Innovations in Public Sector Management: Evidence on Social Protection Programs New Delhi, India January 7
India s Varied Experience b with Social Protection Programs Innovations in Public Sector Management: Evidence on Social Protection Programs January 7 Delhi, India
Implementation of Social Protection Programs in India Jennifer Bussell, UC Berkeley
Social protection programs: Three considerations Innovation What programs are introduced? Incentives What pre-existing factors affect how policies are implemented? Information What do we know about the effects of the programs? P O VE R TY AC TI O NLAB.O R G 4
Social welfare spending has expanded in India For most individuals, accessing public services is the predominant mode of citizenstate interaction Welfare benefits Utilities Identity documents Yet public service provision is often highly flawed, as reflected in inaccessible, expensive, or inappropriate services Ration card office, Kerala, India P O VE R TY AC TI O NLAB.O R G
Innovation India has introduced many unique and pioneering programs State E-governance initiatives and National egovernance Plan Aadhaar Midday Meal, NREGS P O VE R TY AC TI O NLAB.O R G 6
Incentives Policies do not necessarily take all political and bureaucratic incentives into account Actual implementation of policies can be dramatically shaped by the incentives of relevant politicians and bureaucrats These actors may not be involved with the design of policies but are fundamental to their successful implementation Mismatched incentives can threaten even the most promising policy initiative P O VE R TY AC TI O NLAB.O R G 7
Example: Common service centers One-stop citizen service centers Computerized Private sector support Fundamental government services Driving licenses, birth certificates, land titles, welfare benefits, tax payments Considerable variation across Indian states P O VE R TY AC TI O NLAB.O R G 8
Number of services available at centers (2010) 45 40 35 30 25 20 15 10 5 0 P O VE R TY AC TI O NLAB.O R G 9
Explaining variation in reforms: Costs and benefits Benefits - Improved efficiency and reduced cost of service access Citizens Bureaucrats Politicians - Efficiencies in internal processes - Direct: Electoral rewards - Indirect: Potential new rents from contracts Costs - Limited - Costs of learning new processes - Reduced illicit income from bribes - Reduced illicit income from bribes P O VE R TY AC TI O NLAB.O R G 10
Implications and outcomes High levels of petty corruption are associated with less robust reforms Later adoption of service center policies Availability of 14 fewer services, on average Inclusion of fewer bribe-prone services High levels of grand corruption are associated with increased adoption of public private partnerships P O VE R TY AC TI O NLAB.O R G 11
Information: Did reforms improve service quality? Many remaining limitations in the system In the mid-late 2000s across India, 30-54% of people went outside the official system to get services to which they are entitled Bribes, influence of friends/family, influence of bureaucrat, political influence, middlemen, etc. This included ~40% of below poverty line households who paid bribes in dealing with departments such as Housing, Land, and the Police Other and more recent efforts such as Aadhaar may face similar incentive-based constraints in implementation and related limits to delivery of benefits P O VE R TY AC TI O NLAB.O R G 12
Understanding the quality of policy implementation requires better information Estimating the true effects of technology-based programs egovernance, Aadhaar is limited by lack of rigorous, randomized evaluation programs Some experimental evidence that service centers can improve service delivery Promising evaluations in other areas e.g. work programs, health sector provide strong evidence of the benefits offered by rigorous evaluation Most types of social protection programs have the potential to be evaluated in a similar manner P O VE R TY AC TI O NLAB.O R G 13
Kerala Himachal Pradesh Haryana Rajasthan Orissa Chhattisgarh Andhra Pradesh WestBengal Karnataka Punjab Tamil Nadu Maharashtra Uttar Pradesh Uttarakhand Gujarat Delhi Sub-national variation in corruption can help explain regional variation in outcomes 1.00 0.80 0.60 0.40 0.20 0.00 Grand Corruption Petty Corruption P O VE R TY AC TI O NLAB.O R G 14
Kerala Himachal Pradesh Gujarat Andhra Pradesh Maharashtra Chhattisgarh Punjab West Bengal Orissa Uttar Pradesh Uttarakhand Delhi Tamil Nadu Haryana Rajasthan Karnataka Sub-national variation in petty corruption 1 0.8 0.6 0.4 0.2 0 P O VE R TY AC TI O NLAB.O R G 15
