A macroeconomic investigation into the relationship between worldwide income inequality and food security

(1)

1

A macroeconomic investigation into the relationship between

worldwide income inequality and food security

Alexandru-Marius Savu BSc Thesis Economics & Finance Student ID: 10621636 Academic Year: 2015-2016

Supervisor: Ron van Maurik Semester 2, periods 2 and 3

(2)

2

Statement of Originality

This document is written by Alexandru-Marius Savu who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

Abstract

This evaluation is carried out with the purpose of better understanding the relationship between income inequality and food security at a global, macroeconomic level. Various researchers in developmental literature argue that an unfair income distribution negatively influences the affordability of food items, as well as their nutritional value. This study aims to investigate to what extent these claims are supported empirically by engaging in a panel-data regression analysis on a sample of ninety countries throughout three years, 2012, 2013 and 2014. The assessment finds a significant negative correlation between income inequality and both affordability and nutritional value assuming both random and fixed-country effects and controlling for education, population distribution and openness to trade. Consequently, the findings of this analysis strengthen the positions of current literature on the subject and also bring support to the idea that inequality reducing policies are beneficial for society from a developmental point of view.

Key Words: income inequality, food security, affordability, nutritional value /

(4)

4

I. Introduction

A reasonably fair income distribution and sufficient availability, affordability and nutritional quality of food across the world are perhaps two of the core objectives that, in an ideal world, human societies would achieve. In fact, in the recently published Sustainable Development Goals (UNDP, 2016, p. 2), the United Nations Development Programme [UNDP] includes the need to eliminate world hunger and the target of reducing worldwide inequalities as its second and tenth objectives for 2030 respectively.

Nevertheless, in a report on the state of food security worldwide (FAO, 2015, p. 3), the Food and Agriculture Organization of the United Nations [FAO] reveals that approximately 795 million individuals are globally undernourished. Moreover, researchers from the International Monetary Fund [IMF] (Dabla-Norris, Kochhar, Ricka, Suphaphiphat, & Tsounta, 2015, pp.4-7) posit that rising income inequality is of growing concern in modern times, and that sustainable policies should increasingly focus on the requirements of the middle and poor societal classes in order to combat various detrimental effects of income inequality, among which slower GDP growth and suboptimal human resources use.

Intriguingly, despite the fact that the relationship between different factors such as supply availability and food security are rather well-studied and documented (see FAO, 2015, pp. 8-17), the same cannot be said about the influence of income distribution inequality within a given society on the extent to which households have access to safe and affordable nutrition. In other words, although one can understand from reports such as those cited above that a reduction in income inequality and a rise in worldwide food security are critical developmental targets, a causal macroeconomic relationship between the two aspects has seldom been discussed in specialized literature. If such a relationship were to be significant, it would further strengthen the case made for focusing on sustainable, inequality-reducing policies.

In order to be able to evaluate this aforementioned relation, the two main variables of interest have to be clearly defined. Throughout this paper, the term „food security‟ is used, as argued by Timmer (1999, p. 93), to represent a societal state in which all individuals have secure physical and economic access to enough nutrition to meet their dietary requirements for a productive and healthy life. The researcher further distinguishes three main dimensions of food security which are of relevance here: the availability of sufficient amounts of food of acceptable

(6)

6

quality, the affordability or access by households to enough economic resources to acquire appropriate nutrition and the utilization or nutritional value of available food.

Further on, the definition of „income inequality‟ is taken from Deininger and Squire (1996, pp. 565-568). In their paper, the researchers define inequality as the discrepancy in income availability between individuals within different countries. In other words, the higher the percentage of income accrued only to a small proportion of the nation‟s population, the higher the income inequality in that given society.

As explained in-depth in section II.1, according to existing literature (see e.g. Mussa, 2014; Medozza, 2011; Kawachi, 1999), income inequality has a negative effect on two of the three key-dimensions of food security outlined above, namely affordability and nutritional value. These are the two main channels evaluated in this paper. In theory, a larger proportion of the population in societies with higher levels of income inequality will have difficulties affording nutritious food of appropriate quality, compared to those societies where income is more evenly distributed. Furthermore, since income inequality is an economic phenomenon that has become more accentuated in recent years (Dabla-Norris, 2015, p. 6), the focus of this paper is limited to three recent years, 2012, 2013 and 2014, for which data is suitably available.

Given the above, the purpose of this paper is to determine to what extent an unfair distribution of income has had a significant impact on worldwide food affordability and nutritional quality in the period 2012-2014.

In order to provide useful insights into the above query, panel-data regression analysis with both random and fixed country effects is utilized on a large sample of nations throughout recent years such as to determine the extent to which the effects of income inequality on food affordability and nutrition are both significant and persistent over time. Included in this sample are both developed and developing economies. The purpose of this broad approach it to offer an inclusive general view that is not subject to inherent limitations of external validity. The exact methodology, including operationalized variables, is presented and discussed in detail in section III.

It is important, however, to accentuate from the start that the purpose of the following analysis is not to provide an exact quantitative relationship between income inequality and food security. Such a goal is too ambitious and falls outside the scope and possibilities of this limited research. Rather, the evaluation present in this paper is to serve as an (albeit minute) piece of

(7)

7

evidence that can be used to strengthen the claim presented above that inequality-reducing policies do have positive impacts on overall societal well-being, even from a developmental point of view.

The reminder of this paper is structured as such: The following section presents a comprehensive literature review in which current research on the topic is brought to the fore and evaluated in order for one to properly understand the relationship between income inequality, food affordability and nutritional value. Furthermore, throughout this section, alternative variables that have been identified as having a significant impact on food security are discussed. These will later serve as control variables in subsequent analyses. This section concludes by presenting the main working hypothesis of the present research. Further on, as mentioned above, the methodology of the evaluation is offered. In this section, all the relevant variables are operationalized and potential implications and limitations of such transformations are assessed. The research method, as well as its inherent limitations, is also stressed in this section. Subsequently, the regression results are presented and an in-depth discussion on their implications is carried out. Lastly, the paper concludes by offering a summary of the main findings, outlining various limitations and offering recommendations for future research.

II. Literature Review

This section of the paper addresses three main aspects of the research. First, in sub-section II.1, the main findings of existing literature on the relationship between income inequality and food security are presented and discussed in order to comprehensively outline two main theoretical causality channels. Sub-section II.2 evaluates alternative macroeconomic variables that can be relevant in explaining worldwide food insecurity. These elements will be used in later chapters as control variables. Lastly, sub-section II.3 clearly states the paper‟s main working hypothesis.

II. 1. Income inequality as a determinant of food insecurity: existing knowledge II. 1. a. Income inequality and food affordability

As previously mentioned, since this is a relatively new subject of interest in the field of development, specialized literature on the topic is still lackluster. Nevertheless, existing studies

(8)

8

do provide a great deal of useful insights in appraising the relation between inequality and affordability.

Perhaps the main channel through which inequality has been found to exacerbate the issue of food affordability is the so-called poverty penalty, defined by Mendoza (2011, p. 1) as the relatively higher cost incurred by poor individuals, when compared to those richer, in their participation in certain markets. In other words, when income inequality in a given society is high, affording goods and services among which basic necessities such as food becomes more difficult for those in the lower part of the income distribution. This leads to decreased food affordability in that given community.

There are four main theoretical channels through which the poverty penalty functions, as summarized by Mussa (2014, p. 3). First of all, serving the poor is simply more costly for producers, who must adjust their distribution, production and promotion capabilities in order to reach impoverished communities. Second of all, since the poor often suffer from greater liquidity constraints, they must buy food in smaller quantities and thus forgo quantity discounts. Such a process ultimately results in higher unit prices. Furthermore, these aforementioned liquidity constraints, combined with a lack of proper storage facilities in most poor communities results in the poor acquiring food at suboptimal times, when the prices are high due to exogenous reasons. Lastly, the poor might also incur higher search costs compared to the non-poor when trying to access appropriate nutrition.

The existence of the poverty penalty has also been evaluated in empirical on-the-ground studies. In a research effort conducted in Malawi, Mussa (2014, pp. 8-11) finds evidence of a poverty penalty in the maize market, one of the country‟s core staple foods. The researcher argues that the price schedule for maize is downwards sloping and that poor households (in both rural and urban areas) have to pay more for this specific food item. He posits that the main explanation for this phenomenon relates to the above-mentioned liquidity constraint faced by the poor.

In a similar effort, Weerdt, Dilon and O‟Donoghue (2012, pp. 3-6), evaluate the issue of bulk discounts in Tanzanian markets and find that, overall, poor households often cannot take advantage of such opportunities and must acquire goods in small weekly increments that overall increase the unit price. In a study realized in Colombia, Attanasio and Frayne (2006, pp. 21-22),

(9)

9

find that bulk discounts on rice, carrots and beans are all significant, yet difficult to access for those households finding themselves in the lower part of the income distribution.

Overall, according to both theoretical considerations and field analyses, the poverty penalty provides an important connection between income inequality in society and reduced food affordability.

II. 1. b. Income inequality and nutritional value

Appraising the relationship between income inequality and nutritional value (or food utilization) is a daunting task to engage in because of the difficulty in understanding and measuring nutritional value itself. Unlike affordability, which directly relates to prices, quantities and discounts, food utilization ties in with aspects such as sufficiency of access, meeting nutritional benchmarks and preventing health issues that arise from improper consumption (Timmer, 1999, p. 93).

In a comprehensive conceptual framework on the determinants of food security, Pieters, Guariso and Vandeplas (2013, p. 11), when discussing food utilization, posit that contemporary society is confronted with a so-called double burden of malnutrition, encompassing both a widespread prevalence of undernourishment, especially among young people in developing nations, and a rise in obesity, diabetes rates and other related chronic diseases. According to the same researchers, such a phenomenon can be explained by the relatively low price of food items that are rich in calories, yet contain little nutritious value. Such items are especially accessible to poorer households.

Further on, as stated by Ruel (2002, as cited in Pieters et al., 2013, p. 11), if households become richer, they tend to switch to more diversified diets, richer in vegetables, fish and dairy, which contain an increased number of essential nutrients. Summarizing these observations, those finding themselves in the lower part of income distributions in highly unequal societies will often only have access to lower quality, sub-optimal nutrition from the point of view of essential nutrients and will either be subject to undernourishment or, contrarily, obesity and related illnesses. This represents a negative consequence of inequality on food utilization.

This relationship between income inequality, improper nutrition and increasing health concerns is also appraised by Kawachi (1999, pp. 215-217). In his study, the researcher evaluates the claim made by Wilkinson (1996, as cited in Kawachi, 1999, p. 215), that the larger the gap

(10)

10

between the incomes of those rich and those poor, the worse the health status of citizens and

finds that authorities concerned with the well-being and health of society should indeed direct their focus to and engage in inequality-reducing policies.

Lastly, Pickett et al. (2005, pp. 670-671) find income inequality, as measured by the income share of the richest 20% to that of the poorest 20%, to be positively correlated with the percentage of obese men (r=0.48, p=0.03), the percentage of obese women (r=0.62, p=0.003), as well as diabetes mortality rates per one million people (r=0.46, p=0.04), further strengthening the hypothesis proposed by Pieters et al. that income inequality does indeed exacerbate health issues related to improper nutrition.

All things considered, income inequality appears to indeed exhibit a negative relationship with food utilization, in that the higher a society‟s level of inequality, the higher the proportion of people who have access exclusively to subpar, low quality nutrition that will eventually result in the worsening of the community‟s overall health.

II. 2. Alternative determinants of food security

Given that food security is one of the main sustainability goals present in the field of international development, numerous factors, besides income inequality, have been proposed and evaluated as being significant determinants of food security. If one wishes to appraise the relationship between income inequality and food security in an unbiased manner, these factors must be accounted for.

Foremost, both Pieters et al. (2013, pp. 8-12), as well as Charles et al. (2010, p. 816) identify the overall education level of society as playing a vital role in ensuring proper food security. According to the former researchers (Pieters et al., 2013, pp. 8-9), higher education plays a crucial role in diminishing several of the aforementioned causes of the poverty penalty, through the enhancement of an individual‟s cognitive abilities which ultimately results in higher employability, increased marginal productivity and diminished liquidity constraints. Moreover, according to the latter scientists, education also has a positive impact on nutrition and health, through the spreading of best practices relating to food storage, waste management and general societal awareness on the importance of proper nutrition. Thus, a higher level of education seems to enhance both food affordability and utilization.

