How do non-reciprocal trade preferences affect the food exports and food availability per capita of Global South countries?

We investigate the historical effects of non-reciprocal preferential trade agreements (NRPTAs) on food exports and food availability per capita in 112 countries in the Global South to address concerns about their potential non-trade effects. Our empirical analyses use FAO Food Balance Sheet data for the years 1961 – 2013, covering 14 food categories and 91 product groups. We assess the link between NRPTA intensity - measured at the country level as the annaul sum of NRPTAs a country has in place - and the two outcomes using fixed effects dummy variable regressions. Our findings show that NRPTA intensity has a positive effect on food export performance and on food availability per capita, with heterogeneities across least developed, transition, and developing country groups, and its export effects do not jeopardize food insecurity.


Introduction
Embeddedness in world trade is considered a cornerstone for transitioning to a developed country (UNCTAD, 2022).Therefore, since 1971, developed countries have granted trade preferences to Global South countries through the Generalized System of Preferences (GSP) on a non-reciprocal and voluntary basis to improve economic development and alleviate poverty (UNCTAD, 2020).In contrast to reciprocal trade preferences, beneficiaries of non-reciprocal trade preferences do not have to liberalize their markets in return (European Commission, 2023). 1 The agro-food sector plays a vital role in the economic performance of Global South countries (Lin, 2018).Accordingly, non-reciprocal preferential trade agreements (NRPTAs) always cover agro-food products (Hoekman et al., 2005).However, the voluntary nature of NRPTAs makes them somewhat unpredictable.Products and countries can be excluded from NRPTAs, such as the GSP, by the donor country at any time (Kishore, 2017).In addition, the US GSP expires periodically and has always been renewed eventually, so an end is always likely (Hakobyan, 2020;Busch, 2021).In a sense, this uncertainty makes the schemes a tool for unregulated protectionism (Edjigu et al., 2023).In addition, trade wars and protectionism (World Bank, 2023), on the one hand, and the increasing number of reciprocal preferences in the form of Regional Trade Agreements (WTO, 2023a) and European Union (EU) Economic Partnership Agreements (European Commission, n.d.), on the other, potentially undermine the effectiveness of NRPTAs.Nevertheless, a recent study by Ridley and Shirin (2024) highlights the importance of NRTPAs for agricultural trade.In particular, the preference margin (i.e., the difference between the most-favored-nation tariff and the NRPTA tariff) provides an incentive for importers from the preference granting country to source agricultural products from countries in the Global South.
Global food security has improved steadily since the 1960s.This is reflected, for example, in the increase in calories available per capita per day or the decrease in the prevalence of undernourishment as a percentage of the total population (Hoddinott, 2021).Nevertheless, the 2007/2008 price shocks for agricultural commodities (Tangermann, 2016) or droughts due to global warming (Ahmed, 2020) show how vulnerable countries in the Global South are in terms of food security.The reduction or elimination of protection and subsidies in developed countries is considered to have a positive impact on food security in Global South countries (Díaz-Bonilla and Ron, 2010;Martin 2017).However, according to the academic literature, improved access to developed country markets can have a "twin" role.On the one hand, if developing countries export more food as a result of NRPTAs, this could make food scarcer in these countries and increase the prevalence of undernourishment (Mary, 2019).On the other hand, revenues from food exports could generate income and investment capital to increase agricultural and industrial productivity, which could increase the employment rate of the domestic population (Smith and Glauber, 2020).In particular, increased agricultural productivity could have a positive impact on food security.Accordingly, the aim of this study is twofold: First, we assess whether NRPTAs are effective in fostering the food export performance of beneficiary countries.Second, we evaluate the impact of NRPTAs on the food security of beneficiary countries.We measure export performance as the share of food exports in relation to the total food supply of a country.Using this approach, we reveal how much of the available domestic supply is used for exports.In addition, we proxy food security as the food availability per capita (i.e., food available for human consumption) of a country-one of the four dimensions of food security2 (FAO, 2008).
Our empirical analysis takes a historical approach spanning the years 1961-2013.We calculate measures of export performance and food availability per capita at the country and food category levels using the Food Balance Sheets provided by the United Nations' (UN) Food and Agriculture Organization (FAO, 2023).The NRPTA intensity at the country level is measured as the annual sum of NRPTAs based on an economic integration agreements database maintained by Baier and Bergstrand (2021).We further control for the intensity of other trade arrangements (i.e., reciprocal preferential and free trade agreement, customs union, common market, and economic union), for membership in the World Trade Organization (WTO), for labor and capital productivity in agriculture (USDA, 2022), and for gross domestic product (GDP) and population (Bolt and van Zanden, 2020).To investigate the link between NRPTA intensity and the two selected outcome variables, we apply fixed effects dummy variable regression.
Although many studies have investigated the relationship between NRPTAs and trade creation, indicating rather mixed effects (Seyoum, 2006;Herz and Wagner, 2011;Ritzel and Kohler, 2017;Ornelas and Ritel, 2020;Fernandes et al., 2023), less is known about their impact on the food security of beneficiary countries.The counterfactual (ex-ante) simulations of Aghajanzadeh-Darzi et al. (2015) consider the period from 2015 to 2025.The results of an applied general equilibrium model show that the removal of EU trade preferences will have a negative impact on beneficiaries' exports and macroeconomic performance.As the contribution of export gains and higher incomes is rather indirect, the observed impacts on food and nutrition security indicators are limited.Kersten (2018) analyzes the impact of reciprocal preferential trade agreements (i.e., regional and bilateral trade agreements) on food security in 93 lowand middle-income countries for the period 1990-2014.The results show heterogeneous impacts: While bilateral trade agreements have a negative impact on food security, regional trade agreements have a positive impact.To the best of our knowledge, a global and historical ex-post evaluation of the impact of NRPTAs on export performance and food security is lacking in the academic literature.Accordingly, our contribution to the literature investigating the impact of trade integration on export performance and on the food security of countries in the Global South is threefold.First, this is the first ex-post study to consider both dimensions (i.e., export performance and food security).Second, the data used, covering the years 1961-2013, allow us to provide evidence of the historical effect of NRPTA intensity on the two selected outcome variables.Third, our data basis enables us to capture the effect of all NRPTAs in force, thereby allowing us to provide general policy recommendations regarding the effectiveness of NRPTAs.
The remainder of this article is organized as follows: In Section 2, we present the databases, data and measurement issues, and the methods used.In Section 3, the results for export performance and food availability per capita are provided, and in Section 4, we discuss them.In Section 5, we conclude the paper and provide policy implications.

