Cue-Taking in Legislative Voting

The scaling of legislative roll call votes has a long and rich history in political science. However, until recently, the most commonly used models use only very limited information (the votes themselves) when attempting to recover the underlying latent dimensions. Recent developments have added some "strategic" richness by including additional covariates or a modified data generating process. I extend these results further by adding "behavioural" richness. I construct the first model of ideal points to incorporate "cue-taking", i.e. the phenomenon where MPs look to their colleagues to decide how to vote on particular bills. I show how different models of cue-taking (e.g. looking to committee experts or party leaders) can capture different intuitions or stylised facts about legislative politics. Methodologically, I show that these models can be easily and rapidly implemented via a novel application of an Expectation-Maximization (EM) algorithm. I apply these cue-taking models to two highly distinct legislatures: the US House of Representatives (1946-2014) and the Canadian House of Representatives (2004-2016). Despite the important differences between these two bodies, an analysis that incorporates cue-taking reveals new insights into individual and party legislative behaviour. This suggests that roll call votes may be informative about legislative behaviour in parliamentary systems if richer models for scaling votes are employed.