A spatial model of internal displacement and forced migration

Jon Echevarria-Coco, and Javier Gardeazabal

Journal of Conflict Resolution, Volume 65, Issue 2-3 (2021), pages 591-618



The authors develop a spatial model of internal and external forced migration due to conflict. Conflict is modeled as a shock in a particular location that causes the displacement of people—some of whom become refugees and others who become Internally Displaced Persons (IDPs). The model predicts how the total number of displaced people, number of refugees and number of IDPs, as a proportion of a country’s population, varies with the intensity of armed conflict as well as three spatial factors: country area, terrain ruggedness (orography) and distance to neighboring countries. For a conflict of a given magnitude, the model predicts that:

  • Displacement as a fraction of the total population tends to be larger in countries with a smaller geographical area.
  • Displaced people will travel shorter distances in countries with a rougher orography, potentially generating fewer displaced people overall and fewer refugees.
  • The more distant neighboring countries are, the lower the number of refugees, while total displacement should be independent of the distance.

The predictions of the model are tested against real data using a panel of 161 countries covering the period 1995-2016. Data includes the number of refugees and IDPs by country of origin (from UNHCR), total population (World Bank), conflict measures (UCDP-PRIO conflict database) and geographical measures (taken from Nunn and Puga (2012) and CEPII). Additional control variables include measures of civil liberties and political rights (Freedom House), GDP per capita (World Bank), colonial relationships, and the number of countries with which a source country shares an ethno-linguistic group in common with at least nine percent of the population. The authors also conduct robustness checks that control for conflict duration and type (extracted from UCDP-PRIO), the percentage of land area that is desert (from Nunn and Puga (2012)), the percentage of the land area within 100 km of an ice-free coast (also from Nunn and Puga (2012)), and whether the country is an island or part thereof.

The empirical evidence is mostly in line with the predictions of the model. Specifically:

  • Both total displacement and the number of refugees relate positively to conflict intensity and negatively to land area and ruggedness, as predicted by the model.
  • While the number of refugees is related positively to proximity, in line with the model’s prediction, total displacement also responds positively to proximity, contrary to the nil effect predicted by the model.
  • The partial effect of conflict intensity on IDPs is decreasing in proximity and significant for most of the range of proximity values in the sample. The partial effects of area and ruggedness on IDPs are increasing functions of proximity and significant for most of the range of proximity values in the sample. The partial effect of proximity on IDPs is an increasing function of ruggedness and is significant only for high values of ruggedness.

Overall, the evidence is in favor of the predictions of the model, with the only exception of the partial effect of proximity on total displacement, which turns out to be positive while the model predicts it should be nil.


Big Data