Call Us: US - +1 845 478 5244 | UK - +44 20 7193 7850 | AUS - +61 2 8005 4826

political opinion data for greater Khartoum

The second set of robustness checks shown in panel B indicates that our results are not driven by particular modeling decisions. In column (4), we show that our results do not depend on efficiency gains associated with our stratified sampling design. Column (5) addresses the concern that we have ignored those who ‘‘don’t know’’ about their support for unity or separation by coding such responses as an intermediate value and estimating an ordered probit model. Again, the result persists. Finally, column (6) responds to the worry that the listwise deletion of observations for which any data is missing could introduce bias in our estimates. We multiply impute missing values on any of the control variables, perform estimation on a now larger set of observations, and obtain an estimate of essentially the same size and with about the same standard error.