A review of international experience

Learning from the Early Experience of 
India’s Matching Defi ned 
Contribution Scheme
Robert Palacios and Renuka Sane
In an attempt to expand pension coverage, in 2010 the government of India introduced a 
matching defined contribution scheme. Preliminary analysis of the New Pension Scheme 
strongly suggests that women are more likely than men to participate; that higher income 
and more education are positively associated with take-up; and that having other potential 
sources of income for old age, including land, housing, and sons, reduces take-up, with 
other factors held constant. The data do not allow exploration of many others factors that 
may contribute to take-up, including the importance of the credibility of the intermediary 
(aggregator) and the role of information.
T
he challenge of expanding pension coverage is immense in India, where less than 
10 percent of the labor force participates in formal pension schemes, according to 
most estimates. Private, formal sector workers are covered by both defined contribution 
and defined benefit schemes run by the Employee Provident Fund Organization (EPFO). 
Until recently, government employees were covered by the civil service pension system 
inherited from the British—a defined benefit system financed directly from the budget. 
Although a new defined contribution scheme was introduced in 2004, it applied only to 
new entrants, leaving the defined benefit scheme in place for many workers for the next 
few decades.
Like many countries, India has tried to expand coverage of formal sector pension 
schemes. These attempts have largely failed. In the 50 years since the Employees’ Prov-
ident Fund Act (1952) was introduced, coverage increased from roughly 1 percent to 
5 percent of the labor force. This level of coverage is typical of countries at India’s income 
level (figure 12.1).
Moreover, EPFO coverage is highly concentrated among upper-income workers. 
Only a very small proportion of lower-middle-income or poor people, who make up 
the majority of India’s population, have ever contributed to a formal pension scheme 
(figure 12.2).
One response to the failure to expand pension coverage was the introduction of the 
National Old Age Pension Scheme in 1995. This cash transfer scheme targets people age 
65 and older living in households deemed to be below the poverty line. Currently, roughly 
The authors would like to thank Kshetriya Grameen Financial Services (KGFS) for its cooperation 
and support and Parthasarathi Edupalli for research assistance.

244 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
one in every five elderly people in India is receiving a cash transfer of about Rs 200–300 
a month. The National Old Age Pension Scheme benefit can at best supplement other 
sources of support required to survive.
1
FIGURE 12.1  Coverage of contributory pension schemes, by income level 
y = 0.0002e
1.8456x
R
2
 = 0.75747 
0
20
40
60
80
100
120
140
160
2.0 2.5 3.0 3.5 4.0 4.5 5.0 
active members as of labor force
log(income per capita [thousands]) 
SOURCE: Pallares-Miralles, Romero, and Whitehouse 2011.
FIGURE 12.2  Pension system coverage in India, by income level, 2010 
b. Urban household
a. Rural household
0 
10 
20 
30 
40 
50 
60 
70 
Mar.
2009 
Dec.
2009 
Sept.
2010 
June
2011 
% of rural households 
rich 
higher middle income 
middle income 
lower middle income 
poor 
0 
10 
20 
30 
40 
50 
60 
70 
% of urban households 
Mar.
2009 
Dec.
2009 
Sept.
2010 
June
2011 
SOURCE: Center for Monitoring Indian Economy.