Curbing Leakage in Public Programs with Direct Benefit Transfers Prabhat Barnwal, Michigan State University
Missing 40 million households with LPG connections Comparison of LPG (cooking fuel) households from different data sources Large gap in number of registered LPG connections vs. recorded LPG users Overall, about 40 million missing LPG using households Total LPG subsidy outgo: Rs. 48,000 Cr (US$8 billion)in 2013-14 Rs. 36,000 Cr (US$6 billion) in 2014-15 Losing sleep over subsidy leakage, not subsidy itself : Pranab Mukherjee (2012) P O VE R TY AC TI O NLAB.O R G 17
Transforming the welfare delivery mechanism to reduce leakage: Direct Benefit Transfers Leakage: Transfers to non-beneficiaries through illegal means Ghost beneficiaries, critical role of local intermediary officials Isolated databases, difficult and costly de-duplication Traditional enforcement (audit, penalty) is often ineffective DBT launched on January 1, 2013 in pension and scholarship programs LPG subsidy is the first major program under DBT, started in June 2013 DBT seeks to provide benefits directly into the bank account of verified beneficiaries Utilizes a centralized payments infrastructure Aadhaar used for verification of beneficiaries JAM trinity (Jan-Dhan, Aadhaar, and Mobile) have gradually become the backbone of the DBT platform P O VE R TY AC TI O NLAB.O R G 18
LPG distribution and black markets Source: http://indiatoday.intoday.in Setting: an in-kind transfer program leading to market segmentation Universal subsidies for household domestic cooking Commercial usage of the same fuel is taxed Subsidized fuel (i.e. domestic LPG) diverted to the black market using ghost and duplicate accounts Commercial users purchase diverted subsidized LPG in black markets, instead of paying commercial LPG price How big is the price difference between subsidized domestic LPG and nonsubsidized commercial LPG? Jan 2014: about 75% Jan 2016: about 50% P O VE R TY AC TI O NLAB.O R G 19
Direct Benefit Transfer for LPG (DBTL) Policy change: Subsidy transfers directly only to the verified beneficiaries into their bank account Pre-DBTL: The intermediating officials play an important role in enforcement (i.e. LPG distributor and delivery man) Difficult and costly to de-duplicate beneficiary list Post-DBTL: The role of intermediating officials is minimized De-duplication and elimination of ghosts The impact of enforcement would be undermined if Agents find new ways to manipulate the system Technology fails to deliver Due to displacement in fraud and changes in social norms P O VE R TY AC TI O NLAB.O R G 20
How to measure the impact of DBTL? DBTL was introduced in about 300 districts in six phases before getting terminated in early 2014 Phase 1 20 districts Phase 2 34 districts, and so on Difference-in-differences using two quasi-experiments: Compare outcomes in DBT and non-dbt districts before and after policy change Phasing-in of the policy across districts Unexpected termination of the policy Audit surveys in black market in 89 districts covering DBTL On & Off periods Supply side (LPG delivery men) Demand side (Small businesses) LPG transactions data on 4 million customers and about 3000 distributors P O VE R TY AC TI O NLAB.O R G 21
Predictions if DBTL is effective DBTL enforced DBTL terminated Domestic LPG purchase Black market LPG price Commercial LPG purchase P O VE R TY AC TI O NLAB.O R G 22
DBTL s impact on domestic LPG purchase 11 14% reduction in domestic fuel purchase by households P O VE R TY AC TI O NLAB.O R G 23
Removal of enforcement: DBTL termination Domestic fuel purchase reverts to original level, suggesting an increase in diversion of subsidized fuel Positive supply shock in the black market: price decreases by 13 19% Commercial firms return to the black market: reduces fuel purchase through formal channels by 6 9% P O VE R TY AC TI O NLAB.O R G 24
DBTL termination and black market LPG prices Increased supply of diverted LPG refills brings black market prices down quickly No such effect observed in black markets in DBTLtransition or non-dbtl districts P O VE R TY AC TI O NLAB.O R G 25
Reimplementation of DBTL after 2014 elections DBTL was terminated in March 2014, before general elections Reintroduced in November 2014 in 54 districts and has now covered all 676 districts since April 2015 Similar impact on domestic and commercial LPG purchases i.e., reduction on domestic LPG purchase is associated with increase in commercial LPG sales P O VE R TY AC TI O NLAB.O R G 26
