One of the great challenges in any research on voting behavior is to base conclusions on self-reported

voting measures found in surveys. There are a number of explanations for this, most of which relate to

the accuracy of recall in telephone surveys. Generally, surveys ask respondents to report behavior that is either misremembered, or misreported because of social desirability. Further, survey respondents generally are more likely to participate in the political process than those who do not, thus researchers are left with an over-representation of likely participants and a dearth of non-participants. In some instances, fielding a survey of likeliest participants is a beneficial circumstance. However, when one is trying to study the entire population including those who do and do not participate, this overrepresentation may prove to be problematic. In this paper, I examine the difference in the accuracy of self-reported vote found in two modes of the 2008 National Annenberg Election Survey (NAES): the NAES Online Panel and the NAES Telephone Post-election panel. In doing so, I pose one hypothesis and one research question. I hypothesize that ongoing survey contact increases the likelihood of both self-reported voting and actual voting. While addressing the repeated contact hypothesis, I also answer the research question, which mode of interviewing produced more accurate levels of self-report of voting: Telephone or online. The method I follow to address the research is to match the self-report vote data from the two surveys against a matched list from Catalist, a voter database company.