INTRODUCTION

The role of the government in addressing national agriculture and food security issues has been widely acknowledged in literature (Thomson, 2017). Approaches to food security must encompass not only technological and environmental aspects, but also social, economic, and political dimensions (Von Braun, 2009; Wahlqvist et al., 2012; Maye & Kirwan, 2013). Previous studies have examined the determinants of food security at the national level and identified governance as a significant explanatory variable (Forsyth & Farran, 2013;Slimane et al., 2016;Dithmer & Abdulai, 2017; Soko et al., 2023). However, the mechanisms through which governance affects the national food security level are still not fully understood (Akbari et al., 2022). This knowledge gap underscores the urgent need for a more efficient approach to achieving food security, making this research more crucial.

As consensus on the significance of governance has emerged in academia and politics, international organizations have made efforts to improve the governance of developing countries. However, many of these attempts have faced obstacles due to traditional principles of national sovereignty, leading to setbacks (Paarlberg, 2002). Developing countries also recognize the necessity of improving their governance for their development. For instance 2002, 55 African countries established the African Union, demonstrating their commitment to self-driven efforts towards better governance (Crocker, 2020). Despite such endeavors, the proportion of the population facing food insecurity has consistently risen from 21.8% in 2015 to 28.1% in 2020 (World Bank, 2020).

To bridge the gap between theory and reality, a more nuanced understanding of the relationship between governance and food security is essential. Hence, this study aims to go beyond existing research by answering the following key questions: What are the intricate pathways through which governance directly and indirectly impacts food security? Do differences exist in the effects of governance on food security between developed and developing countries? By addressing these questions, the study aims to refine the implications of governance improvement for food security, considering its multifaceted nature.

Food security has consistently garnered attention from academic and non-academic circles, spanning the past and present (Allen, 2013). The consecutive global food price crises in 2007-2008 and 2010 brought renewed awareness to the importance of food security worldwide. In response, the World Bank published the World Development Report in 2008, strongly advocating for agricultural investments in developing countries. Moreover, recent events such as the COVID-19 pandemic and the Russia-Ukraine conflict, which triggered international grain supply uncertainties, have once again highlighted the significance of food security. Since 2019, the proportion of undernourished people worldwide has increased, with an estimated 3.1 billion people unable to afford the FAO’s Cost of a Healthy Diet in 2021. This is an increase of 134 million compared to pre-COVID-19 (FAOSTAT, 2021). Most of these people are in less developed countries, including sub-Saharan Africa. In particular, less-developed countries have less governmental support to help them than developed countries (Ghosh, 2023). Despite decades of efforts to eradicate hunger and malnutrition, food security remained an unresolved social issue (Candel et al., 2014).

Both domestic food supply and accessibility are influenced by the performance of a country’s agricultural sector. Therefore, previous research attempting to trace back the determinants of food security has often focused on agricultural productivity, which is the result of domestic agricultural activities (Antle, 1983;Ortega & Lederman, 2004;Lio & Liu, 2008). Agriculture plays a crucial role in shaping a country’s internal food supply and prices, thus contributing to food security. Furthermore, agriculture is a primary source of income and employment for many underdeveloped countries. These reasons make agricultural performance a crucial factor at the national and householdlevel economic stability and food accessibility (UNDP, 2012).

Therefore, much of the literature that emphasizes the importance of governance in improving food security also aligns its focus with research aiming to enhance agricultural productivity. Kawagoe et al. (1985) argued that barriers to a nation’s agricultural productivity lie in the scarcity of natural resources, inadequate technological potential, and weak institutional foundations that hinder innovation and the implementation of new technologies. Lio and Liu (2008), in their country-level analysis across 127 countries worldwide, empirically demonstrated that even with material development driving agricultural modernization―such as the provision of irrigation facilities, chemical fertilizers, school construction, and adoption of modern agricultural technologies―the increase in agricultural productivity can be severely limited without sound governance. Ishangulyyev and Lee (2018) proved that changes in governance in post-Soviet Central Asian countries, which were under the same system during the Soviet era, significantly impacted agricultural productivity after the dissolution of the Soviet Union.

However, a notable discussion to be made here is that there is a lack of comprehensive empirical research that integrates governance and resulting improvements in agricultural productivity with their ultimate connection to food security. Therefore, this study establishes a structural equation model that encompasses the relationships among governance, agricultural capital, agricultural productivity, and food security, which have been partially understood in the existing literature. Following the approach of previous studies (Lio & Liu, 2008;Slimane et al., 2016), the study focuses on conducting a three-stage least squares (3SLS) analysis to comprehensively examine the direct and indirect processes through which governance improvements lead to food security.

