PHILIP O’KEEFE, ROBERT PALACIOS

T

The imperative for addressing the “coverage gap” for unorganised workers through innovative approaches is therefore clear. At the same time, the international experience suggests that sustainable expansion of social security coverage is a major challenge for countries at India’s current levels of income, urbanisation, and institutional capacity. Also, India’s previous national initiatives to expand health insurance, for example, have failed to achieve significant penetration. The policy question which GoI therefore faces is how to achieve sustainable coverage expansion in a manner consistent with India’s constraints. It is not an issue of objectives – these are shared across the political spectrum and by donors – but rather one of whether the objectives can be achieved with a “big bang” or whether a more sequenced approach is appropriate. The international experience provides some guidance on this question.

Economic and Political Weekly August 12, 2006

Figure 1: Publicly Mandated Pension Coverage and GDP PerFigure 2: Life Insurance Spending as Share of PC GNI andCapita, Various Countries, Late 1990s GNI PC, 2004

-. . -= .y = –0.17x2 + 8.07x – 2.08 R2 = 0.95

0 5 10 15 20 25 30 Income Per Capita (in thousands)

=R2 = 0.64

2 0

This article looks at selected features of the NCEUS proposals in light of experience in developing countries with expansion of social security. It is very selective, and there remain a number of “big picture” issues (e g, governance arrangements), and design specifics which are not addressed. The structure is as follows: First the overall scope of NCEUS proposals is reviewed in the light of international experience. The sections that then follow look at specific elements of the proposals related to old-age pensions, health insurance, and administration, before conclusions for consideration as the policy discussion continues.

The NCEUS proposal is ambitious in that it seeks to offer insurance for several risks to 300 million informal sector workers in a span of five years. The ultimate objective of universal coverage is shared by many developing countries as well as international donors like the Intenational Labour Organisation and the World Bank.

But internationally, social security coverage expansion has been a gradual process. When Germany became the first country to offer social insurance (SI) in the late 19th century, less than one-third of workers were covered, almost all in urban centres. Twenty years after its inception in 1935, the old age and survivors’ pension of the US social security programme covered only about half the labour force, when income per capita was around $8,000 at current prices. In terms of benefits provided, disability coverage followed only in 1957, and Medicare (health insurance for the elderly) almost a decade later. Canada did not have a mandated public pension scheme until 1966.

Sources: IRDA; WDI for 80 countries

Even today, no low income country covers more than half of their labour force through mandated social insurance schemes, as shown in Figure 1. While the figure refers to pension coverage, in many cases the programmes include health, unemployment and other benefits. A similar linkage between country income level and life insurance spending can be observed in Figure 2, which shows life insurance (LI) spending per capita as a share of gross national income (GNI). The relationship can also be seen for Employees Provident Fund Organisation (EPFO) coverage rates and gross state domestic product (GSDP) per capita in India itself. While the figures do not explain the underlying causes of coverage, they suggest that no country has been able to leapfrog the coverage gap to date.

Social health insurance (HI) has proven even more difficult to mandate. In Latin America, where social security agencies typically collect contributions for pensions and health from the same payroll tax, coverage closely follows income levels as in Figure 1. In east Asia, where premiums typically finance a smaller proportion of programmes, universal HI has generally been introduced at per capita income levels between $5,000 and 10,000, and in most cases eventually on a mandatory basis even for rural populations.1 Two examples from that region which have achieved universal coverage of HI are Taiwan and Korea, both richer and more urbanised than India (Taiwan was 57 per cent and Korea 66 per cent urban when universal HI was introduced). However, even in those cases, universal HI coverage was achieved gradually, with for example Korea taking around 12 years to achieve full coverage from initiation of its expansion effort. In addition, only Taiwan had a really comprehensive package of health services insured. Thailand – a frequently cited case for expansion of health coverage, and with GNI per capita of around $2,500 in 2004

– took from the mid-1970s to 1993 to expand HI to 50 per cent of the population, and following its 1993 reforms had covered 69 per cent of the population by 2000. The remaining population was covered only from 2001 by the tax-financed 30 baht scheme.

The experience thus suggests that neither mandates nor voluntary social insurance have achieved broad coverage in low income countries to date, though with exceptions such as pre-reform China. While historical experience need not define the trajectory of every country, no country at India’s level of income, urbanisation and informality has achieved the ambitious goals laid out in the NCEUS proposals. As discussed below, subsidies will be required to encourage some workers to join voluntarily. Also, it will be necessary to control the ratio of administrative costs to contributions if the scheme is to be viable. In these areas and others, the emerging international experience may be useful. The next two sections refer to some of these lessons.

