Measuring the Public's Ideological Preferences in the 50 States: Survey Responses Versus Roll Call Data

Abstract

for some time now, our team has reported states' ideological preferences as the mean ideological selfidentification of respondents in CBS News/New York Times polls (Wright, Erikson, and Mclver 1985; Erikson, Wright, and Mclver 1993). l Recently, Brace et al. (2002, 2004) have reported measuring state ideological self-identification using data from the American National Election Studies (NES) and the General Social Survey (GSS) surveys.2 Using these two independent datasets, both teams of researchers have concluded that aggregated state-level ideological preferences are overwhelmingly stable over time, at least in recent decades. Theory does not demand that be constant. In fact, we know that ideology's sister variable, state-level party identification, does vary over time both in absolute and relative terms (Erikson, Wright, and Mclver 2006).3 But there can be little doubt that when measuring state ideological preferences as the mean self-report of preferences, the absolute and relative positions of the states have been very stable from year to year, going back at least to the mid-1970s. Contrary to this overwhelming evidence from surveys, Berry et al. (2007) insist that state-level ideological preferences are actually fluid, with states frequently shifting back and forth between liberal and conservative. Their argument relies on their index of citizen ideology (Berry et al. 1998), which they find superior to measuring ideological preferences directly from public opinion polls. Unlike ideological identification in polls, their scores change over time. This distinction is crucial for them because they argue that state could causally influence other state-level variables only by changing from year to year and that it could not do so if it is static. Berry et al. imagine that, for example, residents of a state might hold