H N P D I S C U S S I O N P A P E R Public and Private Roles in Health Theory and Financing Patterns Philip Musgrove July 1996 PUBLIC AND PRIVATE ROLES IN HEALTH Theory and Financing Patterns Philip Musgrove July 1996 Health, Nutrition and Population (HNP) Discussion Paper This series is produced by the Health, Nutrition, and Population Family (HNP) of the World Bank's Human Development Network (HNP Discussion Paper). The papers in this series aim to provide a vehicle for publishing preliminary and unpolished results on HNP topics to encourage discussion and debate. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliated organizations or to members of its Board of Executive Directors or the countries they represent. Citation and the use of material presented in this series should take into account this provisional character. For free copies of papers in this series please contact the individual authors whose name appears on the paper. Enquiries about the series and submissions should be made directly to the Editor in Chief. Submissions should have been previously reviewed and cleared by the sponsoring department which will bear the cost of publication. No additional reviews will be undertaken after submission. The sponsoring department and authors bear full responsibility for the quality of the technical contents and presentation of material in the series. Since the material will be published as presented, authors should submit an electronic copy in a predefined format as well as three camera-ready hard copies (copied front to back exactly as the author would like the final publication to appear). Rough drafts that do not meet minimum presentational standards may be returned to authors for more work before being accepted. The Editor in Chief of the series is Alexander S. Preker (apreker@worldbank.org); For information regarding this and other World Bank publications, please contact the HNP Advisory Services (healthpop@worldbank.org) at: Tel (202) 473-2256; and Fax (202) 522-3234. ISBN 1-932126-23-6 © 1996 The International Bank for Reconstruction and Development / The World Bank 1818 H Street, NW Washington, DC 20433 All rights reserved. ii Health, Nutrition and Population (HNP) Discussion Paper Public and private roles in health Theory and Financing Patterns Philip Musgrovea aPrincipal Economist, Health, Nutrition and Population (HNP), the World Bank, Washington, DC, USA (at the time of writing; currently Chief Economist, Disease Control Priorities Project, Fogarty International Center, National Institutes of Health, Bethesda, MD, USA Paper prepared for the World Bank, Washington, DC, 1996 Abstract: The appropriate role of the state in health is complex both in economic theory and in practice. Theory identifies three reasons for state action: public goods or services with large externalities (involving efficiency); poverty (involving equity); and failings peculiar to insurance markets for health care (where both inefficiency and inequity arise). The insurance domain presents the most costly and difficult problems, and explains why--in contrast to other sectors--governments tend to finance an increasing share of health care as incomes rise. Regulation, mandates and provision of information are also crucial public instruments; public provision of care is less important. As income rises, governments finance an increasing share of health care, displacing out-of-pocket expenditure by either tax-financed care or social security contributions. Private voluntary insurance is rare in most countries. Keywords: public health, state and market in health, economic theory of health, health financing Disclaimer: The findings, interpretations and conclusions expressed in the paper are entirely those of the authors, and do not represent the views of the World Bank, its Executive Directors, or the countries they represent. Correspondence Details: Philip Musgrove; Fogarty International Center, National Institutes of Health, Bldg. 16, 16 Center Drive, MSC 6705 Bethesda, Maryland 20892-6705; Tel: (301) 402-0334; Email: musgrovp@mail.nih.gov. iii iv Contents Foreword............................................................................................ x Acknowledgements................................................................................ xiv Executive Summary............................................................................... xvi PART I: INTRODUCTION...................................................................... 1 Why The Public Role In Health Care Matters.......................................... 1 Choices For State Intervention............................................................ 2 A Road Map.......................................................................... 4 PART 2: A CONCEPTUAL BASIS FOR PUBLIC AND PRIVATE ROLES.......... 5 The Three Domains Of Health Care.................................................... 5 Intervention for Public Goods in Health..................................... 10 The Public Role In Low-Cost Private Interventions........................ 11 Risk-Sharing For Catastrophically Costly Private Goods................. 14 Market Failure And Health Care Needs................................................ 19 Dealing with poverty...................................................................... 19 A Minimal State Role In The Absence Of Poverty........................... 19 How Poverty Complicates Public Roles...................................... 21 Summary: justifications and risks of state intervention.............................. 24 PART 3: EMPIRICAL PATTERNS AND EXPLANATIONS........................... 26 Economic issues and health system objectives......................................... 26 Overall Health Spending And Public/Private Composition.......................... 28 What is known about health expenditure...................................... 28 How expenditure level and composition are related to income............ 29 Explaining Health Outcomes...............................................................32 Life Expectancy At Birth......................................................... 34 Childhood Mortality.............................................................. 35 Summary: Expenditure And Results............................................ 40 Out-Of-Pocket Spending Versus Private And Public Insurance...................... 41 The Income-Related Expansion Of Insurance................................. 41 The Two Forms Of Publicly-Financed Insurance............................. 44 State Intervention In The Insurance Domain............................................ 47 Why Social Insurance Is So Widespread....................................... 48 Non-Financial State Intervention............................................... 50 PART 4: CONCLUSIONS........................................................................ 51 The Appropriate Public Role In Health.................................................. 52 What Government Should Not Do.............................................. 52 What Government Should Do................................................... 54 How To Spend Public Money On Health Care.......................................... 59 REFERENCES...................................................................................... 63 STATISTICAL ANNEX........................................................................... 77 Figures Page 1 Three Domains of Health Care 6 2 Need, Demand and Supply for Health Care 20 3 Share of GDP Spent on Health, Related to Income 30 4 Public Expenditure on Health as a Share of Total Health Expenditure, Related to Income 31 5 Share of GDP Devoted to Health Expenditure, and its Distribution between Public and Private Spending Countries Classified by Income 33 6a Life Expentancy and Health Expenditure: Countries Classified by Income 36 6b Life Expentancy and Health Expenditure: Countries Classified by Share of Public Expenditure 37 7a Child Mortality and Health Expenditure: Countries Classified by Income 38 7b Child Mortality and Health Expenditure: Countries Classified by Share of Public Expenditure 39 8 Distribution of Health Expenditure among Out-of-Pocket Spending, Private Insurance and Social Insurance: Countries Classified by Income 42 9 Distribution of Health Expenditure among Private Spending Mandated Social Insurance and Directly-Financed Public Expenditure: Countries Classified by Income 46 10a Appropriate Use of the Instruments of Public Intervention in Health Care 56 10b Typically Inappropriate Use of the Instruments of Public Intervention in Health Care 57 FOREWORD In few areas of the economy are the relations between the state and the market as complex as they are in the health sector. Substantial market failures provide reasons for much greater public involvement that in many other activities. At the same time, public intervention gas often given rise to what may be called government failures. Economic theory provides a valuable but incomplete guide to what governments ought to do to improve efficiency and equity in health care and in the mechanisms, particularly insurance, to finance it. Conceptual argument therefore needs to be complemented by empirical study of how countries organize health financing and delivery, and how those decisions are related to a population's health status, to the level of expenditure on health, to people's satisfaction with the resulting system and to the medical and financial equity with which it operates. This paper was written primarily for World Bank staff and their colleagues in borrowing countries. It summarizes both the theoretical discussion of appropriate public and private roles in health, and factual information about what is spent on health care, how health systems are organized, and which financing mechanisms