Abstract

A vast literature documents growing ideological divisions between the parties in the contemporary U.S. Congress based on estimates from roll-call voting behavior (such as DW-NOMINATE). We revisit theoretical and empirical claims about the nature of partisan polarization by addressing concerns raised in recent scholarship about the comparability and interpretation of roll-call estimates over time. We leverage data from candidate surveys that allow us to hold the policy agenda constant from 1996 to 2008. We show that the replacement of relatively moderate legislators with more ideologically extreme legislators, particularly among Republicans, explains virtually all of the recent growth in partisan polarization. We further demonstrate that these patterns are explained mostly by increased polarization over social and environmental issues and link our findings to changes in the congressional agenda. Our results have important substantive and methodological implications for evaluating sources of legislative polarization and using roll-call measures in empirical applications.