailored clothes, ready-made clothes, rain clothes, underwear, 04 baby clothes, diapers, hats, shoes, boots, etc.) Last 6 months 05 Personal care (soap, toothpaste, razor, sanitary napkins, haircut, manicure, etc.) Last 1 month Transportation (personal transport equipment, operation of transport equipment, 06 maintenance and repair of equipment, gasoline and diesel for own transportation, fees Last 1 month for public transport, etc.) Communication (postage stamps, fax, telephone and internet phone charges, cell 07 phones, phone cards, internet charges etc.) Last 1 month

08 Domestic salaries (servant's salary, hired labour for cleaning, laundry, cooking etc.) Last 12 months Recreation (entertainment services, recreational goods and supplies, tourist travel, 09 hotel accommodation) Last 12 months Personal effects (costume/gold jewelry, handbags, wallets, wristwatch, clocks, 10 Last 12 months umbrella) Gambling (lottery, sports and animal betting: casino gambling, card games, football, 11 boxing, cockfighting etc.) Last 12 months 12 Miscellaneous items (special occasions as funeral rituals, weddings, parties, cash Last 12 months gifts, charity, etc.)

ERIA AND CARD HOUSEHOLD SURVEY/Page 10

6. INCOME

6.1 INCOME FROM RICE PRODUCTION: Please provide information on cost and income from rice production in 2013 crop season

Plot number Plot number Plot number Plot number Plot number Plot Particulars 01 02 03 04 05 GROWING AREAS AND VARIETY 1 Area (ha) (Copy from Table 1) 2 Rice variety 3 Method (1=transplanting seeding rice, 2=paddy-sown field) COSTS Land cleaning 4 Family labour (days) 5 Hired labour (days) 6 Daily wage (Riels) Land preparation 7 If mechanical was used, then rental charge (Riels) 8 Family labour (days) 9 Hired labour (days) 10 Daily wage (Riels) Crop establishment 11 Amount of seed (kg/ha) 12 Cost of seed (Riels) 13 Family labour (days)

14 Hired labour (days) 15 Daily wage (Riels) Fertiliser application 16 1st application: type (1=urea, 2=NPK, 3=Chemical, 4=Organic) 17 1st application: amount/ha 18 2nd application: type (1=urea, 2=NPK, 3=Chemical, 4=Organic) 19 2nd application: amount/ha 20 3rd application: type (1=urea, 2=NPK, 3=Chemical, 4=Organic) 21 3rd application: amount/ha 22 Total cost of Urea (Riels) 23 Total cost of NPK (Riels) 24 Total cost of other chemical fertilisers (Riels) 25 Total cost of organic fertiliser/green manure (Riels) 26 Family labour (days) 27 Hired labour (days) 28 Daily wage (Riels) Irrigation 29 Irrigation cost (Riels) 30 Irrigation fee paid to government (Riels)

ERIA AND CARD HOUSEHOLD SURVEY/Page 11

146

6.1 INCOME FROM RICE PRODUCTION: Please provide information on cost and income from rice production in 2013 crop season (continued) Weeding (Hand) 31 Cost of herbicide (Riels) 32 Family labour (days) 33 Hired labour (days) 34 Daily wage (Riels) Pesticide 35 Total cost of pesticide (Riels) 36 Family labour (days) 37 Hired labour (days) 38 Daily wage (Riels) Harvesting 39 If mechanical was used, then rental charge (Riels) 40 Family labour (days) 41 Hired labour (days) 42 Daily wage (Riels) Threshing, drying and transporting

43 Machine rental for mechanical threshing 44 Family labour (days) 45 Hired labour (days) 46 Daily wage (Riels) Other cash and in-kind expenditures, e.g. land rental, fuel, oil, etc. 47 Other cash expenditures, if any, for cultivation (Riels) 48 Other in-kind expenditures (imputed in Riels)

ERIA AND CARD HOUSEHOLD SURVEY/Page 12

147

6. INCOME

6.1 INCOME FROM RICE PRODUCTION (continued) Plot number Plot number Plot number Plot number Plot number Plot Particulars 01 02 03 04 05 CHANGES OF INPUT USE DUE TO FLOODS 2011/2013 How much differ would the following input used in crop year 2011 (before flood 2011), crop year 2012 (after flood 2011) and the upcoming crop year 2014 (after flood 2013)? crop crop crop crop crop crop crop crop crop crop crop crop crop crop crop year year year year year year year year year year year year year year year 2011 2012 2014 2011 2012 2014 2011 2012 2014 2011 2012 2014 2011 2012 2014 Seed variety (1=improved quality, 2=reduced 48 quality, 3=same)

49 Amount of seed (1=more, 2=less, 3=same)

50 Amount of fertiliser (1=more, 2=less, 3=same)

Amount of herbicide/pesticide (1=more, 2=less, 51 3=same)

148 52 Amount of irrigation (1=more, 2=less, 3=same)

53 Amount of labour (1=more, 2=less, 3=same) Amount of machine used - Organic (1=more, 54 2=less, 3=same)

RICE PRODUCTION INCOME

55 Total amount of crop produced (kg)

56 Total amount consumed within the household Total amount stored for future consumption 57 within the household

58 Total amount of crop sale

59 Average moisture level (%)

60 Price per kg (Riels)

Total income from rice (total amount of crop 61 produced*price) Enumerator: Cross check with household to confirm the amount of total income (note that this is not just from crop sale but from all production)

ERIA AND CARD HOUSEHOLD SURVEY/Page 13

6. INCOME RICE PRODUCTION 6.2 BUSINESS AND WAGE LABOUR ACTIVITIES: Please provide information and income of household members (code in Table 4) who are engaged in salary earning activities, paid farm labour, other casual/wage labour, business and self-employed

Years of How much How many months What's total cost per month in Member experience on he/she earn over the past 12 working on this job? Actual activity code this job per month? month he/she worked?

1 =Salary earner (regular pmt) 2 =Casual work (irregular pmt) 3 =Wage from farm labour 4 =Wage from forestry activities Cost includes only operational costs (fuel, goods, hiring labour, (See from 5 =Farm business/enterprise (Years) (Riels/month) (Month) transportation, etc.). This exclude Table 4) 6 =Non-farm fixed cost or capital. business/enterprise7 = Small (Riels/month) petty trading (goods not produced by household) 8 =Other (specify)

6.3 OTHER INCOME: Please provide information on all other sources of income your whole household earned over the past 12 months

Code Income source Riel

[h1] Income from agriculture (excluding cost)

M11-h1-2 Income from other crop production Income from livestock Income from fish cultivation/trapping of aquatic product Income from forestry and hunting [h2] Rental income M11-h2-1 Income from renting agricultural asset M11-h2-2 Income from renting non-agricultural asset [h2] Pension, remittances and others M11-h2-1 Pension M11-h2-2 Remittances from relatives/friends not currently a household member M11-h2-3 Government scholarships/stipends M11-h2-3 Other government/NGO welfare programs M11-h2-5 Income from lottery and gambling M11-h2-6 Other (specify) Code Expenses source Riel

M11-h3-1 Total expenditure of other agricultural production M11-h4-2 Other (specify) ERIA AND CARD HOUSEHOLD SURVEY/Page 14

149

7. FINANCIAL TRANSACTIONS 7.1 BORROWING: Does your household current have outstanding debt? This include debts from past borrowing that have not been paid in full and debt from borrowing over the past 12 months (since April 2013)? 1=Yes 0=No >> go to 7.2

Reason for not repaying in full (for Source of loan Reason for borrowing Collateral outstanding debt only) 1=Relatives 1 =Buy agricultural inputs 0 =No need 1 =Not due yet 2=Friends/neighbours 2 =Buy agricultural land, equipment 1 = Land 2 =Wait to sell production 3=Landlord/miller 3 =Invest in non-agricultural business 2 =Other asset Loan outstanding Interest 3 =Lost/reduced 4=Trader When did you 4 =Education expenditure 3 = Production as of today rate PER production 5=Money lender obtain this loan? 5 =Household consumption and durable 4 =Other person (principle + YEAR 4 =Need other use of Total amount of loan interest) 6=Saving/credit group you borrow 6 =Health shock: Illness, injury, accident 5 =Group money 7=Cooperatives 7 =Cope with agri shocks (disease, flood) 5 =Debt postpone/restructure 8=Other farmer group 8 =Rituals (marriage ceremony, funeral) 6 =Debt relief 9=Agricultural bank (gov.) 9 =Housing purchase/improvement 10=Commercial bank (private) 10 =Repay existing debts

150 11=Village funds, govt program 11=Other (specify) 12=NGO (non-profit and profit) Riels (mm/yy) Percent/year Riels For outstanding debt resulted from loan taken BEFORE April 2013 01 02 03 For outstanding debt resulted from loan taken during the past year (April 2013-April 2014) 04 05 06 Borrowing/credit constraint:

Would you like to borrow more? (0=No 1=Yes) If you would like to borrow more, can you borrow? (0=No, 1=Yes) What might be key reason of difficulty of accessing loan? (1=too much debt already, 2=already default loan, 3=no collateral/guarantor, 4=cumbersome with paperwork, 5= very few FN institutions, 6= do not know how to obtain credit, 7=likely get rejection anyway, 8=other(specify))

ERIA AND CARD HOUSEHOLD SURVEY/ Page 15

7.2 LENDING: Does your household have any outstanding receivables today? This could result from lending before or during April 2013. 1 = Yes 0 = No >> go to 7.3

Borrower Purpose for which loan Interest Cumulative receivable 1 =Relatives When did you lend out this taken rate PER outstanding Code as reason for YEAR (principle + interest) 2 =Friends/neighbours Total amount of loan loan? 3 =Other (specify) you lend (Riels) (mmyy) borrowing above Percent/year For outstanding loan lent out BEFORE

April 2013 01

151 02

03 For outstanding loan lent out during (April 2013-April 2014) 04 05 06

ERIA AND CARD HOUSEHOLD SURVEY/Page 16

7.3 SAVING: Does your household have accumulated saving as of today? This could result from saving before Does your household participate in saving group or ROSCA (Rotating Organisation for Saving and Credit)? 1 =Yes 0 =No >> go to 7.4

April 2013 or during April 2013‒2014. 1 =Yes 0 =No >> go to the left Saving group ROSCA Saving institution Total amount of Interest Name of group 1 =Saving/credit group saving as of 2 =Cooperatives rate PER How many people participate in the group 3 =Other farmer group today What are the 3 main occupations of 4 =Agricultural bank (gov.) YEAR 5 =Commercial bank (private) participants? 6 =Village fund (1 =farmers, 2=salary earner, 3=casual work, 4=farm labour, Riels Percent/year =Other govt/NGO programs 5=farm business/enterprise, 6=non-farm business/enterprise, 8 =Other (specify) 7=small petty trading) 1st 2nd 3rd 1st 2nd 3rd

For saving accumulated BEFORE April 2013 When did you join and leave the group?

01 (mmyy_mmyy)

02 What are frequency of payment/saving contributions (1=weekly, 2=fortnightly, 3=monthly, 4=other (specify) 03 How much you have to contribute/pay per time?

