Smartcards: Identifying General Equilibrium Effects from Randomized Controlled Trials

Transcription

1 Smartcards: Identifying General Equilibrium Effects from Randomized Controlled Trials Karthik Muralidharan UC San Diego Paul Niehaus UC San Diego February 28, 2014 Sandip Sukhtankar Dartmouth College JEL codes: D50, D73, H53, J38, J43, O18 Keywords: public programs, general equilibrium effects, rural labor markets, NREGA, India We are grateful to officials of the Government of Andhra Pradesh, including Reddy Subrahmanyam, Koppula Raju, Shamsher Singh Rawat, Raghunandan Rao, G Vijaya Laxmi, AVV Prasad, Kuberan Selvaraj, Sanju, Kalyan Rao, and Madhavi Rani; as well as Gulzar Natarajan for their continuous support of the Andhra Pradesh Smartcard Study. We are also grateful to officials of the Unique Identification Authority of India (UIDAI) including Nandan Nilekani, Ram Sevak Sharma, and R Srikar for their support. We thank Tata Consultancy Services (TCS) and Ravi Marri, Ramanna, and Shubra Dixit for their help in providing us with administrative data. This paper would not have been possible without the continuous efforts and inputs of the J-PAL/IPA project team including Kshitij Batra, Prathap Kasina, Piali Mukhopadhyay, Raghu Kishore Nekanti, Matt Pecenco, Surili Sheth, and Pratibha Shrestha. Finally, we thank the Omidyar Network especially Jayant Sinha, CV Madhukar, Surya Mantha, Ashu Sikri, and Dhawal Kothari for the financial support that made this study possible. UC San Diego, JPAL, NBER, and BREAD. kamurali@ucsd.edu. UC San Diego, JPAL, NBER, and BREAD. pniehaus@ucsd.edu. Dartmouth College, JPAL, and BREAD. sandip.sukhtankar@dartmouth.edu. 1

2 Randomized control trials (RCTs) have gained prominence amongst empirical work, particularly in development economics, over the last decade. While RCTs have no doubt improved the standard of identification in development, they are not without their critics (see, for example, Bardhan (2013)). Acemoglu (2010) among others has voiced concern that the bulk of empirical work using microdata, particularly in development economics, engages in partial equilibrium comparisons and thus cannot capture first-order general equilibrium effects. In this (proposed, almost complete) paper we demonstrate how randomized evaluations and empirical work using microdata can be used to address these concerns. We use theory to guide us in predicting which general equilibrium effects are important, and this in turn guides our choice of experimental methods. By randomizing large geographic units into treatment and control arms, we are able to capture general equilibrium and spillover effects of an ambitious intervention aimed at improving public program performance, as well as compare the magnitudes of partial and general equilibrium effects. Our context is an evaluation of a large-scale rollout of payments infrastructure ( Smartcards for short) integrated into the public employment program in India (NREGS). Working with the Government of the Indian state of Andhra Pradesh, we randomized the order in which 158 sub-districts introduced Smartcards to make biometrically authenticated wage payments on NREGS. The experiment randomized the form of payments over a universe of about 19 million people, with randomization conducted over entire sub-districts, making it one of the largest randomized controlled trials ever conducted. NREGS is one of the two flagship welfare schemes in India, and guarantees every rural household 100 days of employment doing manual labor at minimum wages. In theory, employment guarantee schemes such as the NREGS are expected to affect equilibrium in private labor markets (Dreze and Sen, 1991, Murgai and Ravallion, 2005). A truly guaranteed public-sector job puts upward pressure on private sector wages by improving workers outside options. As Dutta et al. (2012) puts it,...by linking the wage rate for such work to the statutory minimum wage rate, and guaranteeing work at that wage rate, [an employment guarantee] is essentially a means of enforcing that minimum wage rate on all casual work, including that not covered by the scheme. Indeed, the existence of such a program can radically alter the bargaining power of poor men and women in the labor market... by increasing the reservation wage... Consistent with this prediction, recent non-experimental analyses have estimated that the NREGS rollout may have raised rural unskilled wages by as much as 5-6% (though this is debated). 1 The NREGS, however, is beset by implementation problems, including delayed and missing payments, and large-scale corruption (Niehaus and Sukhtankar, 2013a,b). Improving the functioning of the NREGS by modernizing government payments delivery could thus have spillover impacts on local labor markets. Indeed, our evaluation of the effects of the biometrically-authenticated payment mechanism on direct program outcomes shows that 1 Imbert and Papp (2012) and Berg et al. (2012) estimate average effects on the order of 5-6%, while Zimmermann (2012) finds no average effects using different methods. The important point for our purposes is that existing estimates are no higher than 5-6%. 2

