A Copula Approach to Missing Data
Missing data is a very prominent feature of conflict related data sets. However, most studies ignore this problem and simply row-wise delete observations with missing observations. In fact, this procedure is so commonly used that applied scholars are not aware that this biases their estimates and may lead to false empirical conclusions. This paper highlights that computational power and advanced statistical methods make missing data imputation feasible for most civil conflict related projects. We present three approaches to missing data imputation and compare them in regard to their trade-off in computational power needed, bias, and variance of the estimated missing data. Our results demonstrate that most applied researchers can achieve great improvements over row-wise deletion by implementing a copula based imputation approach, which is the fastest and easiest to use.