Databases
Our empirical analyses focus on countries in the Global South, which are considered underdeveloped or economically disadvantaged countries.During the five decades that comprise our databases and empirical analyses, the economic situation of a country can improve, and its status can therefore change, for example, from a transition to a developed country.Accordingly, we do not select countries that were classified as a developed economy in the last year of our analysis period (i.e., 2013).Following the UN (2013), we exclude the following countries or country groups: developed economies, including the USA, Canada, Japan, Australia, and New Zealand, the member states of the EU-15, the new member states of the EU, and other developed countries from the European continent (i.e., Iceland, Norway, and Switzerland).We also exclude Czechoslovakia, Yugoslavia, and North Korea, owing to long-term trade sanctions.The complete list of countries can be found in Table A1 in the Appendix.Our dataset consists of 15 transition, 64 developing, and 33 least developed countries.
To construct our variables that measure export performance and food availability per capita, we use data from the Food Balance Sheets for the years 1961-2013.From 2014 on, the Food Balance Sheet methodology has changed.Among others, the key difference between the new and old food balances methodologies is the absence of a balancer variable (FAO, n.d.).To avoid inconsistencies, we do not include the years from 2014 onward in our analyses.We select volumes of domestic production, imports, and exports (each measured in 1000 tons).The data cover 18 food categories, with each food category consisting of individual or multiple product groups.However, we exclude alcoholic beverages because they are not relevant for food security.Infant food was also excluded because production volumes are missing for many countries.In some cases, we combined food categories that are topically related or that consist of a single product group (i.e., 'sugar crops, sugar, and sweeteners,' 'pulses and tree nuts,' 'meat and edible offal,' 'milk and eggs,' and 'fish and aquatic products').Thus, we obtain 14 food categories for a total of 91 product groups, as presented in Table A2 in the Appendix.
The export performance of country c and product group p in year t is calculated as shown in Equation (1).

