Assessing Losers Consent in Mixed Systems: Public Perceptions of Taiwan’s 2008 Legislative Yuan Election

  • Rich, Timothy S. (Department of Political Science, Western Kentucky University)
  • Published : 2016.05.31


A large literature identifies perceptual differences between citizens that support winning parties versus losing parties. These analyses fail to capture the complexity of mixed member legislative systems where one can be a winner or loser at both the national and district level. This paper proposes a two-level framework for the analysis of mixed system. An analysis of Taiwan’s first election under a mixed system provides evidence that both support for a national winner and district level winner produce boosts in positive perceptions of the electoral system. National success appears to be a larger motivator of perceptions.


Losers’ Consent

Ultimately the survival of any electoral system requires a minimum level of public approval. Electoral institutions shape citizen attitudes and alter perceptions of accountability and trust,3 important aspects of civic culture necessary to sustain a liberal tradition.4 While evidence suggests that proportional systems produce higher public satisfaction with democracy, there is little research on how mixed systems shape perceptions5 In addition, many works view mixed systems as a subclass of proportional representation, rather than to argue that the configuration creates unique incentives and behaviors.6 Furthermore, in regards to public perceptions, few studies compare mixed systems to pure systems. Aarts and Thomassen (2006) for example use the Comparative Study of Election Systems (CSES) data and find that mixed systems do worse than both majoritarian and proportional systems in public perceptions of accountability, representation, and democratic satisfaction.7 Cursory evidence from 33 countries in the World Values Survey (WVS) also shows a lower percentage identifying as satisfied in mixed systems (33.71%) compared to non-mixed systems (46.45%).8

More generally, how electoral winners and especially losers view electoral institutions is critical to the viability of democracies.9 Electoral systems do not have a uniform effect on public attitudes as the system in large part determines winners and losers.10 The perceptions of political winners are certainly important, yet one cannot risk ignoring the perceptions of those who lose in elections, as “losers’ consent” in accepting the outcome and the policies that follow is critical to democratic viability.11 Support of democratic principles is crucial in systems where the losers have been consistently out of power. For example, high support of democratic principles for parties beyond the Liberal Democratic Party (LDP) and their traditional coalitional partner (Komeito) arguably explains the country’s long-term democratic stability. In line with prospect theory, election results have a greater impact on losers as losses are more heavily weighted than gains psychologically.12 Thus a citizen’s status as an electoral winner or loser plays an important intervening variable between electoral institutions and perceptions.

One naturally expects those who voted for the winning party/candidate to have more favorable views of the electoral system than losers, both in terms of government performance and how the system works.13 In 29 of the 33 countries studied in Anderson et al. (2005), winners declared more positive evaluations of the political system, with larger gaps between winners and losers in the newer democracies of Eastern Europe. Afrobarometer too shows a gap in institutional trust between winners and losers in all African democracies except Zambia, while the average rating of electoral fairness among winners were higher (and statistically significant) than the losers’ average.14

Cho and Bratton (2005) find support for the government increased after instituting a mixed system in Lesotho.15 Partisan differences mediated these perceptions, with supporters of parties who performed better post-reform more supportive of the government and democracy in general than supporters of parties whose electoral position worsened. Esaisson (2010) showed a disparity in satisfaction with democracy between winners and losers in three mixed systems: Germany (1998), New Zealand (1999 and 2002) and Bulgaria (1990).16 According to pre-election survey data from the Japanese Election Study (JESIII), supporters of the LDP (expected at the time to have a landslide victory) and their coalitional partner Komeito positively correlated with satisfaction with politics (p <.05 and .10 respectively).17 Yet all of these examples conflate a citizen’s vote with whether or not this party (alone or within a traditional coalition) gains a majority of seats. This fails to capture the complexity of mixed systems. For example, anecdotal evidence from multiple mixed systems suggest that supporters of opposition parties that won local SMDs perceived the electoral system as less unfair than those who supporters a losing district candidate. Similarly, the public in multiple electoral systems show higher rates of approval for their local representative over the legislature in general. How quickly perceptions change and whether partisan perceptual divisions are common among other mixed systems remains unclear.