12.  LEARNING FROM THE EARLY EXPERIENCE OF INDIA’S MATCHING DEFINED CONTRIBUTION SCHEME 
245
For most older people in India, income support in old age comes from family. Co-
residence rates are high, with elderly women especially dependent on their families for 
support. Poor families find it difficult to support the elderly, and there is some evidence 
that they do not always provide support even when they are able to do so.
The gaps in family support may increase with the decline in fertility rates, increases 
in life expectancy, and urbanization. Although there is wide regional variation across the 
country, India’s population is aging. Over the last two decades, policy makers have begun 
to introduce new features and to reform the system to prepare for the future.
In 2001, the government set up a special committee to provide recommendations to 
address the increasingly obvious gap in pension coverage. The Dave Committee proposed 
a new approach that would allow workers in the informal sector (known in India as the 
“unorganized sector”) to contribute to a defined contribution plan managed by the state 
(Shah 2006). This New Pension Scheme (NPS) was first adopted for civil servants in a 
major reform implemented in 2004. Under the reformed civil service retirement system, 
all new central government employees are required to contribute 10 percent of their wages 
to a defined contribution account. This mandatory contribution is then matched with an 
equivalent contribution from the government.
In addition to controlling the liability of the government for benefits through a tran-
sition to a defined contribution structure, an important feature of the new scheme was the 
unbundling of recordkeeping and asset management. Over the next few years, most state 
governments adopted this model.
2
 An interim regulator, the Pension Fund Regulatory 
and Development Authority, was set up along with a new recordkeeping infrastructure, 
known as the Central Recordkeeping Authority. A competitive bidding process resulted 
in contracts with three public sector asset managers; private firms were brought in later.
The second phase of NPS implementation was intended to use its infrastructure 
to extend coverage to informal sector workers throughout the country. In 2009, the 
NPS was officially opened to any worker not covered by a formal pension scheme.
3
 Post 
offices, public sector banks, and other entities were enlisted to allow workers to open NPS 
accounts. In addition to its regulatory and oversight role, the Pension Fund Regulatory 
and Development Authority undertook efforts to inform workers about the system and 
encourage their participation.
Despite these efforts, take-up was extremely low the first year, largely as a result of 
the lack of incentives for providers and individuals. Other saving products received more 
favorable tax treatment, and the cost of opening and maintaining accounts made the NPS 
unattractive to most informal sector workers. Although investment returns were good 
and the systems were gradually coalescing for account administration for civil servants, by 
2010 the Indian press was widely reporting the failure of the NPS to achieve meaningful 
coverage expansion among informal sector workers.
To address this challenge, in the spring of 2010, the government of India announced 
that it would deposit matching contributions of Rs 1,000 a year into the NPS accounts of 
individuals who made contributions of Rs 1,000–Rs 12,000 per fiscal year (April 1–March 
31).
4
 These accounts were called NPS-lite, as they differed from those of civil servants in 
several ways. Civil servants were able to view their accounts online and had other services 
that were not available to NPS-lite members. Most important, fees for basic retirement 
savings accounts were significantly lower than the fees on civil servants’ accounts. The fee 

246 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
for opening an account was Rs 70, and the first 12 transactions were free. In contrast, the 
account opening fee for civil servants was Rs 350. (Annex A displays the range of charges, 
which include asset management fees.) Other changes made to the tax treatment of NPS 
accounts put them on an equal footing with similar saving products, although restrictions 
on withdrawals (intended to ensure that the savings were preserved for retirement income) 
were more restrictive.
5
 These changes did not address the lack of outreach or incentives for 
providers to market the NPS to the informal sector, and the increase in take-up remained 
relatively modest.
In August 2010, the government set up a committee to review the performance of 
informal sector pensions. The committee made various recommendations in areas ranging 
from marketing to reduction of costs. It paid special attention to incentives for providers, 
including pension fund managers, and to “points of presence” (places where subscribers 
can open accounts).
In late 2010, the government launched a new initiative. To extend coverage of the 
system, it established incentives to encourage the introduction of entities that would serve 
as account “aggregators.” These aggregators were mostly nongovernmental organizations 
that had met predefined criteria that qualified them to undertake outreach, marketing, 
and enrollment functions (see annex B). These aggregators are paid at least Rs 50 per 
enrollee and a volume bonus of Rs 10 up to 120,000, Rs 17 up 200,000, and Rs 22 up to 
500,000. It remains to be seen whether these incentives are adequate to motivate aggrega-
tors to actively pursue enrollment and whether the reduced costs of NPS-lite combined 
with the matching contribution will be sufficient to encourage low-income informal 
workers to enroll in the system on a large scale.
Early Experience with Matching Defined Contributions in India
The NPS concept that emerged in 2001 began as an attempt to reach the massive infor-
mal labor force in India with a suitable vehicle for saving for old age. Arguably, however, 
the two key ingredients for successful implementation of such an approach—particu-
larly in regard to the inherent limitations faced in reaching a predominantly low-income 
and largely rural target population—were missing. The first was a financial incentive to 
encourage workers with limited liquidity to tie up their limited savings over the extended 
period required to produce meaningful retirement income. The second was a supply-side 
incentive to motivate entry into the market and effective marketing of the product to 
inform people and make it easy to enroll in the program. These pieces of the puzzle were 
put into place only in late 2010.
What does the limited evidence since late 2010 show? No systematic evaluation 
process is yet in place to study what determines take-up of NPS-lite. Collectively, aggrega-
tors enrolled about 300,000 workers in 2011. It is not possible to estimate the take-up 
rate from these figures without information on the potential number of participants, but 
clearly it represents a tiny fraction of the potential pool. At the same time, the number of 
enrollees has expanded sixfold since the new incentive mechanisms were introduced.
A study of the determinants of enrollment patterns as the NPS unfolds and an 
analysis of how different segments of India’s very heteroge neous unorganized sector are 
likely to respond to the design of the program would help policy makers understand the 