Policy summary Improving the design of public programs using DBT to reduce leakage in social protection programs Recurring savings in LPG subsidy is huge Fiscal savings being used to improve LPG access to the poor Need to ensure last mile access and minimal genuine exclusion LPG is an urban fuel, used mainly by the middle and higher class DBT for kerosene and food need more efforts and coordination to improve end user s access and experience P O VE R TY AC TI O NLAB.O R G 27
Using Secure Payments to Improve Delivery of Public Welfare Programs in India Karthik Muralidharan, UC San Diego
High costs of delivering government transfers P O VE R TY AC TI O NLAB.O R G 29
Secure payments as state capacity P O VE R TY AC TI O NLAB.O R G 30
Several grounds for optimism as well as skepticism Optimism Reduce leakage; improve payment experience? Expand feasible set of anti-poverty policies? Leapfrog literacy constraints to financial inclusion? Will be a game changer for governance (both UPA & NDA) Skepticism Complex implementation challenges? Subversion by vested interests? Exclusion errors? Make benefits more difficult to access? P O VE R TY AC TI O NLAB.O R G 31
The AP Smartcard Program P O VE R TY AC TI O NLAB.O R G 32
Evaluation strategy MoU between J-PAL and Government of Andhra Pradesh to randomize the phased roll-out of Smartcards across mandals (blocks) Areas are identical except for Smartcards allowed us to credibly measure program impact (treatment vs. comparison) Detailed household surveys matched to official records Measure access, work done, payments process, leakage, opinions Compare key outcomes in treatment and comparison areas Detailed process report on implementation insights P O VE R TY AC TI O NLAB.O R G 33
Rollout was randomized at sub-district level in 8 districts with ~20 million people P O VE R TY AC TI O NLAB.O R G 34
Treatment and control mandals were identical on average except for the Smartcard program Variable Treatment Mean P O VE R TY AC TI O NLAB.O R G 35 Control Mean Difference (p-value) Population 43,846 43,807 38 (0.99) Pensions Per Capita 0.12 0.12 0.00 (0.68) Jobcards Per Capita 0.54 0.55-0.01 (0.86) Literacy Rate 0.45 0.45 0.00 (0.95) SC Proportion 0.19 0.19-0.00 (0.88) ST Proportion 0.10 0.10 0.00 (0.98) Proportion of Population Working 0.53 0.52 0.01 (0.47) Proportion Male 0.51 0.51 0.00 (0.94) Proportion of Pensioners under Indiramma 0.70 0.69 0.01 (0.52) Proportion of Old Age Pensions 0.60 0.62-0.02 (0.20) Proportion of Weaver Pensions 0.01 0.01-0.00 (0.64) Proportion of Disabled Pensions 0.13 0.13 0.00 (0.76) Proportion of Widow Pensions 0.26 0.25 0.02 (0.13)
Significant positive program impacts on several dimensions P O VE R TY AC TI O NLAB.O R G 36
Other main results No change in official outlays, but households got paid more Highlights importance of matched household surveys and randomized evaluation No reduction in access In fact, access and work done went up (fewer inflated muster rolls) Gains were broad-based (no one was worse off) Biometrics was key to reducing leakage Using local BCs was key to improving payments process Highly cost-effective Value of time savings alone (at NREGS wages) exceeds cost Leakage reduction was ~9x the cost (2% commission) P O VE R TY AC TI O NLAB.O R G 37
Strong user support for Smartcards New system 91% NREGA New system 93% SSP Old system 3% Old system 3% Neutral 6% Neutral 4% P O VE R TY AC TI O NLAB.O R G 38
Summary of lessons learned Large potential gains from biometric payments Achievable in as little as two years, very popular with voters Implementation quality really matters Stable and empowered project leadership with political backing One district, one bank (insurance against poor implementation) Invest in last-mile financial inclusion and not just biometrics Focus on beneficiary experience more than fiscal savings Key to political success was focusing on making biometric Smartcards convenient but not mandatory Large leakage reductions anyway, but eliminated exclusion errors Strongly recommend similar approach for next few years Value of high-quality independent evaluations P O VE R TY AC TI O NLAB.O R G 39
Getting the financial architecture of social programs to work for the poor: Some learnings from experiments on MGNREGS Rohini Pande, Harvard
