IVs and Democracy and Development Lecture 2

1
IVs and Democracy and DevelopmentLecture 2

• Mistake should have read Luo et al. (2007)
  instead of Li and Rozelle (that is for Monday
• Mondays assignment (Fixed Effects)
• Shaban
• Li and Rozelle
• Make up Luo et al. (2007)

2
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)

3
Basic Notation

• Y a0 a1X a2Z 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

4
Basic Notation

• Y a0 a1X-hat a2Z e
• second stage equation
• X b0 b2Z ?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 ? ? 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

5
Go to Angrist and Krueger
6
Village Governance and Transforming Rural Chinas
Communities

• Scott Rozelle, Stanford University
• Linxiu Zhang, Jikun Huang, Renfu Luo, Chengfang
  Liu, CCAP

7
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)?
8
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

9
Our conceptual framework

• Following Rosenzweig and Foster, we believe that
  elections will affect the spending of Chinas
  village leadership
• from expenditures that benefit the elite (qingke
  chifan salaries)
• to expenditures that benefit the villages
  (infrastructure)

10
Farmer Believing there has been Improvements in
Infrastructure since 1998 (from PRA focus
groups)(??????)
Percent of farmers
11
Farmer Dis-satisfaction with Current Status of
Infrastructure(from PRA focus groups Surveys)
Percent of farmers saying not satisfied
12
Also willing to spend resources
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

• Question 1
• If your village received a grant of 50,000 yuan,
  what would you spend it on?
• Answer 85 said roads, irrigation or drinking
  water

a. infrastructure b. environmental
preservation c. welfare payments d. investing in
the leadership office facilities
13
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

14
Objectives of Presentation

• Understand the contours of investment in Chinas
  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?

15
Plan for Talk

• Getting the facts right! collect data
• Measure differences across communities (quantity
  and quality)
• A sketch of Chinas 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?

16
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
17
2000 village survey Distribution of sample
counties in China, 2003 Rural China Investment
Survey Total number of villages 2459
Jilin
Hebei
Gansu
Jiangsu
Shaanxi
Sichuan
18
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

19
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

20
Creating the data set
6/7 years of data
Observation 2 During 3 year election cycle, how
many projects were implement During 3 year
election cycle was leader elected? yes or no
Observation 1 During 3 year election cycle, how
many projects were implement During 3 year
election cycle was leader elected? yes or no
Also move to use total investment
N7041
21
Creating the data set
6/7 years of data
Observation 1 During 3 year election cycle, how
many projects were implement/year During 3 year
election cycle was leader elected? yes or no
Observation 2 During 1 year of term, how many
projects were implement/year During 3 year
election cycle was leader elected? yes or no
Observation 3 During last 2 year of term, how
many projects were implement/year During 3 year
election cycle was leader elected? yes or no
Obs. 4
IV does not change over time though it is
constant across towns
22

• What are the trends of Chinas public investments
  into its villages over time and across space?

23
Composition of investment projectspublic vs.
development
Total number of projects in sample
Development Projects
9,138 projects (in 2459 sample villages)
Most of projects in rural China now focus on the
provision of public goods different than the
1980s
Public goods investments
Nearly 4 per village during the study period or
almost 1 per year
24
Scope of public goods investment projects
25
Type of PUBLIC GOODS projects
Around 2/3rds of public goods investments into 5
types of projects
11
26
Funding sources of public goods investment
projects in rural China
Percent of Number of Public Goods Projects
27
Percent of Total Investment Amount in Public
Goods by Source
Person days per project
47
53
28
Rising OvertimePublic Goods Investment Projects
Projects per year
29
Rising But Still Not Enough!Total Spending
China versus Japan/Korea
Total Inv./Capita (US dollars, PPP terms
30
Distribution of public goods projects, number of
projects and amount of investment
Units number of projects
Units 10K yuan
Mean size of investment project about 50,000
yuan
Mean number of project 3 to 4 for sample period
While some villages have lots of projects
others do not!
31
Determinants of Investment Quantity

• Determinants of Quantity (number of projects /
  amount of investmentyuan)
• Many factors
• Village locations
• Structure of village economy
• Ethnic Make up
• Policies (e.g., Tax Reform)
• Village governance reform

