Decline spills over—makes conflict inevitable

Figure 6 shows the response in terms of the number of conflict events per one million people to a commodity ToT shock in the full sample of Low- and Middle-Income Countries. A deterioration of the commodity ToT, on average, increases the frequency of conflicts. Specifically, we find that a negative ToT shock equivalent to one percent of GDP increases the likelihood of experiencing a conflict by 0.05 events per one million people in the year following the shock. The estimate of 0.05 events per one million people is equivalent to or explains 1.6 percent of the average country’s conflict events per year in our sample. Conflicts remain more likely for 2-3 years after a shock to income before the impact gradually decays. As discussed above, the literature highlights the potential for different directions of the effect of an income shock on conflicts. The opportunity-cost and state-capacity channels predict that a negative income shock increases conflict, whereas the predation effect would work in the opposite way. Previous studies trace the significance of each channel in specific economic sectors. For instance, the predation effect is often reported in oil and other natural resources sectors (e.g., Ross, 2004; Adhvaryu et al., 2021). Dal Bó and Dal Bó (2011) theoretically demonstrate that output price increases in capital-intensive sectors lead to fights over rent through predation, whereas the opportunity cost channel prevails in labor-intensive sectors. Our estimated negative response in the baseline implies that the former two channels should be dominant at the aggregate level. This is consistent with our use of a broad-based commodity ToT index that de-emphasizes idiosyncratic effects in specific sectors. Figure 6: Responses of the Number of Conflict Events to a Commodity ToT Shock, Baseline Specification Notes: The sign of the estimated coefficients is flipped and multiplied by 100. A positive response in the figure indicates an increase in conflict events following a negative commodity ToT shock. Shaded areas indicate the 68 and 90 percentiles of the responses (same below). INTERNATIONAL MONETARY FUND 19 4.2 Evolution of Conflict dyads We further analyze the impact of the commodity ToT shocks on the evolution of underlying conflict dyads. To do so, we decompose conflict dyads, defined as a pair of two opposing actors that are involved in conflict events, into those starting in a given year (“entry”) and those continuing from a past year (“continuation”). We count a dyad as “active” and therefore an “entry” if it causes at least one conflict event in a given year, but was not active in the previous year as a new conflict dyad. We define “continuation” of a conflict dyad if a dyad in a given year was active in the previous year too. Then, we use the number of entry and continuation of dyads as an outcome variable in the regression. The results in Figure 7 illustrate that negative commodity ToT shocks increase the number of continued conflict dyads and prolong existing conflicts in Panel (a). 11 This said, the effect is not strong enough to trigger an entry of a new dyad in Panel (b). The results imply that the increase in the total number of conflict events shown in Figure 6 can be attributed to the intensification of violent activities by existing dyads. It echoes a narrative that conflicts arise from complex contexts, including political, ethnical, and social factors. Economic factors may not be the main driver of a new occurrence of conflict but seem to play an important role in determining the duration of violent activities. Figure 7: Responses of Continuation and Entry of Conflict dyads (a) Continuation (b) Entry 4.3 Countries at Risks Given the statistically significant response of a commodity ToT shock on conflict intensity, a natural next question relates to the type of countries that are most at risk of conflicts when an economic shock hits. To answer this question, we first look at the differences in impact across country groups. Then we examine the relevance of each economic channel. Income level. Previous studies often associate income level with the occurrence of conflicts (e.g., Blattman and Miguel, 2010). However, it is less clear how the impact of a Terms-of-Trade shock differs across countries’ 11 The choice of outcome variables here is closely related to the studies that focus on the onset and duration of conflict of (e.g., Collier and Hoeffler, 2004; Brückner and Ciccone, 2010; Collier et al., 2004; Buhaug et al., 2009). These studies tend to use a binary variable that indicates the onset and presence of conflict at the national level, and thus do not differentiate dyads within countries. INTERNATIONAL MONETARY FUND 20 income levels. To see this point, we modify the regression equation (2) by accommodating differential impacts through interaction terms: 𝑦𝑦𝑖𝑖,𝑡𝑡+𝑠𝑠 − 𝑦𝑦𝑖𝑖,𝑡𝑡−1 = 𝟏𝟏𝑖𝑖,𝑡𝑡,𝑗𝑗 �𝛽𝛽𝑠𝑠,1 Δ𝑡𝑡𝑡𝑡𝑡𝑡𝑖𝑖,𝑡𝑡 + �𝛿𝛿𝑘𝑘𝑘𝑘,1 𝐾𝐾 𝑘𝑘=1 Δ𝑡𝑡𝑡𝑡𝑡𝑡𝑖𝑖,𝑡𝑡−𝑘𝑘� + (1 − 𝟏𝟏𝑖𝑖,𝑡𝑡,𝑗𝑗) �𝛽𝛽𝑠𝑠,0 Δ𝑡𝑡𝑡𝑡𝑡𝑡𝑖𝑖,𝑡𝑡 + �𝛿𝛿𝑘𝑘𝑘𝑘,0 𝐾𝐾 𝑘𝑘=1 Δ𝑡𝑡𝑡𝑡𝑡𝑡𝑖𝑖,𝑡𝑡−𝑘𝑘� +�𝛼𝛼𝑙𝑙𝑙𝑙 𝐿𝐿 𝑙𝑙=1 Δ𝑦𝑦𝑖𝑖,𝑡𝑡−𝑙𝑙 + 𝜇𝜇𝑖𝑖,𝑠𝑠 + 𝜆𝜆𝑡𝑡,𝑠𝑠 + 𝜈𝜈𝑗𝑗,𝑡𝑡,𝑠𝑠 + 𝜀𝜀𝑖𝑖,𝑡𝑡,𝑠𝑠 ⋯ (3) where 𝟏𝟏𝑖𝑖,𝑡𝑡,𝑗𝑗 is an indicator function that takes a value of one for given characteristic 𝑗𝑗 and zero otherwise. The IR of each group is represented by�𝛽𝛽𝑠𝑠,1� and �𝛽𝛽𝑠𝑠,0� for 𝑠𝑠 = 0, 1, 2, …, respectively. We split the sample into LICs and lower MICs (GNI per capita below US$4,045 in 2020) and upper MICs according to the World Bank’s income classification. We complement the classification by using eligibility to the concessional lending resources of the IMF (Poverty Reduction and Growth Trust, PRGT), which is based on an assessment of the level of income per capita, market access, and the severity of short-term vulnerabilities. The lower-income groups include 81 countries according to the World Bank’s income classification and 69 PRGT-eligible countries. The country classification is presented in Annex I. Figure 8 displays quite different patterns across income groups. The responses in LICs/lower-MICs and PRGTeligible countries (red) are considerably larger. The impact peaks at 0.10 conflict events per one million people for both low-income country groups, almost double that for the whole sample shown in Figure 6. The impact can explain 2.5 percent of the average conflict incidents in these country groups, implying that not only the number of countries experiencing conflicts is higher among LICs but also that these countries are more sensitive to economic shocks. Another notable feature is that the impact is highest in the second year after an initial shock and is persistent for more than five years, while the first-year response stays close to that of higher income groups (blue). By contrast, the responses in the higher income groups remain muted over the estimation period. In combination with the number of conflict events shown in Figure 3, this evidence implies that while several upper MICs, for example Mexico, suffer a large number of domestic violent events, the impact of economic shocks is relatively weak compared to that found in LICs. Figure 8: Responses of the Number of Conflict Events, Comparison of Income Groups (a) LICs and lower MICs (b) PRGT-eligible countries Note: Red lines show the IRs of LICs and lower MICs in panel (a) and of PRGT-eligible countries in panel (b). INTERNATIONAL MONETARY FUND 21 FCS. Figure 9 shows the impact of a commodity ToT shock for FCS and non-FCS. As a part of international efforts to support countries suffering conflict, the World Bank releases annually a list of FCS.12 Noting the complexity of fragility and conflict, it relies on two considerations for identifying FCS: (i) countries with high levels of institutional and social fragility, assessed through measures of the quality of policies and institutions, and manifestations of fragility; and (ii) countries affected by violent conflict, identified based on a threshold number of conflict-related deaths relative to the population. The World Bank’s list for 2023 contains 37 economies, most of which are LICs. We use the FCS classification as an alternative country grouping in the heterogeneity regression equation (3). As the list is available only after 2006, we use the 2006 classification for earlier years. As can be expected, our results presented in Figure 9 indicate that FCS display a significantly larger number of conflict events in response to a ToT shock. The initial impact is close to that of LICs shown in Figure 8 and there is no unwinding of the effect even 5 years after the initial shock. In what follows, we look more closely into several dimensions of country characteristics to disentangle potential transmission channels. Figure 9: Responses of the Number of Conflict Events, FCS and non-FCS Note: Red line shows the IRs of FCS. 4.4 Tracking the Transmission Channels What economic channels account