IFPRI-IGIDR Workshop on Implementation of MGNREGA in India A Review of Impacts for Future Learning - Targeting and Implementation - Sudha Narayanan, Upasak Das, Krushna Ranaware
Presented at a one day workshop jointly organized by Indira Gandhi Institute of Development Research (IGIDR), International Food Policy Research Institute (IFPRI), Cornell University, with funding from International Initiative for Impact Evaluation (3ie) titled 'Implementation of MGNREGA in India: A Review of Impacts for Future Learning'.
The main objective of the workshop was take stock of the current scenario of MGNREGA, assess the impacts it has made over the past decade and emerge with knowledge as to the areas under MGNREGA that still need to be studied and can be opened up with more research.
Similaire à IFPRI-IGIDR Workshop on Implementation of MGNREGA in India A Review of Impacts for Future Learning - Targeting and Implementation - Sudha Narayanan, Upasak Das, Krushna Ranaware
Similaire à IFPRI-IGIDR Workshop on Implementation of MGNREGA in India A Review of Impacts for Future Learning - Targeting and Implementation - Sudha Narayanan, Upasak Das, Krushna Ranaware (20)
IFPRI-IGIDR Workshop on Implementation of MGNREGA in India A Review of Impacts for Future Learning - Targeting and Implementation - Sudha Narayanan, Upasak Das, Krushna Ranaware
1. Theme 1:
Targeting and implementation
Implementation of MGNREGA in India:
A Review of Impacts for Future Learning
Featuring work completed by IFPRI, Cornell University, and
IGIDR with funding from 3ie
2. Introduction
• MGNREGA is ostensibly a demand-driven program with
local level implementation at its core
• “Self targeting” mechanism, but which individuals actually
work on MGNREGA is an indication of how well the
program is reaching out to the intended beneficiaries
(poor and marginalized)
• Flows of money to particular geographic areas also tell us
something about the quality of implementation and
governance record
• Are the safeguard measures put in place to bolster
transparency of implementation (social audits, publicly
available data, etc.) actually working?
3. Review of literature
• Targeting and rationing
• Dutta et al. (2012)
• Liu and Barrett (2013)
• Narayanan, Das, Liu, Barrett (2015)
• Narayanan and Das (2014)
• Implementation and governance issues
• Niehaus and Sukhtankar (2013 a, b)
• Gupta and Mukhopadhyay (2014)
• Zimmermann (2013)
• Maiorano (2014)
• Sheahan, Liu, Barrett, Narayanan (2014)
• Chau, Liu, Soundararajan (in progress)
• Narayanan and Ranaware (in progress)
• Overcoming implementation challenges
• Afridi and Iversen (2014)
• Muralidharan, Niehaus, Sukhtankar (2014)
• Raabe et al. (2010)
4. Review of literature: Targeting and rationing
• Some degree of rationing in unavoidable given fixed budgets at the
highest levels, however widespread rationing implies implementation
issues and low quality service delivery
• Dutta et al. (2012) look at MGNREGA admin data and NSSO data
from 2009-10 to study targeting and rationing
• Individuals in the poorest states have the highest demand for MGNREGA work
• But poorest states also have highest rationing levels. Why?