(11)

11

Furthermore, literature on the subject also identifies the distribution of population in a certain country between urban and rural centers as being a significant determinant of food security. Unfortunately, unlike education, insights into this aspect are mixed, as explained by Tacoli, Bukhari and Fisher (2013, p. iv, and pp. 6-7). According to these researchers, since most individuals and households are net buyers of food, they often benefit from an urban establishment where employment is more readily available and wages are on average higher compared to rural areas. On the other hand, urban locations present their own sets of risks, in that urban residents are more sensitive to external economic shocks, changes in labor demand and liquidity short-comings. Thus, whether or not a higher urban proportion of the population has a positive or negative effect on food security is still a topic of debate.

A final factor often brought to the fore in specialized literature concerns a society‟s degree of openness towards trade with other nations. As explained by Diaz-Bonilla et al. (2010, pp. 29-32), openness to trade can benefit a nation‟s food security as long as trade practices are fair and do not severely disrupt agricultural facilities. This observation is particularly acute for developing nations where agricultural activities still represent a crucial factor in the community‟s overall income.

II. 3. The influence of income inequality on food insecurity: A working hypothesis

Taking all the insights discussed and evaluated throughout sub-sections II.1 and II.2 into consideration, one can conclude that, in theory, income inequality should have a significant negative impact on the two key dimensions of food security, namely affordability and nutritional value/utilization. This represents the main hypothesis of this research that shall be tested in the remainder of the paper.

One can furthermore observe that three main additional factors, namely a society‟s overall education level, distribution between urban and rural centers and degree of openness towards trade all exhibit a potential (however, not clearly established) positive impact on food security, although such implications fall outside the scope of this paper.

An econometric interpretation of this hypothesis, alongside the manner in which the required variables are measured and operationalized is discussed in the following section.

(12)

12

III. Methodology

This section outlines in detail the research methods utilized in this paper. First, sub-section III.1 discusses the manner in which the involved variables are measured and operationalized, alongside a justification for using these specific measures. This part concludes with a comparison between two countries for exemplification purposes. Subsequently, in sub-section III.2, the working hypothesis discussed above is transformed into a testable econometric form, which will later serve as the basis for analysis in section IV. Lastly, sub-section III.3 illustrates the econometric method employed, together with statistical characteristics such as sample size and inherent limitations. This sub-section also presents a discussion on the adequacy of the regression specification.

III. 1. Understanding the variables of interest

III. 1. a. Measuring and operationalizing the variables of interest

As previously mentioned above, when discussing specialized literature, measuring the two dependent variables, food affordability and nutritional value is a difficult task that has most of the time been conducted at a microeconomic, individual-household level (see e.g. Agwu, & Oteh, 2014, which study the effect of inequality on food security in a small state in South-Eastern Nigeria by using individual questionnaires). Since this research aims to offer macroeconomic, international insights on the topic, such measures are sadly inappropriate. Instead, this paper requires standardized, universally applicable measurements for these two dimensions that can be utilized in cross-country comparisons.

Fortunately, starting in October 2012, The Economist‟s Intelligence Unit has initiated the so-called Global Food Security Index, through which it tries to measure worldwide food security, precisely utilizing the three dimensions discussed by Timmer (1999, p. 93). In this sense, the program provides, for each of the three variables, an index number ranging from 0 to 100, which encompasses a great array of relevant characteristics. Theoretically, a country with an index score of 0 represents a nation characterized by prohibitively expensive, low quality nutrition, which is almost impossible to procure. Contrarily, all the citizens of a country with a score of 100 would have easy and affordable access to food items of the highest-quality. Nevertheless,

(13)

13

these represent theoretical scenarios that do not exist in the reality. Thus, the index, for the purposes of this paper, is useful as it allows comparisons between nations and not necessarily absolute assessments.

On the one hand, as explained by the researchers behind this project, the food

affordability index (AFF) takes into account and standardizes, among other insights, dimensions

such as food consumption as a share of household expenditure, gross domestic product per capita, access to finance for farmers and the presence of food safety net programs. In other words, the index provided by The Economist addresses some of the defining causes behind the poverty penalty described above, among which liquidity constraints and the existence of proper storage facilities. For this reason, although imperfect, it provides a useful measurement for the purposes of this paper.

Further on, the nutritional value index (which The Economist terms „Quality and Safety Index‟) (UTL) encompasses health-related dimensions such as diet diversification, micronutrient availability, protein quality and food safety, characteristics once again related to the literature insights described in section II above.

Although no index can entirely capture the full extent and complexity of food affordability and utilization worldwide, these measurements provided by The Economist work as decent proxies and have the undeniable advantage of allowing for macroeconomic, cross-country evaluations which have seldom been realized before in developmental studies, especially related to their connection with income inequality. For this reason, they are used here.

In terms of the main independent variable, measuring and operationalizing income inequality, while perhaps more intuitive than accounting for food security, is characterized by its own set of difficulties.

The most popular and utilized measure of income inequality in specialized literature is the so-called Gini Index provided by the World Bank, which measures the extent to which the distribution of income among individuals or households deviates from a perfectly equal distribution (The World Bank). In other words, the higher the Gini Index is, the more severe income inequality in society become. Unfortunately, while an appealing macroeconomic measure, the Gini Index cannot be used in this paper for the simple reason that data on this

(14)

14

measurement for recent years is not yet available for most countries, to an extent that it prohibits any valuable econometric evaluations.

Fortunately, the UNDP provide an alternative measure termed the Inequality-adjusted Human Development Index [IHDI], which basically “discounts” the Human Development Index [HDI] measure of different countries for dimensions related to social inequalities, most notably income, education and health (UNDP, 2015, pp. 1-2). In other words, given perfect equality, the IHDI measurement does not differ from the HDI. However, as inequality starts to increase, the IHDI falls below the HDI for that specific nation. The percentage difference between these two dimensions is termed by the UNDP the human development cost of inequality [HDCI] or the loss

to human development due to inequality. This percentage difference will be used in this paper as

a proxy for income inequality and thus as the main independent variable.

It should be stressed at this point that, most likely, the HDCI is not an ideal measure of income inequality as it also relates to another two key dimensions of societal inequality, namely education and health. Nevertheless, the HDCI and Gini Index should be sufficiently correlated to still allow for valuable insights to materialize.

The remaining control variables are treated as follows: the education level of a country is measured by the Education Index (EDUC) provided by the UNDP (2015, pp. 216-219), which once again aggregates data for numerous countries into one universal, standardized index based on mean and expected years of schooling. It is measured on a scale from 0 to 1. Further on, the rural against urban distribution (UR) of population is measured by the percentage of individuals who find themselves in urban establishments as evaluated by The World Bank. Lastly, the degree of openness to trade (TRD) of a nation is approximated by that country‟s exports of goods and services as a percentage of GDP, once again measured by researchers at The World Bank.