For saving accumulated during April 2013-April 2014 How much did/will you get from the group?

152 04 Overall, what's the main benefit to you of this saving

05 scheme? (1=saving, 2=credit for investment, 06 3=insurance, 4=other (specify)

Were these groups affected by flood? (0=No,1=Temporary stop,2=Permanently dissolved) z 7.4 INSURANCE: Does your household have active insurance contract as of today? 1 =Yes 0 =No If Yes, What type of insurance policy do you have? (1=health, 2=car, 3=homeowner, 4=crop, 5=livestock, 6=other, specify) 1st 2nd 3rd Who pay for the premium (1=household, 2=government/NGO subsidise, 3=government/NGO provide for free) 1st 2nd 3rd If No, are you interested and willing to pay for insurance policies? (0=No, 1=interested in health insurance, 2=in car insurance, 3=in home insurance, 4=in crop/livestock insurance, 5=Other, specify)

ERIA AND CARD HOUSEHOLD SURVEY/Page 17

8. ASSET Please provide information on total amount and current value of your productive and durable asset owned by your household

No Asset Total Total No Asset Total Total value if number value if No Asset Total Total value number sold now sold now numb if sold now (Unit) er (Riels) (Unit) (Riels) (Riels) (Unit) Durable assets

Livestock Non-agricultural equipment (productive asset) 1 Cattle House (excluding 26 Radio/stereo Buffaloes 21 land) 2 27 Television (TV) 3 Pigs 22 Shop 28 Telephone 4 Goats Van/car/jeep/pick 23 up 5 Chickens 29 Cell phone 24 Motorcycle Ducks 6 30 Computer 7 Other (specify) 25 Bicycle 31 Camera (picture/video) 26 Boat Agricultural equipment (productive asset) 32 Satellite dish Cart (pulled by 27 Fishing machine 8 animals) 28 Other 1 (specify) 33 Sewing machine Plough machine 9 (small 4 wheel) 29 Other 2 (specify) 34 Refrigerator/freezer

153 Plough machine 10 (large 4 wheel) Non-agricultural land (Ha) (Riels) 35 Electric Kitchen/Gas Stove

11 Tractor 30 Land for housing 36 Washing machine Instruments for Land for non- irrigation (pump, 31 agricultural activities 37 Vacuum cleaner 12 etc.) 38 Electric fan Planting 13 machine 39 Air conditioner Harvesting 14 machine 40 Sofa set 15 Sprayers 41 Dining set (dining table + Threshing chairs) 16 machine 42 17 Winnower Bed sets (Bed, Mattress…) 43 Rice mill/crop Other 1 (specify) 18 storage 44 Livestock/fishery Other 2 (specify) 19 facilities

20 Other (specify)

ERIA AND CARD HOUSEHOLD SURVEY/Page 18

9. LAND OWNERSHIP AND DISPUTE

9.1 LAND OWNERSHIP: I would now like to ask you about all agricultural land owned or operated by your household. These include plots household operated (reported in Table 1) and plots household owns but rent out or does not operate. Ownership If you own the plot If rent in or Have you made sharecrop any investments status on this plot since In what year How did you acquire it? Do you have and what type Can you use If rent out, how much If rent in or you operate? did you start of paper to certify your this plot as rent you charge or pay sharecropping, how much to own this 1 =Given by the government ownership? collateral for per month? And how rent do you pay per In what year How much Plot plot? (social land concession) loan? many months of renting month? And how many 0 did you first 1 =Own and use would it cost =None 2 =Inheritance or gift from size move to and 2 =Own but rent 0 = Do not have contract? months of renting to buy a plot 1 =Digging well relatives 2 =Digging start to use out 1 = Land investigation paper 0 =No Rent Term contract? like this in 3 =Rent 3 =Bought it from a relative 1 =Yes ditch/canal this plot? 2 = Certificate (title) from Rent Term this village 4 =Sharecropping 4 =Bought it from a non- 2 =Don’t 3 =Building flood government today? (pay Year relative know/not dyke (Year) (Riels/month) (Months) (Riels) 5 =Cleared land/occupied for 3 = Paper from local 4 =Drainage rent by production) sure (Riels/month) (Months) free authority construction (ha) 5 =Use for free 6 =Donated by friend 4 = Application receipt 5 =Soil reclamation

7 =Given by government 6 = Establish fruit 5 = Other (specify) (other program: specify) 6 = Don't know/not sure and nut trees 8=Other (specify) 7 =Other (specify) For plots that your household currently operates (refer to specific plot number in table 1)

154

5 For plots that your household currently own BUT rent out or do not operate 6 7 8 ERIA AND CARD HOUSEHOLD SURVEY/Page 19

9.2 LAND CONFLICT: Please fill out the detailed information on conflicts on the current plots your household own or operate and on plots already lost in the past 5 years

Type of conflict Conflict results In your opinion, what is the likelihood What year 1=Grabbed by that this plot will be with conflict and did it authorities taken away? Have you happen? 2=Grabbed by soldiers/ How long ever had armed officials did it take 1 =Very likely (>80%) (Year) 3=Grabbed by wealthy to resolve 2 =Likely (50%-80%) conflict on elites 3 =Maybe (30%-50%) the this plot? 4=Ownership conflict 4 =Less likely but possible (0-30%)

with non-relatives conflict? 5 =Not possible 0 =No 5=Ownership conflict 6 =Do not know 1 =Yes with relatives (Years) 6=Other (specify) Plot Land not able to 155 size Land lost operate on (ha) (ha) (ha)

Flor plots current own or operate on (refer to specific plot number from the above table)

2 3

4 5 6

7 8 For plots already lost due to conflict over the past 5 years

9 10

ERIA AND CARD HOUSEHOLD SURVEY/Page 20

9. LAND OWNERSHIP AND DISPUTE

9.3 SUBJECTIVE EXPECTATION OF RISK OF LAND INSECURITY: What's the likelihood that your land will in each of your plots will be grabbed by government or lost due to conflicts?

We will once again give you 10 coins. You will be asked to assign them to many different land grabbing situations to reflect your thought of chances these situations will happen. Again, the situation with larger number of coins reflect the situation that you feel most likely happen over the next 10 years.

Please assign the coins to the following land insecurity events based on your opinion about the likelihood that they will occur with your plots in the next 10 years future

No land will be 1/4 of land will be 2/4 of land will be 3/4 of land will be All land will be grabbed by grabbed by authority grabbed by authority or grabbed by authority grabbed by authority authority or taken away due 156 or taken away due taken away due to or taken away due to or taken away due to Plot to conflict

number to conflict conflict conflict conflict 1 (coins) (coins) (coins) (coins) (coins) 2 (coins) (coins) (coins) (coins) (coins) 3 (coins) (coins) (coins) (coins) (coins) 4 (coins) (coins) (coins) (coins) (coins) 5 (coins) (coins) (coins) (coins) (coins)

ERIA AND CARD HOUSEHOLD SURVEY/Page 21

10. SOCIAL CAPITAL, PARTICIPATION IN GOVERNMENT PROGRAMS AND OTHER SHOCKS

10.1 RELATIONSHIP WITH LOCAL OFFICALS AND POLITICIANS: Does any of your household member have relationship with member of the local officials/politicians? Governmental ranking Name of the Can he/she ask for Have he/she done Member Type of of the person he/she person he/she some favour from any favor to this code relationship knows knows this person? person? 1 =Village head 1 =Closed relative 0 =No 0 =No 2 =Commune head 2 =Distant relative 1 =Yes 1 =Yes 3 =Provincial head 3 =Friend 2 =Not sure 2 =Not sure 4 =Member of village 4 =Acquaint committee 5 =Business partner (See Table 4) 5 =Member of commune 6 =Other (specify) committee 6 =Member of provincial committee7 = Other (specify)

10.2 PARTICIPATION IN GOVERNMENT PROGRAM: Have your household and/or members participated in any of the government programs? When has your When has your household Program Have you benefited from the Government program household stopped participation? list program? participated? 0 =No (mmyy) (mmyy) 1 =Yes 2 =Not sure 1 Economic land concession 2 Social land concession

3 Have an IDPoor Card

4 Have other Social Equity Card

5 6

10.3 OTHER SHOCKS: Has this household faced any other shocks/accidents (not related to rice production) since 2004? Please tell me the most important shocks seriously affected your household since 2004. Choose from the code below Code of shocks: 1= Dead of earning member 2004 2010 2= Serious illness/major accident of member 3=Unemployment of earning member 2005 2011 4=Other natural disasters (typhoons, cyclone, etc.) 2006 2012 5=Livestock mortality and other covariate shocks 6=Theft/fire affecting property 2007 2013 7=Decline in production prices 8=Increase in consumption prices 2008 2014 9=Increase in input costs 2009

ERIA AND CARD HOUSEHOLD SURVEY/Page 22

157

11. SOCIAL CAPITAL, PARTICIPATION IN GOVERNMENT PROGRAMS AND OTHER SHOCKS

11.1 SUBJECTIVE MEASURE OF FOOD INSECURITY: Please answer the following questions

Did your family use iodised salt, yesterday? Ask the respondent for a teaspoon full of cooking salt and test for iodine (1 = Iodine present, 2=No iodine, 3=No salt in the household) In the last 12 months, has this household had enough food all days or were there days and weeks with very little or no food so that the household had almost starved (‘was hungry’)? (1=Enough food all the last 12 months (move to 10.2), 2=Not enough food) For those with not enough food: How many of the last 52 weeks did the household have so little food that it was hungry and went to bed without food (or ‘almost starving’)? (weeks)

11.2 OBJECTIVE MEASURES OF FOOD INSECURITY: Now I would like to measure health condition of you and children living in your household.

