Many models of representation, congressional roll call voting, and congressional policy making are premised on some level of citizen knowledge of the representative's actions in Congress. This seemingly simple premise disguises a complex set of relations between constituents and legislators. There are many ways that citizens are made aware of the activities of their representatives in Congress: newspapers and mass media coverage, interest groups, interpersonal communications, direct mailings from congressional offices, and contacts from campaign and party organizations. And citizens know different things about their members, from a general, diffuse reputation for helpfulness and trustworthiness (Bianco 1996; Fenno 1978) or broad partisan and ideological position, to, at the other extreme, specific activity (casework) that members engage in to aid constituents in the district (Cain, Ferejohn, and Fiorina 1987) or specific votes that members may have cast on the floor (Kingdon, 1989). Even this list is grossly oversimplified. Casework can be used to foster generalized reputations, roll call votes can be devices to build trust for incumbents so they don't have to explain themselves next time (Bianco 1996; Kingdon 1989; Fenno 1978), and both can serve as fodder for opponents' attacks on character.
We hope to shed light on one facet of this complex set of relations between legislators and constituents: awareness by citizens of roll call votes cast by their members. Minimally, most models of representation imply at least some tie between the roll call votes and constituency opinion. This further implies some level of awareness by constituents of votes. We assess under what conditions respondents claim that they know their legislators' positions. If citizens don't know specific votes (as we know they often do not), we show how they may infer the position their representative might have taken.
Thus, while our study bears on representation, it is at its heart a study of political information, where the piece of information we are interested in concerns roll call voting. For reasons elaborated in the next section, we narrow our inquiry to particularly high profile roll call votes, hypothesizing that here, with media coverage, member concern, and public interest at their maxima, we should get some sense of the limits of public knowledge.
We approach the question of political information in three ways. First,
we describe the raw level of knowledge about roll call votes. What do
citizens know and how accurate is that knowledge? Second, we compare
``knowers'' --- those respondents who claim they remember how a member
voted --- to ``guessers'' --- those who admitted they do not remember,
but are willing to hazard a guess. At first blush, we might expect that
knowing is superior to guessing, but the accuracy levels among guessers
is surprisingly high. This leads to an obvious third query: what are
the determinants of knowing and of guessing, and how do the two
processes differ?
As in our past work, we show that both individual and
contextual characteristics account for the patterns of knowing,
guessing, and accuracy.
In this paper, we examine whether respondents know, or are willing to guess the roll call positions taken by their U.S. Senators on the the January, 1991 vote to allow the use of force in the Persian Gulf conflict and the October, 1991 vote to confirm Clarence Thomas as Justice in the Supreme Court. In the next section of the paper, we discuss the rationale behind choosing high profile roll call votes, along with a description of the two votes in this study. Next, we compare citizen knowledge of these two votes. In the final section, we explore the determinants of responses across the two votes. The results show that the proportion of the sample as well as the accuracy rates across each vote are roughly comparable. We take this as a recommendation for further replications of this line of inquiry. Second, we find that the size and statistical significance of variables in the ``knowing'' model are also stable across the two votes, but that the determinants of accuracy differ in important ways. We close the paper by speculating on the reasons for these differences, and suggest further paths for research.