Table 3 in the appendix of this paper presents the summary statistics of the data utilized for the variables described above. From this figure, one can observe that large variations exist between countries both in terms of the food security measures considered, as well as the HDCI. The average affordability and utilization scores are 53.25 and 56.79 respectively, while the mean discount of the HDI due to inequality is 31.2%. Once again, it should be reminded that these

(15)

15

index scores are useful only as terms of comparison between different nations, so having a reference of their average allows one to understand how individual countries rank.

Given the above discussion, the main working hypothesis of this paper can be reworded as follows: The higher the level of income inequality in a given country, as measured by the HDCI defined above, the lower that country‟s food affordability and food utilization scores, as operationalized by the index levels provided by The Economist‟s Intelligence Unit.

III. 1. b. Measuring and operationalizing the variables of interest: a facilitating illustration

Since explanations concerning the paper‟s methodology can quickly become too technical and difficult to follow, this section allows one to better understand the dimensions described above by presenting an exemplifying comparison between two countries included in the study on different extremes, as seen below:

Table 1 – A comparison on relevant variables using 2014 data

Country HDI IHDI HDCI (%) AFF UTL EDUC UR (%) TRD (%)

Angola 0.532 0.335 58.684 32 33.5 0.474 43.27 48.75

Belgium 0.890 0.820 8.569 87.5 82.2 0.812 97.818 83.96

The table above, while not a piece of statistical evidence for the issue at hand, serves as a useful illustration that helps one better understand the involved dimensions. First of all, it is readily apparent that 2014 inequality in Angola is significantly higher than that present in Belgium – in the former nation, the HDI is 58.684% higher than the IHDI, compared to the 8.569% in the latter, meaning that “discounting” based on the above-mentioned societal inequalities is much stronger in Angola compared to Belgium.

One can further observe that Belgium scores higher than Angola in both food security dimensions provided by The Economist, having an affordability index score of 87.5 and a quality and safety index score of 82.2, compared to Angola‟s 32 and 33.5 respectively.

Lastly, the three control variables, as measured by the education index, the percentage of urban inhabitants and exports-to-GDP ratio are all higher in Belgium than Angola, meaning that

(16)

16

Belgian citizens, in expectation, enjoy more years of schooling and are more likely to find themselves in an urban establishment, compared to citizens of Angola. Furthermore, Belgium is more open to international trade.

III. 2. The relationship between income inequality, affordability and utilization: an econometric representation

The working hypothesis described in section II.3 above can be transformed in the following regression equations:

(1) AFFit = αi + β1 INit + β2 EDUCit + β3 URit + β4 TRDit + εit

(2) UTLit = γi + τ1 INit + τ2 EDUCit + τ3 URit + τ4 TRDit + Δit

Above, the two dependent variables, affordability and utilization, are represented by the AFF and UTL symbols respectively. The independent variables are measures of income inequality (IN), education level (EDUC), urban vs. rural distribution (UR) and the degree of openness to trade (TRD). α and γ are constant terms, while ε and Δ are error variables which serve to capture any omitted variables and random effects.

According to the theory analyzed, the position of this paper is that both β1 and τ1, the

coefficients on the level of inequality, will be significantly below zero. The other coefficients, on the control variables, will most likely be positive, although this is of little importance to the query at hand.

III. 3. The research method employed

As mentioned in the introduction of this paper, in order to provide relevant macroeconomic insights on the topic investigated, panel-data regression assuming both random and fixed-country effects is utilized. Included in the sample are a total of 90 countries listed in the appendix for which data on all the relevant variables are measured for the years 2012, 2013 and 2014.

(17)

17

First of all, all available information on the two food security indexes, income inequality, education, population distribution and degree of openness has been taken from their respective sources mentioned in sub-section III.2 and organized in panel data format. Subsequently, the data were imported to STATA (a statistical software) and the respective regressions were performed using the program‟s functions. Results are presented in a summarized form in the following section and are also illustrated fully in the appendix of this paper.

The main advantage of using this specific econometric technique, as compared to a simple cross-country evaluation for instance, is that panel-data regression allows one to draw conclusions related not only to the relationship between income inequality and food insecurity, but also to understand whether or not such effects are persistent over time. Furthermore, including a fixed-country effects specification is recommended in order to control for country differences that are persistent over time. Since a very diverse group of countries is utilized in the evaluation of this paper, such an approach decreases the probability of a bias in coefficient estimates quite significantly. In Figure 7 of the appendix, a so-called Hausman-test is realized which shows that, indeed, fixed-country effects seem to be an appropriate assumption.

The large number of data points included in the evaluation represents another strong point of such a technique, as this paper specifically aims to offer a macroeconomic, cross-country perspective on the issue, as compared to most microeconomic, local-level evaluations present in existing literature, in order to diminish the lack of external validity that many of the aforementioned evaluations exhibit.

Nonetheless, several disadvantages and limitations pertaining to the subsequent evaluation exist which must be acknowledged from the beginning.

Foremost, while the number of nations included in the study is sufficiently high, the fact that comprehensive data is only available for three years does limit the statistical power of panel-data regression. In other words, the low number of years on which this analysis is based does lead to an increase in the inaccuracy of results. Fortunately, as mentioned in the first section, this paper does not aim to provide an exact quantitative relationship between income inequality, food affordability and food utilization. Because of this, the heightened inaccuracy resulting from a limited number of years present in the regression is not a major concern. However, it should still be noted when interpreting the results.

(18)

18

Second of all, it is also important to note a concern of multicollinearity present in the regression specifications. As the HDCI measurement also accounts for education dimensions in its computation, it is quite likely that it exhibits a strong correlation with the Education Index itself. Consequently, standard errors on these two variables might increase. This, however, does not represent a sufficient argument to eliminate the education level as a control variable, because, while related, it is necessary for the education index to capture any changes in food security exclusively related to education. Eliminating this variable would ultimately lead to omitted variable inconsistencies and, hence, biased results in the coefficients of income inequality/HDCI.

Before presenting and discussing the findings of the regression evaluation, several tests on the adequacy of the regression technique employed are conducted on the data set. These are available in Figure 4 of the Appendix.