• Weaker governance capacity
• Less empowerment of the poor
• Participation rates are still high for the poor, implying that rationing isn’t
completely undermining the self-targeting of the marginalized (STs and OBCs)
• Find a massive discrepancy in rationing rates between data sources
• MGNREGA admin data: 99% of demand was met
• NSSO data: 56% of demand was met
• Lots of variation across states (80% of the individuals who demand work in
Himachal Pradesh, Rajasthan, and Tamil Nadu receive it)
5. Liu and Barrett (2013)
• Extend the Dutta et al. (2012) story with specific interest in
the extent to which the program is operated as “pro-poor”
across all of India and specifically by individual states
• Use the same data (NSSO 2009-10) to estimate the
relationship between per capita expenditures and 3
MGNREGA targeting measures
1. Participation
2. Job-seeking
3. Rationing
Project paper
6. Liu and Barrett (2013)
• Results: rationing profile
1. Greater rates of self-selection
into the program by poorer
and disadvantaged
households
2. Rationing of MGNREGA jobs
is not pro-poor but rather
exhibits a “middle-class bias”
3. Self-selection dominates
rationing, therefore program is
more pro-poor than not
4. Does not reach poor female-
headed households, both self-
selection and rationing
Project paper
7. Liu and Barrett (2013)
Project paper
• Results: rationing profile
• Considerable differences
across states
• Using participation and
rationing both as metrics, find
that about half of the states
exhibit pro-poor targeting
while the other half do not
• 5 states with exemplary pro-
poor targeting (mostly in the
NE): Manipur, Mizoram, Rajasthan,
Sikkim, and Tripura
• 8 states with almost
exemplary pro-poor targeting:
Andhra Pradesh, Chhattisgarh,
Himachal Pradesh, Madhya
Pradesh, Meghalaya, Nagaland,
Tamil Nadu, and West Bengal
8. Liu and Barrett (2013)
• Appropriate policy responses
• Limited participation by the poor due to low rates of MGNREGA job-
seeking
• Address lack of knowledge about “right to work”
• Identify and address administrative impediments
• Free labor supply constraints (for example, for women)
• Tackle worker discouragement
• High rates of rationing among poor
• Identify and address administrative failures
“Clearly, there is room for improvement and perhaps much to be
learned from an in-depth comparative analysis of MGNREGA
programme implementation across states that have
demonstrated greater or lesser success in targeting the poor
with job opportunities.” (p. 53)
Project paper
9. Narayanan, Das, Liu, Barrett (2015)
Project paper
Revisits rationing and
targeting in light of the
decline in the scale of
the MGNREGA
between
2009-10 (66th Round
of NSS) and 2011-12
(68th Round of NSS)
- What explains this
trend?
- What is the extent of
rationing in 2011-12?
- Is rationing pro-poor?
- What explains
rationing itself?
10. Narayanan, Das, Liu, Barrett (2015)
• The “collapse” of the
MGNREGA
• Two views:
• MGNREGA “has done its
job” and no longer relevant.
• Unmet demand, poor
implementation
discourages workers from
seeking work.
• Research questions
• Demand side issues
(drivers of worker interest)
• Supply side issues (poor
implementation – rationing
and delay in payments)
66th
Round
(2009-10)
%
68th
Round
(2011-12)
%
“Demand” 43.5 30
Rationing Rate 44.4 23.1
Participation Rate
(total worked/total
hhds)
24.2 23
Project paper
11. Evidence on unmet demand
• Drèze and Khera (2011):
•Only 13% of the survey households in the six Hindi speaking states secured
100 days of work.
• Dreze and Khera (2014): PEEP Survey in ten states
• When job card holders were asked how many days of employment they
would like to have over the year, assuming that they are paid on time, an
overwhelming majority (83 per cent) answered ‘100 days’—the maximum
entitlement.
• Himanshu, Mukhopadhyay and Sharan (2015): Primary survey in
Rajasthan
•Decline in MGNREGA demand in Rajasthan, mostly due to loss of worker
interest.
•Das (2015) and Dey (2010)
• Unmet demand in parts of West Bengal. Finds workers get only 10% of the
desired number of days.
•Srinivasan, et. al (in progress) In Surguja, 32.74% of sample reported
that they faced problems getting work.
In both, rationing is on the intensive margin.
Project paper
12. Narayanan, Das, Liu, Barrett (2015)
• A strong “discouraged worker effect”
• Comes mainly from administrative rationing
• To a lesser extent from delay in wage payments (not robust)
• Data is incomplete
• People factor in the delays and are not discouraged as long as payment is
certain.
• Competing explanations matter, but not entirely responsible for
lower demand
• Favourable weather conditions matter
• Better opportunities in the labour market do not matter
• Wage differentials between MGNREGA and alternatives do not matter
• Growth in MPCE associated with lower demand.
Conclusion: Poor implementation, especially administrative rationing is a
deterrent. Where improvements in incomes are higher, demand has
declined but there is no relationship with the growth in non-agricultural
work participation.
Project paper
14. Narayanan, Das, Liu, Barrett (2015)
Correlates of rationing and pro-poor rationing
• Political economy factors
• Clientelism
• Poor implementation capacity
• Overwhelming demand (covariate shocks /catastrophic
shocks)
• Rationing is more related to shocks and perhaps
administrative infrastructure rather than political affiliation
• Post election year rationing is higher.