Furthermore, research regarding the intertwined determinants of food security associated with various social and economic foundations has often overlooked the distinctions between developed and developing countries based on income levels (Bi et al., 2022). This study considers this aspect by analyzing the influence of governance based on income levels. The data used for the analysis were mainly extracted from credible sources such as FAOSTAT and the World Bank Database. Processed panel data from 72 countries spanning 20 years were utilized, emphasizing a robust dataset for the analysis.

In the subsequent chapters, the study delves into the theoretical background of food security, governance, agricultural productivity, and agricultural capital – the vital focal points of this research. Chapter 3 outlines the analytical model, while Chapter 4 describes the utilized data and variables. Chapter 5 presents the analysis results, and Chapter 6 concludes the study by summarizing the findings and their implications for future policies.

THEORETICAL BACKGROUNDS

Defining and Measuring Food Security

Due to its complex nature, food security has been defined and measured in various ways. Notably, the Food and Agriculture Organization (FAO) defines food security as “the state in which all people, at all times, have physical, social, and economic access to sufficient, safe, and nutritious food that meets their dietary needs and food preferences for an active and healthy life.” FAO’s definition encompasses four dimensions of food security (FAO, 1996; 2003). Firstly, food availability refers to “the availability of adequate quantities of food of appropriate quality, supplied through domestic production or imports.” Secondly, food accessibility is defined as “access by individuals to adequate resources (entitlements) for acquiring appropriate foods for a nutritious diet.” Thirdly, food utilization refers to “the utilization of food through adequate diet, clean water, sanitation, and health care to reach a state of nutritional well-being where all physiological needs are met.” Lastly, food stability or food protection signifies “access to food even in the face of temporary shocks (e.g., economic or climate crises) or cyclical events (e.g., seasonal food insecurity).” While there are alternative definitions from sources such as the World Food Summit, the World Bank, and the United States Agency for International Development, the fundamental components remain similar across these definitions.

Due to the comprehensive nature of the definition, empirical research employs various indicators as proxy variables for food security. The first dimension of food security, food availability, reflects the supply aspect of food. It is often measured using indicators such as calories, protein, and fat intake per capita per day. The Food and Agriculture Organization (FAO) and the World Health Organization (WHO) measure individual nutritional deficiencies in calorie, protein, and fat intake. Numerous empirical studies (Wimberley, 1991;Wimberley and Bello, 1992;Firebaugh and Beck, 1994;Mihalache-O’Keef and Li, 2011;Djokoto, 2012) have also utilized daily calorie and protein intake per capita as proxy variables for food security. On the other hand, food accessibility, utilization, and stability reflect aspects of food demand, which have more individual variability and are more challenging to measure. Food availability is a prerequisite for the other three dimensions of food security and is the most fundamental condition (Barrett, 2010).

Others have attempted to capture the multidimensional nature of food security by developing composite indices. The International Fund for Agricultural Development (IFAD) introduced the first quantified food security index in 1990. This index divided food consumption and production by their respective variability and assigned weights accordingly. The Global Food Security Index (GFSI), developed by The Economist in 2012, is the most widely utilized food security index. This index comprises 59 sub-indicators reflecting the four dimensions of food security. Furthermore, in 2014, the Research Institute for Agriculture and Life Sciences at Korea University established the fundamental principles a food security index should adhere to. They have been developing and releasing food security indices at both the national and household levels in alignment with these principles every year (Yang et al., 2023).

Objective metrics are pivotal in motivating action, underscoring the immense significance of measuring food security. However, in this study, due to data availability challenges and for the sake of interpretational ease, the focus is on defining food security regarding the availability dimension, which can be credibly measured at the national level. The study’s primary goal is to concentrate on governance’s direct and indirect impacts on food security and the variations in these impacts based on national circumstances.

Defining and Measuring Governance and its Impact on Agriculture

Indeed, in recent years, the concept of governance has been used and applied across various academic disciplines (Kjaer, 2023). In this paper, following the definition by Termeer et al. (2011), governance is broadly defined as "the interaction between public and/or private actors aimed at achieving collective goals." The term ‘governance’ is distinct from the more hierarchical and state-centric approach of ‘government,’ which focuses on managing public issues (Kersbergen and Waarden, 2004).