First of all, as one can observe in Figures 4a and 4b, the results of the so-called White hetero-skedasticity test show that, when assessing the relationship between income inequality and food affordability, the assumption of homogeneity cannot be rejected (p>0.1). On the other hand, when considering nutritional value as the dependent variable, the data do seem to exhibit a degree of hetero-skedasticity (p=0.0128), meaning that the variance in utilization differs for various levels of inequality. Nevertheless, such an inconvenient occurrence should theoretically only lead to an increase in the variation of results. As this paper does not wish to provide an exact numerical relationship, the method employed is still valid.

Further on, Figure 4c addresses the issue of multicollinearity in the data. From this, one can observe that income inequality as measured by the HDCI is negatively correlated with all three control variables employed in the given sample. Of relevance here is the correlation between income inequality and the education level which is particularly strong (r=-0.8061), implying that some degree of multicollinearity does exist in the data between these two specific variables, as expected, since the HDCI indeed also accounts for aspects of education inequality. Nevertheless, despite the fact that this does not dramatically influence the subsequent evaluation, one should still be aware of the inherent increase in variability brought about by this negative correlation.

Overall, while the above insights do not diminish the validity of the regression findings discussed below, one should be aware of such technical details if the results are to be fully understood and replicable.

(19)

19

IV. Results & Analysis

The purpose of this section is twofold: First, in sub-section IV.1 the outputs of the regressions outlined above are presented in a summarized table form and discussed. Subsequently, sub-section IV.2 contains a comprehensive evaluation of the results at hand which confronts them with the theoretical underpinnings outlined in the literature review above. Uncertainties pertaining to the data and limitations on the extent to which these findings can be generalized are also included here. This section also contains some insights relating to the most important implications of the regression results for developmental policy.

IV. 1. a. The influence of income inequality on food security: regression results

AFF UTL (1) (2) (3) (1) (2) (3) IN / HDCI -0.0899*** (0.0232) -0.0404** (0.0191) -0.0327* (0.0184) -0.0972*** (0.0202) -0.0571*** (0.0171) -0.0482** (0.171) EDUC 72.15*** (8.47) -55.66* (29.21) 61.06*** (8.62) -3.58 (27.07) UR 0.395*** (0.0636) 1.96*** (0.374) 0.288*** (0.0656) 0.276 (0.347) TRD 0.0533 (0.0401) -25.99 (24.24) -0.0079 (0.0392) 0.0141 (0.057) R2 0.5838 0.8687 0.9945 0.5578 0.7906 0.9935

Note: Above, one can observe a total of 6 panel-data regressions realized on sample of n=90 countries through the years 2012,

2013 and 2014. The first three regressions, in the column titled AFF, appraise the influence of the HDCI on food affordability in a simple base specification (1) and including controls for education (EDUC), urban vs. rural population distribution (UR) and openness to trade (TRD) - with random effects (2) and fixed country effects (3). The latter three regressions, in the column titled

UTL, evaluate the influence of the HDCI on the nutritional value of food. They can be interpreted analogously.

Standard errors are included in parentheses.

Statistical significance is as follows: *** significant at the 1% level; **significant at the 5% level; *significant at the 10% level; no asterisk implies a lack of statistical significance.

(20)

20

IV. 1. b. The influence of income inequality on food security: presentation of results

In terms of understanding the above findings, several key observations can be made. They are discussed here.

First of all, as theorized, income inequality, operationalized by the HDCI has a significant negative relationship with both food affordability and food utilization, as measured by the indexes provided by The Economist, in all the regression specifications utilized. In other words, according to this particular data set, the higher the level of inequality in a society, the lower the two dimensions of food security.

In terms of affordability, increasing the gap between the HDI and the IHDI by one percentage point leads to a decrease of 0.09 in the affordability index. This difference falls by more than half to 0.04 when using random effects and correcting for the other factors of influence discussed in specialized literature. When furthermore assuming fixed country effects, the coefficient estimate drops again to a still-significant 0.033. One can observe that the R2 measure is reasonably high in the second and third regression specifications at 0.8687 and 0.9945, implying a good statistical fit of the variables selected.

Furthermore, the results pertaining to the influence of inequality on utilization are similar in nature, in that, again, a significant negative correlation can be observed between the HDCI and the quality and safety index (-0.097 in the base form; -0.057 when using controls and random effects; -0.048 with controls and fixed country effect). In fact, when introducing the control variables, the relationship between inequality and nutrition remains more statistically significant (p<0.01 in both cases) compared to the relationship between inequality and affordability (p<0.05 and p<0.1). Whether or not this result arises from a stronger influence of inequality on utilization compared to affordability or is simply a statistical occurrence cannot be established based on the results above alone.

In terms of the controls utilized, the results are to an extent inconclusive. The random effects regressions suggest that both a society‟s level education, as measured by the education index, and the proportion of individuals living in urban establishments exhibit a strong positive correlation with both measures of food security (p<0.01 in both cases). Further on, the degree of openness, as operationalized by the exports-to-GDP ratio appears to have a negligible effect on the two dimensions (p>0.1 in all regression specifications). On the other hand, when introducing

(21)

21

fixed country effects, the influence of education changes, in that it becomes negative and not as strongly significant. Such a finding is quite intriguing as it goes against the position of current literature and, thus, requires further analysis. The effects of urban inhabitants and trade openness, however, do not change qualitatively between random and fixed effects regressions.

IV. 2. A comprehensive discussion on the relationship between inequality and food security

Overall, the hypotheses proposed in this paper have been confirmed by the set of data evaluated. As one can observe above, after controlling for factors related to education, distribution of the population and openness to trade, income inequality/ the HDCI is significantly negatively correlated with the two dimensions of food security analyzed - affordability and utilization at a macroeconomic, cross-country level. The findings above have several insightful implications discussed below.

First of all, the above regression results support the idea proposed by Medoza (2011, p. 1) and elaborated on by Mussa (2014, p. 3) related to the existence of a poverty penalty, whereby those finding themselves on the lower end of the income distribution in a society with heightened inequality will have difficulty accessing affordably priced goods and services, among which basic food items. Indeed, countries with high income inequality seem to suffer, on average, from worse levels of food affordability compared to societies where income is more evenly distributed.

Consequently, it appears to be wise for policy makers and economic analysts who wish to engage in developmental initiatives to understand to what extent the four causes behind the poverty penalty identified above (liquidity constraints, search costs, etc.) are of material significance in their nation. Such an evaluation can potentially aid with the elaboration of well-targeted, efficient policies that have the purpose of reducing national food insecurity and hence enhancing overall well-being.

