Who Participates in MGNREGA? Analyses from Longitudinal Data

Abstract

The Mahatma Gandhi National Rural Employment Guarantee Act
(MGNREGA) was enacted in 2005 and has completed a little over
a decade in India. It is the largest public employment programme
in the world and has promoted a wider participation from rural
households across the country. This paper examines the issue of
programme participation in MGNREGA holistically by looking
at household and individual-level participation and controlling
for regional heterogeneity, using a unique panel data from the
nationally representative India Human Development Survey. Using
a binary logistic model and fixed effects models at the state and
village level, the paper finds that poor households with a low asset
base and those belonging to the Scheduled Caste (SC)/Scheduled
Tribe (ST) categories are more likely to participate in the programme,
but the support base of MGNREGA is not just limited to these
groups and is rather broad-based. It also shows that as compared
to other types of work, women suffer less disadvantage than men,
thereby providing empowerment opportunities to women.

Full text

ISSN 0972-2661
Review of
Development & Change
Volume XXII Number 1 January - June 2017
Madras Institute of Development Studies
Review of Development & Change: Vol. XXII No.1 Jan – June 2017
S. Mahendra Dev The Problem of Inequality
Susie Tharu Women Writing in India Reconsidered
Edwin Dickens How the Federal Reserve Caused the
Great Ination and Stagation
Mamata Swain Transition in Agrarian Structure in
Odisha
Omkar Joshi, Sonalde Desai Who Participates in MGNREGA?
Reeve Vanneman and
Amaresh Dubey
Sandeep Kumar Kujur Globalisation, Energy Efciency
and Material Consumption in India’s
Pulp and Paper Industry
Aseem Prakash The Hybrid State and Regulation of
Land and Real Estate

REVIEW OF DEVELOPMENT AND CHANGE
Madras Institute of Development Studies
79, II Main Road, Gandhinagar, Adyar, Chennai 600 020
Committed to examining diverse aspects of the changes taking place in our
society, Review of Development and Change aims to encourage scholarship that
perceives problems of development and social change in depth, documents them
with care, interprets them with rigour and communicates the ndings in a way
that is accessible to readers from different backgrounds.
Editor Shashanka Bhide
Editorial Committee Ajit Menon, L. Venkatachalam
Editorial Advisory Board
Sunil Amrith, Harvard University, USA
Sharad Chari, University of California, Berkeley, USA
John Robert Clammer, Jindal School of Liberal Arts and Humanities, India
Devika J, Centre for Development Studies, Thiruvananthapuram, India
Niraja Gopal Jayal, Jawaharlal Nehru University, India
Sisira Jayasuriya, Monash University, Australia
K.P. Kalirajan, Australian National University, Australia
Ravi Kanbur, Cornell University, USA
Anirudh Krishna, Duke University, USA
James Manor, University of London, UK
Mike Morris, University of Cape Town, South Africa
David Mosse, University of London, UK
Keijiro Otsuka, National Graduate Institute for Policy Studies, Japan
Cosmas Ochieng, Boston University, USA
Barbara Harriss-White, Oxford University, UK
Publication Ofcer R. Suresh
Publication Support A. Arivazhagan
Copyright
Copyright of material published in the journal rests with the authors concerned.
The authors, and not RDC or MIDS, are responsible for facts presented and
views expressed.

A note from the Editor ii
The Problem of Inequality
S. Mahendra Dev 1
Women Writing in India Reconsidered
Susie Tharu 44
How the Federal Reserve Caused the Great Ination and
Stagation: A Political Economic Approach
Edwin Dickens 56
Transition in Agrarian Structure in Odisha
Mamata Swain 81
Who Participates in MGNREGA?
Analyses from Longitudinal Data
Omkar Joshi, Sonalde Desai,
Reeve Vanneman and Amaresh Dubey 108
Globalisation, Energy Efciency and
Material Consumption in India’s Pulp and
Paper Industry (1980–81 to 2009–10)
Sandeep Kumar Kujur 138
The Hybrid State and Regulation of Land and
Real Estate: A Case Study of Gurugram, Haryana
Aseem Prakash 173
Review of
Development & Change
Volume XXII Number 1 January - June 2017

ii
A Note From the Editor
Review of Development and Change, a half-yearly journal from
Madras Institute of Development Studies (MIDS), invites articles
on problems of development and social change. The journal seeks
to encourage multi-disciplinary scholarship in particular because
development crosses disciplinary bounds. Multi-disciplinary
engagement may achieve a more holistic understanding of how
development works. Issues of focus include rural and urban
development, environment and sustainable development, social
sectors and human development, and poverty, inequality and
development.
We have now reconstituted our Editorial Board of the journal
keeping in mind our aim to make it more global in its coverage,
while retaining the multi-disciplinary feature of the journal.
We acknowledge with gratitude contributions of the members
of the previous Editorial Board comprising Professors
Barbara Harriss-White, Chandan Mukherjee, G. Haragopal,
M.V. Nadkarni, Rajan Gurukkul, U. Sankar and C.P. Chandrasekhar
for nurturing the journal so far.
We look forward to the guidance and support from the reconstituted
Board of Editors, whose members are listed on inside cover page.

108 Omkar JOshi, sOnalde desai, reeve vanneman and amaresh dubey
Review of Development & Change, Vol. XXII No.1, Jan-June 2017, pp. 108-137
Omkar Joshi is a Doctoral Student, Department of Sociology, University of Maryland, College
Park. Email: ojoshi@terpmail.umd.edu; Sonalde Desai is a Professor of Sociology at University of
Maryland, College Park, and Senior Fellow, National Council of Applied Economic Research; Reeve
Vanneman is a Professor of Sociology at University of Maryland, College Park; Amaresh Dubey is
a Professor of Economics at Jawaharlal Nehru University, New Delhi.
Who Participates in MGNREGA?
Analyses from Longitudinal Data
Omkar Joshi, Sonalde Desai,
Reeve Vanneman and Amaresh Dubey
ABSTRACT
The Mahatma Gandhi National Rural Employment Guarantee Act
(MGNREGA) was enacted in 2005 and has completed a little over
a decade in India. It is the largest public employment programme
in the world and has promoted a wider participation from rural
households across the country. This paper examines the issue of
programme participation in MGNREGA holistically by looking
at household and individual-level participation and controlling
for regional heterogeneity, using a unique panel data from the
nationally representative India Human Development Survey. Using
a binary logistic model and xed effects models at the state and
village level, the paper nds that poor households with a low asset
base and those belonging to the Scheduled Caste (SC)/Scheduled
Tribe (ST) categories are more likely to participate in the programme,
but the support base of MGNREGA is not just limited to these
groups and is rather broad-based. It also shows that as compared
to other types of work, women suffer less disadvantage than men,
thereby providing empowerment opportunities to women.
Keywords: MGNREGA; programme participation; public works
programme; social safety net

