HOUSEHOLD RESPONSES TO PUBLIC HEALTH SERVICES: COST …documents.worldbank.org/curated/en/... · shift from self-care to the modern health sector. If the objective of policy im-provements

HOUSEHOLD RESPONSES TOPUBLIC HEALTH SERVICES:COST AND QUALITY TRADEOFFS

Harold AldermanVictor Lavy

The effectiveness of government investments in health care depends on thepublic's response to price and quality as well as on whether these expendi-tures actually improve health outcomes. Consumers, even those in low-income households, are willing to pay fees for better health care if the feestranslate into improved access and reliability. But when prices rise without aconcomitant improvement in services, malnutrition and child mortality ratesincrease. The availability of basic health care has a relatively greater impacton households with low incomes or low education, or both, than does theprovision of more specialized services. This article describes the types ofservices for which households indicate they are willing to pay increased fees.It also indicates the potential gains from improving these services, as well asthe consequences of moving faster on cost recovery than on providing im-proved or better-targeted services.

Whether households benefit from government expenditures on healthcare depends on the quality of the services delivered and how house-holds respond to that quality. Similarly, the long-run sustainability

of public investment in health services depends on the ability to finance futureimprovements in quality by increasing revenues through higher user fees. Toassess the efficacy of such investments from a biomedical perspective, one needsto understand how the quality of service delivery improves its effectiveness.From an economic perspective, it is important to know how improvements inquality affect consumer behavior; health outcomes depend not only on whatservices are provided but also on what services are used.

Households are not passive recipients of government programs. The benefitsthey derive depend in part on what other resources and information they com-bine (Mosley and Chen 1984). A household is constrained both by its budget(which includes a time budget, because time spent obtaining health care can be

The World Bank Research Observer, vol. 11, no. 1 (February 1996), pp. 3-22© 1996 The International Bank for Reconstruction and Development / THE WORLD BANK 3

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

wb350881

Typewritten Text

76551

wb350881

Typewritten Text

traded off against competing uses of time, both for labor and leisure), and byavailable medical technology. Changes in quality (making clinics more acces-sible, say) lower the cost of acquiring health care, and, as new processes aredevised that permit increases in efficiency and productivity previously precludedby a lack of information, the technology constraints are relaxed. Thus house-hold welfare can be increased by lowering the costs of achieving a given level ofhealth. Governments do not invest in health services, however, to free up house-hold resources for other purchases but to increase human capital by improvingaccess to health care. Governments can do this by expanding publicly subsidizedhealth services and by substituting private-sector services where these are cost-effective.

In this article we review recent research that asks whether consumers willtake advantage of higher-quality health services if out-of-pocket payments areincreased. We also investigate what kinds of improvements matter most to con-sumers, particularly low-income consumers, and whether complete cost recov-ery for these improvements is possible. Our interest in this question is not merelyto guide cost recovery for a system of delivery but to find out, first, how policychanges affect that system's contribution to public health and, second, whichservices affect two indicators of community health—anthropometric measuresof the nutritional status of children and child mortality rates.1 Despite the ev-idence that consumers are willing to pay more for better health care, price in-creases have measurable consequences for indicators of health, in part becausequality improvements do not always match price movements, and in part be-cause the poor cannot afford to take advantage of quality improvements with-out additional support.

The Effect of Quality on Demand for Health Care

Because those who use public medical facilities in developing countries areoften charged nothing, measures of the availability of service, such as the dis-tance or travel time to the nearest health facility, are often used as proxies forthe demand for particular health services. Most studies of developing countries,however, consistent with the results in industrial economies (Manning and oth-ers 1987), suggest that where people pay for health services, the price of caredoes constrain demand (Alderman and Gertler 1989; Mwabu 1989; Gertler andvan der Gaag 1990). Occasionally, user fees do not decrease demand (Akin andothers 1988; Schwartz, Akin, and Popkin 1988) and in some cases even increaseit (de Ferranti 1985). One reason for the diverse results, of course, is the diver-sity of settings. Moreover, several studies do not allow for different responsesamong income groups. Where the data permit this calculation, demand fromlow-income households tends to drop more when prices are increased than doesdemand among wealthier groups (Gertler and van der Gaag 1990). (Most ofthese estimates do not take into account the quality of the service, however.)

4 The World Bank Research Observer, vol. 11, no. 1 (February 1996)

Similarly, studies that measure travel time often fail to indicate what kinds ofservices are available at the destination (Strauss 1990). Distant, but morecomprehensive, health centers may offer better care (facilities, staff, supplies)that compensates for the increased travel time. Moreover, the studies often basethe cost of services on expenditures per medical visit as reponed by patients. Butthe sum paid by a consumer per visit depends not only on the price charged bythat provider for a standard treatment, but also on the type of treatment and thequality of service chosen by the consumer. Estimated effects of service availabil-ity or prices can be biased unless the study takes into account differences in thequality of the service. For example, if prices increase but quality does not, theuse of the service or facility can be expected to decline—a direct response to theincrease in price. But if higher prices are associated with better quality and pa-tients are willing to pay more for that improved quality, it is likely that theestimated price response will be understated. The studies reviewed here avoidthat bias by explicitly addressing the role of quality in health care choice. All thestudies are based on household surveys complemented by full surveys on thecharacteristics of the health facility. Many of the studies rely on data from Af-rica (Ghana, Kenya, and Nigeria). We also present related evidence fromCameroon based on an experiment that observed the effect of changes in priceswhen none of the other factors varied significantly.

Measuring Quality

Many definitions of quality for health care services have been suggested inthe literature. Examples include the degree to which the health facility's actualperformance or achievements correspond to established standards (Wouters,Adeyi, and Morrow 1993) or its capacity to perform certain specific health in-terventions, combined with some indicator of how well these interventions arecarried out (Roemer and Montoya-Aguilar 1988). These criteria, however, donot provide a tractable approach to measuring the multiple dimensions of qual-ity in the supply of and demand for health care services. An alternative measureis based on patients' perceptions of quality. In this approach, objectively mea-surable characteristics of health care facilities are linked to households' subjec-tive assessment of the probable outcome. Thus attributes that are easy to mea-sure serve as proxies for those that are unobserved. These observable servicecomponents, which include physical facilities, number of staff members andlevel of supervision, availability of essential drugs and equipment, and provisionof basic health services, are highly correlated with quality indexes (Garner,Thompson, and Donaldson 1990), thereby allaying concerns that the resultsmay be sensitive to the choice of approaches.

