Responding to the challenges of social health insurance in African countries

Abstract

With most sub-Saharan African countries facing problems of raising revenues for financing the delivery of an essential package of health services, there has been growing interest in social health insurance (SHI) as shown by efforts in Ghana, Kenya, Lesotho, Nigeria, Rwanda, South Africa, Swaziland, Tanzania, Uganda, Zambia and Zimbabwe. While these health financing constraints are linked to broader public affairs of slow, or even negative macroeconomic growth, civil strife and political instability, there is little consensus on what should be the appropriate institutional arrangements and policies to mobilise resources for effective provision of essential health services. Given an observed inclination towards SHI, this paper provides some answers to the challenge of making it work in African countries. The paper discusses a number of policy choices and trade-offs that health planners may consider when implementing SHI to generate financing for the provision of essential health care benefits.

Keywords:

• health financing
• health system reform
• social health insurance
• user charges

1. Introduction

It is well documented that the amount of public funds generated on the back of slow or negative macroeconomic growth is inadequate to support delivery of an essential package of health care to African populations. According to the African Union (2006), only 19 out of 52 African countries spend the recommended annual threshold of $34 per capita for delivering essential health services. Considering sub-Saharan African countries, only 12 countries spend above the recommended threshold. Nearly half of African countries spend less than 5% of GDP on health care, 22% spend between 5 and 8%, 3% spend 8 to 10%, and 2% spend about 10%. Projections of health financing data to 2015 suggest that 80% of West and East African countries and 50% of Central African countries will spend less than $34 per capita on health care. By 2030, over 50% of East African countries and a quarter of West African countries will still be spending less than $34 per capita. The situation is actually worse than this, since by 2015 the annual spending threshold will have increased to $38 per capita.

Sound macroeconomic policies, political stability and economic growth are a pre-requisite for health system development by means of determining the absolute pot of revenues available. Yet existing systems and institutional arrangements restrict the potential pot of revenues that could be mobilised at any level of economic growth without disruptions to the insurance function, i.e. protecting African communities from the financial risks of ill health. The situation in African countries is such that some pluralistic, mixed system of funding is inevitable. For this pluralistic approach, health planners have the following means of mobilising health financing: (1) general and earmarked taxes derived from direct sources such as personal and corporate income tax or indirect sources such as value added tax (VAT) and sales tax, (2) social health insurance (SHI), (3) private health insurance (PHI), (4) voluntary community-based health insurance (CBHI), (5) user charges, (6) medical saving accounts (MSAs), sometimes referred to as health savings accounts (HSAs), and (7) external donor funding.

Of the 52 countries studied in the African Union's 2006 report, 17 countries showed a ‘high’ level of dependence on donor funding exceeding 25% of total health expenditure. Notably in two countries, Benin and Eritrea, reliance on donor funding was at least 60% of total health expenditures. In 23 countries, however, dependence on donor funding was less than 10%. In general, across African countries, there is a disproportionate reliance on user charges, with a growing reliance on CBHI and SHI (and to a lesser extent PHI) and a marginal or non-existent role for HSAs/MSAs. As reported by the African Union (2006), out-of-pocket payments account for 80% of all private spending on health care and 50% of overall country-level health care expenditures. Given these expenditure patterns, the challenge is to find the appropriate mix of health financing mechanisms that will maximise the potential pot of revenues at a given macroeconomic growth level and maintain the insurance function.

In response to this challenge, Ghana, Kenya, Nigeria, Rwanda and Tanzania have instituted structures for SHI, while Lesotho, South Africa, Swaziland, Uganda, Zambia and Zimbabwe are considering it (Lem, 2009; Mathauer et al., 2011). On the basis that SHI appears to be the preferred health financing instrument, this paper conducts a normative (qualitative) analysis of how to make it work, taking into account the costly consequences of poor SHI design or implementation, and the less than encouraging outcomes in some countries. Lloyd-Sherlock (2006), for instance, documents experiences from Argentina and Mexico of how SHI tied to the formal labour market undermines the solidarity principle and promotes distribution away from ‘uninsured’ to ‘insured’ individuals where SHI coexists and competes with deteriorating public health care systems for funding. In addition, SHI can cause the proportion of formal sector workers to fall, while the difficulties of extending insurance coverage beyond the formal sector lead to the creation of separate, downgraded second-class systems for the poor.

This paper therefore aims to offer responses to most but not all of the concerns expressed about SHI (in African countries). I follow Kutzin's (2001) recommendations to focus on the functions of health systems (i.e. collection and pooling of revenues, purchasing and service provision) while paying attention to the existing institutional arrangements and policies in any country. For convenience, I categorise the functions into two broad streams: (1) revenue mobilisation and pooling, and (2) purchasing and service provision. But, irrespective of the institutional arrangements for resource mobilisation, pooling, purchasing and service provision, operating these functions will be of little value in the absence of measures that ensure efficient geographical distribution of service providers and appropriate public transport networks and policies that ensure consumers have physical access to health facilities. Figure 1 captures these issues diagrammatically.

Figure 1: The analytical framework

Using Figure 1 and health economic principles, I catalogue a number of options and trade-offs in offering risk protection and generating adequate revenues through SHI. It is not my intention to advocate a specific health care system structure or policy, and for brevity reasons I do not present a systematic review of empirical evidence on the positives and negatives of SHI. By clearly describing the reasoning behind the options presented, health planners, given the mandate bestowed on them by their respective citizenry, can adopt and adapt the options best suited for their domestic contexts on the basis of feasibility studies (including assessments of the views, interests and power of different stakeholders) and scenario analyses that forecast the possible impacts, followed by benefits-incidence analysis to determine whether the resources mobilised have been spent wisely.

Note that in this paper an ‘essential’ health care package refers to a defined minimum set of interventions (i.e. medical inputs and services), for which there is current clinical and economic evidence of the marginal health benefits (net of costs). This will invariably cover interventions that have been assigned an ‘essential’ status by the World Health Organization and, depending on the social context, herbal medicines and traditional healing practices that are known to be ‘effective’ on the basis of cultural experience and observational scientific evidence. This essential package of health care benefits will, by definition, be an indispensable component of all consumers' preference schedules – insurance protection for which should be demanded for and by all members of a community.

Another point to note is that there is some confusion as to what exactly SHI means. As discussed by McIntyre & Van den Heever (2007), we can think of SHI as a variant of mandatory health insurance where only certain groups (for example, formal sector workers) are legally required to enrol. The other variant is national health insurance (NHI), where the legal requirement to enrol is universal and coverage is for the whole population irrespective of whether individuals contribute or not. In this paper, SHI is synonymous with NHI: I focus on mandatory health insurance designed to achieve universal coverage with community-rated premiums collected for an essential or prescribed minimum package of benefits.1

The paper is structured as follows. Section 2 presents the rationale behind my support for implementing SHI via non-competing social risk pools and assesses ways of organising an essential quantity of health insurance protection for individuals working in the formal sector as well as for those in the informal sector. The section also looks at the role of financial equalisation in reducing the costs or risks of social health insurer insolvency to taxpayers. Section 3 evaluates policy developments in Ghana's National Health Insurance Scheme (NHIS) against the insights developed, and Sections 4 and 5 present discussion and conclusions.

