서베이앤헬스폴리시리서치센터

Kim, S. W., Traugott, M., Park, S. H., and Lee, S. K. (2011). “Why did the pre-election polls in South Korean local elections go all wrong? : assessing the source of errors using dual-frame landline/cell phone post-election survey,” Paper presented at the 64th annual conference of the World Association for Public Opinion Research, Amsterdam.

 The Panel on Pre-election Polls, 2011 WAPOR

(21-23 September, Amsterdam) 

WAPOR Panel Proposal: Pre-election Poll Estimation Problems in Comparative Perspective

While pre-election polls around the world have generally been improving their accuracy in general elections, they continue to encounter difficulties in sub-national elections.  Recent polling for primary elections in the United States, gubernatorial elections in Mexico, mayoral elections in Korea, provincial elections in Canada, and elsewhere in the world have exhibited these problems. This panel will explore the common issues that pollsters face in these circumstances as well as the unique aspects of the electoral context in each country that contribute to estimation errors in such elections. We expect to have four papers on the panel.

1. Korea

Why Did the Pre-election Polls in South Korean Local Elections Go All Wrong?: Assessing the Source of Errors using Dual-Frame Landline/Cell Phone Post-election Survey

Before the 2010 South Korean local elections for big-city mayors and provincial governors, often seen as a midterm referendum on the president, the leading media groups and research firms predicted that the ruling Grand National Party (GNP) would take sweeping wins in pre-election opinion polls, especially including the metropolitan areas such as Seoul and Incheon. But the GNP won only six of 16 crucial races, while its main rival, the Democratic Party, won seven and the remaining races were won by independents and a small party. Although there has been much debate on the issues of survey methodological problems or a phenomenon known as the “spiral of silence” related to poll failures, there is always a substantial lack of scientific evidence to support them. We first describe the survey methodology adopted in most pre-election polls, and then introduce the dual frame landline/cell phone survey design used in our post-election study to investigate what they missed. Based on the results obtained from the dual frame survey, we show the potential source of errors at geopolitical or local levels and discuss some of the best ways to improve poll accuracy.      

2. Canada

Polls at the Subnational Level: The Canadian Case

3. Mexico

The Unnatural Left-Right Coalitions: Challenges for Preelection Polls in Mexico

4. USA

Media Coverage as a Contextual Explanation for Estimation Errors in Pre-Primary Polls in the United States