A Two-Level Process

In every electoral system a gap exists between what the electorate expects and what the system can realistically deliver. New systems (especially comparatively more complex ones) are no exception. Unlike pure proportional systems, the enactment of a mixed system creates multilevel winners and losers. Building off the losers’ consent literature, I suggest that perceptions of the system are a result not only of whether a voter’s party was a “winner” or “loser” nationally, but whether the district candidate was also successful.

Cursory analysis suggests that the public quickly adapts to the institutional incentives of mixed systems, even if surveys suggest a lack of understanding. Public knowledge is generally limited throughout the stages of electoral reform, leaving much of the burden of understanding a new system to the trial and error of actually participating in the electoral process. Cursory evidence suggests many misconceptions regarding the technical aspects of mixed systems, even in early adopters such as Germany, 18 New Zealand, 19 and Japan.20 Paradoxically, despite claims to not understand their mixed system, evidence of strategic voting is commonplace in these same systems as voters analyze both the SMD and PR tier and vote to maximize coalitional seats. 21 This suggests perhaps that citizens who claim they do not understand their mixed system are responding more to an expectations gap, potentially similar to the underlying motivations within the losers’ consent literature.

Mixed legislative systems also create a hierarchy of winners and losers that likely shape perceptions, yet the concept of losers’ consent has seldom been applied to analyses to mixed systems nor has it captured the mixed context. The implementation of SMDs creates greater distance between losers and legislators in competitive districts, especially where they replace SNTV systems, creating at least a temporary disconnect between public expectations of representation and reality.

Table 1 separates the two levels. I expect where a voter’s preferred party and candidate were both successful voters will have the greatest positive perceptions of the system. This is for two reasons. First is a sense of efficacy in that their vote mattered, even if no individual vote is likely to determine an electoral winner. Secondly, dual winners are likely to receive positive gains both in having their preferred policies enacted and in constituency service. Even if citizens only have vague policy preferences, the assumption is that partisans expect their preferred party to be more similarly minded than other parties. In contrast, where a citizen’s district candidate lost and the preferred party was not in the majority coalition, the expectation is for less positive perceptions. Dual losers expect policies directly counter to their own preferences. Those individuals whose preferred contestant were successful in only one area fall someplace in the middle, with national success weighted more than local success (B>C). The priority of the nation over local success again relates to national policy shaping abilities. District success may generate some desired constituency services with policy implications. However, unless the legislator is part of the majority party or coalition, the influence on national policy is likely to be quite limited and their ability to bring pork to the district is potentially limited. Following the table one would expect positive perceptions highest in the upper left quadrant (labeled A) and lowest in the bottom right (labeled D).

Table 1:Matrix of Types of Winners and Losers in a Mixed System

While this analysis focuses on mixed systems, the logic behind this matrix should apply to any electoral system with district elections. For example, the expectation is that Republican supporters would have more positive views after an election when a local Republican won in a district election even if the party fared poorly nationally compared to where Republicans fared poorly both locally and nationally. Furthermore a two-level analysis intends to bridge the gap between individual citizen action (voting) and broad perceptions of national politics by identifying whether a localized effect, district competition, influences these perceptions.

Of course the success or failure of preferred parties alone does not determine citizen perceptions of electoral systems. Just as the decision calculus behind individual votes includes sociodemographic and attitudinal influences, so too should perceptions be influenced by individual level factors. Similarly, some countries (e.g. South Korea) have a history of low evaluations of elected bodies, even if democracy remains entrenched. That said, if electoral systems create perceptual divisions between winners and losers, this division presents a challenge to democratic ideals.


