2. Background
Debate in policy circles about increase in income
inequality, and a fall in poverty ratio.
“Talk of inclusive growth, rich getting richer, faster:
Report,” 01 February 2011, Economic Times.
4. Literature
Debate on Beta and Sigma Convergence
Conflicting evidence in the literature whether
regional inequality has increased, or decreased
over time.
Cashin and Sahay (1996), Aiyer (2001), and
Purfield (2006), find evidence in favor of
convergence.
Rao et al. (1999), Bajpai and Sachs (1996), and
Kurian (2000) find evidence in favor of divergence.
5. Literature (Cont.)
Impact of reforms on income inequality
About, Ahluwalia (2002) finds it has fallen.
Bhattacharya and Sakthivel (2004), find it has
increased.
All these studies mostly have used state level data,
and use growth accounting framework as
suggested in Barro and Sala-i-Martin (1992).
6. Agreement
Evidence of Sigma Convergence after reforms.
Mixed evidence about Beta Convergence.
Pockets of deprivation in progressive states.
Neighborhood effect is important.
7. Motivation
Twin Peaks Hypothesis: Quah’s (1996) and Jones (1997).
Variation in steady-state income level at a sub-state
(district) level.
OLS estimate are conditional average, and fail to capture
potential observational interaction across region.
Failure to capture such neighborhood effect can result in
major model misspecification (Anselin, 1988).
8. Why this Paper?
We use district-level data and examine inter-regional
income inequality.
We look at the factors that affect district-level income
distribution.
We quantify the neighborhood effect using spatial
econometric techniques.
9. Empirical Model
Is there any significant change in median adjusted
per-capita income density?
Factors responsible for district-level income growth.
Quantifying the own effect, direct and indirect
neighborhood effect?
10. Data
District income data (per-capita) covering 536
districts (out of 627 districts) in India.
Time period covered: 1999/2000, 2001/02, and
2004/2005.
11. Data (Cont.)
Development Indicators
No. of factories per one lakh popn.
No. of hospitals and dispensaries per one lakh popn.
Percentage of household with electricity connection,
availing banking services, drinking tap water, and
with close drainage system.
School enrollment as a percentage of total
population.
Gini Coefficient.
Number of Murder committed in 2001.
13. Evidence of Twin Peaks
Kernel Density Plots of Log Per capita Trade&Hotel
Kernel Density Plots of Log Per capita Communication
Median adjusted
0
0
.1
.2
Density
Density
.2
.3
.4
.4
.5
.6
Median adjusted
4
6
8
10
2
4
log(Rs)
6
8
log(Rs)
Per capita Tr&Hotel-1999
Per capita Tr&Hotel-2005
Per capita Communication-1999
Per capita Communication-2005
Kernel Density Plots of Log Per capita Service
Kernel Density Plots of Log Per capita Banking&Insurance
Median adjusted
0
0
.1
.2
Density
.2
Density
.4
.3
.6
.4
Median adjusted
2
4
6
log(Rs)
Per capita Banking&Insurance-1999
Per capita Banking&Insurance-2005
8
10
7
8
9
log(Rs)
Per capita Srv-1999
Per capita Srv-2005
10
11
14. Tests for significance in mean Income
1999/00 and
2004/05
(without Gujarat and
Delhi)
2001/02 and
2004/05
T-test of Mean Difference: Income
19.41 (0.00)a
16.08 (0.00)
T-test of Mean Difference: Log Income
23.22 (0.00)
22.11 (0.00)
Z-Value of sign test of median: Income
6.87 (0.00)
4.98 (0.00)
6.78 (0.00)
4.99 (0.00)
Z-Value of sign test of median: Log Income
a P-values
are in the parenthesis
15. Tests of Distributional Difference of median
adjusted Log Income
Test Statistics
Kolmogorov-Smirnov (KS)
one sided test statistics
a P-values
are in the parenthesis
1999/00 and 2004/05
(without Gujarat and Delhi)
0.042 (0.38)a
2001/02 and 2004/05
0.036 (0.48)
16. All India Moran Indices for income and growth
w y
n
Moran Index (I):
Iy : W
ij
i , j 1
i
y
y
j
y
trace(W 'W ) * y
Variable
Moran
Index (I)
t-statistics
'Per-capita Income 1999/2000'
0.54
19.74
'Per-capita Income 2001/02'
0.53
20.33
'Per-capita Income 2004/05'
0.48
18.51
'Per-capita Annualized Income Growth 1999/002004/05'
0.38
13.88
'Per-capita Annualized Income Growth 2000/012004/05'
0.26
10.05
18. Equation 1
Log income 2004/05
Dependent Variable
Equation 2
Log income 2001/02
Cross Equation Correlation
0.919
R-bar Square
0.6791
0.6850
No. observations, No. Variables
485,
485,
Variable
Coefficient
19
t-stat
19
Coefficient
t-stat
Constant
8.3647**
86.48
8.2919**
93.94
No. of factories total
0.0004*
2.38
0.0004*
2.44
Gini coefficient
0.6409*
2.31
0.6116*
2.42
Murder
0.0003
0.76
0.0003
0.98
0.003*
2.28
0.0031*
2.54
Closed drainage
0.0057**
2.95
0.0044*
2.52
School enrolment
0.009**
3.72
0.0085**
3.83
Hospitals and dispensaries
0.0034**
3.65
0.0031**
3.67
Banking services
0.0064**
2.92
0.0062**
3.12
Tap drinking water
0.0029**
2.81
0.0023*
2.41
W*No. of factories total
0.0002**
4.37
0.0002**
3.18
0.108
1.20
0.1064
1.30
0
-0.23
0
0.29
W*Electricity connection
0.0004
1.16
0.0008**
2.63
W*HH closed drainage
0.0003
0.40
-0.0002
-0.27
W*School enrolment
-0.0002
-0.40
-0.0002
-0.38
W*Hospitals and dispensaries
-0.0003
-0.80
-0.0003
-0.83
W*Banking services
-0.002**
-3.53
-0.0019**
-3.56
W*Tap drinking water
-0.0003
-1.32
-0.0004
-1.78
0.096*
10.48
0.094*
10.10
Electricity connection
W*Gini coefficient
W*Murder
ρ 1 , ρ2
*
Indicates the coefficient is significant at a 2.5 per cent level, and
**
indicates the coefficient is significant at a 1 per cent level.
19. Regression Results
In terms of direct own effect all the development
indicators, sans murder, affect income.
In terms of direct neighborhood effect, only
factory, electricity and bank, affect income.
Evidence about cross income correlation.
Indirect neighborhood effect is 10%.
21. Policy Analysis
A 1% increase in the number of factories increase
income for Urban Bangalore by 0.18 per cent
Similarly, a 1% increase in electrification, closed
drainage, school enrolment, banks, and drinking water,
resulted in an increase in per capita income by 0.30
per cent, 0.41 per cent, 0.25 per cent, 0.17 per cent,
and 0.27 per cent, respectively.
22. Conclusion
There is no evidence in support of emergence of
clusters: clustering of the rich-income districts, and
clustering of poor-income districts.
Private sector is taking initiative in moving to
districts with lesser input costs.
Income growth is spatially correlated.
Human capital, physical, and social infrastructure,
are significantly contributing to the Indian growth
story.