Household head Child 1 Child 2 Child 3 Child 4 Weight (kg): Height (cm): MUAC (cm)

ERIA AND CARD HOUSEHOLD SURVEY/Page 23

158

12. PREFERENCES 12.1 TRUST: 1. Would you say that most of the time people ______(1=Try to be helpful, 2=Are just looking out for themselves, 3=No idea) 2. Generally speaking, would you say that most people ______(1=Can be trusted, 2=Can't be trusted, 3=No idea) 3. Do you think most people ______(1=Would take advantage of you, 2=Would try to be fair, 3=No idea)

4. Now I would like to know how much you trust different groups of people. How much do you feel you can trust the people in the group? (For each, choose among 1=They can be trusted, 2=They cannot be trusted, 3=No idea) 4.1 People in your family 4.5 Business owner/traders you buy things from or do business with 4.2 People in your village/neighbourhood 4.6 Village/local government 4.3 People from other tribes/ethnic groups/religions living in the same community you 4.7 Judges/courts/police belong to

159 4.4 People from other tribes/ethnic groups/religions living outside community you belong to 4.8 Government services (e.g., education, health, electricity, water) 12.2 RELATIVE ECONOMIC POSITION IN THE VILLAGE, SATISFACTION AND HAPPINESS: 1. Relative position: In your own assessment what category would you put your household, compared to the conditions of other households in your village? (1='Rich', 2='Average', 3='Poor', 4='Very poor') 2. Life satisfaction: All things considered, how satisfied are you with your life these days as a whole? If 1 means you are completely dissatisfied on this scale and 10 means you are completely satisfied, where would you put your satisfaction with your household's life? 3. Financial satisfaction: How satisfied are you with your financial situation of your household? If 1 means you are completely dissatisfied on this scale and 10 means you are completely satisfied, where would you put your satisfaction with your household's financial situation? 4. Happiness: Taking all things together, would you say you are (1=Very happy, 2=Quite happy, 3=Not very happy, 4=Not at all happy)

EXPERIMENTS TO BE CONDUCTED AMONG GROUP OF HOUSEHOLDS BEFORE THE INTERVIEW Framed experiments: 12.3 ALTRUISM (through subjective dictatorship game) 12.4 RISK PREFERENCE (through framed Binswanger game framed to reflect seed options with different risk and return - need to know mean return) 12.5 TIME PREFERENCE (through simple framed discounting game) 12.6 PUBLIC GOODS GAME (framed to contribution to irrigation/water project and another one to be decided) Game experiments: 12.7 RISK PREFERENCE (through simple card game)

12.8 CONVEX TIME BUDGET (through a more complicated series of card games)

ERIA AND CARD HOUSEHOLD SURVEY/Page 24

Appendix 2: Test results excluding China and random and fixed effects results in Chapter 4

Table A4.1: OLS results over 1990‒2011 (excluding China) Dependent variable Ln PU Ln PFI Ln DFD Ln ADESA Independent variables Ln property rights protection (PRP) -0.404*** -0.267*** -0.216 0.046*** (0.097) (0.083) (0.132) (0.017) Ln household final consumption -0.540*** -0.531*** -0.941*** 0.112*** expenditure per capita (HHFCEPC) (0.059) (0.053) (0.090) (0.010) Ln population (POP) -0.178*** -0.193*** -0.386*** 0.045*** (0.033) (0.029) (0.057) (0.006) Agriculture share of GDP -3.344*** -3.671*** -6.548*** 0.688*** (AGCOM) (0.600) (0.552) (0.942) (0.108) Percent of agricultural land -0.0005 -0.002 -0.005*** -0.0001 (AGRILD) (0.001) (0.001) (0.001) (0.0002) Latitude (LATIT) -0.875*** -0.737*** -1.260*** 0.196*** (0.299) (0.232) (0.345) (0.044) Ln openness (OPEN) -0.279*** -0.259*** -0.505*** 0.043*** (0.083) (0.071) (0.115) (0.014) Ln human capital (HC) -0.544*** -0.661*** -1.257*** 0.103*** (0.184) (0.149) (0.246) (0.032) Ln inflation (INFL) -0.036 -0.005 0.015 -0.007** (0.028) (0.018) (0.025) (0.003) Ethnolinguistic fractionalisation 0.210* 0.224** 0.425*** -0.052** (ETHNO) (0.112) (0.097) (0.154) (0.026) Year & Region Effects Yes Yes Yes Yes N 531 531 531 531 Countries 56 56 56 56 R2 0.663 0.716 0.715 0.759 Notes: ***, **, * denotes a significance level at 1%, 5% & 10%, respectively. Robust standard errors in parentheses. PU is prevalence of undernourishment; PFI is prevalence of food inadequacy; DFD is depth of food deficit; ADESA is average dietary energy supply adequacy.

160

Table A4.2: Between Estimator panel results (1990‒2011) (excluding China) Dependent variable Ln PU Ln PFI Ln DFD Ln ADESA Independent variables Ln property rights protection (PRP) -0.857** -0.660* -0.749 0.148** (0.382) (0.336) (0.536) (0.068) Ln household final consumption -0.368** -0.318** -0.599** 0.073** expenditure per capita (HHFCEPC) (0.176) (0.155) (0.248) (0.031) Ln population (POP) -0.074 -0.097 -0.212** 0.021 (0.072) (0.064) (0.102) (0.013) Agriculture share of GDP -1.276 -1.496 -2.362 0.333 (AGCOM) (1.235) (1.088) (1.733) (0.221) Percent of agricultural land -0.001 -0.002 -0.007 -0.0001 (AGRILD) (0.004) (0.004) (0.006) (0.0007) Latitude (LATIT) -0.524 -0.508 -0.595 0.177 (0.740) (0.652) (1.039) (0.132) Ln openness (OPEN) -0.125 -0.132 -0.204 0.022 (0.276) (0.243) (0.387) (0.049) Ln human capital (HC) -0.07 -0.232 -0.369 0.022 (0.608) (0.535) (0.853) (0.109) Ln inflation (INFL) -0.028 0.041 -0.050 -0.025 (0.235) (0.207) (0.330) (0.042) Ethnolinguistic fractionalisation -0.244 -0.150 -0.123 0.041 (ETHNO) (0.301) (0.265) (0.423) (0.054) North Africa & the Middle East -0.720** -1.032*** - 0.170*** dummy (DNAME) (0.332) (0.293) 1.545*** (0.059) (0.467) Sub-Saharan African dummy 0.445 0.232 0.208 -0.078 (DSSA) (0.301) (0.265) (0.423) (0.054) East & SE Asian dummy 0.313 0.261 0.523 -0.032 (DESEA) (0.343) (0.302) (0.482) (0.061) South Asian dummy 0.301 0.256 0.558 -0.052 (DSA) (0.364) (0.321) (0.511) (0.065) N 531 531 531 531 Countries 56 56 56 56 R2 0.675 0.700 0.691 0.735 F 6.078 6.820 6.551 8.104 Notes: ***, **, * denotes a significance level at 1%, 5% & 10%, respectively. Standard errors in parentheses. Region fixed effects are controlled for by adding regional dummies. Latin America & the Caribbean dummy (DLAC) is used as the base, so it is dropped from the model in the estimations. PU is prevalence of undernourishment; PFI is prevalence of food inadequacy; DFD is depth of food deficit; ADESA is average dietary energy supply adequacy.

161

Table A4.3: Robustness OLS results: Prevalence of undernourishment and property rights measures Dependent variable Ln prevalence of undernourishment (PU) Independent variables Period 1990 ‒ 2011 1995‒2011 2007‒2011 2005‒2011 2003‒2011 2005‒2011 2007‒2011

Ln property rights protection (PRP) -0.404*** (0.097) Ln property rights (PR) -0.202*** (0.050) Ln physical property rights score 0.162 (PPRS) (0.220) Ln property rights & rule-based -0.421** governance (PRRG) (0.189) Ln regulatory restrictions on sale of 0.114** property (RRRP) (0.055) Ln registering property (REGPR) 0.132

162 (0.104) Ln international property rights -0.912***

index (IPRI) (0.250) ………………….....

(table continues below) N 531 763 175 172 416 347 175 Countries 56 61 41 28 60 60 41 R2 0.663 0.644 0.749 0.549 0.650 0.651 0.771 F n.a. 19.96 21.14 5.322 23.69 29.78 23.01 Notes: ***, **, * denotes a significance level at 1%, 5% & 10%, respectively. Robust standard errors in parentheses. The F-statistics in Column (1) are not available and so are not reported. The number of countries in observations and the time periods vary depending on the availability of data for respective indicators of property rights.

Table A4.3: Robustness OLS results: Prevalence of undernourishment and property rights measures (Continued) Dependent variable Ln prevalence of undernourishment (PU) Independent variables Period 1990 ‒ 2011 1995 ‒2011 2007‒2011 2005‒2011 2003‒2011 2005‒2011 2007‒2011 (continued from table above)

Ln household final consumption -0.540*** -0.649*** -0.700*** -0.604*** -0.616*** -0.579*** -0.498*** expenditure per capita (HHFCEPC) (0.059) (0.046) (0.106) (0.088) (0.056) (0.063) (0.089) Ln population (POP) -0.178*** -0.072*** -0.151*** -0.069** -0.115*** -0.069*** -0.123** (0.033) (0.021) (0.049) (0.030) (0.028) (0.025) (0.056) Agriculture share of GDP -3.344*** -2.310*** -4.373*** -1.641*** -2.247*** -1.908*** -4.109*** (AGCOM) (0.600) (0.332) (1.177) (0.453) (0.425) (0.462) (1.099) Percent of agricultural land -0.0005 -0.003*** 0.005 -0.0005 0.0001 -0.0003 0.005 (AGRILDPC) (0.001) (0.0007) (0.003) (0.001) (0.001) (0.001) (0.003) Latitude (LATIT) -0.875*** -0.267 -1.632*** -0.667** -0.474* -0.285 -1.728*** (0.299) (0.215) (0.513) (0.289) (0.274) (0.287) (0.478)

162 163 Ln openness (OPEN) -0.279*** -0.115* -0.257** -0.205* -0.277*** -0.144* -0.156

(0.083) (0.064) (0.121) (0.110) (0.087) (0.085) (0.125) Ln human capital (HC) -0.544*** -0.182 -0.433 0.287 -0.327 -0.196 -0.477 (0.184) (0.143) (0.336) (0.243) (0.217) (0.234) (0.314) Ln inflation (INFL) -0.036 -0.155*** 0.666* 1.668*** 1.291*** 1.754*** 0.688* (0.028) (0.046) (0.392) (0.368) (0.318) (0.316) (0.367) Ethnolinguistic fractionalisation 0.210* -0.178** 0.422** 0.121 0.049 -0.009 0.464*** (ETHNO) (0.112) (0.070) (0.168) (0.116) (0.104) (0.113) (0.175) Year & Region Effects YES YES YES YES YES YES YES N 531 763 175 172 416 347 175 Countries 56 61 41 28 60 60 41 R2 0.663 0.644 0.749 0.549 0.650 0.651 0.771 F n.a. 19.96 21.14 5.322 23.69 29.78 23.01 Notes: ***, **, * denotes a significance level at 1%, 5% & 10%, respectively. Robust standard errors in parentheses. The F-statistics in Column (1) are not available and so are not reported. The number of countries in observations and the time periods vary depending on the availability of data for respective indicators of property rights.

Table A4.4: Random effects panel results (1990‒2011) Dependent variable Ln PU Ln PFI Ln DFD Ln ADESA

Independent variables Ln property rights protection (PPR) 0.063 0.028 0.022 0.0003 (0.069) (0.051) (0.082) (0.012) Ln household final consumption -0.370*** -0.310*** -0.485*** 0.081*** expenditure per capita (HHFCEPC) (0.099) (0.085) (0.122) (0.019) Ln population (POP) -0.060 -0.079 -0.122* 0.032*** (0.051) (0.049) (0.073) (0.009) Agriculture share of GDP -1.358 -1.222 -2.151* 0.282 (AGCOM) (0.984) (0.818) (1.164) (0.181) Percent of agricultural land -0.001 -0.0006 -0.001 0.0002 (AGRILD) (0.002) (0.002) (0.003) (0.0004) Latitude (LATIT) -1.411* -1.582** -2.297** 0.344*** (0.732) (0.692) (1.125) (0.129) Ln openness (OPEN) -0.125 -0.079 -0.155 0.013 (0.110) (0.080) (0.120) (0.021) Ln human capital (HC) -0.939* -0.76* -0.880 0.225*** (0.489) (0.388) (0.555) (0.079) Ln inflation (INFL) -0.981* -0.789* -0.935 0.232*** (0.513) (0.405) (0.587) (0.082) Ethnolinguistic fractionalisation 0.006 0.013 0.049 -0.001 (ETHNO) (0.018) (0.021) (0.044) (0.005) N 545 545 545 545 Countries 57 57 57 57 R2 n.a. n.a. n.a. n.a. Notes: ***, **, * denotes a significance level at 1%, 5% & 10%, respectively. Robust standard errors in parentheses. F-statistics are not available, so they are not reported. PU is prevalence of undernourishment; PFI is prevalence of food inadequacy; DFD is depth of food deficit; ADESA is average dietary energy supply adequacy.

