I have been asked to comment on Peter Smith's contribution from the perspective of a developed country with an established, universal system of public health care insurance. In the case of my own country, Canada's public insurance covers fully medically necessary physician and inpatient hospital services free of charge, but imposes user charges in sectors such as pharmaceuticals and home care. User charges in Canada always have been and will continue to be fiercely debated. As Peter notes, the reference point for the debate – full coverage – differs from that in many of the developing country settings on which Peter focuses, where the baseline is either no or very limited coverage for many services. By grounding his analysis in the core objectives of a system of health care finance, however, Peter's analysis usefully establishes a common ground for analysis relevant to all settings.

Peter identifies the three basic objectives as: (a) reduction of individuals’ financial risk; (b) redistribution of income in society; and (c) improving health through increased access to and use of health care services. Compared with full coverage, the impact of user charges on (a) and (b) is not contested: user charges unavoidably increase the financial risk that individuals bear and mitigate the redistribution of income from the healthy and wealthy to the sick and the poor. Governments may judge that the promised benefits of user charges out-weigh these negative effects. But such a judgement can be made only after detailed empirical analyses to assess the extent to which user charges increase financial risk for members of society and redistribute income towards the healthy and wealthy, analysis that is surprisingly often lacking. As Bob Evans has emphasized, the nature and persistence of policy debate about user charges in public systems of finance can be explained in part by the desire in some quarters of society to reverse the redistributive effects of public finance (Evans, Reference Evans1997).

I want to focus on Peter's (correct) observation that user charges are, at root, about rationing. Analysis of the core rationing problem faced in health systems requires greater consideration of market failure in health care service markets. Given the third objective to improve health through access and use of health care, optimal rationing would have only those in need receive care, and those in need receive only effective health care given their health condition. Against this standard, it is easy to see why traditional user charge policies fail – they ration solely on the basis of willingness-to-pay.

The dominant historical argument for user charges derives from Pauly's influential analysis of moral hazard (Pauly, Reference Pauly1968) in which he argued that full insurance leads to the inefficient over consumption of health care and that user charges were the optimal policy response to correct this moral hazard. Of course, in Pauly's world of a perfect health care service market, no income effect, no equity concerns and welfarist analysis, rationing by willingness-to-pay through user charges is the first-best instrument to combat moral hazard, and the optimal user charge depends on the price-elasticity of demand. The problem, of course, is that real world health care service markets are characterized by multiple sources of market failure (e.g. informational asymmetry, provider market power, externalities), a positive income elasticity of demand and equity concerns (Hurley, Reference Hurley2000; Nyman, Reference Nyman2003); further, as exemplified by Peter's third objective, many analysts adopt an extra-welfarist emphasis on health as the outcome of concern. These considerations both reduce the magnitude of moral hazard and lead to the conclusion that traditional user charges that ration on the basis of willingness-to-pay alone are the wrong policy prescription. They are wrong because they cause people, especially the poor, to reduce both necessary and unnecessary services, with the consequent negative health effects. From this alternative perspective, user charges are likely to decrease rather than increase efficiency.

A second argument – the focus of Peter's analysis and one that has become more prominent in recent years in developed countries with universal systems of public finance – is a purely fiscal argument about rationing government expenditure. Health care, the argument goes, is eating up too much government expenditure, displacing expenditure on other valued services such as education, transportation, social services and the environment. User charges are needed to control public health care expenditures. This purely fiscal motive calls for raising revenue while imposing the least efficiency and equity cost on society. This argument does not preclude a concomitant concern about moral hazard, but the two motivations can lead to differing approaches to the design of user charge policies.

As a practical matter, market failure and equity concerns in health care markets substantially undermine this fiscal argument for user charges. Reduced consumption of effective preventive and therapeutic care can cause people to develop wholly preventable conditions or allow mild conditions to become more serious and expensive to treat than they would be if they were caught early. The ultimate costs to the public health care system across the full range of health services can actually increase as a result of user charges. Limiting the magnitude of user charges in an attempt to reduce these adverse effects unfortunately also limits their revenue potential. Further, equity concerns dictate that low-income individuals either be exempt from charges or be liable for a maximum amount of annual out-of-pocket expenditure. Such provisions require costly administrative mechanisms to assess eligibility and to track expenditures against any specified expenditure limits, and the associated administrative costs can substantially reduce the net revenue gain to the public insurer. In the end, when all effects are properly counted, real financial savings to the system often end up far less than promised (or than suggested by counting just the user charge revenue).