Focus of the paper
32
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

33
(No Transcript)
34
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

35
Villages that Directly Elect Village Leaders
is growing over time
Percent
83
69
36
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

37
What is going on?Chinas rural election
Though the election have been implemented for
many years, there are still villages that do not
elect their village leader directly
38
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

39
Quite decentralized (how they run elections?)
But there are factors that affect the nature of
the elections

• Election committee governance
• (in gt80 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 wont have elections
  they will only nominate one person the township
  will appoint a leader the towns only let one
  person run for village leader

40

• So is there a relationship between elections and
  public goods investment?
• -- Descriptive
• -- Multivariate

41
Relationship between election of village leader
and total investment
42
Investment Projects and Elections
Total investment
Investment from village itself
Investment (Thousand Yuan)
Village funded investment (Thousand Yuan)
E A
Yes
Yes
No
No
E A
Yes
E A
E A
E A
E A
Yes
No
Yes
No
Yes
No
Yes
No
Road
Drinking Water
Irri.
Drinking Water
Road
Irri.
43
Determinants of Investment Quantity

• Determinants of Quantity (number of projects /
  amount of investmentyuan)
• From paper (by type by source
• using OLS IV)
• From new work (total investment by source
  using Fixed Effect GMM)

44
Road project Probit (Tobit) analysisInvestment
a0 a1D-election a2other factorsProvincial
fixed effects
45
Road project Probit (Tobit) analysis(Focusing on
the D-election coefficient)
46
Irrigation project Probit (Tobit)
analysisInvestment a0 a1D-election
a2other factorsProvincial fixed effects
47
Irrigation project Probit (Tobit) analysis(Focus
on the D-election)
48
Multivariate Analysis Elections and Investment
Quality in Chinas Villages
Dependent variables
Election variable






Roads




Election variable
irrigation
drinking water
Election variable






All projects



49
Problem with OLS regression

• Unobserved heterogeneity
• Need IV (use both 2SLS and bivariate probit)
• Investment a0 a1election a2other factors
  e
• Election b0 b1IV b2other
  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 villages
  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

50
RHS Controls and IVs Correlations?
51
Over-identification test and hausman test
52
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 (r0.89)

53
Accounting for endogeneity of electionsRoad
project IV estimation(IVs rules on candidate
approval during nomination number of meetings
for organizing village election attended by
upper-level government officials)
54
Accounting for endogeneity of electionsIrrigation
project IV estimation(IVs rules on candidate
approval during nomination number of meetings
for organizing village election attended by
upper-level government officials)
55
The multivariable analyses (OLS) Total
investment a0 a1D-election a2other
factors e
56
Results from the multivariate analyses(Focusing
on the coefficient of the direct election dummy
variable)
57
Fixed Effects Analysis Investment a0
a1D-election a2other factors Village effects
58
Results from the multivariate analyses(Focusing
on the coefficient of the direct election dummy
variable)
59
IV-GMM Analysis Investment a0 a1D-election
a2other factors (IVlag value of D-election)
N4587 not 7041
60
What is the mechanism?Why do elected leaders
invest more?
61
May be a Reverse Relationshipbetween
investments and re-election
62
What happens after an elected term?

• Reformulate the sample
• Throw away all leaders that were appointed
• 4 categories
• Keep leaders that were elected (n5000 or so)
• Dont use sample of village leaders that turned
  into party secretaries
• Dont 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

Made into party secretaries
Did not rerun or resigned early
Incumbents that won
Incumbents that lost
All leaders that were elected
63
What happens after an elected term?

• Reformulate the sample
• Left with two groups
• Incumbents that ran and won (n3100)
• Incumbents that ran and lost (n800)
• N about 3900

Did not rerun or resigned early
Incumbents that won
Incumbents that lost
All leaders that were elected
64
Reverse Regression (fixed effects)Reelected a0
a1number of projects a2other factors V e
65
The evidence of election incentive 2(Focus on
public investment with village fixed effect)
More public goods project, more chance of
Re-election
66
Is there the same effect when village leader were
appointment?
Nothing is significant when leader is reappointed
therefore his performance in the village does
not matter
67
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

68
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)
• Openness ? more concentrated effort on helping
  the poor (using state level data) newspaper
  circulation
• Worry about endogeneity of newspapers IV media
  ownership concentration

n.a.
69
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

70

• Thank you