Holy roll calls: religious tradition and voting behavior in the U.S. House *.

Introduction

As research into mass political behavior has demonstrated, religious affiliations and beliefs affect a wide variety of political attitudes and activities (see Jelen, 1998, for a review). Yet, despite evidence that religion affects decisions at the mass level, research into congressional voting decisions has, for the most part, ignored the possible effects of religion on vote choice.

This omission is surprising given that there is evidence that religious beliefs and roll call voting are related. Though there has been very little research into religion's effect on congressional voting, there are some exceptions that demonstrate a relationship between the two. These studies either focus on a single year or two, making generalizations difficult, or worse, they employ invalid measures of religion, measures that hide the full impact of religious tradition. Even with these flaws, much of the early literature on Congress did find some relationship between religious affiliation and voting, despite the use of simplistic measures of religion such as Protestant and Catholic (Fenton, 1960; Rieselbach, 1966).

Since then, researchers have shown that religion, when measured more carefully, affects voting. For example, using in-depth interviews with eighty members during the 96th Congress, Benson and Williams (1982) develop a rich religious classification of members of Congress. Because Benson and Williams (1982) have very detailed information on the beliefs and religious orientations of members, they are able to study the relationship between beliefs and votes in ways that are nearly impossible for other researchers. (1) Yet, their analysis of this relationship relies on simple correlations, with no control even for party. Moreover, because their analysis is based on interviews, the number of members included in the study is limited.

This same limitation is faced by Page et al. (1984), who examine the effect of constituency policy preferences on roll call behavior. Using the 1978 National Election Study (NES), Page et al. test for the effect of constituency opinion on members' voting behavior in 108 congressional districts (25 percent of the House). Their models include a simple measure of the member's religious affiliation (Catholic or non-Catholic), which does not withstand controls for district issue positions, though it does have an effect on the abortion issue, a finding mirrored in research since (Cresanthis, Gilbert, and Grimes, 1991; Tatalovich and Schier, 1993).

In a more recent study, Green and Guth (1991) demonstrate that members' votes represent both their own religious affiliation and the religious groups within their districts. Using a finer measure of religious affiliation, Green and Guth demonstrate that religion, either because of personal beliefs or representational concerns, influences voting behavior.

We build off of these studies in three ways. First and most important, we are careful to use the most comprehensive measure of religion available, one that uses the richness of data on denominational affiliation available for each member. Second, we examine voting over a long time span, 1959-94, rather than during a single congress. (2) Finally, we demonstrate religion's impact both on a specific issue and on broad voting patterns. In doing so we hope to add to the literature on religion in Congress (see Fowler and Hertzke, 1995:123-28; Wald, 1997:146-49, for literature reviews) and to the literature on congressional behavior (see Davidson and Oleszek, 1996:261-66, for a review). To date, there is insufficient longitudinal evidence of religion's impact using theoretically valid measures of religion. (3)

Our purpose in this project then is twofold: we hope to add to the theoretical justification for studying religion's impact and we hope to convince other congressional scholars of the practical importance of including religion for complete model specification. To this end, we employ a new data set with comprehensive religious affiliation information, demographic variables, and voting behavior measures of members of Congress from 1959 to 1994. We also present an empirical analysis that supports the hypothesis that religion has an effect independent of party, other demographic characteristics, and district partisanship. We find that religion does affect roll call voting, in part mirroring trends demonstrated in the masses.

Religious Traditions and Voting

Clausen (1973) aptly depicts members of Congress as typical of other politically active citizens. Members think the same way other people think. They decide on votes the same way. Research has also shown that in addition to the influences of party, president, and constituents, congressional vote choice is affected by the members' gender (Berkman and O'Connor, 1993; Thomas, 1991; Vega and Firestone, 1995; Welch, 1985), and race and ethnicity, though the results of these last studies are decidedly mixed (Hero and Tolbert, 1995; Swain, 1993; Welch and Hibbing, 1984). Since religious characteristics, like demographics, affect mass political behavior, some may consider findings on religion's impact on members' votes as clear and obvious, yet religion has regularly been neglected, as were demographics until recently.

Religion and Voting over Time. While we expect that a relationship between religion and congressional voting exists, measuring it over time is problematic. We would prefer to know the precise beliefs of each member, but surveying members for specific beliefs is difficult to do for one year and virtually impossible to do over a long period of time. Even if such interviews could be done, linking specific beliefs to specific votes would be like trying to find the proverbial needle in a legislative haystack. Connecting specific beliefs to particular policy positions is fruitful for a few issues and doctrines (e.g., Guth et al., 1995, on the environment), but it does not provide the best model for studying the connection between religion and particular policy stances, since it requires that the researcher know a priori which religious attitudes correspond to which political attitudes.

Because of the complexity of religious beliefs and practices, the best way to study religion in Congress is to use the denominational affiliation that virtually every member has made public since the 1950s. There have been a handful of researchers who have used similar data in predicting specific votes, but such analysis often categorizes religion into two categories--Catholic and not-Catholic (e.g., Page et al., 1984; Tatalovich and Schier, 1993). (4) Such simple classification schemes ignore the variety of denominations …