3 NREGS performance dramatically improved: time to collect payments was 19% faster, payment delays were reduced by 29%, beneficiaries earned 23% more on the program, and leakage and corruption declined (Muralidharan et al., 2014). We find that improved implementation of the NREGS triggered large general equilibrium impacts. Columns 1 and 2 of Table 2a report the impact of Smartcards on the average daily wage earned by workers who report positive amounts of private-sector wage employment during June. Smartcards raised June private-sector wages in treated mandals by 5-6% of the control group mean an effect identical in magnitude to the largest existing estimates of the impact of the NREGS itself. Figure 1 shows that agricultural wages are higher throughout the year in treatment areas, with effects somewhat larger during times of the year when NREGS participation is highest. These results suggest that by making the NREGS more attractive as an outside option, Smartcards forced private employers to pay more to attract labor; other data suggest that these results do not reflect an upward shift in the demand for labor. The most direct evidence on this point comes from data on workers reservation wages at the time they worked in the private sector. Columns 3 and 4 of Table 2a show that reservation wages increased by almost as much as wage realizations; the treatment effect on the difference is not statistically significant (p = 0.22). This is inconsistent with a pure labor demand shock, which would increase realized wages and workers surplus without affecting reservation wages. A corroborating piece of evidence is that we find no impacts on earnings from self-employment or enterprise (Columns 6 and 7 of Table 1a). We next examine real effects of higher reservation wages on the allocation of time to the private sector. This effect is theoretically ambiguous: participation should decrease if labor markets are perfectly competitive, but could increase if labor markets are oligopsonistic. In the data, NREGS workers increased their supply of labor to the private sector by a meaningful but statistically insignificant 5.2% (Columns 5 and 6 of Table 2a). This increase comes at the expenses of self-employed work and idle time; both decrease, albeit insignificantly. We can reject the joint hypothesis of perfect competition and a change in the quantity of labor demanded of -6%; we can reject a joint hypothesis that includes any decrease in the quantity of labor demanded at the 23% level. This is hardly conclusive evidence of imperfect competition, but does at least suggest that any distortionary effects of the NREGS are limited. Higher wages and (weakly) higher labor supply imply that beneficiary earnings should have increased. We verify this in Table 1a. We estimate a large increase in total income: Rs. 8,653, which is 12% of control group income or about a quarter of the rural poverty line. Strikingly, only 11% of this increase (Rs. 982) comes directly from increased NREGS income; the remaining 89% is due to increases in private sector earnings, primarily agricultural labor (Rs. 3226) and other physical labor (Rs. 3192). In other words, the direct benefits that Smartcards yield for NREGS participants appear quantitatively less important that their indirect effects on wages in general equilibrium. This bears out the insight in Murgai and Ravallion (2005) that spillover effects to nonparticipants could account for a large portion of the impact on poverty reduction of a scheme like the NREGS. Taken together, these results underscore the value of policy evaluation at scale. NOTE: The data collection and analysis in this paper is done, and the completed paper will include further results on labor market outcomes, household consumption and finances. 3

4 References Acemoglu, Daron, Theory, General Equilibrium, and Political Economy in Development Economics, Journal of Economic Perspectives, 2010, 24 (3), (document) Bardhan, Pranab, Little, big: Two ideas about fighting global poverty, Boston Review, May (document) Berg, Erlend, Sambit Bhattacharyya, Rajasekhar Durgam, and Manjula Ramachandra, Can Rural Public Works Affect Agricultural Wages? Evidence from India, CSAE Working Paper Series , Centre for the Study of African Economies, University of Oxford Dreze, Jean and Amartya Sen, Hunger and Public Action number In OUP Catalogue., Oxford University Press, (document) Dutta, Puja, Rinku Murgai, Martin Ravallion, and Dominique van de Walle, Does Indiaï œs Employment Guarantee Scheme Guarantee Employment?, Policy Research Working Paper Series 6003, World Bank (document) Imbert, Clement and John Papp, Equilibrium Distributional Impacts of Government Employment Programs: Evidence from Indiaï œs Employment Guarantee, Working Paper Series 14, Paris School of Economics Muralidharan, Karthik, Paul Niehaus, and Sandip Sukhtankar, Payments Infrastructure and the Performance of Public Programs: Evidence from Biometric Smartcards in India, Technical Report, University of California, San Diego (document) Murgai, Rinku and Martin Ravallion, Is a guaranteed living wage a good anti-poverty policy?, Policy Research Working Paper Series 3640, The World Bank June (document) Niehaus, Paul and Sandip Sukhtankar, Corruption Dynamics: the Golden Goose Effect, American Economic Journal: Economic Policy, 2013, 5. (document) and, The marginal rate of corruption in public programs: Evidence from India, Journal of Public Economics, 2013, 104, (document) Zimmermann, Laura, Labor Market Impacts of a Large-Scale Public Works Program: Evidence from the Indian Employment Guarantee Scheme, IZA Discussion Papers 6858, Institute for the Study of Labor (IZA) September

