1 India’s Social Pension Schemes

In India, government pensions are provided to the poor elderly, poor widows and the severely disabled. Pension payments are largely funded by the centre, but are administered by the states. Pensions for elderly and widows’ are intended only for the poor. Until recently participation criteria were defined by the states, and were intended to ensure only coverage of the very poor or “destitute”. Recently, however, criteria have been relaxed, and now any elderly person or widow with a below poverty line (BPL) card is entitled to the pension.

Monthly rates vary from state to state, depending on the cocontribution which the state governments make, but the amounts of money involved are small. During 2005 and 2006, in the two states where the survey was carried out, the standard pension payment was Rs 100 a month in Karnataka and Rs 200 in Rajasthan. These amounts have since increased to Rs 400 in both states. The mode of payment also varies from state to state. Both Karnataka and Rajasthan rely mainly on money orders, which can be cashed at a bank (or delivered to the beneficiary via the postal system), but both states also use cash payments at government offices (more common in Rajasthan) and payments into bank accounts (used in Karnataka).

64

The 2004-05 NCAER-University of Maryland India Human Development Survey (IHDS) found a national coverage rate relative to the number of elderly of about 7% for old-age pensions (just under five million recipients) and similar coverage, relative to the number of widows, for the widows’ pensions (almost three million recipients). Coverage rates also vary from state to state. Coverage rates in Karnataka, are much higher: 20% for old-age pensions, and 27% for the widow’s pensions. Those in Rajasthan are closer to the national average, at 7% and 10%, respectively.

2 Data

To analyse the targeting and coverage of the social pension schemes, we rely primarily on analysis of the 2004-05 nationwide 41,000-household IHDS reported in Ajwad (2006). To get at issues of leakage, and for more detailed analysis of the social pension schemes, we analyse special-purpose household surveys carried out in Karnataka (in 2005) and Rajasthan (in 2006).

The two state surveys are representative samples of households with at least one elderly person or a widow.2 In Karnataka, both rural and urban areas were included in the survey, though not the capital Bangalore. The Rajasthan survey was focused on rural areas.3 A detailed description of the sampling method is provided in Murgai (2006) for Karnataka, and Dutta (2008) for Rajasthan. In both states, 20 blocks (taluks) were selected and within these a number of smaller administrative areas.4 Based on a second-stage unit listing operation of households with at

least one elderly person or Table 1: Sample Sizes for Karnataka and Rajasthan Social Pension Surveys

widow (referred to as eld

Full Sample Pensioner Non-pensioner erly households, and widow Households Households

Total 695 The Rajasthan survey also included households

strata were randomly selected 1,566 871

for interview. The fi rst stra

with persons with disabilities. Those households have been excluded from the analysis in this

tum was over-sampled as

paper. For explanation of the “neither” category in Table 1 shows. Karnataka, see footnote 2.

A short questionnaire was administered to each household, with common questions on household characteristics addressed to both groups and separate modules on information related to the pension schemes designed for the two groups. For pensioner households, the special module solicited information on programme eligibility criteria and various transaction costs associated with the application process. For non-pensioner households, the special module also collected information on whether the potential beneficiary met the programme criteria, besides information on failed tries to get a pension or reasons why a potential beneficiary chose to never apply for a pension.

Finally, the Karnataka study also entailed an examination of the treasury department’s computerised pension database to examine leakage. A similar exercise in Rajasthan was not feasible because pension records were only partially computerised.

december 25, 2010 vol xlv no 52

SPECIAL ARTICLE

3 Performance of the Social Pension Schemes

We use three criteria to judge the performance of the social pension schemes: Coverage: what percentage of the targeted population actually benefits from the scheme? Targeting: do the schemes benefit the poor and vulnerable? Compliance: is the scheme run according to guidelines; if not, at what cost?

Of these three, coverage is the least useful for judging performance, since it depends on the budget made available to the programme. Depending on performance in relation to the other two criteria – targeting and compliance – the budget of the scheme concerned can be scaled up or down. However, it is important to note that changing coverage can, in turn, have an impact on targeting and leakage. We explore this issue later in the paper.