Another concern is that measures of quality (or even availability) are liable tobe inaccurate. For example, Thomas, Lavy, and Strauss (forthcoming) comparetwo measures of staffing in Cote d'lvoire: the number of staff members listed inofficial records, and the number who were actually present in the twenty-four

Harold Alderman and Victor Lavy 5

hours preceding the interview. The actual number of doctors present had a fa-vorable and significant effect on child health; the number on the books wasirrelevant. The failure to take inaccuracies in official records into account maylead to incorrect inferences about the efficacy of public policies. Table 1 showsthe various characteristics of quality included in the three studies that are thebasis of the policy simulations below.

How Changes in Quality Affect Consumers

Lavy and Germain (1994) measured the quality of health care in Ghana interms of infrastructure (electricity and running water); personnel (number ofdoctors and nurses); basic adult and child health services, including the avail-ability of a laboratory and the ability to vaccinate children and to provide pre-natal, postnatal, and child-monitoring clinical services; and the availability ofessential drugs (ampicillin, chloroquine, paracetemol) and an operating room.2

Table 1. Quality of Care as Measured in Three Studies on Health Care(average values)

Variable Private Public

GhanaProportion of facilities stocked with drugsNumber of nurses and doctorsProportion of facilities with electricity and waterProportion of facilities with adult and child servicesProportion of facilities with an operating roomDistance to facility (km)Price of consultation (100 cedis)Proportion of consumers treated

NigeriaProportion of facilities stocked with drugsPhysical condition of facility (l=good, 3=poor)Expenditure on care per person in populationPrice of consultation (naira)Proportion of consumers treated

KenyaTypes of drugs availableNumber of medical staffDays a year without antibiotics (log)Days a year without malaria drugs (log)Days a year without aspirin (log)Distance to facility (km)Price of consultation (Kenya shillings)Proportion of consumers treated

Source: Ghana—Lavy and Germain (1994); Nigeria—Akin, Guilkey, and Demon (1995); Kenya-Mwabu, Ainsworth, and Nyamete (1993).

The World Bank Research Observer, vol. 11, no. 1 (February 1996)

0.880.920.590.370.363.421.060.35

1.001.161.362.220.13

10.401.400.670.670.145.98

14.200.25

0.662.000.500.760.355.640.430.14

0.721.293.041.190.35

8.0014.403.663.830.458.00

. 0.000.36

The first four quality factors had large and significant effects on demand. Toillustrate, if the percentage of public facilities in Ghana with drugs availableincreased from 66 percent to 100 percent, the use of such facilities would in-crease by 44 percent (table 2). Similar changes would occur with other indepen-dent changes in infrastructure, services, and personnel. Simultaneous improve-ments in drugs, infrastructure, and services would increase the likelihood thatpatients would use public health facilities by 127 percent. Part of this changereflects shifts from the private sector to the public sector, and part represents ashift from self-care to the modern health sector. If the objective of policy im-provements is to usher more people into the modern health care system, thenquality changes alone could be expected to reduce the probability of self-treatment by about 14 percent.

Use of public health clinics in Ogun State, Nigeria, was affected by severalquality variables, including operational costs per capita, the physical conditionof the facility, the availability of drugs, and the number of functioning x-raymachines and laboratories. The number of support personnel, nurses, and doc-tors per capita did not have a significant effect. When both public and private

Table 2.(percent)

Simulation

Simulated Response to Improved Public Health Services

Increase in patient use of:

Self-care

Publicfacilities

Privatefacilities

Ghana\jnana

Improve quality of public facilitiesDrug availability increasedInfrastructure improvementsService improvementsPersonnel increased to three peopleDrugs, infrastructure, services brought up to

javpra (re IPVPI

1.2-0.9-0.6-0.5

43.733.024.918.9

-7.0-5.3-4.0-3.1

average level -3.5 127.6 -19.5

Improve all quality factors simultaneously(public and private)

Drugs, infrastructure, services brought up toaverage level -14.0 61.0 60.0

NigeriaDrug availability increased (public and private) -8.0 25.0 -45.0Expenditure per person doubled (public and private) -4.0 7.0 0.0

KenyaIncrease two types of drugs at public facility -4.1 3.6 -4.1

Note: Simulations are projected over the entire population, using point estimates at sample mean.Source: Ghana—Lavy and Germain (1994); Nigeria—Akin, Guilkey, and Dehton (1995); Kenya-

Mwabu, Ainsworth, and Nyamete (1993).

Harold Alderman and Victor Lavy

operational expenditures per person were doubled, some patients moved fromself-care to public care, with no change in private-care usage (see table 2).

Improving the availability of drugs elicited a substantial response. Full avail-ability of drugs in both public and private sectors induced many patients tomove from self-care and the private sector to public care. Conversely, if drugavailability were to fall to 50 percent for both types of facilities (a very low ratecompared with existing availability—72.9 percent in the public sector and 99.6percent in the private sector), both private and public clients would likely forgocare, with the private sector, where both availability and use were previouslyhighest, recording a larger drop. If the quality of the public sector were im-proved to nearly match that of the private sector and all prices were doubled,the number of consumers who rely on self-care and private care would drop, butthose using public care would increase.