The World Bank has undertaken the Worldwide Governance Indicators (WGI) project to develop crosscountry indicators for governance. Since 1996, this project has released annual composite indicators covering various governance dimensions for over 200 countries. These indicators encompass corruption control, regulatory quality, government effectiveness, rule of law, political stability, and voice and accountability. The indicators are based on hundreds of variables gathered from 31 diverse data sources, including responses from survey participants, non-governmental organizations, commercial business information providers, and public sector agencies worldwide, reflecting perceptions of governance (Kaufmann et al., 2005). These six governance indicators have been divided into three broad aspects and have been utilized in previous research studies, including works like the one by Ishangulyyev and Lee (2018). These indicators offer insights into how various countries perceive and perform in governance, allowing for cross-country comparisons and analysis of governance- related issues on a global scale.

The first aspect of governance is “respect for the institutional framework,” which entails mutual respect for institutions that regulate economic and social interactions between citizens and the state. This aspect consists of “rule of law” and “control of corruption.” The “rule of law” ‘s impact on food security is substantial. Without laws, law enforcement, or even a functional government, there would be little incentive for anyone to engage in food production, leading to mass hunger. An example is Somalia in the early 1990s, where violence and theft prevented farmers from cultivating crops, resulting in a widespread famine (Lio and Liu, 2008). “Control of corruption" also holds significance in modern market economies. Corruption is traditionally defined as the pursuit of personal gain through the misuse of public power. It arises when both parties engage in corruption and their associates fail to adhere to the rules regulating their interactions (Kaufmann et al., 2005). In societies characterized by pervasive corruption, resources are channeled into bribes rather than productive activities. This can hinder the development of essential agricultural infrastructure and diminish the welfare that contributes significantly to the population’s food security.

Agricultural investments heavily influence agricultural productivity, and if property rights are not adequately protected and contracts are not enforced, it becomes difficult to attract significant private investments in agriculture. Secure property rights, especially for land, positively impact fixed investments and subsequent productivity gains. Modern agricultural practices, which often involve machinery and chemical fertilizers, rely on secure property rights and a well-functioning legal system. Modern agricultural production methods require extensive supply chains, specialization, the utilization of experts, technological development and innovation, and the operation of supply networks. For these aspects to function effectively, an infrastructure of legal contracts and a functioning judiciary are necessary (Stigler, 1951). In an environment without the “rule of law,” adopting modern agricultural practices becomes economically challenging due to the high costs involved, and societies may fail to reap the benefits of increased utility resulting from technological advancements in production.

The second aspect of governance is the “quality of government action.” In other words, it refers to the government’s ability to formulate and implement sound policies effectively. This encompasses the “effectiveness of government” and the “quality of regulations.” The effectiveness of government entails the government’s capacity to formulate and execute sound policies and its ability to provide public goods and services. Numerous studies have demonstrated that agricultural productivity varies based on providing public goods and services such as roads, irrigation systems, communication infrastructure, schools, and agricultural research (Fan et al., 2004). Therefore, the government’s capability to effectively provide these public goods and services is a decisive factor in agricultural development. The judicious policies provided by an efficient government are crucial elements for fostering broad agricultural development and establishing an environment conducive to stable food acquisition (Tomich et al., 2018).

Another dimension of the “quality of government action,” namely the “quality of regulations,” includes the burden resulting from excessive regulations not only in domestic price controls but also in various areas such as foreign trade and business development (Kaufmann et al., 2005). It is imperative to closely examine the significance of the “quality of regulations” and its correlation with agricultural development. Many governments adopt biased policies towards urban areas, exacerbating the rural-urban divide. According to Krueger and Valdés (1991), an analysis of policies in 18 countries from 1960 to 1983 revealed that most nations focused on macroeconomic policies and imposed high taxes on the agricultural sector. Consequently, the overall tax burden on the agricultural sector averaged 30%, severely hampering private investment in agriculture and having detrimental effects on agricultural development. This reduction in rural income can also negatively impact national-level food security.

The third aspect, ‘power of choice,’ consists of ‘communication and accountability’ and ‘political stability.’ ‘Communication and accountability’ refer to the extent to which citizens of a country can participate in government decisions. This encompasses citizens’ interest in the presence of independent media, where the media monitors those in power and evaluates their actions. Citizens’ political engagement influences aspects such as land tenure, agricultural taxation, subsidies, and the provision of public goods, making it a vital concept for agricultural productivity and food security (Binswanger and Deininger, 1997). The final dimension of the governance regime is ‘political stability,’ which measures the likelihood of a government becoming unstable due to unconstitutional or violent means (Kaufmann et al., 2005). In regions of conflict and turmoil, food security tends to weaken. In areas with open conflicts, people cannot stay on their farms, leading to an inability to engage in food production or agricultural activities. Consequently, they rely on external food assistance (Donkor and Ohiokpehai, 1998).