Export performance cpt =
Exports cpt × 100 where Exports cpt is the export volume of product p from country c in year t, Imports cpt is the import volume of p from c in year t, and Domestic production cpt is the volume of p produced in c in year t.The values of Export performance cpt range between 0 and 100%.
The food availability per capita (in kg) of country c and product group p in year t is computed as in Equation (2).
where Total population ct is the population of country c in year t.All other variables remain as defined in Equation (1).
For data on NRPTAs, we used the dataset constructed by Baier and Bergstrand (2021).The dataset also covers the period 195 country pairs, and contains six economic integration agreements-defined following Frankel (1997)-which are coded as follows: 1 = non-reciprocal (one-way) preferential trade agreement, 2 = reciprocal (two-way) preferential trade agreements, 3 = free trade agreements, 4 = customs unions, 5 = common markets, and 6 = economic unions.The overall dataset thus includes a sample of 2,572,440 observations.3 A detailed overview of NRPTAs, such as the GSP, the GSP+, the Everything But Arms regime, and the African Growth and Opportunity Act, can be found in a database hosted by the World Trade Organization (WTO, 2023b).
Our main variable of interest is the NRPTA intensity, which we calculated as the sum of NRPTAs per (beneficiary) country c in year t: To calculate NRPTA intensity ct , we use dummies of country pairs indicating "1 = non-reciprocal (one-way) preferential trade agreement."We summed up the NRPTA dummies for country c (i.e., an exporting country benefiting from non-reciprocal trade preferences) at year t.The economic integration agreements dataset provides no indication for NRPTAs offering market access to multiple countries.For instance, in 2013, the EU's GSP offered market access to its 27 member states.This implies that the NRPTA intensity for a certain beneficiary country exclusively exporting via EU NRPTAs takes the value of 27 in 2013.
We generate another variable that captures the intensity of other trade arrangements (i.e., OTA intensity ct ).It is constructed according to Equation (3) using economic integration arrangements coded as 2, 3, 4, 5, and 6.The WTO membership dummy variable is derived from the Centre d'Etudes Prospectives et d'Informations Internationales gravity database (Conte et al., 2022).
Data on gross domestic product (GDP) and total population are derived from Maddison's historical statistics (Bolt and van Zanden, 2020).The database contains data on GDP (in 2011 US$) and total population for 169 countries dating back to the eighth century.Based on data from the US Department of Agriculture (USDA, 2022), we compute two partial productivity measures (Federal Statistical Office, n.d.).First, agricultural labor productivity captures the efficiency with which human resources are used in the agricultural production process.The agricultural labor productivity of country c in year t is calculated by dividing the quantity of total agricultural output (in US$) by the quantity of total labor in agriculture (per 1000 persons economically active in agriculture) of country c in year t.Second, capital productivity measures the efficiency with which capital is used in the agricultural production process.The agricultural capital productivity of country c in year t is computed by dividing the quantity of total agricultural output (in US$) by the quantity of total agricultural capital stock (in US$) of country c in year t.
After excluding export performance values greater than 100 (2469 observations) and merging NRPTA and OTA intensity as well as further control variables, the dataset relies on 298,585 observations.The summary statistics for the variables used for the econometric analyses are presented in Table 1.

Data sources and measurement issues
We use data from the FAO Food Balance Sheets, which have two major advantages.First, trade and food security issues can be addressed using data from the same source.Second, data covering the years 1961-2013 allow for historical assessments of trade policy effects on trade and food security.As a limitation, the historical dimension may, in some cases, come with impreciseness.In this context, most imputation modules for missing data rely on measurements made in the past, so errors are likely (GSARS, 2017).This might explain why for the outcome 'export performance,' we had to drop 2469 observations greater than 100%.For cash crops, such as sugar, we observe very high values for food availability per capita up to 9700 kg.In such cases, conducting robustness checks by excluding or including food categories is unequivocally necessary.