for the differential impacts of the commodity ToT shocks on conflict intensity? Indeed, income levels are known to be highly correlated with various country characteristics in the development context. For instance, Blattman and Miguel (2010) emphasize that the aggregate income level serves as an indicator for both the opportunity costs of individuals to participate in conflict activities and state capacity to control conflicts, thus not identifying the channel in effect. Here we use measures of inequality and fiscal space as an indicator of country characteristics, 𝟏𝟏𝑖𝑖,𝑡𝑡,𝑗𝑗, in the regression equation (3) to examine the relevance of the opportunity-cost and state-capacity channels.13 Finding heterogenous effects across inequality levels would support the opportunity cost hypothesis, which states that individuals with limited income sources are more easily involved in violent activities. Analysis of heterogeneity across fiscal space would likewise indicate that the provision of public goods and services is constrained and governments cannot properly manage the economy during periods of adverse shocks. 12 The latest list is found here: https://www.worldbank.org/en/topic/fragilityconflictviolence/brief/harmonized-list-of-fragile-situations 13 We focus on these two channels as they imply increases of conflicts responding to an adverse income shock in line with our empirical result, while setting aside the predation channel which works in the other direction. These channels’ relevance is often highlighted by policy makers (e.g., WB and UN, 2018). INTERNATIONAL MONETARY FUND 22 Inequality. We consider three measures of inequality from the World Bank database: (i) Gini coefficient, (ii) unemployment rate, and (iii) the share of the population exposed to food insecurity. 14 Note that while the unemployment rate is not a direct measure of inequality, a higher unemployment rate implies that more unemployed individuals are vulnerable to recruitment into violent forces as their alternative options are likely to be less profitable than those of employed individuals. We identify countries with higher inequality by splitting the sample along the top 25th percentile of each measure and compare the IRs of the two groups. Each variable is lagged by one period from the shock to isolate its impact. The left side of Figure 10 displays the result. For ease of presentation, the figure shows the peak response of the higher inequality group (red) and that of the other group (blue). The estimated responses are considerably larger in countries that are in the highest quantile in each of three inequality measures, meaning that countries with higher inequality experience more severe violence in response to a deterioration of their ToT. Inequality increases the likelihood of conflicts in a range of 0.10 to 0.15 conflict events per one million people; these effects are even stronger than those for the LICs in the sample. By contrast, the estimates for countries with low levels of inequality are close to zero. These results are consistent with the opportunity cost hypothesis presented by the literature. They are also consistent with the hypothesis of the economics of crime (Becker, 1968) that high inequality increases the benefit of committing a crime or engaging in violent activities. Figure 10: Heterogenous Responses in Inequality and Fiscal Capacities Notes: The responses of the number of conflict events per one million population 1 year after a negative commodity ToT shock. Red markers show the response of country groups with higher inequality or lower fiscal space. Fiscal Capacity. Fiscal policies can mitigate macroeconomic shocks. To assess the effect of fiscal policies as a tool for mitigating the economic fallout from ToT shocks, we identify countries with limited fiscal space using gross external debt and the fiscal balance in percent of GDP from World Bank data and the July 2022 vintage of the IMF’s World Economic Outlook Database, respectively. We use the third quartile cutoff (top 25th 14 The indicator is defined as the percentage of people in the population who live in households classified as moderately or severely food insecure. A household is classified as moderately or severely food insecure when at least one adult in the household has reported to have been exposed, at times during the year, to low quality diets and might have been forced to also reduce the quantity of food they would normally