109
Who Participates in MGNREGA?
1. INTRODUCTION
In modern welfare states, provision of employment is often one of the
important mandates of the government. But the government has often
played two roles – as a provider of employment through public works
as well as a guarantor of employment rights. Although government-
provided employment is not meant to substitute for market employment,
public works programmes can provide a strong social safety net for the
unemployed and underemployed. In this paper, we examine the issue
of participation in the Mahatma Gandhi National Rural Employment
Guarantee Act (MGNREGA),1 the largest employment programme in
the world (World Bank 2015), which guarantees right to employment
in India. The MGNREGA, enacted in 2005, promises not less than 100
days of wage employment in a nancial year to every rural household
whose adult members are willing to engage in unskilled manual work.
Initially, the Act was intended to cover 200 backward districts in India,
but subsequently its scope was extended to all rural areas. Two critical
objectives of the Act and subsequent implementation guidelines are (a)
ensuring livelihood security for the most vulnerable people living in rural
areas,2 through providing employment opportunities for unskilled manual
work and (b) aiding in the empowerment of marginalised communities,
especially women, Scheduled Castes (SCs) and Scheduled Tribes (STs).
MGNREGA is a universal public works programme, but it has a strong
underlying targeting mechanism because entry into the programme is by
self-selection. Self-selection could affect the take-up of the programme
and in turn inuence labour market dynamics. The Government of India
has made a massive nancial commitment to this programme. In 2016–17,
Rs 38,500 crore (0.32 per cent of GDP as per the revised estimates)
was allocated to MGNREGA, and a total of 2.35 billion person-days
of employment were created (CBGA 2017; Sarkar and Islary 2017).
However, despite the promise of 100 days of employment, the average
number of days of employment provided under MGNREGA peaked at
54 days in 2009–10 and has been declining since. It has been hovering
below the 50 days mark for the most part (Desai et al. 2015). An analysis
of programme participation into MGNREGA can provide both theoretical
insights into how employment guarantee programmes work in general, as
well as wider policy lessons. Specically, we ask two questions: (a) which

110 Omkar JOshi, sOnalde desai, reeve vanneman and amaresh dubey
households are more likely to participate in this programme and what are
their prior characteristics? (b) what are the intra-household dynamics of
participation, especially with respect to gender and age? Answers to these
two questions provides an assessment of the success of the objectives of
MGNREGA mentioned previously. Our analysis offers new insights by
using a nationally representative longitudinal data set, the India Human
Development Survey (IHDS), with information from both the pre- and
post-MGNREGA periods for the same households.
2. BACKGROUND AND REVIEW OF LITERATURE
Since MGNREGA was passed, there has been a voluminous literature on
its features, design and impact on several outcomes of economic importance.
Both quantitative and qualitative studies have examined the performance
of MGNREGA in various geographic regions, ranging from a single state
to three or four states to an all-India level. Thematically these studies
have covered programme features and challenges (Dreze and Khera 2011;
Roy 2015), rationing (Dutta et al. 2012, 2014; Das 2015), its impact on
employment and wages (Azam 2012; Berg et al. 2012; Desai et al. 2015;
Imbert and Paap 2015; Zimmerman 2015), incomes (Jha et al. 2009),
welfare (Deininger and Liu 2013; Imbert and Paap 2015), migration (Liu
and Barret 2012; Novotny et al. 2013; Imbert and Paap 2014), agriculture
(Bhargava 2014; Varshney et al. 2014), children’s education (Afridi et al.
2016) and women’s empowerment (Khera and Nayak 2009; Sudarshan
2011; Desai et al. 2015).
We engage with those studies in this burgeoning literature on
MGNREGA that are linked closely to research questions regarding
programme participation. There is well-documented evidence that
welfare programmes suffer from the phenomenon of ‘elite capture’
(von Barun 1995; Barett and Clay 2003) and that elite capture can have
negative consequences (Besley et al. 2004). At the same time, studies
that examine the issue of targeting in welfare programmes suggest that
programmes that have self-selection and demand-driven features work
well, avoiding the problems associated with targeting (Besely and Coate
1992; Ravallion 2003).

111
Who Participates in MGNREGA?
In case of MGNREGA, evidence of e lit e capture in the allocation
of work has been documented. Jha et al. (2009), using primary data of
900 households from Andhra Pradesh and Rajasthan, show the capture of
MGNREGA by landed classes in Andhra Pradesh. Niehaus and Sukhtankar
(2012) also nd evidence of political clientelism at work. However, other
studies have argued that MGNREGA has been quite successful in targeting
marginalised sections of society. Ghosh (2009) argues that MGNREGA
involves more women, SCs and STs as workers. Deininger and Liu (2013)
offer evidence of pro-poor targeting of the programme. Sudarshan (2011)
also nds that in Kerala there has been some shift out of agriculture into
MGNREGA mainly for female workers because of the higher wages
paid under the programme. The administrative data released by the
Ministry of Rural Development (MoRD) and National Sample
Sur vey Office (NSSO) Survey Reports (NSSO 2011) also suggest
that MGNREGA is successful as a self-targeting programme, with a high
degree of participation from marginalised groups. At the national level, the
share of SCs and STs in the work provided under MGNREGA has been
high, at 40–50 per cent across each year of the scheme’s implementation.3
Not only have the marginalised sections been participating more in the
programme, but they have derived more benets as well. Several studies
have investigated the impact of MGNREGA on the welfare of the poor.
Berg et al. (2012) nd the wage effect of MGNREGA to be positive
across different implementation stages even after controlling for district
and time-xed effects, rainfall and the implementation phase. Klonner
and Oldiges (2014) nd large, season-specic effects among a traditionally
deprived sub-group of the rural population, whose incomes are particularly
dependent on agricultural wage labour. Ghosh (2009), Dreze and Khera
(2011), Sudarshan (2011) and Dutta et al. (2012) examine the welfare
gains of marginalised communities. They argue that since a majority of the
world’s poor live in rural areas and the poorest of the poor are agricultural
wage workers, a rural public work programme like MGNREGA in India
constitutes an important anti-poverty tool for all rural labour via both
its direct employment effects and its indirect wage effects.
The other important issue linked to participation is that of the
extent of participation and regional variation. The edited volume by
Khera (2011) and Dreze and Khera (2011) document several qualitative

112 Omkar JOshi, sOnalde desai, reeve vanneman and amaresh dubey
case studies, highlighting the positive impact of MGNREGA as well
as the challenges faced in implementing the Act. Two issues emerging
from this review are particularly noteworthy. First, MGNREGA has
become an important component of employment for rural Indians,
although it remains as a supplement to other work, with only 3.45
per cent of the households engaging in the full 100 days of work per
year (Ministry of Rural Development 2014). Second, the distribution
of MGNREGA participation is highly uneven across states. As far
as this uneven variation in MGNREGA participation across states is
concerned, Roy (2015) suggests a variety of factors – commitment of
local elites, geographical variations and political economy of programme
implementation – that are responsible for this. Chopra (2015) analyses
the puzzle of differing performance across states using a qualitative study
of the programme implementation in four Indian states (Bihar, Andhra
Pradesh, Chhattisgarh and Assam) and links varying performance to
differing political commitment in each state. Reddy et al. (2010) too,
in their study, nd that commitment, capacity and preparedness of local
governance structures impact the effectiveness of the programme.
Some studies point at corruption as the reason for lacklustre
implementation of MGNREGA and low participation. Kapur (2010) and
Niehaus and Sukhatankar (2012) report instances of underpayment of
wages on account of over-reporting and wage skimming by administrators.
Muralidharan et al. (2016) show the positive impact of biometrically
authenticated payment infrastructure on wage payments in MGNREGA.
The unevenness in programme participation would be a cause
for concern if richer households or richer villages were able to
disproportionately capture MGNREGA work, thereby leaving out the
poor. Dutta et al. (2012) examine the unmet demand for MGNREGA work
using the National Sample Survey (NSS) 66th Round data. They too
nd that MGNREGA participation rates vary across states, as observed by
Dreze and Khera (2011). They suggest that this variation is on account of
two effects: the rst is an indirect effect of greater poverty via higher demand
for MGNREGA work and the second is the direct effect of having greater
unmet demand for work on the programme. They show that rationing of
work takes place on the ground. However, they conclude that despite the
unmet demand of poor families and rationing, the self-targeting mechanism