• Pro-poor rationing is correlated with the identity of the party.
Project paper
15. Narayanan and Das (2014)
• Patterns of Rationing for Women (NSS 68th Round, 2011-
12)
• The MGNREGA has been considered a “women’s programme”
• One-third mandate
• Equal wages
• Other programme features supportive of women’s participation
• In general, the performance has been good with women’s
participation consistently above the mandated one-third of all workers.
• Disaggregated analysis of rationing rates too suggest women are not
disadvantaged.
• Yet, vulnerable groups among women face problems at all stages –
possessing job cards, seeking work and being rationed out.
• For example, widows, single women households, mothers of young
children.
• Considerable variation across states.
Project paper
16. Narayanan and Das (2014)
Project paper
Groups Rationing rate (All-India)
Female headed households 0.19
Female headed households with no adult
males
0.19
Widows 0.20
Females from households belonging to the
Scheduled Castes or Tribes
0.26
Females from households with children (0-5
years)
0.26
17. Narayanan and Das (2014)
Conclusion
• Differentiated nature of
women’s experience in
accessing the MGNREGA
• Where there is pro-women
rationing, states need to play
supporting role and address
higher order concerns.
• Wwomen’s participation is
weak and rationing indicates
some sort of administrative
discrimination, policies have to
focus on enabling women to
access work and sensitizing
implementing staff.
Project paper
States Rationing
rate for
males
Rationing Rate
for females
Andhra Pradesh 0.19 0.16
Rajasthan 0.40 0.26
Tamil Nadu 0.14 0.07
Chattisgarh 0.10 0.10
Karnataka 0.41 0.41
Maharashtra 0.65 0.65
Gujarat 0.46 0.48
Bihar 0.43 0.58
Madhya Pradesh 0.39 0.42
India 0.28 0.25
18. Narayanan and Ranaware (in progress)
• Delay in payments
• Occur at many points in the
workflow
• Worker experience
Chhattisgarh (Nov, 2014): In
Surguja, proportion who reported
that they faced problems regarding
timely payments 47.6%
• 41.8 in “low” GPs
• 50.2 in “medium” GPs
• 51.5 in “high” GPs
PEEP Survey (2013)
66% over waited over 15 days
• 59.7% in “Leaders”
• 79% in “Learners”
• 61.5% In “Laggards”
Project paper
19. Narayanan and Ranaware (in progress)
• Proximate correlates of delays in wage payments
• Data
• Delay in payments from MIS (variable quality, dynamic, April 2014)
• Compute average delay, using proportion of musters delayed as
weights.
• Combine data on rainfall, banks, post offices, elections, etc. from
various sources.
• Qualitative research on the last mile problem.
Project paper
20. Banks and post offices
Narayanan and Ranaware (in progress)
Project paper
21. Review of literature: Implementation
• Governance (rent-seeking, leakages)
• When statutory daily wages increased in 2007, officials report more fictitious
work on wage projects in Orissa, however theft on piece-rate projects declined
(with Andhra Pradesh as control state) (Niehaus and Sukhtankar 2013 a)
• Very high leakage rate among wages paid at that time: wages in official
government data increased as expected, but households report receiving the
same wages as before (Niehaus and Sukhtankar 2013 b)
• Election-effects (politics)
• In Rajasthan, funds allocated were 22 percent higher in blocks where the INC
seat share was less than 39 percent in the previous election (Gupta and
Mukhopadhyay 2014)
• Only true when the MP of the district, who approves the block fund allocation, is from
INC
• Government benefited from MGNREGA in 2009 (Zimmermann 2013)
• However only talking about improving the plight of the poor is not sufficient to ensure
electoral success, also requires good implementation
• Also finds that districts were not perfectly align with the phase they should have been
assigned based on the known classification system/algorithm
• Political commitment to results has been key to good implementation in Andhra
Pradesh, but evolved into a top-down system instead of a demand-driven one
as a result, e.g. role of Field Assistants (Maiorano 2014)
22. Sheahan, Liu, Barrett, Narayanan (2014)
• Extend qualitative findings from Maiorano (2014) to study the
relationship between fund allocated at the mandal level by
fiscal year and election outcomes in Andhra Pradesh
• All data from publicly available sources (1063 mandals across
22 districts)
• MGNREGA spending amounts from government website
• Indian Population Census from 2001
• Indian Agricultural Census from 2005/06
• Geo-referenced rainfall data via NASA
• Disaggregated voting data from assembly constituency level via
