Workfare as an Effective Way to Fight Poverty: The Case of India’s NREGS

Highlights

• We study the impact of India’s workfare scheme (NREGS) on poor rural households.

• Panel data and exogenous household selection for work form the basis of analysis.

• We find NREGS significantly increased expenditure on food and non food consumables.

• NREGS also increased food security by significantly reducing meals forgone per week.

• The probability of holding savings also went up for households under NREGS.

Summary

This paper analyzes the impact of India’s National Rural Employment Guarantee Scheme (NREGS) on poor rural households. In particular, we study the impact of the program on food security, savings, and health outcomes. We have a panel data of 1,064 households from 198 villages of Andhra Pradesh, over two years. In the early stage of the program, several households that applied for work were denied employment due to shortage of work. We exploit this exogenous variation to calculate triple-difference estimates of the impact of the program. Our results indicate that the NREGS significantly increased the monthly per capita expenditure on food and non-food consumables. The program also improved food security by a significant reduction in the number of meals foregone by households per week. The program raised the probability of holding savings and reduced the incidence of depression among rural households.

Introduction

This paper measures the welfare impact of the National Rural Employment Guarantee Scheme (NREGS) on the participating households. In doing so, it aims to contribute to the empirical literature on the impact of workfare schemes in low-income countries (Devereux & Solomon, 2006 and Subbarao, 1997, Subbarao, 2003 provide good overviews). We have a panel data of 1064 ultra-poor rural households from 198 villages of the Medak district, in Andhra Pradesh, over two years. In particular, we compare over time, the difference in average outcomes of households that applied for and received jobs under NREGS with the average outcomes of households that applied for and did not receive jobs. The main reason reported for denying applicants was lack of worksites or existing worksites with limited jobs available. Incidence of rejection was higher in the early years of NREGS roll-out and in the subsequent years this reduced considerably as the program scaled up in size and scope. We exploit this exogenous variation in participation by households to calculate triple-difference estimates of NREGS impact on extremely poor rural households. In particular, we proceed in two steps. In the first step, we estimate the difference in difference of average outcomes over time, of households that applied for jobs and received them under the scheme with households that applied for jobs but were denied due to shortage of work. In the second step, we refine this by calculating a triple-difference estimate, by further differencing out, from both these group of households, the average outcomes of households that are matched based on propensity scores.

Unlike a conventional difference-in-difference set up, in our analysis the NREGS was already in operation when the first survey (baseline) was collected in 2007. This means that instead of the treatment effect of the program introduction, the comparison of post-intervention data results in the estimate of the longer term impact of the program on participating households. In other words, with the double-difference estimator, we cannot estimate a possible first-stage shift in outcome brought about by the introduction of the program but we can track differences in the growth rates in the outcomes afterward. Our empirical strategy is, therefore, to compare the households that demanded and received employment with households which applied for work but was denied due to shortage of available work.

While developed countries increasingly lean on workfare programs¹ as a means to reduce work disincentives provoked by their far reaching social security systems, the concept of cash-for-work has gained importance in less developed countries as well.²

Looking back to a long history of food-for-work programs in times of economic distress, developing countries increasingly run public works programs not only to better target benefits to the poor but also to use the emerging labor force to build up the rural economic infrastructure. In the fore of that development is India where the ambitious NREGS came into force in 2006. Since then, each rural household is guaranteed 100 days of unskilled wage employment per year.

The basic merit of a workfare program as a means for direct transfers to the most needy is that it helps in targeting the beneficiaries. It is widely accepted that the non-poor are less likely to accept low wages and heavy labor to obtain benefits (Dev, 1995, Haddad and Adato, 2001, Mujeri, 2002, Webb, 1995). There have been recent studies that explore the impact of NREGS on different outcomes. Azam (2012) explores the positive effect of the program on agricultural wages, Imbert and Papp (2013) shows that the indirect benefits of the program in the form of higher private earnings are almost of the same order as the direct benefits in the form of NREGS wages. Afridi et al. (2013) show that rising women’s participation in the NREGS improves children’s educational outcomes. The study that comes closest to ours is Liu and Deininger (2010) which shows that NREGS resulted in a significant and positive impact on consumption expenditure, intake of energy and protein, and asset accumulation.

Our study is distinct from the existing literature in that we assess the impact of the NREGS on extreme poverty as measured by food security, financial inclusion, and health outcomes. Food security is a fundamental concern of the poor and we try to capture this through the monthly per capita consumption expenditure and the number of meals foregone by the household as a whole at the member level. Previous literature on food security (Folke Larsen & Bie Lilleør, 2014) has used “Household Hunger Scale” which is based on multiple question asking whether due to lack of resources anyone in the household (1) went to sleep at night hunger, (2) has no food to eat of any kind in the household and (3) went a whole day and night without eating. We use similar metrics to gauge food security of household members in addition to monthly per capita consumption expenditure.

To assess extreme poverty through the lens of financial wellbeing, we analyze a poor household’s probability of holding any savings as well as the amount of reported savings. The third set of outcomes of interest are health measures which include physical and mental health. While the relationship between physical health and poverty are well established in the literature (Banerjee & Duflo, 2011), the link between mental health and poverty is being proposed by a recent but growing literature which includes Banerjee, 2000, Bertrand et al., 2004 and Mullainathan and Shar (2010). These works explore the link between poverty and mental strain which affects behavioral decision making which includes desperation, entrepreneurial ability, and performance.

Our main results indicate that the NREGS had significant impacts on extreme poverty within the first few years of implementation. The participation in the workfare program improved sharply over the initial years, and our analysis shows significant increase in the monthly per capita expenditure on food by Rs. 25.8 (9.6%) and on non-food consumable by Rs. 11.17 (23%). The program also improved food security by a significant reduction in the number of meals foregone by households per week. The program raised the probability of holding savings for a poor rural household by 21% and the per capita amount saved increased by Rs. 18.6. The health outcomes impacted by the program include a significant reduction of 12% in the incidence of reported depression. Other indicators of mental health have also shown significant improvements over time. There were no significant impacts on physical health outcomes that we measured.

The rest of this paper is structured in the following way: Section 2 outlines the context for this study, where we describe India’s poverty and the rationale for the NREGS, Section 3 discusses the empirical strategy we use for the analysis; Section 4 has details on data where subsection 4(a) outlines the sampling methodology used to generate a sample of 1064 ultra-poor households from 198 villages in Medak, subsection 4(b) has a discussion on attrition in the sample over time, subsection 4(c) describes the nature of job rationing in our sample which is crucial for this study and subsection 4(b) describes the summary statistics of the data. Section 5 has results and Section 6 concludes.