A study of a rural district in Kenya (Mwabu, Ainsworth, and Nyamete 1993)looked at treatment-related measures of quality, including the availability of avariety of drugs and diagnostic equipment (although the size of the sample lim-ited the number of variables used). The study found that demand was lower atfacilities lacking aspirin. Curiously, demand was higher at facilities lackingantimalarial drugs. The likely explanation is that high demand depleted the stock,and not that stock shortages encouraged demand. The probability of a visit to apublic facility was positively related and most sensitive to the availability of abroad number of drugs. For example, making two additional drugs availablein government health facilities led to a 3.6 percent increase in the use ofpublic facilities and a 4.1 percent reduction in the use of private providers (seetable 2).3

These results are similar to, but larger in magnitude than, Hotchkiss's (1993)findings on the choice of obstetric care in Cebu, Philippines. His study includedthe following measures of quality: the availability of medical supplies (numberof drugs available to treat diarrhea), practitioner training (doctor or midwife),service availability, facility size, and waiting time. Again, the quality of servicesprovided had a significant effect on users' choices. For example, drug availabil-ity, waiting time, and the availability of doctors to perform deliveries were signifi-cant determinants of choice, although drug availability affected the choice ofwell-educated women, but not other consumers.

Many of these changes in health care choice involve users who shift betweenproviders, rather than between seeking professional help or doing without care.This is also the case with respect to changes in distance and price. For example,if the average distance to the nearest facility in Ghana were reduced by 50 per-cent, demand at those clinics would nearly double; self-care would decline byonly 2.6 percent (table 3). Thus much of the change would reflect a shift fromprivate care. (This pattern was also observed in Pakistan by Alderman and Gertler1989.)

Simultaneous price increases in public and private facilities are also possible.An experiment in Indonesia found that a change of fees in public facilities in-


Table 3. Simulated Response to Changes in(percent)

Study

GhanaPublic facilities

Reduce distance 50 percentIncrease user fees 50 percentIncrease user fees 100 percent

Private facilitiesReduce distance 50 percentIncrease user fees 50 percentIncrease user fees 100 percent

Public and private facilitiesReduce distance 50 percentIncrease user fees 50 percentIncrease user fees 100 percent

NigeriaDouble public and private prices

KenyaReduce distance 20 percent (public)Increase public fees from zero to

Kenya shillings 10

Price and Access to Health Care

Increase in patient use of:

Self-care

-2.60.10.3

-3.01.02.0

-5.01.02.0

7.0

-1.8

20.9

Public facilities

95.9-5.8

-11.3

-15.07.0

15.0

72.02.03.0

0.0

1.6

-18.2

Source: Ghana—Lavy and Germain (1994); Nigeria—Akin, Guilkey, and DentonMwabu, Ainsworth, and Nyamete (1993).

Private facilities

-14.91.01.9

20.0-9.0

-17.0

5.0-8.0

-15.0

-23.0

-1.8

20.9

(1995); Kenya—

duced a similar change in private facilities (Gertler and others 1995). Table 3indicates that if private fees in Ghana were to go up simultaneously by the samepercentage as public fees, substitution would lead to increased use of publicclinics as well as increased self-care.

Theory anticipates that, in deciding whether to seek care and which providerto consult, households base their choice on many factors, including the avail-ability of drugs, doctors, hours, and clinical services; the adequacy of equip-ment; and the physical condition of health care facilities. Empirical studies alsohelp predict the costs that can be recovered, but several factors can bias theresponse. First, if the study relies on respondents who have actually chosen theservice in question, self-selection becomes a problem. Second, use may appearto decrease with improved quality if the services provided are effective in treat-ing illness and thus in reducing overall population morbidity. Finally, some mea-sures of quality (especially stocks of supplies) actually capture both supply anddemand behavior. If demand is high and stocks are quickly depleted, it mayappear that people are using a low-quality facility when, in fact, high demandhas depleted available supplies.


These three issues could cloud the interpretation of results, but the studiescited are not particularly sensitive to the first concern because the quality dataare collected from a sample of facilities rather than from users. The secondconcern is likely to lead to an underestimation of the demand for quality, pro-viding a bound for estimates. A similar argument is plausible regarding the fail-ure to include supply response. Both these a priori arguments, however, as wellas econometric techniques used to control for selection biases, often have lim-ited persuasive power. Supportive evidence can be provided if results derivedfrom alternative methodologies converge on the econometric results from cross-section surveys. Thus we compare the results reported above with those fromthe field experiment described below.

In an experiment in the Adamaoua province of Cameroon, researchers mea-sured consumers' willingness to pay higher fees for improved health services. Inthis study, revolving funds that use revenues from sales to replenish drug sup-plies and subsidize other services were established at each health center. Litvackand Bodart (1993) report that the probability of using health centers increasedsignificantly in the treatment areas compared with those in control areas. Whendrugs became available at the local health center, the increase in the value of theservice far outweighed the fee charged compared with the time, transportation,and treatment costs formerly borne by the patient, and thus the number of peopletreated rose.

Willingness to Pay for Quality Improvements

Can consumer response to quality be used to guide pricing policy? How muchare individuals willing to pay for better care in existing public facilities or toreduce the travel time and costs to obtain treatment? Table 4 indicates the feesconsumers in Ghana said they were willing to pay for various quality factors.The most important single factor is distance: households were willing to pay 2.6percent (155 cedis in 1989) of their monthly income to reduce the distance (ortravel time) to the nearest clinic by half. Second in importance is drug availabil-ity: people were willing to pay 53 cedis to ensure that the clinic is reliably stockedwith basic drugs. The importance of these two factors is also evident from asimulation (not shown in the table) in which consumers are equally far frompublic and private facilities, and both facilities have the same drugs available. Inthis case the probability that patients would consult the public facility increasedby 300 percent (because fees are lower), and the amount that people were will-ing to pay in such a scenario amounted to 5 percent of monthly income.