3.1 Coverage

Coverage can be defined in relation to either the specifi c eligibility criteria of the schemes or their broad objectives, namely to support the elderly and widowed poor. Performance in relation to the former is in fact low, as states do not follow their own eligibility criteria. However, especially given that eligibility criteria have changed since the conduct of the survey, this is less important than whether the pensions reach the poor. We analyse eligibility in Section 3.3 in the context of a discussion about compliance. In this section, we focus on the extent to which pensions provide support for the elderly and widowed poor.

We draw on Ajwad (2006) for IHDS survey-based estimates of national coverage. Households are ranked into quintiles based on an index which combines data on household ownership of various assets and dwelling characteristics through principal components to proxy for household wealth (Filmer and Pritchett 2001). Figure 2 shows the coverage of the target households (those containing at least one elderly person, and those containing at least one widow, respectively) for the two schemes by wealth quintile. The downward sloping nature of the curves in Figure 2 reflects the progressivity of the schemes, which is discussed in the next sub-section. But the low values on the vertical axis show the very low coverage of the schemes, which reaches only 10%, even among the poorest quintile.

Figure 2: National Coverage Rates of Households with Elderly or Widows by Wealth Quintiles

(% pensioners among target households)

12

10

8

6

4

2

0 Poorest Q2 Q3 Q4 Richest Source: Ajwad (2006), based on 2004-05 IHDS.

Old-age pension (among elderly households) Widow’s pension (among widow households)

To achieve full coverage of the schemes would require a massive expansion, at least given the new guidelines. Earlier guidelines called for coverage of half (the poorest half) of the elderly and widowed poor. The new guidelines call for full coverage of

Economic & Political Weekly

EPW

december 25, 2010 vol xlv no 52

all poor. About one-third of Indian households have a BPL card. Full coverage for the old-age pension would thus correspond roughly to ensuring that every household in the bottom two quintiles with an elderly member receives an old-age pension. This, in turn, would require an increase in coverage from almost five million to almost 30 million on the (unrealistic) assumption that only households in these quintiles received a pension, and, without any improvement in targeting, would require an increase in coverage to 55 million.

It is hard to judge the extent to which the binding constraint on coverage is limited central government funds (the new guidelines explicitly state that the scheme will expand to meet demand from eligible applicants), but the state-level surveys are useful for understanding some of the reasons why coverage is not higher. They show that awareness is high but for universal. In both Karnataka and Rajasthan, about 70% of potential benefi ciaries were aware of the old-age pension, but in almost half the areas sampled awareness of the existence of both pensions was less than 50%. In addition, the Rajasthan study indicated the level of knowledge was low – even among households that were broadly aware of the social pension schemes, only a small minority was aware of the details of the application process. Among potential beneficiaries, who had heard of the schemes, 35% knew virtually no details of the scheme, 58% knew about the level of benefi ts only, while only about 8% knew details of the eligibility criteria and the process of application. Qualitative work also indicated high awareness of the scheme, but not of the application process including required documents, eligibility criteria and sanctioning authority, even among several sarpanches (village government heads) interviewed.

In both states, one in three non-recipients had applied for a pension. Paying a bribe to become a pensioner is common – in Karnataka 28% of unsuccessful pension applicants reported paying a bribe. The average waiting time is a year in Karnataka, and six months in Rajasthan. The complexity of the process was a major deterrent to applying in both states.

3.3 Compliance and Leakage

There are at least five possible types of non-compliance. First, duplicate records in the administrative database of pensioners lead to the possibility that a pensioner is overpaid, or that someone else is cashing in one of the money orders. Second, if enrolled pensioners are not receiving payments because they are “missing” (either do not exist, or have moved or died) there is the risk that fraudulent pension recipients are receiving the pension in their stead. Third, enrolled pensioners might not be missing, but still might not be receiving their pension in full, or even in part. Fourth, pensioners might have to pay bribes at the time of joining or during the year to receive the pension. Fifth, pensioners might be enrolled, but may not be eligible given scheme guidelines.

We are able to address the third and fifth of these issues using both state surveys, and the others using the Karnataka survey (in large part because of the state’s computerised pension recipient lists).