164

Table A4.5: Fixed effects panel results (1990‒2011) Dependent variable Ln PU Ln PFI Ln DFD Ln ADESA

Independent variables Ln property rights protection (PRP) 0.113 0.0721 0.151 -0.008 (0.099) (0.070) (0.119) (0.014) Ln household final consumption 0.0135 -0.071 -0.004 0.006 expenditure per capita (HHFCEPC) (0.175) (0.153) (0.214) (0.026) Ln population (POP) 1.094* 0.607 1.074 -0.152 (0.618) (0.503) (0.744) (0.105) Agriculture share of GDP -0.659 -0.798 -1.343 0.154 (AGCOM) (0.960) (0.853) (1.200) (0.173) Percent of agricultural land -0.001 -0.001 -0.001 0.0003 (AGRILD) (0.002) (0.002) (0.003) (0.003) Latitude (LATIT) (dropped) (dropped) (dropped) (dropped)

Ln openness (OPEN) -0.068 -0.023 -0.062 -0.002 (0.098) (0.072) (0.109) (0.018) Ln human capital (HC) -0.580 -0.450 -0.567 0.053 (1.008) (0.798) (1.252) (0.168) Ln inflation (INFL) 0.029 0.031 0.078 -0.007 (0.028) (0.027) (0.056) (0.007) Ethnolinguistic fractionalisation (dropped) (dropped) (dropped) (dropped) (ETHNO) Year & Country Effects YES YES YES YES N 545 545 545 545 Countries 57 57 57 57 R2 0.387 0.333 0.229 0.544 F 8.424 6.186 3.326 10.66 Notes: ***, **, * denotes a significance level at 1%, 5% & 10%, respectively. Robust standard errors in parentheses. Year and country fixed effects are controlled for. PU is prevalence of undernourishment; PFI is prevalence of food inadequacy; DFD is depth of food deficit; ADESA is average dietary energy supply adequacy.

165

Table A4.6: Random effects panel results (1990‒2011) (excluding China) Dependent variable Ln PU Ln PFI Ln DFD Ln ADESA Independent variables Ln property rights protection (PPR) 0.064 0.028 0.024 -0.00003 (0.070) (0.051) (0.083) (0.012) Ln household final consumption -0.371*** -0.311*** - 0.079*** expenditure per capita (0.103) (0.088) 0.484*** (0.019) (HHFCEPC) (0.127) Ln population (POP) -0.071 -0.095* -0.148* 0.033*** (0.055) (0.052) (0.079) (0.010) Agriculture share of GDP -1.339 -1.199 -2.124* 0.288 (AGCOM) (1.003) (0.832) (1.180) (0.183) Percent of agricultural land -0.001 -0.0006 -0.001 0.0002 (AGRILD) (0.002) (0.002) (0.003) (0.0004) Latitude (LATIT) -1.498** -1.717** -2.514** 0.360*** (0.756) (0.719) (1.177) (0.136) Ln openness (OPEN) -0.131 -0.083 -0.163 0.014 (0.114) (0.082) (0.124) (0.021) Ln human capital (HC) -0.970* -0.771* -0.913 0.233*** (0.513) (0.405) (0.587) (0.082) Ln inflation (INFL) 0.007 0.014 0.051 -0.001 (0.019) (0.021) (0.045) (0.005) Ethnolinguistic fractionalisation 0.037 0.039 0.177 -0.002 (ETHNO) (0.281) (0.247) (0.366) (0.056) N 531 531 531 531 Countries 56 56 56 56 R2 n.a. n.a. n.a. n.a. Notes: Notes: ***, **, * denotes a significance level at 1%, 5% & 10%, respectively. Robust standard errors in parentheses. F-statistics are not available, so they are not reported. PU is prevalence of undernourishment; PFI is prevalence of food inadequacy; DFD is depth of food deficit; ADESA is average dietary energy supply adequacy.

166

Table A4.7: Fixed effect panel results (1990-2011) (excluding China) Dependent variable Ln PU Ln PFI Ln DFD Ln ADESA Independent variables Ln property rights protection (PRP) 0.117 0.0748 0.156 -0.009 (0.101) (0.071) (0.121) (0.015) Ln household final consumption -0.004 -0.082 -0.007 0.003 expenditure per capita (HHFCEPC) (0.176) (0.157) (0.219) (0.027) Ln population (POP) 1.117* 0.622 1.077 -0.147 (0.639) (0.513) (0.766) (0.108) Agriculture share of GDP -0.571 -0.745 -1.323 0.166 (AGCOM) (1.000) (0.868) (1.235) (0.178) Percent of agricultural land -0.001 -0.0008 -0.001 0.0003 (AGRILD) (0.002) (0.001) (0.002) (0.0003) Latitude (LATIT) (dropped (dropped (dropped (dropped) ) ) ) Ln openness (OPEN) -0.073 -0.027 -0.071 -0.001 (0.101) (0.073) (0.110) (0.019) Ln human capital (HC) -0.622 -0.475 -0.585 0.051 (1.021) (0.805) (1.266) (0.169) Ln inflation (INFL) 0.029 0.032 0.079 -0.007 (0.028) (0.028) (0.057) (0.006) Ethnolinguistic fractionalisation (dropped (dropped (dropped (dropped) (ETHNO) ) ) ) Year & Country Effects YES YES YES YES N 531 531 531 531 Countries 56 56 56 56 R2 0.384 0.329 0.226 0.535 F 7.538 6.014 3.230 9.518 Notes: ***, **, * denotes a significance level at 1%, 5% & 10%, respectively. Robust standard errors in parentheses. Year and country fixed effects are controlled for. PU is prevalence of undernourishment; PFI is prevalence of food inadequacy; DFD is depth of food deficit; ADESA is average dietary energy supply adequacy.

167

Appendix 3: Descriptive statistics for indicator of security in land property rights by village and province

Indicator of security in land property rights by village Village ID Mean Standard Deviation # of plots 01 8.692 2.719 13 02 9.130 1.938 23 03 5.083 5.142 12 04 9.882 0.4851 17 05 7.917 3.343 12 06 9.143 1.590 21 07 8.077 4.019 26 08 8 3.343 18 09 10 0 8 10 10 0 17 11 9.444 1.867 27 12 5.045 4.123 22 13 9.231 2.717 26 14 9 2.646 15 15 9.875 0.5 16 16 8.107 3.862 28 17 5 4.601 13 18 9.454 1.507 11 19 9.286 1.488 21 20 9.636 0.492 22 21 7.941 3.976 17 22 8.667 3.279 9 23 10 0 11 24 7.5 4.287 18 25 6 4.619 16 26 9.692 0.630 13 27 8.261 3.532 23 28 7.172 4.098 29 29 9.882 .4851 17 30 6.667 4.091 21 31 8.273 3.289 11 32 10 0 17 Total 8.407 3.255 570

Indicator of security in land property rights by province Province ID Mean Standard Deviation # of plots 001 8.459 3.185 172 002 8.296 3.316 115 003 8.294 3.289 143 004 8.55 3.283 140 Total 8.407 3.255 570

168

Appendix 4: Summary of estimation results for the three empirical chapters

Table A5.1: List of estimation results Chapter 2: Land property rights and household food security

I. Basic estimated equations LHS variable: Food insecurity

∗∗ ∗∗ (1) 퐹퐼ℎ = ⋯ − 0.020 퐿푃푅ℎ − 0.008 푉푇퐷 − 0.072 퐴퐴푃푅 ∗∗ ∗∗∗ (2) 퐹퐼ℎ = ⋯ − 0.019 퐿푃푅ℎ − 0.028 퐶퐿푆 − 0.044 푆푃 − 0.002 푌푅퐶퐸

− 0.026 퐶퐹 + 0.009 푅푄푖 − 0.006 푉푇퐷 − 0.031퐴퐴푃푅 ∗∗ ∗∗ (3) 퐹퐼ℎ = ⋯ − 0.019 퐿푃푅ℎ − 0.028 퐶퐿푆 − 0.045 푆푃 − 0.002 푌푅퐶퐸

− 0.025 푁퐶 − 0.008 푆퐻 + 0.004 푅푄푖 − 0.005 푉푇퐷 − 0.029 퐴퐴푃푅 ∗ ∗∗∗ (4) 퐹퐼ℎ = ⋯ − 0.018 퐿푃푅ℎ − 0.028 퐶퐿푆 − 0.045 푆푃 − 0.002 푌푅퐶퐸

− 0.029 퐶퐹 − 0.008 푆퐻 + 0.013 푉푀푇 − 0.091 퐺푇 − 0.042 푅푄푖 − 0.005 푉푇퐷 − 0.031 퐴퐴푃푅 ∗ ∗∗∗ (5) 퐹퐼ℎ = ⋯ − 0.017 퐿푃푅ℎ − 0.028 퐶퐿푆 − 0.030 푆푃 − 0.001 푌푅퐶퐸 − 0.034 퐶퐹 − 0.008 푆퐻 + 0.012 푉푀푇 − 0.087 퐺푇 − 0.091 푆푅푆

− 0.040 퐿퐷 − 0.026 푅푄푖 − 0.005 푉푇퐷 − 0.027 퐴퐴푃푅 ∗∗∗ (6) 퐹퐼ℎ = … . . − 0.028 퐶퐿푆 − 0.015 푆푃 − 0.001 푌푅퐶퐸 − 0.042 퐶퐹 − 0.009 푆퐻 + 0.023 푉푀푇 − 0.095∗ 퐺푇 − 0.063 푆푅푆 − 0.064 퐿퐷

− 0.019 푅푄푖 − 0.003 푉푇퐷 − 0.027 퐴퐴푃푅 ∗∗ ∗∗∗ (7) 퐹퐼ℎ = ⋯ − 0.018 퐿푃푅ℎ − 0.028 퐶퐿푆 − 0.033 푆푃 − 0.001 푌푅퐶퐸 − 0.036 퐶퐹 − 0.008 푆퐻 + 0.014 푉푀푇 − 0.084 퐺푇 − 0.070 푆푅푆