The table shows that households will pay smaller amounts to ensure thatchild care and immunization and laboratory services are available and to doublethe number of doctors and nurses. More important, they are willing to pay morefor combined or simultaneous quality improvements than the sum of the dis-crete improvements. For example, a household at mean income is willing to pay353 cedis (6 percent of monthly income) for simultaneous improvements in in-

1 0 The World Bank Research Observer, vol. 11, no. 1 (February 1996)

Table 4. Willingness to Pay for ImprovedIncome Category(prices in cedis: 175 cedis=$l)

Improvement

DistanceDistance reduced by 50 percent

PricePrice reduced by half

Price = zero

QualityDrug availability at maximum level

Services at maximum level

Infrastructure at maximum level

Medical personnel set to three individuals

Drugs services and infrastructure set tomaximum

Health Services

One standarddeviation below

mean income(40,000 cedis)

106(3.19)

4(0.13)9(10.28)

13(0.39)10(0.31)3(0.09)3(0.08)91(2.72)

Note: Percentage of monthly income in parentheses.Source: Lavy and Germain (1994).

in Ghana,

At meanannualincome(70,242cedis)

1S5(2.65)

9(0.15)18(0.31)

53(0.91)42(0.71)20(0.18)11(0.18)353(6.03)

by Household

One standarddeviation

above meanincome

(100,000cedis)

178(2.14)

12(0.14)23(0.27)

117(1.04)68(0.81)33(0.40)17(0.20)580(6.95)

frastructure and in drug and service availability—substantially more than thesum of each intervention (115 cedis).

Much of the policy debate about recovering the costs of public health servicesrevolves around the concern that rising prices may reduce the overall utilization.Can part of the negative effect of prices on services be offset by improving qual-ity? What if providers simultaneously improved the quality of care and raisedfees? In the population referenced in Ghana, the demand for quality is so highthat if the availability of drugs and services and the physical condition of publicfacilities were improved by 100 percent, the percentage of individuals choosingtreatment in a public clinic would not decline unless prices were raised by morethan 1,200 percent.

Similarly, the government could build fewer facilities, which would increasethe average distance to the nearest public facility, but could offset this effect byimproving quality. Consumer choice in Ghana is unaffected when the meandistance to the nearest public health provider is doubled if services are simulta-neously improved 100 percent. In Cebu, Philippines, prices could double and

Harold Alderman and Victor Lavy 11

still the probability of choosing a public hospital would actually increase by 1.8percentage points for poor households and 4.4 percentage points for nonpoorhouseholds if crowding were simultaneously reduced to zero and the availabilityof drugs were doubled (Hotchkiss 1993).

Another way to assess willingness to pay is based on a survey of what con-sumers say they are willing to pay for certain programs. Researchers usuallyconcede that the technique is experimental and interpret their findings with cau-tion (see Griffin and others 1995). This approach was used, for example, in astudy of rural households in the Central African Republic, which found thatconsumers were willing to pay for improved services and increased access todrugs (Weaver and others 1993). Not surprisingly, higher-income groups tendedto be willing to pay more for quality improvements, but a willingness to pay wasnot confined to these groups. Similarly, in Tanzania, Abel-Smith and Rawal(1992) found that among low-income respondents, 45 percent were willing topay 200 Tanzanian shillings (in 1989,192 Tsh = US$1) for improved services ingovernment hospitals or health centers, close to the 60 percent that the highest-income group was estimated to be prepared to pay. Moreover, a higher propor-tion of the wealthier households than low-income households were unwilling topay any additional amount for quality improvements. In a comparison of thetwo types of quality improvements, more people said that they were unwillingto pay to reduce the waiting time to less than an hour than were prepared to payfor assured drug supplies.

These studies draw a very similar picture to the simulations mentionedearlier; user charges will not reduce utilization if the revenue generated isused to improve the quality of services. Conversely, the negative price re-sponse observed in most studies implies that, if the fees go into general rev-enues without generating higher-quality services, households will shift pro-viders. Thus policymakers should focus on those quality improvements thatwill ultimately reduce the financial burden on the poor, such as reduced travelcosts and assured drug supplies.

Who Benefits from User Fees?

Do quality improvements financed by increased user fees primarily ben-efit wealthier households? Not necessarily. In some cases the poor have beenshown to take greater advantage of simultaneous increases in health carequality (drug availability) and fees than have the wealthy (as in the experi-ment cited above by Litvack and Bodart 1993). Often, however, willingnessto pay for quality rises with income, so user fees may dissuade poor house-holds, yet not substantially decrease overall demand. Table 4 shows thatsuch a situation is possible; the price a household in Ghana will pay for agiven quality of health care increases with income. There are combinationsof prices and quality that will be chosen by the average consumer, yet deemedunaffordable by the poor. This choice is more pronounced with quality vari-


ables (drug availability, infrastructure, services, and personnel) than withdistance or price.

Furthermore, as a percentage of income, willingness to pay for accessibilitydecreases as income rises, while in the case of quality factors, it increases intandem with income. For example, low-income households in Ghana are willingto pay 3.2 percent of their monthly income to reduce the distance to the nearestpublic facility by half, but richer households are prepared to pay only 2.1 per-cent. In contrast, the poor are willing to pay only 2.7 percent of monthly incomefor improvements in all measured indicators of health service quality, comparedwith 7 percent in the case of the nonpoor. These results suggest that in low-income communities priority should be given to improved accessibility whileensuring some appropriate level of quality (especially the availability of drugs).In areas with more affluent populations, quality considerations, especially ad-equate availability of equipment and basic services, might be a greater priority.

Are Quality Improvements Feasible?

Are the fees households are willing to pay for better health care in line withthe costs of providing such services? The willingness to pay for raising publichealth facilities to the level of those in the private sector in Ghana was equal toalmost 1 percent of monthly income (60 cedis), or about two-thirds of the av-erage fee in the private sector. When the distance was also equalized, the will-ingness to pay exceeded the average private-sector fee of 110 cedis by 10 cedis.Akin, Guilkey, and Demon (1995) found that when quality improved, the de-mand for public health facilities in Nigeria increased even when fees were raisedto the level charged by private facilities.

A further perspective comes from comparing the willingness to pay for qual-ity improvements with estimated costs per patient. Figures for Ghana from theWorld Bank and the Ministry of Health (Government of the Republic of Ghana1991, 1994) indicate that the per patient cost in a rural health facility was 400cedis in 1988; full cost recovery was obtained if twenty-five patients a day weretreated. The amount consumers were willing to pay for quality services in sucha center was 225 cedis. The corresponding unit cost for an urban facility was600 cedis and thirty-eight patients a day, and the willingness to pay was 355cedis. These comparisons suggest a high potential for cost recovery.