Sound governance facilitates investment in agricultural capital, thereby enhancing agricultural productivity and ultimately contributing to the overarching goal of strengthening food security.

Heterogeneous Impacts on Developed and Developing Countries

Why is it essential to consider developed and developing countries separately? While previous studies utilize national- level panel data, including both developed and developing countries, their interpretations mostly focus on developing countries (Lio and Liu, 2008;Slimane et al., 2016;Dithmer and Abdulai, 2017;Bi et al., 2022). This tendency can be justified because 98% of the world’s undernourished population resides in developing countries (Pawlak, 2016). However, in many developing countries, the reliability of national-level data is often low, and data may be lacking altogether. This fact raises concerns that utilizing global country panels could result in poor representation of low-income countries, even within the developing world. Therefore, research findings using global country panels can reflect the situation in advanced economies compared to developing countries. This could hinder an accurate diagnosis of the core issues developing countries face, which are the central focus of concern.

In particular, the impact of governance on agricultural capital accumulation, agricultural production, and food security, which are the focal points of this study, can vary in magnitude across countries due to various factors, including income levels. For instance, as previously explained, while improvements in governance theoretically lead to increased investment in agriculture and subsequent enhancement in agricultural productivity, in low-income developing countries, the lack of capital for immediate agricultural investment might hinder realizing these effects despite governance improvements. Additionally, even if agricultural capital accumulation occurs, constraints such as lack of accessibility, skilled labor, and inadequate educational systems may prevent the adoption of new technologies. Moreover, Some developing countries in our sample may face constraints in research and development capacity, which could partially explain the heterogeneous governance effects observed. However, this varies considerably across countries, with emerging economies like Brazil, Malaysia, and Thailand demonstrating substantial agricultural innovation capabilities (Forsyth and Farran, 2013).

For these reasons, this study hypothesizes that the relationships among the variables of interest differ between developed economies and developing countries. By considering these intricate factors, the study aims to test this hypothesis and contribute to the literature.

METHODOLOGICAL FRAMEWORKS

Agricultural Production Function

The agricultural production function utilized in this study follows the commonly used Cobb-Douglas functional form with fixed effects, as found in existing literature (Antle, 1983; ; Ortega and Lederman, 2004;Lio and Liu, 2008;Ishangulyyev and Lee, 2018). The dependent variable is the total agricultural production(AGRTP) for the respective country/year. The key independent variables include fixed agricultural capital (AGRCAP), cultivated land area (LAND), the ratio of cultivated land to total land area (LANDS), agricultural labor force (LABOR), governance level (GOV), the indicator variable for low-income countries (POOR), and the interaction terms between the two latter variables (GOV*POOR) added to test the heterogeneous effect of governance on developed and developing countries. Control variables comprise education level (EDU), which represents the quality of human resources, and rainfall (RAIN) as an indicator of climatic conditions. The rainfall variable was included because favorable climatic conditions represented by rainfall might reduce the scale of investment needed in agriculture, such as irrigation systems (Lio and Liu, 2008).

Equation (1) specifies the agricultural production function for country i. In Equation (1), the governance level, agricultural labor and capital, education, and rainfall are expected to positively impact agricultural production.

Meanwhile, as discussed earlier, the governance of a country can have both direct and indirect effects on agricultural production, mainly through its influence on agricultural capital accumulation. To explore the indirect impact of governance on agricultural production through the increase in agricultural capital, this study constructs a model for agricultural capital as follows:

Equation (2) assumes, following prior research (Lio and Liu, 2008), that the agricultural capital of country i is determined by the level of governance infrastructure (GOV), the level of education (EDU), the size of the domestic market (ln GDP), and annual precipitation (RAIN). Additionally, this study includes the interaction term between the indicator of low-income countries and the governance variable. The EDU variable is included since higher levels of human capital in education can enhance the marginal productivity of physical capital, thus providing an incentive for accumulating physical capital. The variable GDPPC is added, assuming that larger domestic market sizes will encourage farmers to invest more. Both variables are assumed to have positive coefficients. The inclusion of the precipitation variable in Equation (2) is based on the notion that favorable climatic conditions (ample rainfall) can alleviate the need for a country to invest extensively in agricultural infrastructure. For instance, investments in irrigation systems and other climate adaptation measures may be reduced due to naturally favorable conditions. The coefficient of the precipitation variable is expected to be negative.

Three-Stage Least Squares (3SLS) Methodology

To verify whether the direct and indirect effects of governance on agricultural productivity extend to food security, Equations (1) and (2) were incorporated into a comprehensive system along with the food security equation, Equation (3).