Secondly, the above evaluation also provides certain macroeconomic empirical support for the ideas put forward by researchers such as Pieters et al. (2013, p. 11) and Kawachi (1999, pp. 215-217) who argue that diminishing to an extent income inequality in a given society has direct and indirect benefits on the health of that community‟s citizens. In this paper, a latter indirect link has been evaluated and evidence has been provided that individuals belonging to a

(22)

22

country characterized by higher HDCI will on average have access to lower-quality, nutrition-wise suboptimal food. This in turn can potentially exacerbate issues such as undernourishment, diabetes and other illnesses related to a highly improper diet, although an exact link remains to be established.

These findings further strengthens the case made by Dabla-Norris et al. (2015, pp. 4-7), who, on behalf of the IMF, argue that policy-makers wishing to enhance overall societal well-being should focus a part of their efforts on reducing the consequences of an unfair income distribution. Such actions, as appraised above, can have positive effects not only on GDP growth and optimal human resources use as argued by the same researchers, but also on a critical developmental target pertaining to health and proper nutrition.

In addition, the findings presented above, at least when considering random effects, strengthen the claim made by Pieters et al. (2013, pp. 8-12) that better education in society positively influences the food affordability index through an enhancement of the cognitive abilities of individuals. This ultimately results in better employment and lower liquidity constraints. Note that a potential positive feedback channel exists here, in that higher education increases cognition which ultimately affects liquidity constraints and thus diminishes the consequences of the poverty penalty. This in turn may lead to higher rates of education and lower income inequality which will once again diminish the poverty penalty and increase food security in society. Nevertheless, these insights are not supported by the fixed-country effects regression.

Furthermore, the data analyzed above also provide empirical support for the argument according to which a higher level of urban population is indeed positively correlated with both food affordability and nutrition (Tacoli et al., 2013, p. 6-7). The same cannot be said, however, for the relationship between a country‟s degree of openness to trade and its levels of food security as a significant relationship has not been found between this variable and either of the two food security dimensions appraised. Further investigation in this regard is, thus, necessary.

While the above evaluation does provide useful insights into the importance of accounting for income inequality when wishing to enhance a nation‟s level of food security, it does suffer from some inherent limitations that one should be aware of.

Foremost, as already mentioned in the introduction of this paper, the relationship described in this research‟s findings is highly unlikely to be quantitatively exact, due to several

(23)

23

econometric considerations among which a potential issue of reverse causality. In other words, while this paper has analyzed the theoretical influence of income inequality on food affordability and nutritional value, a reverse causality channel might also exist. In this sense, one might argue that members belonging to a society who has access to high-quality, affordable nutrition have more means of achieving personal and financial prosperity and thus bridge to a certain extent the gap in income distribution. If this is the case, a bias will materialize in the coefficient of the HDCI which needs to be investigated more in-depth.

Moreover, as already briefly discussed in the methodology section, the use of only three years of data, while sufficient to provide a rough understanding of the discussed relationship, poses a problem of statistical prowess for the panel-data evaluation conducted above. Again, this assessment should be continuously augmented as new data becomes available for the estimated coefficients to more accurately reflect reality.

Further on, the main purpose of this research from the onset has been to provide a macroeconomic view on an issue that has seldom been discussed in specialized literature. Despite the fact that this represents a strong point of the above evaluation, it also embeds some issues. Among these difficulties, the most important is that the results discussed above, while true on average to a certain extent cannot be applied as such to each individual country in the sample. Otherwise stated, a certain reduction in income inequality in Angola is unlikely to have the same impact on affordability and utilization as a reduction in Belgium of identical magnitude. For this reason, economic analysts should augment the above regressions with country-specific factors that will give them more exact insights into their society‟s relationship.

In order to address this specific limitation, two additional regressions present in Figures 5a and 5b of the appendix to this paper have been conducted. These regressions are similar to those discussed and evaluated above, only augmented to account for the difference between OECD1 and non-OECD member countries. In other words, the purpose of these additional regressions is to determine whether the influence of inequality on the two dimensions of food security considered is stronger or weaker in developing nations, compared to industrialized countries.

According to the results, when assuming random-effects in the data, the difference is significant. The influence of inequality on both food affordability and utilization is lower in

1

(24)

24

countries belonging to the OECD compared to non-member states. To some extent, this is an intuitive finding, as because of strong past industrialization in these nations, food has become more affordable and of higher nutritional standards with little regard to the inequality in society. On the other hand, given fixed-country effects, these differences appear to be insignificant. This result, most likely, originates from the regression specification, as a fixed-effects model already accounts for persistent differences between economies.

Furthermore, in Figures 6a and 6b of the appendix, two additional regressions are present, which augment those in Figures 5a and 5b with additional interaction terms between the control variables utilized throughout the analysis and a country‟s belonging to the OECD. The purpose of these two evaluations is to determine whether being a developed, industrialized nation still has a significant effect on food security, even when accounting for interactions with the controls utilized. As one can observe, this is indeed the case when assuming random effects, whereas a non-significant effect is observed when assuming fixed-country effects. In other words, the findings of figures 5a and 5b are strengthened with additional evidence.

While not part of the main research initiative of this paper, this differentiation between OECD and non-OECD member countries does provide an additional insight into the issue at hand which can be the focus of future, more sophisticated studies.

V. Summary & Conclusion

The aim of this paper has been to offer an insight into the degree to which an unfair distribution of income in society negatively impacts the community‟s levels of food security. As most literature related to this subject is the result of small-scale, microeconomic evaluations on the topic, this research wished to offer an international, cross-country perspective on the issue, something seldom realized before. In order to achieve this aim, the following structured process has been followed.

First of all, a number of recent relevant studies on the relationship between income inequality and food insecurity have been presented and critically discussed. Briefly put, according to specialized developmental literature, income inequality negatively affects two key dimensions of food security, namely affordability and utilization. The former link theoretically materializes itself through the poverty penalty channel, whereby numerous poor individuals in

(25)

25

highly unequal societies incur extra costs when deciding to participate in various markets. Further on, the impact of income inequality on nutritional value is explained by the fact that poorer individuals will only have access to suboptimal nutrition when inequality is high. This results in a double burden of malnutrition that negatively impacts societal health.

