Explaining the 2004 exit poll error as a product of response bias and not fraud: A comparison of the 'spiral of silence' and 'political ambivalence' perspectives as heurism for analyzing the causal structure of differential response in the exit polls

Leaked exit poll data on the 2004 Presidential election suggested that John Kerry was ahead in several key states and was winning the election. Yet the official vote tallies showed that George W. Bush easily secured his second term. The unprecedented degree of discrepancy between the exit poll estimates and the official tallies fuelled a fierce controversy. Since no particular methodological shortcomings had been identified, some advanced the "reluctant Bush respondent" hypothesis, arguing that the exit polls overstated the Kerry vote because Bush supporters were less willing to express their views by participating in the exit polls. Since no information existed on those alleged reluctant Bush supporters who refused to participate in the exit poll, this hypothesis relied on rather fragmented, circumstantial evidence. Also, the scope of the reluctant Bush respondent hypothesis was seriously limited: it remained virtually unaccounted why Bush voters would have been reluctant to participate. Due to its circumstantial empirical standing and limited scope, the hypothesis was criticized and deemed implausible. This criticism had far-reaching implications: in lieu of hands on evidence on what made the exit polls overestimate the Kerry vote, some alleged that not the exit polls but the vote counts might have been inaccurate.This dissertation advances the thesis, that poll data cannot constitute evidence of fraud. Even though evidence of response bias may be established only in a circumstantial manner, two theoretical frameworks, the spiral of silence and political ambivalence, may explain why differential response occurs in pre-election and exit polls. The reluctant Bush respondent hypothesis is scrutinized in light of these two theoretical frameworks as well as the 2004 National Annenberg Election Survey (NAES). The findings constitute evidence that, first, relative to Kerry voters, Bush supporters were less outspoken regarding politics during the last month of the campaign. This decreased public outspokenness could have spilled over into decreased participation rate in the exit polls. Second, the findings shed light on the causal structure of the reluctant Bush respondent phenomenon: elements of the public opinion dynamics of the 2004 Presidential election campaign are found to be consistent with the spiral of silence theory.