Methods
To estimate the effect of NRPTA intensity on export performance and food availability per capita, we use a fixed effects dummy variable estimator.The major challenge hereby is to identify an unbiased effect of our variable of interest on the two outcomes.The omitted variable (and selection) bias is the major source of endogeneity facing studies estimating the effects of (reciprocal) free trade agreements on trade flows.Trade partners select themselves into free trade agreements, implying that unobserved factors determine the selection into free trade agreements.Thus, unobserved factors also affect trade flows; thus, studies estimating the effect of free trade agreements on trade flows usually yield biased estimates (Baier and Bergstrand, 2007).By contrast, NRPTAs are given exogenously to beneficiary countries by donor countries (Ito and Aoyagi, 2019;Ritzel and Kohler, 2017;Ritzel et al., 2018;Panda, 2020).This implies that NRPTA donor countries can decide upon country and product coverage, as well as on the period in which NRPTAs are granted.In this context, it is highly unlikely that countries in the Global South would withhold economic development just to benefit from NRPTAs.Consequently, NRPTAs can be considered an exogenously given treatment, enabling us to estimate the unbiased effect of NRPTA intensity on relevant outcomes.
We estimate a fixed effects dummy variable regression model (Wooldridge, 2012) using the Poisson pseudo-maximum likelihood (PPML) estimator.The PPML estimator provides consistent estimates in the presence of heteroskedasticity and is a natural way to deal with zero values of the dependent variable (Santos Silva and Tenryro, 2006).The regression equation takes the functional form shown in Equation (4): where Y represents the dependent variable (i.e., export performance and food availability per capita).Equation ( 1) is separately estimated for (i) export performance and (ii) food availability per capita.β 0 depicts the intercept, and β 1 captures the effect of the variable NRPTA intensity ct .X represents a vector of further control variables affecting export performance and food availability per capita.In particular, we control for the intensity of other trade arrangements (OTAs) aside NRPTAs, WTO membership, GDP, and population, as well as for labor and capital 3 There are some cases (coded as 7 = no country) defined as a country-pair/ year cell in which at least one of the two countries in a pair either does not exist or does not have independence.We drop these observations, which make up about 21% of the dataset.
productivity in agriculture.GDP and population are used as measures of economic size (Head and Mayer, 2014).GDP also captures the political stability of a nation, as both variables are positively correlated (Cervantes and Villaseñor, 2015).Labor and capital productivity are directly related to the production component of our two outcomes (Fuglie, 2018).Additionally, we include country fixed effects γ to account for time-invariant country characteristics (e.g., island or landlocked country), food category fixed effects θ to control for, for example, differences in prices and quantities across food categories and year fixed effects λ capturing, for example, shocks such as the 2007 and 2008 food crisis affecting all of the considered countries.ε cpt denotes the error term.Due to the exogeneity of NRTPAs, we assume that β 1 and ε cpt are not correlated, enabling us to estimate the unbiased effect.For NRPTA and OTA intensity, as well as for the WTO membership dummy variable, we compute average marginal effects.The remaining control variables are expressed in natural logarithms.Equation ( 4) is estimated for the aggregate of all food categories and country groups.To exploit country group heterogeneity, we separately estimate Equation (4) on sub-samples for the least developed countries, transition countries, and developing countries.This means that for the independent variables considered in Equation (4), we obtain an average effect of the estimators at the country group level.Additionally, to exploit food category heterogeneity, we separately estimate Equation (4) for the 14 individual food categories.

Results
In Section 3.1, we present descriptive results on the development of export performance and food availability per capita at the country level.In Section 3.2, we present the results of the PPML regression for all country groups (Section 3.2.1),for individual country groups (Section 3.2.2.), and for individual food categories (Section 3.2.3).As we are mainly interested in the effect of NRPTA intensity, we do not interpret and discuss the results of the control variables (i.e., results for OTA intensity, WTO membership, GDP, population, labor, and capital productivity).However, in many cases, the variables have the expected signs and meaningful magnitudes.The complete results tables can be found in the Appendix.

Table 1
Summary statistics for variables used for the econometric analyzes for all countries in the Global South.Fig. 1.Average export performance across all food categories for each of the selected countries between 1961 and 2013.