In Karnataka, an analysis was undertaken of administrative lists of pensioners to assess the extent of duplication of records.9 Note that full addresses are not provided (and often do not exist in rural areas), so that the exact extent of duplication is unknown. Records contain data on name, father’s/husband’s name, and post office, as well as an identifier, which is meant to be but may not be unique for each household. An algorithm was run to assess the closeness of different records. Only 0.2% of records were virtually identical (for example, all details the same except the identity card).10 But 1% of records had a very high similarity score (0.9 or above) and 6% a high score (0.8 or above).

We were unable to check through fieldwork the probability that records assigned a particular duplication score were in fact duplicates. In the analysis which follows we simply assume that pairs with a score of 0.9 or above were duplicates, and pairs with a score below 0.9 are not. This gives a total of 1% of records as duplicates.

There is a concentration of likely duplicates in a small number of areas: half the duplicates were in two of the 20 blocks. It is unclear what pension payments are made to duplicates, though, if programme guidelines are followed, money orders are printed for all records. We came across no cases of households receiving double-pensions (or, at least, none who admitted it). It is possible that duplicate pensions are siphoned off by postal deliverers, or at some other stage of the process.

could be found but reported that the pensioner did not exist).11 As with duplicates, the problem is concentrated geographically: 60% of cases were in the worst third of sampled blocks.

Households which were wrongly listed as containing an enrolled pensioner were not sampled, and we do not know the fate of pensions intended for these missing pensioners. This percentage will never go to zero, as pensioners will continue to die, and it will take some time, even in a perfectly-performing system, for this to be recorded. Nevertheless, it is reasonable to regard pensions intended for but not going to missing enrolled pensioners as a form of leakage.

Analysis of the extent of receipt of pension payments was collected for both Karnataka and Rajasthan. On average in Karnataka, enrolled pensioners receive 96% of their pension, and in Rajasthan 93%. There are two distinct problems in this regard. A large number of pensioners – one in five in Karnataka, and one in four in Rajasthan – report paying small bribes to the postman and government offi cials.12 A small number of pensioners in both states report receiving much less than they are entitled to, or no pension at all. Note again the geographical concentration of the problem. If the two worst-performing blocks in Karnataka are excluded, the percentage shortfall falls from 4% to just 1%.

On bribes paid to join, as already mentioned, those who applied for a pension, but unsuccessfully, paid on an average fee of Rs 200 (in Karnataka). If we assume that successful recipients paid the same bribe on average, and amortise the fees both they and the unsuccessful applicants pay over a 20-year period (assuming a 6% discount rate), then this adds another 2% to leakage. The Karnataka survey also found that 8% of participants had paid a median amount of Rs 100-Rs 200 in the last 12 months in fees to various officials in order to receive a pension. This adds less than 1% to total leakage.13

Finally, we turn to the issue of eligibility. The criteria Karnataka and Rajasthan use for selecting poor widows and old people were, at the time of the survey, similar but not identical. The Karnataka criteria were more restrictive. Both Karnataka and Rajasthan restricted receipt of old-age pension to those without adult children (adult family members in the case of Rajasthan) to support them. They both placed limits on own-income support and on household support for old-age and widow pensions. However, the demographic eligibility criteria for old-age pensions were more stringent in Karnataka (only those without a spouse are eligible). The reverse was true for widow pensions, where Rajasthan further restricted receipt to those without adult family members. An important difference was that in Rajasthan the BPL

Economic & Political Weekly

EPW

december 25, 2010 vol xlv no 52

status overrode all other income support and demographic (i e, family member) criteria. In other words, at the same of the survey, Rajasthan, but not Karnataka, applied what is now the nationally mandated practice of providing pensions to all elderly and widow with a BPL card.14

Were these criteria enforced in practice? In Karnataka, the criteria were honoured almost entirely in the breach – only 9% of recipients of the old-age pension and 1% of recipients of the widow pension met all the demographic and destitution criteria. Pensions went entirely to the elderly and widows. But many old people in receipt of a pension have adult children or grandchildren, and have income or income support in excess of what is allowed. The situation was not much different in Rajasthan, but the BPL clause allowed a much higher share to be eligible. Only 26% of elderly pensioners and 9% of widowed pensioners would have been eligible in Rajasthan were it not for the BPL clause. Table 3: Percentage of Pensioners and Non-Pensioners Satisfying Eligibility Criteria

Destitution criteria refer to lack of own income or income support in the case of Karnataka and to lack of own income in the case of Rajasthan. In the case of Rajasthan to be eligible a pensioner must either have a BPL card or meet the demographic and destitution criteria. In the case of Karnataka, both demographic and destitution criteria must be met. Source: Karnataka and Rajasthan Social Pension Surveys.

