Family or Money? Campaign Spending and Family Politics in Taiwan’s Legislative Election
Chia-hung Tsai
(Election Study Center and Graduate Institute of East Asian Studies, National Chengchi University)
Lu-huei Chen
(Election Study Center and Department of Political Science, National Chengchi University)
Chi Huang
(Department of Political Science and Election Study Center, National Chengchi University)
International Conference on Family Politics and Elections in Southeast Asia
Aug. 3, 2018
Motivation
Whether campaign spending or political family affects candidates’ vote shares in Taiwan?
Weak party institutionalization tends to have Family politics. But Taiwan’s two-party system has consolidated.
The current mixed-member majoritarian system inherits person vote due to the legacy of SNTV in previous legislative elections. Political families are expected to exert influence on the election outcomes.
Background
One of the common characteristics of electoral politics in South east Asia is family politics (Teehankee, 2002, Fukuoka, 2013; Purdey, 2016).
It is revealed that 65 percent of candidates whose relatives holding office won seats in all of the national and local elections between 2001 and 2016, while 70 percent of Philippine Congress members are from political families.
Why Taiwan?
Taiwan has undergone democratic transition and a two-party system has emerged since 1990s.
As many local factions faded away, political families in which several members by blood or marriage hold elected offices are active in local and national politics.
If political families are brought by weak party institutionalization, why are they influential in Taiwan where the two-party system is consolidated (Wong, 2005)? Would they continue to play a role in Taiwan politics?
Between 2001 and 2016, 65.5 percent of candidates whose relatives holding office are elected and 43.9 percent of candidates who has no relative holding office were elected.
Political Dynasty
“The political family has an entrenched place within the modern political systems of South East Asian states.”(Purdey, 2016)
Dal Bo, Dal Bo and Snyder (2009) argued, “power begets power.” Using instrumental variables, they found that legislator’s terms would increase the probability of his/her relative entering Congress.
Coronel (2007) provided a list of seven advantages of political families in Philippines: money, machine, media, marriage, murder or mayhem, myth, and mergers.
Mendoza et al. (2012) found that successful dynasties appear in provinces with lower aggregate levels of health, education, and income.
“Indonesian’s oligarchs were eternally grateful to Suharto for their very existence and for the decades of order and stability he provided during their aggressive and often illegal entrenchment.” (Winters, 2011)
Campaign Spending
Mayhew (1974, 40-41) states, “Of campaign resources, one of the most vital is money.”
Scholars suggested that high-quality challengers consider to join the race only when national conditions are favorable to them . Campaign spending indeed influences the entry of a high-quality challenger (Jacobson and Kernell, 1983; Box-Steffensmeier, 1996).
We can infer that politicians with political family background are likely to participate in the election since they are well financed. The other candidates may either withdraw themselves or receive limited vote shares.
We cannot observe the performance of candidates who did not finish the competition, so we may over-estimate of the effect of spending on vote shares.
Functional Form
The U-shaped curve suggests the following functional form:
\[\begin{eqnarray}
E(Y|X)&=&\beta_{0}+\beta_{1}\text{Political Family}+\beta_{2}\text{ln(Spending)}+\beta_{3}\text{ln(Spending)}^{2}
\end{eqnarray}\]
Assume that we have a linear function that has quadratic term:
\[\begin{eqnarray}
y_{i} &=& \beta_{0}+\beta_{1}\mathrm{ln}x_{i}+\beta_{2}\mathrm{ln}(x_{i})^{2}
\end{eqnarray}\]
We can take derivative with respect to \(x\):
\[\begin{eqnarray}
\frac{\partial \mathrm{ln}y_{i}}{\partial \mathrm{ln}x_{i}}
& =&\beta_{1}+2\beta_{1}\mathrm{ln}x_{i}
\end{eqnarray}\]
Eq (2) means an increase in \(x_{i}\) will have “stronger” impact in \(y_{i}\) for more extreme values
Interaction model
In addition to the effect of political family and campaign spending, we want to estimate the effect of the interaction term. If the effect of campaign spending multiplied by political family is significant, political dynasties extend their influence on vote shares by waging the campaign.
\[\begin{eqnarray}
E(Y|X)&=&\beta_{0}+\beta_{1}\text{Political Family}+\beta_{2}\text{ln(Spending)}+\beta_{3}\text{ln(Spending)}^{2}+\beta_{4}\text{Political Family}\times \text{ln(Spending)}
\end{eqnarray}\]
Finally, we control for the effect of nomination by KMT, DPP and NPP.
\[\begin{eqnarray}
E(Y|X)=&\beta_{0}+\beta_{1}\text{Political Family}+\beta_{2}\text{ln(Spending)}+\beta_{3}\text{ln(Spending)}^{2}+ \\
& \beta_{4}\text{Political Family}\times \text{ln(Spending)} +\beta_{5}\text{KMT}+\beta_{6}\text{DPP}+\beta_{7}\text{NPP}
\end{eqnarray}\]
|
|
|
Vote shares
|
|
|
|
Model 1
|
Model 2
|
Model 3
|
|
Intercept
|
141.477***
|
142.477***
|
72.004***
|
|
(-23.222)
|
(-23.896)
|
(-19.531)
|
|
Political family
|
4.431**
|
11.976
|
-16.034
|
|
(-1.946)
|
(-41.197)
|
(-31.945)
|
|
Log of spending
|
-27.598***
|
-27.765***
|
-14.144***
|
|
(-3.445)
|
(-3.57)
|
(-3.024)
|
|
Log of spending squared
|
1.300***
|
1.306***
|
0.678***
|
|
(-0.126)
|
(-0.131)
|
(-0.117)
|
|
Political family X Log of spending
|
|
-0.463
|
|
|
(-2.525)
|
|
KMT nomination
|
|
|
13.797***
|
|
(-2.159)
|
|
|
|
|
|
|
DPP nomination
|
|
|
28.397***
|
|
(-2.239)
|
|
NPP nomination
|
|
|
14.359***
|
|
(-4.558)
|
|
N
|
262
|
262
|
262
|
|
R-squared
|
0.707
|
0.707
|
0.827
|
|
Adj. R-squared
|
0.704
|
0.703
|
0.822
|
|
Residual Standard Error
|
12.282
|
12.305
|
9.515
|
|
F-statistics
|
207.647
|
155.160***
|
173.408***
|
|
|
Findings
In terms of a change in unlogged spending, when spending increases by 10%, Y is expected to change about -27.598*0.095=-2.65.
Regarding the higher order term, the value of X at which the predicted value of Y reaches the lowest point of the function is - \(\beta_{2}/2\times\beta_{3}\) (Aiken, West, and Reno, 1991)
When log spending is 27.598/2.6 or 10.753, the predicted value of vote shares is the minimum point of the model.
family background and the interaction term between family background and log spending are not significant predictors.
log of campaign spending and its squared term still influence vote shares while the party nomination of three parties: KMT, DPP, and NPP, have significant effects.
Conclusion
Political families add some advantage to their candidates.
Campaign spending affects vote shares only after about NTD 46,630.
There is no multiplicative effect of political dynasty and campaign spending; political families do not have more leverage with more campaign spending.
Thank you for your attention