2. Making social health insurance work

2.1 Setting the stage

As noted above, a pluralistic mixed system of funding is necessary if African nations are to spend at least the recommended annual threshold on essential health care – and there is a growing trend towards implementation of SHI. But prior to this, most African countries adopted user charge policies to correct revenue shortfalls in public budgets allocated for the supply of essential health care. We can depict the health care financing situation in a hypothetical African country using Figure 2, which shows a social demand curve (D _social ) made up of aggregated individual private demand curves (D _private ) and a marginal external benefit (MEB) curve from consumption of essential health care. The tax subsidy (So) curve refers to possible combinations of tax funding and alternative revenue sources. Before user charge policies, macroeconomic constraints meant that available tax-funded budgets allocated for essential health care (represented by the rectangle ABQ0P0) only allowed Q0 of care to be provided ‘free’ of charge – which is far from the social optimum Q* given (current) health care production costs Co. With access costs P0, this creates an unsatisfied demand of Q* – Q0.

Figure 2: Mixing methods of health care financing

The thinking behind user charges is as follows. If revenues mobilised via user fees are hypothecated for services provision (and this does not lead to cuts in the allocated health care budgets), a user charge P1 – P0 will enable provision and consumption of services at Q1, which is closer to Q*. A user charge of P2 – P0, itself a subsidised price that will not survive in a competitive health care provider market, will satisfy both private and social demands with no excess demand (i.e. point O, with ACOP2 being equal to ABQ0P0). Besides the empirical difficulty of health planners locating point O (the ‘optimal’ user fee, tax subsidy and quantity combination), an even greater problem with user charges, paid at the point of health care consumption, is the lost welfare gains from insurance protection covering unpredictable demands for Q* – Q0 of services – albeit some insurance protection is implicitly provided by making Q0 of services ‘free’. Indeed, in the absence of health insurance, it is appropriate to charge less than P2 – P0 or even abolish user charges to reduce the financial risks consumers face. To offer some implicit insurance protection (and reduce inequalities in health care consumption for low-income people, those with high medical needs and high access costs), the public health system will have to supply less than Q* of care.

It makes sense, therefore, to transform health care consumption at the higher user charge of P2 – P0 into premiums for multi-period, time-consistent or guaranteed-renewable health insurance. This is the source of the (renewed) interest in SHI. To the best of my knowledge (or at least across the sample of African countries implementing or considering SHI), this means mandatory participation in non-competing risk pools with (income-related) community-rated premiums for a defined package of essential health care benefits. The obvious problem with such SHI pools is the welfare costs of zero competition. However, from this welfare cost we have to subtract the gains from avoiding adverse selection that lead to unsteady patterns of health insurance coverage; an outcome that is in direct conflict with the objective of insured universal access to essential health care.

Yet another welfare gain from non-competing SHI pools with (income-related) community-rated premiums is that they have built-in guaranteed renewability properties. They offer not just insurance protection within a single period but also multi-period protection against what is called premium or reclassification risk; that is, the uncertain increases in premiums (and exposure to catastrophic out-of-pocket payments if insurance become ‘unaffordable’) associated with the worsening of health risks over an individual's lifetime. The way a non-competing SHI pool provides this multi-period insurance protection is by resupplying the pool with low risks by making participation mandatory (with no opt-out provisions) and because if we have a large ‘stable’ SHI pool, then any increase in community-rated premiums takes place more slowly when people become high risks. The picking and shifting of risk types that occur under competition will fragment SHI pools and distort the resupply of low risks.

There are alternative solutions to insuring reclassification risk when competitive SHI premiums are risk rated. Cochrane (1995), for instance, proposes the use of savings accounts and lump-sum ‘severance’ payments, while Herring & Pauly (2006) describe the ‘optimal’ schedule of frontloaded but declining premiums where the ‘loading’ on single-period premiums covers the expected costs due to the worsening of health risks over an individual's lifetime. With perfect capital markets and perfect foresight, frontloaded multi-period insurance will charge a single lump-sum premium for lifetime coverage. Although it allows consumers to switch health insurers freely, Cochrane's (1995) proposal involves indemnity payments that require specification of all possible states of nature. On the other hand, insurance with frontloaded premiums may be unaffordable and, if affordable, associated with ‘high’ switching costs and locked-in consumers in a way that is not significantly different from mandatory participation in non-competing SHI pools.

Obviously, with competing SHI pools and community-rated premiums, the welfare losses from adverse selection and distortions to multi-period insurance protection can be reduced by risk equalisation systems, but then such complex systems are far from perfect (Van de Ven, 2011) and are unlikely to fit well within the institutional, capacity and resource constraints of most African nations. The issue is that there are fixed (start up) and variable costs associated with operating a risk equalisation system to reduce the scope for risk selection. The more intense the competition is, the greater the shifting and picking of risk types and the higher the variable costs of risk equalisation. Risk equalisation will have to be not only prospective at the beginning of a specified time period but also concurrent with ex post retrospective cost-based payments to compensate for the leavers and joiners in any given risk pool. The complexity and efficiency tradeoffs involved in equalising risks are perhaps the reason why insurer competition is not on the reform agenda of African countries.

Whatever the exact reasons are, it is clear that African countries considering SHI prefer to start or work with non-competing risk pools. However, the welfare costs of non-competing SHI pools need not be large, as the two key functions of health insurers, (1) risk pooling and (2) control of moral hazard and cost containment, are separable. Risk pooling involves the administrative tasks and costs of collecting premiums and providing information to consumers, verifying and processing provider-payment claims (within the shortest possible time), dealing with fraud and abuse, and investing premium revenue where applicable. These tasks do not directly affect the costs, quantity and quality of health care provided. The second function of moral hazard control and cost containment largely involves getting health care providers to supply services at reasonably low costs, although some demand-side measures (i.e. consumer cost-sharing and information provision) are desirable. The second function can be executed independently of having competing insurers, not to mention the observation that there is no automatic translation of efficiency incentives generated from insurer competition to health care providers. Non-competing social health insurers can therefore be ‘efficient’ by delegating moral hazard control and cost containment functions to another agency while focusing their efforts on administrative efficiency in pooling risks. In effect, the non-competing social health insurers will act as ‘financers’ of health care expenses while the other agency acts as a ‘strategic purchaser’ charged with the task of improving the efficiency of health care production and consumption.

Health planners will, however, have to find ways of making a mandate for SHI work, knowing that a 100% enforcement effort is unlikely. Figure 2 offers a possible solution. Mandatory enrolment could be made more attractive by removing or reducing the tax subsidy on user fees such that consumers who opt out of SHI are faced with, at least, the full cost (≥ Co – P0). If desired, the user fees charged could be (re)structured to include a mark-up as a penalty for non-enrolment or delays in enrolment. Consumers who opt out should be faced with the full unsubsidised cost, since we cannot tell whether imperfectly informed preferences for self-insurance are ‘rational’ when consumers are not faced with the full costs of their choices. Of course, removing the tax subsidy will ‘hurt’ mostly consumption by the poor and those with average incomes, while the rich will probably be unaffected if they already make private purchases at prices above Co – P0, even though they have paid their taxes towards publicly provided health care. Yet charging zero or lower user fees to reduce the financial risks of illness only makes sense where there is no insurance protection. With full-cost charges for those who opt out, consumers will be able to see that tax-subsidised SHI offers them better value than being uninsured.

