Who Benefits from a Low Turnout in the 2009 Local Election?
III. Swing Voters made the differences?
From the swing voters perspective, the poor performance of the central
government is the major reason that some previous ruling party supporters swing their votes to the opposition party. Such argument is based on the retrospective voting theory—that is, voters tend to cast their votes to either reward or punish the ruling party after evaluating the government performance (Key 1966; Fiorina 1978).
According to the theory, it is straightforward to think that voters would take President Ma Ying-jeou’s performance into consideration when casting their votes to the incumbent president in the next presidential election. Yet, in the two local elections, such as the 2009 and 2010 elections, did voters still vote for/against the ruling party based on the policy implementation of the Ma administration? Did the retrospective voting theory also explain the linkage between the performance of the central government and results of the local governments?
Previous scholars have adopted the concept of retrospective voting theory to develop the so-called referendum voting model to explain voting behavior in the (midterm) elections that between the two major (national) elections. The model suggests that even in local elections where none of the candidates takes direct responsibility for the success/failure of policy implementation of the central
government, the performance of the ruling party in the central government still play a significant role in determining the results of local elections simply because voters may take local elections as a venue to exercise a vote of confidence in the ruling party (Piereson 1975; Simon, Ostrom, and Marra 1991).
For example, empirical studies on US gubernatorial elections suggest that presidential approval rate (or a president’s popularity) more or less has an impact on the outcomes of gubernatorial elections (King 2001).8 Additionally, as the state of the economy is a common indicator that researchers use to assess the performance of the central government, it has been shown that a country’s economic situation will
8 King (2001) provided a thorough review regarding how national factors influence gubernatorial elections.
8
directly affect the central ruling party's performance in local elections (Niemi, Stanley and Vogel 1995 ; Svoboda 1995; Partin 1995; Stein 1990; Remmer and Gelineau 2003). In short, on a basis of referendum voting model, scholars analyzing either micro- or macro-level data have found that the outcome of local elections, particularly the winning odds for the ruling party, tend to be tied with the performance of the central government (Simon 1989).
Taiwan's relevant literature, mainly in focus on presidential and legislative elections, has provided mix empirical evidence regarding retrospective voting: while some studies found that one’s assessment of the economic situation will affect his/her vote choice (Wang 2004; Chen 1998), some found no significant linkage between the two variables (Hsieh et al. 1998). And very few studies have paid attention on the linkage between the performance of central government and the outcomes of local elections. One of the exceptions would be Huang and Cheng’s (2005) research that explored aggregate level data and found that national economic well-being (measured by unemployment rate) has a greater impact than local economic situations on the ruling party’s prospect in magistrate elections.
Prior to the 2009 election, the ruling party KMT’s performance clearly did not match the expectation of the general public. From a viewpoint of objective assessment, Taiwan’s economic condition has been bad as the economic growth rate has been negative for the past three seasons; from a viewpoint of subjective assessment,
President Ma’s approval rate was around 30% while about 45% of survey respondents were dissatisfied with President Ma’s performance. After all, the Ma administration did not provide a satisfying score card right before the 2009 election. Yet, the extent to which the performance of the central government affects the outcome of local election should be an empirical question.
Was the loss of the ruling party KMT in the 2009 local election mainly due to KMT’s lack of voter mobilization or in fact because of swing voters? The following section utilizes pre-election survey data to answer this question. The survey took place one month prior to the election and used telephone survey method to collect a random sample of 869 respondents across 15 counties/cities. It did not reach any respondents in the two island counties, Kinmen and Liangchiang.
First, we explore whether voters who voted for the KMT president nominee Ma Ying-jeou were more likely to vote for the KMT candidates in the 2009 magistrate
9
election. Table 1 shows a two-way analysis of the 2008 and 2009 vote choices.9 In Table 1, we found that about 48% of the Ma Ying-jeou supporters are still willing to vote for the KMT magistrate candidates while only about 9% of them tend to swing toward the DPP candidate. Additionally, 43% of the Ma Ying-jeou supporters will not go to vote (or have not made the decisions). On the other hand, among those who support the DPP presidential candidate Frank Hsieh, 51% of them are willing to vote for the DPP candidates in the 2009 local election, only 7% will shift to support the KMT candidates, and 42% have not decided yet. While the majority of the DPP supporters (more than 50%) in the 2008 presidential election tend to maintain their partisan preferences when they consider who they will vote for in the 2009 local election, less than the majority of Ma Ying-jeou’s supporters tend to maintain their partisan support across the two elections. It seems that the degree of cohesion among the DPP supporters is better than that among the KMT supporters. Yet, the difference between the two parties in cohesion seems marginal (i.e., 51% vs. 48%).
Table 1: Vote Choices of the 2008 and 2009 Elections
Voting Intention in the 2009 Local Election Vote Choice in the 2008
Presidential Election KMT DPP Others Total (# of respondents) KMT: Ma Ying-jeou 47.8% 8.8% 43.4% 100% (464) DPP: Frank Hsieh 7.3 50.8 41.9 100% (124)
Others 13.2 10.7 76.2 100% (281)
Note1: “Others” in the 2008 presidential election includes those respondents who answered “forget”,
“not eligible to vote”, “did not vote”, “cast an invalid ballot”, and “refuse”; “Others” in the 2009 local election includes those respondents who answer “vote for any other candidate”,
“undecided”, “not willing to vote”, “may cast an invalid ballot”, “it depends”, “no comments”,
“don’t know”, and “refuse”. We assume that those respondents who did not answer their vote choices are likely to not go to vote.
Note 2: The sample size is 869.
Additionally, a very similar proportion of partisan supporters in the 2008 election tend to change their supports toward the opposite party candidates in 2009 (KMT 8.8% vs. DPP 7.3%). And the respondents who have not decided prior to the 2009 election also almost equally distributed between the two party supporters (KMT 43.4% vs. DPP 41.9%). In short, in Table 1 we are not able to detect whether a
9 For the 2008 presidential election, the wording of the question is: “Who did you vote for in last year’s (2008) presidential election?” For the 2009 local election, the question becomes: “Who would you vote for if tomorrow were the election date?”
10
considerable proportion of voters tend to swing their votes from the KMT to the DPP.
And we are also not able to conclude whether Ma Ying Jeou’s supporters are more hesitant to decide who they will vote for in the 2009 local election.
Table 2: Vote Choices of the 2008 and 2009 Elections（Independent Voter Only）
Voting Intention in the 2009 Local Election Vote Choice in the 2008
Presidential Election
KMT DPP Others Total
(# of respondents) KMT: Ma Ying-jeou 15.3% 10.2% 74.5% 100% (98) DPP: Frank Hsieh 6.7 23.3 70.0 100% (30)
Others 9.0 7.9 83.1 100% (177)
Note1: “Others” in the 2008 presidential election includes those respondents who answered “forget”,
“not eligible to vote”, “did not vote”, “cast an invalid ballot”, and “refuse”; “Others” in the 2009 local election includes those respondents who answer “vote for any other candidate”,
“undecided”, “not willing to vote”, “may cast an invalid ballot”, “it depends”, “no comments”,
“don’t know”, and “refuse”. We assume that those respondents who did not answer their vote choices are likely to not go to vote.
Note 2: The sample size is 305.
How about independent voters? Did they tend to switch from one party to the other? We then only collect the sample of “independent voters” (those who insist they do not have any partisan attachment) in the survey and conduct the same two-way analysis. Table 2 shows that among those independent voters who support Ma
Ying-jeou in 2008, only about 15% of them are willing to vote for the KMT candidate in the 2009 local election. And more than 10% of them are ready to switch to support the DPP candidates. On the other hand, among those independent voters who support Frank Hsieh in 2008, over 23% of them are still willing to vote for the DPP candidate in 2009. And only less than 7% of them tend to change their support toward the KMT candidates. This cross-tabulation analysis seems to suggest that the independent voters were more likely than partisan voters to shift from the KMT to DPP. Yet, due to the very small sample size (only 305 independent voters), we are not able to assure such finding.
The above analyses seem to suggest that only a small proportion of voters may switch their support in the 2009 local election. Next, we further explore why
Taiwanese voters remain/change their support between the 2008 and 2009 elections.
