In the study of corporate political activity in the United States, scholars have consistently relied on samples comprised entirely or principally of large firms. While scholars have raised the issue of bias in these samples, there have been no systematic examinations of the consequences for causal inference. We address this issue directly by comparing the results of comprehensive models that examine corporate lobbying using both large-firm and randomly-generated samples. We find that while there are some notable differences, they are certainly not so large as to lead us to question fundamentally the results of decades of scholarship. In short, the results generated using a random sample lead to causal inferences largely consistent with those in the theoretical and empirical literature. In particular, firms' resources and interactions with government condition both their decisions to lobby and the level of their activity.