There might be an unpleasant trade-off here. Suppose that under tax-subsidised charges and a less than fully enforceable SHI mandate, there is W number of uninsured people. Unless full-cost charges have no effect on SHI enrolment, the number of uninsured will be αW where 1 – α is the fraction who responded to the incentive to enrol. Now uninsured consumers may forego medical care, leave conditions untreated and suffer deterioration in health outcomes because they cannot afford full-cost charges or from being exposed to unaffordable catastrophic expenses. There are therefore losses either way, but we will expect full-cost charges to be more favourable to social welfare since αW < W and the greater the risk pooling we have, the more stable the flow of revenues for producing the right quantity and quality of essential health care.

However, when a mandate is not 100% enforceable and enrolment in SHI becomes ‘voluntary’, the cumulative effects of various household and community-level factors – for instance, income, education, confidence in or distrust of insurers, proximity to health facilities and health care quality (see Mwabu et al., 2002) – may strengthen or undermine the intended effects of full-cost charging. What we can say is that removing the tax subsidy (holding all else constant) should alter the marginal benefit-cost calculus in favour of mandatory enrolment in SHI.

2.2 Institutional and governance structures

In the subsection above I argued that, given mandatory community-rated premiums collected by non-competing SHI pool(s), it is necessary to separate risk pooling functions from moral hazard control and cost containment in order to reduce the welfare cost of zero insurer competition. How could this be implemented? One possible option is to define SHI pools according to geographical regions, bearing in mind that the size of the risk pool should be large enough but without the diseconomies of scale arising from complexity in managerial, purchasing and stewardship functions. This fits with the typical trend for governance and administration of public policy and services to be planned along geographical lines. Unlike a national-level insurance pool, geographically defined SHI pools will foster local accountability and offer the ‘scope’ for yardstick (not direct) insurer competition on the basis of designated tasks and responsibilities, specifically the functions of risk pooling.

Using Figure 1, I separate the functions of the public body responsible for health care and health in my hypothetical African country, i.e. the Department or Ministry of Health (MoH), into two streams: one for ‘revenue mobilisation and pooling’ (RMnP) and the other for ‘purchasing and service provision’ (PnSP). I give each stream of the MoH clearly defined country-specific operational roles, objectives and performance targets related to regulation, information provision, and monitoring and enforcement of policies affecting the functions listed on either side of the dotted line in Figure 1. (See also Figure 3.) That is, the RMnP agency will be responsible for the geographically defined SHI pools while the PnSP agency governs health care providers.

Figure 3: Fitting the pieces together

The administrative tasks of the geographical SHI pools that are overseen by the RMnP agency will include (1) collecting premiums and adopting measures to prevent insurance abuse and fraud, and (2) operating under the lowest feasible administrative costs. The SHI pools acting as ‘financers’ of a package of essential services in quantities corresponding to Q* (not just Q* – Q0) will be required to provide out-of-geographical area coverage in emergency situations, with the appropriate insurers being reimbursed for the expenses incurred. This is to avoid (1) the development of parallel health systems that create layers of ‘new’ agency relationships on top of existing ones; (2) the duplication of administrative, information, premium or tax collection and health care provider payment systems; and (3) situations where the welfare gains that are possible depend on the inefficiencies in each parallel health system and how they interact with each other.

The PnSP and RMnP agencies together will act as the ‘strategic purchaser’ (a hub for researching, developing and implementing strategies and policies for ensuring efficiency in health care production and consumption). This is a countermeasure to correct the presumably low-powered efficiency incentives for moral hazard control and cost containment that come with no insurer competition. The PnSP agency will work with the RMnP agency to (1) design policies and interventions to ensure appropriate cost-containing procurement of medical inputs, and use of the ‘right’ mix and quantity of medical inputs in the production of health care; and (2) apply measures for reducing access costs, P0. As the ‘strategic purchaser’, the PnSP and RMnP agencies will also work together to determine credible premiums for each geographical SHI pool. Note that I use the word ‘credible’ to describe premiums that have sound actuarial foundations or that are derived using statistical and probabilistic methods.

In the context of Figure 2, these credible premiums should, at least, be set and revised according to (1) changes in the aggregate loss probability of each geographical population covered, (2) robust forecasts of health care costs for each SHI pool adjusted for inflation in prices of medical inputs etc., and (3) the tax-funded subsidy ABQ0P0. It will be prudent to compile and track trends in health care expenditure in periods t₀ and t₁ and, using a combination of all available data (and actuarial judgement), to forecast costs and premiums for period t₂, t₃, t₄ and so on. This could be done using the health-insurance simulation software SimIns, which is designed to assess the feasibility of mixing health care financing methods and to estimate premiums, health care resource use and costs that will ensure the financial equilibrium of SHI systems. See Mathauer et al. (2011) for an application of SimIns to assess the financial feasibility of the proposal to implement SHI in Lesotho.

2.3 The formal SHI ‘market’

With a payroll to pay premiums from, the formal sector, however ‘small’, will be the platform for country-wide implementation of SHI. The formal sector is also where enforcement of a mandate is likely to approach 100%. I highlight below three issues for health planners to consider.

The first is whether mandated premiums should be paid by individuals or households. This paper favours premium payments on the basis of households. That is, an individual's membership of the SHI pool depends on bringing all household members along. This potentially accommodates skewed income distributions (individuals may be poor, but ‘rich’ when included in a household unit) and potentially avoids social exclusion of non-working individuals (the elderly and children). Dong et al. (2004), for example, report that household heads and other members exhibit different willingness-to-pay (WTP) for health insurance. Where health insurance is purchased voluntarily on an individual basis, differences in WTP open up the possibility that a household (when considered as a unit) will purchase inadequate insurance protection – especially where individuals are risk tolerant or estimate their health risks incorrectly, and when there are conflicts and non-cooperative behaviour between household members and egoistic objectives override altruistic concerns.

One would expect that an SHI mandate with households as the insured units would increase coverage for essential health care, not least by reducing the number of zero insurance demands by individuals who believe ‘they will never get sick’ but then become burdens on their immediate and extended families. It should also reduce attempts by individuals to ‘crook the system’ by using eligibility cards belonging to household members and other individuals. For example, insured consumers may feign illness in order to access medical care on behalf of their uninsured household members. Yet if a mandate is not 100% enforceable, it is possible that households will enrol mostly members with high health risks and/or risk aversion. I believe that if even crude census or polling registration data are available, identifying households and their members to check this form of adverse selection should be reasonably straightforward. Where there are information gaps, I suggest conducting demographic surveys to support the household mandate.

