Commodity Price Shocks and Civil Conflict

Background to the Paper

First paper was descriptive, the second gave you a causal link. 

While the larger picture is true, if the country is doing poor economically, the chances that it will move to civil is higher. There are other effects of economic shocks which you miss when inspecting the large relation. This paper is a top 5 journal in review of economic studies.

Dube is in Chicago University (NYU at the time), Vargas is in Colombia at a university somewhere there.

While we expect this negative relation at the macro level, if you look at the micro level there are many other things going on. In particular, the relationship between income and conflict to have two main components.

"Commodity Price Shocks and Civil Conflict: Evidence from Colombia" - O. Dube and J. Vargas (2013)

• Paper focuses on the channels through which income shocks perpetuate armed conflict.

• Relationship between income and conflict is theoretically ambiguous.

1. Opportunity Cost Effect: higher wages may lower conflict by reducing labour supplied to appropriation, and lower wages does the contrary. So if wages improve, you decrease the probability of conflict, and when you aggregate this up this is what gives you the relationship that economic growth gives you lower levels of conflict.

• Consistent with evidence that growth reduces the risk of civil war:

  • Collier and Hoeffler (1998, 2004)

  • Fearon and Laitin (2003)

  • Miguel, Satyanath, Sergenti (2004)

2. Rapacity Effect: rise in contestable income may increase violence by raising the gains from appropriation. Think of conflict as something being produced by a firm, needing finances, labour, etc. If a country has natural resources (valuable), and are capture, then that behaves as your means of funding. Discovery of oil like in the last paper will increase the value of R, spurring the groups to try and capture the oil so that they can fund their operations. Something that is increasing economic growth can actually end up fuelling conflict, contradicting the opportunity cost of conflict channel, making the effect of economic growth theoretically ambiguous. 

Supported by:

• Hirshleifer (1991)

• Grossman (1999)

• Fearon (2005)

• This paper demonstrates that some types of income shocks increase conflict, while others reduce it.

• The basic idea is the focus on exogenous (price) income shocks induced by movements in world commodity prices, and estimate how these shocks have affected civil war dynamics in Colombia.

• Some prices shocks increase conflict, some reduce so there is heterogeneity. 

• In economics, there is an emphasis of looking at within countries as when you are making comparisons across countries there are too many things which can effect your causal interpretation. Always good to go more micro.

• There are multiple instrumental variables used in a panel data analysis.

Institutional Context: Conflict in Colombia

The Colombian civil war started during the launch of a communist insurgency in the 1960s. During the period of analysis, it involved three sets of actors:

• left wing guerrillas

• government - supposed to be neutral but sometimes supported the right wing paramilitaries

• right wing paramilitaries 

Guerrillas (FARC and ELN) fight today with the stated aim of overthrowing the government and claim to represent the rural poor by supporting aims such as land redistribution.

The conflict remained low intensity throughout the 1980s, when it effectively served as a Cold War proxy, but escalated sharply after the 1990s.

Paramilitarism had its roots in anti-insurgent self-defence groups that were organised in the late 1980s by rural landowners and drug barons in response to guerrilla extortion. However, these groups did not emerge as an organised third force with strong presence until the mid 1990s (AUC).

The conflict was 3-sided during the 1990s with all groups fighting one another, although there is extensive evidence of paramilitaries teaming up with the government military against the guerillas.

Both the paramilitaries and guerrillas are financed by the cocaine trade, as well as kidnapping, extortion, and predation on public funds. Qualitative evidence that predation is particularly high in regions with natural resources.

These groups also appropriate government revenue, as audits show that oil and gas royalties are often missing from municipal coffers where the paramilitaries exert influence. - (Human Rights Watch, 2005). 

One journalistic account: "Colombias main oil pipeline... has so many holes in it that it is known as the flute." - (McDermott, 2004). This brings in the finances

Both the guerrillas and the paramilitaries recruit from the ranks of rural workers. Although FARC typically does not pay regular salaries, some former combatants reported receiving occasional payments. This brings in the labour

Mechanisms: Commodity Prices & Civil War

The "opportunity cost - rapacity" framework generates two sets of testable predictions.

1. First, a rise in the price of agricultural goods:

• Such as coffee should increase work hours in the productive sector and increase wages relative to contestable municipal revenue

• Thus reducing conflict differentially in regions that produce these goods more intensively. 