113
Who Participates in MGNREGA?
of the Act works well, enabling it to reach relatively poor and backward
families. Narayanan et al. (2016) also highlight administrative rationing,
leading to ‘discouraging worker effect’ and variation in participation rates.
Public works programmes involve huge commitments of nancial
resources. This gives rise to the possibility of uneven distribution, namely
either richer households or richer villages disproportionately capturing
MGNREGA work, leaving out the poor. This poses serious challenges to
the programme design.
One problem encountered while examining the effectiveness of the
programme with regard to participation, in general, and that of SCs/STs
or women, in particular, is that programme participation is often affected
by factors correlated with caste and gender. For example, seasonal work
available through MGNREGA may be more attractive to marginal farmers
who do not have year-round work, but may hold little attraction for people
who have a steady job in a nearby town. However, both landownership as
well as farm productivity are correlated with caste (Desai et al. 2010).
Adivasis living in remote areas may have few other job opportunities,
while people belonging to forward castes living in the more developed
regions may have little need to rely on MGNREGA. In this case, the
higher participation of STs in MGNREGA should also be examined in
comparison to that of their forward-caste counterparts living in similar
areas in order to examine programme targeting at the local level.
Another important issue is that of the income status of the participant
households. Since MGNREGA provides employment and thereby income
generation for poor households, their current poverty and employment
status is endogenous to the issue of programme participation. Hence in
examining programme participation issues, it is necessary to look at the
income of the households before their participation in the MGNREGA
programme.
Also, even when a marginalised household participates in the
MGNREGA programme, we do not know who participates within
the household and whether their gender and age have any bearing on
participation, since the 100-day limit is operative at the household level.
Existing literature does not tell us anything about this intra-household
dynamics of participation.

114 Omkar JOshi, sOnalde desai, reeve vanneman and amaresh dubey
This paper lls the gap in the existing literature by addressing all
the above-mentioned issues related to participation by utilising a unique
panel data set from IHDS. We look at the issue of programme participation
holistically and compare participation of marginalised SC/ST households
to that of forward and other caste households. We also look at the pro-poor
targeting aspect of the programme by including the previous income of
the participating households, thereby avoiding the conation of current
income and current participation. Since there is evidence of regional
variation in programme participation, we also control for unobserved
state- and village-level heterogeneity. Lastly, we examine the choice of
intra-household participation by individuals, by looking at the gender and
age of individuals of the participating households. Thus, our results provide
a more robust picture of participation as compared to previous studies.
3. DATA
The data for this study was taken from the nationally representative multi-
topic IHDS. This panel survey was conducted during two rounds: IHDS-I
in 2004–05 and IHDS-II in 2011–12. IHDS-I and IHDS-II are part of a
collaborative research programme between the National Council of Applied
Economic Research (NCAER) and the University of Maryland with the
goal of documenting changes in the daily lives of Indian households in
the context of a society undergoing rapid transition. The surveys were
conducted in all the states and union territories (UTs) in India except
the UTs of Andaman and Nicobar Islands and Lakshadweep. IHDS thus
gathered detailed village-, household- and individual-level information
about a range of socio-economic and demographic variables, viz. income,
employment, consumption expenditure, education, gender relations, social
networks, marriage, youth, health and fertility.
The unique feature of the IHDS data set is that the same households
were visited during both rounds of the survey. This paper utilises the
household panel from both rounds of IHDS data covering about 42,000
households, of which about two-thirds are rural. We have included only
rural households for our analysis as MGNREGA is operational only in rural
areas. The aggregate re-contact rate of IHDS for the rural areas between
the two rounds is 90 per cent. There was a loss of 2,754 households due

115
Who Participates in MGNREGA?
to attrition in rural areas, but attrition does not signicantly affect the
representativeness of our sample.
Analyses of IHDS-I (Desai et al. 2010) and IHDS-II reveal that
on most major parameters like poverty, literacy and work participation,
the national rates derived by IHDS are comparable to those of NSS and
the Census. In the context of this paper, it would sufce to note that the
MGNREGA participation rate obtained by using the IHDS panel is
comparable to the participation rate obtained from NSS data (NSSO
2011; Dutta et al. 2012). Dutta et al. (2012) calculated the MGNREGA
participation rate to be 24.9 per cent during 2009–10 while the IHDS nds
it to be 24.4 per cent during the preceding year, 2011–12 (Table 1).
Table 1: MGNREGA participation and household characteristics
Household characteristics % of sample
households
% Participating in
MGNREGA
All India 100.0 24.4
Caste and ethnicity (2011–12)
Forward castes/others 23.3 17.0
OBCs 42.2 20.7
SCs 24.1 36.0
STs 10.4 28.8
Religion (2011–12)
Hindu 84.5 25.2
Muslim 9.8 21.1
Others 5.8 17.1
Income quintiles (2004–05)+
Lowest 20.4 27.7
2nd quintile 21.7 30.5
3rd quintile 21.1 26.8
4th quintile 19.2 22.0
Top 15.4 11.8
Contd..
+ Quintile calculations are based on rural income distribution.

116 Omkar JOshi, sOnalde desai, reeve vanneman and amaresh dubey
Household characteristics % of Sample
households
% Participating in
MGNREGA
Highest education of adult member in the household (2004–05)
None 28.2 33.0
Primary (1–4 std.) 9.5 29.7
Secondary (5–9 std.) 33.7 23.9
10–11 std. 12.4 17.4
12 std./some college 8.24 15.1
Graduate/diploma 8.0 10.0
Landownership (2004–05)
No landownership 44.5 25.4
Marginal (0–1 hectare) 32.8 26.5
Small (1–2 hectares) 10.8 21.8
Medium (2–5 hectares) 9.3 18.2
Large (5 and more hectares) 2.7 12.9
Household income source (2004–05) (Based on source of maximum
income)
Non-agricultural wage 19.1 29.1
Agricultural wage 21.6 35.6
Monthly salaried 13.1 16.3
Business 10.7 17.5
Farm cultivation 22.0 22.3
Animal care 5.5 15.5
Remittances/other income 8.1 17.0
Village infrastructure (2004–05)
More developed villages 46.1 20.4
Less developed villages 54.0 27.8
Contd..

117
Who Participates in MGNREGA?
Household characteristics % of Sample
households
% Participating in
MGNREGA
States
Jammu and Kashmir 1.2 16.7
Himachal Pradesh 0.8 39.0
Uttarakhand 1.8 29.6
Punjab 2.0 10.9
Haryana 1.9 5.4
Uttar Pradesh 15.6 20.8
Bihar 8.3 11.4
Jharkhand 4.6 7.7
Rajasthan 5.7 38.7
Chhattisgarh 3.3 60.5
Madhya Pradesh 5.5 25.8
North-east 1.2 34.7
Assam 2.4 19.9
West Bengal 8.6 44.9
Odisha 4.2 11.2
Gujarat 4.0 3.4
Maharashtra and Goa 7.4 2.9
Andhra Pradesh 8.3 47.4
Karnataka 4.7 13.2
Kerala 3.1 16.2
Tamil Nadu 5.4 41.4
Rural sample only. N=28,205.
Source: Indian Human Development Surveys 2005 and 2012.

