International trade and the stability of food supplies in the Global South

For data on caloric consumption, crop production, exports, and imports, we rely on FAO food balance sheets (FAOSTAT 2015). We calculate domestic food supply for each crop as production + imports − exports in each year. Data on biophysical redundancy is taken from Fader et al (2016), which estimates the amount of land that is currently unused for agriculture but is suitable for growing crops. We use the most recent available data, which in this case is from 2012. In our exposure to imported variance analyses, we use data from FAOSTAT for country-level crop yields and harvested area (FAOSTAT 2019). Trade matrices for 1985–2010 are derived from bilateral trade matrix estimates of the FAO for wheat, maize, and rice. Included in the analysis are all countries from Africa, Asia, and the Americas (excluding the US and Canada).

Here, we use 'stability' to imply the absence of variability. Stability is a key temporal determinant of food security (Gross et al 2000), and affects the physical flow of food, associated with three dimensions: availability, accessibility, and utilization. We furthermore use the import dependency ratio (IDR) to estimate the relative importance of imports. IDR is calculated as the ratio of imports to total domestic supply (sum of production and imports, which also include food aid, net of exports, in tons). The caloric dependency ratio (CDR) is calculated as the ratio of calories derived from a crop to the of total per capita calorie supply. This is the only metric that is based on calories and provides an indication of the relative importance of a crop for the consumption of an average household of a country. We use the CDR to identify countries for which the respective crop is a relevant source of calories (here defined as all countries with a CDR > 10%). Both ratios are dimensionless.

We calculate crop yield, import, and export anomalies by removing the long-term trend in each country using a low-frequency Gaussian filter with a kernel standard deviation of three years, which is similar to a nine-year running mean. Absolute anomalies are calculated as deviations from this long-term trend, which represents changes in management and technology.

To calculate exposure to imported variance, we assume that crop production anomalies in an exporting country are distributed amongst importers proportionally to the fraction of exported crop sent to each country. This is equivalent to a fixed percent reduction on exports to all importing nations. We use the time-averaged trade matrix for each time period to determine the import/export relationships. We then use the historical record of crop production anomalies and the fixed trade matrices to calculate imported production anomalies in each year for each country, from which we can calculate imported production variance for each country.

To analyze the effect of evolving trade patterns, we compute the variance of total supply of the key crops analyzed under two theoretical scenarios: one in which exposure to supply variance is calculated assuming international trade relations remain fixed as they were during the 1985 to 1990 period (averaged), and one in which supply variance is calculated assuming international trade as it was from 2005 to 2010. This analysis is designed to isolate the effect of the changing patterns of trade by using the average trade matrix from six-year periods in combination with the observed crop failures over the entire period, thus asking what a response to the same crop failure events would be with different trade networks. Note that we do not include derivative products in this portion of the analysis. The six-year averages are used to limit the influence of one-off trades/outliers to create a robust network representative of that time period.

In our analysis, we use metrics that are based on the standard coefficient of variation (CV), i.e. standard deviation over the mean: (i) Imported CV, i.e. coefficient of variation based on imported variability, defined as standard deviation of detrended imported anomalies over the expected mean domestic supply quantity; and (ii) Total CV, i.e. coefficient of variation of total supply, defined as standard deviation of total domestic supply over the expected mean domestic supply quantity. These metrics normalize the variation by the mean, the main difference to the standard CV being that the different variations (imported, domestic production, total supply) are all normalized by the same mean, namely the expected mean domestic supply quantity. The resulting metrics are dimensionless and allow to compare across countries.

In figure 2, we show how much of the variance in a country's domestic food supply is a result of its variance in imported food for maize, wheat, and rice. We find that for rice, nations that have the highest total supply variability are import dependent nations in which variability of rice imports accounts for nearly all of the variability of total domestic supply, while for wheat and maize countries with high total supply variance tend to only import a fraction of that variance (figure 2). Generally, we observe a geographical clustering of countries with similar profiles, which differ depending on the crop.

Zoom In Zoom Out Reset image size

For maize, exposure to supply variance comes as a mix of domestic production variance and imported variance. Countries with higher total supply CV (> 0.2) are mostly located in Southern Africa (figure 2(a)). Both reliance entirely on local production and reliance entirely on imports can lead to high domestic food supply CV. We see countries with high total CV along both the 1:1 line (i.e. all variation in domestic supply comes from imports) and along the x-axis (i.e. almost all variation in domestic supply comes from variations in local production, as is the case for Paraguay, Ethiopia, and Georgia).

Zimbabwe is interesting in that it has a high total CV of maize supply, indicating unstable domestic supply of food, and yet it has ostensibly diversified its source of maize by both importing and producing domestically. Upon closer inspection, however, the source of maize is not diverse from an abiotic stress perspective. Zimbabwe sources the vast majority of its maize from South Africa. Synoptic-scale droughts driven by, for example, the El Niño Southern Oscillation are large enough to affect both regions simultaneously, which is why major crop failures occur simultaneously throughout the region (Funk et al 2018, Anderson et al 2019). An analysis of the crop yield anomalies and food supply anomalies confirms this (see Figure S1 (stacks.iop.org/ERL/15/074005/mmedia)). Diversifying domestic food supply, therefore, requires that a country import from somewhere that does not tend to experience crop failures in the same years.