The second issue is whether formal-sector workers should pay premiums (on behalf of their households) from their pre- or after-tax income. I can think of two reasons why premiums should be paid out of pre-tax income. One is to make SHI enrolment attractive by leaving low-income earners with adequate spending money for leisure and other non-health-care consumption, and the other is to ensure that high-income earners offer political support to income-related community-rated premiums. The inequitable, higher marginal increase in real disposable incomes for high-income people will be the ‘cost’ of ensuring the political acceptability and survival of SHI. Considering this necessary inequity and that the tax deduction on the portion of income used to pay SHI premiums means lower general revenues for the provision of other public goods and services, health planners will want to limit the tax deduction to only essential health care benefits that are of ‘higher’ social value. It should not be applied to purchasing insurance for supplementary non-essential health care.

The third issue is whether the responsibility for premium payments should rest solely with formal workers and their households with no mandatory contributions from employers (I will call this an ‘individual’ mandate) or whether employers should be mandated to make contributions on behalf of their workers (i.e. an ‘employer’ mandate). An individual mandate avoids distortions in individuals' job choices, while an employer mandate creates a ‘job lock’. Employees will be reluctant to move to new jobs where they will be most productive if the total compensation offered by prospective employers, i.e. higher wages and employer SHI contributions (above the mandated minimum) is no more than what they already have.2 Furthermore, an individual mandate avoids distortions in firms' demand for labour and answers concerns that mandatory employer contributions for SHI will increase labour costs and job losses, provoke labour strife and negatively affect foreign direct investments – as argued by the Federation of Kenyan Employers against the National Health Insurance Fund in Kenya (Carrin et al., 2007). With an employer mandate, a political compromise to address these concerns is to exclude employer contributions to premiums from taxation, but this will only increase the costs of tax concessions needed to make SHI feasible.

Under an employer mandate, possible distortions in labour markets include laying off workers, or downsizing or outsourcing in attempts to accommodate increasing labour-compensation costs. This is particularly the case when employers cannot tell how rival firms will adjust wages to reflect mandatory premium contributions, where there are rival firms free of such mandates, and where there is rigidity in (public sector) wages. If sacking workers is the ex post response from employers in the short run, or mandated contributions force employers to demand ex ante less than the ‘optimal’ amount of labour (employer profits may fall in the process), this will increase short-run unemployment that in the long run may lead to lower wages, as employers need not offer prevailing or higher wages to hire new workers. However, this will be accompanied by transitory social costs of foregone revenues from a shrinking (formal) tax base. Alternatively employers may, over each short-run period, reflect their premium contributions in wages (or perquisites) to maintain jobs or in the form of increased prices for goods and services if this is feasible without a fall in profits. That is, sooner or later, the costs of mandated employer contributions will have to be shifted to consumers and/or to employees.

What makes an individual mandate appealing is that it does not stop employers from voluntarily financing or part-financing employees' premiums in order to attract, retain or motivate staff. If employers' contributions to SHI premiums are voluntary, their response to competition in product-, services- and labour-markets will be voluntary adjustments of wages (or total compensation) with no reason to lay off workers. This applies especially to small and medium enterprises, who presumably cover non-trivial segments of rural (non-farm) workforce in African nations and whose cash flows and operating margins are much more susceptible to variation (Kinda & Loening, 2010). Requiring mandatory employer contributions from these enterprises may have the unintended negative consequences of decreasing their profit margins, slowing their growth and increasing their risk of going out of business.

The economic thinking is that the net effect on wages will be almost neutral under an individual mandate, but without the labour market distortions of employer mandates (Krueger & Reinhardt, 1994). The effects of an individual mandate will be almost neutral, as adjustments to an employer mandate by consumers, employers, employees and government will not be instantaneous or continuous. An individual mandate means that SHI will have only minor ramifications for the wider economy. It means consumers are more likely to be cost-conscious, since any short- or long-run adjustments to premiums to incorporate the costs of moral hazard will be reflected directly in wages (not partly or artificially absorbed by employer contributions).

There is, however, one good reason why health planners may want to implement employer mandates and this has to do with situations where mandatory enrolment in SHI is not politically feasible or where there is a social distaste for anything that appears as coercion and restricting voluntary choice. By adding health insurance benefits to employees' compensation, the job-lock problem created becomes a de facto commitment to a given risk pool (Crocker & Moran, 2003). The effect is analogous to mandatory participation and, because of friction in labour mobility, these ‘stable’ employment-based risk pools to an extent also provide multi-period protection under community-rated premiums – though only as long as individuals do not lose their jobs or their employers do not go out of business. In adopting an employer mandate, however, one has to determine whether these benefits are worth the economic distortions and the costs of the inevitable political compromise.

2.4 The informal SHI ‘market’

For anyone concerned about SHI in African countries, the big question is how to mobilise premiums from (households consisting solely of) informal-sector workers, who have seasonal fluctuations in their income earnings and who tend to constitute the largest segment of the population in African countries. I suggest: (1) mandatory savings accounts to serve as the informal ‘payroll’ from which premiums will be paid, and (2) a social-reinsurance facility to support voluntary insurance demands already expressed through existing not-for-profit or for-profit CBHI schemes. I refer to these mandatory savings accounts as ‘health insurance accounts’ (HIAs) or ‘medical insurance accounts’ (MIAs), in that they provide a safe storage place to reserve money (earmarked) for paying SHI (or even PHI) premiums. I consider HIAs/MIAs to be the more appropriate option for households in urban and peri-urban areas and CBHI for rural households.

To avoid any confusion now and in the future, I must point out that the HIAs/MIAs proposed here are different from HSAs/MSAs designed to encourage households to build up precautionary savings as self-insurance against uncertain health care expenditures. The HSAs/MSAs commonly referred to in the literature do not pool risks; they do not implement an insurance contract. What is suggested here is the use of household income held in HIAs/MIAs to pay certain premiums. Given that the minimum savings needed is for a stream of annual premiums (and perhaps modest co-payments for a selective set of routine services that are most susceptible to moral hazard), a high cultural propensity to save, as in Singapore, is not necessary. High saving rates are only necessary if the accumulated deposits are for self-insuring what might be catastrophic medical bills (Dixon, 2002).

In the African context, a health savings experiment in rural Kenya, in which 113 ROSCAs (Rotating Savings and Credit Associations) were randomly allocated to informal and formal saving devices, showed a 93% uptake of HSAs/MSAs within six months of follow-up, and 97% after 12 months (Dupas & Robinson, 2011). The study also showed a reasonably high long-run usage of these saving devices: 53% were still using their HSAs/MSAs after 33 months of follow-up, and of those still saving in their HSAs/MSAs, 73% had made withdrawals and most of these withdrawals were for health emergencies. So if individuals in rural Kenya could reserve enough income in HSAs/MSAs, withdraw a portion of these savings for catastrophic health care expenditures, and still continue usage of HSAs/MSAs after almost three years, then paying a stream of certain but relatively small premiums via HIAs/MIAs should be affordable in spite of any liquidity costs of earmarking savings for SHI.