Taiwan’s Election Reforms

While the Kuomintang (KMT) conducted local elections throughout the authoritarian era (1949-1987), Taiwan only effectively moved towards democratization with direct competitive elections for the Legislative Yuan in 1992. Prior to this year, the last election for Legislative Yuan seats occurred in 1948, with the body representing all of China. Supplementary elections replaced legislators who died in office while occasionally increasing the size of parliament for greater representation of Taiwan. Reforms under President Lee Teng-hui forced legislators representing China to retire, leaving the Legislative Yuan directly elected by the Taiwanese populace.

Prior to 2008, Taiwan employed the single non-transferable vote (SNTV) for legislative elections, much as Japan before 1996. In such systems voters have one vote for multimember districts, with district magnitude as high as thirteen in Taoyuan County in the election prior to reform. In the larger multimember districts in particular, only a sliver of the vote guarantees a candidate’s election. Using the Taoyuan Country district as an example, a candidate at best only needed 7.69% of the vote to be elected, with this percentage declining if one candidate was exceptionally popular.

Scholars are well aware of the pathologies of SNTV in terms of party proliferation and the difficulties in strategic voting.22 Parties run multiple candidates based on their expectations of party and personal appeals in the district, attempting to balance support across candidates in order to elect the maximum number party candidates. Identifying the proper number of candidates to run remained particularly cumbersome. If a party nominated too few candidates, it risked having a smaller legislative contingent, but nominating too many would decrease the likelihood they all would be elected. Candidates themselves had no incentive to follow the party’s goals in balancing support across candidates, but rather wanted to maximize support for themselves. Similarly, being too popular of a candidate may encourage voters to strategically vote from another member of the party and in the process inadvertently sabotage a strong candidate’s hopes to get elected. Despite these problems, the system traditionally produces rather proportional results.

Although the KMT initially maintained an advantage in SNTV due to effective vote rationing among party candidates within the same district, this advantage eroded over time. While changing the system had been considered for years, only in 2005 did Taiwan adopt reforms, agreeing to a parallel mixed member majoritarian (MMM) system. Voters cast two ballots—one for a district candidate in SMDs (73 total) and one for the national party list (34 seats). In addition the number of seats was cut in half, a move initially supported by members of even the smaller parties such as the People’s First Party (PFP) and Taiwan Solidarity Union (TSU) which risked extinction with its implementation. This rare downsizing of a legislative body echoed the general public sentiment regarding the Legislative Yuan.23

Whereas SNTV encourages a proliferation of district candidates and strategic defection from both the strongest and weakest candidates, the move to SMDs leads voters to generally concentrate around two candidates. As a measure of strategic voting, I use the second to first loser ratio (SF), the difference between the second runner up divided by the first runner up. The ratio drops from over .60 for the last four elections under SNTV, suggesting a large percentage of the vote for non-viable candidates, to just over .10 in 2008. This is despite the largest average number of candidates per district in sixteen years (3.81). Using the effective number of districts candidates, a weighted measure based on the amount of support, found an average of 2.1 effective candidates in 2008, remarkably consistent with Duverger’s Law. Furthermore, the percentage of the vote received by the top two candidates averaged 95.8 percent, again consistent with Duverger. District results thus suggest an understanding of the mechanical effects of the mixed system.

The DPPs poor showing in many district races, in part due to a nomination process that favored extremist candidates, led to only thirteen district seats. Both the KMT and DPP used a nomination formula in districts that combined a general public survey with that just of party officials. While the public survey aspect weighed heavier in this nomination process (70%), the sampling frame differed between parties. The KMT surveyed a sample of all eligible voters in the district, while the DPP only surveyed those who identified as likely to vote for the DPP or their traditional coalitional partner the TSU. These survey decisions resulted in many districts a DPP candidate popular among strong partisans, but further away from the median district voter.