− 0.038 푅푄푖 − 0.005 푉푇퐷 − 0.025 퐴퐴푃푅 ∗∗ ∗∗∗ (8) 퐹퐼ℎ = ⋯ − 0.098 퐿푃푅ℎ − 0.351 퐶퐿푆 − 0.162 푆푃 − 0.006 푌푅퐶퐸 − 0.197 퐶퐹 − 0.039 푆퐻 + 0.128 푉푀푇 − 0.583∗ 퐺푇 − 0.387 푆푅푆

− 0.446 푅푄푖 − 0.024 푉푇퐷 − 0.095 퐴퐴푃푅 ∗∗ ∗∗∗ (9) 퐹퐼ℎ = ⋯ − 0.060 퐿푃푅ℎ − 0.194 퐶퐿푆 − 0.091 푆푃 − 0.005 푌푅퐶퐸 − 0.124 퐶퐹 − 0.015 푆퐻 + 0.064 푉푀푇 − 0.345∗ 퐺푇 − 0.218 푆푅푆

− 0.233 푅푄푖 − 0.005 푉푇퐷 − 0.061 퐴퐴푃푅

169

LHS variable: Intensity of food insecurity ∗∗∗ (1) 퐼퐹퐼ℎ = ⋯ − 0.067 퐿푃푅ℎ − 0.146 퐶퐿푆 + 0.235 푆푃 + 0.015 푌푅퐶퐸 − 0.514 퐶퐹 + 0.064 푆퐻 + 0.753∗ 푉푀푇 + 0.222 퐺푇 − 0.510 푆푅푆 ∗ − 1.770 푅푄푖 − 0.061 푉푇퐷 − 0.152 퐴퐴푃푅 ∗ ∗∗∗ (2) 퐼퐹퐼ℎ = ⋯ − 0.349 퐿푃푅ℎ − 1.526 퐶퐿푆 − 0.109 푆푃 + 0.017 푌푅퐶퐸 − 1.761 퐶퐹 + 0.228 푆퐻 + 1.721 푉푀푇 − 1.029 퐺푇 − 1.922 푆푅푆

− 3.287 푅푄푖 − 0.163 푉푇퐷 − 0.097 퐴퐴푃푅 ∗∗∗ ∗∗∗ ∗ ∗∗∗ (3) 퐼퐹퐼ℎ = ⋯ − 0.056 퐿푃푅ℎ − 0.394 퐶퐿푆 − 0.245 푆푃 + 0.016 푌푅퐶퐸 − 0.538∗∗∗ 퐶퐹 + 0.066∗∗ 푆퐻 + 0.773∗∗∗ 푉푀푇 + 0.061 퐺푇 ∗∗∗ ∗∗∗ ∗∗∗ − 0.448 푆푅푆 − 1.784 푅푄푖 − 0.056 푉푇퐷 − 0.065 퐴퐴푃푅

II. Potential transmission channels ∗∗ ∗ (1) 퐶퐴ℎ = + 0.024 퐿푃푅ℎ + 0.011 퐶퐿푆 + 0.080 푆푃 − 0.005 푌푅퐶퐸 + 0.033 퐶퐹 ∗∗ + 0.020 푆퐻 + 0.026 푉푀푇 − 0.040 퐺푇 + 0.052 푆푅푆 + 0.483 푅푄푖 + 0.001 푉푇퐷 + 0.028 퐴퐴푃푅 ∗∗∗ ∗∗ (2) 퐶푂퐿ℎ = ⋯ + 0.038 퐿푃푅ℎ − 0.022 퐶퐿푆 − 0.038 푆푃 + 0.001 푌푅퐶퐸 + 0.076 퐶퐹 − 0.011 푆퐻 + 0. 162∗ 푉푀푇 − 0.127∗∗ 퐺푇 + 0.031 푆푅푆

− 0.150 푅푄푖 + 0.004 푉푇퐷 − 0.017 퐴퐴푃푅 ∗∗ ∗∗∗ (3) 푅퐶푅ℎ = ⋯ + 0.318 퐿푃푅ℎ + 0.510 퐶퐿푆 − 0.281 푆푃 + 0.002 푌푅퐶퐸 + 1.154 퐶퐹 + 0.183 푆퐻 + 0.869 푉푀푇 − 1.174 퐺푇 + 0.666 푆푅푆 ∗∗∗ − 1.266 푃푅푖 + 0.096 푉푇퐷 − 2.197 퐴퐴푃푅

Chapter 3: Flood risk, irrigation, productivity, and household food security LHS variable: Yield per plot ∗∗ ∗∗∗ (1) 푌푝 = ⋯ − 0.718 퐹푅 + 0.398 푆푅푆 + 0.623 푅푄푖 + 2.947 퐶퐿푆 ∗∗∗ ∗∗∗ ∗∗∗ + 0.381 푆푃 − 0.032 푌푅퐶퐸 + 1.682 퐶퐹 + 4.495 퐶퐶퐼푅푃푝 ∗∗ − 0.534 퐶푂퐼푅푃푝 − 0.012 푉푇퐷 − 1.457 퐴퐴푃푅

170

∗∗ ∗∗∗ (2) 푌푝 =. . … − 0.734 퐹푅 + 0.349 푆푅푆 + 0.595 푅푄푖 + 2.958 퐶퐿푆 ∗∗∗ ∗∗∗ ∗∗∗ + 0.372 푆푃 − 0.009 푌푅퐶퐸 + 1.443 퐶퐹 + 4.619 퐶퐶퐼푅푃푝 ∗ ∗∗ − 0.562 퐶푂퐼푅푃푝 − 0.009 푉푇퐷 − 1.443 퐴퐴푃푅 − 0.073 푆퐻 ∗∗ ∗∗∗ (3) 푌푝 = ⋯ − 0.683 퐹푅 + 0.335 푆푅푆 + 0.617 푃푅푖 + 2.966 퐶퐿푆 + 0.426 푆푃 ∗∗∗ ∗∗∗ ∗∗ − 0.029 푌푅퐶퐸 + 1.645 퐶퐹 + 4.366 퐶퐶퐼푅푃푝 − 0.530 퐶푂퐼푅푃푝 − 0.002 푉푇퐷 − 1.451∗∗ 퐴퐴푃푅 − 0.084 푆퐻 + 0.077∗ 퐿푃푅 ∗∗ ∗∗∗ (4) 푌푝 = ⋯ − 0.663 퐹푅 + 0.274 푆푅푆 + 0.663 푅푄푖 + 2.973 퐶퐿푆 + 0.475 푆푃 ∗∗ ∗∗∗ ∗∗∗ − 0.026 푌푅퐶퐸 + 1.433 퐶퐹 + 5.012 퐶퐶퐼푅푃푝 ∗∗ ∗∗ − 0.628 퐶푂퐼푅푃푝 − 0.001 푉푇퐷 − 1.423 퐴퐴푃푅 − 0.102 푆퐻 + 0.089∗∗ 퐿푃푅 + 0.982∗ 푅퐷퐶 − 0.366 푅퐿푃 + 0.863 푈푃 ∗∗ ∗∗∗ (5) 푌푝 = ⋯ − 0.689 퐹푅 + 0.294 푆푅푆 + 0.629 푅푄푖 + 2.967 퐶퐿푆 + 0.443 푆푃 ∗∗ ∗∗∗ ∗∗∗ − 0.026 푌푅퐶퐸 + 1.358 퐶퐹 + 4.993 퐶퐶퐼푅푃푝 ∗∗ ∗∗ − 0.623 퐶푂퐼푅푃푝 − 0.001 푉푇퐷 − 1.421 퐴퐴푃푅 − 0.103 푆퐻 + 0.084∗∗ 퐿푃푅 + 0.643 퐼푃

LHS variable: Per-hectare rice yield at household level ∗∗ (1) 푌ℎ = ⋯ − 0.375 퐹푅 + 0.010 푆푅푆 − 0.010 푅푄푖 − 0.081 퐶퐿푆 − 0.064 푆푃 ∗∗∗ ∗∗∗ − 0.021 푌푅퐶퐸 + 1.377 푁퐶 + 1894 퐶퐶퐼푅푃ℎ ∗∗∗ + 0.743 퐶푂퐼푅푃ℎ + 0.006 푉푇퐷 − 0.087 퐴퐴푃푅 ∗∗ (2) 푌ℎ =. . … − 0.370 퐹푅 + 0.027 푆푅푆 − 0.000 푅푄푖 − 0.085 퐶퐿푆 − 0.062 푆푃 ∗∗∗ ∗∗∗ − 0.022 푌푅퐶퐸 + 1.374 퐶퐹 + 1.852 퐶퐶퐼푅푃ℎ ∗∗∗ + 0.753 퐶푂퐼푅푃ℎ + 0.006 푉푇퐷 + 0.092 퐴퐴푃푅 + 0.025 푆퐻 ∗ (3) 푌ℎ = ⋯ − 0.344 퐹푅 + 0.020 푆푅푆 + 0.011 푅푄푖 − 0.081 퐶퐿푆 − 0.034 푆푃 ∗∗∗ ∗∗∗ − 0.021 푌푅퐶퐸 + 1.351 퐶퐹 + 1.724 퐶퐶퐼푅푃ℎ ∗∗∗ + 0.769 퐶푂퐼푅푃ℎ + 0.009 푉푇퐷 − 0.097 퐴퐴푃푅 − 0.020 푆퐻 + 0.039∗ 퐿푃푅

171

∗ (4) 푌ℎ = ⋯ − 0.303 퐹푅 − 0.028 푆푅푆 − 0.049 푅푄푖 − 0.078 퐶퐿푆 + 0.014 푆푃 ∗∗∗ ∗∗∗ ∗ − 0.018 푌푅퐶퐸 + 1.172 퐶퐹 + 2.404 퐶퐶퐼푅푃ℎ ∗∗∗ + 0.666 퐶푂퐼푅푃ℎ + 0.012 푉푇퐷 − 0.068 퐴퐴푃푅 + 0.001 푆퐻 + 0.049∗∗ 퐿푃푅 + 1.201∗∗∗ 푅퐷퐶 + 0.378∗ 푅퐿푃 + 0.503 푈푃 ∗∗ (5) 푌ℎ = ⋯ − 0.350 퐹푅 − 0.023 푆푅푆 + 0.024 푅푄푖 − 0.080 퐶퐿푆 − 0.016 푆푃 ∗∗∗ ∗∗∗ ∗∗ − 0.017 푌푅퐶퐸 + 1.049 퐶퐹 + 2.383 퐶퐶퐼푅푃ℎ ∗∗∗ + 0.671 퐶푂퐼푅푃ℎ + 0.013 푉푇퐷 − 0.065 퐴퐴푃푅 − 0.001 푆퐻 + 0.046∗∗ 퐿푃푅 + 0.676∗∗∗ 퐼푃