5 Wage T C Wages for Agricultural Labor, Differenced Male Female Control Treatment 10 0 January February March April May June July August September October November December Month January February March April May June July August September October November December Figure 1: Wages for agricultural labor by month The outcome shown in these figures is the difference between endline and baseline wages at the village level, along with standard error bands (clustered at the mandal level). The wage reports come from the villagelevel (as opposed to household-level) survey, with one observation per village for the endline survey (three per village at baseline). 5

6 Table 1: Income increases (a) NREGS Total income NREGS Ag labor Other labor Farm income Business Misc (1) (2) (3) (4) (5) (6) (7) (8) Treatment (3655) (3640) (579) (1456) (1282) (2189) (1300) (2090) BL GP Mean.043 (.048) District FE Yes Yes Yes Yes Yes Yes Yes Yes Adj R-squared Control Mean N. of cases (b) SSP Total income SSP Ag labor Other labor Farm income Business Misc (1) (2) (3) (4) (5) (6) (7) (8) Treatment (3610) (3578) (124) (1603) (787) (1510) (1013) (1852) BL GP Mean.027 (.064) District FE Yes Yes Yes Yes Yes Yes Yes Yes Adj R-squared Control Mean N. of cases This table analyzes household level income (with top 0.5% trimmed in both treatment and control groups). All amounts are in Rupees (Rs. 1 = US $ in August 2012). Total income is full household annual income in 2012 from all sources. Columns 3-8 break down this figure into its components. Column 3 shows income from NREGS (SSP in table (b)). Columns 4 and 5 show income from physical labor other than NREGS, with column 4 referring to income from agricultural labor and column 5 income from other physical labor. Column 6 shows income from farming or livestock, while column 7 shows business/entrepreneurial income. Column 8 groups together all other remaining sources of income. Standard errors clustered at mandal level in parentheses. Statistical significance is denoted as: p < 0.10, p < 0.05, p <

7 Table 2: Decomposing June income gains (a) Private sector employment in June Daily wage (Rs.) Daily res. wage (Rs.) Days worked Days self-employed Days idle (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Treatment (3.7) (3.7) (3) (2.9) (.6) (.57) (.44) (.44) (.56) (.54) BL GP Mean (.057) (.057) (.086) (.051) (.05) District FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Adj R-squared Control Mean N. of cases Survey Both Both Both Both Both Both Both Both Both Both (b) NREGS employment in June Days worked Hours worked Hourly wage (1) (2) (3) (4) (5) (6) Treatment (.66) (.64) (3.4) (3.3) (.77) (.8) BL GP Mean (.042) (.041) (.037) District FE Yes Yes Yes Yes Yes Yes Adj R-squared Control Mean N. of cases Survey NREGS NREGS NREGS NREGS NREGS NREGS This table analyzes beneficiary level employment outcomes during the month of June, with Panel (a) focused on private sector employment and Panel (b) on NREGS employment. Days worked refers to total paid employment, self-employed refers to total self or other unpaid employment, while days idle corresponds to days not spent working, including weekends. Wage is the average daily wage (in rupees), while res wage is the reservation wage, or the lowest daily wage the beneficiary would have accepted for labor. The respondents were asked if they would have been willing to work for someone else for a daily wage of Rs. 20, with the amount increasing by Rs. 5 until the respondent answered affirmatively. Days worked and hours worked in Panel (b) refer to days and hours worked on NREGS project during the study period (baseline in May 31 to July 4; endline in May 28 to July 15). Standard errors clustered at mandal level in parentheses. Statistical significance is denoted as: p < 0.10, p < 0.05, p <