IVs … and
Democracy and Development
Lecture 2
• Mistake: should have read Luo et al.
(2007) … instead of Li and Rozelle (that is
for Monday …
• Monday’s assignment (Fixed Effects):
– Shaban
– Li and Rozelle
– Make up: Luo et al. (2007)
Today
• Angrist and Kreuger: IV analysis
– Notation
– Three uses of IVs:
• Simultaneity
• Measurement Error
• Omitted Variable (Unobserved Heterogeneity)
– Returns to Human Capital
– Other examples from the appendix
• Does democracy lead to development in
China?
– Luo et al. (2007)
Basic Notation
Y = a0 + a1*X + a2*Z
+e
where Y is the independent variable / outcome var
X is the endogenous variable
or variable of interest
Z is vector of observable control vars.
e is the error
[including factors that are unobserved that affect Y]
The problem: cov(x,e) ≠ 0
 Estimates of a1 are biased
Basic Notation
Y = a0 + a1*X-hat + a2*Z
+e
[second stage equation]
X = b0 +
+ b2*Z + γ*IV
+u
[first stage equation …]
where control variables, Z, are the SAME in both equations
IV is a vector of instrumental variables
two characteristics
a.) IV has an effect on X (or γ ≠ 0)
b.) IV does not affect Y, except through it indirect effect
on X
X-hat is the prediction from the first stage equation and in a
multivariate framework ONLY contains 1 type of net
information  part of X that is affected by IV
Go to Angrist and Krueger
Village Governance and Transforming
Rural China’s Communities
Scott Rozelle, Stanford University
Linxiu Zhang, Jikun Huang, Renfu Luo, Chengfang Liu, CCAP
Large Investments are Needed to
Improve the Quality of Life in
Rural Areas
Yet some villages have better infrastructure (and have enjoyed
a higher “quantity” of investment in recent years) … What
role does local governance play (ceteris paribus)?
Interest in the literature
• Chattopadhyay and Duflo (2004)
– Will election rules (e.g., reservation seats for women) affect investment
into type of investment that women want …
– Randomly assigned reservations … natural field experiment using village
level data
• Besley and Burgess (2001)
– Openness  more concentrated effort on helping the poor (using state
level data) … newspaper circulation …
– Worry about endogeneity of newspapers … IV: media ownership
concentration
• Rosenzweig and Foster (2003)
– Will democracy (election of local leaders through an popular vote-based
election) help the poor … by giving them a bigger voice?
– Governance structure  type of investment using village data
– Worry about endogeneity: use log of village population + dummy
variables indicating the passage of legislation relevant to local
governance
Our conceptual framework
• Following Rosenzweig and Foster, we
believe that elections will affect the
spending of China’s village leadership:
from expenditures that benefit the elite
(qingke chifan; salaries)
to expenditures that benefit the villages
(infrastructure)
Farmer Believing there has been Improvements
in Infrastructure since 1998
(from PRA focus groups)
（农户小组访谈）
Percent of farmers
Health
Schools
Drinking
Water
Irrigation
Roads
100
80
60
40
20
0
Farmer Dis-satisfaction with Current Status of
Infrastructure
(from PRA focus groups + Surveys)
Percent of farmers saying “not satisfied”
Health
Schools
Drinking
Water
Irrigation
Roads
100
80
60
40
20
0
Also willing to spend resources
Question #1:
If your village received
a grant of 50,000
yuan, what would
you spend it on?
a. infrastructure
b. environmental preservation
c. welfare payments
d. investing in the leadership office
facilities
Answer: 85% said
roads, irrigation or
drinking water
Question #2:
If village leader asked
you to contribute 20
yuan/capita to improve
the infrastructure in
your village, would you
be willing to
contribute?
Answer: 79% said “yes”
Main hypotheses
#1: Village elections will induce greater
spending on public goods
#2: Because villagers demand greater
spending on public goods, leaders that
deliver more public goods during their
term will have a greater probability of
being re-elected
Objectives of Presentation
• Understand the contours of investment in
China’s villages and examine differences
across villages
• What is the role of governance changes?
– How far has local democracy progress?
– Do villages with leaders that were elected
investment more projects / higher investment?
• Why?
– Do village leaders that invest more in public goods
get rewarded by the villager electorate?
Plan for Talk
Getting the facts right! [collect data]
Measure differences across communities (quantity and quality)
A sketch of China’s shift towards democracy
Empirical analysis:
Does having village elections  greater levels of investment?
Why?
Can projects be designed  ‘better’ investments?
Does having village elections  ‘better’ investment?
Getting Data Right
[“2000 village study”]
• Collected the data ourselves (Sept. 2003 …
revisit in 2005)
• 6 provinces, 1 from each region of China
• From each province randomly selected:
– 6 counties / 6 townships / all villages
– 6 x 6 x 6 = 216 townships
– All villages in each town
Collect information on:
Investments (6 years)
Village elections (6 years)
Other village characteristics
”2000 village survey”
Distribution of sample counties in China,
2003 Rural China Investment Survey
Jilin
Total number of villages: 2459
Hebei
Gansu
Jiangsu
Shaanxi
Sichuan
Broad coverage of surveys …
[“it was like Typhoon hit our village”]
•
•
•
•
•
•
•
•
•
Village leader
Party secretary
Village accountant
Small group leaders
2000 Farm Households
Focus groups
Engineering project evaluations
School / clinic evaluations
“Guerrilla” key informant interviews
Blocks of Survey
• Public goods investments / sources
• Engineering-based evaluations (key projects)
• Project design, villager satisfaction and investment
choices
• Fiscal data (Tax-for-Fee reform)
• Village elections (Election procedures) … and more
(other part of governance structure of village)
• Village Characteristics and Other
Creating the data set
6/7 years of data
Observation 1:
Observation 2:
During 3 year
election cycle, how
many projects were
implement
During 3 year
election cycle, how
many projects were
implement
During 3 year
election cycle …
was leader elected?
[yes or no]
During 3 year
election cycle …
was leader elected?
[yes or no]
N=7041
Also move to use
total
investment
Creating the data set
6/7 years of data
Observation 1:
Observation 2:
Observation 3:
During 3 year
election cycle,
how many
projects were
implement/year
During 1 year of
term, how many
projects were
implement/year
During last 2
year of term,
how many
projects were
implement/year
During 3 year
election cycle
… was leader
elected? [yes or
no]
During 3 year
election cycle
… was leader
elected? [yes or
no]
Obs. 4
During 3 year
election cycle
… was leader
elected? [yes or
IV does not change over
no]
time … though it is
constant across towns
• What are the trends of China’s public
investments into its villages over time and
across space?
Composition of investment projects
public vs. development
Total number of projects in
sample
9,138 projects
(in 2459 sample villages)
Development Projects
13%
87%
Public goods investments
Nearly 4 per village during the study period …
or almost 1 per year
Most of projects in
rural China now focus
on the provision of
public goods …
different than the
1980s
Scope of public goods investment projects
Project
N
Total investment (10k yuan)
Build road or bridge
Build school
Build clinic
Drinking water
1266
850
163
636
11.2
9.9
2.5
7.5
Irrigation system
Drainage system
Loudspeaker for village
committee
625
194
6.6
6.3
379
6.0
Land improvement
52
11.0
Watershed management
Land Leveling
151
124
29.8
13.6
Beautify environment
Forest closure
Eco-forest
157
140
55
2.4
3.4
3.4
Grain for green
892
6.7
Build pasture
Recreation center
Other public project
19
262
10
13.4
5.0
24.4
5,975
10.8
N
Type of PUBLIC GOODS projects
Other
public
projects,
37
Roads,
21
Schools,
14
Clinic, 3
Drinking 11
water, 11
Irri., 14
Around 2/3rds
of public
goods
investments
into 5 types of
projects
Funding sources of public goods investment
projects in rural China
Percent of Number of
Public Goods Projects
50
40
30
20
10
0
Funded fully from
above
Jointly funded
Funded fully by
village
Percent of Total Investment Amount in
Public Goods by Source
Person days per project
1200
1000
800
From
above
By village
47%
53%
600
400
200
0
Corvee Labor
Rising Overtime:
Public Goods Investment Projects
Projects per year
1.2
1
0.8
0.6
0.4
0.2
0
1998-2002
2003-2004
Rising … But Still Not Enough!
Total Spending: China versus Japan/Korea
Total Inv./Capita (US dollars, PPP terms
500
400
300
200
100
0
Japan-50s
Korea80s
China2000
China2004
.5
.8
Distribution of public goods projects,
number of projects and amount of
investment
.4
.3
Mean size of investment
project: about 50,000 yuan
.1
0
.2
.1
.3
.4
Mean number of
project: 3 to 4 for
sample period
Units: 10K yuan
.2
Density
.5
.6
.7
Units: number
of projects
5
10
15
20
25 30 35 40
investment volume
45
0
0
0
1
2
3
4
5
6
7
8
9 10 11 12 13
total number of PUBLIC projects at village level between 1998-2003
14
Kernel density estimate
Normal density
While some villages have lots of projects … others do not!
50
55
60
Determinants of Investment Quantity
Determinants of Quantity (number of
projects / amount of investment—yuan)
– Many factors
•
•
•
•
Village locations
Structure of village economy
Ethnic Make up
Policies (e.g., Tax Reform)
• Village governance reform
Focus of
the
paper
Village Elections in China
• They are now into 6th to 8th round … on average,
there has been 5 elections held in the typical
village … [there have been more than 5 million
“elections” in rural China since mid-1980s]
• BUT, a lot of heterogeneity
• Despite the progress in holding elections,
almost no one is going to claim that they are
perfect
• Some leaders still are appointed … procedures
are poor …
Village Elections in China
• They are now into 6th to 8th round … on average,
there has been 5 elections held in the typical
village … [there have been more than 5 million
“elections” in rural China since mid-1980s]
• Election of leaders is rising over time
• Some leaders still are appointed … procedures
are poor …
Villages that “Directly Elect” Village
Leaders … is growing over time
Percent
100
80
60
69
83
40
20
0
1995 to 1997
2002 to 2004
Village Elections in China
• They are now into 6th to 8th round … on average,
there has been 5 elections held in the typical
village … [there have been more than 5 million
“elections” in rural China since mid-1980s]
• Election of leaders is rising over time
• Some leaders still are appointed … and it varies
across space …
What is going on?
China’s rural election
Appointed village leader
Elected village leader
Frequency
Percent
Frequency
Percent
Jiangsu
379
28.03
973
71.97
Gansu
213
22.88
718
77.12
Sichuan
114
10.63
958
89.37
Shaanxi
211
19.36
879
80.64
Jilin
79
7.36
995
92.64
Hebei
492
32.33
1,030
67.67
Total
1,488
21.13
5,553
78.87
Though the election have been implemented for many years, there
are still villages that do not elect their village leader directly
Village Elections in China
• They are now into 6th to 8th round … on average, there
has been 5 elections held in the typical village … [there
have been more than 5 million “elections” in rural China
since mid-1980s]
• Election of leaders is rising over time
• Some leaders still are appointed … and it varies across
space …
• No one claims elections are perfect … but, signs that
they are getting better
Quite decentralized (how they run
elections?) But there are factors that
affect the nature of the elections
• Election committee governance
– (in >80% of village, who runs the elections?: the
village party secretary … that is the leader of the
village Communist Party cell … though there are
difference in participation of township/county officials)
• Nominations (not transparent … often subject to
“approval” of town officials
• Many times villages just won’t have elections …
they will only nominate one person … the
township will appoint a leader … the towns only
let one person run for village leader
So is there a relationship between
elections and public goods
investment?
-- Descriptive
-- Multivariate
1
100
0.8
80
Thousand Yuan
# of projects/year
Relationship between election of
village leader and total investment
0.6
0.4
60
40
0.2
20
0
0
Elected
Appoint
Elected
Appoint
Investment Projects and Elections
Village leader directly elected?("Yes"or"No")
Village leader directly elected?("Yes"or"No")
Total investment
45
45
Village leader directly elected ? ("Yes"
or "No")village
Investment
from
Village funded investment (Thousand
itself
Yuan)
12
Investment (Thousand Yuan)
40
40
10
35
35
30
30
8
25
25
6
20
20
15
15
10
10
E A
Yes
Yes
No
No
Yes
Yes
E A
E A
E A
4
E A
No
Yes
2
5
5
Yes
Yes
0
0
E A
Road
Road
Road
No
Irrigation
Irrigation
Irri.
No
No
0
School
School
Drinking
Water
Road
Road
Irrigation
Irri.
School
Drinking
Water
Determinants of Investment
Quantity
Determinants of Quantity (number of
projects / amount of investment—yuan)
– From paper (by type; by source …
using OLS + IV)
– From new work (total investment; by source
… using Fixed Effect + GMM)
Road project Probit (Tobit) analysis
Investment= a0 + a1*D-election + a2*other factors+Provincial fixed effects
All funding sources
Village funded only
Above funded only
Have or no road
project
(Probit)
Road project
investment level
(tobit)
Have or no road
project
(Probit)
Rroad project
investment level
(tobit)
Have or no road
project
(Probit)
Road project
investment level
(tobit)
0.14
(3.38)***
5.14
(3.20)***
0.15
(2.74)***
5.76
(3.13)***
0.16
(2.20)**
12.40
(2.01)**
0.12
(3.19)***
4.76
(3.10)***
-0.02
(0.45)
0.18
(0.10)
0.14
(2.19)**
12.40
(2.30)**
-1E-06
(0.02)
-0.004
(1.96)**
2E-04
(2.54)**
-0.001
(0.64)
-1E-04
(0.71)
-0.001
(0.05)
-4E-09
(0.31)
1E-06
(2.96)***
-1E-08
(0.74)
2E-06
(4.08)***
-5E-08
(1.28)
-5E-06
(1.56)
3E-05
(1.34)
0.0021
(2.80)***
-3E-05
(1.14)
0.0008
(0.97)
8E-07
(0.02)
-0.0032
(1.03)
0.002
(2.45)**
0.07
(2.16)**
0.001
(1.34)
0.03
(0.73)
0.004
(3.39)***
0.27
(2.69)***
Per capita land in 1997(mu)
-0.017
(1.66)*
-0.64
(1.55)
-0.033
(1.98)**
-1.03
(1.71)*
-0.010
(0.58)
-1.10
(0.77)
Percentage of effectively
irrigated land in 1997(%)
0.001
(1.12)
0.088
(3.57)***
0.001
(1.53)
0.092
(3.32)***
-0.001
(0.54)
0.047
(0.53)
Hilly land over 25 degree in total land
in
the village in
1997(%)
0.002
(2.87)***
0.072
(2.35)**
0.001
(0.73)
0.027
(0.75)
0.004
(3.44)***
0.211
(2.04)**
The distance of the nearest
road to the village in 1997(Km)
-0.002
(1.12)
-0.11
(1.64)
-0.006
(2.51)**
-0.28
(3.02)***
-0.002
(0.53)
-0.07
(0.27)
The farthest distance between two
small
groups in this village
in
1997(Km)
-0.007
(0.99)
-0.48
(1.79)*
-0.002
(0.29)
-0.07
(0.24)
-0.022
(1.57)
-1.67
(1.45)
The distance between village
and township seat in 1997(Km)
0.0021
(0.56)
0.15
(0.99)
-0.0015
(0.31)
-0.05
(0.31)
-0.0043
(0.63)
-0.31
(0.54)
Number of fellow villagers working
in
township
or
county
governments(person)
0.006
(2.92)***
0.37
(4.41)***
0.002
(0.54)
0.07
(0.74)
0.007
(1.90)*
0.63
(2.00)**
The illiterate rate of village
-0.13
(1.00)
-6.64
(1.28)
-0.72
(3.92)***
-31.70
(4.54)***
0.44
(2.33)**
31.80
(2.01)**
(yes=1, no=0)
-0.05
(1.18)
-1.01
(0.66)
0.01
(0.16)
0.79
(0.47)
-0.11
(1.47)
-9.05
(1.54)
Province Dummy
Yes
Yes
Yes
Yes
Yes
Yes
-1E+00
(7.23)***
-126
(7.37)***
7041
0.02
Village leader elected
directly(yes=1,no=0)
Dummy of rural Tax for Fee
reform(before=0,after=1)
Net per capita income
in 1997(yuan)
Net per capita income
square in 1997
Total population in 1997
(person)
Percentage of minority
population in 1997(%)
labor force in 1997(%)
major road passing by village
-35.6
(8.72)***
7041
7041
0.03
0.02
Absolute value of t statistics in parentheses, * significant at 10%; **
Constant
Observations
Pseudo R2
Data source: Authors’ survey.
-6E-01
(5.75)***
-58.3
(11.90)***
7041
7041
0.04
0.03
significant at 5%; *** significant
-2E+00
(11.6)***
7041
0.05
at 1%
Road project Probit (Tobit) analysis
(Focusing on the D-election coefficient)
All fund source
Have or no
Investment
project
level
(Probit)
(tobit)
Village leader
elected Directly
(yes=1,no=0)
Village funded only
Have or no
Investment
project
level
(Probit)
(tobit)
Above funded only
Have or no
Investment
project
level
(Probit)
(tobit)
0.14
5.14
0.15
5.76
0.16
12.40
(3.38)***
(3.20)***
(2.74)***
(3.13)***
(2.20)**
(2.01)**
Irrigation project Probit (Tobit) analysis
Investment= a0 + a1*D-election + a2*other factors+Provincial fixed effects
All fund source
Village funded only
Above funded only
Have or no
irrigation project
(Probit)
Irrigation project
investment level
(tobit)
Have or no
irrigation project
(Probit)
Irrigation project
investment level
(tobit)
Have or no
irrigation project
(Probit)
Irrigation project
investment level
(tobit)
0.14
(3.01)***
4.7
(3.23)***
0.15
(2.49)**
2.42
(2.48)**
0.08
(1.09)
1.95
(0.63)
-0.24
(5.54)***
-7.7
(5.54)***
-0.18
(3.09)***
-3.12
(3.39)***
-0.15
(2.20)**
-7.19
(2.37)**
-0.0001
(1.02)
-0.003
(1.61)
0.00010
(1.38)
-0.0004
(0.34)
0.0001
(0.66)
0.005
(0.92)
1E-09
(0.09)
0.0000004
(1.20)
-3E-09
(0.21)
5E-07
(2.61)***
-5E-08
(1.57)
-2E-06
(1.42)
-6E-05
(2.67)***
0.000
(0.45)
-0.0001
(4.43)***
-0.001
(2.75)***
-2E-05
(0.67)
-0.0004
(0.27)
0.000
(0.44)
0.08
(2.31)**
-0.002
(0.89)
-0.05
(1.52)
0.003
(2.24)**
0.16
(2.45)**
-0.02
(1.24)
-0.76
(1.55)
-0.036
(1.37)
-0.26
(0.67)
-0.014
(0.64)
-0.60
(0.64)
irrigated land in 1997(%)
0.0036
(5.19)***
0.084
(3.72)***
0.005
(5.24)***
0.08
(4.96)***
-0.002
(2.13)**
-0.10
(2.04)**
Hilly land over 25 degree in total
land in the village in
1997(%)
-0.003
(2.99)***
-0.09
(3.00)***
-0.0034
(2.31)**
-0.06
(2.51)**
-0.0020
(1.41)
-0.07
(1.04)
road to the village in 1997(Km)
-0.0056
(2.30)**
-0.20
(2.54)**
-0.010
(2.04)**
-0.14
(1.90)*
-0.005
(1.34)
-0.15
(0.99)
The farthest distance between two
small
groups in this village
in
1997(Km)
-0.022
(2.26)**
-1.03
(3.23)***
-0.015
(0.99)
-0.30
(1.24)
-0.02
(1.10)
-1.04
(1.59)
and township seat in 1997(Km)
-0.005
(1.06)
-0.17
(1.09)
-0.003
(0.44)
-0.10
(0.85)
-0.009
(1.09)
-0.08
(0.24)
Number of fellow villagers
working
in township or
county governments(person)
0.004
(1.62)
0.11
(1.36)
-0.006
(1.41)
-0.13
(1.87)*
0.003
(0.72)
0.16
(0.96)
-0.46
(2.85)***
-1E+01
(2.26)**
-0.32
(1.32)
-3.85
(1.01)
-0.26
(1.14)
-8.29
(0.81)
Village leader elected
directly(yes=1,no=0)
Dummy of rural Tax for Fee
reform(before=0,after=1)
Net per capita income
in 1997(yuan)
Net per capita income
square in 1997
Total population in 1997
(person)
Percentage of minority
population in 1997(%)
Per capita land in 1997(mu)
Percentage of effectively
The distance of the nearest
The distance between village
The illiterate rate of village
labor force in 1997(%)
Province Dummy
Constant
Observations
Pseudo R2
Yes
Yes
Yes
Yes
Yes
Yes
-0.75
(6.67)***
-28.90
(7.89)***
-1.53
(9.87)***
-25.48
(9.67)***
-1.32
(7.08)***
-66.0
(7.41)***
7041
0.06
7041
0.02
7041
0.13
7041
0.08
7041
0.03
7041
0.02
Absolute value of t statistics in parentheses, * significant at 10%; ** significant at 5%; *** significant at 1%
Data source: Authors’ survey.
Irrigation project Probit (Tobit) analysis
(Focus on the D-election)
All fund source
Village leader elected
Directly (yes=1,no=0)
Village funded only
Above funded only
Have or no
project
(Probit)
Investment
level
(tobit)
Have or no
project
(Probit)
Investment
level
(tobit)
Have or no
project
(Probit)
Investment
level
(tobit)
0.14
4.7
0.15
2.42
0.08
1.95
(2.48)**
(1.09)
(0.63)
(3.01)*** (3.23)*** (2.49)**
Multivariate Analysis: Elections and
Investment Quality in China’s Villages
Dependent
road project (total)
variables
Roads
Election
Village
leader elected
directly (yes=1,
variable
no=0)
drinking
water
Election
Village leader elected
directly (yes=1,
variable
no=0)
Election
Village leader elected
variable
directly (yes=1,
no=0)
Above funded road only
(1)
(2)
(3)
(4)
(5)
(6)
Probit
Tobit
Probit
Tobit
Probit
Tobit
+
0.08
(1.90)*
0.13
+
(3.06)***
+
3.11
(1.91)*
irrigation project (total)
irrigation
Village funded road only
4.09
+
(3.11)***
school project (total)
+
+
0.08
+
(1.42)
+
3.18
0.16
(1.72)*
+
(2.10)**
+
12.3
(1.95)*
Village funded irrigation only
Above funded irrigation only
0.13
1.74
0.05
1.38
(2.19)**
(1.98)**
(0.81)
(0.45)
+
Village funded school only
+
+
Above funded school only
+
+
+
0.14
18.2
0.18
4.13
0.17
15.2
(2.42)**
(2.94)***
(2.18)**
(2.42)**
(1.79)*
(2.32)**
Dependant variable
All projects
Number of public goods projects per term (All public goods projects)
(1) ols
(2) Tobit
+
The way village leader acceded to office
(0=appointed,1=elected directly)
(3) Fixed effect
+
+
0.178
0.293
0.314
(4.60)***
(4.90)***
(4.85)***
Problem with OLS regression
• Unobserved heterogeneity
• Need IV (use both 2SLS and bivariate probit):
Investment = a0 + a1*election + a2*other factors + e
Election = b0 + b1*IV
+ b2*other factors + u
Where, IV is specified to be two variables (in paper):
a.) measure of rule: does slate of village leader candidates have to be
“approved by the township” as part of the election process
b.) number of meetings held for a village’s election that was attended by
officials from county and township
[pass exclusion restriction/over-ID tests … there is no obvious evidence in
descriptive statistics that suggest rules and investments are related …
RHS Controls and IVs: Correlations?
The candidate should be
approved by upper government
Meetings hold by county and
township for term turn over
More than
Less than fifth
fifth
Yes
No
1.9
2.3
1.9
2.1
1621
1352
1751
1412
The illiterate rate of village labor force in
1997
Number of fellow villagers working in
township or county governments
% of hilly land over 25 degree in total land
in the village in 1997
0.1
0.1
0.08
0.11
5.6
4.9
6
5
21
25
21
24
The distance of the nearest road to the
village seat in 1997
The distance between village and township
seat in 1997
The farthest distance between two small
groups in this village in 1997
6.5
5.6
5.2
6.7
5.3
5.2
5.0
5.4
2.7
2.5
3.2
2.4
Per capita land in 1997
Net per capita income in 1997
Over-identification test and hausman test
Dependant variable
Have or no road project
Have or no Village funded road project
1.98
0.295
Hansen-Sargan
Chi-sq(1)
over identification
P-value
0.1598
0.587
Chi-sq(1)
0.0526
1.503
P value
0.8226
0.2202
Have or no irrigation project
Have or no Village funded irrigation
Test
Hausman Test
Dependant variable
project
Hansen-Sargan
Chi-sq(1)
2.63
0.237
over identification
P-value
0.105
0.626
Chi-sq(1)
1.029
0.014
P value
0.311
0.906
Test
Hausman Test
Are our IV measure measuring
“dysfunction”?
• Correlation of villages within a township
(high) … if they have to approve the
nominations of slates of candidates in one
village in the township, the probability is
high that the township has to approve the
slate of candidates of the other village.
• Same is true with number of meetings held
with each village (r=0.89)
Accounting for endogeneity of elections
Road project IV estimation
(IVs: rules on candidate approval during nomination; number of meetings for
organizing village election attended by upper-level government officials)
Direct election of
village leader
(yes=1,no=0)
F test value of the
instrument
variable
All Funding
source project
Village funding project
Other controls
0.86[0.15]
1.30[0.07]
(1.79)*
(5.07) ***
Geography,
Society and
economic factors
17.65
16.45
Marginal effect in bracket; Absolute value of t statistics in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%
Accounting for endogeneity of elections
Irrigation project IV estimation
(IVs: rules on candidate approval during nomination; number of meetings for
organizing village election attended by upper-level government officials)
Direct election of
village leader
(yes=1,no=0)
F test value of the
instrument
variable
All Funding
source project
Village funding project
Other controls
1.22[0.1]
0.35[0.02]
(5.05)***
(0.47)
Geography,
Society and
economic factors
33.37
30.68
Marginal effect in bracket; Absolute value of t statistics in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%
The multivariable analyses (OLS)
Total investment = a0 + a1*D-election + a2*other factors + e
The way village leader acceded to office
(0=appointed, 1=elected directly)
Yearly public goods projects in latest term
Village
All sources
Yearly public goods investment in latest term (Log)
Above
Co-funded
funded
Village
All sources
funded
Above
Co-funded
funded
funded
0.09
0.06
0.01
0.02
0.77
0.39
0.13
0.25
(4.24)***
(5.20)***
(0.78)
(1.66)*
(4.98)***
(4.20)***
(1.10)
(2.56)**
Tax-for-Fee reform
-0.34
-0.06
-0.17
-0.11
-3.50
-0.77
-1.70
-1.40
(after=1,before=0)
(11.96)***
(3.96)***
(8.71)***
(7.51)***
(16.64)***
(6.17)***
(10.62)***
(10.53)***
Target factors
net per capita income in 1997(1000 yuan)
Percentage of minority
population in 1997
Hilly land over 25 degree in total
and area in the village in 1997
Total population in 1997(1000 ren)
The illiterate rate of village
abor force in 1997
-9.19
51.31
-38.59
-21.91
-29.05
461.89
-290.40
-213.38
(0.75)
(7.67)***
(4.66)***
(3.45)***
(0.32)
(8.65)***
(4.26)***
(3.77)***
-0.0002
-0.00001
-0.0003
0.00004
0.001
-0.001
0.002
0.0004
(0.66)
(0.05)
(1.10)
(0.22)
(0.30)
(0.78)
(0.74)
(0.25)
0.0006
0.0001
0.0002
0.0004
0.01
0.0001
0.0005
0.0060
(1.67)*
(0.27)
(0.82)
(1.87)*
(2.43)**
(0.03)
(0.22)
(3.50)***
37.3
5.3
21.5
10.5
278.5
65.3
258.0
92.1
(3.51)***
(0.91)
(2.99)***
(1.90)*
(3.58)***
(1.41)
(4.36)***
(1.87)*
-0.03
-0.05
-0.01
0.02
-0.4
-0.5
-0.01
0.4
(0.58)
(1.47)
(0.20)
(0.69)
(1.02)
(1.92)*
(0.04)
(1.64)
-0.010
The distance between village committee
-0.001
0.001
-0.0004
-0.001
-0.008
0.004
0.010
and township seat in 1997
(0.67)
(0.70)
(0.37)
(1.55)
(0.61)
(0.52)
(1.01)
(1.21)
The distance of the nearest road
-0.002
-0.002
0.0001
-0.0001
-0.01
-0.01
-0.003
-0.005
(2.05)**
(3.70)***
(0.15)
(0.25)
(2.06)**
(3.76)***
(0.64)
(1.37)
o the village seat in 1997
Demand factors
The number of village/group
enterprise in 1997
Per capita land in 1997 (mu)
Percentage of effectively
0.02
0.02
0.0005
-0.0002
0.11
0.16
0.03
-0.003
(2.94)***
(5.32)***
(0.10)
(0.05)
(2.04)**
(4.74)***
(0.81)
(0.09)
-0.010
-0.006
-0.003
-0.001
-0.02
-0.03
-0.03
0.03
(2.06)**
(2.19)**
(0.97)
(0.38)
(0.63)
(1.70)*
(1.02)
(1.16)
-0.001
-0.0001
-0.0002
-0.001
-0.003
0.001
0.001
-0.006
(3.96)***
(0.44)
(0.72)
(6.21)***
(1.33)
(0.94)
(0.32)
(4.13)***
The farthest distance between two small
0.005
-0.001
0.002
0.003
0.02
0.002
0.01
0.02
groups within this village in 1997
(1.53)
(0.52)
(1.15)
(1.99)**
(0.89)
(0.12)
(0.49)
(1.59)
rrigated land in 1997
Other Factors
Number of fellow villagers working in
ownship or county governments
Age of village leader
Education of village leader
Prior occupation of village leader
(pure farmer=1,others=0)
The time of the village leader
acceded to office
0.005
0.000
0.004
0.001
0.03
0.001
0.028
0.008
(4.19)***
(0.40)
(4.88)***
(1.29)
(3.53)***
(0.27)
(4.17)***
(1.35)
0.001
0.002
-0.001
0.001
0.02
0.02
0.003
0.01
(1.22)
(2.88)***
(1.15)
(0.82)
(1.92)*
(3.26)***
(0.44)
(1.70)*
0.02
0.02
-0.01
0.01
0.04
0.21
-0.10
0.07
(1.52)
(3.56)***
(1.63)
(1.31)
(0.43)
(3.83)***
(1.45)
(1.11)
-0.04
-0.03
-0.02
0.01
-0.33
-0.20
-0.23
0.12
(2.12)**
(2.54)**
(1.81)*
(0.95)
(2.48)**
(2.50)**
(2.22)**
(1.47)
0.07
0.01
0.03
0.03
0.44
0.06
0.19
0.26
(15.82)***
(3.45)***
(10.11)***
(13.67)***
(14.50)***
(3.25)***
(8.32)***
(13.60)***
Results from the multivariate analyses
(Focusing on the coefficient of the direct election dummy
variable)
Yearly public goods projects in latest Yearly public goods investment in
term
latest term (Log)
All
Village
Above
All
Village
Above
Co-funded
Co-funded
sources funded
funded sources funded
funded
Ols
0.09
0.06
(4.24)*** (5.20)***
Fixed
effect
0.16
0.04
0.01
(0.78)
0.08
0.02
0.77
0.39
(1.66)* (4.98)*** (4.20)***
0.04
0.95
0.26
0.13
0.25
(1.10)
(2.56)**
0.63
0.36
(4.67)*** (2.55)** (3.38)*** (2.40)** (3.53)*** (1.79)* (3.26)*** (2.25)**
Fixed Effects Analysis
Investment = a0 + a1*D-election + a2*other factors + Village effects
Yearly public goods projects in latest term
Village
Above
All sources
Co-funded
funded
funded
The way village leader
0.16
0.04
0.08
0.04
acceded to office
(0=appointed,1=electe (4.67)*** (2.55)** (3.38)*** (2.40)**
d directly)
Tax-for-Fee reform
-0.30
-0.05
-0.14
-0.11
(after=1,before=0)
Age of village leader
Education of village
leader
Yearly public goods investment in latest term (Log)
Village
All sources
Co-funded Above funded
funded
0.95
0.26
0.63
0.36
(3.53)***
(1.79)*
(3.26)***
(2.25)**
-3.41
-0.70
-1.60
-1.39
(9.63)***
-0.01
(10.20)***
0.01
(9.99)***
-0.001
(3.36)***
0.001
(7.12)***
-0.003
(7.31)*** (14.75)*** (5.65)***
0.0005
0.00
0.02
(0.44)
0.005
(1.59)
-0.005
(2.23)**
-0.005
(0.51)
0.015
(0.33)
-0.07
(2.54)**
-0.05
(1.22)
-0.10
(0.82)
0.10
(0.20)
-0.027
(0.46)
-0.0005
(0.36)
-0.033
(1.34)
0.006
(0.41)
-0.36
(0.56)
-0.11
(0.80)
-0.22
(0.99)
-0.01
(1.40)
(0.34)
(1.32)
(0.78)
(1.11)
(0.04)
0.03
0.03
0.46
0.07
0.19
0.27
Prior occupation of
village leader
(pure
(0.77)
(0.03)
farmer=1,others=0)
The time of the village
0.06
0.01
leader
acceded to office
(15.33)*** (4.29)***
Constant
-129.1
-17.6
(9.42)*** (14.08)*** (14.40)*** (4.22)***
-52.1
-59.3
-925.5
-148.0
(8.20)***
-379.2
(14.24)***
-542.9
(15.29)*** (4.27)***
7041
7041
(9.36)*** (14.08)*** (14.43)*** (4.33)***
7041
7041
7041
7041
(8.23)***
7041
(14.36)***
7041
Observations
Number of village id
2448
2448
2448
2448
2448
2448
2448
2448
R-squared
0.06
0.01
0.03
0.05
0.06
0.01
0.03
0.05
Results from the multivariate analyses
(Focusing on the coefficient of the direct election dummy
variable)
Yearly public goods projects in latest Yearly public goods investment in
term
latest term (Log)
All
Village
Above
All
Village
Above
Co-funded
Co-funded
sources funded
funded sources funded
funded
Ols
0.09
0.06
(4.24)*** (5.20)***
Fixed
effect
0.16
0.04
0.01
(0.78)
0.08
0.02
0.77
0.39
(1.66)* (4.98)*** (4.20)***
0.04
0.95
0.26
0.13
0.25
(1.10)
(2.56)**
0.63
0.36
(4.67)*** (2.55)** (3.38)*** (2.40)** (3.53)*** (1.79)* (3.26)*** (2.25)**
IV-GMM Analysis
Investment = a0 + a1*D-election + a2*other factors
(IV=lag value of D-election)
Yearly public goods projects in
latest term
Yearly public goods investment in
latest term (Log)
All
Village Co-fund Above
All
Village Co-fund Above
sources funded
ed
funded sources funded
ed
funded
The way village leader
acceded to office
(0=appointed,1=elected
directly)
Observations
N=4587 not 7041
0.11
0.02
(1.80)* (1.74)*
4587
4587
0.06
0.03
(1.59)
(0.86)
4587
4587
0.62
0.15
0.64
(1.81)* (1.59)* (1.91)*
4587
4587
4587
0.26
(0.96)
4587
What is the mechanism?
Why do elected leaders invest more?
May be a “Reverse Relationship”
between investments and re-election
# of projects/year
1
0.8
0.6
0.4
0.2
0
Re-elected
Not
Reelected
What happens after an elected term?
• Reformulate the sample:
Made into party
secretaries
– Throw away all leaders that
were appointed
Did not rerun or
– 4 categories
resigned
– Keep leaders that were
early
elected (n=5000 or so)
Incumbents
that won
• Don’t use sample of
village leaders that “turned
into” party secretaries
• Don’t use elected village
leaders that decided not to
run / resigned early
– Left with two groups:
• Incumbents that ran and
won
• Incumbents that ran and
lost
Incumbents
that lost
All leaders that were elected
What happens after an elected term?
• Reformulate the
sample:
Left with two groups:
Did not rerun or
resigned
early
Incumbents
that won
• Incumbents that ran
and won (n=3100)
• Incumbents that ran
and lost (n=800)
N = about 3900
Incumbents
that lost
All leaders that were elected
Reverse Regression (fixed effects)
Reelected = a0 + a1*number of projects + a2*other factors +
V+e
Table 11 The yearly public projects number (investment level) in latest tem and the re-election
(re-appointment) of village leader
Re-election (1=yes,0=no)
Public projects number
Re-appointment (1=yes,0=no)
0.05
0.06
-0.02
-0.07
(2.54)**
(2.30)**
(0.54)
(1.36)
Above funded public projects number
Village funded public projects number
Co-funded public projects number
Public investment level
0.03
0.14
(0.47)
(0.88)
0.15
-0.42
(2.59)***
(3.49)***
0.07
-0.06
(1.82)*
(0.72)
0.003
0.003
(1.42)
(0.68)
Average investment level per project
-0.002
0.009
(0.66)
(1.44)
Average investment level of above
0.004
-0.001
funded public projects
(0.92)
(0.10)
Average investment level of village
funded public projects
-0.01
0.02
(1.94)*
(2.20)**
Average investment level of co-funded
-0.005
0.01
public projects
(1.47)
(1.51)
Age of incumbent
-1.43
-1.43
-1.43
-1.42
-0.75
(14.67)*** (14.70)*** (14.68)*** (14.55)*** (4.35)***
Education of incumbent
Year Dummy
Constant
-0.74
-0.75
-0.72
(4.29)***
(4.37)***
(4.23)***
0.03
0.04
0.02
0.02
0.45
0.46
0.44
0.45
(0.18)
(0.22)
(0.16)
(0.14)
(1.34)
(1.36)
(1.31)
(1.34)
yes
yes
yes
yes
yes
yes
yes
yes
5.57
5.61
5.57
5.49
3.03
2.98
3.10
3.13
(4.33)***
(4.11)***
(4.24)***
Observations
(10.63)*** (10.69)*** (10.61)*** (10.44)*** (4.15)***
3718
3718
3718
3718
875
875
875
875
R-squared
0.19
0.18
0.19
0.19
0.13
0.13
0.14
0.18
The evidence of election incentive 2
(Focus on public investment with village fixed effect)
Re-election (1=yes,0=no)
Public projects number
Above funded public projects number
Village funded public projects number
Co-funded public projects number
Public investment level
0.05
(2.54)**
0.06
(2.30)**
More public goods
project, more chance of
Re-election
0.03
(0.47)
0.15
(2.59)***
0.07
(1.82)*
0.003
(1.42)
Average investment level per project
-0.002
(0.66)
Average investment level of above
0.004
funded public projects
(0.92)
Average investment level of village
funded public projects
-0.01
(1.94)*
Average investment level of co-funded
-0.005
public projects
(1.47)
Is there the same effect when village
leader were appointment?
Public projects number
-0.02
(0.54)
Re-appointment (1=yes,0=no)
-0.07
(1.36)
Above funded public projects number
0.14
(0.88)
-0.42
(3.49)***
-0.06
(0.72)
Village funded public projects number
Co-funded public projects number
Public investment level
Average investment level per project
0.003
(0.68)
0.009
(1.44)
Average investment level of above
funded public projects
Average investment level of village
funded public projects
Average investment level of co-funded
public projects
Nothing is significant when leader is
reappointed … therefore his performance in the
village does not matter
-0.001
(0.10)
0.02
(2.20)**
0.01
(1.51)
Summary
• Elections in a number of different regressions …
using types of investments … sources of
investment … and total investment (# of projects
/ investment levels) … PLUS different estimating
strategies (IV / FE / GMM)  find positive
impact
• Mechanism: Pork barrel politics, with Chinese
characteristics
Support for the literature
• Rosenzweig and Foster (2003)
– Will democracy (election of local leaders through an popular vote-based
election) help the poor … by giving them a bigger voice?
– Governance structure  type of investment using village data
– Worry about endogeneity: use log of village population + dummy
variables indicating the passage of legislation relevant to local
governance
• Chattopadhyay and Duflo (2004)
– Will election rules (e.g., reservation seats for women) affect investment
into type of investment that women want …
– Randomly assigned reservations … natural field experiment using village
level data
• Besley and Burgess (2001) n.a.
– Openness  more concentrated effort on helping the poor (using state
level data) … newspaper circulation …
– Worry about endogeneity of newspapers … IV: media ownership
concentration
Plus LOTS more to do
• Using new IV (county level protocols)
• Using “quality of election” and protocols to
see whether or not the type of election
matter …
• Looking at other aspects of governance
(how about the Communist Party Cell?)
• Effect of elections on “quality of
investments”
Thank you