LHS variable: Revenue at plot level ∗∗ ∗∗ ∗∗∗ (1) 푅퐸푉푝 = ⋯ − 0.586 퐹푅 + 0.289 푆푅푆 + 1.780 푅푄 + 2.366 퐶퐿푆 ∗∗∗ ∗∗∗ ∗ + 0.376 푆푃 − 0.024 푌푅퐶퐸 + 1.388 퐶퐹 + 2.837 퐶퐶퐼푅푃푝 ∗∗ − 0.247 퐶푂퐼푅푃푝 − 0.010 푉푇퐷 − 1.165 퐴퐴푃푅 ∗∗ ∗∗ ∗∗∗ (2) 푅퐸푉푝 =. . … − 0.601 퐿푅 + 0.244 푆푅푆 + 1.754 푅푄 + 2.376 퐶퐿푆 ∗∗ ∗∗∗ ∗ + 0.369 푆푃 − 0.023 푌푅퐶퐸 + 1.396 퐶퐹 + 2.950 퐶퐶퐼푅푃푝 ∗∗ − 0.273 퐶푂퐼푅푃푝 − 0.008 푉푇퐷 − 1.152 퐴퐴푃푅 − 0.067 푆퐻 ∗∗ ∗∗ ∗∗∗ (3) 푅퐸푉푝 = ⋯ − 0.562 퐿푅 + 0.234 푆푅푆 + 1.771 푅푄 + 2.382 퐶퐿푆 ∗∗ ∗∗∗ ∗ + 0.409 푆푃 − 0.201 푌푅퐶퐸 + 1.361 퐶퐹 + 2.758 퐶퐶퐼푅푃푝 ∗∗ − 0.248 퐶푂퐼푅푃푝 − 0.002 푉푇퐷 − 1.158 퐴퐴푃푅 − 0.075 푆퐻 + 0.059∗ 퐿푃푅 ∗∗ ∗∗ ∗∗∗ (4) 푅퐸푉푝 = ⋯ − 0.546 퐿푅 + 0.184 푆푅푆 + 1.811 푅푄 + 2.386 퐶퐿푆 ∗ ∗∗ ∗∗∗ ∗∗ + 0.446 푆푃 − 0.019 푌푅퐶퐸 + 1.145 퐶퐹 + 3.346 퐶퐶퐼푅푃푝 ∗∗ ∗ − 0.338 퐶푂퐼푅푃푝 − 0.000 푉푇퐷 − 1.124 퐴퐴푃푅 − 0.091 푆퐻 + 0.068∗∗ 퐿푃푅 + 0.806∗ 푅퐼 + 0.416 푅퐿퐼 + 0.730 푈푃퐼 ∗∗ ∗∗ ∗∗∗ (5) 푅퐸푉푝 = ⋯ − 0.567 퐿푅 + 0.195 푆푅푆 + 1.782 푅푄 + 2.383 퐶퐿푆 ∗ ∗ ∗∗∗ ∗∗ + 0.426 푆푃 − 0.018 푌푅퐶퐸 + 1.095 퐶퐹 + 3.399 퐶퐶퐼푅푃푝 ∗∗ ∗ − 0.335 퐶푂퐼푅푃푝 + 0.001 푉푇퐷 − 1.130 퐴퐴푃푅 − 0.093 푆퐻 + 0.065∗∗ 퐿푃푅 + 0.595∗ 퐼푃

172

LHS variable: Per-hectare rice revenue at household level ∗ ∗∗∗ (1) 푅퐸푉ℎ = ⋯ − 0.283 퐿푅 + 0.013 푆푅푆 + 1.585 푅푄푖 − 0.060 퐶퐿푆 ∗∗∗ ∗∗∗ − 0.036 푆푃 − 0.016 푌푅퐶퐸 + 1.119 퐶퐹 + 1.036 퐶퐶퐼푅푃ℎ ∗∗ + 0.515 퐶푂퐼푅푃ℎ + 0.005 푉푇퐷 − 0.077 퐴퐴푃푅 ∗ ∗∗∗ (2) 푅퐸푉ℎ =. . … − 0.028 퐿푅 + 0.022 푆푅푆 + 1.590 푅푄푖 − 0.062 퐶퐿푆 ∗∗∗ ∗∗∗ − 0.521 푆푃 − 0.016 푌푅퐶퐸 + 1.118 퐶퐹 + 1.014 퐶퐶퐼푅푃ℎ ∗∗ + 0.521 퐶푂퐼푅푃ℎ + 0.004 푉푇퐷 − 0.080 퐴퐴푃푅 + 0.013 푆퐻 ∗ ∗∗∗ (3) 푅퐸푉ℎ = ⋯ − 0.260 퐿푅 + 0.017 푆푅푆 + 1.599 푅푄푖 − 0.059 퐶퐿푆 ∗∗∗ ∗∗∗ − 0.014 푆푃 − 0.015 푌푅퐶퐸 + 1.100 퐶퐹 + 0.915 퐶퐶퐼푅푃ℎ ∗∗ + 0.533 퐶푂퐼푅푃ℎ + 0.007 푉푇퐷 − 0.083 퐴퐴푃푅 + 0.009 푆퐻 + 0.030∗ 퐿푃푅 ∗∗∗ (4) 푅퐸푉ℎ = ⋯ − 0.228 퐿푅 − 0.019 푆푅푆 + 1.629 푅푄푖 − 0.058 퐶퐿푆 ∗∗∗ ∗∗∗ + 0.022 푆푃 − 0.013 푌푅퐶퐸 + 0.946 퐶퐹 + 1.464 퐶퐶퐼푅푃ℎ ∗∗ + 0.450 퐶푂퐼푅푃ℎ + 0.009 푉푇퐷 − 0.057 퐴퐴푃푅 − 0.005 푆퐻 + 0.037∗∗ 퐿푃푅 + 0.933∗∗∗ 푅퐷퐶 + 0.342∗ 푅퐿푃 + 0.308 푈푃 ∗ ∗∗∗ (5) 푅퐸푉ℎ = ⋯ − 0.265 퐿푅 − 0.019 푆푅푆 + 1.609 푅푄푖 − 0.058 퐶퐿푆 ∗∗ ∗∗∗ + 0.002 푆푃 − 0.013 푌푅퐶퐸 + 0.849 퐶퐹 + 1.464 퐶퐶퐼푅푃ℎ ∗∗ + 0.451 퐶푂퐼푅푃ℎ + 0.010 푉푇퐷 − 0.057 퐴퐴푃푅 − 0.008 푆퐻 + 0.036∗∗ 퐿푃푅 + 0.563∗∗∗ 퐼푃

LHS variable: Food insecurity ∗∗ ∗ (1) 퐹퐼ℎ = ⋯ − 0.007 푌ℎ − 0.018 퐿푃푅 + 0.015 푅푄푖 − 0.011 퐶퐿푆 − 0.041 푆푃 − 0.002 푌푅퐶퐸 + 0.003 퐶퐹 − 0.009 푆퐻 − 0.005 푉푇퐷 − 0.042 퐴퐴푃푅 ∗∗∗ ∗∗∗ (2) 퐹퐼ℎ = ⋯ − 0.053 푌ℎ − 0.015 퐿푃푅 + 0.016 푅푄푖 − 0.030 퐶퐿푆 − 0.054 푆푃 − 0.002 푌푅퐶퐸 + 0.045 퐶퐹 − 0.007 푆퐻 − 0.004 푉푇퐷 − 0.037 퐴퐴푃푅 ∗∗ ∗ (3) 퐹퐼ℎ = ⋯ − 0.009 푅퐸푉ℎ − 0.018 퐿푃푅 + 0.038 푅푄푖 − 0.012 퐶퐿푆 − 0.004 푆푃 − 0.002 푌푅퐶퐸 + 0.001 퐶퐹 − 0.009 푆퐻 − 0.005 푉푇퐷 − 0.041 퐴퐴푃푅

173

∗∗∗ ∗ ∗∗∗ (4) 퐹퐼ℎ = ⋯ − 0.059 푅퐸푉ℎ − 0.016 퐿푃푅 + 0.118 푅푄푖 − 0.029 퐶퐿푆 − 0.051 푆푃 − 0.002 푌푅퐶퐸 + 0.041 퐶퐹 − 0.007 푆퐻 − 0.004 푉푇퐷 − 0.034 퐴퐴푃푅 ∗∗∗ ∗∗ (5) 퐹퐼ℎ = ⋯ − 0.099 푌ℎ − 0.055 퐿푃푅 + 0.038 푅푄푖 − 0.051 퐶퐿푆 − 0.150 푆푃 − 0.008 푌푅퐶퐸 + 0.131 퐶퐹 − 0.029 푆퐻 − 0.012 푉푇퐷 − 0.098 퐴퐴푃푅 ∗∗∗ ∗ ∗∗∗ (6) 퐹퐼ℎ = ⋯ − 0.227 푌ℎ − 0.053 퐿푃푅 + 0.109 푅푄푖 − 0.204 퐶퐿푆 − 0.208 푆푃 − 0.006 푌푅퐶퐸 + 0.223 퐶퐹 − 0.026 푆퐻 − 0.010 푉푇퐷 − 0.086 퐴퐴푃푅 ∗∗∗ ∗∗ (7) 퐹퐼ℎ = ⋯ − 0.114 푅퐸푉ℎ − 0.057 퐿푃푅 + 0.115 푅푄푖 − 0.057 퐶퐿푆 − 0.144 푆푃 − 0.008 푌푅퐶퐸 + 0.123 퐶퐹 − 0.029 푆퐻 − 0.013 푉푇퐷 − 0.104 퐴퐴푃푅 ∗∗∗ ∗∗ ∗∗∗ (8) 퐹퐼ℎ = ⋯ − 0.254 푅퐸푉ℎ − 0.054 퐿푃푅 + 0.261 푅푄푖 − 0.203 퐶퐿푆 − 0.196 푆푃 − 0.006 푌푅퐶퐸 + 0.205 퐶퐹 − 0.027 푆퐻 − 0.011 푉푇퐷 − 0.086 퐴퐴푃푅

LHS variable: Length of food insecurity ∗∗∗ ∗∗∗ ∗∗∗ (1) 퐼퐹퐼ℎ = ⋯ − 0.225 푌ℎ − 0.013 퐿푃푅 − 1.888 푅푄푖 − 0.158 퐶퐿푆 + 0.222 푆푃 + 0.015∗∗∗ 푌푅퐶퐸 − 0.087 퐶퐹 + 0.075∗∗ 푆퐻 − 0.040∗∗∗ 푉푇퐷 − 0.053 퐴퐴푃푅 ∗∗∗ ∗∗∗ ∗∗∗ (2) 퐼퐹퐼ℎ = ⋯ − 0.325 푌ℎ − 0.014 퐿푃푅 − 1.752 푅푄푖 − 0.423 퐶퐿푆 + 0.145 푆푃 + 0.017∗∗∗ 푌푅퐶퐸 − 0.061 퐶퐹 + 0.080∗∗∗ 푆퐻 − 0.039∗∗∗ 푉푇퐷 − 0.035 퐴퐴푃푅 ∗∗∗ ∗∗∗ ∗∗∗ (3) 퐼퐹퐼ℎ = ⋯ − 0.270 푅퐸푉ℎ − 0.016 퐿푃푅 − 1.524 푅푄푖 − 0.173 퐶퐿푆 + 0.235 푆푃 + 0.015∗∗∗ 푌푅퐶퐸 − 0.092 퐶퐹 + 0.076∗∗∗ 푆퐻 − 0.041∗∗∗ 푉푇퐷 − 0.060 퐴퐴푃푅 ∗∗∗ ∗∗∗ ∗∗∗ (4) 퐼퐹퐼ℎ = ⋯ − 0.317 푅퐸푉ℎ − 0.019 퐿푃푅 − 1.430 푅푄푖 − 0.422 퐶퐿푆 + 0.170 푆푃 + 0.017∗∗∗ 푌푅퐶퐸 − 0.123 퐶퐹 + 0.077∗∗ 푆퐻 − 0.041∗∗∗ 푉푇퐷 − 0.056 퐴퐴푃푅