2. Second, an increase in the price of natural resources:

• Such as oil should increase municipal resource revenue

• But not result in offsetting wage increases

• Thus increasing conflict differentially in the natural resource region.

Data on Conflict 

Conflict Analysis Resource Center (CERAC) dataset: event-based and includes more than 21,000 war-related episodes in over 950 Colombian municipalities from 1988 to 2005. 

Based on media reports in 25 major newspapers, and supplemented by reports from Catholic priests who describe cases of political violence in nearly every municipality  in the country (even in remote regions). The priests are regarded as neutral actors in the conflict, and often used as negotiators between the two sides. 

The events are also cross-checked against several other official sources, including a dataset by the National Police and reports by Human Rights Watch and Amnesty International.

News from government may have biases, which news sources may not. That is why many research papers on conflict do not use government sources as main source, it is simply used to validate the findings.

Distinguish between clashes (both parties are fighting) and attacks (one sided onto the other) (massacres and non-massacres). 

Data on Commodities

Agricultural commodities: hectares of land used for cultivating that crop in a given year. For example, a coffee census records the amount of land used for growing coffee in each municipality (950) in 1997. This is done in the middle of the period which is a limitation. 

The main oil measure is the average barrels of crude oil produced per day in each municipality in 1988, the beginning of the sample period. This was done in the beginning of the period so less problematic, this is fixed, given, and not affected by conflict; whereas the coffee production could have been affected by conflict.

Coca cultivation (used for making cocaine) data sources: Direccin Nacional de Estupefanientes (DNE) has a measure of land used for coca cultivation in each municipality in 1994; an equivalent measure is available over 1999-2005 from the United Nations Office for Drug Control (UNODC).

Data on Prices of Coffee & Oil

Data on coffee prices comes from the National Federation of Coffee Growers (NFCG), the quasi-governmental institution which oversees the taxation of coffee exports and sets the international price of coffee paid to growers.

Should we use the national price vs International price? Maybe national because that is what determines profitability in growing coffee in the domestic market and that is what would have an effect on the wages of workers. Yes, but the problem is that the national prices may be affected by ongoing conflict (we always worry about feedback effects), so their strategy is:

Instrument the international price of commodities with the export volume of other nations if Colombia ranks among the major (top ten) exporters over this period. If they are not there then use international price as a proxy. Uses data on the export volume of the three other leading coffee exporting nations from the International Coffee Organisation (ICO) since Colombia is the second largest coffee exporter over 1988-2005. Cannot simply use International price as Colombia themselves is one of the 'big players in the international market.

Uses the international price of crude oil directly (from the International Financial Statistics) as Colombia falls below the top 20 exporters.

Estimation Strategy 

The point that they are trying to make is:

• In municipalities which grow a lot of coffee

• If there is a positive increase in the price of coffee

• That should reduce conflict as coffee becomes more profitable

• Hence you hire more workers at an acceptable wage

• These workers work on the coffee plantation

• Hence not recruited by rebel groups

• Therefore conflict is lower. 

At the municipality level you need information on the price, how much coffee they produce. The point is that the price (potentially) changes from year to year but the amount of coffee grown. You have some data on land used for coffee growing but that comes from just one census in 1997, in the middle the period. In terms of oil production per municipality, you have that data from 1988 making it a good proxy as it not affected by conflict (exogenous). 

Time variation stems from movements in annual prices. There is no variation in the quantity of production, that is fixed for a municipality for the entire period and varies across municipalities.

Prices of oil (international): exogenous to Colombia's production as the country does not rank among the leading oil exporters.

Price of coffee (international): endogenous to Colombia's production as is among the ten largest coffee exporters.

Hence, you instrument the internal coffee price by Colombian producers with the coffee export volume of the three other leading coffee exporting nations: Brazil, Vietnam, and Indonesia.

Cross-sectional variation based on commodity distribution across municipalities. The main oil variable is production from 1988 as there is no feedback from that

Coffee production intensity is measured in 1997, the middle of the sample period as that may be contaminated by the effects of conflict prior to that. For coffee production you require a certain environment. This predicts what would be the optimal amount of coffee production based on the features of the environment in each municipality. 

Hence, instrument coffee production intensity (amount of coffee grown in a municipality) with this with rainfall and temperature factors, which captures the latent coffee production capability of a municipality. This tries to limit any potential feedback loops.