Old-Age Pensions

The NCEUS proposal to address the risk of destitution in old age is twofold. For those BPL, a flat Rs 200 monthly pension (indexed) is proposed over age 60, essentially as an expansion of the current National Old Age Pension Scheme (NOAPS) which provides “social pensions” for destitute elderly over 65, poor widows and disabled people. The idea of providing a transfer to those who cannot afford to participate in a contributory scheme is

Economic and Political Weekly August 12, 2006

sensible and many countries have chosen to implement this type of programme.2 Coverage here depends on the availability of fiscal resources and the competence of administration of the programme. The non-contributory nature of the scheme and the historical problems with BPL targeting accuracy raise further questions that are beyond the scope of this paper.

An even greater challenge is enticing informal sector workers (and their employers) to contribute to a contributory pension scheme. Here, the NCEUS proposes a subsidised defined contribution (DC) scheme with a guaranteed nominal 10 per cent rate of return for APL workers. Is the DC approach better than alternatives? What are the implications of a guaranteed return?

On the first question, it is clear that the standard defined benefit formula – whereby a formula is applied to a known reference wage to calculate the pension – is not suitable for the unorganised sector. This leaves two options, both of which are being tried in different developing countries. First, a flat contribution and benefit schedule is specified, often for specific occupational groups like farmers. A nearby example is that of Sri Lanka where a scheme for farmers and fishermen was implemented in 1990 and now covers around 6,50,000 members.

A second approach is to offer a defined contribution scheme and to subsidise it through a government match. Subsidised DC schemes have been recently introduced in Mexico and also through legislation in Vietnam and the Dominican Republic. Several provinces are considering this approach in China. Interestingly, the longest experience with this approach can be observed within India itself. In West Bengal, the state government matches the provident fund contributions of close to 7,00,000 unorganised sector workers earning less than Rs 3,500 per month.

Admittedly, the experience to date is short. Nevertheless, comparing the two approaches leads to several conclusions relevant to the NCEUS proposal. First, the flat payment scheme approach, while initially simple, results in huge uncertainty both for the covered worker and the government as sponsor. Sri Lankan experts have estimated that the promised pension at retirement would be practically worthless under moderate inflation rates while low inflation could lead to a large government liability [Eriyagama et al 2003]. In comparison, the account balance in a DC scheme increases every year depending on the interest rate. As long as this is tied to a reasonable benchmark or the result of appropriate investments, it should produce a reasonable pension. From the government’s point of view, the subsidy is upfront, explicit and transparent. The DC approach also allows potential portability as workers move between sectors or geographic regions.

The NCEUS proposal includes a guaranteed nominal return of 10 per cent annual in its DC scheme. Because of their popularity across the world, a good deal of work has been done on DC guarantees with lessons relevant to India. The first point is that both relative and absolute rate of return guarantees are observed. In the former case, the guarantee might be related to a benchmark (e g, government bond yield) or, as is common in Latin America, the average return for private pension fund providers. With a few exceptions, guarantees of absolute nominal returns are found in countries where inflation and interest rates are relatively stable. Singapore and Malaysia, for example, promise their provident fund members a 2.5 per cent rate of return as a minimum. Switzerland recently lowered the guaranteed return on its DC schemes from 4 to 2.5 per cent nominal.

In principle, however, a nominal rate of return guarantee can be very costly to the guarantor. Its value can be calculated on the basis of historical interest rates in the same way as an option. And the longer the duration of the guarantee, the costlier it becomes. A commitment to crediting returns for one year (as done each year in the EPFO at a lower interest rate) implies a much lower liability than a 30-year guarantee of the same return. In the US, when proposals to guarantee a new DC scheme were costed, the estimated cost of absolute nominal guarantees was found to be fiscally unsupportable. In the case of India, preliminary cost estimates arrive at a similar conclusion. Fiscally, a guaranteed rate of return of 10 per cent is neither transparent nor is it likely to be sustainable.

The financial sustainability of a pension scheme is paramount to its success and credibility. The NCEUS presents cost estimates only for the next five years. In light of global ageing trends, most countries now formulate pension policies using long-term projections (typically at least 40-50 years). While there is great uncertainty as to how many people would take up the NCEUS scheme, projections should be generated under upper and lower bound scenarios for more than five years (over and above costs related to the guarantee). Also, the cost of the pension for BPL individuals should be included and reflect population ageing.