Considering the growth of rural agricultural banking and microcredit, microfinance or microsavings initiatives (Kloeppinger-Todd & Sharma, 2010), HIAs/MIAs should be reasonably easy to implement – offering the choice of paying premiums to CBHI schemes or to the geographical SHI pools already serving the formal sector. However, with seasonal income fluctuations of informal, particularly agrarian, workers, some flexibility in how premiums are paid is needed. This can be achieved by estimating annual flat premiums that, if necessary, could be paid in full or in small sums over defined time periods, say monthly or quarterly. Individuals would then be able to shift premium payments from periods of low income to periods of high income. To effect this, it may make sense to allow withdrawals from the HIAs or MIAs even when the balance is zero or negative – up to some defined debt limit.

This is consistent with arguments by Ahuja & Jutting (2004) that the poor living above the poverty line do not necessarily need subsidised premiums or exemptions. What is more important is access to (micro)credit/savings and borrowing options such that poor consumers can borrow money to pay premiums with the borrowed-money burden being compensated by a restored capacity to generate income ex post of an illness episode. Of course, borrowing money to cover premiums may not be feasible if health (and productivity) is not fully restored or restored in a shorter period of time than the loan repayment period. The problem, however, is amplified with user charges since it requires near perfect foresight to know the amount of borrowed money or savings needed to cover an unpredictable stream of health care costs.

One benefit of relying on HIAs/MIAs is that it will be no different from an individual mandate for formal-sector workers and it will avoid the inequities in disposable incomes (net of SHI premiums) between formal and informal workers as it will be under employer mandates. If SHI premiums for formal workers are to be paid out of pre-tax income, money held in HIAs/MIAs could also be subject to tax deductions up to some defined maximum limit (this will be at least the flat annual premium), above which taxes apply. This together with an individual mandate will ensure participation dynamics in the formal and informal markets do not differ by much. Using savings accounts to implement SHI should help check free-riding by informal workers and generate the maximum revenues possible over time periods where increasing and steady economic growth shifts more informal workers into the formal sector. It should address concerns raised by health maintenance organisations in Kenya that free-riding with income-related payroll premiums for equal benefits under the National Health Insurance Fund will place an extra burden on formal workers (Carrin et al., 2007).

Of course, for households living on or below the poverty line, exemptions from premium contributions will be needed. For any African country with a large segment of the population living below the poverty line, there will be significant hurdles to overcome when funding an essential package of health care. Here economic growth, direct income transfers and improved standards of living are the key issue, not institutional arrangements. Since one cannot wait for a favourable state of affairs, the next best option is exemption from premiums. Similar to user fee exemption, premium exemption means supplying either less than Q* of essential health care or, if Q* of care is to be provided, demanding higher premiums from the non-exempted and increasing the tax-funded health care budget. In both cases, the welfare gain is that with a certain premium3 one is not faced with an unstable cohort of people needing to be exempted. In the case of user charges, a catastrophic illness can easily make health care unaffordable to apparently high-income households deemed not eligible for user fee exemptions.

Nevertheless, co-existing formal and informal SHI ‘markets’ may introduce some difficulties with collecting premiums. If we consider the general population in our hypothetical African country, there will be two broad groups of individuals: those who will be paying premiums and those who will be exempted. Of the paying group, individuals will be distributed into households consisting solely of formal or informal workers, and households consisting of a mix of worker types. Collecting premiums from homogeneous households will be fairly straightforward but not so for heterogeneous households. The simplest solution is for members of the latter type of household to pay their premiums separately (i.e. from their HIAs or MIAs or formal payrolls) but as a household unit.

The second measure I consider appropriate is a social reinsurance facility for existing CBHI schemes. This will be provided by SHI pools serving the formal sector. We thus have an SHI system with a ‘wholesale’ sector where the insured units are entities involved in the business of pooling and bearing risks and a ‘retail’ sector where the insured unit is consumers. Reinsurance is a much needed platform for CBHI schemes to benefit from administrative assistance, actuarial support and the economies of scale in risk pooling. This last is of prime importance since CBHI schemes are usually small risk pools with a low capital base and facing voluntary insurance demands from relatively homogeneous localised consumers whose illness probabilities are more likely to be interdependent. For any given quantity of insured health care benefits, CBHI schemes are susceptible to long-run financial instability problems and a tendency to disintegrate (Dror & Preker, 2002). As far as I know, only the Nigerian NHIS reinsures CBHI schemes (Lem, 2009:26) but, following Bennett (2004), clarity is needed as to the roles of the CBHI schemes to be reinsured. Should they offer risk protection against cost-sharing arrangements, insure Q* – Q0 of care, or provide risk protection for (non-essential) health care of higher quality (quality that is not necessarily linked to clinical value)?

I suggest prohibiting complementary insurance for cost-sharing payments as this imposes a negative externality on health-system efficiency: the incremental welfare costs of increased consumption to QS from a point between Q* and QS will not be entirely borne by CBHI schemes (or their members). Given the objective of ensuring universal access to essential health care, this paper supports only CBHI schemes insuring consumption of Q* – Q0 of essential health care benefits. Yet, faced with a low capital base and the possibility that payouts will exceed premium revenues, it may be difficult to find CBHI schemes willing to insure Q* – Q0 of care. The mark-ups to baseline premiums that are necessary if CBHI schemes are not to suffer losses may inadvertently reduce insurance demands. As well as reducing the risk of going out business, social reinsurance for CBHI schemes will provide the financial capability to insure Q* – Q0 of essential health care that will otherwise not be supplied, as it requires, at least, a much larger pool of capital reserves.

That said, reinsurance may prevent CBHI schemes from making adequate efforts to control moral hazard. For instance, if reinsurance premiums are calibrated according to the expected health care costs of supplying inpatient and outpatient services to CBHI members, [(P2 – P0) (Q* – Q0)], the real social costs could be greater by as much as [0.5 (QS – Q*) (Co – P0)]. This can be reasonably checked if the reinsurer incorporates the costs of wasteful consumption in credible reinsurance premiums. Indeed with credible premiums charged for the publicly provided reinsurance facility, CBHI schemes could protect their net premium revenue by ceding only rare, catastrophic (inpatient) health care costs (that are of higher variance, near zero price demand elasticity and less susceptible to moral hazard) to the geographical (re)insurers serving the formal market – who in turn need not worry too much about moral hazard spillovers.

To sum up: I envisage a reliance on HIAs/MIAs linked to microfinance initiatives, and social reinsurance for existing CBHI schemes should allow expansion of SHI to poor households in the informal sector, especially those living above the poverty line. Considering that not all informal workers are high risks or low-income consumers, getting the informal sector to make some contributions is not just about preventing free-riding on formal workers. Achieving this may require: (1) education campaigns on the value of committing to save and/or demanding credit for the purchase of health insurance; (2) easing supply- and demand-side constraints to accessing credit and savings options (see Mpuga, 2010); and (3) helping CBHI schemes adopt meaningful investment portfolio strategies (for instance, in government treasury bonds) to capitalise efficiently on any net premium revenues.

