Accounting for Candidate Obfuscation and Electoral Context when Modeling Issue Voting under Uncertainty.

Spatial theories of voting argue that voters compare their policy preferences to the issue positions of the candidates and then choose the candidate who best represents their preferred policies (Enelow and Hinich 1984). Voters, however, are expected to have only imperfect information about where the candidates stand on issues. This uncertainty can have multiple sources, including candidate obfuscation (Downs 1957, Page 1978) as well as the differential costs of acquiring information (Alvarez and Franklin JOP 1994), and is often considered to be a ubiquitous part of the political environment. Uncertainty also has important consequences for how voters choose among candidates during an election, operating as a "cost" that reduces the likelihood that a voter will support a given candidate (Alvarez 1997).Our paper seeks to expand upon the empirical evidence that uncertainty operates as a cost to voters when choosing between candidates. We use an estimation strategy applied by Bartels (AJPS 1985) in which predicted probabilities of nonresponse to candidate placement questions (constructed from a first-stage analysis) serve as measures of uncertainty in a (second-stage) regression of proximity voting. We expand upon Bartels' approach in two ways. We first seek to empirically distinguish between the effects of external sources of uncertainty attributed to candidate ambiguity and the internal sources of uncertainty due to voter attention to campaigns. Examining differences in levels of uncertainty across different presidential campaigns provides leverage over the question of candidate ambiguity.The internal sources of uncertainty are assessed by the effect of information on uncertainty, which is expected to be constant across election cycles. We also consider whether levels of uncertainty due to candidate ambiguity have declined in recent years due to the resurgence of partisanship (Bartels AJPS 2000). Second, we seek to expand Bartels' work by examining whether the cost associated with uncertainty varies across elections and types of candidates. One possibility is that the presence of an incumbent president in the election increases the cost associated with uncertainty. Finally, we also examine the effect of group-specific issue-taking by candidates on the level of external uncertainty, recognizing the incentives that candidates have to attract a larger number of voters by "sprinkling" their policy positions across the issue space (Downs 1957). ..PAT.-Unpublished Manuscript [ABSTRACT FROM AUTHOR]