An important first step was to understand the overall bivariate relationship between key political variables and dispositions. Most critical, of course, were any possible relationships between vote-share and disposition. These relationships for people saying they voted for George W. Bush are summarized in Figure 1.

Using independent samples t-tests, all comparisons within the figure are significant, with the exception of “four call attempts” and “all call attempts” to “completes from first attempt and NA.”  As shown in the bar representing “all call attempts,” the overall results of the survey found that 52 percent of respondents who said they voted said they voted for Bush. This should not be surprising, since it is well-known that more people say they voted for the winner than actually did when surveyed in postelection polls. Within a typical preelection poll—that is, by limiting the call attempts—vote-share increases by the modest amount of 2 percent. However, this is starkly contrasted by Bush’s vote-share among completed interviews within five or more call attempts, which stands at 46 percent.

Similarly, when completed interviews at four or fewer call attempts were investigated, we found a 3 percent difference from all completes, in favor of Kerry, while first-attempt completes and all “no-answer-only” completes at four or fewer calls had a 4 percent swing in the other direction, in favor of positive Bush vote-share. Overall, then, the difference in Bush vote-share between “primary” dispositions (first attempts and no answers) and “secondary” dispositions (refusal conversions, language barriers, privacy managers, and callbacks) was 7 percent.

A second exploration by party identification, had, unsurprisingly, similar results. In conjunction to vote-share for Bush versus Kerry, there was a positive relationship between the likelihood to identify as a Democrat and the number of call attempts. Also in conjunction with vote-share, identifying as a Democrat was more substantially associated with secondary dispositions than with primary dispositions.

Third, dispositions were analyzed by voter registration, with results summarized in Figure 2.

Here the results were even more divergent, with significant results found for all t-test comparisons, with the exception of “four call attempts” and “all call attempts” to “completes from first attempt, CB, and NA.” Overall, 81 percent of respondents claimed they were registered—a result that again falls in line with the inflation caused by social desirability in most postelection polls. The difference between call attempts, split after four, was substantive, with more respondents who were completed with fewer than five attempts saying they were registered, as compared to respondents who completed at five calls or more, for an overall difference of 9 percent.

Similarly, respondents who completed at the first call attempt or at fewer than five attempts and with only no-answer dispositions throughout were seven percentage points more likely to be registered than respondents captured through secondary dispositions.