Furthermore the mixed parallel system, despite a party list, does not ensure overall proportionality, especially if one party wins a disproportionate number of district seats. In the Taiwan case, the KMT received a slight majority of the party vote (51.2%), yet obtained 72.6% of the seats. Despite only a 14.3% difference in party vote between the two parties, a nearly 50% discrepancy in seats arises. This is not unique to Taiwan, with similar patterns evident in Japan and South Korea. For example, in Japan’s 2003 House of Representatives election, the Democratic Party of Japan (DPJ) received 37.4% of the party vote compared to the LDP’s 34.9%, yet the LDP captured 49.4% of seats compared to 36.9% for the DPJ. In 2005, the LDP received 38.2% of the party vote, but 61.67% of total seats. A similar spread is seen with the DPJ victory in 2009: with 42.41% of the party list vote and 64.17% of seats. A similar boost is seen for the largest party (and to a lesser extent the second largest) in South Korea in the past three elections. While DPP supporters may not have expected a majority in the Legislative Yuan, the poor showing likely reduced positive evaluations of the new system.



To analyze public perceptions of Taiwan’s mixed system, I use a survey from the Election Study Center (ESC) at National Chengchi University (NCCU). Conducted in June of 2009, a year after the election, this survey includes questions regarding party preference as well as the party of one’s district vote.24 The survey also includes three questions regarding perceptions of democracy and the electoral system:

Overall, to what extent are you satisfied with democracy in Taiwan?

Do you think that this election system is a fair system to all of political parties?

Do you think this election system is able to elect members of quality to the Legislative Yuan?

Each of these three variables is measured on a four point scale with higher scores indicating positive values. Table 2 collapses these down to dichotomous variables to show variation by party identification. Supporters of the KMT consistently show higher support for satisfaction with democracy and perceptions of the new system’s fairness and ability to elect quality legislators, a pattern that endures when restricted to just those who voted in the Legislative Yuan election. These initial findings largely are consistent with expectations from the losers’ consent literature, but fail to address the multilevel nature of mixed systems.

Table 2:Summary Statistics of Positive Perceptions

As each of these variables potentially measures a distinct dimension of perceptions of the mixed system, each will be used as the dependent variable in a series of ordered logit regressions, using the original four point scales. I include controls for age, gender (female), and education, as well as a four-point measure of interest in politics. To capture elements of losers’ consent, two dummy variables measure status as a winner. First, I measure whether a respondent’s preferred party was a national winner (defined as support for the KMT 25 ), which resembles operationalization in previous research on the topic. The survey unfortunately does not ask for the respondent’s vote choice on the party list. As such, I presume that respondents vote sincerely in the party list tier, consistent with literature on mixed systems, and use an included question on partisan identification for this measure. Secondly, a dummy measure captures whether the respondent voted for a winning district candidate, regardless of party affiliation. Finally I test each model under two specifications. The first includes all respondents with the second including only those who stated they voted in the legislative election. This distinction is to identify whether direct participation (in this case voting) shifts perceptual patterns.

Tables 3 through 5 present the logistic regression results. Starting with satisfaction with democracy (Table 3), results are consistent with the losers’ consent literature. Support for a national winner positively correlated (p. = .001) with satisfaction with democracy in both models with the largest coefficients. Meanwhile support for a district winner similarly correlated with satisfaction (p. = 0.05) in the total sample, but fails to reach significance when restricted just to voters. Coefficients in both models further suggest that support for a national winner creates a greater boost in positive views than a district winner. Moving to perceptions of the fairness of the system (Table 4), again support for both national and district winners correlated with more positive views at p. =.001 and 0.01 respectively across both models. Again the coefficient on the national winner variable is much larger than the district variable, consistent with expectations. In regards to the quality of legislators elected (Table 5) again both types of winners in both models positively correlate with more positive views at the .001 level. Furthermore, these models show with the smallest differential in coefficients on the national and district winner variables of across the logistic regression models.