For rice, countries with higher total supply CV (> 0.2, figure 2(b)) generally import their variability. These countries are either from arid regions in the Middle East and North Africa, such as Oman or Djibouti, or are small island states, such as Cape Verde or Sao Tome and Principe. Importantly, there are no countries with higher total supply CV that comes through domestic production instead of imports. This indicates that rice production in self-sufficient and exporting countries of the Global South has been generally stable over time. Interestingly, supplies have also been comparatively stable (total supply CV < 0.2) in countries in Western Africa, a region that is particularly prone to food supply shocks (Bren d'Amour et al 2016), with the exception of Gambia and Mauritania in Northwest Africa.

For wheat, countries in Central Asia tend to have the highest CV due to a mix of imported variance and variance of domestic supply (figure 2(c)). In Central Asia, both countries that rely exclusively on domestic production and those that import between a fraction or most of their food supply have CVs above 0.5. However, while the total CV is comparable for most of the countries from Central Asia, the CV of imports varies substantially. While this is unique to wheat, there are similarities to e.g. rice for countries with a total supply CV > 0.2 and < 0.5. Countries that fall into this category are generally highly import dependent and import all of their variability. Countries from North Africa are notably absent in the sense that they have total supply CV < 0.2. This is noteworthy since these countries have been identified as very vulnerable to wheat supply shocks (Bren d'Amour et al 2016).

The heterogeneity in our results indicate that there is no clear answer to our research question of what are the consequences of increased trade for the stability of wheat, maize, and rice supply. In some cases, imports are responsible for high overall variability in supply. In other cases, the answer is less clear. Here, countries might have a high IDR and import most of their variation. Still, the overall variability in supply might be low. These findings raise the question of what can countries do to stabilize food supplies? For instance, could countries that largely import their variability in supply grow more of their own food or do they import by necessity?

To explore these questions, we analyze a country's potential for agricultural production, which depends on its biophysical capacities, i.e. the amount of available land, available water and climate of a country relative to what is needed to grow food.

We find that a country's biophysical capacity does not generally explain high import dependency and/or high variability in total supplies (figure 3). In fact, with the possible exception of rice, there is no clear relationship between the biophysical capacity of a country, its reliance on imports, and the variability of total food supply. For instance, we find that countries with high import dependencies (> 0.5, indicated by the bubble size) do not necessarily have low biophysical capacities, as the examples of Botswana and Namibia (figure 3(a)) show. For maize, we further find that very high biophysical capacities do not lead to stable supplies, as exemplified by Georgia and Paraguay.

Zoom In Zoom Out Reset image size

For rice, the countries with the highest import dependency have close to no biophysical capacities, such as Djibouti, indicating that they are doing so largely out of necessity (figure 3(b)). Further, we find that countries from West Africa that are import dependent (IDR > 0.5) have ample biophysical capacities. However, with the exceptions of Mauritania and Gambia, the total supply variability is comparatively low (< 0.2), indicating that the necessity to substitute imports by domestic production is low.

We find that many countries that are importing wheat are not necessarily doing so out of necessity (figure 3(c)). Countries from Asia and Africa that have low biophysical capacities import more than 50% of their supplies, but Botswana, for example, also imports more than 50% of its supplies despite having sufficient biophysical resources. Furthermore, high biophysical capacity does not lead to more stable total supplies, as the spread of countries with a biophysical capacity index > 0.75 indicate.

In theory, high biophysical capacity should enable countries with high IDR (> 0.5) and higher total supply variability (> 0.2) to import less and increase domestic production for domestic consumption. Countries that fall into the same category but have low biophysical capacity, however, do not have this option. These countries could reduce exposure to supply variance by re-examine trading partners in light of exporting countries production variance.

To get an understanding of the stability of exports, we compared the variability of yields and exports of the top ten exporters (table 1). The exports markets of these crops are highly concentrated: the top ten exporting countries combine for 90%, 88%, and 78% of all exports for maize, rice, and wheat respectively. The stability of exports coming from these countries is hence highly important for import dependent countries and we can see a lot by just looking into the top ten.

Our findings show that exports of the top ten exporters have been comparatively unstable over time (table 1). With the exception of North American and some Western European countries, the CV of exports has been high (> 0.2) for most countries for both maize and wheat. We further find that the variation in export quantities cannot always be explained by the variability of yields. Brazil, for example, has exceptionally stable maize yields (with a standard deviation of percent yield anomalies of 7%) yet has amongst the highest variation in exports (CV of export of 0.74). This could indicate that big producers such as Brazil react to market dynamics, for example by substituting the production of maize with soybeans.