118 Omkar JOshi, sOnalde desai, reeve vanneman and amaresh dubey
The IHDS data are well suited for analysis of MGNREGA for several
reasons, as listed below:
• Since the initial survey was conducted in 2004–05, just before the
initiation of MGNREGA, IHDS could examine the impact of the
initial household conditions in shaping programme participation.
• IHDS contains more detailed economic information than is
available in the NSS.
o IHDS collected not only consumption data as is done for
the NSS but also detailed income data. For instance, IHDS
collected separate data on income from farm and non-farm
business activity, enabling us to investigate the role of levels
and sources of income on MGNREGA participation.
o IHDS also collected data on household assets, a predictor
of long-term economic status (Filmer and Pritchett 2001),
which the analyses presented in later sections show to be the
strongest economic predictor of MGNREGA participation.
• The panel structure is particularly useful in avoiding problems
of endogeneity for evaluating the targeting of low-income
households. Cross-sectional analyses entail an additional risk
of confounding the effects of MGNREGA participation with
its causes.
• Unlike NSSO surveys, IHDS also contains information on village
infrastructure, allowing us to compare the role of MGNREGA
in the developed and less developed villages.
4. METHODOLOGY AND EMPIRICAL STRATEGY
This paper examines the issues of programme participation and success
of MGNREGA in providing opportunities to groups disadvantaged by
caste and religion, viz. SCs, STs; by economic level, viz. households with
low incomes and few assets prior to programme initiation; and by gender,
viz. women. Although MGNREGA doesn’t have specic provisions for
Muslims, we know that they too like the SCs and STs suffer from socio-
economic disadvantage. Hence, we look at their programme participation
as well. We analyse MGNREGA participation at both the household as

119
Who Participates in MGNREGA?
well as individual levels, taking into account the initial household income,
income composition and initial employment status of various household
members, particularly women.
For household-level analyses, we use a binary logit model with
the MGNREGA participation status of the household as the dependent
variable.4 The variables on the right-hand side include a mix of continuous
and categorical variables serving as indicators of household characteristics
such as assets, total income, educational qualications, sources of income,
caste and religion, among other things.
Targeting of disadvantaged households may occur in two ways: (1)
since the disadvantaged households often live in poor areas, geographic
targeting (as with the phased implementation of MGNREGA) may
indirectly help the disadvantaged households and (2) there may be direct
targeting of marginalised households within any given area.
In order to examine the targeting, we undertake three sets of analyses
at the household level:
(1) First, we estimate models of whether a household participates in
MGNREGA at an all-India level without including any state-level control
variables. This allows us to examine total targeting of disadvantaged
households at an all-India level. The basic specication takes the following
form:
Pr (MGNREGAi)/Pr (1 – MGNREGAi) = exp(α0 + αjXij + βkXik)
(1)
where the probability of participation in MGNREGA by the i-th household
in 2011–12 is a function of j variables measured in 2004–05 and k
variables measured in 2011–12. Since economic status is both a function
of and determinant of MGNREGA participation, the variables denoting
economic status – household income, sources of income, ownership of
consumer assets and education level – are drawn from IHDS-I conducted
in 2004–05 before MGNREGA was implemented. Household structure
variables – the number of adults in the household and social group – are
drawn from the 2011–12 survey.
(2) Then, we model the effect of state of residence and whether
the village is more developed or less developed in order to account for
geographical variation. For India, as a whole, disadvantaged households

120 Omkar JOshi, sOnalde desai, reeve vanneman and amaresh dubey
may enjoy greater access to MGNREGA because they live in disadvantaged
states or in poorer villages within those states. Comparisons of the
household coefcients in the two models will reveal how much of the
targeting was accomplished simply by focusing on the poorer states and
less developed villages. This model is specied as:
Pr (MGNREGAi)/ Pr (1 – MGNREGAi) = exp (α0
+ β jXij
+ β kXik
+
βsXis)
(2)
where s denotes state of residence and whether the village can be classied
as less developed or more developed. Less developed villages are dened
as those that have six or fewer infrastructure facilities from a list of 10
facilities listed in Table A.1.
(3) However, our controls for village characteristics and state of
residence do not encompass the full range of potential differences between
different geographical areas and individual characteristics may act as a
proxy for these regional differences. For example, it is well recognised
that STs are often located in poorer districts and areas without good
transportation access. In order to control for this geographical diversity
and look at differences by caste and income within the village, we estimate
a village-level xed effects model:
Pr (MGNREGAiv)/Pr (1–MGNREGAiv) = exp(α0+βjXij+βkXik
+
βsXis
+
μv)
(3)
where μv reects the intercept for each village. This is equivalent to adding
a dummy variable for each village. These models are estimated by using
STATA command xtlogit with a village-level xed effects specication.
The village-level xed effects model allows us to control all
unobserved vi llag e characteristics to focus only on the impact of
household characteristics within the villages. There may still be some
elite capture within some villages where the more powerful households
manage to secure the limited work available. The xed effects model
tests whether, even within a village, the more disadvantaged households
have greater access to MGNREGA work.
Since households are constrained to a total of 100 days per year of
work in MGNREGA, gender, age and marital status may determine which
person within the household is chosen to participate in the programme.

121
Who Participates in MGNREGA?
As MGNREGA pay does not allow gender or age wage differentials, we
expect women, the elderly and the very young to be disproportionately
selected by households for MGNREGA work. This would then free prime-
age males, who are better paid in the market economy, to seek market
work. These age and gender differences have been estimated by using a
household-level xed effects regression.
Pr (MGNREGApi)/Pr (1 – MGNREGApi) = exp (α0 + βpXpi + μi)
(4)
where the probability of MGNREGA participation for person p in
household i is a function of age, education, gender and marital status
of the individual within the household. This framework is similar to the
village-level xed effects model estimated in Equation (3).
4.1 Description of variables
One of the strengths of the IHDS survey is the wide range of social and
economic data collected for each household. In our analytic models, we
test for participation along several dimensions of social and economic
disadvantages while holding constant other household and individual
characteristics that might also inuence MGNREGA participation. We
focus on the following sets of variables:
Social background: The caste group (forward castes, other backward
classes (OBCs), SCs and STs) and religion (Hindu, Muslim or other)
of the household head.
Long-term economic status as measured in 2004–05: A count of
up to 30 assets owned by the household, including possessions
ranging from basic assets like a table or a chair to more modern and
expensive possessions like a refrigerator or a washing machine.
Household assets reect the underlying economic well-being of
that particular household, as possessions accumulated over many
years shed better light on a household’s long-term economic standing
than do annual measures like income, which tend to be volatile.
Assets also constitute the economic measure that is most likely to
be visible to others in the village, which could act as a deterrent
to MGNREGA participation.

122 Omkar JOshi, sOnalde desai, reeve vanneman and amaresh dubey
Household income and occupation as measured in 2004–05:
Household income quintile; sources of household income,
including participation in wage labour, farming and business; and
landownership.
Household demographics: The number of adult members in the
household and the highest level of education attained by any adult
household member.
Individual characteristics: Sex, age and marital status of the
individual members of the household.
5. RESULTS AND DISCUSSION
In this section, we present the results of our analyses. We begin by
highlighting the participation across different socio-religious and economic
axes in MGNREGA. This is followed by different variants of econometric
models described above.
5.1 Descriptive statistics
Descriptive statistics for all household variables are included in
Table 1. The nal column of the table shows the proportion of each category
participating in MGNREGA. The overall household participation rate is
24.4 per cent, but Table 1 shows that participation varies widely by caste,
religion, income level and other household characteristics. Even wider
variation by states is seen, with only 2.9 per cent of the rural households
in Maharashtra and Goa reporting MGNREGA participation as compared
to a corresponding gure of 60.5 per cent in rural Chhattisgarh. However,
it is important to note that the smaller IHDS samples within each state
mean larger standard errors for these estimates.
All indices of social or economic position show greater MGNREGA
participation among the least privileged. SC and ST households show
higher participation rates than OBC households, which, in turn, show
slightly more participation than forward-caste households. Households
from the bottom three income quintiles participate in the programme at
equivalent but higher rates (about 30 per cent) than the top two quintiles.
In that sense, MGNREGA is proving to be an important anti-poverty

