###########################################################
# This file replicates Table 7 in Hong and Park (2015) 
# If you use these data, please cite this paper. 
#
# Authors: Ji Yeon Hong and Sunkyoung Park
# Last updated: 6/9/2014
# Software: R 3.0.2
###########################################################

# use "BJPS_Hong_Park_fraud_dataset.dta"

library(foreign)
library(Matching)
fraud.data <- read.dta('BJPS_Hong_Park_fraud_dataset.dta')
attach(fraud.data)


###########################################################
##                   Model 1                             ## 
###########################################################

## Calculate a propensity score

glmapp_dum <- glm(app_dum ~ eligible + rural + urban+ turnout + gn + y1973 + y1978 + y1981 + y1985 + y1988, family=binomial)
rrapp_dum <- Match(Y = fraud, Tr=app_dum, X=glmapp_dum$fitted)
m7 = Match(Y = fraud, Tr = app_dum, X = glmapp_dum$fitted)

## Matching
MatchBalance(app_dum ~ eligible + rural + urban+ turnout + gn + y1973 + y1978 + y1981 + y1985 + y1988, match.out =rrapp_dum, nboots = 1000)

## Genetic Matching 

X7 <- cbind(eligible + rural + urban+ turnout + gn + y1973 + y1978 + y1981 + y1985 + y1988)
BalanceMatrix7 <- cbind(eligible + rural + urban+ turnout + gn + y1973 + y1978 + y1981 + y1985 + y1988)
genapp_dum <- GenMatch(Tr = app_dum, X = X7, BalanceMatrix = BalanceMatrix7, pop.size = 1000, ties=FALSE)

mgenapp_dum <- Match(Y = fraud, Tr = app_dum, X = X7, Weight.matrix = genapp_dum)
summary(mgenapp_dum)



###########################################################
##                   Model 2                             ## 
###########################################################

## Calculate a propensity score
glmbeg_dum <- glm(beg_dum ~ eligible + rural + urban+ turnout + gn + y1973 + y1978 + y1981 + y1985 + y1988, family=binomial)
rrbeg_dum <- Match(Y = fraud, Tr=beg_dum, X=glmbeg_dum$fitted)
m8 = Match(Y = fraud, Tr = beg_dum, X = glmbeg_dum$fitted)

## Matching
MatchBalance(beg_dum ~ eligible + rural + urban+ turnout + gn + y1973 + y1978 + y1981 + y1985 + y1988, match.out =rrbeg_dum, nboots = 1000)
 
## Genetic Matching 
X8 <- cbind(eligible + rural + urban+ turnout + gn + y1973 + y1978 + y1981 + y1985 + y1988)
BalanceMatrix8 <- cbind(eligible + rural + urban+ turnout + gn + y1973 + y1978 + y1981 + y1985 + y1988)
genbeg_dum <- GenMatch(Tr = beg_dum, X = X8, BalanceMatrix = BalanceMatrix8, pop.size = 1000, ties=FALSE)

mgenbeg_dum <- Match(Y = fraud, Tr = beg_dum, X = X8, Weight.matrix = genbeg_dum)
summary(mgenbeg_dum)


###########################################################
##                   Model 3                             ## 
###########################################################

## Calculate a propensity score
glmcom_dum <- glm(com_dum ~ eligible + rural + urban+ turnout + gn + y1973 + y1978 + y1981 + y1985 + y1988, family=binomial)
rrcom_dum <- Match(Y = fraud, Tr=com_dum, X=glmcom_dum$fitted)
m9 = Match(Y = fraud, Tr = com_dum, X = glmcom_dum$fitted)

 ## Matching
MatchBalance(com_dum ~ eligible + rural + urban+ turnout + gn + y1973 + y1978 + y1981 + y1985 + y1988, match.out =rrcom_dum, nboots = 1000)

## Genetic Matching  
X9 <- cbind(eligible + rural + urban+ turnout + gn + y1973 + y1978 + y1981 + y1985 + y1988)
BalanceMatrix9 <- cbind(eligible + rural + urban+ turnout + gn + y1973 + y1978 + y1981 + y1985 + y1988)
gencom_dum <- GenMatch(Tr = com_dum, X = X9, BalanceMatrix = BalanceMatrix9, pop.size = 1000, ties=FALSE)
mgencom_dum <- Match(Y = fraud, Tr = com_dum, X = X9, Weight.matrix = gencom_dum)
summary(mgencom_dum)