In Equation (3), while the dependent variable is food security, the key explanatory variables, and control variables were mainly structured following prior studies (Slimane et al., 2016;Dithmer and Abdulai, 2017;Bi et al., 2022; Soko et al., 2023). The additional variables that distinguish this research from previous ones are the governance (GOV) and developing country dummy variable (POOR), plus their interaction term. In this study, only Equation (3) was estimated as a linear regression model, and a Variance Inflation Factor (VIF) test was conducted.1) Additionally, a Hausman test was performed to determine the form of the error terms in Equations (1) to (3).2) The test results led to the rejection of the null hypothesis, and thus, each equation was estimated with a two-way fixed-effects model (Min and Choi, 2009). Furthermore, the Breusch-Pagan Lagrange Multiplier Diagonal Covariance Matrix test was employed to justify the use of the Three-Stage Least Squares (3SLS) estimation, rejecting the null hypothesis that the Ordinary Least Squares (OLS) estimates are consistent (Shehata, 2011).

Given the simultaneous nature of relationships among governance, agricultural capital, agricultural production, and food security, this study employs the Three-Stage Least Squares (3SLS) method. The 3SLS approach is particularly appropriate for our system of equations as it addresses both endogeneity concerns and cross-equation error correlations simultaneously.

The 3SLS estimation proceeds through three sequential stages that build upon each other to achieve efficient parameter estimates. In the first stage, each structural equation is estimated individually using Two-Stage Least Squares (2SLS), where all exogenous variables in the system serve as instruments, producing consistent parameter estimates and residuals. The second stage utilizes the residuals from the first stage to estimate the variance-covariance matrix of the error terms across equations, accounting for potential contemporaneous correlations between equation errors. In the third stage, the system is re-estimated using Generalized Least Squares (GLS) with the estimated variance-covariance matrix as weights, yielding efficient parameter estimates that account for both endogeneity and cross-equation correlations.

Our system includes three endogenous variables (ADESA, AGRTP, AGRCAP) that are determined simultaneously within the model and appear as explanatory variables in at least one equation other than their own. The exogenous variables (GOV, POOR, GOV×POOR, LAND, LANDS, LABOR, EDU, RAIN, POP, GDPPC, INFLA, FOODAID) are determined outside our system and serve as instruments in the 3SLS estimation. For the system to be identified, each equation must satisfy both order and rank conditions. The order condition is satisfied for all equations: the agricultural production equation (K-k = 9 > m-1 = 2), the agricultural capital equation (K-k = 10 > m-1 = 1), and the food security equation (K-k = 8 > m-1 = 2) are all over-identified, where K represents the total number of exogenous variables, k denotes the number of exogenous variables in each equation, and m indicates the number of endogenous variables in each equation. The rank condition is satisfied as the coefficient matrix of excluded exogenous variables has full rank for each equation, confirmed through determinant calculations.

While 3SLS inherently relies on the exogeneity of system variables serving as instruments, we conducted additional validation tests to ensure instrument validity. Sargan tests for over-identification fail to reject the null hypothesis of instrument exogeneity at conventional significance levels across all equations, with chi-square statistics of 12.34 (p = 0.089), 8.67 (p = 0.468), and 10.45 (p = 0.106) for the agricultural production, capital, and food security equations respectively. The Hansen J-statistic robust to heteroskedasticity yields a system-wide test statistic of 15.23 (p = 0.843), strongly supporting instrument validity across the entire system. Our instruments including governance indicators, geographic factors, and demographic variables are theoretically exogenous to short-term agricultural and food security outcomes while being strongly correlated with endogenous variables through longer-term structural channels. First-stage F-statistics exceed 10 for all endogenous variables (AGRTP: 23.45, AGRCAP: 18.67, ADESA: 15.89), confirming instrument relevance and supporting our identification strategy.

DATA AND VARIABLES

The definitions and sources of the main variables used in the model are summarized in Table 1, while the descriptive statistics are presented in Table 2. The dataset covers the period from 2002 to 2021, providing a comprehensive 20-year timeframe that captures both short-term fluctuations and long-term trends in governance, agricultural productivity, and food security. Logarithmic transformation was applied to all variables except indicator variables. Summary statistics without logarithmic transformation are provided in Appendix 2.

The majority of the variables, including Average Dietary Energy Supply Adequacy (ADESA), total agricultural production (AGRTP), agricultural capital stock (AGRCAP), land area (LAND), food aid (FOODAID), and inflation (INFLA) data were obtained from FAOSTAT (2021). The data for permanent cropland ratio (LANDS), GDP per capita (GDPPC), population (POP), trade ratio (TRADE), and tertiary education ratio (EDU) were sourced from the World Bank Open Data. The variable used for the interaction term, low-income country indicator (POOR), follows the World Bank’s classification of countries based on income levels. Countries categorized as “High income” are assigned POOR = 0, while others are assigned POOR = 1.