Additional areas where the NCEUS proposal would benefit from a review of international experience in pension design include: (i) maximum age for joining the system (there is none in the report, though nearly all DC systems in developing countries have some maximum age above which workers cannot join); (ii) allowed withdrawals for certain contingencies (which in other developing countries have tended to undermine the financial protection objective of the pension); (iii) design of disability and survivors insurance; and

(iv) modes of pension payout after retirement (where a third option not allowed for in the NCEUS proposal – scheduled withdrawals of accumulation – seems worth considering). These and other issues on pensions warrant further attention.3

Health

NCEUS proposes voluntary HI for hospitalisation expenses and sickness cover up to modest defined limits, covering both the contributor and dependents. A full discussion of the challenges of rolling out HI to unorganised workers is not warranted here, though there are threshold issues of a HI model where the public delivery system is in principle free, and of demand-side constraints in the absence of basic quality care in many areas. This section focuses on a couple of lessons of international experience with expansion of HI which suggest that particular caution is needed in attempting to expand HI relative to other forms of social insurance. What experience makes clear is that there are significant risks in expanding HI coverage which – if not carefully attended to in design and implementation – can both undermine the financial protection benefits of HI for workers and risk, increasing the financial burden of healthcare for the uninsured.

Unlike pensions, voluntary HI faces two fundamental problems – adverse selection and moral hazard. Adverse selection takes place when those more likely to need medical services join the scheme while healthier individuals do not. Internationally as well as in India, attempts to deal with it include mandatory group or community participation or, alternatively, heavy subsidisation from the state budget. These

Economic and Political Weekly August 12, 2006 solutions are not, however, contemplated in the NCEUS proposal. Nor are there any provisions for dealing with the increasing price of services due to increased demand from the new scheme.

Evidence of increased use of health services and resulting medical price inflation has been documented in several east Asian countries following HI expansion. This has been analysed for Indonesia, Taiwan, Korea and Philippines. Gertler and Solon (1999) in Philippines for example found an over 23 per cent price mark-up for insured patients. Several countries have tried to control this impact through imposition of price controls, but this has not prevented provider-induced cost escalation for the insured. Thus while health spending in aggregate increased (as intended), the medical benefits for workers of the increased spending were significantly less than proportional due largely to moral hazard issues on the part of both insured patients and health service providers.

Moral hazard in purchase of health services will be greater where price elasticity of demand is higher. Gertler and Hammer (1997) find that price elasticity of demand for health services in developing countries is considerably higher than for developed countries (between -0.5 and -1.0 in developing countries, against an estimate for developed countries of around -0.2). The cross-country finding has been confirmed in household data on health spending following introduction of insurance, for example in rural China, Indonesia, and Jamaica. While countries have tried in several cases to control patient level moral hazard through co-payments, this has not typically controlled spending escalation as hoped. In the NCEUS proposal, there are no measures contemplated to mitigate these effects.

Interestingly, such effects were not found in Singapore, which used an unusual HI model of individual medical savings accounts combined with risk pooling for catastrophic care. The so-called Medisave account is an individual account which must first be exhausted before the beneficiary accesses the larger risk pool of the system (i e, Medishield, which covers cases of serious illness and is financed from an additional premium payment), thus providing individual incentives to control unnecessary usage of health services.

An additional lesson of developing country experience with expansion of HI coverage is that some form of public subsidy has typically been used to induce participation in voluntary HI systems (see the table). Some countries have achieved remarkable success in expansion of coverage under such schemes, e g, Colombia, which moved from 23 to 62 per cent HI coverage in 10 years following introduction of HI subsidies for the poor in 1993, though with other reforms also. An early example of an insurance-like system was China’s Cooperative Medical System, which at its peak in the 1970s was estimated to cover 90 per cent of the rural population using a combination of compulsory prepayment by the residents, village contributions, and general revenues. At the same time, Asian systems have typically also had significant co-payments from participants.

There are therefore a range of factors in design of a HI scheme for informal sector workers which international experience suggests need more attention as discussions on SI reform progresses. This would also suggest reviewing the experience of non-insurance approaches such as the Thai 30 baht scheme for service coverage expansion. A range of factors, including moral hazard, administrative complexity, and current supply side and provider payment systems suggest that HI is the most challenging and high risk element of the NCEUS proposal.

One strength of the NCEUS report is recognition of the administrative challenges of reaching a large, heterogeneous, and disbursed population, and the risk of high transaction costs that this implies. However, having recognised the challenge, the proposed solution on cost containment is a somewhat blunt one of capping on administrative fees of 5 per cent of central and state government contributions. What evidence is there that such a figure is realistic?

In neighbouring Sri Lanka, the Farmers and Fishermen Pension Fund offers fewer benefits yet ran administrative costs over 25 per cent of contribution income during its first decade. Kannan (2002) cites the Kerala experience, where welfare funds operated with costs much higher than 5 per cent and in some cases spent more on administration than benefits. Even the large and long-established central Beedi Workers’ Welfare Fund has had administrative costs of 7-8 per cent, despite cess-financing.