Furthermore, specialized literature also identifies additional factors as macroeconomic determinants of food insecurity. The most important of these elements are a society‟s overall

level of education, which is believed to positively impact both the affordability and nutritional

value of food through an enhancement of average cognition in society, the distribution of the population between urban and rural establishments, according to which a higher proportion of urban inhabitants increases employment possibilities and thus boosts food security, and a country‟s degree of openness to trade, which should also benefit affordability and nutritional value as long as the negative impacts of unfair trade practices are diminished. All these factors serve as control variables for the present analysis.

Consequently, the purpose of this paper has been to evaluate whether the two channels described above do in fact hold at a cross-country, international level. Otherwise stated, the position of this paper can be reformulated in a working hypothesis as follows: After correcting for education level, population distribution and degree of openness, the higher the level of income inequality in a given country, the worse off that specific nation will be in terms of both food affordability and utilization.

In order to empirically test the above claim, a panel-data regression evaluation is used on a large sample of 90 countries throughout three years, 2012, 2013 and 2014. Such a large sample of nations allows several generalizable findings, discussed above, to materialize.

For each of these countries, the main variables of interest had to be operationalized. In this sense, income inequality was measured as the percentage difference between the HDI of a country and the IHDI of that nation, the HDI “discounted” for inequalities in terms of income, education and health. Food affordability and utilization were both operationalized using two index measures provided by The Economist‟s Intelligence Unit. The control variables were approached using the education index, the percentage of urban population and the exports-to-GDP ratio.

In large, the findings of the regression analysis bring support to the arguments brought forth in the evaluated literature. In this sense, the results presented above show that income

(26)

26

inequality does indeed exhibit a significant negative correlation with both food affordability and utilization, after correcting for the influence of the other factors. Furthermore, the level of education in society and the proportion of urban population positively influence the two dimensions of food security, while a country‟s degree of openness to trade appears to have an insignificant role.

The results of the econometric evaluation also have a number of implications for policy makers and economic analysts, the most important of which being that policies targeting a decrease in income inequality will have a positive effect on a key developmental area, the food security of society. For this reason, one should also take this relationship into consideration when weighing the advantages and drawbacks of putting forth inequality reducing measures.

Nevertheless, while these findings do offer some valuable theoretical and empirical insights into the issue, they are still of limited value as such, because of several inherent limitations of the analysis.

Foremost, one has to consider the issue of construct validity when appraising the relationship. As previously mentioned, the HDCI is not an ideal measure for income inequality, as it also accounts for education and health related issues. In essence, this means that the results above do not actually isolate a perfect relationship between income inequality and food security. Once data becomes available, further research might be able to conduct a similar analysis using more exact measures such as the Gini Index that specifically account for income distribution imperfections and disregards other unnecessary aspects.

Similarly, it is difficult to say to what extent the indexes provided by The Economist accurately depict food affordability and nutritional values in the countries considered. Unfortunately, obtaining such data on a macroeconomic level in a standardized fashion is a daunting task. For the moment, these measures are a good working proxy that could be successfully used in future studies.

Another limitation of the present research is that no serious differentiation is made between developing and industrialized nations, meaning that the uncovered relationship cannot be assumed universally applicable for each nation in the sample. While a minor differentiation is made between OECD and non-OECD countries at the end of section IV.2, it is of superficial nature. An alternative to this approach would be to consider each continent individually. If such an effort were to be conducted, one might discover that the influence of inequality on food

(27)

27

security to be stronger or weaker in one cluster of countries compared to the other. Consequently, valuable insights could be derived as to how policy has to be adjusted according to a country‟s overall level of development.

Lastly, as already discussed, the use of only three years of data in the analysis implies difficulties for the statistical reliability of results. If one hopes to arrive as close as possible to an exact causal relationship between the two dimensions, further research should integrate more comprehensive data as it becomes available, in order for the relationship to become more precise.

(28)

28

Bibliography

Agwu, N. M., Oteh, O. U. (2014). Analysis of Income Inequalities and Food Security among Farmers in Abia State, South Eastern Nigeria. Scientific Papers Series Management,

Economic Engineering in Agriculture and Rural Development, 14(3), 7-14.

Attanasio, O. P., Frayne, C. (2006). Do the Poor Pay More? Institute for Fiscal Studies, 1-25. Charles, H. J., Beddington, J. R., Curte, I. R., Haddad, L., Lawrence, D., Muir, J. F., Pretty, J.,

Robinson, S., Thomas, S. M., Toulmin, C. (2010). Food Security: The Challenge of Feeding 9 Billion People. Science, 327(812), 812-818.

Dabla-Norris, E., Kochhar, K., Suphaphiphat, N., Ricka, F., Tsounta, E. (2015). Causes and Consequences of Income Inequality: A Global Perspective. International Monetary

Fund Strategy, Policy and Review Department, 1-39.

De Weerdt, J., Dillon, B., O‟Donoghue, T. (2012). Paying More for Less: Why Don‟t

Households in Tanzania Take Advantage of Bulk Discounts? University of Washington

Center for Studies in Demography and Ecology, 1-13.

Deininger, K., Squire, L. (1996). A New Data Set Measuring Income Inequality. The World

Bank Economic Review, 10(3), 565-591.

Diaz-Bonilla, E., Ron, J. F. (2010). Food Security, Price Volatility and Trade: Some Reflections For Developing Countries. International Center for Trade and Sustainable Development,

Issue Paper no. 28, 1-57.

Food and Agriculture Organization of the United Nations. (2015). The State of Food Insecurity

in the World 2015. Meeting the 2015 international hunger targets: taking stock of uneven progress. Retrieved from: http://www.fao.org/hunger/en/

Jahan, S. (2015). Human Development Report 2015. United Nations Development Programme, 1-274.

Kawachi, I., Kennedy, B. P. (1999). Income Inequality and Health: Pathways and Mechanisms.

HSR: Health Services Research, 34(1), 215-227.

Kawachi, I., Kennedy, B. P., Lochner, K., Prothrow-Stith, D. (1997). Social Capital, Income Inequality and Mortality. American Journal of Public Health, 87(9), 1491-1498. Mendoza, R. U. (2011). Why do the poor pay more? Exploring the poverty penalty concept.

(29)

29

Mussa, R. (2014). Do the Poor Pay More for Maize in Malawi? Munich Personal RePEc

Archive, 1-21.

Mussa, R. (2014). Food Price Heterogeneity and Income Inequality in Malawi: Is Inequality Underestimated. UN-WIDER Development Conference Helsinki, Finland. Retrieved from: http://www1.wider.unu.edu/inequalityconf/sites/default/files/IGA/Mussa_0.pdf Pieters, H., Guariso, A., Vandeplas, A. (2013). Conceptual Framework for the Analysis of the

Determinants of Food and Nutrition Security. FoodSecure, Working Paper no. 13, 1-45. Pickett, K. E., Kelly, S., Brunner, E., Lobstein, T., Wilkinson, R. G. (2005). Wider Income Gaps,

Wider Waistbands? An Ecological Study of Obesity and Income Inequality. J Epidemiol

Community Health, 59, 670-674.

Tacoli, C., Bukhari, B., Fisher, S. (2013). Urban Poverty, Food Security and Climate Change.

International Institute for Environment and Development Human Settlements, Working Paper no. 37, 1-29.

Timmer, C. P. (1999). The Macro Dimensions of Food Security: Economic Growth, Equitable Distribution, and Food Price Stability. International Agricultural Trade Research

Consortium, 91-119.

The Economist Intelligence Unit. (2015). Global Food Security Index. Retrieved from: http://foodsecurityindex.eiu.com/

United Nations Development Programme. (2015). Calculating the human development indices,

Technical Notes. Retrieved from: http://hdr.undp.org/sites/default/files/hdr2015_

technical_notes.pdf

United Nations Development Programme. (2016). Sustainable Development Goals (SDGs). Retrieved from: http://www.undp.org/content/undp/en/home/sdgoverview/post-2015- development-agenda.html

(30)

30

Appendix

In this appendix, the reader can find an alphabetical list of the 90 countries used in the above analysis and the exact STATA regression outputs mentioned throughout the text. By having access to this information, one can potentially replicate the study and modify it accordingly in order to provide additional insights into the topic2. Lastly, the reader also has access to a so-called test of data stationarity and two robustness tests of the data.

Table 2 – The nations included in the sample for analysis

Angola, Argentina, Australia, Austria, Azerbaijan, Bangladesh, Belarus, Belgium, Benin, Bolivia, Botswana, Brazil, Bulgaria, Burkina Faso, Burundi, Cambodia, Cameroon, Canada, Chad, Chile, Colombia, Congo, Costa Rica, Cote d‟Ivoire, Czech Republic, Denmark, Dominican Republic, Ecuador, Egypt, El Salvador, Ethiopia, Finland, France, Germany, Ghana, Greece, Guatemala, Guinea, Haiti, Honduras, Hungary, India, Indonesia, Ireland, Israel, Italy, Japan, Jordan, Kazakhstan, Kenya, Madagascar, Malawi, Mexico, Morocco, Mozambique, Nepal, Netherlands, Nicaragua, Niger, Nigeria, Norway, Pakistan, Paraguay, Peru, Philippines, Poland, Portugal, Romania, Russia, Rwanda, Senegal, Serbia, Slovakia, Spain, Sweden, Switzerland, Tajikistan, Tanzania, Thailand, Togo, Turkey, Uganda, Ukraine, United Kingdom, United States, Uruguay, Uzbekistan, Venezuela, Vietnam, Zambia.

Note: Bolded countries belong to the Organization of Economic Co-operation and Development / OECD

Table 3 – Summary Statistics of the Data

2

(31)

31

(32)

32

(33)

33

(34)

34

Figure 4 – Appraising the Adequacy of the Regression Technique Employed

Figure 4a – White Test for Hetero-Skedasticity (Affordability Regression)

Above, one can observe the fixed-country effects regression of affordability on inequality augmented with the squares of the original variables. According to econometric theory, if hetero-skedasticity is indeed present in the data, the coefficients on these quadratic variables should be jointly significantly different from 0. As one can observe in the F-test conducted above, this is not the case (p=0.4125), meaning that the assumption of homogeneity in data is not violated.

(35)

35

Figure 4b – White Test for Hetero-Skedasticity (Utilization Regression)

The results presented above should be interpreted in a similar manner with those discussed in Figure 4a. In this scenario, however, the coefficients on the added quadratic variables do in fact appear to be jointly significant (p=0.0128), implying that some degree of hetero-skedasticity is present when discussing the relation between utilization and inequality.

(36)

36

Figure 4c – Correlations Table between the Variables Considered

In this table, one can observe the coefficient of correlation between the four independent variables utilized in the regression analysis. Normally speaking, if the absolute value of such correlations is higher than 0.7, a case of multicollinearity in the data set can be inferred. In this case, this happens in two instances, namely in the relationship between the HDCI and the education level (r=-0.8061) and the education level and urban versus rural population distribution (r=0.7769).

(37)

37

Figure 5 - Differences between OECD and non-OECD nations through HDCI

(38)

38

The two STATA outputs present in Figure 5a distinguish the influence of income inequality on affordability in nations belonging to the OECD and non-member states, assuming random-effects and fixed-country effects respectively.

The additional variable OHDCI is an interaction term between the HDCI and a dummy variable which takes the value 1 when a country is part of the OECD and 0 otherwise.

As one can observe, in the first regression, this variable has a significant positive coefficient (p=0.016), meaning that when a country is indeed an OECD member, the influence of inequality on food affordability becomes less negative, although still statistically significant.

On the other hand, when accounting for fixed-country effects, the coefficient on this additional variable is insignificant (p=0.418), most likely as a result of the regression specification utilized.

(39)

39

(40)

40

The two STATA outputs on the previous page distinguish the influence of income inequality on nutritional value in nations belonging to the OECD and non-member states, assuming random-effects and fixed-country effects respectively. They can be interpreted analogously to those present in Figure 5a.

(41)

41

Figure 6 – Overall impact of a differentiation between OECD and non-OECD nations

(42)

42

(43)

43

The two STATA outputs on the previous two pages distinguish the influence of income inequality on affordability in nations belonging to the OECD and non-member states, assuming random-effects and fixed-country effects respectively.

Unlike the regressions in Figure 5, here, there are four additional variables, namely OHDCI, OEDUC, OUR and OTRD, which all represent interaction terms between the HDCI and the control variables, and a dummy variable which takes the value 1 when a country is part of the OECD and 0 otherwise.

As one can observe from the F-tests above, when assuming random effects, the difference between OECD and non-OECD countries remains significant even with the addition of the new interaction terms (F=49.67, p<0.01), whereas this difference does not materialize when assuming fixed-country effects (F=0.22, p=0.9259).

(44)

44

(45)

45

A macroeconomic investigation into the relationship between worldwide income inequality and food security