2.5 Financial equalisation

Typically, the purpose of risk equalisation systems is to level the playing field for competing social or private health insurers. Since we are concerned here about non-competing SHI pools, there is no need to level the playing field to eliminate adverse risk selection. There are, however, good reasons why a financial equalisation system should be considered. Let us assume that for each geographical SHI pool the premium rate for the formal insurance market is λ and the revenue base is B, absolute flat-premium in the informal insurance market is ω with a revenue base M, and μ is the modest cost-sharing for a selective set of services (Qs) most susceptible to moral hazard. Total expected payouts for supplying Q* of services to each geographical risk pool will amount to Co times Q* and the ‘break-even’ condition for each short run period, t, – say each financial year – may be defined as:

(E indicates the expected value or estimate.) Based on audited accounts in time period t0, and the expected premium revenues and payouts in t₁ for Q* of services, SHI pools will receive a positive or negative tax subsidy. Equation (1) thus represents, on average, a pro-poor distribution of the tax-funded subsidy. An alternative arrangement is for the MoH to pay for Q0 of services using tax-funded subsidy, with the SHI providers paying for Q* – Q0 services. This, however, creates parallel health systems requiring a financial equalisation agent that collects and redistributes premium revenue.

It might be argued that erosion of reserves accumulated in good years (i.e. where premium revenue exceeds payouts) will undermine the financial stability of social health insurers. But rarely will one find social health insurers holding large reserves to hedge against unanticipated disruptions to their expected payouts. They will be bailed out in time of crisis by tax-funded revenues. Indeed, sustained use of taxpayers' money to reinsure the deficits of SHI pools (obras sociales) in Argentina is thought to have contributed to the country's financial collapse in 2001 (Lloyd-Sherlock, 2006). Besides achieving a pro-poor distribution of the tax subsidy, financial equalisation is a necessary tool to recover some or all of the financing for bail-outs that will be paid eventually by taxpayers in any African country. This is not to say that a lax environment should be allowed to flourish: the objective of the ‘strategic purchaser’ is to ensure fiscal discipline in both health financing and service provision.

Given that the critical parameter to be estimated is Co·Q*, this financial equalisation objective can be easily integrated into existing budgetary (resource) allocation formulae where available. Depending on institutional capacity and information available, the process of deriving E[Co·Q*] may involve risk adjustments according to reliable variables known to help predict health care demands and expenditures, for example demographic and socioeconomic data, and diagnostic measures of health status. Although the principles are the same, this risk adjustment process should not have the same complexity and costs associated with risk equalisation as it applies to stable non-competing SHI pools.

3. Social health insurance in Ghana

In this section, I draw parallels between my normative analysis and suggestions and actual decisions made by health planners under the NHIS in Ghana. The reasoning behind the introduction of the scheme follows the story told by Figure 2. Given shortfalls in general tax funds (generated from personal income tax, corporate tax, VAT on goods and services, petroleum tax, import tax, etc.), in the 1990s Ghana introduced a user charge policy referred to as a ‘cash and carry system’. Inadequacy of health care financing and recognition of the lost welfare gains from insurance protection led to implementation of the NHIS, which, according to Gyapong et al. (2007), was designed to cover user fees representing only a fifth of the total costs of providing essential health care (in quantities below Q*).

The Ghanaian NHIS defines a minimum essential health care benefits package that covers inpatient and outpatient services and it is made up of 170 district mutual health insurance schemes (DMHIS), which are geographical risk pools to match the administrative districts in the country. The NHIS is separated into formal and informal ‘markets’ where premiums for formal sector workers are defined as 2.5% of payroll income. For administrative simplicity, the payroll deductions are passed through the Social Security and National Insurance Trust Fund, which manages pension benefits. In principle, premiums for informal workers are graduated according to income but because of administrative difficulties in assessing incomes, a flat annual premium per adult of GH¢72 000 (GH¢7.20 after re-domination of the cedi) is charged.4 The premium obligation lies with individuals rather than households, with exemptions put in place for children under 18 and the elderly (Gyapong et al., 2007:44–5; McIntyre et al., 2008).

The NHIS exemptions for children as of the year 2006 applied when both parents had paid their premiums but, as I understand it, there is now a decoupling of premium exemptions for children and enrolment of parents. However, in the absence of a household mandate this might lead to adverse selection of high-risk household members into each DMHIS, as a mandate that is not 100% enforceable in practice means ‘voluntary’ enrolment. Indeed, recognising the threat of adverse selection, the DMHIS for Nkoranza District opted to link premium exemptions for children to enrolment of households as a unit. (Note that there is some variation in how districts implement the NHIS, and the Nkoranza District has the experience of running voluntary insurance schemes prior to formalisation of the NHIS.) Rajkotia & Frick's (2012) evaluation of the requirement to enrol as a family unit (in exchange for premium exemptions) showed that it had contained adverse selection but not eliminated it. It is not likely that the requirement will affect selective enrolment from households with young people above 18 years of age. A household mandate that ties enrolment of any individual (eligible for premium exemptions or not) to bringing along all other household members should help address the problem of adverse selection.

The NHIS does not explicitly state that an individual mandate for SHI contributions is the guiding principle but this is apparent from the status quo. I infer that this is the case since Gyapong et al. (2007) report that the next phase of reforms may involve mandatory contributions from employers equal to the 2.5% deductions from employees. Although the distortions an employer mandate creates in labour markets are well noted, health planners in Ghana must consider the inequities in disposable incomes between formal and informal workers as well as the costs of political compromise with employers who obstructively oppose SHI reform. It appears that in Tanzania private sector employers, faced with mandated insurance contributions, lobbied for the creation of a risk pool for their employees (the National Social Security Fund) as a separate entity from the National Health Insurance Fund, which is compulsory for public servants (McIntyre et al., 2008). To the extent that forming a separate risk pool transfers more control to private employers with regards to managing moral hazard and labour compensation costs, fragmentation is a rational response to employer mandates.

The Ghanaian NHIS is fragmented into parallel health systems. The tax subsidy for Q0 services is distributed, according to a needs-based resource allocation formula, to regional and district health facilities, while the DMHIS are expected to fund Q* – Q0 of what is largely the same package of essential services from premium revenues. Under the current parallel arrangements, the DHMIS employ an alternative provider payment system in the form of fee-for-service. The wasteful duplication in parallel health systems is evident from the observation that by March 2006, 33% of the GH¢114 billion given to the NHIS was spent on administrative and start-up costs – a figure above the expected maximum of 20% (Gyapong et al., 2007). Furthermore, it does not take much to see that fee-for-service payment with insurance coverage and no measures to control moral hazard puts the long-run financial health of the DMHIS at risk. The exclusion of even modest cost sharing from the NHIS may reflect the notion that high access costs P0 (related to travelling, waiting times and queuing) will serve as an implicit ‘market price’ to ration excess demand. However, to exert meaningful control over moral hazard, we need explicit price- and information-based measures.