The simulations and field experiments reported here show that there is roomfor quality improvements financed—at least in part—by an increase in user fees.Obviously, a literal interpretation of the quality-price tradeoff would be an over-simplification. Drug and equipment availability, infrastructure, and physicalconditions are almost certainly proxies for other quality characteristics. Qualityin general would need to be significantly increased for the predicted shift to thepublic sector to occur.

Because the costs of various quality improvements were not analyzed in thedemand studies reviewed, it is unclear whether the revenues raised would be

Harold Alderman and Victor Lavy 1 3

sufficient to make the quality changes needed to prevent any loss of usage. Butthe results strongly suggest that, if prices for public health services were raisedand the revenues used to increase the quality of care, the use of both public andmodern health care in general might, in fact, increase.

The Effect of Quality on Nutritional and Mortality Outcomes

The degree of use of government clinics provides little indication of the effectthat government investments have on the health status of the poor or of thegeneral population. To some extent the subject goes beyond the scope of thisarticle; the evidence is too limited and variable to provide strong guidelines with-out a fair amount of discussion. A few general patterns, however, provide aperspective on the discussion of pricing policies and the demand for quality.First, in addition to determining which service provider is chosen, prices areimportant determinants of mortality and nutritional status. Second, the impactof service quality on these two health indicators depends on the level of educa-tion or income—or both. And third, the cross-sectional studies contain few regu-larities that can be used to prioritize improvements in services, but the diverseeffects of quality variables hint at a need to consider decentralization and changesin incentives.

Prices and Outcomes

With few exceptions, the shorter the travel and waiting time for a service, themore it will be used. But the net effect on nutrition or mortality is less clear.Moderate changes in use of public facilities may reflect shifts from private pro-viders; the net increase in the number of people using modern medical servicesmay be fairly small. In such circumstances households benefit from a reductionin travel or in direct costs, but their overall health status may be unaffected.

What evidence is there about the consequences of price changes or travel timeon mortality or nutritional status? Benefo and Schultz (1994) find that increaseddistances to clinics lead to higher mortality rates for children in Cote d'lvoireand Ghana. (In Ghana, however, this was observed only for the children ofmore-educated women.) A similar study using the same data noted that mortal-ity increased with greater distances to clinics in rural areas of Ghana (Lavy andothers, forthcoming) but found no measurable impact from either increased avail-ability of health facilities or services for children at clinics in urban areas (whereservices were in any case more accessible).4

Similarly, the price of antibiotics was a significant determinant of mortalityin Ghana. Benefo and Schultz (1994) found that if drug prices were doubled,child mortality would increase by 50 percent.5 Because the nutritional status ofchildren is sensitive to disease (Alderman and Garcia 1994), it is consistent tofind that nutrition is also affected by drug prices. Taking availability of drugs as


a proxy for the effective price of those drugs, similar evidence comes from Coted'lvoire, where communities that did not have access to antibiotics had signifi-cantly worse measures of nutritional status. Drug supplies are associated withan increase in nutrition equal to a third of the gap between the average childheight and international standards (Thomas, Lavy, and Strauss forthcoming).

Because both distance to clinics and the price of drugs tend to reduce thedemand for health services, any increase in price without a concomitant in-crease in quality is likely to have a detrimental effect. It is not possible to inferprecisely the joint effect of increased prices and better care, but it is only com-mon sense to note that cost recovery fees should be used to ensure quality im-provements. It cannot be assumed that, because the average consumer's de-mand for health does not change much in response to price changes, communityhealth indicators do not change. The average is often not indicative of the mostvulnerable population. The available evidence then implies either bounds forcost recovery or the need to implement parallel measures to protect this popula-tion simultaneously with the reform of health care, or both.

The Interaction of Education and Services

Just as price effects differ by income, the effect of health care services oftendiffers appreciably according to the education of the family. For example,Frankenberg (1995) found that government health services had only a smalleffect on the average mortality rate in Indonesia, but she noted that the effectwas three times larger for the children of uneducated women. In contrast, thereduction in child mortality attributed to the provision of both maternity clinicsand doctors was higher if the mother had at least a primary education than if shewere uneducated.

In Cote d'lvoire, the availability of antibiotics or immunization servicesbenefited children in households with no education more than in those with aneducated parent (Thomas, Lavy, and Strauss forthcoming). This finding is pairedwith an observation that parallels Frankenberg's study—childbirth services pro-vided greater benefits to households whose members had some education. InGhana, only households with little education benefited appreciably from im-proved sanitation (Lavy and others forthcoming); the impact of clinic serviceson nutrition did not vary among education subgroups. The observation thatsanitation—but not better health facilities—benefited households with little edu-cation parallels Barrera's (1990) results for the Philippines and Sastry's (1994)observations for the northeast of Brazil.

Educated households also have more of other resources, and not all the stud-ies controlled for assets and family background. Thus the most prudent inter-pretation of differences in the level of education completed is not derived fromeducation policy but from the differential impact of investments on differentincome groups. Thus Thomas, Lavy, and Strauss (forthcoming) emphasize thatthe significantly greater beneficial effect of basic services on less-educated house-


holds suggests that investing in immunizations or in the provision of pharmaceu-tical products will substantially improve the health of the poor.6 Similarly,Frankenberg's (1995) results for Indonesia show that investments such as gov-ernment health clinics, in contrast to doctors or even maternity clinics, implicitlytarget poorer households. Stated somewhat differently, prioritizing services forlow-education (low-income) communities can achieve much higher marginalreturns than a similar investment that is not targeted. From a targeting perspec-tive, the positive association between the effect of doctors and maternity clinicswith levels of education or the similar association with specialized services im-plies that these services are, in effect, subsidies that benefit primarily upper-income groups.