Value-based user charges have emerged largely as a response to the failures traditional user charge policies. By linking the size of the user charge (including the possibility of zero charge) to the cost-effectiveness of a service, the charge scheme better targets reductions in utilization on low-value care. To the extent that such a scheme can selectively reduce low-value utilization, it can enhance the efficiency of utilization and, by avoiding the problem of higher long-run costs caused by reductions in the use of effective services, also generate fiscal savings.

As Peter notes, value-based user charges are in a sense consistent with coverage principles that underlie most publicly financed systems, and, one could argue, merely represent a refinement and extension of those principles. Canadian Medicare, for example, calls for covering all medically necessary physician and hospital services. Evidence of effectiveness is widely agreed upon to be a necessary condition for a service to be medically necessary (Hurley et al., Reference Hurley, Stoddart, Birch and Torrance1997). But for a few exceptions, thus far this principle has been implemented through only binary, 0–1 coverage decisions: a service is either fully covered or it is not covered. Value-based coverage extends this to allow multiple potential levels of coverage depending on evidence of cost-effectiveness.

While value-based coverage policies offer considerable advantages over traditional user charge policies, in the end they can at best only partially address the core rationing challenge faced by public insurers. As Peter observes, we lack essential information on the effectiveness and cost-effectiveness of a large majority of health care services provided in modern health care systems. As a practical matter, this severely limits the scope for such policies. But more fundamentally, even if we had such information, I would argue that value-based coverage policies apply to the wrong unit of analysis – a health care service or procedure – to address the rationing problem effectively.

The rationing challenge, to reiterate, is to ensure that citizens have access to all cost-effective services given their medical needs, and that their use be limited to only cost-effective services. The use of services that are wholly ineffective is a minor source of inefficiency and expenditure in modern health care systems. After all, the set of services that are wholly ineffective is very small (which is why de-listing exercises aimed at eliminating their coverage find so little savings). The largest source of inefficiency is the provision of otherwise effective services in contexts where they are not effective – inappropriate utilization. The fundamental rationing challenge then, is to restrict use of services to contexts in which they are appropriate. For most services the information required to assess appropriateness is revealed only in the clinical encounter (Hurley et al., Reference Hurley, Stoddart, Birch and Torrance1997), and varies from patient to patient (or even encounter to encounter for a single patient). Setting the level of a user charge based on evidence of the average effectiveness or cost-effectiveness of a service is too crude; the charge needs to be specific to the clinical situation. In principle it is possible to design case-specific conditional coverage policies, and it is even practical for selected services, but it is beyond our current abilities to do so on a wide scale across a broad array of services (the revolt against such conditional coverage policies during the managed care era in the United States certainly provides a cautionary message).

Demand-side financing policies are the wrong instrument to solve this problem. Effectively addressing this rationing challenge requires policies to affect the clinical encounter and the decisions of providers working in concert with patients. Essential to this effort are better funding models, better delivery models, better alignment of funding and delivery, better quality assurance programmes, and greater accountability. Key themes within emerging models internationally include blended funding arrangements, elements of pay-for-performance, multi-disciplinary care, vertical and horizontal integration, and multi-pronged strategies of financial and non-financial mechanisms working in concert.

How does all of this bear on Peter's call for a more nuanced debate on user charges, especially in settings working to achieve universal coverage? The single biggest lesson from developing countries with universal systems of public insurance is that, contrary to functions ascribed to it in formal models of insurance and much literature in public finance, a public insurer can't just be a financial intermediary – a revenue collector and a ‘bill-payer’. To view the public insurer as purely a financial intermediary is to restrict its policy options to choosing the optimal subsidy rate. Instead, a public insurer has to be an active manager of the health care system, a manager that deploys the full range of policy levers at its disposal to shape the behaviours of providers, patients and other organizations to achieve its system goals.

Will this, even if done well, resolve the fiscal challenges Peter outlines? No. But then, another key lesson from developed systems is that neither will user charges. The fiscal tension is an unavoidable, inherent element of a system of public finance. The question is not whether people are willing to pay in aggregate for the growing costs in modern health care systems – they are. The question, as Peter rightly observes, is whether they are willing to do so through a system of public insurance that ensures access for all and that redistributes income from the wealthy and healthy to the poor and sick. In the end, this is at root a political question rather than an economic one.