174

Chapter 4: Property rights and food security in developing economies

LHS variable: Food (in)security indicators OLS estimates (1) 퐿푛 푃푈 = ⋯ − 0.449∗∗∗ 퐿푛 푃푅푃 − 0.518∗∗∗ 퐿푛 퐻퐻퐹퐶퐸푃퐶 − 0.205∗∗∗ 퐿푛 푃푂푃 − 0.307∗∗∗ 퐿푛 퐴퐺퐶푂푀 − 0.077∗∗∗ 퐿푛 퐴퐺푅퐼퐿퐷 − 0.032 퐿푛 퐿퐴푇퐼푇 − 0.164∗∗ 퐿푛 푂푃퐸푁 − 0.513∗∗∗ 퐿푛 퐻퐶 − 0.026 퐿푛 퐼푁퐹퐿 + 0.030∗∗∗ 퐿푛 퐸푇퐻푁푂 + 푌푒푎푟 & 푅푒푔푖표푛 퐸푓푓푒푐푡푠 (2) 퐿푛 푃퐹퐼 = ⋯ − 0.313∗∗∗ 퐿푛 푃푅푃 − 0.519∗∗∗ 퐿푛 퐻퐻퐹퐶퐸푃퐶 − 0.222∗∗∗ 퐿푛 푃푂푃 − 0.361∗∗∗ 퐿푛 퐴퐺퐶푂푀 − 0.079∗∗∗ 퐿푛 퐴퐺푅퐼퐿퐷 − 0.023 퐿푛 퐿퐴푇퐼푇 − 0.156∗∗ 퐿푛 푂푃퐸푁 − 0.600∗∗∗ 퐿푛 퐻퐶 − 0.005 퐿푛 퐼푁퐹퐿 + 0.026∗∗ 퐿푛 퐸푇퐻푁푂 + 푌푒푎푟 & 푅푒푔푖표푛 퐸푓푓푒푐푡푠 (3) 퐿푛 퐷퐹퐷 = ⋯ − 0.289∗∗ 퐿푛 푃푅푃 − 0.911∗∗∗ 퐿푛 퐻퐻퐹퐶퐸푃퐶 − 0.461∗∗∗ 퐿푛 푃푂푃 − 0.620∗∗∗ 퐿푛 퐴퐺퐶푂푀 − 0.173∗∗∗ 퐿푛 퐴퐺푅퐼퐿퐷 − 0.031 퐿푛 퐿퐴푇퐼푇 − 0.326∗∗∗ 퐿푛 푂푃퐸푁 − 1.066∗∗∗ 퐿푛 퐻퐶 − 0.038 퐿푛 퐼푁퐹퐿 + 0.035∗∗ 퐿푛 퐸푇퐻푁푂 + 푌푒푎푟 & 푅푒푔푖표푛 퐸푓푓푒푐푡푠 (4) 퐿푛 퐴퐷퐸푆퐴 = ⋯ + 0.061∗∗∗ 퐿푛 푃푅푃 + 0.109∗∗∗퐿푛 퐻퐻퐹퐶퐸푃퐶 + 0.041∗∗∗ 퐿푛 푃푂푃 + 0.080∗∗∗ 퐿푛 퐴퐺퐶푂푀 + 0.003 퐿푛 퐴퐺푅퐼퐿퐷 + 0.025∗∗∗ 퐿푛 퐿퐴푇퐼푇 + 0.021 퐿푛 푂푃퐸푁 + 0.129∗∗∗ 퐿푛 퐻퐶 − 0.007∗ 퐿푛 퐼푁퐹퐿 − 0.004∗ 퐿푛 퐸푇퐻푁푂 + 푌푒푎푟 & 푅푒푔푖표푛 퐸푓푓푒푐푡푠

Between estimator (BE) estimates (1) 퐿푛 푃푈 = ⋯ − 0.863∗∗ 퐿푛 푃푅푃 − 0.408∗∗ 퐿푛 퐻퐻퐹퐶퐸푃퐶 − 0.171 퐿푛 푃푂푃 − 0.211 퐿푛 퐴퐺퐶푂푀 − 0.097 퐿푛 퐴퐺푅퐼퐿퐷 − 0.019 퐿푛 퐿퐴푇퐼푇 − 0.068 퐿푛 푂푃퐸푁 − 0.239 퐿푛 퐻퐶 − 0.033 퐿푛 퐼푁퐹퐿 − 0.021 퐿푛 퐸푇퐻푁푂 − 0.823∗∗∗ 퐷푁퐴푀퐸 + 0.250 퐷푆푆퐴 + 0.262 퐷퐸푆퐸퐴 + 0.124 퐷푆퐴

175

(2) 퐿푛 푃퐹퐼 = ⋯ − 0.688∗∗ 퐿푛 푃푅푃 − 0.371∗∗ 퐿푛 퐻퐻퐹퐶퐸푃퐶 − 0.189∗∗ 퐿푛 푃푂푃 − 0.259 퐿푛 퐴퐺퐶푂푀 − 0.102 퐿푛 퐴퐺푅퐼퐿퐷 − 0.020 퐿푛 퐿퐴푇퐼푇 − 0.081 퐿푛 푂푃퐸푁 − 0.359 퐿푛 퐻퐶 + 0.046 퐿푛 퐼푁퐹퐿 − 0.018 퐿푛 퐸푇퐻푁푂 − 1.129∗∗∗ 퐷푁퐴푀퐸 + 0.071 퐷푆푆퐴 + 0.228 퐷퐸푆퐸퐴 + 0.070 퐷푆퐴 (3) 퐿푛 퐷퐹퐷 = ⋯ − 0.806 퐿푛 푃푅푃 − 0.676∗∗∗ 퐿푛 퐻퐻퐹퐶퐸푃퐶 − 0.386∗∗ 퐿푛 푃푂푃 − 0.388 퐿푛 퐴퐺퐶푂푀 − 0.197∗ 퐿푛 퐴퐺푅퐼퐿퐷 − 0.003 퐿푛 퐿퐴푇퐼푇 − 0.162 퐿푛 푂푃퐸푁 − 0.440 퐿푛 퐻퐶 − 0.021 퐿푛 퐼푁퐹퐿 − 0.029 퐿푛 퐸푇퐻푁푂 − 1.663∗∗∗ 퐷푁퐴푀퐸 + 0.040 퐷푆푆퐴 + 0.502 퐷퐸푆퐸퐴 + 0.225 퐷푆퐴 (4) 퐿푛 퐴퐷퐸푆퐴 = ⋯ + 0.155∗∗ 퐿푛 푃푅푃 + 0.081∗∗ 퐿푛 퐻퐻퐹퐶퐸푃퐶 + 0.029 퐿푛 푃푂푃 + 0.055∗ 퐿푛 퐴퐺퐶푂푀 + 0.008 퐿푛 퐴퐺푅퐼퐿퐷 + 0.020 퐿푛 퐿퐴푇퐼푇 − 0.001 퐿푛 푂푃퐸푁 + 0.074 퐿푛 퐻퐶 − 0.018 퐿푛 퐼푁퐹퐿 + 0.005 퐿푛 퐸푇퐻푁푂 + 0.193∗∗∗ 퐷푁퐴푀퐸 − 0.033 퐷푆푆퐴 − 0.011 퐷퐸푆퐸퐴 − 0.030 퐷푆퐴

Random effects (RE) estimates (1) 퐿푛 푃푈 = ⋯ + 0.068 퐿푛 푃푅푃 − 0.324∗∗∗ 퐿푛 퐻퐻퐹퐶퐸푃퐶 − 0.129 퐿푛 푃푂푃 − 0.068 퐿푛 퐴퐺퐶푂푀 − 0.071 퐿푛 퐴퐺푅퐼퐿퐷 − 0.143∗∗ 퐿푛 퐿퐴푇퐼푇 − 0.065 퐿푛 푂푃퐸푁 − 0.939∗ 퐿푛 퐻퐶 + 0.001 퐿푛 퐼푁퐹퐿 − 0.026 퐿푛 퐸푇퐻푁푂 (2) 퐿푛 푃퐹퐼 = ⋯ + 0.032 퐿푛 푃푅푃 − 0.278∗∗∗ 퐿푛 퐻퐻퐹퐶퐸푃퐶 − 0.126∗ 퐿푛 푃푂푃 − 0.078 퐿푛 퐴퐺퐶푂푀 − 0.048 퐿푛 퐴퐺푅퐼퐿퐷 − 0.152∗∗ 퐿푛 퐿퐴푇퐼푇 − 0.031 퐿푛 푂푃퐸푁 − 0.760∗ 퐿푛 퐻퐶 + 0.009 퐿푛 퐼푁퐹퐿 − 0.023 퐿푛 퐸푇퐻푁푂 (3) 퐿푛 퐷퐹퐷 = ⋯ + 0.028 퐿푛 푃푅푃 − 0.435∗∗∗ 퐿푛 퐻퐻퐹퐶퐸푃퐶 − 0.189∗ 퐿푛 푃푂푃 − 0.142 퐿푛 퐴퐺퐶푂푀 − 0.075 퐿푛 퐴퐺푅퐼퐿퐷 − 0.223∗∗ 퐿푛 퐿퐴푇퐼푇 − 0.074 퐿푛 푂푃퐸푁 − 0.880 퐿푛 퐻퐶 + 0.043 퐿푛 퐼푁퐹퐿 − 0.032 퐿푛 퐸푇퐻푁푂

176

(4) 퐿푛 퐴퐷퐸푆퐴 =. . . −0.001 퐿푛 푃푅푃 + 0.074∗∗∗ 퐿푛 퐻퐻퐹퐶퐸푃퐶 + 0.045∗∗∗ 퐿푛 푃푂푃 + 0.018 퐿푛 퐴퐺퐶푂푀 + 0.013 퐿푛 퐴퐺푅퐼퐿퐷 + 0.038∗∗∗ 퐿푛 퐿퐴푇퐼푇 + 0.002 퐿푛 푂푃퐸푁 + 0.225∗∗∗ 퐿푛 퐻퐶 − 0.0004 퐿푛 퐼푁퐹퐿 + 0.007 퐿푛 퐸푇퐻푁푂

