Liberals should pray for rain: weather, opportunity costs of voting and electoral outcomes in South Korea

ABSTRACT

Studies suggest that rainfall on an election day benefits the Republican Party in the United States and conservative parties in Western Europe. A common explanation is that marginal voters, whose turnout decisions can be affected by weather conditions, are more likely to be supporters of liberal parties. This paper shows that bad weather on an election day instead benefits the liberal parties in South Korea, where election days are designated as special holidays. Young voters, who tend to prefer liberal parties, often plan other activities whose feasibility is contingent on good weather conditions; as a result, they are more likely to turn out, and thus to provide additional electoral support to liberal parties, when weather is bad.

KEYWORDS:

• Weather
• electoral outcomes
• opportunity costs of voting
• South Korea
• turnout

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. For example, in Australia (https://goo.gl/rG2QSE), Canada (https://goo.gl/icCMyj), Japan (https://goo.gl/gfXWIQ), the United Kingdom (https://goo.gl/PO1tuU) and the United States (https://goo.gl/GivCfb).

2. Earlier efforts to examine the link between weather and turnout include Gatrell and Bierly (2002), Knack (1994), Merrifield (1993), Scala (2003) and Shachar and Nalebuff (1999). However, these studies often suffer from poor estimates of meteorological variables (e.g., Merrifield 1993; Shachar and Nalebuff 1999).

3. Others, however, suggest that the costs associated with rainfall are marginal, so weather’s effects on election outcomes would be negligible (e.g., Knack 1994; Persson, Sundell, and Richard 2014).

4. Kang (2016) also reports similar findings. He examines the proportional tier of three legislative elections in South Korea in 2004, 2008, and 2012 and demonstrates that rainfall benefits liberal parties. This paper moves the focus of analysis to the single member district (SMD) tier and presents comprehensive analysis of weather’s electoral effects in South Korea, considering effects of temperature and rainfall together. Moreover, this paper provides an opportunity to examine the heterogeneous effects of rainfall on turnout conditional on voter age directly, using turnout by age group data measured at the district level. Despite the constraint of the aggregate level data, the analysis of turnout by age group allows for a limited but more precise test of the heterogeneous effect of rain conditional on voter age.

5. The Regulation on Public Office Holidays (Article 2) designates election days followed by the termination of office as a holiday for public offices. Private enterprises generally abide by this regulation.

6. See JooangIlbo, April 9 2014, https://goo.gl/36oO3X, accessed November 11 2016.

7. See Asiatoday, April 13 2016 https://goo.gl/J7rErL; Nocut News, May 31 2006, https://goo.gl/n0jAHN, accessed November 16 2016.

8. For example, see Chosun Biz, June 4 2014, http://goo.gl/YIB6cS; Yonhapnews, April 8 2012, http://goo.gl/Q6lWXz; The Korea Times, April 8 2014, http://goo.gl/nnB8PJ, accessed November 10 2016.

9. See Hankook Ilbo, December 24 2012, goo.gl/K2vqcC, accessed November 10 2016.

10. For example, Kang (2003) divides voters into three cohorts: the pre-386 generation, the 386 generation, and the post-386 generation, where the 386 label refers to people who were born in the 1960s, went to college in the 1980s and were in their 30s in the 1990s. Chung (2012) divides voters into four cohorts: the pre-democratization generation, the democratization generation, the post-democratization generation, and the X generation.

11. During the four elections examined in this paper, the seat shares for the two parties were 91% (2000), 91% (2004), 78% (2008) and 93% (2012).

12. Citizens older than 19 are automatically registered. A voter list is managed by a government agency, the National Election Committee.

13. The liberal party group includes: Democratic Labor Party (2004, 2008), Uri Party (2004), Creative Korea Party (2008), Unified Democratic Party (2008), Unified Progressive Party (2012), and Democratic United Party (2012). The conservative party group includes: Grand National Party (2004, 2008), Millennium Democratic Party (2004), People’s United 21 (2004), United Liberal Democrats (2004), Liberty Forward Party (2008, 2012), Pro-Park Coalition (2008), and Saenuri Party (2012).

14. See Gomez, Hansford, and Krause (2007) and Persson, Sundell, and Richard (2014) for further details on how to interpolate weather estimates in each administrative unit.

15. In 2008, monthly average rainfall in April was lower than in other years. In most areas, rainfall on Election Day was recorded as a monthly high. See Korea Meteorological Administration (2010).

16. While South Korea is a relatively small country, the distribution of rainfall is quite varied in the dataset. One standard deviation of rainfall is about 7.5mm, which is slightly bigger than the rainfall variation in Gomez, Hansford, and Krause (2007).

17. The size of each geographic unit is estimated from the shapefiles to calculate the number of polling stations per km² and the population density.

18. Alternatively, I consider a least-squares mixed-effects approach, which includes random effects for each legislative district and fixed effects for each election year. The results are robust across different specifications. Refer to Table A and Figure A in the Supporting Information section for details.

19. For example, the 2004 legislative election was considered an opportunity to find out how people felt about the impeachment of President Roh Moo-Hyun. On the other hand, the 2008 legislative election received relatively less attention because it was held in the honeymoon period after the election of President Lee Myung-Bak (inaugurated in January 2008). Widespread support for the new president and his party made legislative races less competitive in general and reduced turnout.

20. The results are substantively identical when using the continuous measure of rainfall. See Table B in the Supporting Information section.

21. I also examine the model with the squared term for weather estimates, to address the possibility that weather estimates and electoral outcomes have a non-linear relationship. See Table C and Figure B in the Supporting Information. The results are substantively identical.

22. I also consider a model examining the effect of rainfall in a binary format. The mere presence of rain increases liberal party candidates’ vote share by about 2.9 percentage points, and reduces conservative party candidates’ vote share by about 2.1 percentage points. See Table B in Supporting Information.

Additional information

Notes on contributors

Woo Chang Kang

Woo Chang Kang is an assistant professor in the Department of Political Science and International Relations at Korea University. His research interests are in political behavior and quantitative political analysis, with particular attention to Korea and the United States.