123
Who Participates in MGNREGA?
programme as it attracts more poor than rich households. However, its
appeal is broader than its appeal to the poor alone, as middle-income
households also participate in MGNREGA work at signicant rates.
In addition, landless households have higher participation rates (25.4 per
cent) than large landowners (12.9 per cent). Households depending on farm
or non-farm wage labour exhibit higher participation rates than those with
salaried incomes. Illiterate households show higher participation rates than
those with adults who are graduates or have acquired secondary school
level education. Work offered under MGNREGA is low-skill, manual
labour. Given this feature of the programme, we note that education,
work and income variables have the intended distribution as shown in
Table 1 follow the intended lines.
Although descriptive statistics are suggestive, econometric analysis
is needed to explore the relationship between participation choice and
other variables. Therefore, we run a logistic regression model regressing
actual participation on the full range of variables.
5.2 Determinants of household-level participation
First, we estimate a household-level logistic regression model for all
households given their initial household-level characteristics (that is, in
2005, before MGNREGA began). Table 2 gives the coefcients for a logistic
regression at the all-India level, indicating the likelihood of MGNREGA
participation. Here the economic variables of direct interest are consumer
durables owned in 2004–05 and income in 2004–05. The asset coefcient
is negative and signicant; households with a relatively large asset base
do not participate in the kind of work offered by MGNREGA.
Table 2 also delineates the relationship between household income
in 2004–05 and MGNREGA participation. Here the results are intriguing.
Although families from the top-income quintile are least likely to participate
in MGNREGA (even after holding constant their asset level), the bottom
80 per cent are more or less equally likely to participate.
There is a strong link between levels of education and poverty. A better
educated household has more job opportunities and is in a better position
to escape poverty. Since MGNREGA offers only casual unskilled labour
work on a temporary basis, a less educated household is more likely

124 Omkar JOshi, sOnalde desai, reeve vanneman and amaresh dubey
Table 2: Logistic regression results (household level)
MGNREGA participation status Coefcients Standard error
Round II (2011–12) Variables
No. of adults 0.050** 0.016
Caste and ethnicity (‘forward castes’ omitted)
OBCs 0.053*** 0.061
SCs 0.652*** 0.068
STs 0.301*** 0.082
Religion (‘Hindus’ omitted)
Muslims –0.052 0.075
Others –0.450*** 0.105
Round I (2004–05) Variables
No. of consumer durables owned –0.046*** 0.006
Household income in 2004–05 (bottom quintile omitted)+
2nd quintile 0.141** 0.062
3rd quintile 0.104 0.065
4th quintile 0.082 0.070
Top quintile –0.271*** 0.091
Education level of the household (no education omitted)
Primary (1–4 std.) –0.064 0.071
Secondary (5–9 std.) –0.210*** 0.054
10–11 std. –0.381*** 0.081
12 std./some college –0.470*** 0.102
Graduate/diploma –0.748*** 0.119
Sources of income (farm/own cultivation omitted)
Non-agricultural wage 0.004 0.065
Agricultural wage 0.229*** 0.063
Salary –0.225** 0.082
Business –0.252** 0.088
Animal care –0.507*** 0.106
Remittances/other –0.359** 0.103
Constant –0.879*** 0.099
Sample size (N households) 28,129
***<0.001, **<0.01, *<0.5 and + 0.5 to 0.05 respectively.
+ Quintile calculations are based on rural income distribution.
Source: Authors’ calculations from IHDS I and II.

125
Who Participates in MGNREGA?
to turn to MGNREGA as a source of employment even holding constant
their economic level. In Table 2, the education effect appears to be quite
linear with lower levels of participation seen for people with each higher
level of education.
MGNREGA operates in rural areas that remain predominantly
agrarian even today despite a falling share of agriculture in national
income and employment. Since agricultural activities are seasonal in
nature, it is vital for the workforce employed in agriculture to have access
to alternative channels of employment. The regression results show that
households with agricultural wage workers are more likely to participate
in MGNREGA than households who cultivate their own land.5 It thus
appears that MGNREGA does act as a source of alternative employment
and provider of safety net.
Historically, SCs and STs have lagged behind in overall development
and hence have rightly been t h e focal point for this programme. As
compared to the forward-caste households, SC and ST households are
characterised by less education, income and household assets and are thus
favoured for targeting in this programme, which is essentially meant for
poor and less educated households. The results in Table 2 show higher
S C an d S T participation even beyond these economic and educational
characteristics. This conrms that MGNREGA is succeeding as a self-
targeting programme for the most disadvantaged communities.
In contrast to Dalits and Adivasis, Muslims have no signicantly
different participation rates from economically equivalent majority Hindus.
More detailed analyses, not shown here, have shown that Christians tend
to have somewhat higher participation rates and Sikhs lower participation
rates, but these differences are entirely a function of the high participation
rates in the North-East and low rates in Punjab.
5.3 State and village xed effects logistic models
The ndings depicted in Table 2 thus clearly indicate that participation
in MGNREGA is largely from the marginalised groups. However, some of
these results might be due to the strong geographical variation noted in Table 1,
which is not captured by this model. For example, since caste and income
vary by place of residence and its economic opportunity structure, the

126 Omkar JOshi, sOnalde desai, reeve vanneman and amaresh dubey
successful targeting of disadvantaged households may derive primarily
from targeting disadvantaged regions and villages. Regional targeting could
be particularly important for Adivasis since they are most likely to live
in remote areas where few alternative job opportunities exist but which
were earlier targeted for inclusion in MGNREGA. Adivasis integrated
into more developed, mixed-caste rural areas, on the other hand, might not
enjoy the same access to MGNREGA employment. In order to estimate
how much of the targeting is only a result of the geographical targeting of
poorer states and villages, we calculate two additional models that control
rst for state-level and then village-level differences.
Interestingly, instead of attenuating caste effects, the geographic
controls have little or no impact on the estimates of participation levels
by Dalits and Adivasis. State-level controls in Table 3 actually slightly
raise the estimates of Adivasi participation while slightly lowering the
estimates for Dalits. The more extensive village controls reverse those
changes but leave the estimates almost equal or even somewhat stronger
than in the initial estimates from Table 2. This indicates that the higher
participation rate of Dalits and Adivasis is not a function of Dalit and Adivasi
areas getting more programme access but rather of greater programme
participation even in mixed locations.
Similar results are also observed for the estimates of economic
effects. Both asset levels and income coefcients become larger after
state-level controls are added. This suggests that contrary to the initial
impressions, MGNREGA participation levels are not especially high in
the poor states. Like the poorer states of Chhattisgarh and Rajasthan, the
better-off states of Tamil Nadu, Andhra Pradesh, and the North-East also
register high participation rates. States like Jharkhand, Bihar and Odisha,
while showing moderate levels of MGNREGA participation (see Table
1), have negative coefcients once household-level economic standing
is controlled for. Hence, rather than low-income households having high
MGNREGA participation rates because they are located in poor states,
it appears to be more likely that within each state, it is the low-income
households that participate.

127
Who Participates in MGNREGA?
Table 3: State and village xed effects logistic models
MGNREGA
participation status
State xed effects++ Village xed effects++
Coefcient Standard
error Coefcient Standard
error
Round II (2011–12) variables
No. of adults 0.137*** 0.018 0.166*** 0.015
Caste and ethnicity (forward caste omitted)
OBC 0.142* 0.069 0.260*** 0.068
SC 0.587*** 0.074 0.740*** 0.072
ST 0.486*** 0.098 0.335** 0.102
Religion (Hindus omitted)
Muslims –0.182* 0.089 –0.250* 0.110
Others –0.132 0.136 –0.226 0.143
Round I (2004–05) variables
No. of consumer
durables owned –0.088*** 0.008 –0.076*** 0.007
Household income in 2004–05 (bottom quintile omitted)+
2nd quintile 0.096+0.049 0.112+0.060
3rd quintile –0.005 0.051 0.059 0.063
4th quintile –0.087 0.056 –0.056 0.069
Top quintile –0.424*** 0.074 –0.381*** 0.086
Highest adult education level of the household in 2004–05
(no education omitted)
Primary (1–4 std.) 0.024 0.081 –0.039 0.073
Intermediate
(5–9 std.) –0.109+0.060 –0.108* 0.052
10–11 std. –0.189* 0.092 –0.250** 0.075
12 std./
some college –0.203+0. 111 –0.296** 0.090
Graduate/diploma –0.596*** 0.126 –0.646*** 0.108
Contd..