Fixed effects pooled panel estimates (1) 퐿푛 푃푈 = ⋯ + 0.135 퐿푛 푃푅푃 + 0.033 퐿푛 퐻퐻퐹퐶퐸푃퐶 + 0.901∗ 퐿푛 푃푂푃 − 0.021 퐿푛 퐴퐺퐶푂푀 − 0.062퐿푛 퐴퐺푅퐼퐿퐷 ∓ 퐿푛 퐿퐴푇퐼푇 (푑푟표푝푝푒푑) − 0.023 퐿푛 푂푃퐸푁 − 0.501 퐿푛 퐻퐶 + 0.025 퐿푛 퐼푁퐹퐿 ∓ 퐿푛 퐸푇퐻푁푂 (푑푟표푝푝푒푑) + 푌푒푎푟 & 퐶표푢푛푡푟푦 퐸푓푓푒푐푡푠 (2) 퐿푛 푃퐹퐼 = ⋯ + 0.087 퐿푛 푃푅푃 − 0.058 퐿푛 퐻퐻퐹퐶퐸푃퐶 + 0.480 퐿푛 푃푂푃 − 0.061 퐿푛 퐴퐺퐶푂푀 − 0.046 퐿푛 퐴퐺푅퐼퐿퐷 ∓ 퐿푛 퐿퐴푇퐼푇(푑푟표푝푝푒푑) + 0.017 퐿푛 푂푃퐸푁 − 0.376 퐿푛 퐻퐶 + 0.029 퐿푛 퐼푁퐹퐿 ∓ 퐿푛 퐸푇퐻푁푂 (푑푟표푝푝푒푑) + 푌푒푎푟 & 퐶표푢푛푡푟푦 퐸푓푓푒푐푡푠 (3) 퐿푛 퐷퐹퐷 = ⋯ + 0.175 퐿푛 푃푅푃 + 0.027 퐿푛 퐻퐻퐹퐶퐸푃퐶 + 0.883 퐿푛 푃푂푃 − 0.083 퐿푛 퐴퐺퐶푂푀 − 0.074 퐿푛 퐴퐺푅퐼퐿퐷 ∓ 퐿푛 퐿퐴푇퐼푇(푑푟표푝푝푒푑) + 0.006 퐿푛 푂푃퐸푁 − 0.447 퐿푛 퐻퐶 + 0.074 퐿푛 퐼푁퐹퐿 ∓ 퐿푛 퐸푇퐻푁푂 (푑푟표푝푝푒푑) + 푌푒푎푟 & 퐶표푢푛푡푟푦 퐸푓푓푒푐푡푠 (4) 퐿푛 퐴퐷퐸푆퐴 = ⋯ − 0.012 퐿푛 푃푅푃 + 0.005 퐿푛 퐻퐻퐹퐶퐸푃퐶 − 0.113 퐿푛 푃푂푃 + 0.012 퐿푛 퐴퐺퐶푂푀 + 0.013 퐿푛 퐴퐺푅퐼퐿퐷 ∓ 퐿푛 퐿퐴푇퐼푇(푑푟표푝푝푒푑) − 0.011 퐿푛 푂푃퐸푁 + 0.036 퐿푛 퐻퐶 − 0.0006 퐿푛 퐼푁퐹퐿 ∓ 퐿푛 퐸푇퐻푁푂 (푑푟표푝푝푒푑) + 푌푒푎푟 & 퐶표푢푛푡푟푦 퐸푓푓푒푐푡푠

Robustness test results LHS variable: Prevalence of undernourishment (1) 퐿푛 푃푈 = ⋯ − 0.449∗∗∗ 퐿푛 푃푅푃 − 0.518 ∗∗∗퐿푛 퐻퐻퐹퐶퐸푃퐶 − 0.205∗∗∗ 퐿푛 푃푂푃 − 0.307∗∗∗ 퐿푛 퐴퐺퐶푂푀 − 0.077∗∗∗ 퐿푛 퐴퐺푅퐼퐿퐷 − 0.032 퐿푛 퐿퐴푇퐼푇 − 0.164∗∗ 퐿푛 푂푃퐸푁 − 0.513∗∗∗ 퐿푛 퐻퐶 − 0.026 퐿푛 퐼푁퐹퐿 + 0.030∗∗∗ 퐿푛 퐸푇퐻푁푂 + 푌푒푎푟 & 푅푒푔푖표푛 퐸푓푓푒푐푡푠

177

(2) 퐿푛 푃푈 = ⋯ − 0.200∗∗∗ 퐿푛 푃푅 − 0.067∗∗∗ 퐿푛 퐻퐻퐹퐶퐸푃퐶 − 0.171∗∗∗ 퐿푛 푃푂푃 − 0.318∗∗∗ 퐿푛 퐴퐺퐶푂푀 − 0.105∗∗∗ 퐿푛 퐴퐺푅퐼퐿퐷 + 0.026 퐿푛 퐿퐴푇퐼푇 − 0.032 퐿푛 푂푃퐸푁 − 0.258∗∗ 퐿푛 퐻퐶 − 0.196∗∗∗ 퐿푛 퐼푁퐹퐿 − 0.005 퐿푛 퐸푇퐻푁푂 + 푌푒푎푟 & 푅푒푔푖표푛 퐸푓푓푒푐푡푠 (3) 퐿푛 푃푈 = ⋯ + 0.127 퐿푛 푃푃푅푆 − 0.663∗∗∗ 퐿푛 퐻퐻퐹퐶퐸푃퐶 − 0.091∗∗ 퐿푛 푃푂푃 − 0.221∗ 퐿푛 퐴퐺퐶푂푀 − 0.056∗ 퐿푛 퐴퐺푅퐼퐿퐷 − 0.049 퐿푛 퐿퐴푇퐼푇 − 0.052 퐿푛 푂푃퐸푁 − 0.127 퐿푛 퐻퐶 + 1.073∗∗∗ 퐿푛 퐼푁퐹퐿 + 0.018 퐿푛 퐸푇퐻푁푂 + 푌푒푎푟 & 푅푒푔푖표푛 퐸푓푓푒푐푡푠 (4) 퐿푛 푃푈 = ⋯ − 0.438∗∗∗ 퐿푛 푃푃푅퐺 − 0.718∗∗∗ 퐿푛 퐻퐻퐹퐶퐸푃퐶 − 0.129∗∗∗ 퐿푛 푃푂푃 − 0.499∗∗∗ 퐿푛 퐴퐺퐶푂푀 − 0.077∗∗∗ 퐿푛 퐴퐺푅퐼퐿퐷 − 0.043 퐿푛 퐿퐴푇퐼푇 − 0.272∗∗∗ 퐿푛 푂푃퐸푁 − 0.086 퐿푛 퐻퐶 + 1.446∗∗∗ 퐿푛 퐼푁퐹퐿 + 0.024∗∗∗ 퐿푛 퐸푇퐻푁푂 + 푌푒푎푟 & 푅푒푔푖표푛 퐸푓푓푒푐푡푠 (5) 퐿푛 푃푈 = ⋯ + 0.099∗ 퐿푛 푅푅푅푃 − 0.614∗∗∗ 퐿푛 퐻퐻퐹퐶퐸푃퐶 − 0.142∗∗∗ 퐿푛 푃푂푃 − 0.262∗∗∗ 퐿푛 퐴퐺퐶푂푀 − 0.063∗∗∗ 퐿푛 퐴퐺푅퐼퐿퐷 − 0.015 퐿푛 퐿퐴푇퐼푇 − 0.131∗ 퐿푛 푂푃퐸푁 − 0.421∗ 퐿푛 퐻퐶 + 1.250∗∗∗ 퐿푛 퐼푁퐹퐿 + 0.010 퐿푛 퐸푇퐻푁푂 + 푌푒푎푟 & 푅푒푔푖표푛 퐸푓푓푒푐푡푠 (6) 퐿푛 푃푈 = ⋯ + 0.242∗∗ 퐿푛 푅퐸퐺푃푅 − 0.536∗∗∗ 퐿푛 퐻퐻퐹퐶퐸푃퐶 − 0.083∗∗∗ 퐿푛 푃푂푃 − 0.143∗∗ 퐿푛 퐴퐺퐶푂푀 − 0.050∗∗∗ 퐿푛 퐴퐺푅퐼퐿퐷 − 0.023 퐿푛 퐿퐴푇퐼푇 − 0.028 퐿푛 푂푃퐸푁 − 0.160 퐿푛 퐻퐶 + 1.553∗∗∗ 퐿푛 퐼푁퐹퐿 − 0.016 퐿푛 퐸푇퐻푁푂 + 푌푒푎푟 & 푅푒푔푖표푛 퐸푓푓푒푐푡푠 (7) 퐿푛 푃푈 = ⋯ − 1.179∗∗∗ 퐿푛 퐼푃푅퐼 − 0.463∗∗∗ 퐿푛 퐻퐻퐹퐶퐸푃퐶 − 0.072 퐿푛 푃푂푃 − 0.235∗ 퐿푛 퐴퐺퐶푂푀 − 0.066∗∗ 퐿푛 퐴퐺푅퐼퐿퐷 − 0.028 퐿푛 퐿퐴푇퐼푇 − 0.054 퐿푛 푂푃퐸푁 − 0.183 퐿푛 퐻퐶 + 1.060∗∗∗ 퐿푛 퐼푁퐹퐿 + 0.043∗ 퐿푛 퐸푇퐻푁푂 + 푌푒푎푟 & 푅푒푔푖표푛 퐸푓푓푒푐푡푠

where: AAPR Annual average precipitation rate ADESA Average dietary energy supply adequacy AGRILD Agricultural land

178

AGCOM Agricultural share over GDP

CAh Credit or loan access, measured at household level

CCIRPp Cost of chemical inputs in rice production at plot level

CCIRPha Cost of chemical inputs in rice production per hectare

COIRPp Cost of other inputs in rice production at plot level

COIRPha Cost of other inputs in rice production per hectare

COLh Collateral use, measured at household level DNAME Dummy for North African and Middle East DSA Dummy for South Asia DESEA Dummy for East and Southeast Asia DSSA Dummy for Sub-Saharan Africa DFD Depth of food deficit ETHNO Ethnolinguistic fractionalisation

FIh Food insecurity at household level FR Flood risk

HHFCEPC Household final consumption per capita GT Government trust HC Human capital

IFIh Length of food insecurity at household level INFL Inflation rate IP Irrigated plots IPRI International property rights index LATIT Latitude LD Land documents as ownership proof

LPRh Land property rights at household level

LPRp Land property rights at plot level CF Annual crop frequency OPEN Openness PFI Prevalence of food inadequacy POP Population PPRS Physical property rights score

RQ Rice quality

179

PR Property rights PRP Property rights protection PRRG Property rights and rule-based governance PU Prevalence of undernourishment

RCRh Revenue-cost ratio, measured at household level REGPR Registering property rights

REVh Per-hectare household rice revenues

REVp Rice revenues at plot level RDC Reservoir, dyke, or canal irrigation RLP River, lake, or pond irrigation RRRP Regulatory restrictions on sale of real property SH Size of household SRS Short-run shocks SP Soil problem UP Underground or piping irrigation VMT Village-member trust

Yp Rice yield at plot level

Yh Rice yield at household level

180