Table 3:***p<.001, **p<.01, *p<.05

Table 4:***p<.001, **p<.01, *p<.05

Table 5:***p<.001, **p<.01, *p<.05

Controls for age, gender, education, and interest in voting do not appear to have a uniform effect on perceptions. Whereas age negatively correlates with satisfaction with government and views of fairness at the .001 level, age is not statistically significant in explaining perceptions of the quality of legislators. Female respondents were more likely to view the system producing quality legislators, but not significant in another models. Education positively correlated with satisfaction in Model 1, but failed to reach significance elsewhere. Finally, interest in politics negatively correlates with fairness at only the .10 level.

As an additional test, I combined the dependent variables of the original three models to create a new dependent variable that ranges from three to twelve and resembles a normal distribution (regression table omitted for space). With the same independent variables, being a national or district level winner positively correlated with more positive responses at the .001 level. Being a district winner in fact produces a roughly three times as large of a positive effect than district level success. Furthermore national and district success produce the largest and second largest positive coefficients in both models, consistent that electoral success more than demographic factors shape differences in perception of the mixed system.

Taken as a whole, empirical analysis largely supports expectations from the losers’ consent literature. Under three separate measures of perceptions and viewing both total respondents and just those who self-reported as voting, the findings consistently show that supporting a national winner produces a significantly positive boost in perceptions beyond that of supporting a district candidate. Yet, with the exception of one model, identifying as a winner at the national and district level both show a statistically significant positive boost in views. Overall, this analysis suggests the concept of losers’ consent can be applied to mixed systems and capture the differentiation of national and district level success.


Discussion and Conclusion

Losers’ consent remains a useful conceptual tool for analyzing public opinion across electoral systems. However defining winners and losers requires greater clarity, especially in the context of mixed legislative systems. This paper suggests a move in the right direction, a two-stage approach to identifying winners and losers. Empirical analysis confirms that simply winning at the national level is not the only electoral factor to boost positive perceptions of the electoral system. In perceptions of fairness and in the quality of legislators district level success also produces a boost in perceptions. Empirical analysis does not indicate that national and district success deserves equal weighting in understanding perceptions of electoral systems. Consistently identification as a national winner produced a much larger positive influence than district success, consistent with an implicit acknowledgement that national level success is crucial for policy implementation.

Public opinion surveys shortly after the establishment of a new electoral system allows insight into perceptual differences between winners and losers, but gives little indication as to the stability of these perceptions. In the Taiwan case, where the KMT captured over seventy percent of seats, may have polarized citizens more so than in subsequent elections. In contrast, the 2012 election produced far less lopsided results. The DPP increased their showing from 27 seats to 43, while the winners bonus shrunk from 21.4% in 2008 to 12.1%. The DPP also managed a small winners bonus, receiving 35.4% of seats with 34.6% of the party vote. Post-election surveys thus potentially provide a means to assess the stability in opinions between winners and losers.

Future research should apply this two-stage approach to other mixed systems not only to flesh out how electoral systems shape perceptions, but to determine whether systematic variation exists among mixed systems. For example, much of the mixed systems literature assumes the distinction between mixed member proportional (MMP) systems (e.g. Germany and New Zealand) and mixed member majoritarian (e.g. Japan and Taiwan) results in different strategies among voters. If voters interact with the system differently across subclasses of mixed systems, perhaps perceptions also differ in a similar way. Furthermore, since this two-stage approach applies to any system which employs district level elections, future work can directly compare perceptions between pure majoritarian and mixed systems to identify whether district success is consistent across systems.

Finally, the extent to which citizens claim to understand their mixed system likely varies by whether or not they supported a winning candidate and/or party, yet this remains largely unexplored. Rich (2014) finds that supporters of smaller parties in Taiwan were more knowledgeable about technical aspects of the system, such as the electoral threshold for PR seats, but how this relates to losers’ consent remains unexplored.26

Cited by

  1. Winning is not everything: Public perceptions of losers and non-voters in South Africa vol.51, pp.3, 2016,
  2. Losers’ and Non-voters’ Consent: Democratic Satisfaction in the 2009 and 2013 Elections in Germany pp.1477-7053, 2018,