The demographic criteria were evidently too strict. Moreover, household and own-income are difficult to observe, and evidence from Karnataka suggests that own-income for pensioners is inversely correlated with household wealth: poorer pensioners have to work. Family support is also difficult to verify.15

Given the recent reforms and the linking of social pensions to BPL status, non-compliance with the eligibility criteria is of only historical interest. A summary message of leakage can be obtained by aggregating the other four types of non-compliance. In Karnataka, these add to about 17%: 4% from partial or nonreceipt of the pension; 9% from missing pensioners; 3% from joining and ongoing fees or bribes; and 1% from duplicate records. In Rajasthan, only leakage due to partial/non-receipt of the pension is available. This amounts to 7% (relative to 4% in Karnataka).

Overall, the results of this evaluation of the social pensions are moderately positive. The targeting of the scheme is moderately progressive, though this clearly varies from state to state. Equally or more importantly, the social pensions target the vulnerable, and provide an independent income stream for those who would otherwise be dependent and powerless. Leakage, while present, is moderate.

4 Comparison to the Public Distribution System

To put the evaluation of the social pensions into context, it is useful to present a parallel evaluation of India’s largest antipoverty programme, the PDS.16 Data on PDS utilisation was also

68 collected by the IHDS. Coverage is measured in terms of numbers utilising the PDS system. Targeting is measured taking into account not only whether, but how much (in terms of kilograms of grain purchased) households use the PDS. This measure ignores the differential pricing available under the PDS, but analysis suggests that it is not the case that poorer households pay lower prices.

The PDS has a much higher coverage of a much larger target group than the social pensions scheme, as Figure 8 shows:

Figure 8: Converage of PDS and Social Pension Schemes, by Wealth Quintile

(% of quintile receiving benefits) 35

15 20 25 30 PDS usage (among all households)

Old-age pension (among elderly households)

5 10 Widow’s pension (among widow households)

0

Poorest Q2 Q3 Q4 Richest Source: Data from Ajwad (2006) based on 2004-05 IHDS.

The targeting performance of the PDS and the pension schemes are quite similar (Figure 9). All three schemes deliver half of their benefits to the poorest 40%. The pension schemes do better at targeting the poorest 20%, but also provide more to the top 20%. The fact that benefits from the PDS are progressively distributed is surprising, given the trenchant criticisms of the lack of targeting in the PDS, and reflects the success of the government’s reform strategy for the PDS, through the introduction of the targeted public distribution system (TPDS) which aimed to limit the provision of subsidised food to the poor (who were issued with a BPL card, entitling them to food from public distribution shops at subsidised prices).

Figure 9: Concentration of PDS and Social Pension Participation, by Wealth Quintile

(cumulative % of quintile covered)

0 20 40 60 80 100 45-degree line OAP Widow PDS

0 20 40 60 80 100 Source: Data from Ajwad (2006) based on 2004-05 IHDS.

Though we do not have data on this point, it is reasonable to surmise that the PDS like social pensions contributes to a better intra-household allocation of resources, since it forces the receipt of benefits in the form of necessities, rather than cash, which could be used solely or primarily for the benefit of the household head.

An important difference between the PDS and social pensions concern leakage. A full analysis of leakage such as was carried

december 25, 2010 vol xlv no 52

5 Conclusions

This paper sought to evaluate the performance of India’s social pensions, from both an absolute and a comparative perspective. Based on the best data available – a nationwide, generalpurpose household survey, and two, special-purpose statespecific surveys – we found that in general, the social pensions perform well. In particular, the performance of the social pensions scheme appears to outperform the PDS. Though it has lower coverage, both schemes target the poor and vulnerable, and the social pensions schemes have much lower leakage: half the level in Karnataka comparing a comprehensive measure of leakage for social pensions to a partial measure of leakage for subsidised food.