Education can also be seen as a complement to certain services and a substi-tute for others.7 It may, alternatively or in addition, have a direct effect onhealth services—through changes of attitudes or the ability to process instruc-tions. Educated mothers may use additional information or other assets to ob-tain higher returns from the use of community infrastructure. Or they may ownassets or have access to information and services that substitute for publiclyprovided infrastructure and thus have less need for such services than do unedu-cated mothers (Haddad, Hoddinott, and Alderman 1994). Education may alsoshift the status of a woman within the family structure and indirectly contributeto her controlling more resources by shifting her status or position in the house-hold.

Finally, although maternal education does not appear to explain child nutri-tion or mortality in Ghana and Cote d'lvoire, some inputs have a significanteffect in interaction with education in the studies cited. Such an interaction isequally an observation about education as about services; in the West Africanstudies it implies that educating mothers has a limited impact on child health ifthe mothers do not also have access to health services.

The Variability of Results

The specific services that are most effective in improving the health of thepublic are by no means the same in each community. To illustrate, in ruralGhana, the height of children is related to the number of nurses, support staff,and beds as well as to the availability of drugs (Lavy and others forthcoming),but none of these services explains height for age in urban areas of the country(although the number of doctors does). Similarly, although this study used simi-lar methodologies for data collection and analysis as Thomas, Lavy, and Strauss's(forthcoming) study of neighboring Cote d'lvoire, there are as many differencesas common factors in the variables that affect child nutrition.

To a degree differences across studies reflect different variable definitionsand different aggregations used to form quality indexes. Moreover, few studiescan control for program placement, that is, whether programs are locatedwhere they are needed most (which may lead to a false association between


programs and malnutrition) or are placed where the most powerful consum-ers live (which may lead to an exaggerated belief that the programs reducedmalnutrition). Of the cited studies, only Frankenberg (1995) controlled forplacement effects. Her results, like the few other studies in the literature thatlook at health care placement, imply that cross-section results may be under-estimations of the full effects of programs on nutrition and mortality. Yet itseems inappropriate to attribute the lack of compelling patterns to data in-adequacies alone.

One explanation for the low impact of some programs is the availability ofclose substitutes, either in the form of privately provided care or other govern-ment services. For example, maternal clinics and family planning centers mayregularly have large effects on nutrition and mortality because privately pro-vided services are not close substitutes. Conversely, services for children andgeneral facilities are often privately available; thus public facilities that providesuch services mainly benefit households that have less access to private care. Asimilar interpretation may explain the general importance of drug availabilityas well as its limited effect in urban Ghana. It may also explain why some ser-vices substitute for education in some communities and may be complements inothers at different levels of overall development.

Moreover, there often are community effects that appear to explain a fairamount of the variance in community health, yet are not directly observed byresearchers. Benefo and Schultz (1994) found that, even with the inclusion ofcommunity services and prices as well as household resources, the average mor-tality rate of other households in a community explains the probability of childsurvival. They conclude, as did Alderman and Garcia (1994) in regard to similarvariables for community morbidity, that this variable indicates important com-munity factors that have not yet been isolated.

Because the results of studies vary widely and because unspecified commoncommunity factors are important, can one infer that health priorities should bedetermined at a local level? This is the basic thrust of Hanushek's (1995) recentreview of the effects of school quality on performance. He noted that becausemeasured input into education jointly explained little about educational out-comes, and because few inputs were consistently significant in the studies re-viewed, further work should follow an output-based approach to education.Although Kremer (1995) offers a different interpretation of this evidence, hereaffirms the basic theme that we echo: Because different inputs are effective indifferent circumstances, it often is advisable to decentralize decisionmaking fordelivering health services. Local organizations may be in a better position toobserve factors that are not measured by researchers or more distantpolicymakers. If so, it is important both to allow for discretionary allocation ofthe health budget at a decentralized level and to provide the proper incentivesfor targeted services. In some cases this may mean permitting health care cen-ters to retain user fees, while in others performance-based budget support forpublic health measures may be effective.


Conclusion

In many environments significant increases in the share of resources going tothe poor can be achieved by ensuring that services which are intended for thegeneral population, such as primary health care and education, are in fact madeaccessible. To finance such services, many debt-stricken developing countries inAfrica and elsewhere have embarked on policies designed to increase the shareof operating expenses that health care systems raise from user fees. The ev-idence reviewed here confirms that new or increased user fees in public healthfacilities, when accompanied by an improvement in services, can increase use.Moreover, this increase will be positive for both the poor and the nonpoor, butmuch larger for the latter group. In particular, the poor are more likely to usemodern health care that they pay for (if resources are of high quality) than touse subsidized public care that is ineffective.

These results are encouraging from the standpoint of raising user fees to fi-nance a higher share of the recurrent costs of public primary health care and tofree resources for quality improvements. These findings indicate only the poten-tial, however. If quality improvements do not accompany cost recovery mea-sures and if such measures fail to provide for exemptions targeted on the poor,these results imply serious consequences for community health. Unfortunately,in many countries, particularly in Africa, clinics do not retain user fees but arerequired to turn them over to the government. Similarly, few countries havesuccessfully implemented sliding fee structures (Nolan and Turbat 1995). A fur-ther cautionary note comes from the observation that prices of privately pro-vided services may rise in response to changes in prices of public services (Gertlerand others 1995).

In one area—the increased provision of drugs—progress has been noted, es-pecially in francophone Africa (Nolan and Turbat 1995; McPake, Hanson, andMills 1993). Many of these countries, including Cote d'lvoire subsequent to theperiod analyzed above, have been able to increase the supply of drugs to publicclinics, albeit at higher prices.

The finding that drug availability is a priority for consumers is positivefrom a policy point of view. Unlike facilities, equipment, and human re-sources, drugs and vaccines can be readily provided by governments and cangenerate a rapid improvement in public health. The Zimbabwe governmentestablished a standard nationwide system of stock control and doubled theavailability of essential drugs in health facilities over a four-year period (WorldBank 1993).

