In the November 2008 issue of this Quarterly, we published a series of articles dealing with a central problem of legislative research: how best to measure the policy preferences of individual legislators and of legislative parties. Two additional articles in this series appeared in the February 2009 issue and two final articles appear in this issue. It is important to recognize that while identifying legislators’ “ideal points” may appear to be a measurement problem, in fact it raises methodological issues in the broad epistemological sense of that term, issues of the relationship between measures and the concepts of theoretical interest. To understand what a legislature does (and why it does it) we need to know the policy preferences of its members. Is the most widely used measure of those preferences—their roll-call vote—a valid indicator of those preferences? Previous articles in this series have grappled with that question. The two concluding pieces in this issue do so again, in this case comparing two of the most widely used statistical models for identifying legislators’ locations in an abstract policy space from their voting record.
The importance of the issue of measuring legislators’ positions is well illustrated by the lead article in this issue. Eric Schickler and Kathryn Pearson reassess the party cartel theory, which asserts that the majority party controls the agenda of the U.S. House of Representatives through its ability to control the decisions of the House Rules Committee. Schickler and Pearson challenge this conclusion. They do so by questioning the use of roll-call votes as evidence of the Rules Committee’s influence on the floor of the House. For that measure they substitute data on the content of bills, transcripts of committee hearings, members’ statements, and newspaper coverage. Using data from 1937 to 1952, they assess the extent to which the Rules Committee reported bills to the floor that challenged the policies of the majority party. During those years, the Democrats had a majority in Congress but a conservative coalition appeared to have great influence. The authors show that the Rules Committee frequently pushed conservative measures to the floor that the Democratic administration opposed. Schickler and Pearson use this example to demonstrate the inadequacies of roll-call data, noting that these data are not valid measures of member’s preferences at many stages of the legislative process. They also remind us that roll-call data are sparse before a rules change in 1970 that permitted them to be used in the Committee of the Whole. The sparse record, and the evidence that the leaders determine when roll calls occur, shows that reliance on roll calls skews the evidence of what we really want to know.
Ever since Richard Fenno introduced the concept of the “home style” of members of Congress, that concept has been used to analyze the constituency service of legislators, not only in the United States. David C.W. Parker and Craig Goodman examine the effect of the resources that members employ to develop a positive impression among their constituents. The authors use survey data from the American National Election Studies in the 1990s, instead of relying on election results, as indicators of constituents’ satisfaction. They find that the more money members of Congress spend on travel and on mail, the more citizens regard them as serving the constituency. The more bills members sponsor, the more constituents regarded them as policy experts. All told, the “representational allowances” that members spend in their district have a small, though possibly meaningful, influence on constituents’ perceptions of them. In that sense, incumbents who have such allowances have an advantage over challengers, who do not. But expenditure does not influence that aspect of a member’s home style that Fenno regarded as most important—the ability to communicate to constituents that the member is “one of us.”
One of the most controversial provisions of the Voting Rights Act renewed in 2006 is its provision authorizing race-conscious legislative districts to assure that minority groups—notably African Americans and Latinos—are able to win constituencies. In interpreting this law, courts have applied stringent standards for allowing—and disallowing—racial districting. The courts take into account social science research on the probability of electing minority candidates from non-minority districts. David Lublin, Thomas L. Brunell, Bernard Grofman, and Lisa Handley provide evidence that racial districting continues to be important for the election of minority candidates, even though some minority candidates have in recent years won in majority-white districts. The article is designed to provide evidence that without districts in which minorities have a majority of voters, the election of minority candidates at both the state and national levels will be very rare. This is equally true for both African American and Latino candidates.
The concluding articles in this issue examine NOMINATE, which has been the dominant spatial model of legislators’ preferences obtained from roll-call data, and compare it to newer models. The scores derived from this model have been the standard measure used in roll-call research on the U.S. Congress for a generation, ever since Poole and Rosenthal introduced it 25 years ago. In the 1990s, scholars using roll-call data developed other estimators of legislators’ ideal points, among which IDEAL is prominent. Each uses different behavioral models and different mathematical routines to derive statistical estimators of legislators’ positions in an abstract policy space. In the article coauthored by Royce Carroll, Jeffrey B. Lewis, James Lo, Keith Poole, and Howard Rosenthal, the authors identify the differences between these models. They then compare the results of the models as applied to 213 votes in the U.S. Supreme Court and 520 votes in the U.S. Senate. They use Monte Carlo experiments to establish all the plausible circumstances under which the different models would produce different estimates of members’ ideal points. The authors conclude that the differences between the results produced by these two models—and others—rest on differences in some of their underlying behavioral and mathematical assumptions, and that neither is always superior for analyzing a particular legislature. Their analysis, however, reveals the importance of understanding the assumptions underlying a particular spatial model for interpreting its results.
The final article in this issue, and the final article in the series on identifying legislators’ policy preferences, is contributed by Joshua D. Clinton and Simon Jackman, two of the three authors who developed the statistical model that produces the IDEAL scores of legislator preferences. They agree that neither their model nor NOMINATE has clear advantages over the other. They find it heartening that two models with substantially different assumptions yield such similar results in the analysis of roll-call votes. However, they point out that the Bayesian approach underlying their model makes it applicable to a wider range of data and therefore has a flexibility that makes it more applicable to a series of research questions that go beyond those that rely on roll-call data. These final two articles in our series on identifying the policy preferences of legislators reflect the huge edifice of roll-call research that has facilitated research on the U.S. Congress in the last generation, and that has encouraged thinking about alternatives.
Legislative Research Center
Return to November 2009 Titles
Return to LSQ home page