Who Pays for Health St people, particularly the poor, will reduce the amount of services demanded(even to the extent of not demanding a service at all) but will not necessarily be more rational in distin guishing when to demand services or which services they need to demand. Therefore, us ing user charges indiscriminately will indiscriminately reduce demand, hurting the poor in particular. Free-of-charge services do not translate automatically into unjustified over-utilization of services. Services that are free of direct charge are in reality not necessarily free or afford able, particularly for the poor, because of other costs associated with seeking health care, such as the cost of medication(when not available free of charge), under-the-table pay- ments, transportation, or time lost from work (18, 19) Given its potentially negative impact on necessary services, especially for the poor, co- payment should not be chosen as a source of financing except for low-cost relatively pre- dictable needs. Rather, it can be used as an instrument to control over-utilization of specific interventions(when such over-utilization is evident) or to implicitly exclude services from a benefit package when explicit exclusion is not possible Because of the desirability of sepa rating contributions from utilization, out-of-pocket payment should not be used unless no other alternative is available. All prepaid arrangements are preferable, except for low-cost interventions for which the administrative costs involved in prepayment arrangements mig not be worthwhile SPREADING RISK AND SUBSIDIZING THE POOR POOLING OF RESOURCES Pooling is the main way to spread risks among participants. Even when there is a high degree of separation between contributions and utilization, prepayment alone does not guarantee fair financing if it is on an individual basis only- that is, via medical savings accounts. Individuals would then have limited access to services after their savings were exhausted. It is claimed for medical savings accounts, which have been implemented in Singapore and in the United States, that they reduce moral hazard and give consumers the incentive to buy services more rationally, but while there is evidence of reduced expendi ture and of substantial savings among those who receive tax benefits and can afford to save (20), there is no evidence of more rational purchasing And individual financing fosters fee- for-service payment and makes it harder to regulate the quality of provision( 21). Peopl with a high risk of having to use services, such as the sick and the elderly, would be denied access because they could not save enough from their income. On the other hand, the healthy and the young, whose risk is usually low, might prepay for a long time without needing the services for which they had saved. In this case, mechanisms allowing for cross- subsidies from the young and healthy to the sick and old would benefit the former without damaging the latter. Thus, systems as well as people benefit from mechanisms that not only increase the degree of prepayment for health services, but also spread the financial risk among their members Ithough prepayment and pooling are a significant improvement over purely out-of- pocket financing, they do not take questions of income into account. As a result of large pools, society takes advantage of economies of scale, the law of large numbers, and subsidies from low-risk to high-risk individuals. Pooling by itself allows for equalization of contributions among members of the pool regardless of their financial risk associated with service utilization. But it also allows the low-risk poor to subsidize the high-risk rich Soci eties interested in equity are not indifferent to who is subsidized by whom. Therefore, healthWho Pays for Health Systems? 99 people, particularly the poor, will reduce the amount of services demanded (even to the extent of not demanding a service at all) but will not necessarily be more rational in distin￾guishing when to demand services or which services they need to demand. Therefore, us￾ing user charges indiscriminately will indiscriminately reduce demand, hurting the poor in particular. Free-of-charge services do not translate automatically into unjustified over-utilization of services. Services that are free of direct charge are in reality not necessarily free or afford￾able, particularly for the poor, because of other costs associated with seeking health care, such as the cost of medication (when not available free of charge), under-the-table pay￾ments, transportation, or time lost from work (18, 19). Given its potentially negative impact on necessary services, especially for the poor, co￾payment should not be chosen as a source of financing except for low-cost relatively pre￾dictable needs. Rather, it can be used as an instrument to control over-utilization of specific interventions (when such over-utilization is evident) or to implicitly exclude services from a benefit package when explicit exclusion is not possible. Because of the desirability of sepa￾rating contributions from utilization, out-of-pocket payment should not be used unless no other alternative is available. All prepaid arrangements are preferable, except for low-cost interventions for which the administrative costs involved in prepayment arrangements might not be worthwhile. SPREADING RISK AND SUBSIDIZING THE POOR: POOLING OF RESOURCES Pooling is the main way to spread risks among participants. Even when there is a high degree of separation between contributions and utilization, prepayment alone does not guarantee fair financing if it is on an individual basis only – that is, via medical savings accounts. Individuals would then have limited access to services after their savings were exhausted. It is claimed for medical savings accounts, which have been implemented in Singapore and in the United States, that they reduce moral hazard and give consumers the incentive to buy services more rationally, but while there is evidence of reduced expendi￾ture and of substantial savings among those who receive tax benefits and can afford to save (20), there is no evidence of more rational purchasing. And individual financing fosters fee￾for-service payment and makes it harder to regulate the quality of provision (21). People with a high risk of having to use services, such as the sick and the elderly, would be denied access because they could not save enough from their income. On the other hand, the healthy and the young, whose risk is usually low, might prepay for a long time without needing the services for which they had saved. In this case, mechanisms allowing for cross￾subsidies from the young and healthy to the sick and old would benefit the former without damaging the latter. Thus, systems as well as people benefit from mechanisms that not only increase the degree of prepayment for health services, but also spread the financial risk among their members. Although prepayment and pooling are a significant improvement over purely out-of￾pocket financing, they do not take questions of income into account. As a result of large pools, society takes advantage of economies of scale, the law of large numbers, and cross￾subsidies from low-risk to high-risk individuals. Pooling by itself allows for equalization of contributions among members of the pool regardless of their financial risk associated with service utilization. But it also allows the low-risk poor to subsidize the high-risk rich. Soci￾eties interested in equity are not indifferent to who is subsidized by whom. Therefore, health