128 Omkar JOshi, sOnalde desai, reeve vanneman and amaresh dubey
5.4 Individual-level xed effect logistic model
In addition to providing livelihood security for households, MGNREGA
contains a particular focus on women and is expected to ensure that 33
per cent participants are women. In order to examine the success of this
objective, we run th e household-level xed effects model to stu dy t he
intra-household dynamics of MGNREGA participation. For this analysis,
we consider individual demographic characteristics, viz. gender, age and
marital status. The data utilised for individual-level regression has been
taken from MGNREGA-participating households in the second round of
the survey (IHDS-II), and the results are presented in Table 4.
MGNREGA
participation status
State xed effects++ Village xed effects++
Coefcient Standard
error Coefcient Standard
error
Sources of income in 2004–05 (farm/own cultivation omitted)
Non-agricultural
wage –0.033 0.074 0.013 0.067
Agricultural wage 0.172* 0.076 0.252*** 0.066
Salary –0.335*** 0.091 –0.266** 0.079
Business –0.143 0.096 –0.175* 0.084
Animal care –0.456*** 0.113 –0.133 0.100
Remittances/other –0.351** 0.111 –0.493*** 0.094
Village infrastructure in 2004–05 (more developed villages omitted)
Less developed
villages 0.375*** 0.053 -
Constant –1.297 0.131 -
Sample size
(N households) 27,909 18,463
Note: State dummies are included in the state-xed effects model but are not reported.
***<0.001, **<0.01, *<0.5 and + 0.5 to 0.05 respectively.
+ Quintile calculations are based on rural income distribution.
++ Fixed effects models estimated only with cases that contain variation in MGNGREGA
participation within the second level unit.
Source: Authors’ calculations from IDHS-I and II.

129
Who Participates in MGNREGA?
It is necessary to compare the individual MGNREGA participation
results with similar results for overall work participation. We need to understand
not only which individuals participate in MGNREGA, but whether such participation
rates are greater or lesser than general work participation in those households.
While women are less likely than men to participate in MGNREGA, this
Table 4: Household-level xed effect++ logistic model: MGNREGA and overall
work participation
MGNREGA
participation
Overall work
participation
Coefcients Standard
error Coefcients Standard
error
Round II (2011–12) variables
Gender
Female –0.550*** 0.031 –2.183*** 0.026
Age category (18–29 years) (Omitted)
30–39 0.925*** 0.057 1.320*** 0.040
40–59 0.969*** 0.051 1.464*** 0.034
60–75 –0.003 0.071 –0.374*** 0.042
76 and above –1.767*** 0.222 –2.489*** 0.086
Marital status (currently married omitted)
Never married –0.812*** 0.069 –1.825*** 0.038
Widowed –0.405*** 0.080 –0.869*** 0.048
Separated/
divorced –0.590*** 0.202 –0.476** 0.147
Married, no gauna –1.049** 0.456 –0.176 0.272
Sample size
(N individuals) 19,432 60,226
Note: ***<0.001, **<0.01, *<0.5 and + 0.5 to 0.05 respectively.
+ Quintile calculations are based on rural income distribution.
++ Fixed effects models estimated only with cases that contain variation in MNGREGA participation
within second level unit.
Source: Authors’ calculations from IDHS-I and II.

130 Omkar JOshi, sOnalde desai, reeve vanneman and amaresh dubey
disadvantage is smaller than for overall work participation. The female
coefcient is –0.55 for MGNREGA participation but –2.18 for overall
work participation. This conrms that f e m a le w o r ker s f i n d
g r e a te r f a vo u r i n MGNREGA work than in other types of work.
This could be because MGNREGA is the only sector where wage equality
is enforced; in all other work, women earn substantially less than men.
Another explanation for the higher female participation in MGNREGA
could be the 33 per cent reservation for female workers in the programme.
The second variable of interest is the age of individual MGNREGA
workers. MGNREGA participation shows an inverse U relationship with
the youngest and oldest workers least likely to participate in it. However,
once again, when compared to the inverse U relationship with overall
work participation in the panel, we see that for MGNREGA work, the
inverse U is more muted. This suggests that while the youngest and oldest
workers are left out of the other sectors, MGNREGA is more open to
them. Similarly, though married individuals are most likely to participate
in MGNREGA, widows/widowers face fewer disadvantages in MGNREGA
work as compared to other types of work.
These results are interesting in that they highlight ways in which
labour market opportunities shape household work participation decisions.
Adult males who have access to better labour force opportunities than
women are not as attracted to MGNREGA as are women and the elderly
who have limited opportunities in other kinds of work.
6. CONCLUSION
We began by asking which households are more likely to participate in
the MGNREGA programmes, what their characteristics are and what the
intra-household dynamics are with regard to participation patterns. Our
results provided an intriguing answer. We found that for the categories of
households who otherwise are excluded socially and for those which the Act
denes as being marginalised, that is, Dalits and Adivasis (SCs and STs),
participation rates are higher. This relationship persists even when we
controlled for prior income and consumer assets, sources of income and
place of residence, all of which are themselves related to MGNREGA
participation. Women are not quite on par with men in MGNREGA

131
Who Participates in MGNREGA?
participation but the disadvantage they faced is much less than what they
experienced in ot her types of wor k. However, higher participation rates are not
observed for groups that have not been explicitly identied, like Muslims.
Like the targeting of socially excluded groups, the economic targeting
of MGNREGA seems to have been moderately effective, though there is
a broader band of people falling in the low to middle economic groups
who have taken advantage of MGNREGA. While the highest income and
education groups seem to self-select themselves out of MGNREGA work,
there is little gradation among the population in the bottom 60 per cent.
This actually speaks about the broad-based support that MGNREGA as a
programme enjoys. Moreover, MGNREGA work seems most attractive to
agricultural wage labourers, who are familiar with the hard, manual labour
offered by MGNREGA.
Although the paper conrms some of the observations made using
other data-sets and that of micro-level studies, it strengthens these earlier
observations by reinforcing those ndings and adding to them using a
longitudinal data set that allows us to look at before and after MGNREGA
incomes for the same set of households These results provide a useful
starting point given current discourses around reforming MGNREGA.
Concerns about the high costs of the social programme and a belief that
MGNREGA has failed to reach the poorest drives the call for reforms in
MGNREGA (Dutta et al. 2014). In this paper, we offer somewhat different
evidence and suggest that despite being plagued by various deciencies,
MGNREGA’s achievements are moderately effective in terms of its
programme participation and that it would be worthwhile to build on this
success instead of reinventing the wheel.
NOTES
1 The National Rural Employment Guarantee Act (NREGA), 2005, was later renamed
as the Mahatma Gandhi National Rural Guarantee Act (MGNREGA).
2 This focus on the vulnerable population was enhanced through phased
implementation of the programme in 2006 with the rst 200 districts being chosen
on the basis of their backwardness.
3 See www.mgnrega.nic.in and MoRD, 2012.
4 Although there are differences of opinions among researchers about suitability
of logit models, for a binary dependent variable such as the present (whether