According to Gyapong et al. (2007), some DMHIS are already running cash flow deficits, which in turn has led to delayed reimbursements and declining confidence in the NHIS. This premium-revenue deficit is unlikely to be the outcome of unbridled moral hazard and appears to be due to the slow enrolment of informal workers. Witter & Garshong (2009) report that the 2.5% earmarked VAT levy accounts for 70 to 75% of NHIS revenue, reflecting the fact that the large majority of enrolled NHIS members are those exempted from premium contributions: 34% of the total Ghanaian population, 64% of the total of registered NHIS members (as of the year 2008). So in reality the extent of premium-revenue deficit is worse than it appears. To correct this deficit, it could be tempting to increase premiums charged to those who have already enrolled. But for people with low incomes, those paying user charges below Co – P0, and those able to access free charitable care from religious and non-governmental organisations, increasing the premiums may lead to non-enrolment or non-renewal of NHIS membership, especially for informal workers. The alternative is to increase the VAT levy. Given that the VAT levy is earmarked for the NHIS, the excess tax burden this might impose may be negligible but this may not be the case if widespread premium-revenue deficits demand large increases in the VAT levy.

There are ways to get informal workers to pay, as discussed in Section 2.4, via HIAs or MIAs and removal of the tax subsidy on user charges to make enrolment in the DMHIS more attractive than self-insurance. It seems that this approach has been followed in Ghana with diagnosis-related group (DRG) pricing5 of health services that has doubled the user charges faced by uninsured cash-holding consumers. This is not a surprising outcome since the previous user charges were subsidised and not fully reflective of actual health care costs. Witter & Garshong (2009), however, describe this as a ‘pro-rich’ bias and question whether the introduction of SHI in Ghana is ‘something new or something old’, since the doubling of user charges increases the financial risks borne by cash-holding consumers. However, it is this same ‘pro-rich’ bias that makes health insurance more valuable – at tax-subsidised premiums. In fact, if the tax-subsidy on user charges is removed or reduced, it will be costly for cash-holding consumers to stay out of the NHIS, considering the insurance contributions they make via the VAT levy.

Another issue of concern is the discretionary loophole given to the DMHIS to use premium revenues in excess of payouts to provide a non-essential package of services (Gyapong et al., 2007:45). This directive is not sustainable, as it requires the DMHIS to consistently have good years where premium revenue exceeds payouts. Even worse, it undermines the purpose of a financial equalisation system, as noted in Section 2.5, which is to reduce the risk and costs of bail-outs to taxpayers.

A yet more worrying observation is that the NHIS premiums are not credible: they have no sound actuarial foundations. According to Agyepong & Adjei (2008), the current premium schedule was inferred from rates charged by existing CBHI schemes that insure only inpatient care, with a very limited set of people being eligible for exemptions. This is not necessarily a bad thing, since the decision to ‘borrow’ CBHI premiums was to avoid losing the political window of opportunity for reform. The focus should therefore be on estimating credible NHIS premiums. It may turn out that these credible estimates are no different from the CBHI premiums borrowed – in which case resources spent on proper actuarial calculations may seem a waste. But from this resource cost we have to subtract the welfare gains to risk-averse DMHIS who do not want to be exposed unnecessarily to premium-revenue deficits and consumers who do not want to see their insurance protection disrupted.

Instead of pursuing credible premium estimates, the NHIS debate has rather focused on proposals for a one-time life premium. The origins of this one-time premium are unclear but prior to its proposal the previous government, the National Patriotic Party, tinkered with the inferred NHIS premium – changing it from a premium for annual coverage to a one-off premium for five-year coverage and then back to an annual premium. According to Alfers (2009), this caused a lot of confusion. As one would expect, risk-averse consumers will demand limited intertemporal variations in premiums, with such fluctuations reflecting non-poolable risks such as forecasting errors and inflationary increases in health care costs. I suspect that the one-time life premium, since it comes with zero variation of any sort, was put forward by the government that won the 2008/09 elections, the National Democratic Congress, as a competing strategy to win or maximise public votes. Theoretically, a one-time life premium, reflecting the expected discounted present value of all future premiums (an extreme form of frontloading) offers both single- and multi-period insurance protection. However, if we are blessed with the foresight to correctly estimate a one-time life premium then we might as well design SHI benefits as health-state-specific indemnity payments.

The practical difficulties of indemnity insurance are well known: it is hard to specify ex ante all possible health states, how many people will suffer a particular illness, how much care is needed in each health state by each individual, and predict the future costs of health care, driven by technological advances. This applies equally to estimating a single lump-sum lifetime premium. It is thought that the one-time life premium will be GH¢150 (GH¢1 500 000, roughly US$100). But even if this is correct, non-existent or imperfect capital markets means the lump-sum premium will be unaffordable to many (Alfers, 2009). One way of rationalising the one-time life premium is to say it is part of a transitory move towards a fully tax-funded public insurance (Agyepong et al., 2011); in which case whether the one-time life premium is correct and affordable or not does not really matter. This begs the question of where the money will come from, but even if it is available, there will be competing non-health-care claims on public finances. Considering the commendable willingness of Ghanaians to pay premiums for Q* – Q0 of services, a move towards a wholly tax-funded system will be of little advantage – not to mention the excess tax burden that this might involve. Since roughly three quarters of NHIS revenue comes from the VAT levy, a move towards tax financing may appear to be the simpler option, but this ignores the fact that the status quo still does not allow Q* of care to be supplied. Indeed, if credible premiums reflecting consumption at unit prices of P2 – P0 are charged, Ghana could create a self-sustaining NHIS with contributions from the VAT levy targeted towards households (or their members) who are eligible for premium exemptions.

The last issue I want to comment on is the rather perverse provision for formal workers to opt out of the NHIS for PHI. Provisions to opt in and out will undermine the multi-period protection provided by SHI pools by allowing, for instance, high-income low-risk people to opt out only to return as low-income high risks. If the rationale is to allow people with high or adequate income to express their heterogeneous preferences for more than essential health care services, then this can be accommodated by allowing such people to purchase supplementary PHI. There is no need for people to opt in and out of the NHIS. I argue that in the next phase of reforms health planners in Ghana should consider: (1) consolidating the parallel health systems, (2) reversing the provision to opt out of the NHIS, (3) an individual mandate with households as the insured units, (4) HIAs/MIAs as an informal payroll, (5) developing credible NHIS premiums, (6) social reinsurance for existing CBHI schemes, and (7) a financial equalisation system.

4. Discussion

Using Figures 1 and 2, I have presented a number of policy options that health planners could consider when implementing SHI. If we are to put the options together (in an unconstrained environment that may not be found in real life), we will arrive at the health care financing arrangement depicted in Figure 3. Granted, there are concerns about SHI that I want to address. Wagstaff (2010), for instance, notes from his re-examination of SHI that tax financing does not leave the larger informal or agrarian segment of the population with no or inferior insurance benefits while the health system staggers slowly down the road to universal coverage. Tax funding avoids labour market distortions associated with payroll taxes of SHI and raises revenues in a more equitable fashion, and on top of this the incremental costs of collecting additional taxes may be lower than collecting premiums. Wagstaff (2010) further argues that if existing tax systems are regressive, SHI as an alternative vehicle for income redistribution may mask reasons for tax reform. Wagstaff & Moreno-Serra (2009) have also cautioned health planners who are contemplating shifting from general tax revenues to SHI that this might not lead to the expected health outcomes. Their study covering countries of central and eastern Europe and central Asia over a 15-year period failed to show positive impacts of higher health care expenditures on aggregate health outcomes.