Table 1 does, however, demonstrate some remarkable differences between major exporting nations that could affect the stability of food supplies importing from these countries. For wheat, for example, Australia and Central Asian exporters have much higher yield variance and variations in exports as compared to exporters in North America and Europe. This may contribute to why countries in Central Asia have such high total supply variance. It would also imply that countries with high total supply variance that import from Australia or Central Asia could diversify their supply by importing from regions with more stable crop yields and exports.

Up to this point we have identified where supplies are unstable, and how both imports and domestic production can lead to supply instability (figure 2). Owing to the contribution of domestic production to food supply instability, it is not immediately clear how an increasing dependence on wheat, maize, and rice imports (figure 1) affects food supply stability.

In figure 4 we model the CV of total supply for two scenarios using different trade networks, each representing a network averaged over a different time period (figure 4). The 1985–1990 period of trade represents a time with more countries producing a greater fraction domestically. The 2005–2010 period represents a more import-dependent world.

Zoom In Zoom Out Reset image size

Holding all else equal, changes in trade networks have generally increased exposure to domestic food supply variance in the Global South, although the effect has been heterogeneous (figure 4). In some cases, for example maize in Africa and Asia, the median overall food supply variances remain largely unchanged (figure 4(a)). The median total supply variance of maize in the Americas, on the other hand, has increased notably, which indicates that changes in trade patterns have exposed the region to greater food supply variability. The median total supply variation has also increased for rice in Africa and for wheat in Asia. Figure 4 demonstrates that changing patterns of global trade have led to an increased supply variance for some crop/region combinations, even after accounting for changes in the frequency of crop failures over time.

The implications of food supply variability are far-reaching. The absence of stable supplies has important ramifications for food security (Wheeler and von Braun 2013), mostly by threatening food availability. It is important to recognize, however, that food supply variability also affects other dimensions of food security and may be amplified or mitigated by any number market forces. Supply variations, in particular supply shortages, may affect prices (Kornher and Kalkuhl 2013) as well as price volatility (Timmer 2008, Tadasse et al 2016), and therefore the accessibility dimension of food security. High prices can further affect production by determining what producers grow (Haile et al 2015), and might therefore also contribute to variabilities in supply. Faced with supply shortages, consumers might substitute crops with other crops (Benson et al 2008, Dorosh et al 2009). These implications highlight the importance of finding stabilizing measures in an increasingly globalized food system.

A number of measures can help to reduce variabilities in domestic food supply (table 2). Depending on factors such as biophysical capacities, countries can try to (i) stabilize their domestic production, (ii) increase their imports to decrease their dependence on unstable domestic production, (iii) increase their domestic production to decrease dependencies on imports, (iv) change their import suppliers, or (v) introduce measures to increase the resilience of the food system.

Implementing measures to stabilize domestic production (ad i) is primarily an option for countries with sufficient biophysical resources. These countries, exemplified by Ethiopia, are typically largely self-sufficient in their supply of the respective crop (here maize; see figures 2(a) and 3(a)), and not import dependent in most years. Rather, they depend on their domestic production, which is highly variable. Generally, measures such as deficit or supplemental irrigation promise more stable yields (Oweis et al 1998, Zhang and Oweis 1999), as do improved varieties. However, while there is irrigation potential in Ethiopia and SSA in general (Xie et al 2018), there are several scale-up constraints, including inadequate funding, human capacity constraints and limited private sector involvement (Awulachew 2019), and associated sustainability concerns (Rosa et al 2018).

Some of these factors can also constrain a country's option to increase imports in order to decrease the dependence on unstable domestic production (ad ii). Generally, the ability of countries like Ethiopia to engage in international trade may be limited by high transaction costs (Smale et al 2011), a lack of foreign currency reserves (Van Ittersum et al 2016), and poor institutional settings (Hatzenbuehler 2019).

At the same time, there are countries that are highly import dependent and import the variabilities in their supplies despite having sufficient biophysical resources (ad iii). This is true for Cote d'Ivoire and Liberia (see figures 2(b) and 3(b)), both of which import a large share of their rice supplies (Balasubramanian et al 2007). Since the vast majority is produced on rainfed drylands in both countries (78% in Cote d'Ivoire and 92% in Liberia, see Balasubramanian et al 2007), any effort to increase domestic production would have to also addresses (i). Potential measures could include improved rice varieties (Lançon and Erenstein 2002), or increasing the amount of irrigated wetland rice production.

High variability, import dependent countries with insufficient biophysical resources, could reconsider their import suppliers (ad iv) provided the large discrepancy in the stability of exports between even the most major exporters (see table 1). In order for imports to stabilize domestic supply, importing countries need access to stable supply of exports. These countries should therefore seek to import from suppliers with stable domestic production, stable exports, and uncorrelated crop failures.

Additional strategies to stabilize food supplies can be measures that promote food system resilience (ad v), such as the introduction of regionalized food distribution networks, stocks and grain reserves, or waste reduction (Schipanski et al 2016, Laio et al 2016).