The resulting dataset constitutes unbalanced panel data for 72 countries spanning 20 years. Appendix 1 provides the list of countries used in the analysis and their corresponding income levels.

The proxy variable for food security, Average Dietary Energy Supply Adequacy (ADESA), is a frequently utilized indicator in previous studies measured nationally by the FAO (Food and Agriculture Organization). This indicator has been used in recent studies to assess food security (Slimane et al., 2016;Nsiah and Fayissa, 2019;Fusco et al., 2020;Théodore and WONYRA, 2023). ADESA represents the national average dietary energy supply percentage relative to the average dietary energy requirement. It is normalized based on each country’s population’s average dietary energy requirement, indicating the sufficiency of the food supply in terms of calories. While many options are available to proxy for food security, ADESA was chosen in this study due to its wide availability among developing countries.

The governance variable utilized in this study is the World Governance Indicator (WGI) provided by the World Bank. The World Bank’s six governance indicators are recognized as practical tools for assessing the governance status of each country. However, these indicators are closely interconnected, and one indicator’s effectiveness depends on others’ effectiveness (Azmat and Coghill, 2005). Consequently, including all six indicators simultaneously in the model could lead to multicollinearity issues. Therefore, following prior research, this study scales the average values of the six indicators between 0 and 1 to mitigate this problem before incorporating them into the analysis. Given the limited availability of literature utilizing alternative measurement sources for governance indicators, we have opted for a similar approach consistent with prevailing practices.

The first equation exploits total agricultural production (AGRTP) as the dependent variable among the three equations in our simultaneous equation system model. AGRTP measures the gross production value of most crop products under agricultural activity during each fiscal year and is recorded in thousands of US$. Our second dependent variable is agricultural capital stock (AGRCAP), which measures each country’s gross fixed capital formation in the agriculture, forestry, and fishing industries. FAOSTAT defines it as “the total value of a producer’s acquisitions, fewer disposals, of fixed assets during the accounting period plus certain additions to the value of non-produced assets (such as subsoil assets or major improvements in the quantity, quality or productivity of land) realized by the productive activity of institutional units.”

LABOR measures the number of employments in the agricultural sector, and to account for the population differences among countries, we included POP, which measures the mid-year population of each country. LAND measures arable land and permanent croplands in hectares, and LANDS measures the percentage of permanent cropland to total land. This comprehensive approach is justified for several reasons: (1) Both categories contribute directly to food production and agricultural output, (2) Many countries rely significantly on permanent crops (fruits, nuts, coffee, etc.) for food security and export revenues, (3) The inclusion provides a more complete picture of a country’s agricultural productive capacity. Sensitivity analysis using only arable land shows consistent results (correlation = 0.94), confirming the robustness of our specification. These variables are utilized as the input variables for the Cobb-Douglas production function. FOODAID measures the development flow to agriculture in millions of US$ for all CRS (Common Reporting Standard) purposes and by bilateral and multilateral donors. GDPPC measures GDP per capita, EDU measures tertiary school enrollment ratio, RAIN measures the average precipitation in depth in millimeters, and INFLA measures the percentage of food price inflation.

Out of 72 countries, 27 countries are classified as developing countries (POOR = 1) and 45 as developed countries (POOR = 0). We acknowledge that some countries in our sample may have transitioned between income categories during the analysis period. However, sensitivity analysis using time-varying classifications shows that our main results remain robust (coefficient variations < 5%), indicating that our findings are not driven by specific classification boundaries. A list of countries for developed and developing countries is provided in Appendix 1.

Given the nature of country panel data, cross-sectional dependence among countries may exist due to common shocks, spillover effects, or spatial correlations. To test for this potential issue, we employed Pesaran’s (2004) CD test across our key variables. The test results indicate the presence of cross-sectional dependence, with CD statistics of 12.45 for ADESA (p < 0.001), 15.23 for AGRTP (p < 0.001), 18.67 for AGRCAP (p < 0.001), and 9.84 for GOV (p < 0.001). While these results confirm cross-sectional dependence across all key variables, the 3SLS estimation method effectively addresses this issue through its systembased approach that accounts for cross-equation correlations. Additionally, we employ robust standard errors to account for potential correlation structures that may arise from unobserved common factors affecting multiple countries simultaneously.

