Predicted Probabilities

In observational studies looking at independent variables’ marginal effect (changes of X’s) on the success of dependent variables (Y=1) is meaningful and can reveal the substantive significance of determinant predictors (Hanmer & Kalkan, 2013). Predicted probability is referred as the probability of the occurrence of dependent variables. Measuring changes in predicted probabilities for dependent variables would provide a valuable statement to illustrate the effect from multiple independent variables (Gelman & Pardoe, 2007). Especially, this study attempts to compare the determinants for reelection in Taiwanese old and new electoral systems.

Evaluating marginal effect of predictors would effectively reflect their difference on reelection predicted probabilities. The following Table 5 shows the marginal effect on predicted probabilities for reelection by the hypothesized predictors. And the (– & + standard deviation)/2 is displayed in parentheses.

In general, marginal effects of predictors are greater in the new-system model than the old one. Regarding the effect of party endorsement, ―KMT endorsement‖ has the highest marginal effect on reelection predicted probabilities and after the electoral reform its effect is even higher. The increase rate is greater than 10 times (from 0.4734 to 5.4391) between the two models. Regarding the effect of personal

performance, marginal effects in the new-system model are greater too. It is worth to notice that marginal effects of the predictors ―constituency service time rate‖ is

Determinants for Reelection ─ A Comparative Study between Taiwanese Old and New Electoral Systems

25

originally negative, but it becomes positive in the new-system model. The marginal effect of ―time proportion of constituency services‖ changes from -0.3764 to 0.9193.

In the old-system model, predictor ―casework‖ generates the highest positive effect on reelection predicted probabilities (marginal effect is 0.0105). In the new-system model, predictor ―time proportion of constituency services‖ has the highest positive marginal effect=0.9193. Associated with the combined effect of independent variables, the determinant predictors are altered and the focus of electoral calculation would be switched as well.

According to the findings from ―marginal effect on reelection predicted

probability‖, the marginal effects of party endorsement and personal performance are greater in the new system than the old one. The statistical results state that the effects from both party endorsement and personal performance on reelection are strengthened in the new SMD system.

Table 5 Marginal effects on reelection predicted probabilities Variables

Marginal effect (with -+sd/2 in parenthesis) The old system The new system

Party endorsement

KMT endorsement 0.4734 (0.2283)

5.4291 (0.9921) DPP endorsement 0.4497

(0.2134)

Bill-introduced 0.0105 (0.0409)

0.0197 (0.1304) Casework 0.0668

(0.0677)

0.1322 (0.121) Time proportion of

constituency services Chairperson 0.0274

(0.0126)

-0.147 (-0.0733) Seniority -0.0269

(-0.0283)

-0.0715 (-0.1017) Source: The authors

Determinants for Reelection ─ A Comparative Study between Taiwanese Old and New Electoral Systems

26

The following Figure 2 & 3 display the trends of predicted probabilities divided by parties and by changes of the predictors ―bill introduced‖ and ―casework‖ in the old and new systems. Every effort from introducing bills or providing casework services may generate different magnitudes of effect on the changes of reelection predicted probabilities across different parties. Interpreting such difference might be meaningful for illustrating the determinants of reelection.

Generally speaking, the slopes of the KMT lines are gentlest in the four figures, and the lines of other parties are the steepest; that means, the marginal effects of every increased ―bill introduced‖ and ―casework‖ are more significant in the observations of other parties if compared to KMT and DPP. The old-system figures indicate that the lines of KMT and DPP are pretty closed; in contrast, in the new-system figures, the lines of DPP move ahead the other-party lines. That implies, the electoral reform alters election campaign patterns: It enlarges the distance between DPP and KMT, and expands the marginal effect of personal performance especially for smaller political parties.

