1. Employment and Earnings in Rural India: 2004-2012
Shantanu Khanna, Deepti Goel and René Morissette 
NOPOOR POLICY SEMINAR, March 11, 2016

2. What do we Study?
Document the change in rural earnings distribution between 2004/05 to 2011/12
Study EARNINGS and not wage rates
Focus on change in inequality measures over this period

3. Motivation
Studies on earnings in urban India (Kijima 2006; Azam 2012) and on consumption expenditure (Sen and Himanshu 2004; Cain et al 2010; Motiram and Vakulabharanam 2013; Subramanian and Jayaraj 2015) We don’t know of similar analysis for rural earnings
Document changes at various percentiles: Other studies have found that inequality has risen because of changes in the top end
Methodology gives factor contribution in explaining the change over time at the unconditional quantiles

4. Preview: Earnings in Rural India 2004/05 to 2011/12
Real Earnings increased at all percentiles
Sharp fall in earnings inequality
Gini: to 0.395
Much of the drop is coming from changes in the top half of the earnings distribution
Decline in inequality is explained by changes in returns to characteristics
Had returns to characteristics remained fixed, inequality would have risen
Education has an inequality increasing effect

5. Data
Employment Unemployment Surveys of the National Sample Survey Organisation
2004/05 (61st round) and 2011/12 (68th round)
Analysis sample: Wage earners, in the age group, in 23 major states of rural India
Wage earners make up a little less than 30% of adult population
44,500 in 2004/05 and 36,000 in 2011/12
Dropped 0.1 percent of observations at both ends

6. Methodology
Fortin, N. M., T. Lemieux, and S. Firpo. 2011, Decomposition Methods in Economics, Chapter in the Handbook of Labor Economics
Concept of Unconditional Quantile Regression
Unconditional quantile regressions (UQR, Firpo et al., 2009) help us examine the marginal effects of covariates on the unconditional quantiles of an outcome variable.
Recentered Influence Function Decomposition based on above
Allows detailed decomposition of Structure and Compositions Effects at unconditional quantiles
Closest in spirit to the Oaxaca-Blinder Method for decomposition at the mean

7. Kernel Density Estimates of Log of Real Weekly Earnings; Rightward shift; More Compressed;
Median earnings increased from 265 to 458 rupees per week (from 2.5 times to 4 times the 2004/05 poverty line)

8. Absolute Earnings increase smaller at lower percentiles:
Real Weekly Earnings (in rupees) at each percentile
Absolute Earnings increase smaller at lower percentiles:
98, 193 and 307 at first, median and ninth deciles.

9. Change in LOG Real Weekly Earnings (in rupees) at each percentile
Percentage increase: 89, 72 and 44 percent at 1st, 5th and 9th deciles. Shows inequality decreased over the seven year period. 25/10 ratio of raw earnings not much change, 90/75 ratio sharp drop.

10. Decomposition: Accounting Exercise Not a Causal Analysis
Total Change in any distributional statistic broken into two parts
Structure Effect: Due to changes in Returns to Characteristics
Composition Effect: Due to changes in Distribution of Characteristics

11. Complete List of Characteristics we look at
Age
Male: 70% to 75%
Married
Muslim
Caste
Education: Illiterate- 47% to 36%
Industry: Agriculture 61 to 49% & Manufacturing 21 to 33%
Occupation
State of Residence

12. Unconditional Quantile Regressions: Male, 2011/12

13. Unconditional Quantile Regressions: College & Beyond, 2011/12

14. Re-centered Influence Functions Aggregate Decomposition Results
Based on characteristics 2004/05 and returns 2011/12; both have sig. positive contri in overall increase at each percentile except at top; structure eff dominant; Ineq decreased mainly coz returns to charac increased much more for lower percentiles than for higher. If returns unchanged, inequality would have increased.

15. Aggregate Decomposition of Total Change Structure Effect -0.393 -0.095
Decomposition of Changes in Inequality Measures from 2004/05 to 2011/12
90-10
50-10
90-50
Gini
Value in 2004/05
1.855
0.881
0.974
0.461
Value in 2011/12
1.575
0.792
0.783
0.395
Total Change
-0.280
-0.089
-0.191
-0.066
Aggregate Decomposition of Total Change
Structure Effect
-0.393
-0.095
-0.298
Composition Effect
0.113
0.006
0.107
0.029
Detailed Decomposition of the Composition Effect
Education
0.098
0.009
0.090
0.024
Experience
0.027
0.001
0.008
Male
-0.018
-0.013
-0.005
-0.006
Married
0.003
States
0.002
Muslim
0.000
Caste
-0.002
-0.001

16. Rural Real Earnings, 2004/05 to 2011/12
Increase in real earnings at all percentiles suggests that economic growth has been accompanied by decline in poverty
In rupee terms was higher at higher percentiles
In percentage terms was higher at lower percentiles
Sharp fall in earnings inequality in rural India
Gini of in 2004/05; and in 2011/12 (rural)
Gini of in 2004/05; and in 2011/12 (urban)
At the aggregate level, decline in inequality explained by changes in returns to characteristics; if these returns had not changed, inequality would have risen on account of changes in the distribution of characteristics
Education has a significant inequality increasing effect
Gini : Brazil-53.9 (2009); Russia-39.7 (2009); India-33.9 (2009); China-42.1 (2010) and South Africa-63 (2008).

17. Concluding Remarks Rural and Urban inequality dynamics are different
Our work shows that earnings inequality declined in recent times in rural, while it remained stagnant in urban
To inform policy, the next step should be to try to unearth the mechanisms that explain this decline
Our work shows that it is not the overall change in characteristics but the change in returns to these characteristics that is explaining the decline in inequality
Education has an inequality increasing effect
Feminization of workforce also has an inequality increasing effect
Perhaps, measures such as MGNREGA could be contributing to the decline (conjecture)