서베이앤헬스폴리시리서치센터

Kim, S. W., Traugott, M. W., Kwak, N. J., Choi, E. H., and Lee, H. N. (2014). “Using two-wave dual frame RDD telephone pre-election poll in the 2012 Korean presidential election,” Paper presented at the 67th annual conference of the World Association for Public Opinion Research, Nice, France.

Abstract

The outcome of the presidential election is predicted with a small margin of error by the pre-election polls before the election. However, the poll estimates can vary, due to events or media coverage during the campaign. For example, the surprising move of the candidates or the unexpected withdrawal of their candidacy may affect people’s opinion on whom they plan to vote for. Thus, the powerful way to incorporate such dynamic elements in the polls would be necessary, and the panel design interviewing the same individuals at different points in time would be useful for the purpose since it provides direct evidence of changes in attitude or behavior or decision or awareness of voters.

We conducted a pre-election panel study with a national random sample of respondents before and after the official campaign period in the 2012 Korean presidential election. This study utilizes the dual frame RDD design of both landline and cell phone numbers to eliminate the coverage bias due to cell-only populations. This design uses list-assisted RDD sampling based on 100-banks of phone numbers for a landline sample as well as RDD sampling based on 10,000 banks for a cell phone sample. The weight variables were developed to avoid the overlap and overrepresentation problem in the dual frame RDD designs.

We present the AAPOR (WAPOR) response rates in the landline, cell, and dual frames at two different points in time. The demographic distributions of respondents in each wave are compared with those of the population of voters. To assess the validity of this study we measure how well poll estimates in the two waves match actual election outcomes. We take a look at how the sudden departure of a candidate to be a strong rival affected people’s decisions during the campaign. We also present the level of public awareness of the leading candidates’ policies in the election. In addition, we show the evidence that the voter turnout, which had an impact on the presidential election result, was highly influenced by campaign activities (including encouraging others to vote or talking with family) of some age groups. These findings will encourage the use of dual frame telephone panel surveys in a pre-election study.