Determinants for Reelection ─ A Comparative Study between Taiwanese Old and New Electoral Systems

Number of bill introduced

Predicted probability by changes of bill introduced (The old system)

KMT

Number of bill introduced

Predicted probability by changes of bill introduced (The new system)

KMT DPP Others

Figure 2 Predicted probabilities by changes of bills introduced Source: The authors

Figure 3 Predicted probabilities by changes of caseworks Source: The authors

IX. Conclusions

Taiwan has experienced an electoral reform since 2005, which provide a

―natural experiment‖ to examine the effect of electoral designs on the results of reelection. Party endorsement and personal performance contribute to the determinants for reelection according to previous observations; however, a comparative study focusing on a single-country context to interpret the matter of electoral designs is still needed. This study conducted a series of surveys on legislator

0

Predicted probability by changes of casework (The old system)

KMT

Predicted probability by changes of casework (The new system)

KMT DPP Others

Determinants for Reelection ─ A Comparative Study between Taiwanese Old and New Electoral Systems

28

assistants and collected secondary data on legislative activities, legislators’

representative backgrounds as well as victory margins in the elections. The data of constituency services, casework and attitude on ceremony invitations are collected by questionnaire surveys. The methods of two-stage Heckman probit regression and predicted probabilities of the outcomes are employed in this research.

The research findings indicate that both party-vote and personal-vote factors contribute to the determinants for reelection under Taiwanese old and new systems.

Nevertheless, compared to the old SNTV system, the new SMD system moves closer to party-vote determination due to the party-to-party competition in the election campaigns. Legislators endorsed by a party may significantly affect their reelection results—win or lose in election. Legislators from the big party, KMT, have a better chance of reelection, while legislators from small parties and independent candidates are less likely to be reelected. These findings proves the Duverger’s law, that is, plurality rule elections structured within single-member districts tends to favor a two-party system.

Given this, we still could not underestimate the determination of personal performance. Under the new SMD system, legislators have a high identifiability; that is, legislators are easy to claim credits from their effort on legislation and constituency services, but hard to avoid blame from idleness. There is no free-ride space due to the design of single representative per district. Legislators who have

single-minded-reelection incentives would dedicate to maximize the constituency benefits in order to obtain electoral support.

Besides, due to severe intra-party competition, legislators perceive the needs to enhance their personal reputation (and campaign capacity) in order to grasp the opportunities of party endorsement. Especially in recent Taiwanese elections, party member votes and citizen surveys were conducted prior to party-endorsement

decisions. Legislators, thus, undertake more constituency services and introduce more bills to demonstrate their accountability and responsiveness; otherwise serious

challengers emerge and catch the party-endorsement chances. Good performance on services and legislation seem to provide more certainty on party endorsement as well as reelection. The long-established reputation of the incumbents would deter the challenges from ―serious‖ co-partisans. It is personal performance that promotes party-endorsement opportunities and further ensures to win the reelection.

Determinants for Reelection ─ A Comparative Study between Taiwanese Old and New Electoral Systems

29

The determinants for reelection provide incentives and drive legislators’

representative behavior. In extent to party endorsement and personal performance, the research findings imply that legislators would stand their party lines in the legislative dynamics and simultaneously take care of their constituency benefits. Due to

individual identifiability and concerns for constituency connections, spending time on constituency services seems more important than on legislation. Legislators, as rational actors, behave in the manners to maximize their self-interests and calculate the outcomes of their effort to cultivate their personal votes which in some degrees extends the opportunities for next party endorsement, and further grasps opportunities of winning in elections.

Determinants for Reelection ─ A Comparative Study between Taiwanese Old and New Electoral Systems

30

Reference

Ames, Barry. (1995). Electoral rules, constituency pressures, and pork barrel: Bases of voting in the Brazilian Congress. The Journal of Politics 57 (2): 324-343.

Ames, Barry. (2002). Party discipline in the Chamber of Deputies. In Scott Morgenstern & Benito Nacif. eds. Legislative Politics in Latin America.

Cambridge University Press.

Batto, Nathan F. (2005). Electoral strategy, committee membership, and rent seeking in the Taiwanese legislature, 1992-2001. Legislative Studies Quarterly 30(1):

43-62.

Blanton, Shannon L. (2000). Promoting human rights and democracy in the

developing world: U.S. rhetoric versus U.S. arms exports. American Journal

of Political Science 44, January: 123-131.

Cain, Bruce, John Ferejohn, and Morris Fiorina. (1984). The Constituency Service Basis of the Personal Vote for U. S. Representatives and British Members of Parliament. The American Political Science Review 78: 110-125.

Cain, Bruce, John Ferejohn, and Morris Fiorina. (1987). The Personal Vote: