Can We Learn the Causes of Congressional Decisions from Roll-Call Data?

Scholars frequently use individual-level roll-call data to ascertain the causes of congressional behavior. However, the explanation of policy position variation across members does not explain aggregate policy decisions by Congress. Moreover, while these data are appropriate to explain the causes of policy-position variation across members, the attempt to econometrically isolate members' ideology from other causes has not been successful because important causes of decisions have not been included in the regression equations. Finally, roll-call votes are not a random sample of congressional decisions. Inferences made from roll calls have three sources of error. First, the processes that determine committee and voice-vote decisions are different from those that determine roll-call decisions. Second, the estimates of factor effects on decisions generated from roll-call data will be much too large, because they do not take account of the probability that a proposal receives committee approval and is subject to a roll call. Finally, the unknown causes of roll-call occurrence are correlated with the unknown causes of the actual