Correcting for Selection Bias in Social Movement Research

Longitudinal and cross-national research on new social movements often relies on newspaper reports for data on the frequency and type of events. Reporting by newspapers is, however, known to be strongly affected by selection bias. Newspapers report events that they find "newsworthy." Surprisingly, the effect of these selection biases on data analyses has received scant attention, and the methodology to correct for these biases hardly has been discussed. Based on data covering events in four cities in Switzerland we propose two types of corrections for the selection bias. The first is based on a simple weighting scheme for events reported in local newspapers. These weights are likely to provide useful corrections even in other contexts. The second correction attempts to model explicitly the selection bias of the media. This truncated regression approach is shown to be a useful strategy when the selectivity by the newspapers is severe, and the factors affecting this selection are largely known.