eat because of a lack of money or other resources. INTERNATIONAL MONETARY FUND 23 percentile) to split the sample and compare the IRs in the two country groups, as we do for the inequality measures described above. The hypothesis is that countries with a higher level of external debt or with a higher fiscal deficit are more likely to face fiscal constraints when hit by a negative ToT shock, thereby limiting the government’s ability to organize an effective response to support the economy. The right-hand side of Figure 10 indicates that the estimated impact tends to be larger in countries with relatively low fiscal space. The estimated coefficient is substantially larger for counties with higher external debt, even if the standard error is somewhat large. The result could reflect the fact that most LICs and lower MICs rely on external financial support to implement their budget. In these cases, an elevated level of external debt can become a binding constraint for fiscal policies. Overall, the analysis in this section suggests the presence of both opportunity-cost and state-capacity channels in determining the impact of commodity ToT shocks on conflict. These channels have often been discussed in the literature. Our innovation is that we separately test each of these hypotheses using proxy indicators and find the expected results. This said, we note that these proxies may also capture deep-rooted structural issues. For instance, limited fiscal capacities may be a consequence of intensive conflicts in past years, which may also account for the large standard error in the regression analysis. We conduct further robustness checks in Section V. 4.5 Testing for Spillovers Across Countries Consequences of conflicts can spread to neighboring countries. For instance, Buhaug and Gleditsch (2008) document that the eruption of civil war in the Democratic Republic of Congo (then Zaire) was influenced by events in neighboring countries. With such spillover of conflicts, the overall impact of a macroeconomic shock could be larger than that captured in a single country. We examine this hypothesis by constructing a distanceweighted index of other countries’ commodity ToT: �𝜅𝜅𝑖𝑖′Δ𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡−𝑚𝑚,𝑖𝑖′ 𝑖𝑖′≠𝑖𝑖 where 𝜅𝜅𝑖𝑖′ is the distance between country 𝑖𝑖 and country 𝑖𝑖′, and Δ𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡−𝑚𝑚,𝑖𝑖′ is the commodity ToT change in country 𝑖𝑖′ at time 𝑡𝑡 − 𝑚𝑚. We use the distances from the country of interest to all other country pairs to construct the weights. The distance data are from United States International Trade Commission. We add the one- to two-year lagged terms, i.e., 𝑚𝑚 = 1, 2, to the baseline regression equation (1). The results are presented in Figure 11. The IRs become significant three years after the occurrence of the initial ToT shocks, implying that the spillover effect occurs with some time lag. The literature highlights various potential mechanisms that could drive the spillover of conflicts. First, there could be a direct expansion of violent activities or advancement of armed forces into neighboring countries. Another possible explanation is that information on a country’s conflict can trigger conflict events in other countries by influencing people’s beliefs. Furthermore, refugee flows can weigh on budgets (e.g., through the provision of shelter or healthcare) and weaken their fiscal capacity in host countries. In LICs and MICs, which host 74 percent of refugees, 15 pressure on already stretched government budgets can lead to a deterioration in the provision of public services essential for human capital development, unless humanitarian and development support is scaled up to both meet these needs (Rother et al., 2016; Bohmelt, Bove and Gleditsch, 2018) and enable policies that 15 https://www.unhcr.org/refugee-statistics/ INTERNATIONAL MONETARY FUND 24 raise the economic welfare of the local and displaced communities (IMF, 2022b). In these contexts, support for state capacity is critical (e.g., Salehyan and Gleditsch, 2006). Finally, ToT shocks can generate negative economic spillovers that potentially lead to conflicts in neighboring countries as well. Although our empirical results are agnostic with regard to the precise channels at work, the presence of spillovers implies that economic shocks can have an impact on conflicts that extends beyond the borders of the countries that were hit at first.