From a different perspective, to the degree that drug availability is a commonindicator of the overall quality of public services (McPake, Hanson, and Mills1993; Jarrett and Ofosu-Amaah 1992), drug supplies may indirectly affect mortal-ity by encouraging consumers to seek out health services. As such, cost recovery


and pricing policies for drugs will be enhanced or undermined in accord with thecomplementary improvements of other services.

But such services—training and placing clinic personnel, say—take consid-erably longer to accomplish than measures to increase supplies of drugs andvaccines. Although the availability of drugs may offset the shock of imposinguser fees, the potential for delays in implementing other changes in the healthenvironment suggests that these fees may need to be phased in. Moreover, afterimplementing initial cost recovery measures, authorities often fail to followthrough with promised fee waivers and sliding fee scales to reflect income dis-parities. Thus the policy challenge is not how to keep health costs low but howto provide value for the fees charged.

NotesHarold Alderman is a senior economist with the Poverty and Human Resources Di-

vision, Policy Research Department, the World Bank. Victor Lavy is a professor in theDepartment of Economics at the Hebrew University of Jerusalem. This article wasoriginally prepared for the Workshop on the Effects of Service Quality and Cost onEducation and Health, World Bank, May 18, 1994. The authors thank Jere Behrman,David Hotchkiss, Emmanuel Jimenez, Reiko Nimii, and Duncan Thomas for commentson an earlier draft.

1. Although anthropometric measures of nutritional status are strongly related to theprobability of mortality (Pelletier, Frongillo, and Habicht 1993), they measure differentdimensions of child welfare that may respond to different interventions (Kielmann, Tay-lor, and Parker 1978).

2. Drug availability can range from 0, in the case where none of the drugs are available,to 1, if they all are present. If only one drug is present, the measure is defined as 0.33.

3. The sample means for public and private facilities are 8 and 10.4, respectively.4. For the dependent variable, Benefo and Schultz (1994) use the proportion of children

in the family who were born at least five years before the survey and died before their fifthbirthday. Lavy and others (forthcoming) use the probability of survival (hazard of mortal-ity) of all children born in the decade. Unlike Lavy and others, Benefo and Schultz do nothave a separate regression for rural populations. They do, however, include a dummyvariable for rural residence and its interaction with education (the former increases andthe latter decreases mortality).

5. Most studies find that nutrition and child survival are responsive to food prices (Al-derman 1993; Behrman, Deolalikar, and Wolfe 1988). Specifically, the studies from WestAfrica that provide much of the evidence on service quality described here also show thatfood prices have a pronounced effect on health indicators (Thomas, Lavy, and Straussforthcoming; Lavy and others, forthcoming; and Benefo and Schultz 1994).

6. Thomas, Lavy, and Strauss (forthcoming) include a measure of household expendi-tures. Thus education is not merely a proxy for unobserved income. They do not test,however, whether the effects of infrastructure differ by income.

7. Cleland and van Ginneken's (1988) review found that the strong inverse relation ofeducation and child mortality generally survives controls for income and assets, althoughnot always without attenuation. For a similar observation on nutrition, see Thomas, Strauss,and Henriques (1991).


ReferencesThe word "processed" describes informally reproduced works that may not be com-

monly available through library systems.

Abel-Smith, Brian, and Pankaj Rawal. 1992. "Can the Poor Afford 'Free' Health Services?A Case Study of Tanzania." Health Policy and Planning 7(4):329-41.

Akin, John, Charles Griffin, David Guilkey, and Barry Popkin. 1988. "The Demand forPrimary Health Care Services in the Bicol Region of the Philippines." Economic Devel-opment and Cultural Change 34:755-82.

Akin, John, David Guilkey, and Hazel Denton. 1995. "Quality of Services and Demandfor Health Care in Nigeria: A Multinominal Probit Estimation." Social Sciences andMedicine 40(11): 1527-37.

Alderman, Harold. 1993. "New Research on Poverty and Malnutrition: What Are theImplications for Research and Policy?" In Michael Lipton and Jacques van der Gaag,eds., Including the Poor. Washington, D.C.: World Bank.

Alderman, Harold, and Marito Garcia. 1994. "Food Security and Health Security: Ex-plaining the Levels of Nutritional Status in Pakistan." Economic Development andCultural Change 42(3):485-508.

Alderman, Harold, and Paul Gertler. 1989. The Substitutability of Public and PrivateHealth Care for the Treatment of Children in Pakistan. Living Standards MeasurementStudy Working Paper 57. Washington, D.C.: World Bank.

Barrera, Albino 1990. "The Role of Maternal Schooling and Its Interaction with PublicHealth Programs in Child Health Production." Journal of Development Economics32:69-91.

Behrman, Jere, Anil Deolalikar, and Barbara Wolfe. 1988. "Nutrients: Impacts and Deter-minants." The World Bank Economic Review 2(3):299-320.

Benefo, Kofi, and T. Paul Schultz. 1994. Determinants of Fertility and Child Mortality inCote d'lvoire and Ghana. Living Standards Measurement Study Working Paper 103.Washington, D.C.: World Bank.

Cleland, John, and Jerome van Ginneken. 1988. "Maternal Education and Child Survivalin Developing Countries: The Search for Pathways of Influence." Social Science andMedicine 27(12):1357-68.

De Ferranri, David. 1985. Paying for Health Services in Developing Countries: An Over-view. Staff Working Paper 670. Washington, D.C.: World Bank.

Frankenberg, Elizabeth. 1995. "The Effects of Access to Health Care on Infant Mortalityin Indonesia." Health Transitions Review 5(2, November).

Garner, Paul, Jane Thompson, and Daryl Donaldson. 1990. "Quality Assessment ofHealth Facilities in Rural Papua New Guinea." Health Policy and Planning 5(1):49-59.

Gertler, Paul, and Jacques van der Gaag. 1990. The Willingness to Pay for Medical Care:Evidence from Two Developing Countries. Baltimore: Johns Hopkins University Pressfor the World Bank.