Election Commission of India (2004 and 2009 elections)
Project paper
23. Sheahan, Liu, Barrett, Narayanan (2014)
• Allocation of funds before 2009 elections:
• Fiscal years include 2006/07, 2007/08, and 2008/09
• No evidence that fund allocation was correlated with political
variables in initial years of program implementation
• We define political variables relative to voting patterns in the 2004
election (the baseline political affiliation of mandals)
• Allocation of funds well correlated with various “need based”
measures
• People: Illiteracy, scheduled caste and tribes, agricultural laborers
(assumed to be casual laborers)
• Place: Unirrigated land, poorer infrastructure (paved road, ag credit
societies, etc.), remoteness
• This suggests no manipulation of approved funds in an effort to win
elections
Project paper
24. Sheahan, Liu, Barrett, Narayanan (2014)
• Allocation of funds after 2009 elections:
• Fiscal years include 2010/11, 2011/12, and 2012/13
• Consistent evidence that fund allocation was correlated with
political variables in years of program implementation directly
following 2009 election
• We define with respect to INC specifically but also UPA-affiliated parties
• However, effects are very small in magnitude and economically
insignificant
• At the same time, still well-correlated with needs-based measures
too, although less strongly than before 2009 election
• For example, not well-correlated with rainfall shock, suggesting that
implementation was not very flexible in dealing with labor market
dynamics
Project paper
25. Sheahan, Liu, Barrett, Narayanan (2014)
(1)
All years
(2)
Pre-2009
(3)
Post-2009
Clientelism 1.0 0.1 2.5
Needs-based: labor-related 14.2 9.9 22.9
Needs-based: land-related 11.3 11.6 16.7
Needs-based: infrastructure-related 14.2 12.0 20.2
Needs-based: rainfall-variability 2.5 2.9 3.9
Election controls 2.2 2.3 3.1
District and year dummies 54.6 61.2 30.7
Project paper
Decomposition of R2 for MGNREGS fund expenditure models
• Even when politics may have influenced fund allocation,
correlation with needs-based measures far exceeds political
measures, even in post-2009 years
• Like Zimmerman (2013), we also find aggregate MGNREGA
spending in the pre-election years is positive and statistically
significantly correlated with the movement of voters towards
UPA candidates
26. Chau, Liu, Soundararajan (in progress)
• Hypothesis: targeting at the household level by the village
local leaders in Andhra Pradesh is less related to need
and more related to politics
• If targeting is not well correlated with needs, then analyze
household political/voting responses to being allocated
work preferentially
• Do higher levels of MGNREGA benefits affect the likelihood of
households shifting parties between elections?
• Does this vary with past political affiliation and intensity of political
participation?
• Use household survey data from 2006 and 2008 matched
with administrative voting records
Project paper
27. Das (2015)
• Hypothesis:
• Explores if political clientelism affect allocation of benefits to the
households under the programme in West Bengal.
• Examines if households, whose heads attend political meetings and
rallies regularly get more benefits out of the programme.
• Survey of 556 households in four GPs of the Cooch Behar district of
West Bengal.
• Findings
• Households that support the local ruling political party have significantly higher
probability of getting work after seeking relative to those supporting the
opposition party.
• They get more days of work and therefore earn more.
• Need to generate awareness and reduce rationing.
28. Review of literature: Overcoming implementation issues
• Social audits
• In Andhra Pradesh, positive but insignificant impact of audits on
employment generation and a modest decline in the leakage amount per
labour related irregularity (Afridi and Iversen 2014)
• However, increase in more sophisticated and harder to detect material-related
irregularities
• Need for follow up and punishment/correction mechanisms
• Biometric smartcards
• In Andhra Pradesh, the new system delivered a faster, more predictable,
and less corrupt payments process without adversely affecting program
access (Muralidharan, Niehaus, Sukhtankar 2014)
• Program participants happy with change too
• Process-influence mapping exercise (Raabe et al. 2010)
• Implemented by a group of researchers 2 districts in Bihar
• Found many nodes of where the details of program implementation were
not working
• A useful process for other areas, especially where there are known
challenges?