are used. It concludes that there are three reasons for the state to interfere in health markets: to ensure the right production and consumption of interventions that provide public goods or externalities; to avoid the waste and inequity of unregulated, private voluntary insurance for costly health interventions; and to protect the poor, who can afford little or no care or insurance out of pocket. Each of these tasks calls for particular instruments; if they are not used, or used wrongly, intervention can make things worse rather than better. Particularly now that health expenditures are a large and growing share of many economies, and many countries are engaged in efforts to reform their health systems, this analysis may be useful to public debate and decisions in this crucial and controversial sector. David de Ferranti Director Human Development Department ACKNOWLEDGEMENTS Catherine Jourdan assembled much of the statistical material on which this paper draws. Joanne Epp completed this task and prepared the first version of Annex Table 1; Yvette Atkins and Akiko Maeda produced the final version. Cathie Boezer prepared the first version of the references, which Linda Kean helped to finish. The regressions used to compare health expenditure with national income and with health outcomes were estimated by Hailu Mekonnen. Kathy Rosen transformed the data and regression results into Figures 3-9, and turned the author's sketches into Figures 1, 2, 10A and 10B. Fabiana Imperatriz helped prepare the final version for publication. The paper was written under the guidance of David de Ferranti, who also commented extensively on the first draft and made a valuable suggestion about the overall structure. Jacques van der Gaag, George Schieber and David Dunlop read early versions of the paper and provided immensely helpful information, suggestion and encouragement. Howard Barnum, José-Luis Bobadilla, Xavier Coll and Helen Saxenian also offered useful comments. Linda Kean edited the paper to take all these comments into account. The final version benefited from review by Nicholas Barr, Ralph Harbison, Elizaabeth King, Maureen Lewis, Samuel Lieberman, Paul Shaw and Verdon Staines, and verbal comments by Jeffrey Hammer. Any remaining errors or confusion are the author's sole responsibility. The author is also grateful to the Worlbank for publishing this report as an HNP Discussion Paper. EXECUTIVE SUMMARY Health systemsare complicatedin principleand extremelyvariedin practice,so it is not easyto reachconclusionsaboutthe appropriateroleof the statein the health sector. Governmentscan employfivedifferentinstrumentsto affectthe outcomesa privatemarket wouldgenerate--theycan createand disseminateinformation,regulateprivateactivity, mandatecertainactionsby individualsor firms, financehealth-relatedservicesor deliver those servicesthroughpublicfacilitiesand staff. Choicesaboutthe state role meandeciding what shouldbe leftto the privatesector,howheavilythoseprivateactivitiesshouldbe regulatedor otherwiseinfluencedbypublicpolicy,and whatactivitiesare best dealtwith by substantialstateinterventionsuch as provisionof care and mandatingor financinginsurance coverage. Boththeoryand the availableempiricalevidencesuggestthat there is a small plateauof somewhatdifferent,but all moderatelysuccessful,combinationsof public and privateactivitieswhichdifferentcountrieshavereached. Over this area, all the major objectivesof a health systemcan be more or less well satisfied,althoughthere are always conflictsbetweenone objectiveand another. Theredoes not seem to be a sharppeak of idealpublic and privaterelationsrisingabovethe plateau. The most importantconclusions or recommendationsfor publicpolicythenhaveto do withassuringa place on the plateau, and avoidingthe numerouschasmswhichsurroundit. Thesechasms,which represent failuresof variouskinds, result fromtoo much, too little or the wrongkind of state intervention. Economictheoryindicatesthree distinctjustificationsfor state interventionin the health care market. Theseare to assurethe optimalproductionof public goods;to corrector offsetfailingsin the marketfor healthinsurance;and to subsidizeconsumerstoo poor to buy insuranceor the inexpensivehealthcare that the non-poorcan financeout-of-pocket. Appropriateinterventionin anyof theseareas can contributeto one or more of the four main generalresultsthat people appearto wantfrom a healthcare system:goodhealth,low cost, satisfactiononthe part of both consumersand providers,and equity,both medicaland financial. This paper developsthe argumentsfor interventionand their limitations;relates themto the instrumentsavailableto governmentsfor affectingmarketoutcomes;and examineshowhealth care is actuallypaid for and providedin a large numberof countries and howthe level, compositionand mechanismsof financeappear to affecthealth outcomes. Thetheoreticaland empiricalanalyseslead to a considerationof why healthsystemshave typicallyevolvedin certainways,whetherthat is consistentwiththe theoreticallyappropriate rolesof the state and the privatesectors,and whatadvantagesor failingsthe systemsshow. Someconclusionscanbe drawnfromtheoryalone; othersemergefrom empirical study; and somefindingsdependonboth kindsof analysis. Theseconclusionsare developed in the body of the paper and are brieflysummarizedhere-- + The health sectoris neitherjust likeany other economicsectornor totallydistinctfrom them. While noneof the reasonsfor stateinterventionis uniqueto the health sector,certain marketfailuresare worsethere thanelsewherein the economy. A largerstate role is 1 2 generallywarrantedthan in other sectors,but still not a role of the sizeand nature the governmentactuallytakes in manycountries. * Empirically,the state role, particularlyin financing,tendsto expandas incomeincreases. The expandedrole for the stateis a consequenceof the increasedimportanceof healthcare interventionswhichcan be catastrophicallycostly; it parallelsthe historicalexpansionof the importanceof insuranceagainstthese risks. * There is a smallbut extremelyimportantcollectionof health-relatedactivitieswhichmust be financedby the stateif they are to be providedat all, or providedat the sociallyoptimum levelof consumption. Theseinterventionsappearto accountfor much of the impactof healthspendingon healthimprovements.Theyprobablyexplainwhypublichealth expenditureis somewhatmoreeffectivethanprivateexpenditurein extendinglife expectancy. * Thesepublichealth activitiesare especiallyimportantat low incomelevels,forboth epidemiologicaland economicreasons,so that publicfinancingmay be particularlycrucial for health inpoor countries. Publicfinancebecomeslessurgent, but not less common,at higher incomes. (Publicprovision,in contrast,becomesless common.) * Despitethe potentialimportanceof stateinterventionin poor countries,there is little relationin thosecountriesbetweenwhat is spenton healthcare and lifeexpectancyor child mortality. This is partly becausehealthstatusat lowincomesdependsmore on factors outsidethe health care sector; oncethe healthgainpossiblefromthose factorsis realized, health expenditureappearsin somerespectsto be moreeffectivein richercountries. * For thoseactivitieswhichdo not haveto be financedby the state,and which canbe providedprivatelyto non-poorconsumers,poolingof risksthroughprivateinsuranceis theoreticallythe appropriatefinancialsolution. However,privateinsurancemarketshave failingsof their own, which haveled in mostcountriesto a largeor evenpredominantshare of insurancebeing financedsocially. * Tryingto do withoutsocialinsuranceand rely onprivaterisk-sharingrequiresextensive stateinterventionof other kinds, if the consequencesof marketfailureare to be avoided. Thusthe state roleremainsimportantevenwhenboth publicgoodsand povertyceaseto be major determinantsof publicintervention. The optionforthe state simplyto stayout of the health care marketrequiresthat a societybe willingto pay substantialcosts in bothefficiency and equity. * The appropriatepublicroledoes not necessarilyincludeprovidinghealthcare; the state need not deliverservicesto makethemavailable. (Theremaybe exceptionsfor a few crucial services,or perhapsfor somegeographicareas, wherecompetitiveprivateprovision is infeasible.) 3 * All peopleface potentiallycatastrophichealthrisks, but somecannotaffordenough insuranceto protectthemselves. Thisaccountsfor much of the staterole in all countries, and affectsthe natureof stateintervention. If it werenot for poverty,publicinterventionin health care could be more limited,and mightrequireless publicfinance. * Overall,the evidenceis that for governmentsto achievethe four distinctobjectivesof cost control,goodhealthoutcomes,equityand popular satisfaction,all fiveinstrumentsof stateinterventionare relevant,and are typicallyapplied. Partlybecauseof conflictsamong the differentobjectivesof a health system,it is not clear whichcombinationof actionsis best. It is clear that "success"dependson such interference,and not simplyon who paysfor or provideshealthcare. Manycountriesin all parts of the worldare nowengagedin effortsto reformtheir health sectors,effortswhichofteninvolvesubstantialchangesin the relationbetweenthe governmentand the privatesector. The discussionhere of whereand how to drawthe boundarybetweenpublicinterventionand privateactivityis meantto aid understandingof the issuesinvolvedand to contributeto betterpublicpolicy. 