There is evidence, on the other hand, that in African countries higher health care expenditures could lead to better population health outcomes (Anyanwu & Erhijakpor, 2009). We need to be careful here though. The link between health financing, expenditures and improved health outcomes is complex and subject to a multitude of health care and non-health-care influences and agents. There is no guarantee that this paper's suggestions will get African countries closer to Q* of services (bearing in mind the empirical difficulty of locating point O in Figure 2) nor will Q* of services necessarily lead to better health outcomes considering also the apparently unshakable grip of poverty on the African continent. Even if SHI increases government expenditures on health care per capita, what matters more is the incidence of spending – where, how and on what the revenues mobilised are spent. It is not surprising that Wagstaff & Moreno-Serra (2009) found an increase in the fraction of salaries of health professionals (i.e. price/cost of labour inputs) as a percentage of government health care expenditures. The iron law of health care financing dictates that part of the money spent on health care will end up as incomes for health workers. The crucial element therefore remains the incidence of spending.

In general, the validity of arguments against SHI depends on the structures and policies in place. Taking into account the excess tax burden and the non-trivial costs of collecting indirect taxes, a reliance on tax financing to achieve universal coverage is just as likely to be a slow process as SHI based solely on payroll taxes. In Sections 2.3 and 2.4 I highlighted ways of dealing with the informal sector – and with an individual mandate there should be near-zero incentive for people to move into the informal sector, thereby narrowing the formal tax base. With a savings-accounts-based SHI, there will be minimal free-riding and individuals choosing to self-insure will not have the option of tax-subsidised user fees. Nevertheless, if there is the slightest chance that SHI may mask the need for good governance and reworking existing tax systems in African countries, the concerns expressed hold some weight. For example, if in any given African country the skewness in income distribution is deemed unreasonable, then SHI (by providing time-consistent protection of incomes and consumption) may overshadow the reason for change. This, however, should not require a shift towards taxation but rather being aware of the masked effects.

I want to emphasise that the implementation of SHI does not give African governments leeway to abscond from increasing budgetary outlays to meet the 2001 Abuja Declaration target of spending at least 15% of tax-generated revenues on health care (African Union, 2006). Satisfying or exceeding the Abuja target means that in Figure 2 more than Q0 of services could be provided using only tax financing and this should translate into an increased likelihood of raising adequate complementary revenues for Q* of services via SHI. This will require introducing progressivity into existing taxes, broadening the tax base and shifting health care expenditures to essential health services, if not done already. That said, the link between the rate of economic growth and building adequate financing for health care reinforces arguments for increasing earmarked donor support to increase health care budgets. SHI will therefore help to shore up donor funding and public budgets for health care provision.

It might be argued that most African countries do not have the capacity to institute SHI arrangements effectively, notably collecting premiums, managing health care provider contracts and costs, and dealing with corruption and poor governance. I believe, however, that the adoption or adaptation of the array of options presented here will be most suited to the African context. For instance, we do not need to worry about implementing risk equalisation systems for competing social health insurers or the distortions in labour markets. The creation of HIAs/MIAs to implement SHI should not be an expensive venture and, without incurring the costs of a social reinsurance facility, CBHI schemes are unlikely to offer much for efforts towards universal coverage for essential health care. Even if we believe that fully tax-funded systems (or perhaps those complemented by voluntary PHI and not SHI) are the answer, African countries will still have to find ways of managing health care provider contracts and controlling costs due to untamed moral hazard.

One (perhaps disappointing) limitation of my analysis is that I have not considered the challenges of policy implementation. This is not to say that this paper has paid no attention to the politics and institutional complexities of reform change, capacity constraints, the existence or evolution of misreporting, cheating and corruption in administrative systems, and so on, or the need to engage with various stakeholders and interest groups. Agyepong & Nagai (2011), for example, document how ‘street level bureaucracy’ can lead to implementation gaps: the success or failure of reform change can be heavily influenced by front-line workers who are responsible for the implementation, daily management and execution of policies and programmes. As Agyepong & Nagai (2011:232) rightly point out, ‘A [technically sound] policy is only as good as its implementation arrangements.’ And if we do not consider how the practical (political) difficulties of policy implementation can offset the potential net gains, we can be drawn into a cyclical replacement of failed policies with newer and better ones, forgetting that the success of new policies is susceptible to the same capacity constraints, lack of leadership, poor governance and other contextual factors that contributed to the failure of old policies.

For brevity reasons, this paper cannot deal extensively with the practical or political feasibility of reform or provide a review or critique of how contextual factors have influenced past efforts at universal coverage in Africa. Consideration of contextual factors is not unique to any reform and the only time health planners need not worry about these issues is when they decide to do nothing about the status quo. No one can reasonably expect reform to happen overnight. It may take several years to implement adequate information, managerial and financial accounting systems, and effective measures for efficient production of health care. Regardless of the specific health care financing arrangement adopted (whether influenced by my suggestions or not), phased-out incremental health reforms, well-balanced policy ownership sharing, long-term commitment and continuity in reform efforts are vital elements for African health systems to achieve insured universal access to essential health care.

5. Conclusions and future research

African countries have many adjustments to make if they are to implement SHI (reforms) to generate adequate complementary revenues to tax financing. This task will require various commitments in the form of financial aid, technical help, capacity building, conflict resolution, democratic politics, and so on. Obviously, I have only presented broad macro-level considerations and the key determinant of reform policy success or failure will be the day-to-day micro-level execution and management of policies and regulations. But it was never my intention to provide all the answers. In this paper I have provided a general exposition of various ways of avoiding or dealing with the adverse consequences of SHI. However, it is critical that the options presented, for example using HIAs/MIAs and reducing the tax subsidy on user charges to increase SHI enrolment rates, and an individual mandate to minimise distortions in labour markets, are subject to an empirical test in the African context. Also, this paper has dealt only with health financing and not issues related to health care and health production, for instance, the provision and retention of adequate human resources for health, and performance-based financing of health care providers. These are topics for future research.

Acknowledgements

The views and opinions expressed here are those of the author, and not those of the author's affiliated institutions, past or present. All errors are of the author's own making, and no external research funding was used in support of this paper. The author has no conflicts of interest to declare.

Notes

¹User charges, or fees as some authors call them, refer to the cash you pay at the point of use for the health care you need. You only pay these charges when sick. Premiums are what you pay whether sick or not, such that when you fall sick health care is ‘free’ at the point of use. It is not really ‘free’ because you would have prepaid the costs involved.

²Some analysts may not agree that job lock is really a problem. People will change jobs regardless of the total compensation package if, for example, they do not like their employer. Even when present, job lock means employers benefit more from on-the-job investments they made in employees' skills.

³In any given year, you will always be certain about the size of the premium you pay, but when you fall sick and have to pay for health care at the point of use, you will never know for sure how much it cost. The premium need not, however, be fixed year after year.

⁴US$1.00 = GH¢1.94.

⁵DRG pricing or reimbursement is the practice of grouping related diseases diagnosed in routine clinical practice, for the purpose of identifying the type of health care needed and setting prices or provider payment on the basis of the average treatment cost for each group.