International experience suggests that complex transactions in a rural setting can be especially expensive. A review of 82 HI schemes in developing countries in the 1990s found that simpler services implied lower administrative demands on the system, suggesting that the proposed NCEUS scheme may be at the higher end of administrative demand.4 Overall, in international experience from informal sector HI schemes, administrative costs have been in a range from 5-17 per cent of income.

On the other hand, there is evidence from mandatory SI programmes that there are scale economies in record-keeping and other standard processes. In this context, the assumed participation in the NCEUS proposal could result in lower cost ratios. However, voluntary take-up of the scheme is by no means guaranteed. Given the need to set up a new bureaucracy, recordkeeping apparatus and monitoring processes, the fixed costs of implementing the scheme could be high. If the optimistic assumptions for participation do not materialise quickly, the 5 per cent figure

Table: Selected Examples of Premium Subsidy and Cost Recovery in Informal Sector HI Schemes

Economic and Political Weekly August 12, 2006

could turn out to be a substantial underestimate.

A parallel risk is that low costs are achieved at the expense of quality of service. Many SI schemes in developing countries today cannot ensure accurate tracking of contributions or process benefit payouts in a timely manner. In pensions, countries that have introduced DC schemes have quickly become aware of the need to track individual account balances on a regular basis or risk losing confidence of members. Delays and errors would lead the new scheme to lose credibility and reduce voluntary participation (completing the link with higher costs through fewer scale economies).

It would be useful for those implementing any new scheme in India to review the international experiences with setting up record-keeping and other systems to see, for example, how long it took, how much it cost and what approach was taken (e g, contracting out of services versus in-house provision). On the other hand, India has two potential advantages that can be brought to bear on both the cost and the quality of administration. The first is the possibility of reaping very large economies of scale due to India’s size. The second is the possibility of harnessing state of the art information and communication technologies available in India as nowhere else in the developing world. However, even with these advantages, the major administrative demands implied by the NCEUS proposals – both within and outside the public sector – raise the more fundamental question of institutional capacity, around which further examination of international and Indian experience seems desirable.

Conclusions

The NCEUS proposals address a pressing need in Indian society and should generate overdue attention to the social security needs of unorganised workers. However, international experience suggests that it is very ambitious and risks missing achievable objectives by over-reaching. The recent experience of Indonesia – a country with around one-fifth India’s population and a similar income level – provides a cautionary tale. After passing a new social security bill in 2004 providing for old-age pension, health insurance, a death benefit and disability insurance for all workers, two years have passed with essentially no progress toward implementation. It is an instructive instance of how setting unrealistic targets can result in implementation paralysis.

The international experience to date is that expanding coverage of contributionbased SI programmes is a gradual process which is inextricably tied to levels of development, industrialisation, and informality. There is no evidence of a magic bullet to achieve high coverage at low levels of income, and thus no precedent for what has been proposed by NCEUS, including in India, where previous initiatives such as universal health insurance security (UHIS) have achieved minimal penetration.5

India is, however, a unique country at a unique point in its history. High growth rates allow for initiatives that could not have been considered a decade ago. India’s scale could prove an advantage to the extent that a uniform platform for everything from regulation to record-keeping can be designed and implemented, even if administered on a decentralised basis. Finally, India has the potential to harness technology and its ample human capital to leapfrog some of the administrative hurdles that face other countries.

An answer may lie in an incremental approach to addressing the coverage gap. Two dimensions – the nature of the risk to be covered and the population targeted

– can be distinguished. Clearly, some types of benefits are easier to administer. Health involves considerations of supply-side constraints and moral hazard that old-age and death benefits do not. It is no accident that more countries have contributory pensions than health programmes. Sequencing the introduction of benefits like LI and pensions could allow time to build institutional infrastructure and credibility before tackling more difficult programmes.

The second dimension to be considered is the heterogeneity of the unorganised sector. The 300 million target population varies in many respects: by income levels, risk attitudes, age and SI needs, sector and location of employment, membership of potential intermediary organisations and other factors. These are key considerations in focusing the efforts of a new, voluntary programme in order to achieve critical mass early.

The objections to such an incremental approach are well taken. The most marginalised workers may receive less immediate attention, and greatly needed financial protection against health shocks comes only gradually. However, a failed attempt to cover everyone very rapidly could result in paralysis and loss of credibility that could make future initiatives for voluntary take-up more difficult. In contrast, an incremental approach could feasibly achieve important benefits for the poor in the near future, and building credibility to provide a sustainable strategy for realisation of more comprehensive gains in their financial protection over time.

EPW

Email: Pokeefe@worldbank.org rpalacios@worldbank.org