Resource Allocation for Family Planning in Developing Countries: Report of a Meeting (1995)

ment Report (World Bank, 1993). Jamison noted that the $4-5 billion currently spent on family planning is not large compared with the $170 billion spent (for both public and private sectors) on health care of other types in poor countries; even the increases called for to meet unmet need would not raise the proportion significantly. But the claims on additional resources are many—achieving and maintaining universal coverage for childhood immunizations would require a doubling of expenditures that are now around $1 billion per year. The former director of the World Health Organization's Global Programme on AIDS has called for $2 billion more per year for AIDS programs. Family planning and other reproductive health services fared quite well in the World Development Report comparisons of cost-effectiveness (in terms of discounted costs per disability-adjusted life-years). Much public investment in family planning, certainly provision of information, can be justified as a low-cost measure for health promotion. But if, for many clients, family planning is merely a useful service that people value and would like to have provided free of charge, then it is one among many such services and there is nothing especially compelling about providing such services at public expense (Demeny, 1994).

Gabriel Ojeda referred to an intangible “understandable attitude” on the part of donors that had supported programs in some countries for long periods, which he likened to “the weariness parents feel when they see their grown-up children stranded indefinitely in their parental home doing little or nothing to earn their keep.” Throughout the meeting, different views were expressed on this point. For the sector as a whole, John Ross argued, more resources are needed and are likely to be forthcoming. The task is not “how to do more with less, but how to do more with more”; the decisions Ross listed include how to pay for family planning services as they now exist, how to scale up the number of clients, how to add reproductive health services, and how to promote a shift to more effective methods as appropriate. The majority opinion appeared to be that programs have to learn “how to do a lot more, with some more.”

One session of the meeting dealt with measuring the effects of program subsidies on the use of contraceptives. In the most general sense, subsidies can refer to any valued service (including provision of information) provided free of charge or at less than cost, with some portion of the costs borne by the public sector. A more restricted meaning is contraceptive supplies priced free or below cost. Sanderson and Tan (1992) showed that, in most family planning programs in Asia, more than half of total expenditures are for subsidized contraceptive supplies. In his remarks, Akin pointed out that one can think of both price subsidies and indirect subsidies (i.e., those that affect providers, either through training or by provision of higher-level administrative services) as affecting ei-

ther the money and time prices faced by consumers or the quality of the service offered for a given price.

Fertility decline is only one of the desired results of family planning programs, yet it looms large in public discussion. If family planning program inputs do not greatly affect the pace of fertility decline, then it is possible that subsidized services from the public sector are merely substituting for sources and methods of fertility regulation that couples would adopt in any event. In this view, some countries have high fertility rates because that is what couples want; modern contraceptives supplied by the programs may substitute for contraceptives couples would have obtained even at market prices, or more commonly for the more risky or unpleasant alternatives like abstinence, withdrawal, abortion, and infanticide. The macro-level evidence on these points, reviewed at the meeting by Paul Schultz, is equivocal. The debate matters in the present context because policy makers will need to form some estimate of the effectiveness of family planning expenditures in deciding how much, and for which groups, they may reduce subsidies. Reductions will not matter so much if program expenditures are not at the margin producing much impact on desired outputs.

Schultz put forward three basic questions that would need to be considered in making decisions on the allocation of resources to family planning: (1) What would be the marginal impact of additional resources? (2) Who would reap the benefits? and (3) How much would the beneficiaries value the prevention (or delay) of a birth? Despite a large evaluation literature, he claimed, we still cannot answer these questions very convincingly, because of some persistent methodological problems that include:

• Use of contraceptive prevalence rather than fertility as an outcome measure. Modern contraceptives can be a substitute for other means of fertility regulation, such as prolonged lactation, periodic abstinence, and withdrawal. Although contraceptive use and fertility are strongly associated in cross-national comparisons, program effects on contraceptive use may well be larger than effects on fertility. Family planning programs may have goals, such as improved women's and children's health and welfare, for which fertility is not the most appropriate outcome measure, but substitution effects need to be considered whatever the outcomes of concern.

• Public-sector family planning programs may be substitutes for private-sector (commercial or nonprofit) distribution of contraceptives. More generally, cross-effects of programs in other social sectors (such as education and health) as well as other providers of contraception need to be taken into account.