Should social pensions, therefore, be scaled up? While this at first glance seems to follow automatically, in fact, it depends on the likely impact of increased coverage on both targeting and leakage. On targeting, the evidence presented in the paper suggests that increased coverage, in the form of expansion in numbers, is in fact likely to worsen targeting (Figure 7). The shift now underway to basing social pensions on BPL status might also worsen targeting given the only mildly progressive distribution of BPL status (Ajwad 2006). However, neither expansion nor a shift to BPL status is likely to render the social pension regressive in its distribution, and its positive impact on vulnerability would remain intact. In any case, the comparison with the PDS suggested that the key attraction of the social pension scheme was its low level of leakage.

What would happen to leakage if the cash transfers schemes are expanded? This depends on the reason for the current low leakage. There are two possibilities. One is that leakage is low because levels of discretion are low. There is a high level of discretion at the time of joining, but payments then follow more or less automatically. Bribes might have to be paid to join the scheme, but once a pensioner is on the list there is little scope for further diversion of funds, at least on a large scale. Compare this to the PDS, where not only do potential participants face the same

Economic & Political Weekly

EPW

discretionary challenges of signing up, but, in addition, they have to persuade the shopkeeper to open his shop and sell to them rather than divert his grain to the open market. This is evidently a difficult task, not often achieved. In summary, social pensions might be a pain to register for, but thereafter the benefi ts fl ow more or less automatically. For the PDS, every extraction of benefits from the system requires effort and is a potential rent-seeking opportunity.

The other possible explanation for the relatively low leakage of the social pension schemes is that both the payments involved and the number of recipients are relatively small. Those after public funds through corrupt means are likely to follow the money and target resource-rich programmes.

These two hypotheses are both plausible. One stresses the institutional, supply-side features which might make the social pensions schemes relatively impervious to corruption. The other stresses demand-side factors which might make the social pensions schemes less attractive to corrupt agents. But they have very different implications for the scaling up of social pensions. If the former is correct, then scaling up is warranted, since scaling up will not affect the design of the social pensions schemes. If the latter is correct, however, then scaling up will compromise performance.

Since we cannot distinguish between these two hypotheses it makes sense to scale up the social pensions schemes progressively and to monitor performance along the way, to try to detect a possible deterioration in performance.

One factor which supports a scaling up conclusion is that it seems from the analysis that such leakage as there is in the social pensions scheme should be tractable to policy action. Thus, if scaling up does indeed increase leakage, corrective measures can be taken to combat this. Evidence from Karnataka suggests that abuse of the scheme – whether in terms of duplicate records, missing pensioners, or non/partial pension payment – is often concentrated in a few problem administrative areas, and should be susceptible to administrative checks and surveys, especially after computerisation.

If the social pension schemes should be scaled up, how should this be done? In the short term, the way forward would be to expand the coverage among widows and the elderly and to increase the size of the pension. This is a path the Government of India has already embarked upon. In the longer term, the policy question is whether it makes a sense to expand the categories to whom social pensions are given. Ultimately, one can imagine a situation in which, say, cash pensions are made to every holder of a BPL card, instead of, or as well as, an entitlement to buy subsidised food and/or to a guarantee to a minimum number of days of public employment. With the National Right to Employment Guarantee Act, and the decision of the government to support a Right to Food Act such a scenario is far-fetched today. Moreover, as stressed above, little is known about how cash transfers would perform if scaled up. Nevertheless, it is reassuring to know that calls for a greater role in Indian social security policy for cash transfers are at least not contradicted by the performance of India’s existing cash transfers.

1 World Bank, forthcoming, Social Protection for a Changing India, New Delhi, draft.

2 This is not strictly true, as households in receipt of one of these pensions could be included in the sample even if the individual receiving the pension was not actually elderly or a widow. In practice, the numbers in this category were negligible: none in Rajasthan, and just four households in Karnataka.

3 The Rajasthan survey was designed to also examine performance of the pension programme for persons with disabilities. In this paper, we draw only on the results for elderly and widow pensions.

4 In Karnataka, the second stage of sampling was based on postal delivery areas and polling stations. In Rajasthan, gram panchayats were used for second-stage selection.

5 “In a major survey on health profiles of older people in India, covering 5,000 households in urban and rural areas, 45% of both men and women in the sample reported chronic illnesses. Smaller studies in India have also indicated that in addition to coronary, muscular and respiratory problems, close to 90% of older people surveyed had visual impairment and more than 40% suffered from some form of depressive illness” (HAI 1999: 12).