132 Omkar JOshi, sOnalde desai, reeve vanneman and amaresh dubey
household participates or not), logit models are better suited unless there is a
compelling reason to use a LPM model.
5 We also use landownership as a control in regression in Table A.2. It shows that
only large landowners opt themselves out of participation in MGNREGA.
REFERENCES
Afridi, F., A. Mukhoopadhaya and S. Sahoo. 2016. Female Labor Force Participation
and Child Education in India: Evidence from the National Rural Employment
Guarantee Scheme. IZA Journal of Labour & Development. 5:7.
Azam, M. 2012. “The Impact of Indian Job Guarantee Scheme on Labor Market
Outcomes: Evidence from a Natural Experiment”. IZA Discussion Paper 6548.
Bonn, Germany: Institute for the Study of Labor.
Barrett, C.B. and. C. Clay. 2003. “Self-Targeting Accuracy in the Presence of Imperfect
Factor Markets: Evidence from Food-for-Work in Ethiopia. Journal of Development
Studies 39(5): 152–180.
Berg, E., S. Bhattacharya, D. Rajasekhar and R. Manjula. 2012. Can Workforce
Increase Wages? Evidence from a Natural Experiment in India. WPS/2012-5.
Oxford: Centre for the Study of African Economies Working Papers.
Besley, T. and S. Coate. 1992. Workfare versus Welfare: Incentive Arguments for
Work Requirements in Poverty-Alleviation Programs. American Economic Review
82(1): 249–261.
Besley, T., R. Pande, L. Rahman and V. Rao. 2004. The Politics of Public Good Provision:
Evidence from Indian Local Governments. Journal of the European Economic
Association, 2(2–3): 416–426.
Bhargava, A. 2014. The Impact of India’s Rural Employment Guarantee on Demand
for Agricultural Technology. Working Paper. Davis, CA: University of California.
Centre for Budget and Governance Accountability. 2017. What Do the Numbers Tell?
An Analysis of Union Budget 2017–18. New Delhi.
Chopra, D. 2015. Political commitment in India’s social policy implementation: Shaping
the performance of MGNREGA. ESID Working Paper No. 50. Manchester, UK:
University of Manchester.
Das, U. 2015. Rationing and Accuracy of Targeting in India: The Case of the Rural
Employment Guarantee Act. Oxford Development Studies.
Deininger, K. and Y. Liu. 2013. Welfare and Poverty Impacts of India’s National
Rural Employment Guarantee Scheme: Evidence from Andhra Pradesh. IFPRI
Discussion Paper 01289. Washington, D.C.
Desai, S., A . Dubey, B . Joshi, M . Sen, A . Shariff and R . Vanneman. 2010.
Human Development in India: Challenges for a Society in Transition. New Delhi:
Oxford University Press.

133
Who Participates in MGNREGA?
Desai, S., P. Vashishtha and O. Joshi. 2015. Mahatma Gandhi National Rural
Employment Guarantee Act: A Catalyst for Rural Transformation. New Delhi:
National Council of Applied Economic Research.
Dreze, J. and R. Khera. 2011. The Battle for Employment Guarantee, in R. Kh er a
(Ed.), The Battle for Employment Guarantee. N ew Delhi: Oxford University
Press, pp. 43–81.
Dutta, P., R. Murgai, M. Ravallion and W.V. Dominique. 2012. Do es India’s Employment
Guarantee Scheme Guarantee Employment? Policy Research Paper. Washington
DC: World Bank.
Dutta, P., R. Murgai, M. Ravallion and W.V. Dominique. 2014. Right to Work? Assessing
India’s Employment Guarantee Scheme in Bihar. Washington, DC: World Bank.
Filmer, D. and L. Prichett. 2001. Estimating Wealth Effects without Expenditure
Data—Or Tears: An Application to Educational Enrollments in States of India.
Demography 38(1): 115–132.
Ghosh, J. 2009. Equity and Inclusion through Public Expenditure: The Potential of
the NREGA. Progoti.
Imbert, C., and J. Papp. 2015. Labor Market Effects of Social Programs: Evidence
from India’s Employment Guarantee. American Economic Journal: Applied
Economics 7(2): 233–263.
Jha, R., S. Bhattacharyya, R. Gaiha and S. Shankar. 2009. Capture of Anti-Poverty
Programs: An Analysis of the National Rural Employment. Journal of Asian
Economics, 20(4): 456–464.
Kapur, D. 2010. Political Economy of the State. In N.G. Jayal and P.B. Mehta (Eds.),
The Oxford Companion to Politics in India. Oxford: New Delhi.
Khera, R. (Ed.) (2011). The Battle for Employment Guarantee. Oxford: Oxford
University Press.
Khera, R. and N. Nayak. 2009. Women Workers and Perceptions of the National
Rural Employment Guarantee Act. Economic and Political Weekly 44(43): 49–57.
Klonner, S. and C. Oldiges. 2014. Can an Employment Guarantee Alleviate Poverty?
Evidence from India’s National Rural Employment Guarantee Act. Discussion
Paper Series No. 564, Heidelberg: University of Heidelberg, Germany.
Liu, Y. and Barrett, C. 2012. Heterogeneous Pro-Poor Targeting in India’s National Rural
Employment Guarantee Scheme. Economic and Political Weekly 48(10): 46–53.
Ministry of Rural Development. 2012. MGNREGA Sameeksha, An Anthology of
Research Studies on the Mahatma Gandhi National Rural Employment Guarantee
Act, 2005, 2006–2012. Edited and compiled by Mihir Shah, Neelakshi Mann and
Varad Pande. New Delhi: Orient Blackswan.
Ministry of Rural Development. 2014. Mahatma Gandhi National Rural Employment
Guarantee Act, 2005 – Report to the People. New Delhi: Government of India.
Muralidharan, K., P. Nielhaus and S. Sukhtankar. 2016. Building State Capacity:

134 Omkar JOshi, sOnalde desai, reeve vanneman and amaresh dubey
Evidence from Biometric Smartcards in India. American Economic Review
106(10): 2895–2929.
Narayanan, S., U. Das, Y. Liu and C. Barrett. 2016. The Discouraged Worker Effect
in Public Works Programs: Evidence from the MGNREGA in India. IFPRI
Discussion Paper 01633.
Niehaus, P. and Sukhtankar, S. 2013. The Marginal Rate of Corruption in Public
Programs: Evidence from India. Journal of Public Economics 104: 52–64.
Novotny, J., J. Kubelkova and V. Joseph. 2013. A Multi-Dimensional Analysis of the
Impacts of the Mahatma Gandhi National Rural Employment Guarantee Scheme: A
tale from Tamil Nadu. Singapore Journal of Tropical Geography 34(3): 322–341.
NSSO (National Sample Survey Organisation). 2011. Employment and Unemployment
Situation in India, 2009–10, NSS 66th Round. New Delhi: Government of India.
Ravallion, M. 2003. Targeted Transfers in Poor Countries: Revisiting the Tradeoffs and
Policy Options, Vol. 3048. Washington, DC: World Bank Publications.
Reddy, D., R. Tankha, C. Upendranadh and A. Sharma. 2010. National Rural
Employment Guarantee as Social Protection. IDS Bulletin 41(4): 63–76.
Roy, I. 2015. Class Politics and Social Protection: The Implementation of India’s
MGNREGA. ESID Working Paper No. 46, University of Manchester, UK.
Sarkar, A. and J. Islary. 2017. Wage and Earnings from Participation in MGNREGA
Works in Jharkhand. International Journal of Rural Management 13(1): 20–53.
Sudarshan, R.M. 2011. India’s National Rural Employment Guarantee Act: Women’s
Participation and Impacts in Himachal Pradesh, Kerala and Rajasthan. Sussex,
UK: Institute of Development Studies (IDS).
Varshney, D., D. Goel and J. Meenakshi. 2014. The Impact of MGNREGA on Cropping
Patterns, Wages and Labour Use. Working Paper. Delhi: Delhi School of Economics.
von Braun, J. (Ed.) 1995. Employment for Poverty Reduction and Food Security.
Washington, DC: International Food Policy Research Institute.
World Bank. 2015. The State of Social Safety Nets 2015. Washington, DC: World Bank.
Zimmermann, L. 2015. Why Guarantee Employment? Evidence from a Large Indian
Public Works Program. Working Paper. Georgia, USA: University of Georgia.