• Program inputs may be allocated in a manner that is not independent of other factors, some unmeasured in data sets, that are associated with contraceptive use. Attempts to evaluate the impact of program inputs have to take account of possible targeting of inputs, either to areas with particular need of services or to areas in which it is easiest to get new services established. The former type of

Schultz noted the limitations of analyses of data aggregated to the national level. Such data do allow some testing of general hypotheses about the effectiveness of family planning program inputs in reducing fertility across very different settings, which can then be studied in detail with household-level data, preferably longitudinal data. In a recent analysis using cross-country data for several periods, Schultz found that treating family planning program effort as endogenous in a multiequation model that included determinants of child mortality and separated effects of women's and men's earning power showed very little independent effect of the program inputs on fertility (Schultz, 1995). Simpler models relating family planning effort and a generalized index of development to fertility decline, in his view, cannot be used to produce unbiased estimates of program effects.

Analyses of nonexperimental data, even those using multivariate techniques, will produce biased estimates of the independent effects of program inputs unless models incorporate some method for correcting the problem of nonrandom allocation of program inputs. Analysts have used various modeling techniques to overcome this problem, such as fixed-effects models (Gertler and Molyneaux, 1994), but these in turn can be very sensitive to the assumptions of linearity and of unchanged effects of the unobserved variables, or to lags in effects of either observed or unobserved variables. In many applications of this method, it is difficult to ensure that enough program change has actually occurred between the times of measurement, and that there is enough variation across areas, to get sufficiently precise estimates of the effects of program inputs. As James Knowles observed, the econometric techniques tend to solve one problem at a time and are not robust when there are errors in measurement, as is typically the case. Knowles called for more work on Monte Carlo simulation of models of contraceptive choice, to determine how much the various problems in specification and measurement discussed here really matter for the inferences that policy makers need. The major implications of these methodological problems for analysts of non-experimental data, in Susan Cochrane 's view, are that the “compensating behaviors” need to be measured and program inputs, including actual prices facing clients, need to be measured carefully.

Cochrane also raised some of the problems with true experiments in family planning, including ethical difficulties and the fact that people react in ways that upset experimental or quasi-experimental designs, for example, by traveling to get preferred services.

Molyneaux and Gertler reported preliminary results of analyses of the simultaneous choice of contraceptive method and provider using recent data from the Demographic and Health Survey and income and expenditure surveys in Indonesia. Using data on both time and money prices of contraceptives, they developed a new method of expressing discounted costs per expected month of use. Their method incorporates information on average durations of use of different methods, as well as information on the local price and availability of methods from different potential suppliers, to allow for detailed estimates of the costs for different methods actually faced by potential users. As had some previous studies in Indonesia, this one found relatively small, though statistically significant, effects of price on contraceptive use. If public-sector price subsidies were removed altogether, Molyneaux and Gertler estimate that contraceptive use would decrease by about 3-4 percent overall. The effect would be largest among the poorest quartile of households, which would reduce contraceptive use by 10 percent.

John Akin presented results from studies of contraceptive choice in Jamaica, Thailand, and the Philippines (Akin and Schwartz, 1988; Schwartz et al., 1989; Akin et al., 1994). The likelihood of choosing a method did indeed respond to the money price in all the countries studied. But he too found that price sensitivity was relatively low for all methods except the condom (and in Jamaica, injectables). For pills, for example, it was estimated that quadrupling the price would decrease the proportion of contracepting couples choosing pills by 0.1 percentage point; in Jamaica, they estimated that doubling the price of the pills would decrease usage by about 9 percentage points. Akin underscored some of the points about data collection made by Molyneaux and Gertler; the data sets he used contained little or no information about the quality of services, and price information had to be inferred in some cases.

It is often asserted in the health sector that user fees are most appropriate for curative care, but not for preventive care, for which divergences between the social value of the service and the immediate perceived need of the individual client are greater. In family planning there are some interesting examples of facilities (like the Marie Stopes clinics) supporting themselves by charging for the services people will pay for (like abortions and menstrual regulation) a price high enough to cross-subsidize services, like contraceptive provision and counseling, for which people are reluctant to pay the full cost.

Given the variation of price elasticities for different contraceptive methods, it may not be wise simply to raise prices by an equal percentage for all methods as a way to recover program costs. If policy makers are concerned about the method mix, for example, or about promoting condom use for control of sexually transmitted diseases, then they would want to consider keeping prices low for particular methods, like condoms. Perhaps less obviously, for a given target revenue, the impact on overall contraceptive prevalence will be lower if the prices are raised more than average for methods with low elasticity of demand