Specifically, on the basis of referendum voting model, we focus on four questions as follows:
11
1. Does the poor performance of the central government cause Ma Ying-jeou’s supporters to swing toward the DPP candidates in the 2009 election, or simply cause Ma’s supporters to not turn out to vote?
2. Did Frank Hsieh’s supporters swing toward the KMT candidates or choose not to vote in the 2009 election simply due to their positive assessment of the central government’s performance?
3. Did the non-voters in the 2008 presidential election come out to vote for the KMT candidate in 2009 because of their positive assessment of Ma Yig-jeou’s
performance?10
4. Did the non-voters in the 2008 presidential election come out to vote for the DPP candidate in 2009 because of their negative assessment of Ma Yig-jeou’s
performance?
Corresponding to the above four questions, we define four dependent variables, and set up four logistic models (Model I ~ IV). Following is the coding scheme for the dependent variables, respectively:11
1. Do Ma Ying-jeou’s supporters still intend to vote for the KMT magistrate candidates? (Ma Yin Jeou’s supporters vote for the KMT = 1; Ma Yin Jeou’s supporters do not vote for the KMT = 0.)
2. Do Frank Hsieh’s supporters still intend to vote for the DPP magistrate candidates?
(Frank Hsieh’s supporters vote for the DPP = 1; Frank Hsieh’s supporters do not vote for the DPP = 0.)
3. Do non-voters in the 2008 vote for the KMT magistrate candidate? (2008
non-voters vote for the KMT =1; 2008 non-voters do not vote for the KMT = 0.) 4. Do non-voters in the 2008 vote for the DPP magistrate candidate? (2008
non-voters vote for the DPP =1; 2008 non-voters do not vote for the DPP = 0.) According to the referendum voting model, a voter’s assessment toward the performance of the central government will determine his/her voting behavior even in local elections. We set up three major independent variables and hypothesize that a voter’s voting intention will be driven by his/her 1) level of satisfaction with Ma Ying-jeou’s job performance; 2) assessment of Ma’s policy impact on personal
economic well-being; and 3) assessment of Ma’s policy impact on national economy.12
10 By non-voters, we assume that those respondents who did not clearly answer their vote choices are very likely to not go to vote. Thus we classified those respondents as non-voters. This is indeed a very strong assumption.
11 Please refer to Table A1 in Appendix for details.
12 The wordings of the three questions are as follows: 1. Are you satisfied with President Ma’s overall performance? 2. Do you think your household economic well being gets better, worse, or no change due to President Ma’s cross-strait economic policy? 3. Do you think national economy gets better,
12
It is important to note that the latter two variables are set to test the so-called
“pocketbook voting” and “sociotropic voting” models, respectively: While the pocket book voting model suggest that a voter may support the ruing party simply because his/her (personal/household) economic well-being is getting better (Campbell et al.
1960), the sociotropic voting model argues that a voter’s assessment of the state of national economy affect his/her intention to vote for the ruling party (Kinder and Kieweit 1979, 1981). Yet, our question wordings are little different from the
traditional ones. That is, we ask whether President Ma’s cross-strait economic policy has positive/negative impact on personal/national economic conditions. As cross-strait economic relationship is the most important policy dimension in Taiwan, we think that this kind of question wording is actually better in the sense that it also measures a respondent’s assessment of the central government performance.
In addition to the above three independent variables, we also include
demographic variables (e.g., gender, age, and education level) and party identification variables as control variables. And as voters actually cast their votes in different local contests, in order to control geographic differences, we include an additional control variable that indicates whether there is a KMT or DPP incumbent running for
reelection. It is argued that an incumbent candidate certainly enjoy some advantage in Taiwan’s electoral context (e.g., name recognition, resource distribution, and so on) when running for reelection (Lo 2001).13
Table 3：Determinants of Changing (or Maintaining) Voting Intentions, 2008-09
Model I Model II Model III Model IV Dependent Variables Consistent
KMT
Consistent DPP
Non-voter to KMT
Non-voter to DPP Independent Variables Coef.
（Std）