Legislative Voting in the Canadian Parliament

The paper applies a Bayesian Markov Chain Monte Carlo estimation methodology to scale roll call votes in the 35 and 38 Canadian Parliaments over a multidimensional policy space. The results clearly demonstrate that policy debates are two-dimensional in Canada. The first dimension represents the classical division between the governing and the opposition parties that has been found in similar Parliamentary systems (Hix and Noury 2007b). We also find a relevant second dimension which captures the opposition between Quebec and the Western provinces of Canada. Our results show that there is a clear separation between the Reform Party (and later the Conservatives) and the Bloc Quebecois in both Parliaments; whereas the Liberals and the NDP occupy the center on this regional division issue. We also note that the newly created Conservative Party has moved closer to the center and the governing Liberals in the 38 Parliament. This results is explained by the recent party merger between the right wing political formations in Canada. ∗Corresponding author: [email protected] 1 Overview Scaling legislative roll call votes in a multidimensional issue space is a common practice in the United States Congress (Poole and Rosenthal 1997a; Poole 2007; Poole 2005; Heckman and Snyder 1997; Clinton, Jackman and Rivers 2004). The idea behind this methodology is that legislators have underlying ideal points over a set of policy alternatives in a low dimensional Eucledian space (Hinich and Munger 1997). Lawmakers will support a new bill if the distance between their own ideal points and the status quo is greater than the distance between their own preference and the proposed policy alternative. Ideal points are usually calculated by aggregating roll call votes to create vote-based scores for each individual legislators. One of the main characteristics of the spatial analysis of roll call voting is that the distribution of ideal point between legislators generally reflects partisan affiliation within the legislature (Clinton, Jackman and Rivers 2004). Party members tend to have similar preferences; hence their votes tend to be clustered together over the course of a legislative session. The second most important finding of the spatial analysis of roll call voting is that much of the voting behavior can be explained by a stable, low-dimensional issue space, which is generally limited to no more than two dimensions. In the United States for example, the first and most significant issue dimension corresponds to the classical left-right ideological continuum; while the second dimension is related to issues of state rights and civil rights (Poole 2007). Of course the analysis of roll call votes is not well suited for every type of legislatures. In the most perfect scenario, the absence of parties, log-rolls, or other factors affecting legislative preferences could theoretically give us the precise location of a lawmaker’s ideal point on a given policy dimension (or a specific policy issue). However, most legislatures have some institutional characteristics that constrain the behavior of its members. Generally, the absence of political parties (Jenkins 1999) allows for the greatest amount of liberty. In this context, legislators have the most opportunity to vote according to their own preferences. But even under these conditions, strategic behavior or position-taking can theoretically lead a representative to vote against his or her own underlying interest. Nevertheless, in countries like the United States where party discipline is weak, we still find that given a sufficiently large number of roll call votes, the scaling of ideal point estimates can actually represent a legislator’s ideology over time (Poole 2007). In Parliamentary systems where party discipline is generally the rule, studies of roll call votes have reached a different conclusion. For instance, in their study of legislative behavior in fourteen Parliaments, Hix and Noury (2007b)