political influence and voter preferences · for roads officially completed before start of 2011...
TRANSCRIPT
POLITICAL INFLUENCE AND
VOTER PREFERENCES: EVIDENCE FROM INDIA AND PAKISTAN
01 June 2016 1
Jacob N. Shapiro
Princeton University
June 1, 2016 India Pakistan
Jonathan Lehne (PSE) Michael Callen (Harvard)
Oliver Vanden Eynde (PSE) Ali Cheema (CERP/IDEAS)
Adnan Khan (Oxford)
Asad Liaqat (Harvard)
Farooq Naseer (LUMS)
Co-Investigators:
Why This Matters
2
Streetlight outside PTI supporter’s home in
Lahore, September 2015
PMLN supporter’s home <100m away,
bears name of the candidate who got the
light placed there using government
resources
Part 1: Political Influence on PMGSY
3 Deterrence with Proxies - Minerva / ONR
Data
4
Variable Source Definition
Contractor
political
connections
PMGSY &
ECI
Change in share of contractors sharing last
name of winner and runner-up
Missing roads 2011
census &
PMGSY
For roads officially completed before start of
2011 census variable = 1 if all villages listed on
road had no road (tarmac, gravel, or wbm) in
2011 census
Cost PMGSY Log(sanctioned cost/km)
Delays PMGSY Actual date of completion – contracted date of
completion (in days)
Cost Overruns PMGSY Total final cost/sanctioned cost
Quality PMGSY Dummy variable = 1 if road “unsatisfactory” or
“in need of improvement” in latest inspection
Traditional RD Plot
5
-0.0
5-0
.03
0.0
00
.03
0.0
5
Cha
ng
e in
the
sh
are
of sam
e n
am
e c
ontr
acto
rs
-.2 -.1 0 .1 .2Win margin
Linear fit Quadratic fit
Varying Sample Bandwidth
6 24 February 2016
So What?
7
Politicians have no formal role in PMGSY, but their
influence on contract allocations is significant:
– Share with winner’s name increases from 4% to 7%
– 1600 road contracts (worth ≈ $ 470 million) affected
– Cost 12% more, 14% more likely to fail inspection
– 73 additional missing roads ≈ 125,000 people
deprived of access to wider Indian economy.
– But there’s no electoral cycling
So why do politicians get away with this?
Part 2: Political Networks and Voting
8
Add for UC-48 Chairman and Vice-Chairman Add for UC-48 General Councilor
Candidates
Pre-Election Experiment
9
Treatment:
– Political connections of candidates
– Spending by the current government
Outcomes:
– Beliefs about connections
– Beliefs about government performance
– Satisfaction with performance
– Support for candidates
Experimental Results
10
Beliefs:
– Movement depends on priors
– Those who underestimate move up & vice versa
Satisfaction:
– Moved up in places government spends a lot
– Not moved by connectedness information
Support:
– Connectedness information moves it up
– Performance information does not
What Happened in the Election?
11
What Happened in the Election?
12
Connections matter, a lot…
– 353 candidates polled >1,000 votes
– 209 more votes if connected to current MPA
– 16% greater chance of winning
But…
– If MPA is PML-N it’s 275 more votes
– 401 less if MPA is PTI
– Ties to Thana/Katcheri matter even more
– And correlation with past service delivery
Conclusion
13
Political influence is costly
And voters are choosing on factors other than
performance
The impact of past performance on voter evaluations is
small vs. prospective assessments of what politicians can
do for them based on politician characteristics
Presents a challenge if we expect local government to
solve accountability gaps