5 PART1: INTRODUCTION The questionof whatgovernmentsoughtto do, and particularlywhen and howthey ought to interveneto changethe outcomesof privateactivities,is one of the perennialissues of development. Much economictheory is devotedto determiningunder what conditions interventionbythe statecan improveonprivateoutcomes. Empirically,it appearsthat excessivestate interferenceis responsiblefor mucheconomicdevelopmentfailure,and that countrieswhichhaveprosperedhavegenerallylimitedsuch interferenceand allowedprivate activityto contributemoreto progress[WorldBank 1991]. If this is largelytrue in industry, agriculture,transportand other sectorsof the economy,it is reasonableto ask whetherthe sameconclusionholdsfor healthcare. This questionof what the stateshouldand shouldnot do in the healthsectorhas beentakenup at intervalsbythe WorldBank[1980, 1987, 1993b],in parallelwith a large and steadyincreasein lendingfor healthprojectsand a growingempiricalknowledgeof the healthsectorsof a varietyof countries[WorldBank 1995]. More generally,the issueof the stateversusthe marketin healthcare has inspired much recent studyand debate [Bennett1991,Birdsalland James 1992,Camposet. al. 1987, Departmentof PublicHealth and Policy1993,McLachlanand Maynard1982,Pan American Health Organization1991,Reinhardt1992]. Particularlyin low- and middle-income countriescontemplatingor undertakinghealth sectorreform, this debate is often characterizedby someconfusionoverthe definitionand scopeof the terms "public"and "private",as wellas relatedterms such as "competition","privatization",and the various meaningsof "market"in healthcare [Saltman1995]. Thispaper synthesizescurrent theoreticalunderstandingof the subjectand providessupportingempiricalinformation referringparticularlyto the questionof howhealthcare is paid for. Why the Public Role in Health CareMatters Health care in about 1990cost at least $ 1.7 trillion, or about 8 percentof world income[Murray,Govindarajand Musgrove1994],makingit one of the largestindustriesin the globaleconomy. On average,60percentof this is public spending. If this spendingis excessiveor otherwiseinappropriate,the consequencesfor the economyand for health outcomescould be substantial. Governmentsalso providea large shareof healthservices, sometimesas large as the sharein spending,and oftenintervenein variouswaysin the privatehealthcare market. Sincemosthealthcare is a privategood, it is surprisingthat so much of it is provided,financedor regulatedby the state. In contrastto whathappensin many other sectorsof the economy,this substantialpublicrole is mostpronouncedin high- incomecountrieswhichare generallyvery market-oriented;the stateusuallyfinancesa smaller shareof in health care in poorer countries. Doesthis patternprovidemodelsfor the futuredevelopmentof the health sectorand the publicrole in it, for low-incomecountries? And is the variationamongcountriesinthe amountand kind of public interventionassociated with differencesin people'shealth,in what is spenton healthcare or in howwell health systemsfunction? 6 Besidesconsuminglargeresources,manyhealth systemsare regardedas inefficientor inequitableor both; theyare oftendescribedas in "crisis", as needing"reform",or as having "failed". Are the supposedfailuresof healthsystemsreal ? If so, theymightbe caused by ~misguidedpublic intervention,so theycouldbe correctedbya smalleror differentpublic role and a greaterrelianceon privatemarkets. Or governmentsmightintervenefor sound reasons,to corrector compensatefor failingsin those markets;that is, outcomeswouldbe evenworseif leftentirelyto the privatesector. Is there an appropriatefrontierbetween privateand public action,and a best combinationof instrumentsfor the stateto use whenit intervenes? Theseand relatedquestionsmustbe confrontedin anyreformeffort [Aaron 1994,Cutler 1994,OECD 1994]. It is relativelysimpleto concludethat governmentsshould do certainthings and shouldleaveothersto privateactivity;but oftenthere is a varietyof possiblesolutionsand no obviouslybest approach. Theorydoes not alwaysprovideclear answers,and the empiricalevidenceis incomplete,extremelyvariedand difficultto interpret. Choices for State Intervention It mattersnot only whethergovernmentsintervene,but also howthey do it: the secondessentialquestionis whatthe publicsectorshoulddo, giventhat someproblemin the privatemarketappearsto warrantsomepublic action. Thisis particularlyimportantbecause governmentfailingsin the health sectorare also common,and oftenresult from intervening in the wrongwaysor withthe wronginstruments. Thereare fivedistinctinstrumentsof public intervention:arrangedfromthe least to the greatestintrusionintoprivatedecisions, these are to-- e inform,whichmay meanto persuade,but doesnot requireanyoneto do anything. Governmentsdo this whentheypublicizethe health risks of smoking,or includehealth and basic hygieneeducationin publicschools. Theseare examplesof informationdirectedat consumers,but governmentsalso informhealthcare providersand suppliersof healthcare inputs,as byconductingresearchand disseminatinginformationon diseasepatternsand on the effectsand risks of medicalprocedures. * regulate,whichdetermineshow a privateactivitymaybe undertaken. Governments sometimesregulatethe medicalprofessionby settingstandardsfor doctorsor accrediting hospitals,althoughthese activitiesmayalso be undertakenbyprivatebodies. And governmentregulationis commonin the insuranceindustry,in the importationof medical equipment,drugsand suppliesand in the protectionof food and waterquality. More generally,governmentscan influenceprivatehealthcare activityin manyways,often combiningregulationwithsomefinancialincentivesto offsetthe costs[Bennettet. al. 1994] but withoutpublic financing. Regulationis usuallypursuantto a law,and is often determinedby an executiveor administrativebody. * mandate,which obligatessomeoneto do somethingand (usually,thoughnot always)to pay for it. Compliancewithregulationscan also implysubstantialprivatecosts; but a mandatedactivityis differentin that it must be performed,whereasa privateproducercan 7 react to regulation by choosing not to undertake the activity. Mandates are usually specified in law, which may subsequently be adumbrated by regulation. The most important mandates, in financial terms, are the requirements that employers provide health services or -insuranceto their employees, or contribute to social insurance funds for that purpose. Governments can also impose mandates on individuals, as by requiring that children entering school be immunized. * finance health care with public funds. Because mandated insurance is effectively paid for by an earmarked, involuntary contribution which is equivalent to a tax, "public" health expenditure is commonly defined to include such costs along with expenditures from public budgets. (Some important differences between these two ways of paying for health care are treated further in Part 3.) The obverse of spending public funds is to tax particular activities or goods, such as alcohol and tobacco, at least partly for health reasons. This issue is not treated here; while taxation which reduces consumption of specific goods may have substantial health effects, it is alwayslimited to very few goods and is not used systematically to promote health. Finally, the state may-- ' provide or deliver services, using publicly-owned facilities and civil service staff. This is what Ministries of Health in most poor countries do; so do various governmental bodies in many countries at all income levels. Once a society has decided to finance health services with public funds, the choice arises of whether to provide them through public facilities or to pay private producers to provide them. The appropriate way to consider this choice is as a standard "make or buy" decision. The issues for a government are the same as for a private firm, and turn on costs--is it cheaper to produce something than to buy from an outside supplier ?--and on the risks and difficulties of enforcing contracts and avoiding fraud when dealing with such suppliers [Coase 1988]. Because public financing requires public resources, it is perhaps the crucial choice about state action. However, all the instruments mentioned have costs; even information is not free. The benefits from any intervention have alwaysto be weighed against these costs. In addition, sometimes two instruments overlap, or one requires the use of another: mandates imply regulation, and public provision usually implies at least partial public finance. (User fees or even private insurance payments to public facilities may cover part of the cost and could in principle cover all of it.) And two instruments can be alternatives: for part of the population, governments can either finance health care or mandate financing by employers or other private institutions. Differences in ability to pay make it natural to operate mixed systems of public intervention, such as mandated coverage for the non-poor and public finance for the poor. There are alternatives even within mandated coverage, such as "play or pay" arrangements in which employers can finance health care directly for their employees or pay into a social security scheme. These instruments can all be applied to the delivery of health care services, but they can also be used to influence the markets for variousfactors or inputs used to produce those services. Thus governments may set standards for medical supplies (especially drugs) and 8 equipment,and sometimesoperatelaboratoriesor factoriesto producethem. State interventionin factormarketsmaygreatlyinfluencewhichservicesare produced,for whom, and at whatcost, evenwithminimalpublic interferencein serviceproduction. Although these marketsare also susceptibleto failings--andgovernmentsalso may failwhen they intervenein them--theanalysishere is limitedto healthcare and insurance. A RoadMap Theessentialquestionto ask in determiningthe appropriaterolesof the privateand public sectorsis whetherhealthservicesdifferfrom othergoodsand services in some fundamentalwaysthatjustify or explainthe extensiveand variedpublic intervention observedin the health sector. The objectis not to dividethe sectorinto separatepublic and privatespheres,but ratherto determinewhat functionsor activitiesare best