6 Both Figures 3 and 4 are defined in relation to the number of beneficiaries not the amount received by each beneficiary. Given that the pension amounts are uniform (at least within a state), and that non and partial receipt of the pension appears to be a relatively minor and distributionneutral problem (Section 4.3), the distribution of beneficiaries across wealth quintiles and the distribution of benefits across quintiles will be about the same. In addition, using the number of benefi ciaries at the national level has the advantage of abstracting from differences in pension payments across states.

7 An eight-state evaluation of the national social assistance programme (including old-age pensions) conducted in 1998 by ORG also found that pension coverage among SC/ST, landless, illiterate households is high, about 40-60% of NOAPS benefi ciaries were women, and more than 95% of benefi ciaries met the age criteria.

8 This question was not asked in the Karnataka survey. 9 Assessing the incidence of duplicate pension records or enrolled pensioners who cannot be traced was not possible in Rajasthan because the administrative database was only partially computerised. 10 In theory, different IDs should indicate different households, but not in practice. 11 A very small percentage of listed pensioners could be found but had never received a pension. They were also counted as missing pensioners. 12 Payments to government officials are more prevalent in Rajasthan since more pensions are distributed through government offices (in cash), and relatively fewer through the postal system (via money orders). 13 It is assumed that these bribes are additional to the ones causing incomplete receipt of the pension. In practice, there might be an overlap. An unpublished study of social pensions in Himachal Pradesh found that there are high transaction costs in the application process, but also that the administration of the scheme is fairly smooth once pensions are sanctioned. For instance, the application process in Himachal Pradesh was fairly lengthy with only 25% sanctioned within six months and the average time taken is slightly over a year. There are high satisfaction levels with the scheme, particularly with respect to selection procedure and regular payments but the amount is perceived as insufficient (the benefit amount at the time of the HP study was approximately Rs 100 depending on pension type). 14 The difference between then and now is that now BPL status is mandated to be the only criterion. At the time of the Rajasthan survey, it was possible to get a pension even without a BPL card provided the other criteria were met.

References

October 30, 2010

For copies write to: Circulation Manager,

320-321, A to Z Industrial Estate, Ganpatrao Kadam Marg, Lower Parel, Mumbai 400 013. email: circulation@epw.in

Effects without Expenditure Data – or Tears: An Application to Educational Enrolments in States of India”, Demography, 38(1): 115-32.

HAI (1999): The Ageing and Development Report: Poverty, Independence and the World’s Older People, HelpAge International.

IDS (2006): “Unconditional Cash Transfers in Africa”, Social Protection, Institute of Development Studies, Issue 1.

Kapur, D, P Mukhopadhyay and A Subramaniam (2008): “The Case for Direct Cash Transfers to the Poor”, Economic & Political Weekly, 12 April, pp 37-40.

Murgai, R (2006): “Do Public Funds Reach the Destitute? Assessment of Leakages and Targeting of Social Pension Programmes in Karnataka”, Manuscript.

ORG (1998): Evaluation of National Social Assistance Programme in Selected States, report prepared by Operations Research Group, Ministry of Rural Areas and Employment, New Delhi.

Pal, S and R Palacios (2006): “Old Age Poverty in the Indian States: What the Household Data Can Say? Draft.

Planning Commission (2005): Performance Evaluation of Targeted Public Distribution System, Programme Evaluation Organisation, Government of India, New Delhi.

– (2008): Eleventh Five-Year Plan 2007-12: Volume II Social Sector (New Delhi: Oxford University Press).

Start, D and P Deshingkar (2006): “Poverty Slides in the Countryside: Shocks, Finance, and Welfare in Varying Contexts” in J Farrington, P Deshingkar, C Johnson and D Start (ed.), Policy Windows and Livelihood Futures: Prospects for Poverty Reduction in Rural India (New Delhi: Oxford University Press).

Weigand, C and M Grosh (2008): “Levels and Patterns of Safety Net Spending in Developing and Transition Countries”, Social Protection Discussion Paper No 0817, World Bank.

World Bank (2009): Conditional Cash Transfers: Reducing Present and Future Poverty, Policy Research Report, Washington DC.

december 25, 2010 vol xlv no 52