12.  LEARNING FROM THE EARLY EXPERIENCE OF INDIA’S MATCHING DEFINED CONTRIBUTION SCHEME 
247
dynamics and scope for expanding coverage. Combining evaluation with national-level 
survey data would help policy makers obtain a more accurate idea of the potential size of 
the program and craft a better strategy. The next section presents a preliminary quantita-
tive analysis of the differences between people who have enrolled in the NPS and people 
who have not. 
DATA
The data presented here are from Kshetriya Grameen Financial Services (KGFS), a finan-
cial services provider that operates in five dis tricts in three states in India.
6
 In addition to 
selling insurance and loan products, KGFS functions as an aggregator for the NPS. 
KGFS institutions are promoted by the Institute for Financial Management and 
Research Trust, a foundation whose mission is to ensure that every individual and 
enterprise has access to financial services. All KGFS institutions have a common parent 
company, which provides equity capital to each. Each institution is an autonomous, self-
contained re gional operation with its own management team hired locally.
When a customer enrolls in the KGFS sys tem, information on his or her demo-
graphics, income, and financial goals is recorded. With the help of an in-house algo-
rithm, a financial well-being report is generated that provides recommenda tions on which 
products the individual should buy. These products include loans (joint-liability loans, 
emergency loans, and gold loans), savings products (money market mutual funds), and 
insurance (personal accident insurance, term life insurance, and livestock insurance). Indi-
viduals often buy products that are different from the ones KGFS recommends.
The data used in this chapter are on all customers in the various locations in which 
KGFS operates. (For detailed descriptive statistics, see annex C.) All customers were told 
about the NPS and shown how to sign up for the scheme. 
Figure 12.3 displays the number of people that en rolled in the NPS each month 
between November 2010 and April 2012. In the early months of January–March 2011, 
there was a surge in participation, perhaps because individuals had to enroll be fore the end 
of March to receive the Rs 1,000 co-contribution (although the peak was in February, not 
March).
Annex C compares people who enrolled in the NPS with people who did not. 
Enrollment is much higher among women (30 percent) than among men (9 percent); 
among women, it is higher among “housewives” (women who do not work outside the 
home) (32 percent) than among wage laborers (25 percent). In the states in which KGFS 
operates, Tamil Nadu has the highest par ticipation rate (25 percent), followed by Orissa 
(21 percent); Uttarakhand (8 percent) lags far behind. The regions in Tamil Nadu are 
better off than the regions in Orissa and Uttarakhand, perhaps explaining their residents’ 
greater willingness to set aside money for old age. Enrollees in the NPS have slightly 
higher median income (Rs 94,625) than non-enrollees (RS 90,000).
The proportion of households subscribing to the NPS increases marginally with 
household income, falling slightly at the highest part of the distribution (figure 12.4). 
Although the sample is not representative of the population even in these three states, it is 
striking that a substantial portion of the poorest people were willing to participate.
The NPS may be the only long-term saving product held by KGFS customers, or it 
may be held along with several other loan and insurance products (table 12.1). Background 

248 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
risk and liquidity constraints are important determi nants of household portfolio choice 
(Angerer and Lam 2009; Heaton and Lucas 2000; Zeldes 1997). 
Purchasers of accident insurance and joint liability group loans—a common type of 
microfinance group loan in India—are more likely to participate in the NPS than people 
who do not buy the two products. The numbers are particularly stark for joint liabil-
ity group clients, 33 percent of whom enrolled in the NPS. In contrast, among people 
FIGURE 12.3  Monthly enrollment in India’s New Pension Scheme, 2010–12 
numb
e
r
0 
500 
1,000 
1,500 
2,000 
2,500 
3,000 
3,500 
4,000 
4,500 
Nov. 
2010 
Jan. 
2011 
March 
2011 
May
2011 
July
2011 
Sept.
2011 
Nov.
2011 
Jan.
2012 
March
2012 
SOURCE: Authors’ calculations, based on KGFS data.
FIGURE 12.4  Participation in India’s New Pension Scheme, by per capita income, 2011
% c
o
n
tr
ib
u
ti
n
g
 to N
P
S
0.00 
0.05 
0.10 
0.15 
0.20 
0.25 
1 2 3 4 5 6 7 8 9 10 
income decile 
SOURCE: Authors’ calculations, based on KGFS data. 