Gertler, Paul, Jack Molyneaux, Carl Sennato, S. Ahmed, S. Indrajaya, and ElizabethFrankenberg. 1995. "Pricing Public Health Care Services: Results from a Social Ex-periment in Indonesia." RAND Corporation, Santa Monica, Calif. Processed.

Government of the Republic of Ghana, Ministry of Health. 1991. "Ghana Health in Brief."Accra. August. Processed.

. 1994. "Design Proposal for Prototype Health Centers." Accra. February. Pro-cessed.

Griffin, Charles, John Briscoe, Bhanwar Singh, Radhika Ramasubban, and Ramesh Bhatia.1995. "Contingent Valuation Meets the Cold Truth of Actual Behavior: Predicting Con-


nections to New Water Systems in Kerala, India." The World Bank Economic Review9(3):373-96.

Haddad, Lawrence, John Hoddinott, and Harold Alderman. 1994. "Intrahousehold Re-source Allocation. An Overview." Policy Research Working Paper 1255. Policy Re-search Department, World Bank, Washington, D.C. Processed.

Hanushek, Eric. 1995. "Interpreting Recent Research on Schooling in Developing Coun-tries." The World Bank Research Observer 10(2):227-46.

Hotchkiss, David. 1993. "The Role of Quality in the Demand for Health Care in Cebu,Philippines." Department of Economics, University of North Carolina at Chapel Hill.Processed.

Jarrett, Stephen, and Samuel Ofosu-Amaah. 1992. "Strengthening Health Services for MCHin Africa: The First Four Years of the 'Bamako Initiative'." Health Policy and Planning7(2):164-76.

Kielmann, Arne, Carl Taylor, and R. Parker. 1978. "The Narangwal Nutrition Study: ASummary Review." American journal of Clinical Nutrition 31:2040-52.

Kremer, Michael. 1995. "How Can Schools in Developing Countries Be Improved?" TheWorld Bank Research Observer 10(2):247-54.

Lavy, Victor, and Jean-Marc Germain. 1994. Quality and Cost in Health Care Choice inDeveloping Countries. Living Standards Measurement Study Working Paper 105. Wash-ington, D.C: World Bank.

Lavy, Victor, John Strauss, Duncan Thomas, and Phillipe de Vreyer. Forthcoming. "Qual-ity of Health Care, Survival and Health Outcomes in Ghana." Journal of Health Eco-nomics.

Litvack, Jennie, and Claude Bodart. 1993. "User Fees and Improved Quality of HealthCare Equals Improved Access: Results of a Field Experiment in Cameroon." SocialScience and Medicine 37(3):369-83.

McPake, Barbara, Kara Hanson, and Anne Mills. 1993. "Community Financing of HealthCare in Africa: An Evaluation of the Bamako Initiative." Social Science and Medicine36(ll):1383-95.

Manning, Willard, Joseph Newhouse, Naihua Duan, Emmet Keeler, Arleen Leibowitz,and Susan Marquis. 1987; "Health Insurance and the Demand for Medical Care: Ev-idence from a Randomized Experiment." American Economic Review 77:251-77.

Mosley, William, and Lincoln Chen. 1984. "An Analytic Framework for the Study ofChild Survival in Developing Countries." Population and Development Review10(Supp.):25-48.

Mwabu, Germano. 1989. "Nonmonetary Factors in the Household Choice of MedicalFacilities." Economic Development and Cultural Change 37(2):383-92.

Mwabu, Germano, Martha Ainsworth, and Andrew Nyamete. 1993. "Quality of MedicalCare and Choice of Medical Treatment in Kenya: An Empirical Analysis." journal ofHuman Resources 28(4):838-62.

Nolan, Brian, and Vincent Turbat. 1995. Cost Recovery in Public Health Services in Sub-Saharan Africa. Washington, D.C: Economic Development Institute of the World Bank.

Pelletier, David, E. Frongillo, and Jean-Pierre Habicht. 1993. "Epidemiologic Evidence fora Potentiating Effect of Malnutrition on Child Mortality." American Journal of PublicHealth 83:1130-33.

Roemer, M. I., and C Montoya-Aguilar. 1988. "Quality Assessment and Assurance inPrimary Health Care." World Health Organization Offset Publication 105. Geneva.Processed.

Sastry, Narayan. 1994. "Community Characteristics, Individual Attributes, and Child Sur-vival in Brazil." Office of Population Research, Princeton University, Princeton, N.J.Processed.


Schwartz, J. B., J. S. Akin, and B. M. Popkin. 1988. "Price and Income Elasticities ofDemand for Health Care: The Case of the Infant Delivery in the Philippines." TheWorld Bank Economic Review 2{l):49-76.

Strauss, John. 1990. "Household, Communities and Preschool Nutrition Outcomes: Evidencefrom Rural Cote d'lvoire." Economic Development and Cultural Change 38:231-62.

Thomas, Duncan, Victor Lavy, and John Strauss. Forthcoming. "Public Policy andAnthropometric Outcomes in Cote d'lvoire." Journal of Public Economics.

Thomas, Duncan, John Strauss, and Maria Helena Henriques. 1991. "How Does Mother'sEducation Affect Height?" Journal of Human Resources 26(2):183-211.

Weaver, Marcia, R. Chapko, R. Ndamobissi, C. Kornfield, C. Blewane, and A. Sathe.1993. "Estimating Willingness to Pay for Quality of Care: Comparison of ContingentValuation and Two-Step Health Expenditure Methods." HFS Smaller Applied ResearchPaper 9. Abt Associates, Inc., Cambridge, Mass. Processed.

World Bank. 1993. World Development Report 1993: Investing in Health. New York:Oxford University Press.

Wouters, Annemarie, Olusoji Adeyi, and Richard Morrow. 1993. "Quality of Health Careand Its Role in Cost Recovery." U.S. Agency for International Development, Washing-ton, D.C. Processed.


Documents

HOUSEHOLD RESPONSES TO PUBLIC HEALTH SERVICES: COST …documents.worldbank.org/curated/en/... · shift from self-care to the modern health sector. If the objective of policy im-provements