Exiting the conflict trap

A dove holding an olive branch is trapped inside three loops of barbed wire

There is a broad consensus on the adverse effects of internal armed conflict on economic development. However, the size of these effects is larger when measured using cross-country comparisons than when exploiting within-country variations.

To bridge part of this gap, Joan Margalef and Hannes Mueller propose a statistical approach in BSE Working Paper 1402, “Caught in a Trap: Simulating the Economic Consequences of Internal Armed Conflict.”

They bridge the gap by introducing the concept of the conflict trap into an otherwise standard macroeconomic growth simulation. The conflict trap is a phenomenon where countries that experience conflict tend to relapse into it generating repeated cycles of violence. In the conflict literature, it is well-established and often blamed for low economic development in the affected countries. By creating a new simulation framework that captures conflict dynamics, Margalef and Mueller estimate the long-term developmental impact of armed conflict which includes a model of the trap. They show that their approach can explain the large macroeconomic development effects noted in the cross-country literature while employing a within-country estimation approach.

The dangerous times after peace

Figure 1 below shows their estimation of the likelihood of a resurgence in internal armed conflict once a conflict has occurred. The likelihood of remaining in conflict is very high, nearly 80 percent. In the first year of peace after conflict, the likelihood of a renewed outbreak is over 20 percent. It then falls to 3 percent after 4 years and stabilizes around 2 percent. This indicates not only that conflict is self-perpetuating but also that the post-conflict period is extremely risky – around half of the countries that escape from conflict will experience a resurgence before reaching 8 years of peace. This illustrates the conflict trap. 

Given these vicious cycles of open violence and the negative relationship between conflict and economic outcomes, entering conflict can create a long-term spiral of economic damage that hinders development. To quantify the overall impact of entering conflict, the authors design and estimate a statistical model that simulates a country’s conflict path and its effect on real GDP per capita.

Figure 1. An illustration of the conflict trap

A model of the conflict trap

The authors model the dynamics of conflict and peace using a Markov Process, which includes a state of conflict, seven post-conflict peace states, and a final state of stable peace. The conflict stage and the multiple post-conflict peace stages embody the conflict trap, as they capture the higher risk of conflict relapse shown in Figure 1. A country that is in the state of conflict must successfully navigate through each of these seven post-conflict stages sequentially to reach stable peace. If a country transitions back to conflict at any point, it restarts the process.

Margalef and Mueller first estimate the transition matrix using the proportion of transitions observed in the data. Their first finding is that, given their estimated transition probabilities, it takes, on average, more than 18 years for countries to escape the conflict trap, meaning, going from conflict to stable peace. Then, the authors estimate the distribution of growth for each state using a fixed-effects regression. They find that being in a state of conflict lowers growth by 3 percentage points compared to stable peace, while the post-conflict peace states do not show a significant difference. 

Given the transition matrix and the distribution of growth for each state, they can simulate growth paths that countries experience as they move from conflict to stable peace. A country starts in the state of conflict and with a level of real GDP per capita. Then, at each period, they draw the new state from the transition matrix and impute its effect on GDP by drawing growth from the estimated distribution of the realized state. 

Furthermore, to account for countries’ characteristics that make them more prone to conflict, beyond their conflict history, the authors use a simple machine learning model to partition the data into two subsets: one composed of a less conflict-prone sample, and another composed of a more conflict-prone one. After re-estimating the model for each sample, they find substantial differences in the estimated transition matrix but similar growth distributions. This indicates that any variation in results between samples can be mainly attributed to differences in conflict dynamics, that is the severity of the conflict trap.

Measuring the cost of conflict

The authors then utilize their model by running simulations to obtain the distribution of real GDP from all simulated paths. Figure 2 shows the distribution of the loss for each sample.


They find that the average loss from the conflict trap in GDP per capita after 30 years is almost 20%. Importantly, there is a large variation across simulations, with the 75th percentile experiencing a decline of 30% while the 90th percentile declines by almost 45%. When comparing the two samples, the higher conflict tendency of the conflict sample leads to significantly greater losses; the average loss is almost double that of the peaceful one. More strikingly, the 90th percentile of the conflict sample reaches losses above 50%. This is the main finding: countries that are worst affected by the conflict trap lose half their GDP potential due to the cycles of peace and violence they are caught in. Losses in the 90th percentile in the peaceful sample are 35%, which highlights the significance of the conflict trap even here. 

There are two takeaways from this. First, addressing conflict trap dynamics seems to be crucial for the long-term development of the economy. Helping countries escape the conflict trap should be a key policy goal. Second, the prevention of conflict outbreaks will have a huge impact on economic development overall. The average effect of 20% GDP loss justifies huge investments in preventative action.