A Quantitative Analysis of Participation in MGREGA in Andhra Pradesh Doug Johnson (with Caroline Larouche, SankarNarayan, Deepak Saraswat, Viveh Shah, and Ajay Tannirkulam)
Overview • Andhra Pradesh has made all MGREGA muster rolls available to the public on AP MGREGA website • AP MREGA data was combined with census data, daily rainfall data, and local election data • All data which can be publicly shared is now available on CMF website
Why MGREGA Matters for Financial Inclusion …allows for easier… Financial Inclusion Delivery of Government Benefits …provides an opportunity for increasing…
MGREGA in Andhra Pradesh * Source: http://nrega.nic.in ** Number of rural households from 2001 census
Finding 1: Wages are pretty similar across castes and genders, but not across locations max difference = 2.66 Rs.
Finding 1: Wages are pretty similar across castes and genders, but not across locations max difference = 20.79 Rs.
Finding 2: The 100 days rule isn’t interpreted all that literally If the 100 days rule was interpreted literally, we would expect to see more households who have worked 100 days and less than those who have worked > 100 days each year. Distribution of Total MGREGA Days Worked per Household, Phase 1 Districts FY0708
Finding 2: The 100 days rule isn’t interpreted all that literally Further, administrators don’t seem to consider the number of other workers in a household in provision of work. Distribution of number of days worked per worker does is relatively similar for households with different number of MGREGA workers. Distribution of Total MGREGA Days Worked per Worker by Number of MGREGA Workers in Household, Phase 1 Districts FY0708
Finding 3: Getting a jobcard doesn’t appear to be a major hurdle If there was a fixed cost to participating in MGREGA (such as a facilitation fee for obtaining a jobcard) , we would expect to see few households working only one week. Distribution of Total MGREGA Weeks Worked, Phase 1 Districts FY0708
Finding 4: MGREGA helps participants cope with bad weather • A key goal of MGREGA is to provide basic income support in times of severe hardship • For rural poor, weather is a major source of risk. • Other attempts to reduce rural poor’s susceptibility to weather risk have had limited success. • If responsive to changes in demand, MGREGA may help participants cope with bad weather. • Yet it is not clear how responsive to changes in demand MGREGA really is. Even if responsive to changes in demand at individual level, it may not be responsive to changes in demand at village level or higher.
Finding 4: MGREGA helps participants cope with bad weather Average Total Rainfall by Month (Phase 1 Districts, 2001-2008) • Only interested in relationship between weather induced changes in income and participation in MGREGA. Yet weather may influence MGREGA other than through changes in income. • To prevent effect of weather on MGREGA other than through income from contaminating results, we look at relationship between rainy season weather and participation in MGREGA in following dry season. Total MGREGA Wages by Month (Phase 1 Districts, 2008) Dry season
Finding 4: MGREGA helps participants cope with bad weather • Participation in MGREGA in dry season strongly correlated with rainfall in previous rainy season • Change in wages explained by change in weather at the village level quite high. • Crude comparison of these figures with estimates of the impact of access to banking services on susceptibility to weather shocks by Jayachandran indicates that MGREGA at least as important as financial access in reducing impact of weather shocks. Distribution of Change in Wage per Worker Explained by Rainfall
Finding 5: Caste and gender of local leaders have little effect on participation • Potential drawback of locally administered programs is that there may be discrimination in provision of benefits • Can assess whether local leaders discriminate in providing work by comparing GPs where sarpanch seat is reserved for particular caste or for a woman with other GPs • Reservations of sarpanch seats to women are completely random. • Reservations of sarpanch seats to SC, ST, and BC are random conditional on the share of these groups in mandal population.
Finding 5: Caste and gender of local leaders have little effect on participation Reservations have only marginal effects. Reserving sarpanch seat for SC increases share of workers who are SC by 2.4 percentage points Reserving sarpanch seat for BC increases share of workers who are BC by 1.9 percentage points