12.  LEARNING FROM THE EARLY EXPERIENCE OF INDIA’S MATCHING DEFINED CONTRIBUTION SCHEME 
249
TABLE 12.1  Participation in India’s New Pension Scheme by people with and without other financial 
assets
Financial asset
Did not buy NPS
Bought NPS
Number of observations
Personal accident insurance
Bought
77
23
79,073
Did not buy 
79
21
18,240
Life insurance
Bought
63
37
39,943
Did not buy 
87
13
57,370
Joint liability group loan (microfi nance)
Bought
67
33
58,448
Did not buy 
92
8
38,865
SOURCE: Authors’ calculations, based on KGFS data.
NOTE: Figures show only KGFS assets. However, given that KGFS opens branches only in areas that are not typically 
served by the formal fi nancial sector, it is unlikely that households had access to insurance products other than those sold 
by KGFS.
without a joint liability group loan, just 8 percent were enrolled. Two factors may account 
for this pattern. On the one hand, people who are very well off do not need to take such a 
loan and may also feel no need to contribute to the NPS. On the other hand, people with 
a joint liability group loan are better able to cope with liquidity demands and are therefore 
better able to set aside money for old age. 
Full analysis of NPS enrollment requires data on persistence (the percentage of peo-
ple who continued to contribute the year after enrolling). Such data are not yet available.
STATISTICAL ANALYSIS OF PEOPLE WHO DO AND DO NOT ENROLL IN THE NEW 
PENSION SCHEME 
Many factors affect the decision to join the NPS. The following logit model is estimated 
to identify them: 
c* = 
β 0 + β1
X
 + 
ε
c
 = 1 if c* > 0
The dependent variable, c, indicates whether the respondent contributed to the 
NPS. The variable X includes demographic indicators (age, sex, marital status, edu cation, 
occupation, and family size and composition).
Of interest is the impact of income and wealth on participation. Household income 
is measured as the sum of the income of all household members. Three sets of variables 
capture household wealth: socioeconomic status (captured by whether the household has 
an electricity connection, whether it uses gas or wood for cooking, and whether it has a 
private toilet), land- and homeownership, and ownership and value of consumer durables 
(ownership of a CD player, a mixer and grinder, a mobile phone, a sewing machine, a 

250 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
television, a refrigerator, a washing ma chine, and a computer and the value of jewelry and 
livestock owned). 
Log values are used for land and housing assets and income, because variation in 
percentage terms is more useful than variation in absolute levels. All values are scaled so 
that observations with a value of zero are not lost. The regression also controls for the 
region in which the household resides (table 12.2).
The results in table 12.2 are generally consistent with international evidence on the 
relationship between pension contribu tions and age, marital status, education level, and 
income. Age, for example, is positively associated with participation, as is education and 
marital status (Munnell, Sundén, and Taylor 2002). The age effect is nonlinear, increasing 
until the early 40s and falling thereafter (figure 12.5), a pattern witnessed in the Organisa-
tion for Economic Co-operation and Development (see chapter 2).
Although income is positively correlated with participation, the correlation between 
participation and regular income is negative (people with regular income are more than 
50 percent less likely to contribute to the NPS). This result reflects the eligibility condi-
tions of the program, which in principle at least, apply only to people not covered by the 
EPFO or other pension schemes. People with regular wage income are more likely to be 
covered by the formal pension scheme. The fact that they already contribute to a pension 
likely reduces their interest in the NPS, regardless of whether such a rule is enforced.
The value of land and housing is also negatively correlated with participation. A 
straightforward interpretation would be that these assets are viewed as a potential source 
of income in old age or as a resource to be used as part of a bequest strategy by which the 
owner extracts support during old age from his or her children (Bernheim, Schleifer, and 
Summers 1985; for empirical evidence in rural Kenya, see Hoddinott 1982).
7
 This inter-
pretation does not apply to wealth held in the form of jewelry or livestock.
The negative correlation between NPS participation and the number of children is 
strong, but only with regard to sons. This result is consistent with Indian cultural norms, 
which assume that sons will be the main source of parental support.
8
 As the number of 
sons increases, there is less of an incentive to set aside money for old age.
The estimations confirm the dramatic difference between the take-up rates of men 
and women shown in annex C. There are several possible interpretations of this result, 
including the higher risk aversion and longer life expectancies of women. It may also 
reflect a desire among women to control at least part of the household’s savings by open-
ing these accounts in their names (Bannerjee and Duflo 2011).
Prospects for Matching Defined Contributions in India
Experience with NPS-lite has been mixed, possibly because of inadequate incentives or 
lack of confidence in aggregators. A rigorous evaluation is needed to assess the determi-
nants of success or failure.
A first step in formulating a frame of reference for such an evaluation is recognizing 
that the unorganized sector is very heterogeneous. It includes tens of millions of subsis-
tence farmers and a similar number of small-scale urban vendors. It includes people at 
the highest and lowest ends of the income distribution as well as people of all ages and 
education levels.

12.  LEARNING FROM THE EARLY EXPERIENCE OF INDIA’S MATCHING DEFINED CONTRIBUTION SCHEME 
251
TABLE 12.2  Logit regressions for determinants of participation in India’s New Pension Scheme
Variable
Estimate
Standard 
error