undertakenby the state and bythe market, whetherthey operateseparatelyin or variousformsof collaboration[vander Gaag 1995]. Part 2 of the paper is devotedto a conceptualdiscussion of the peculiaritiesof health and the cases in whichpublic interventionappears necessaryor at least advisableto improveon the outcomesthat privatemarketswouldgenerate. To simplifythis discussion,it is initiallyassumedthat no one is poor. Povertyis, of course,both an importantcause and a consequenceof ill health [Behnnan1990]. Takingit into accountgreatlycomplicatesthe issueof whatthe stateshouldand should not do in healthcare, so it is appropriateto treatthis issueseparatelyand in somewhatmore detail. The last sectionof Part 2 thereforeis devotedto a discussionof howconcernfor the poor affectsfinancialand institutionalarrangements,boththeoreticallyand in practice. The conceptualdiscussionis necessarilyabstract,and sayslittle about howhealth systemsare actuallyorganizedor function--andin particular,whethergovernmentsin fact play the role that theory suggeststhey shouldplay. Part 3 is thereforeempirical:it presents informationon the leveland compositionof healthcare spendingand its relationto sectoral organization. Financingis emphasizedpartlybecauseit is the easiest typeof interventionto quantify. The presentationin Part 3 is primarilydescriptive,but includessomediscussion both of the historicaldevelopmentof healthsystemsand of the apparentbenefitsand drawbacksof particular financingand institutionalarrangements. The intentionis to explain, so far as possible,whatthe stateactuallydoes in differenthealthsystemsand why that role does or does not correspondto the conceptuallyappropriaterole forpublic action. Finally,Part 4 drawstogetherthe conclusionsthat it seems legitimateto derive from the combinationof conceptualdiscussionand empiricalinformation. As a guide to policy and stateaction, theseare expressedas a set of Dos and Don'ts for governments--things that governmentsgenerallyoughtnot to do, and othersthey shoulddo, to approacha social optimumin health care. Part 4 ends witha discussionof howbest to spendpublicmoney for health care. 9 PART2: A CONCEPTUAL BASIS FOR PUBLIC AND PRIVATEROLES IN HEALTH The health sector is sufficientlycomplicated, and the conditions of countries are sufficiently different, that economic theory by itself is an inadequate guide to where the frontier should be drawn between the private economy and state action, or to which state interventions should be undertaken, and in what degree. Nonetheless, theory is essential, particularly where two issues are concerned. Both of these refer to market failure, or circumstances in which private markets either cannot be expected to function at all, or can be expected to yield undesirable outcomes which appropriate public intervention might improve on. Some such failures may occur in any sector of the economy; traditional public finance theory [Musgrave 1959] explains these cases and provides guidance for public action. Other possible failures arise where insurance is involved, and for reasons specific to the health sector, present particularly acute problems for health insurance [Arrow 1963, 1985]. Since what the health care sector provides to consumers and beneficiaries are specific activities or interventions, it is useful to organize a conceptual basis according to distinctions among these activities. The next section provides this classification; three subsections elaborate on the peculiarities of each area. The Three Domains of Health Care While the activities that promote, protect or restore health are very heterogeneous, they fall into three natural domains, corresponding topublic goods, to low-costprivate interventions and to catastrophically costly private goods. These domains are constructed by classifying health-related activities along two dimensions, as shown in Figure 1: first by the degree to which they are private or public goods, and second, by how much they cost. Both dimensions refer only to characteristics of particular activities or interventions themselves, not to who consumes them or pays for them. As the descriptions of the three domains indicate, private goods are separated into low-cost and high-cost, whereas all public goods constitute one domain, regardless of their cost. The reason for this is that while the costs of public goods matter for deciding whether they should be produced, the issue of market failure related to such goods is independent of costs. With private goods, in contrast, some problems of market failure occur only with those services costly enough to be financed by insurance; and poverty, or the inability to buy even low-cost services, is a distinct reason for public intervention. This classification of health-related activities does not reserve a place for "merit goods", interventions which everyone "ought to have". To exclude this category is not to deny the social or political importance of views about what people "have a right to". The difficulty is that there is no good way to define such goods a priori, and societies make different choices about them. Moreover, so far as these goods are supposed to justify public intervention, there are often other grounds for state financing, mandating or regulating of the goods or services regarded as meritorious. For example, it is widely believed that all Figure 1 ThreeDomains of Health Care 6 Costofan intervention 5 2 Characterof anintervention - ~Public Private > = Publicgoodsor largeexternalities z Low-cost,mostlyprivate tILl Catastrophicallycostlyprivate Examplesof somespecificinterventions 1 Immunization 2 Vectorcontrol 3 Treatmentof tuberculosis 4 Treatmentof minortrauma 5 Normalobstetriccare 6 Surgeryforcancer 11 children have a right to immunization, but public promotion of immunization can also be justified by the market failures involved. A particular intervention occupies a small space in Figure 1; to indicate how different interventions would be classified into the three domains, some typical health care activities are located aproximately. The cost of an intervention can vary depending on many factors, such as how widespread the intervention is; there can also be variation in the public or private good nature of an activity. Because a health system produces a variety of interventions in all three domains, it is spread widely over the space. Figure 1 therefore does not serve to compare different systems. Public goods are goods or services such that one person's consumption does not reduce the amount available for others to consume. T'ypicallythese are goods from which consumers cannot be excluded: if they are made available to anyone, they are available to all, at least locally or temporarily. Since people can consume such goods without having to pay for them, no one will produce them for sale to individual consumers. Therefore they will be produced only if government (or some other source such as a charitable organization) pays for their production. The notion of a public good is no different in health than in any other sector: wherever such goods or services are to be available, they must be financed by government or some other non-market alternative. Control of disease vectors and protection of food and water safety are examples of (nearly) pure public goods in health. Individual action may be ineffective (if one's neighbor's house harbors rats or mosquitoes), costly (water purification) or virtually impossible (testing for food safety). Most activities in this category are preventive, but some curative actions are also partly public. And not all preventive interventions are public goods. For example, at low levels of coverage immunization confers some public good benefits-- because the immunization of part of the population reduces the likelihood that unimmunized people will become infected--but it still produces mostly private benefit for the immunized individuals. However, as immunization coverage approaches 100 percent, the benefit becomes more and more public through the mechanism of "herd immunity": a lone unimmunized individual would be just as well protected as if he had been immunized, and could enjoy this protection without paying for it. When the disease can be eradicated by complete immunization coverage, a pure public good is created. This is an example of how the public or private nature of an intervention may depend on the degree of coverage. When smallpox was endemic, individuals had a strong incentive to be vaccinated, without regard for how many other people were also protected. Now that the disease has been eliminated, everyone benefits. While the distinction between public and private goods is crucial, it does not by itself define the appropriate boundary between private and state action. Moreover, the boundary between public and private goods is not sharply defined, because some interventions provide substantial externalities. (Figure 1 treats such interventions as partly public and partly private, rather than locating them in a separate domain.) In these cases, individuals can and 12 do buy an interventionand benefitfrom it, but they cannotpreventnon-consumersfrom also derivingsomebenefit. Becausethe purchasersdo not captureallthe benefit,they maybe unwillingto pay for all of it: inconsequence,privatemarketscan existbut will produceless of these interventionsthan wouldbe optimalfor societyas a whole. This problemarises mostreadilywithcommunicablediseases,becausethe infected personputs othersat risk. Curingone case thereforealso preventsothers. Tuberculosis control is a clear example:no victimof tuberculosisis likelyto ignorethe disease, so there is no problemof peopleundervaluingthe privatebenefitsof treatment. Rather,the cost of treatment--andthe factthat theymayfeelbetter eventhoughthe diseasehas not been cured-- maylead