Climate, Conflict and Labor Markets

Abstract:

This research aims at linking two strands of the literature: the recent works on climate and conflict (Burke, Hsiang and Miguel, 2015) and the economics of labor coercion (Acemoglu & Wolitzky, 2011, Dippel, Greif and Trefler, 2015). In the first area of knowledge, my research addresses an important knowledge gap on the role that institutional quality plays to help explain the directions and magnitudes of the impact of weather fluctuations on conflict through their effect on economic outcomes.  Regarding the second knowledge area, my research contributes to the creation of new knowledge by bringing rich individual data and the use of satellite-generated information to the analysis of coerced labor. Also, by analyzing the current phenomenon of coca planting and exploitation by non-State armed actors, my research can inform illegal drug policy as well as rural development policies.

 

More precisely, this research aims at understanding the causal effect of weather-induced agricultural shocks on confrontations and forced displacement in the context of the Colombian civil conflict. I first match monthly municipal rainfall to coca leaf yield, to show that rainfall is a strong predictor of coca yield. Standard economic theory predicts that the rising productivity should translate into larger wages. However, when I estimate the effect of the positive productivity shock on rural unemployment and wages in coca areas, I find that the increasing production due to good weather rises labor demand but labor income remains constant.

 

To estimate the effect of the positive productivity shock to coca-suitable areas on labor demand and wages, I estimate OLS equations of (log) wages and the probability of unemployment on individual characteristics that determine labor market outcomes and rainfall, controlling for municipality fixed-effects that absorb time-invariant characteristics, whether observed or unobserved, disentangling the precipitation shock from other sources of omitted variable bias, month*year fixed effects during the period January 2004-June 2010 and a department-specific linear time trend. Coca-suitable municipalities are those where coca has been cultivated in the municipality during at least one year during the period 1999-2010. These estimates are explicitly reduced form, and they focus on the effect of the variation in precipitation in coca-suitable areas on labor market outcomes. Given that precipitation varies plausibly randomly over time, as random draws from the municipality climate distribution, this approach has strong identification properties (Dell, Jones and Olken, 2014).

 

To explain the unexpected result of unchanged rural wages in response to increasing productivity, I draw on labor coercion models, where the role of coercion is to decrease the outside option of the coca farmers (Acemoglu & Wolitzky, 2011, Dippel, Greif and Trefler, 2015). These models hypothesize that an increase in coca productivity should be associated with expansion efforts by the coercive non-State armed groups and a decrease in forced displacement. Forced displacement occurs when farmers are able to leave the coca farming contract. Because the coerced sector yields higher returns with better rainfall, I first test that excess rainfall causes differentially lower forced displacement in coca suitable areas than in non-suitable areas. The results confirm, that in fact, an additional milimeter of precipitation above the municipality mean decreases forced displacement by 1.22% in coca-suitable areas; by contrast in non-coca areas the effect of positive rainfall shocks is approximately ten times smaller and insignificant.

 

A second way to test for coercion in coca suitable labor markets is to test whether high productivity months witness lower forced displacement. I therefore match coca leaf yield data with local violence in a small sample of municipalities with rich harvest data. I employ an Instrumental Variables estimation where changes in forced displacement are explained by coca leaf yield variations, and I instrument changes in coca leaf yield with rainfall. One additional arroba of coca leaf per hectare harvested is associated with a reduction of forced displacement of 4 people on average, and this estimate is significant at the 5 percent level.

 

Finally, I regress coercion efforts measured by confrontations between non-State actors who profit with cocaine and government forces as well as mayor killings on coca yield, instrumented with rainfall. As suggested by the labor coercion models, the IV estimates of the effect of positive productivity shocks on violence in coca areas are statistically significant and economically important, suggesting that the effect of weather shocks on labor markets is mediated by the quality of the institutions where these markets are embedded.

Last updated on 04/21/2015