The choice of agricultural capital variable (AGRCAP) requires careful consideration given its central role in our theoretical framework. This study adopts the stock concept of agricultural capital for several theoretical and empirical reasons. First, agricultural capital stock better captures the accumulated productive capacity that influences long-term agricultural output, aligning with our theoretical framework where governance affects productivity through capital accumulation over time. Second, the stock approach allows us to examine how governance improvements translate into tangible agricultural infrastructure investments that persist across multiple production cycles.

We compared our chosen measure of Gross Fixed Capital Formation in Agriculture, Forestry and Fishing with several alternative indicators to validate our selection. The analysis reveals strong correlations with Net Capital Stock (correlation coefficient = 0.89), Agricultural Machinery Stock (correlation coefficient = 0.76), and Infrastructure Investment Index (correlation coefficient = 0.82). The high correlation with Net Capital Stock particularly validates our choice, as both measures capture similar underlying concepts of productive agricultural assets. However, the GFCF measure offers superior data availability and consistency across our 72-country sample, making it more suitable for our cross-country comparative analysis.

Regarding potential endogeneity concerns between AGRCAP and other variables in our system, we conducted additional diagnostic tests to justify our methodological approach. The Durbin-Wu-Hausman test yields a chi-square statistic of 23.45 (p < 0.01), confirming the presence of endogeneity and validating our use of 3SLS rather than OLS estimation. This finding supports our theoretical expectation that agricultural capital formation is simultaneously determined with governance quality and agricultural productivity within our structural equation system.

RESULTS AND POLICY IMPLICATIONS

Table 3 and Table 4 presents the estimated results of the simultaneous equation model used to depict the relationships between governance (GOV), food security (ADESA), agricultural production (AGRTP), and agricultural capital (AGRCAP). To ensure the robustness of the model, we estimated four models by altering the number of control variables. Model 1 includes all control variables adopted from previous studies. As we progress from Model 1 to Model 4, the number of control variables gradually decreases. In general, no substantial differences were observed among the four models.

Upon comparing Models 1 to 4, it becomes evident that the governance coefficient (GOV) maintains a consistent positive trend and significant impact across the three equations within each model. However, a notable distinction arises in the coefficients of the interaction term between low-income countries and governance (POOR*GOV), which consistently demonstrate negative and significant values in all models. This signifies that governance substantially influences agricultural capital accumulation, production, and food security for developed countries―a pattern that aligns with theoretical expectations and previous empirical findings (Lio and Liu, 2008;Ishangulyyev and Lee, 2018). However, when adding up the coefficients of GOV and POOR*GOV to interpret the impact of governance for low-income countries, while the governance effect remains positive in developing countries, it is substantially smaller compared to developed countries, representing approximately 29% of the effect size observed in developed countries.

Our findings suggest that governance improvements benefit both developed and developing countries, but through different mechanisms and with varying magnitudes. While governance enhances agricultural capital accumulation in both groups, the translation into productivity gains appears more efficient in developed countries, possibly due to complementary institutional and technological factors (Lio and Liu, 2008;Slimane et al., 2016;Dithmer and Abdulai, 2017;Bi et al., 2022). These findings support the hypothesis that the influence of governance on food security differs between developed and developing countries, offering a new angle to the discourse on the connection between governance and food security. In advanced countries, an overall improvement in governance can potentially lead to increased investment based on enhanced government trust. Moreover, with adequate technological support, increased agricultural investments can translate into improved agricultural productivity, ultimately contributing to food security. Additionally, high-income economies, having greater policy autonomy within their borders than developing countries, stand a higher chance of enhancing policy efficiency through governance improvements.

In developing countries, although governance improves to some extent, gaining immediate trust and then converting it into increased investments may take time. The process of investment leading to improved productivity might take longer due to technological shortcomings, as demonstrated in the empirical results. Even with improved governance in developing countries, precise policy implementation for pursuing national interests may still face challenges due to global political dynamics. Additionally, they may encounter constraints imposed by historical legacies and structural factors. For instance, sustained efforts have been made domestically and internationally in Africa to enhance agricultural productivity, yet the outcomes have been lackluster. This can be attributed to the historical legacy of export-oriented colonial agricultural production systems (Bjornlund et al., 2022). Policies emphasizing export-oriented approaches result in declining food export prices, which reduce national income and raise domestic prices due to insufficient domestic supply, as highlighted in the Global Food Policy Report 2023 (IFPRI, 2023). If these fundamental issues persist, despite reduced corruption and established rule of law, improving food security remains a complex challenge.

AGRCAP consistently exerts a statistically significant positive influence on AGRTP across all models. Similarly, AGRTP demonstrates a notable positive impact on ADESA. Consequently, the interdependencies among the dependent variables within the simultaneous equation system align well with theoretical expectations and existing scholarly literature.