135
Who Participates in MGNREGA?
Name of the variable Details
Variables from Round II (2011-12)
No. of Adults (age>18) in the
Household
Mean = 2.86
SD = 1.40
Caste and Ethnicity Forward castes, OBCs, SCs, STs
Religion Hindus, Muslims, Others
Variables from Round I (2004–05)
Number of consumer durables
owned by household (0-30)
Mean =9.97
SD = 5.24
Household income Household Income Quintile Class
Education level category of the
household based on the highest
level of adult education
No education, primary (1–4 std.), secondary (5–9
std.), 10–11 std., 12 std./some college, Graduate/
Diploma
Main source of income for
household
Non-agricultural wages, agricultural wages,
farming/own cultivation, salary, business, animal
care, remittances/other
Village infrastructure
Type of village (more developed/less developed)
based on access to 10 infrastructure facilities,
namely electricity, paved road, grocery shop, bus
stop, landline and mobile telephone, post ofce,
police station, markets and bank
Type of farmers based on land
ownership
None, marginal (0–1 hectares), small (1–2
hectares), medium (2–5 hectares), large (5
hectares and more)
Gender Male, female
Age category 18–29 years, 30–39 years, 40–59 years, 60–75
years, 76 years and above
Marital status Currently married, never married, widowed,
separated/divorced, married, No gauna
Appendix A
Table A.1: Variables – Denitions and details
Source: Authors’ calculations from IHDS-I and II.

136 Omkar JOshi, sOnalde desai, reeve vanneman and amaresh dubey
MGNREGA participation status Coefcient Standard error
Round II (2011-12) Variables
No. of adults 0.136*** 0.017
Caste and ethnicity (forward castes omitted)
OBCs 0.155* 0.069
SCs 0.672*** 0.074
STs 0.524*** 0.098
Religion (Hindus omitted)
Muslims –0.135 0.088
Others –0.116 0.139
Round I (2004–05) Variables
No. of consumer durables owned in 2004–05 –0.097*** 0.007
Household income in 2004–05 (bottom quintile omitted)+
2nd quintile –0.406a0.208
3rd quintile 0.112 0.069
4th quintile –0.016 0.071
Top quintile –0.120*** 0.076
Education level of the household in 2004–05 (no education omitted)
Primary (1–4 std.) 0.004 0.080
Secondary (5–9 std.) –0.143* 0.060
10–11 std. –0.243** 0.091
12 std./some college -–0.275* 0.110
Graduate/diploma -–0.703*** 0.127
Village infrastructure in 2004–05 (more developed villages omitted)
Less developed villages 0.356*** 0.053
Type of farmers
Marginal (0–1 hectares) 0.209*** 0.056
Small (1–2 hectares) 0.233** 0.078
Table A.2: Logistic regression with landownership (household level)

137
Who Participates in MGNREGA?
Note: State dummies are included in the model specication but are not reported.
***<0.001, **<0.01, *<0.5 and + 0.5 to 0.05 respectively.
+ Quintile calculations are based on rural income distribution.
++ Fixed effects models estimated only with cases that contain variation in MNGREGA participation
within second level unit.
Source: Authors; calculations from IHDS-I and II.
Medium (2–5 hectares) 0.277** 0.090
Large (5 and more hectares) –0.094 0.187
Constant –1.460*** 0.124
Sample size 27,909

STATEMENT about ownership and other particulars of the newspaper
REVIEW OF DEVELOPMENT AND CHANGE to be published as
required under Section 19D(b) of the P.R.B. Act read with Rule 8 of the
Registration (Central) Rules, 1956.
Form IV
(See Rule 8)
1. Place of Publication : Chennai
2. Periodicity of its publication : Half-yearly
3. Printer’s Name : Mr S. Baskarapandian
(Whether citizen of India) Indian
Address 112/2, Bells Road
Triplicane, Chennai 600 005
4. Publisher’s Name : Shashanka Bhide
(Whether citizen of India) Indian
Address 79, Second Main Road
Gandhi Nagar, Adyar
Chennai 600 020
5. Editor’s Name : Shashanka Bhide
(Whether citizen of India) Indian
Address : 79, Second Main Road
Gandhi Nagar, Adyar
Chennai 600 020
6. Name and addresses of : Madras Institute of
individuals who own the Development Studies
newspapers and partners or 79, Second Main Road
shareholders holding more Gandhi Nagar, Adyar
than one per cent of the Chennai 600 020
total capital
I, Shashanka Bhide, hereby declare that the particulars given above are
true to the best of my knowledge and belief.
Dated: March 2017 Sd/- Shashanka Bhide
Signature of Publisher

Guidelines for Contributors
1. Papers not already published nor meant for publication elsewhere should be
sent in a word document to editrdc@gmail.com.
2. Articles should be between 8,000 and 10,000 words including tables, charts,
notes and references. Book Reviews should be 1,000 to 1,500 words. But
discretion rests with the editors. The articles should be typed in Times New
Roman Font size 12 and 1.5 line spacing.
3. Each paper must have a title page which will carry the full title of the paper,
the name and address of the author, and institutional afliation, if any, and
should provide a contact phone number and email address. The title page
must also give an abstract of the paper (100 to 150 words), and may include
acknowledgments. The author should not be identied anywhere else in the
paper. The page starting the text of the paper should carry only the title of the
paper.
4. All tables and diagrams should be formatted in black and white. Data used for
tables and graphs must be provided in Excel. All tables and diagrams must
carry numbers for identication and must be clearly referenced in the article.
5. Authors should ensure that the paper is properly edited for language and
consistency before submission.
6. Reference to sources/literature cited should be carried within the text in
brackets giving the name of the author, year of publication and page number,
e.g. (Basu 1967: 35). Notes and list of references (bibliography) must appear
at the end of the text. References should be listed alphabetically by author and
chronologically for each author. Some examples are given below:
Rao, S.R. and J.W. Townsend. 1998. Health needs of elderly Indian women: an
emerging issue, In Krishnaraj, M., Sudarshan R.M. and Shariff, A. (eds),
Gender, Population and Development, New Delhi: Oxford University
Press, pp. 340–363.
Sundararaman, T., I. Mukhopadhyay and V.R. Muraleedharan. 2016. No respite
for public health, Economic and Political Weekly, 51 (16): 39–42.
Tout, K. 1989. Ageing in Developing Countries. Oxford: Oxford University
Press.
7. The journal follows British English spelling (with -ise variant), not American
(e.g. programme, not program; labour, not labor). The Oxford Advanced
Learner’s Dictionary is to be followed for spellings.
8. Quotation marks should be consistently single, except for a quote within a
quote: e.g. Sen summed it up best by saying: ‘The importance of capital in
the production process notwithstanding, a distinction must be made between
“foreign” and “domestic” capital.’
9. Authors will receive a PDF le of their published article.

Regd. No.65834/1996
Review of Development and Change is published twice a year.
Subscription covering two issues of a volume is as follows:
India and other SAARC Countries
Individuals Rs.300
Institutions Rs.500
Foreign (by Air mail)
Individuals US $50 / £40
Institutions US $100 / £80
Account payee Cheques/Drafts to be drawn in favour of
Director, Madras Institute of Development Studies
Published by Shashanka Bhide for Madras Institute of Development Studies,
79, Second Main Road, Gandhinagar, Adyar, Chennai 600 020, India.
Printed by S. Baskarapandian at Mani Offset, 112/2, Bells Road, Triplicane, Chennai
600 005.
Editor : Shashanka Bhide