peopleto abandontreatmentprematurely,withbad consequencesnot only for themselvesbut for others. The rest of societythereforehas an interestin treatingthose with tuberculosis,and assumingat leastpart of the cost. Asymptomaticcommunicablediseases, such as somesexuallytransmittedinfections,also createexternalities;but becausepeople maynot realizetheyare infected,the demandfor care is too lowevenwhencare is free (zero price). There is then an argumentnot onlyfor subsidizingtreatment,but for persuadingthose infectedto seekcare. Publicinterventionto ensureefficientprovisionof true publicgoodsand goodswith significantpositiveexternalitiesusuallyrequirespublicfinance,becausefree-ridingmakesit difficultto get the same resultsby other means. However,at the levelof small communities,it maybe possibleto encouragecollectiveactionwith onlylimitedpublic finance. In the extremecase, a communitycan internalizethe externalitiesand people may contributeeventhoughtheycould ride free, becausetheir immediateneighborsare involved and can apply moralsuasionor sanctions. The samerole can sometimesbe playedby a privateentrepreneur,whetherfor profitor not [Foldvary1994]. Where such smallgroup collaborationis not feasibleand publicfinanceis necessary,the chiefissue forpublic policy is whichactivitiesto promoteor pay for. The averageincomeof a countryobviouslylimits what it can spendon publicgoodsin health, but individualincomesare largelyirrelevant, becausepeopledo not buy the servicesand sometimesdo not evenchooseto consumethem. (Individualincomeis relevantfor manyprivategoodswithexternalities,for whichprivate demandwouldbe too lowbut not zero.) Most healthcare, however,is a (nearly)pureprivategood: Figure 1 reflectsthis by showingpublicgoodsas onlya narrowbandat the left-handside, withprivategoods occupyingmost of the space. Largelyor exclusivelyprivateactivitiesincludemost curative care--especiallyfor non-communicablediseaseswhichpose no threatto others--andall rehabilitativecare, and also somepreventiveor "pre-curative"care (suchas well-babyvisits, and screeningfor hypertension,cervicalcancer,or glaucoma). Theyincludehome treatment,usinghealth-specificpurchasedinputs,as wellas medicalor other professional care. This area shadesintothe myriadactivitiesof dailybehaviorwhichalso affecthealth, such as diet, exercise,safetyprecautions,sexualbehavior,and the use of alcohol, tobaccoor drugs. Amongthese activities,child-rearingis of crucial importance,both for the immediate effecton health--youngchildrenare especiallyvulnerableto infectionsand accidents--andfor 13 the formation of life-long habits. The health effects of these behaviors are usually small on a daily or episodic basis, but can be very large cumulatively. Figure 1 distinguishes among private interventions according as they are cheap or costly. This is not a sharp boundary, because what is affordable for some people is out of reach for others. And activities which are individually not very expensive may have to be repeated often, creating large cumulative expenditures: renal dialysis and physical therapy are examples. Nonetheless it is crucial to distinguish interventions according to whether they can be paid for out-of-pocket, or financed from accumulated savings, or are so expensive as to represent a catastrophic burden. The cost of medical care is catastrophic if a family or individual can meet it only by selling assets, or taking on debt, to such an extent as to leave it permanently poor. As indicated in Part 1, it is initially assumed that no one is too poor to pay for interventions in the "low-cost" domain. In any health systemthere is always some private out-of-pocket spending, corresponding to the band along the bottom of Figure 1. Before the modem understanding of disease and medicine, all health interventions were of this kind, and were paid for by consumers or by charity. Historically, health expenditure began in the lower right corner of the Figure and has expanded into public goods (to the left) and also into very costly private interventions (towardthe upper right) as knowledge, wealth and institutional capacity have increased. While this pattern is general, countries have followed different paths in the expansion. Interventions which are needed unpredictably, because disease strikes randomly, and are also too costly for households' ordinary budgets or savings to finance, define the domain of catastrophically expensive care. The only way to deal with the combination of high cost and uncertainty about needs is by risk-sharing, in which people finance health care collectively by contributions which are related to the expected expense in the group but not to any individual's (unknown) likely consumption of care. Although the boundaries are blurred, the domain of risk-sharing is conceptually quite distinct from the other two. It normally does not extend into the many routine low-cost, health-related activities, because risk is unimportant there. (It is true that famine relief and other responses to unforeseen disasters amount to sharing the risk of inadequate food or other daily needs, but under normal circumstances it is impossible to buy food insurance.) And explicit insurance--a contract between the consumer who agrees to contribute, and someone else who agrees to pay for specified care--is simply not feasible for public goods. Of course, when the government or a private charity pays for public goods, it assumes the financial risk. This can be thought of as implicit insurance, and in this sense all publicly-financed health care is a form of insurance, even if there is no explicit contract and no payment of individual claims. Risk-sharing presents the most numerous and complex issues for public policy. The growth in total health care costs is concentrated in this domain because it includes the 14 catastrophically expensive activities. It also shows the greatest variety of institutional arrangements, including substantial participation by private but non-profit providers and financing institutions [Frank and Salkever 1994]. The reason is that while insurance is the natural solution to the risk of needing costly interventions, private insurance markets tend to fail in ways that affect both efficiency and equity, and different institutions represent different partial solutions to those failures. These problems are quite distinct from the inability of markets to deliver public goods or to assure the right level of production of goods with significant externalities. The question is how and how far governments can and should try to correct for the failings of the insurance market, and whether public finance is necessary or whether other instruments can be substituted for it. Because neither economic theory nor common sense provides as much of an answer as in the domains of public goods and low-cost interventions [Diamond 1992, Zeckhauser 1994], judgements in the domain of risk-sharing are more tentative and depend more on empirical information. Intervention for Public Goods in Health By definition, public goods cannot be sold in private markets and so create a straightforward justification for collective action. However, the goodor servicealso mustbe worththe requiredpublicexpenditure:simplybeing a public goodis an insufficientconditionfor state intervention. Totake an extremeexample, erectinggiantfans to blowawaypollutedair wouldprovidea public good,but wouldalmost surely cost too muchto be justified. The questionsto answerthen are, whichgoodsare sufficientlypublic that privatemarketscannotprovidethemadequately? and how should they be valuedto determinewhetherit is justifiedto pay forthem ? As to the first question,governmentsusuallytry to providesuch indisputablypublic goodsas diseasesurveillanceand sanitaryinspection. Theyoftenerr, however,by tryingto covertoo wide a range of interventions;Ministriesof Healthsometimesappearto regardall of healthas a publicgood. An alternativeexplanationis that they regardall health care as a merit good; the beliefthat everyonehas an unlimitedor ill-definedright to free care is sometimesenshrinedin legislationor nationalconstitutions[Fuenzalida-Puelmaand Connor 1989]. Whenpublicfinancingis insufficientto fulfillthat promise,and particularlywhen publicprovisionis poorlymanaged,the result is likelyto be both inefficientand inequitable. Governmentsmay err in anotherway,by recognizingthat it canbe efficientfor the public sectorto supplya service--thealternativebeing to regulateprivateprovision--butsubsidizeit whenmostusers could pay for it. Watersupplyand sanitationservicesare goodexamples [WorldBank 1992a];theygeneratelarge publichealthbenefitsbut are nonethelessmostly privategoodsfor which non-poorconsumersare willingto pay. Still, the state is not alwayswrongwhenit treatsa largelyprivategood as if it were entirelypublic. Themost strikingexampleis immunization. Had it been leftto private marketsduring the lastfewdecades,it is inconceivablethattoday some80percentof the world'schildrenwouldbe immunizedagainstthe sixmajor vaccine-preventablechildhood diseases[Geoffardand Philipson1994]. Treatingthe ExpandedProgramof Immunizationas a publicgoodmadepossiblehighcoverageevenin verypoor countries[EPI 1993]--often higherthan in the UnitedStates,whichhas reliedmore on privatefinanceand provision[EPI 15 1995,Havemanand Wolfe1993]. This "mistake"doubtlessimposedsomecosts,in the form of public expenditurewhichwasunnecessarybecausesomepeoplewouldhavepaid for immunizationprivately,and in the distortionscausedbythe taxesto pay for the program. But such costsare negligiblein comparisonwiththe healthgains. Andthe public interventionin organizingand largelyfinancingthe EPI did not crowdout, but probably stimulated,much privateparticipationin boththe financingand the deliveryof vaccinations [vander Gaag 1995]. The secondquestionis howto valuea public-goodhealth intervention. This is the naturaldomainof cost-effectivenessand cost-benefitanalyses. Any valuationof effects involvesa seriesof subjectivechoicesas well as objectivemeasuresof results:whetherand howmuchto discountthe future; whetherto makedistinctionsamongbeneficiariesbyage, sex, incomeor other features;howto valuedifferentstatesof health;and howto add effects acrossbeneficiaries. There is no singleanswer;evenwhenthe calculationsare limitedto cost-effectivenessin producingdifferenthealth outcomes,withno allowancefor other benefitssuch as increasedproductivityand incomes,differentchoicesof theseparameters yield somewhatdifferentpriorities [Murray,Lopezand Jamison1994]. If the onlybenefitfroman interventionis improvedhealth,it does not matter whether that is measuredin health terms (livessaved,healthylifeyearsgained)or monetized. When there are also significantcollateralbenefits,differentapproachescan lead to different rankingsof interventions. Thisis the case for education,water supplyand sanitation,and other activitieswhichare valuablefor healthbut also for other reasons--andwhich maynot bejustifiedfor the healthbenefitsalone[WorldBank 1993b]. Cost-benefitor cost- effectivenesscomparisonsare also relevantto public interventionin the other domainsof healthcare, but theyare particularlyimportantforpublic goodsfor whichno privatemarket pricesexist. The issuesof which activitiesto considerpublic,and whethertheydeserveto be financedpublicly,are difficult. Nonetheless,there are severalreasons whythis is the simplestdomainin whichto determinepublicpolicy. There is broad agreementon the substantialbenefitsfrom a fewcrucial interventions,whichare extremelycost-effective [Jamison,Mosley,Meashamand Bobadilla1993]. Individualpovertyis not a major source of problems,as it is withprivategoods: onlysociety'soverallcapacityto pay matters. Individualignoranceor absenceof demandis also of little importance. Finally,this area does not contributemuch to the explosionof health care costs,and its financialimportance declinesas incomerises. The Public Rolein Low-CostPrivateInterventionsThisdomainincludesso many differentactivities,whichare undertakenrepeatedlyand usuallyhavelittle health impactper episode, that continued,universal,direct publicinterventionis simplyimpossible. Governmentscannotbe responsibleforeveryone'sdailylife, and canprobablycontribute mostby improvinghouseholds'capacitiesto lookafter their ownhealth. Promoting developmentgenerally--notonly increasedincomesbut more educationand accessto all 16 kinds of knowledge, goods and services--seems to be the best way to do this [World Bank 1993b, Chapter 2]. How far interference in people's ordinary behavior is justified, depends on whether the health benefits outweigh the curtailment or modification of individual choices, including non-health benefits. Apart from indefensible extreme positions--for example, that only health matters, and is worth any price; or that only people's private appreciation of their their own utility matters, and should be treated as sacrosanct--there is no straightforward answer to this question. Public action cannot be justified simply because of a health improvement; neither can it be rejected just because individual liberty might be limited. Specific public intervention for improved health may be justified under three conditions: ignorance or incomplete knowledge, externalities, and the failure of adults to act as appropriate agents for children. Each of these involves some kind of market failure, or violation of the private- market assumption that rational adults are making informed choices and paying the consequences of their decisions. The first problem is ignorance: people might take better care of their health if they knew how. For example, vitamins are crucial to health but are not observable in food, and people may already believe untruths about diet that help cause vitamin deficiencies [Johns, Booth and Kuhnlein 1992]. In general, ignorance on the part of one or both parties to a transaction is a major source of failure in the health care market. Of course, "perfect decision making is not ever possible, so the real issue is when the government can or ought to intervene in the information market to improve the market's performance" [Beales, Crasswell and Salop 1981]. Moreover, information is not entirely free, and people do not always act on it. Thus while the cost-effectivenessof efforts to make behavior more efficient can be very high, it is also quite variable, and people's reactions to information are hard to predict. Correcting ignorance is not simply a matter of telling people something new, but a larger question of changing beliefs and behavior. Pure cases of ignorance, where simply being told something is enough to affect how people behave, are probably rare. In fact, it is difficult to be sure when people "know" something relevant to their health: if they say they know it but do not change their behavior, they may be expressing an informed choice or they may claim to know what they have not really understood. Where better knowledge alone does not lead to changed behavior, regulation or mandates may also be justified even though they imply more intrusive or coercive intervention. In all such cases, the difficult question is how far it is legitimate to try to change people's views of what they want or what is good for them. Information often complements these other instruments, to reduce opposition to them or improve their effectiveness. The interaction between information and other instruments of behavior change is seen clearly in the successful effort to reduce smoking in the United States [U.S. Dept. of Health and Human Services 1989, 1992]. This situation raises a second problem, of externalities, or interactions among presumably informed adults. Driving while drunk is an example, as is dumping feces or 17 trash in communalwatersupplies. Theseactivitiesimposeboth healthdamageand financial costson others,and individualprotectionmaybe impossibleor verycostly. The chief instrumentfor publicaction is regulation,perhapssupportedby mandates;these instancesdo not typicallyrequirepublicfinanceof healthcare activities. They mayof coursealso require negativemandates,in the formof lawsagainstcertainactivitiesor behaviors. In practice, there is no sharpboundarybetweenthis and the first problem,becausesome of the behavior that imposescostson others mayalso arise from ignorance:thus reducingthe harm from a particular behaviormay requirebothinformationand monetary,legalor other incentives. It is more effectiveto criminalizedrunkdrivingif peopleare also informedof the dangers,and the health damagecan be limitedbymandatingthe use of seatbelts. The third conditionis an agent-principalproblem[Stiglitz1989],and in contrastto the externalitiesjust discussed,is intergenerational.Childrenare not yet informed,sovereign adults; they are vulnerablenotjust to accidentsand diseasebut to the indifferenceand even sadismof their parents. Thisproblemis somewhatsimilarto the situationof doctorsacting as imperfectagentsfor their patients. However,patientsoftencanchooseand contractwith the doctorswho act as their agents,whereaschildrenhaveno choice of who actsfor them. (Similarproblemsarise for adultswhoare mentallyretardedor incapacitatedby somekinds of disease.) Whatshouldthe statedo whenparentsare inadequateagentsfor their children? Requiringthat childrenbe immunizedis relativelyeasy,but it is harder to deal with child- beatingor exposureof childrento secondhandsmoke,and stillharder to confrontparents' beliefsin such mattersas sexualeducation. Where sexualbehavior,vehicleuse and consumptionof alcohol,tobaccoand drugsare concerned,these issuescontinuethrough adolescence. This is an exceptionallycontentioustopic,where it is hard to drawthe frontier betweenpublicand privateresponsibilities. Differentsocietieshaveadopteddifferent solutions,and there is oftenbitterdisagreementwithinsocietiesoverthe rightsand dutiesof parentsand the degreeto which the statecan or shouldinterferein familylife. There are potentiallyverylarge healthgainsat stakein this debate:eightor nineof the ten worldwide leadingcauses of illnessin youngchildrenare substantiallycorrectableat low cost [World Bank 1993,AnnexTableB.6], and four of these--diarrhealdiseaseplus three nutritional deficiencies--canbe largelycontrolledbythe family,with littlepublic expenditure. These diseasesaccountfor about 20 percentof youngchildren'sill health;the totalshare of child healththat dependson parentalbehavioris of course substantiallylarger. In all these instances,the principalinstrumentsof stateaction shouldprobablybe informationand regulation. Mandatesare justifiedfor a fewactivitiessuchas requiring schoolchildrento be immunized,or that foodstuffsbe fortified,and againsta fewother activities. Substantialpublic finance,however,is usuallyjustifiedonly becausesomepeople are too poor to pay for health-relatedgoodsand activities,whetherthese involvemedical care or such necessitiesas food. 18 Risk-Sharingfor CatastrophicallyCostlyPrivateGoods Whenrisks cannotbe fully controlled, and the associated costs may be catastrophic, the only solution is to share the risk. None of the features of this domain is unique to health care, but the magnitude and interaction of certain problems are especially important in health care markets. Moreover, the health risk is only partly associated with income or employment, and the financial risk is hardly associated at all with income or occupation. Health insurance differs sharply from insurance for nonhuman assets such as homes or vehicles, where the value of the asset, and therefore the cost of insurance, is usually related to income. Another fundamental difference is that medical care allows for preventive maintenance and for repair, but not for complete replacement of the damaged capital, which in this case is a human body. Insurance for nonhuman assets operates in just the opposite way, protecting against the loss of the asset but not paying for its upkeep. Insurance may even cover the cost of a temporary substitute for the lost home or vehicle, which is impossible with health insurance. Risks are also harder to estimate for health insurance, both because of the inherently much greater complexity of the body than of nonhuman property and because there are often different possible treatments for a given health problem, with different costs, outcomes and risks. Insurance against health risks raises some well-knowndifficulties [Arrow 1985], leading to various kinds of market failure. One such problem arises because insurance is a contract by which someone other than the patient agrees to pay for his or her health care. As with all contracts, there is an incentive for the insured to behave differently because of the insurance; this is called moral hazard [Pauly 1968]. One consequence is that consumers who do not pay the full cost of health care will consume more of it. This is desirable, since the point of insurance is to let people consume health care they could not otherwise afford. It means, however, that the price must cover the increased demand that results from insurance, and not simply the care that people would otherwise want to buy out-of-pocket. In theory, there are two potentially more worrisome problems associated with this moral hazard. The first is that people may not only consume more medical care generally, but care that costs too much relative to its effectiveness, yielding smaller health gains per dollar spent. The second risk is that people may take poorer care of their health via daily activities, because they pay the full cost of those, but only part or none of the cost of the resulting increased curative care. Both problems imply excess resources being dedicated to health care. They may also imply worse health, if increased curative care does not fully compensate for reduced prevention and protection. Some degree of moral hazard is intrinsic to all kinds of insurance, but it is more limited in the case of nonhuman assets because the insurance does not cover ordinary wear and tear. And cheating the insurer, by burning down one's house or abandoning one's car and reporting it stolen, is illegal, to prevent the insured person from fraudulently collecting cash. Such compensation is generally not possible under health insurance. (Cash payment for permanent disabilities is usually included with life insurance and represents compensation 19 for the lossof part of a life. The only significantmoralhazardfor such insuranceappearsto be suicide,which is often specificallyexcludedfromthe causesof deathfor which compensationwillbe paid.) Moral hazardin health insuranceis independentof how it is financed,so it does not by itselfdeterminewhether insuranceshouldbe paid for privatelyor publicly. A consumer who voluntarilybuys privateinsuranceends up payingfor the additionalmedicalcare consumedby other purchasers,andjudges whetherthis cost isjustifiedby his or her own greater accessto care. Wheninsuranceis paid forbytaxes or mandatorycontributions, however,this choicecannotbe made. Moral hazardmaythenjustify controlson whatpublic moneyis used for, to avoidexpenditureon interventionsof little healthvaluewhich consumerswouldnot voluntarilyagreeto buy for other people [Musgrove1995b]. There is scantempiricalevidenceon the importanceof theseproblems,particularlyas to whetherinsuranceleadspeopleto be morecarelessabouttheir healththan theywouldbe if uninsured. As to the relationbetweeninsuranceand less cost-effectivemedicalcare, the evidencein the UnitedStatesis that higherout-of-pocketcost formedicalcare (higher copaymentsor lowerdeductibleson insurance)does not makeconsumerschoosemore cost- effectiveservices,and mayevenmakepoor consumersforegohighlyjustifiedcare [Lohret. al. 1986, Newhouseet. al. 1993]. Exceptin the latter case, there is little evidencethat makingthe consumerpay highercostsunder insuranceleads to worsenedhealth. Inefficiencyin a competitiveinsurancemarketalso takesthe form of excesspurchases of insurance--thatis, insurancefor interventionswhichcould be more efficientlyfinancedout of pocket [Pauly 1974],or insurancewhichleadsto needlessor unjustifieduse of medical care. This is inefficientto the degreethat it leadsto excessadministrativecostsfor handling numeroussmallclaims,and becauseof the excessconsumptionof healthcare [Feldstein 1973]. Private insurerscanonlypartly controlthis tendencythroughdeductibles(which removesmallrisks fromcoverage,until out-of-pocketpaymentsreach somelimit). This problemarisespartly becauseof ignorance:peopletend to overestimatesmallrisks and may buy too much insuranceevenwhentheypay its fullcost. Moral hazard, however,is a greaterproblem:peoplewho do not pay the fullcost of insurancewill buy too much of it, just as with medicalservices. Market failurein the formof over-insurancehappensprimarilythroughthe tax system. Manygovernmentsallowprivateemployersto treat insurancecost as an expense, but then--incontrastto salaries--donot treatthe valueof insuranceas incometo workers. Subsidythroughthe tax systemis notoriousin the UnitedStates[Pauly1986],both for insurancefor workerswhich is financedbyemployersand forpart of the Medicareinsurance for the elderly. It is estimatedthat employer-financedinsurancewoulddeclineby one-sixth or more in the absenceof this subsidy,and that in consequencethe overalldemandfor medicalserviceswouldfallby aboutfivepercent[Chernick,Holmerand Weienberg1987]. Alternatively,governmentsdirectlysubsidizesocialsecurityhealth benefits(mandated insurance)out of generalrevenue. Generalrevenuesare used to supportsocialsecurity 20 systemsthroughoutEuropeand LatinAmerica [McGreevey1990]. In Chile, payrolltaxes can also be used to financeprivateinsurance. All thesedirectand indirectsubsidiesto insuranceare not onlyinefficient,but highlyinequitablewhenonlypart of the labor force is covered. The poor typicallybenefitonlywhencoverageis (closeto) universal. Whenvariousprofit-maximizinginsurerscompeteto sell insurance,there are two furtherand closelyrelatedproblems,of adverse selection onthe part of consumersand of risk selection on the part of insurers[Arrow1985,Milgromand Roberts1992]. The former refersto selectionof customerswhichwouldbe adverseto the interestsof insurers-- fundamentally,it describesthe dangerof enrollingpeoplewho wouldcost more on average than the insurancecould finance. Thiscan happenbecausethe amountof insurancecoverage people wantand are willingto pay for dependspartly ontheir knowledgeof their ownhealth conditionsand risks. Peoplewho expectto need littlehealthcare are unwillingto pay as much as those who expectto needmuch care, so a policycostlyenoughto coverhigh-risk people will lose out in the marketto a cheaperpolicyadequatefor low-riskpeople. Universalcoverageat the sameprice foreveryonemaythereforebe impossibleto achieve,or may not generateenoughrevenueto financeall the healthcare demanded[Summers1989]. Toprotectthemselvesagainstthe combinationof lowpremiumsand highpotential costs, insurersengagein risk selectionor "cream-skimming": theyspendmore on administration,or createbarriersto enrollment,to screenout high-riskindividuals(suchas the aged)or conditions(suchas cancer). Such"underwriting",as it is called, is particularly costlyfor individualapplicantsfor insurance,and gives rise to largescale economiesbecause whena largegroup is enrolled,the insurerneedsto estimateonlythe averagerisk of needing care [Diamond1992]. Thispracticeis the naturalmarketresponseto the problemof adverse selection. Inefficiencytakesthe formof increasedadministrativecosts,and also increased health risks for thoseexcludedfrom insurance. Particularlywhenpre-existingconditionsare not covered,peoplewithhealthproblemswho are insuredbytheir employerscannotreadily changejobs withoutlosingtheir insurance;other differencesin insurancecoveragemayalso create "job lock" amongworkers. Thislabor immobilityis anothersource of inefficiency,of unknownmagnitude[CongressionalBudgetOffice1994]. One answerto the problemsof adverseand risk selectionis price differentiation accordingto risk, which is theoreticallyefficientin that it allowseveryoneto havethe insurancehe or she is willingto payfor. Suchprice variationis commonto other formsof insurance:for example,ratesfor automobileinsuranceoftenvaryby ageand by the waya vehicleis used. Unfortunately,there are seriousdifficultieswith lettingthe marketcreate comparabledifferentialsin health insurance. One is that somepeopleare willingto pay only a smallamount,becausetheyexpectto needlittle medicalcare. Facedwitha price that wouldcover the cost of care for everyone,theywill not purchaseinsurance. Whenthey drop out of the market, the price of coveragefor those whoexpectto need more care is drivenup becausethe risk is spread overfewerand higher-costpeople. Evenif they are willingto pay more thanthose who anticipateneedingvery littlecare, the price of insurance may rise beyondtheir capacitytopay. Theywill be unableto buy insurance,despitea 21 willingnessto pay more thanthe averageconsumer. Suchfailuresdo not occur in other insurancemarketsbecauserisksare more uniformor predictable,or more closelyrelatedto income. Of course, people'swillingnessto buy insurancedependson their expectationabout futureneeds for medicalcare, and theymayguess wrong. Peoplewhoare youngand healthytoday,and thereforeunwillingto spendmuch on insurance,maywhenthey are older wantmuch more medicalcare than theynowanticipate. But if theybecomewillingto buy substantialcoverageonlylate in life,the cost will be higherthan if it were spreadovera longerperiod, so they maybe unable to pay for insuranceoncetheyrecognizethe need for it. The difficultyof predictinghealthcare needs is exacerbatedbythe rapidityof technical changein this sector[Weisbrod1991]. Anotherproblemwithdifferentialprices is that despitethe importanceof many behaviorsfor specifichealthproblems,ratherlittle of totalhealth risk is under the individual'scontrol. Peoplecannotbe held personallyresponsiblefor much of their ill health since it is geneticin origin, or due to the actionsof others. Oftenthe best that people can do bycontrollingtheir ownbehavioris to postponeproblems,whichis veryvaluablebut does not necessarilysavemoneyovera lifetime[Russell1986]. Behavioralchangemayalso take a longtime to showeffecton the burdenof diseaseor the volumeof treatment[World Bank 1994a]. Becauseso little healthrisk is under people'scontrol,behavior-relatedprices-- whether forhealth care or for insuranceitself--areof onlylimitedvaluein makingmarkets work. One canchargepeoplefor smoking(bytaxation)or for not wearingseat belts (by fines), or rewardthemfor carefuldriving(bylowerinsurancerates), but prices are not feasiblefor most health-relatedbehaviors. Peoplehavesomechoiceof where to live and whetherto drive a car, but no choiceabout inhabitingthe body theywereborn with [Miller 1978]. 