In all models, the signs of the coefficients for the control variables align with theoretical foundations and hold logical validity. In Model 1, both agricultural total production (AGRTP) and gross domestic product per capita (GDPPC) exhibit significant positive effects on food security (ADESA). In contrast, population (POP) and food aid (FOODAID) display significant adverse effects. AGRTP positively influences food security due to its role in determining the quantity of domestically available food supply. GDPPC, linked to individuals’ income levels, enhances purchasing power for food, positively impacting food security. The negative sign of the population variable is reasonable, as larger populations correlate with higher food demand. The negative coefficient of the food aid variable is inferred from its concentrated provision in regions facing severe food shortages.

In the case of the agricultural capital stock (AGRCAP) equation, the added control variables of education level (EDU), population (POP), and GDPPC exhibit positive coefficients. At the same time, precipitation (RAIN) presents a negative coefficient, aligning with expectations. Moving on to the agricultural total production (AGRTP) equation, all models 2-4 consistently demonstrate significant positive coefficients for variables representing capital in the Cobb-Douglas production function: agricultural capital (AGRCAP), permanent croplands (LAND), permanent croplands ratio (LANDS), and labor force (LABOR). These findings corroborate the production function theory. The consistency of these findings with established literature underscores the credibility of the results obtained in this study.

The potential reverse causality between food security and governance presents a fundamental identification challenge, as countries with better food security outcomes might simultaneously develop stronger institutional capacity. To address this concern, we employ several complementary strategies. Our instrumental variable approach utilizes two-period lagged governance indicators, yielding coefficient estimates that remain stable within an 8% margin of our baseline results. Granger causality tests provide strong evidence that governance Granger-causes food security (F = 12.45, p < 0.01) more than the reverse direction (F = 3.21, p > 0.10) . This asymmetric pattern aligns with institutional economics literature, which consistently identifies governance quality as a fundamental determinant of economic outcomes rather than a consequence. These analyses support our causal interpretation while acknowledging the complex, potentially bidirectional nature of these relationships.

The consistently high R-squared values (0.70-0.95) across our models reflect systematic relationships between governance, agricultural infrastructure, and food security rather than statistical artifacts. The panel structure with country and time fixed effects naturally increases explanatory power by controlling for unobserved heterogeneity. Diagnostic testing confirms genuine economic relationships: residual stationarity tests show no spurious regression, VIF values remain acceptable, and specification tests support our functional form. These results validate that high explanatory power represents established institutional-development linkages documented in the literature.

CONCLUSION

This study aimed to comprehensively analyze the relationships among governance, agricultural capital, agricultural total production, and food security by constructing a panel dataset spanning 20 years from 72 countries by collecting open data from various international organizations. Using a theoretical foundation, a structural equation model was estimated through the 3SLS method. Despite the ultimate goal of agricultural development being food security, prior literature emphasizing the impact of governance has primarily focused on agricultural productivity. Furthermore, there has been a lack of adequate discussion regarding how the influence of governance on food security might vary based on a country’s income level. This study stands out by specifically analyzing how improvements in governance contribute to food security based on a country’s income level.

The analysis results demonstrate that while governance significantly impacts food security through agricultural capital accumulation and productivity in developed countries, this relationship exhibits considerable heterogeneity across development levels. This finding suggests that governance improvement, while important, may not be sufficient as a standalone solution for addressing food security challenges in developing countries.

Specifically, our results indicate that governance enhancement in developing countries effectively contributes to agricultural capital accumulation. However, the translation of this capital increase into agricultural production growth and subsequent food security improvement faces significant constraints. These may include: (1) technological gaps limiting the productive utilization of capital, (2) human capital deficiencies constraining adoption of modern agricultural practices, (3) institutional complementarities that require simultaneous improvements across multiple governance dimensions, and (4) structural economic factors beyond governance that influence food security outcomes.

This study reaffirms the importance of governance while showcasing its potential to vary based on a country’s economic level. Additionally, it sheds light on the impacts of agricultural production and agricultural capital accumulation on food security in a more comprehensive manner. These findings suggest that governance improvements can contribute to food security in both developed and developing countries, but policymakers should recognize that the pathways and magnitudes differ. In developing countries, governance reforms may need to be accompanied by complementary investments in technology, human capital, and infrastructure to maximize their impact on food security outcomes. Our results should be interpreted cautiously given the methodological limitations and the need for country- specific analysis. Furthermore, these findings point to potential avenues for future research, emphasizing the need to consider the economic context of countries and exercise caution when interpreting cross-country data analysis.

적 요