Social protection program management Common feature: Centralized financing but decentralized implementation Associated problems: Beneficiary selection is at local level and disconnected from financing (from state or federal exchequer) Funds released in tranches cause short-run mismatches between need and receipt of funds. Concern: Poor financial practices imply excessive fund float and weak implementation Rents available to local actors who determine access Can technology-based innovations to program s financial architecture help? P O VE R TY AC TI O NLAB.O R G 41
Focus on MGNREGS for innovations MGNREGS remains India s flagship social protection program: Employs 100 million rural equivalent to 1 out of every 3 rural adults. Increasing MGNREGS budget is a significant policy response to recent drought Fantastic example of data transparency large scale administrative data made available [Netnrega/nrega-reportdashboard.html] E-payments seen as potential impetus for financial inclusion Households Provided Employment P O VE R TY AC TI O NLAB.O R G 42
Can e-governance improve fund management? (joint with Banerjee, Duflo, Imbert, and Mathew) Study: Field experiment evaluating expenditure-based fund flow reform for MGNREGS in Bihar Policy reform that we evaluate: Reduce number of administrative tiers involved in fund release Funds released on basis of expenditure (wage bill) incurred P O VE R TY AC TI O NLAB.O R G 43
Intervention design Control Treatment P O VE R TY AC TI O NLAB.O R G 44
Decrease in spending Estimated effect = Rs. 230,000 per GP for total of 4.1 million USD nrega.nic.in shows slightly higher number: Rs. 330,000 per GP P O VE R TY AC TI O NLAB.O R G
Reduced float of funds.. Balance in GP accounts P O VE R TY AC TI O NLAB.O R G 46
But no decline in work Weeks worked (household survey)
Decline in officers wealth of over a third Reported wealth of district officials P O VE R TY AC TI O NLAB.O R G
Some learnings of MGNREGs e-reforms: Lesson 1: Data helped distinguish lower demand and corruption decline hypothesis. Short-run roll-back of program but in the longer run national rollout. Lesson 2: Bihar study: Better financial management reduced leakage but did not increase demand. Weaker evidence that local elite Pradhans lost wealth. Lesson 3: AP Study: Reducing Pradhan control at payment delivery reduced leakage and increased access. Significant beneficiary support but less than universal take-up within villages. Lesson 4: Ongoing work in MP: Potentially high returns to improving household ability to access banking system. P O VE R TY AC TI O NLAB.O R G
From reducing leakage to ensuring social protection for the vulnerable emerging points (larger project joint with Field, Moore, Rigol, and Schaner) MGNREGA has made a commitment to moving away from cash to bank payments. Work by Muralidharan et al. show that this is a very positive change But as financial inclusion efforts get underway on a massive scale in India, open question on how G2P payments should be structured We are working to examine whether direct transfer of benefits into a woman s account for MGNREGA work without and with training on account usage matters for both household and women s outcome and the nature of corruption Project is still underway, but we have three emerging lessons on program implementation P O VE R TY AC TI O NLAB.O R G
DBT unless accompanied by significant training may continue to exclude the most vulnerable 1. 95% of bank clients know that they have an account. But, overall, only 24% know their account number. Among those who received financial capacity building, two-thirds can report their account number. P O VE R TY AC TI O NLAB.O R G
The combination of financial capacity building and direct benefit transfers has a powerful effect on financial inclusion 2. Women who receive DBT + training are twice as likely to have a nonzero balance, and those with deposits have balances more than four times those of women who received bank accounts only. P O VE R TY AC TI O NLAB.O R G
Technology alone does not change incentives: Need adequate feedback loops 1) Having a bank account is not sufficient to receive DBT: 25% of male NREGA workers and 34% of female NREGA workers with a bank account still receive public works wages in cash 2) And having wages deposited into a bank account still leaves workers vulnerable to wage skimming: Among workers who received their wages through DBT, 40% report that a local official accompanied them to retrieve their wages on their last visit to the bank. P O VE R TY AC TI O NLAB.O R G 53
R Subrahmanyam Additional Secretary Department of Higher Education 54