Party Incumbency Advantage in South Korea (Legislative Elections, 1988-2016)

forcing variable

• prevmargin: PARTY vote share margin at t-1

main outcome variables

• winner: Party winning probability at t
• vshare: Party vote share at t

plot details

• bin size: 1%
• 3rd degree polynomial

Conservative Party

# Imbens-Kalyanaraman bandwidth selection 
# using D. Caughey's code
rdoptband_catch(con$prevmargin[which(is.na(con$prevmargin)==F)], con$winner[which(is.na(con$prevmargin)==F)], cutpoint = 0)

## Optimal Bandwidth       RD Estimate    Standard Error 
##        0.13645864       -0.18741985        0.07230893

rdoptband_catch(con$prevmargin[which(is.na(con$prevmargin)==F)], con$vshare[which(is.na(con$prevmargin)==F)], cutpoint = 0)

## Optimal Bandwidth       RD Estimate    Standard Error 
##        0.13063236       -0.03437639        0.01395966

# Local-Polynomial RD Estimation with Robust Confidence Intervals
# Calonico, Cattaneo and Titiunik (2014a)
rdrobust(con$winner, con$prevmargin)

## Call:
## rdrobust(y = con$winner, x = con$prevmargin)
## 
## Summary:
##                         
## Number of Obs 1036      
## BW Type       mserd     
## Kernel Type   Triangular
## VCE Type      NN        
## 
##                    Left   Right 
## Number of Obs      471    565   
## Eff. Number of Obs 310    281   
## Order Loc Poly (p) 1      1     
## Order Bias (q)     2      2     
## BW Loc Poly (h)    0.1369 0.1369
## BW Bias (b)        0.2491 0.2491
## rho (h/b)          0.5497 0.5497
## 
## Estimates:
##              Coef    Std. Err. z       P>|z|  CI Lower CI Upper
## Conventional -0.1872 0.0917    -2.0404 0.0413 -0.3670  -0.0074 
## Robust                                 0.0465 -0.4211  -0.0033

rdrobust(con$vshare, con$prevmargin)

## Call:
## rdrobust(y = con$vshare, x = con$prevmargin)
## 
## Summary:
##                         
## Number of Obs 1036      
## BW Type       mserd     
## Kernel Type   Triangular
## VCE Type      NN        
## 
##                    Left   Right 
## Number of Obs      471    565   
## Eff. Number of Obs 323    300   
## Order Loc Poly (p) 1      1     
## Order Bias (q)     2      2     
## BW Loc Poly (h)    0.1452 0.1452
## BW Bias (b)        0.2707 0.2707
## rho (h/b)          0.5365 0.5365
## 
## Estimates:
##              Coef    Std. Err. z       P>|z|  CI Lower CI Upper
## Conventional -0.0335 0.0161    -2.0789 0.0376 -0.0650  -0.0019 
## Robust                                 0.0492 -0.0733  -0.0001

Democratic Party

# Imbens-Kalyanaraman bandwidth selection 
# using D. Caughey's code
rdoptband_catch(dem$prevmargin[which(is.na(dem$prevmargin)==F)], dem$winner[which(is.na(dem$prevmargin)==F)], cutpoint = 0)

## Optimal Bandwidth       RD Estimate    Standard Error 
##        0.13426911       -0.08299221        0.08751580

rdoptband_catch(dem$prevmargin[which(is.na(dem$prevmargin)==F)], dem$vshare[which(is.na(dem$prevmargin)==F)], cutpoint = 0)

## Optimal Bandwidth       RD Estimate    Standard Error 
##        0.14551417        0.01157425        0.02416518

# Local-Polynomial RD Estimation with Robust Confidence Intervals
# Calonico, Cattaneo and Titiunik (2014a)
rdrobust(dem$winner, dem$prevmargin)

## Call:
## rdrobust(y = dem$winner, x = dem$prevmargin)
## 
## Summary:
##                         
## Number of Obs 702       
## BW Type       mserd     
## Kernel Type   Triangular
## VCE Type      NN        
## 
##                    Left   Right 
## Number of Obs      300    402   
## Eff. Number of Obs 164    224   
## Order Loc Poly (p) 1      1     
## Order Bias (q)     2      2     
## BW Loc Poly (h)    0.1258 0.1258
## BW Bias (b)        0.2218 0.2218
## rho (h/b)          0.5671 0.5671
## 
## Estimates:
##              Coef    Std. Err. z       P>|z|  CI Lower CI Upper
## Conventional -0.0870 0.1118    -0.7783 0.4364 -0.3062  0.1322  
## Robust                                 0.3859 -0.3728  0.1441

rdrobust(dem$vshare, dem$prevmargin)

## Call:
## rdrobust(y = dem$vshare, x = dem$prevmargin)
## 
## Summary:
##                         
## Number of Obs 702       
## BW Type       mserd     
## Kernel Type   Triangular
## VCE Type      NN        
## 
##                    Left   Right 
## Number of Obs      300    402   
## Eff. Number of Obs 170    228   
## Order Loc Poly (p) 1      1     
## Order Bias (q)     2      2     
## BW Loc Poly (h)    0.1340 0.1340
## BW Bias (b)        0.2342 0.2342
## rho (h/b)          0.5719 0.5719
## 
## Estimates:
##              Coef   Std. Err. z      P>|z|  CI Lower CI Upper
## Conventional 0.0111 0.0292    0.3810 0.7032 -0.0462  0.0685  
## Robust                               0.7881 -0.0583  0.0768