Descriptive results
In Fig. 1, we visualize the average export performance across all food categories for each country of the Global South over the decades that fall within 1961-2013.The value ranges of the average export performance are presented in decile rankings and colored from light blue (low values) to dark blue (high values).Countries that are not considered are colored gray.
During the 1960s and 1970s, the maximum average export performance was 18.5%.With the start of global trade liberalization in the 1980s (Clapp, 2017), the average maximum export performance constantly increased up to 32.1% for the period 2010-2013.Morocco was the top exporting country in the 1960s (average export performance = 18.5%), whereas Argentina was the top exporting country since the 1970s.From the 1980s onward, Latin and Central American countries, such as Uruguay, Chile, Costa Rica, Guatemala, and Nicaragua, as well as Southeast Asian countries, such as Thailand and Malaysia, increased their export performance.Accordingly, these countries constantly belonged to the upper decile.With the end of the Cold War in the beginning of the 1990s, Russia and satellite states, such as Belarus and Ukraine, increased their export performance.
Fig. 2 visualizes the average per capita food availability across all food categories (in kg) on the world map for each of the selected countries.For sugar, sugar crops, and sweeteners, we observe huge values for the measure of food availability per capita up to 9700 kg.Accordingly, we do not consider the sugar crops, sugar, and sweeteners category in the visualization of the development of food availability per capita presented in Fig. 2.
In contrast to the variable food export performance, we do not observe a constant increase in the maximum value for the average food availability per capita between 1961 and 2013.The maximum levels were reached in the 1970s (46.8 kg), 1980s (47.6 kg), and 1990s (48.0 kg).Afterward, the maximum value for the average food availability per capita decreased.From the 1990s on, we can observe a deterioration in food availability per capita for countries located in East and Southeast Africa, such as Ethiopia, Kenya, Tanzania, Zimbabwe, and South Africa.By contrast, the situation in China improved with the end of the Cold War in the beginning of the 1990s.
In the next section, we empirically test the role that NRPTAs play in the patterns we observe in Figs. 1 and 2.

All country groups
Table 2 shows the PPML regression results for all country groups and food categories.Due to high food availability per capita values, we excluded the food category 'sugar, sugar crops, and sweeteners'.However, this choice does not affect our findings.If we include sugar products, our findings remain the same in direction and statistical significance and only differ slightly in magnitude (see Table A3 in the Appendix).
For both models, we identify a statistically significant positive effect of the variable NRPTA intensity.An increase in NRPTA intensity by one unit boosts the export performance of a beneficiary country by, on average, 5.5 percentage points.An increase in NRPTA intensity by one unit is associated with a rise in food availability per capita by, on average, 0.04 kg per food category.As we show in the next section, the results obtained for the variable NRPTA intensity are driven by country group heterogeneity.

Individual country groups
Fig. 3 visualizes the average marginal effects of the variable NRPTA intensity on export performance and their 90% confidence intervals for different country groups.The results of the control variables can be found in Table A4 in the Appendix.
For the least-developed countries, our estimates reveal a statistically significant negative effect of NRPTA intensity on export performance.Here, a one-unit increase in NRPTA intensity is associated with a decrease in export performance by, on average, 6.8 percentage points.By contrast, for transition and developing countries, the effect of NRPTA intensity is positive and statistically significant.
Fig. 4 shows the average marginal effects of the variable NRPTA intensity and their 90% confidence intervals on food availability per capita (the results of the control variables can be found in Table A5 in the Appendix).Due to the high food availability per capita values for sugar, sugar crops, and sweeteners, we exclude this food category.However, this choice does not affect our findings.If we include sugar products, our findings remain the same in direction and statistical significance and only differ slightly in magnitude (see Fig. A1 and Table A6 in the Appendix).
For all individual country groups, the effect of NRPTA intensity on food availability per capita is statistically significant and positive.The largest effect of the variable NRPTA intensity can be observed for transition countries, where a one-unit increase in NRPTA intensity is associated with an increase in food availability per capita by, on average, 0.36 kg per food category.By contrast, the lowest effect can be identified for developing countries.Here, a one-unit increase in NRPTA intensity leads to an increase by, on average, 0.03 kg per food category.

Individual food categories
Fig. 5 shows the average marginal effects of the variable NRPTA intensity for individual food categories and their 90% confidence intervals for the model variant with the dependent variable export performance.For brevity, we do not present estimates separated by country groups.The regression results for the control variables can be found in Table A7 in the Appendix.
For most of the individual food categories, we identify a positive sign of our variable of interest, that is, NRPTA intensity.Only for oil crops is the effect of NRPTA intensity on export performance statistically significantly negative.Within plant-based food categories, the strongest effect of NRPTA intensity on export performance can be observed for pulses and tree nuts, for which a one-unit increase in NRPTA intensity leads to an increase in the export performance of beneficiary countries by, on average, 14.1 percentage points.The smallest magnitude of the NRPTA intensity estimator is observed for the food category cereals, for which a one-unit increase in NRPTA intensity only leads to an increase in export performance by 2.8 percentage points.Within the animal-based food categories, NRPTAs cause the strongest boost in export performance for fish and aquatic products.Here, a one-unit increase in NRPTA intensity causes the export performance of fish and aquatic products to increase by 10.8 percentage points.By contrast, the effect of NRPTA intensity on the export performance of animal fats is positive but statistically non-significant.
In Fig. 6, we present the average marginal effects of the variable NRPTA intensity for individual food categories and their 90% confidence intervals for the model variant with food availability per capita as the dependent variable.The regression results for the control variables can be found in Table A8 in the Appendix.
For most of the individual food categories, our variable of interest shows a statistically significant positive effect on food availability per capita.In the case of cereals, a unit increase in NRPTA intensity is associated with an increase in food availability per capita by, on average, 0.16 kg.Only for stimulants and spices do we observe a statistically significant negative effect of NRPTA intensity.

Discussion
Our work sheds light on the role of NRPTAs in direct trade and their non-trade effects.This is necessary because many policies have secondorder effects that are often overlooked when considering their first-order effects.We extend existing works by going beyond export performance to also consider food security.
Our main finding is that NRPTAs enhance the export performance and food security of beneficiaries.On export performance, our findings confirm theoretical predictions of the trade-promoting effect of trade agreements (Grossman and Sykes, 2005;Persson, 2015) and are also in line with empirical findings for the aggregate economy (Cirera et al., 2016;Gil-Pareja et al., 2017;Ornelas and Ritel, 2020) and agriculture (Afesorgbor et al., 2023;Agostino et al., 2010;Cardamone, 2011;Cipollina and Salvatici, 2010;Ridley and Shirin, 2024;Ritzel and Kohler, 2017;Scoppola et al., 2018).Our findings differ in the choice of trade performance measure.Existing estimates are based on observed trade values.We, however, measure export performance as observed trade values in relation to domestic food supply (i.e., domestic production plus imports).Using this approach, we reveal how much of the available domestic supply is used for exports.
On food security, we observe, on average, a positive effect of NRPTA intensity on food availability per capita.This implies that gains from embeddedness in world trade through NRPTAs are invested in improved agricultural inputs and production processes.Investments in, for example, water management practices and stress-tolerant varieties, in turn, have a positive impact on food security (Dar et al., 2013).Furthermore, by providing preferential access to specific markets, beneficiaries may see it prudent to expand their range of agricultural production to take advantage of new markets (Scoppola et al., 2018).This diversification can reduce dependence on a few staple crops and enhance food security.Therefore, opening up the economy for trade can help reduce the structural food supply inadequacy prevalent in many developing countries.Nevertheless, the reverse is also true if the preferences target only specific products (Persson and Wilhelmsson, 2016).Our findings are consistent with existing works that test the effect of trade openness on food security.In their global study, Dithmer and Abdulai (2017) find that a one standard deviation increase in trade openness increases dietary energy consumption by 93 kcal.
Beyond these average effects, we also observe heterogeneities across country groups and products.Although NRPTAs enhance the export performance of transition and developing countries, they reduce the export performance of least developed countries.This is indeed a surprising finding, given that LDCs are the target of many NRPTAs; however, it also confirms that the preferences alone may not be sufficient to enhance exports if the corresponding domestic trade infrastructure is weak.Yet, weak domestic institutions and trade facilitation measures coupled with the inability to meet standards in developed countries   characterize many LDCs (Kareem et al., 2023), and these factors may dampen the effectiveness of NRPTAs.Although this introduces further nuance into our main finding, it highlights the importance of going beyond aggregate findings and assessing effects at lower levels.The negative effect is also supported by a literature stream that finds that NRPTAs have marginal, null, or even negative effects on export performance (e.g., Cardamone, 2011;Fernandes et al., 2023;Gradeva and Martínez-Zarzoso, 2016).Our findings confirm the positive effect of NRPTAs on food availability per capita for all country groups.Overall, while the magnitude of the positive effect of existing NRPTAs is heterogeneous across country and product groups, our positive average effects support the consensus view that the observed export performance of beneficiaries may have been worse off without preferences (Grossman and Sykes, 2005).Regarding concerns about whether NRPTAs influence food insecurity, our findings show that, at least in our setting, this is not the case.Although food security has been challenged in recent times by major global events, including COVID-19 and the war in Ukraine, we affirm the role of trade preferences in enhancing food security. 4Our macro-level evidence is also confirmed by micro-level findings from Senegal (Van den Broeck et al., 2018) and multiple developing countries (Van den Broeck and Maertens, 2016) that horticultural exports do not jeopardize the availability of food.

