THE ELECTORAL INCENTIVES FOR LEGISLATOR PERCEPTIONS IN MIXED SYSTEMS: INTEGRATING EVIDENCE FROM EAST ASIA

Abstract

How do mixed legislative systems shape legislator behaviour? This article analyses the three East Asian cases—Japan, Taiwan and Korea—to identify whether these conform to institutional expectations seen elsewhere (e.g., Germany and New Zealand), namely a difference between those elected in single member districts (SMDs) versus proportional representation (PR). Through survey data in all three countries, empirical evidence is largely consistent with the ‘best of both worlds’ thesis with behaviour and preferences varying by seat type.

ACKNOWLEDGEMENTS

This research was funded in part by grants from the National Science Foundation, the Korea Foundation and the Matsuhita Foundation.

Notes

Taiwan and Korea both have unicameral legislatures, while Japan uses this mixed system for the House of Representatives.

Korea established its first mixed system with SMDs and a national party list with seats allocated proportionally in 1996. However, in this year and 2000, a one-vote system was used where one's district vote was aggregated by party to fill the PR seats. A two-vote system like that of Japan was adopted just prior to the 2004 election.

Interviews with legislators, party officials and researchers of electoral politics in both Taiwan and Korea consistently identified Japan as the system most like their own, with the former routinely identifying Japan as the inspiration for their mixed system. In addition, legislators and party officials in both Taiwan and Korea were far less likely to know that the other country used a mixed system.

Besides 34 PR seats, Taiwan also reserves six seats in three multi-seat aboriginal districts.

Dual-listing is prohibited in both Korea and Taiwan, but an explicit rationale for this rule is unclear. In interviews party officials simply repeated claims of different expectations from each tier.

Similar patterns are seen in New Zealand. Correspondence with two legislators identified the list as providing particular skills or representing particular groups, creating greater diversity on gender, ethnicity and sexual orientation. Besides focusing on broad national issues, several list legislators also mirror district legislators as a means essentially to run an opposition campaign for three years.

The SIPS surveys can be acquired through http://www.sips.re.kr/

This distinction fails to reach statistical significance when restricted to analysis of the two largest parties.

The JoongAng Daily legislator surveys were conducted in conjunction with the Korean Political Science Association and Soongsil University. Additional information can be found through http://www.kpsa.or.kr/. I thank Prof. Sangjoon Ka for his assistance in access to these surveys.

The surveys, conducted in conjunction with the University of Tokyo, can be accessed through http://www.masaki.j.u-tokyo.ac.jp/ats/atpsdata.html

All of the DPJ's elected PR members who filled out the survey were dual listed.

Personal correspondence.

I thank Legislator Wu Yu-Sheng () and his staff for assistance in this.

A Wilcoxon rank-sum test between tiers failed to reach significance (p = 0.5135), whereas the same test to measure between parties is significant (p = 0.0002). A simple OLS regression with on the media references with controls for the two outlier legislators and a dummy variable for SMD legislators found the latter to not be significant. Similar results were found when adding an additional control for party, a dummy variable for DPP.

These surveys explicitly target the chief assistant of legislators, providing for a much higher response rate than many counterparts. Potential validity concerns in surveying chief legislative assistants exist; however, such problems should be minimal based on the intimate knowledge of the assistants and that the findings are largely compatible to the previous research. Furthermore, assistants are often the actual responders to traditional surveys of legislators.

A model was also run including interactions between the party and seat type, but several of the interactions dropped out due to collinearity. The resulting model produced similar results to the original model with no inclusions statistically significant. The dependent variable was also recoded into a dichotomous variable for a probit analysis (1= ranking constituency focus first) with largely consistent results.

The surveys ask for the amount of hours per week on each area. While the question stated in a typical 70-hour week, respondents varied widely in the amount of hours indicated. Therefore, this was recorded as the percentage of time spent in each category.

Only one DP PR legislator answered the survey question and thus a direct comparison with the GNP is not made.

List representatives were also asked about the party/district divide; however, the overwhelming majority answered neither option. In contrast over 95% of district respondents chose one of the two options.

The SIPS survey asks the frequency of conflict, with only less than 5% of respondents in either tier claiming to have frequent conflicts with the party. Likewise statistical tests (Pearson's Chi-Square) both to measure between tiers and between the GNP and DP were not statistically significant.

The Pearson's Chi-Square of the distribution of factors across tiers was also significant at the 0.001 level.

The same survey asks about non-policy election factors as well. However, there was no statistical difference among types of legislators within either the LDP or DPJ.