Do, Quy-Toan and Lakshmi Iyer, "Geography, poverty and conflict in Nepal", Journal of Peace Research 47(6):735-748, November 2010.
About the data (from the data appendix of the working paper):
A. Measures of conflict:
These were computed using data obtained from the website of the Informal Sector Service Center (INSEC), a Nepalese NGO which documents human rights abuses across all the districts of Nepal. We obtained data on the number of conflict-related deaths between 13 February, 1996 and 31 December, 2004, caused by the state and by the Maoists separately. We also obtained annual data on the number of conflict-related deaths for the years 1999, 2002, 2003 and 2004 from the Human Rights Yearbook published by INSEC for various years. We use district population data from the 1991 census. The main variables are defined as follows:
Normalized total killed = Total killed between 1996 and 2004 divided by district population
Dummy for more than 100 people killed equals one if the total people killed between 1996
and 2004 is greater than 100.
Normalized total abducted = Total number of people abducted by Maoists between 1996
and 2004 divided by district population (in thousands).
Data on latitude, rainfall and maximum elevation of the district was obtained from Nepal District Profile:1994. Data on proportion of district under forest area dates from March 2001 and was obtained from Japan Forest Technology Association (JAFTA), Information System Development Project For The Management of Tropical Forest: Activity Report of Wide Area Tropical Forest Resources Survey (Kingdom of Nepal).
C. Development indicators:
Poverty figures (indicating the proportion of households in the districts below the poverty level) were obtained from the Nepal Living Standards Survey 1995-96 (NLSS-I) conducted by the World Bank. Literacy rates for 1991 were obtained from the 1991 Census. Infant mortality rates come from the Nepal District Profile:1994.
Data on roads, post offices, schools, banks and health posts come from Nepal District Profile:1994 and Nepal District Profile : 2002. They are normalized by the population of the district.
E. Caste and language diversity:
Caste polarization and fractionalization measures were computed using 2001 district level census data on population by caste (the 2001 nationwide caste proportions are very similar to the pre-conflict 1991 caste proportions; however, we do not have the latter at the district level). We retained castes which make up more than 1% of the district population, castes that make up less than 1% of the district population are classified as “others”. We have 76 distinct categories under this measure. The proportion for each category is computed by dividing total number of individuals in the district falling under that category by respective district population. Proportion of advantaged castes is computed as the proportion of district population belonging to the Brahmin, Chhetri, Thakuri or Newar castes. High caste dominance is a dummy which equals one if the proportion of advantaged castes is between 45 and 90 percent of the district population.
For measures of linguistic diversity, we use data from 2001 census on the number of people
speaking different languages. We retain languages spoken by more than 1% of the population,
and the remaining are put under ”others” category. This gives us 13 different categories.
Linguistic polarization and fractionalization measures are computed similar to the caste polarization
and fractionalization measures.