Conclusions and policy implications
Integrating developing countries into the global trading system is high on the development agenda.Many uni-and multi-lateral efforts acknowledge the role of trade and globalization in economic development.One way developed countries attempt to achieve this policy objective is to offer NRPTAs to their developing country partners.Such NRPTAs often grant developing countries substantial free market access to developed countries' markets.How these trade preferences affect exports has been the subject of empirical scrutiny.However, increasing export potential may come at the expense of food availability per capita in the preference-receiving country.In this paper, we tackle this "twin" role of NRPTAs using historical data.We ask a simple yet policy-relevant question: How do NRPTAs affect export performance and food security (proxied as food availability per capita)?
Our empirical analysis combines data on imports, exports, food production-to compute our outcome variables, export performance and food availability per capita-and NRPTAs for 112 transition, developing, and least developed countries from 1961 to 2013.We then estimate a linear fixed effects model in which we regress our outcome variables on a count of country-specific NRPTAs in separate equations.Due to the exogeneity of NRPTAs-as recipients cannot self-select into being granted beneficiary status-we are able to estimate the unbiased effect of preferences on export performance and food availability.
Our main finding can be summarized as follows: On average, NRPTAs increase export performance and food availability per capita for beneficiaries.That is, beneficiaries of NRPTAs trade more and have better food security outcomes.Specifically, a unit increase in the count of NRPTAs a country enjoys increases its export performance by 5.5 percentage points and food availability per capita by 0.04 kg.However, these average effects are heterogeneous across countries and products.
Our findings are policy-relevant.First, regarding export performance, we show that the role of trade preferences cannot be denied.Over our long study period, they are instrumental in increasing the export performance of the recipient countries.Therefore, it is important that they continue to be used as tools for economic development.This requires relevant measures to ensure that the preferences are stable.Regarding food security, we find that some of the increased production triggered by the new export possibilities also enhances domestic food availabilty.However, we also see that for cash crops, such as spices and stimulants, the effect of NRTPAs on domestic availability is negative.Although these are not necessary to feed the population-and thus, negative effects are not directly detrimental for food security reasons-they could yield more export revenue for beneficiary countries when exported in their processed form.However, limited processing and value-addition potentials exist for these products in recipient countries, partly because higher tariffs are often charged on the processed form of these products.In the future, the policy focus should also target how these preferences can induce domestic value addition, not just the export of raw commodities.

Funding
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.Fig. 6.Average marginal effects the variable NRPTA intensity for individual food for the model variant with food availability per capita as the dependent variable. 4While our work shows that NRPTAs enhance food security, we cannot reach conclusions about the and diversity of available foods.This will nevertheless be an interesting point of departure for future studies in assessing how NRPTAs affect nutritional outcomes.For instance, Geyik et al. (2021) find that trade (more generally and not NRPTAs specifically) does not substantively improve the nutrient adequacy of most low/lower-middle income countries.C. Ritzel and D.-D.D. Fiankor Global Food Security 42 (2024) 100800

Food
availability per capita cpt = ( Domestic production cpt + Imports cpt − Exports cpt )

Fig. 2 .
Fig. 2. Average food availability per capita across all food categories for each of the selected countries between 1961 and 2013.

Table 2
PPML regression results pooled for all country groups and food categories.

Table A2
Fish and aquatic products Other aquatic animals, aquatic plants, cephalopods, crustaceans, demersal fish, freshwater fish, other marine fish, other mollusks, pelagic fishTable A3PPML regression results of the effect of NRTPAs on food availability per capita.The food category 'sugar, sugar crops and sweeteners' is included in this model.**, ** and * denote significance at 1%, 5% and 10% respectively.Robust standard errors are in parentheses.**, ** and * denote significance at 1%, 5% and 10% respectively.Robust standard errors are in parentheses.Note: The food category 'sugar, sugar crops and sweeteners' is excluded from the models. **

Table A6
PPML regression results of the effect of NRPTAs on food availability per capita (including sugar, sugar crops, and sweeteners).PPML regression results for individual food categories for the model variant with the dependent variable export performance.***, ** and * denote significance at 1%, 5% and 10% respectively.Robust standard errors are in parentheses.**, ** and * denote significance at 1%, 5% and 10% respectively.Robust standard errors are in parentheses.PPML regression for individual food categories for the model variant with the dependent variable food availability per capita.***, ** and * denote significance at 1%, 5% and 10% respectively.Robust standard errors are in parentheses.**, ** and * denote significance at 1%, 5% and 10% respectively.Robust standard errors are in parentheses.