1 Notionsof fairnessare involvedin the choiceof howfar to allowor controlprice differentiationin healthinsurance,becausepeopleoftendo not thinkothers shouldbe punishedfinancially,in additionto their physicalsuffering,for bad luck. In addition, the possibilityof deathor substantialpermanentdisabilitysometimesmakestreatmenturgent. Adverseselectionis thereforea problemof equityas wellas of efficiency. The problemsof moralhazardand adverseselectionarise partly fromthe fact that consumers and insurers possess only incomplete information, which causes market failure in the sensethat marketswork perfectlyonlywhenboth buyersand sellerspossessfull information[Arrow1985]. Of course, insuranceis wantedin the first placebecausepeople do not knowwhat will happento their health, and theyagreeto sharerisks whentheydo not knowwhat will happento others'health. A furthercomplicationis that of information asymmetry:informationavailableto onlyone side of a marketreadilyleadsto marketfailure. For example,consumerswhoknowtheir health riskshavean incentiveto concealthemfrom insurersso as to avoidhigherpremiums. Theyalso knowhowtheyhavemodifiedtheir behavior,or mean to do so, becauseof insurance. Insurers,in contrast,generallyknow 22 more than consumersabout averagerisksand aboutcostsof care; consumerignoranceof these matterscan also lead to inefficiency. Unfortunately,it does not followthat the problemsof incompleteand asymmetric infonnationcouldbe correctedjust bysupplyinginformation. Betterknowledgeon the part of consumersabout health risksmaylead to more efficientpurchaseand use of insurance. However,obtainingthe informationneededto restoresymmetrywouldbe impossibleor very costly;too much is stillunknownabouthowmuchpeoplecan controltheir health through behavioralchoices. Evenif it weresymmetricallyavailableto consumers,insurersand providers,more informationmightmakeit easier for insurersto practicerisk selectionand discriminateamongcustomers,and therebyexacerbateinequity. In fact, in an unregulated market, this is the probableconsequenceof the increasingavailabilityof informationlinking geneticendowmentto the likelihoodof developingspecificillnessesor health problems [Houseof Commons1995]. Informationasymmetryalso arises betweenpatientsand doctors,since the latter typicallyknowmuch more aboutmedicalconditionsand treatments. Patientsmayaccept, or evendemand,treatmentstheywouldnot buy if fullyinformed,but which are advantageous, financiallyor otherwise,to medicalprofessionals. Thereis howeverlittle firmevidenceas to howmuch of this potential"supplier-induceddemand"actuallyoccurs [Pauly1988]. In any case, this is not simplya problemof rich countries,wheremostpeople haveinsurance and thereforedo not worry aboutcosts; it has also been documentedin poor societieswhere lack of educationand informationmaymakeit particularlyeasyto exploitconsumers [Bennettet. al. 1994]. In summary,the consequenceof thesemarketfailuresis that in anunregulated, competitiveprivatemarketin third-partyinsurancethosewithchronic conditionsor high health risks will be under-insured,administrativecostswillbe higherthan necessarybecause of insurers' effortsto screenout risksand the costsof processingclaims in a marketwith many insurersand manyproviders,and proceduresof lowor questionablevaluewill be performedbecauseneitherthe providernor the consumerpaysfor them. It is in these specificsensesthat "themarketdoes not work"in healthcare; these are primarilyfailuresof the insurancemarketrather than shortcomingsof the marketfor health care itself. Privatemarketshavedevelopedother formsof insurancewhichreduce, but do not eliminate,theseproblems,such as healthmaintenanceorganizations(HMOs). Under this arrangement,providersalso act as insurersand assumethe risk. Insofaras this controls costsbyshiftingthe burdento suppliersof medicalcare ratherthan to consumersor third- party insurers[Ellisand McGuire 1993],it mayallowmore coverageof the chronicallyill, and may reducethe utilizationof relativelyineffectiveprocedures. However,withoutsome formof publicinterventionsucharrangementswill havelittleeffecton the problemscaused bypovertyand adverseselection. The questionremainswhichkinds and degreesof public interventioncan best mitigatethe problemsinherentto privateinsurancemarketswithout introducingworseinefficienciesor inequities. 23 MarketFailureandHealthCareNeeds Since the unregulated,unsubsidizedprivatemarketis the extremealternativeto governmentinterventionin healthcare, much of the debateas to appropriatepublic and privateroles in the sectorturnson whether,how,and howbadly marketsmayfail. Market failure,as an economicnotion,refersto possiblemismatchesor disequilibriabetweenwhat the marketsupplies,and whatfully-informed,rationalconsumersof healthcare would demand. It does not deal withthe conceptof needfor healthcare, which is theoreticallyan unsatisfactoryconceptbut is also difficultto do without[Culyer1995]. Peoplewanthealth care not for any intrinsicutilitybut becausetheythink theyneed it, that if care is not providedtheir healthwill deteriorateor fail to improve. In contrast,much of the criticism byboth healthcare professionalsand consumersof howhealth systemsoperatedeals explicitlywithneeds. Justas demandand supplymaybe out of balance,there can be imbalancebetween demandand need or betweenneedand supplyof services,as shownin Figure 2. Market failuresin the narroweconomicsenseare amongthe reasonsfor these imbalances(theseare indicatedon the Figureby asterisks). Somefailuresresultfrombarriers to the operationof competitiveprivatemarketsin bringingsupplyand demandtogether. Others distortdemand fromwhat it wouldbe if basedon completeand symmetricinformationand if there wereno publicgoodsor externalities;this causes imbalancebetweendemandsand needs. While competitiveprivatemarketsare generallythe best wayto bring demandand supplytogether, theyare much worsesuitedto makeeither demandor supplymatchpeople'sneeds. Public interventionin the healthmarket, in contrast,is aimedat satisfyingthoseneeds, and runs the correspondingrisk of failingto take accountof demand. Either a purelyprivateor a purely public healthcare systemis likelytocontrol one of the three potentialimbalances,at the cost of failingto control or evenworseningone or both of the others [Musgrove1995c]. This is a major reasonwhy mosthealthcare systemsare far frombeing all privateor all public [Dunlopand Martins 1995]. Dealing with Pbverty In the discussionthus far, it has beenassumedthat no one is too poor to buy a varietyof health interventionsout of pocket. Similarly,it is assumedthat no one is too poor to buy insuranceagainstcatastrophichealthrisks; peopledifferin incomeand in their assessmentof risks, but everyonecan affordsomeinsurance. Thismeansthat the distinction betweenthe "low-cost"domainin Figure 1 and the domainwhere insuranceis neededis roughlythe same for everyone. A Minimal State Role in the Absenceof Poverty Undertheseconditionsthe state'srole in the domainof private,inexpensivehealth-relatedactivitieswouldbe limitedmostlyto informationand regulation;there wouldbe no reasonto financethis kindof health care publiclyif everyonecouldaffordit out-of-pocket.(Mandatesmightstill bejustifiedto deal with some of the externalitiesmentionedabove.) In the domainof risk-sharing,people who Flgure 2 Need,Demandand Supply for Health Care NEED / \ ~Poverty Highcost Highcost Lackof informationC') MarketIncentvies SupieInduced demand() Externalities(9 Lackontmnation of ) ubli goods(7 SUPPLY DEMAND Lackof InformationC) Otherbarriersto competition(i) Non-marketIncentives () Sourcesof marketfailure Source:Musgrove(1995b). 25 chose not to buy insurance,or boughttoo little of it, wouldhaveto pay forcare out-of- pocketor do withoutit. Becausedoingwithoutcare wouldsometimespose the risk of avoidabledeath, some kindsof care--emergencyservices,at least--aretypicallyavailableevento the uninsured. People'swillingnessto let otherssufferthe consequencesof imprudencedoes not usually includelettingthemdie becausetheyboughttoo little catastrophicprotection. Thiskind of imprudenceconstitutesmoral hazard, andunlike somebehavioron the part of insured people, it cannotbe dealtwith bydifferentialpremiums. Motorcyclistswho prefernot to wearhelmetscan in principlebe chargedmorefor their insurance,just as smokerscan; the more difficultproblemis howmuch care to providefor the uninsuredcyclistwhoseinjuries are worsebecauseof failureto weara helmet. This situationprovidesa justificationfor enoughpublicfinanceor mandatedinsurance to coverthe cost of a few crucialservicesto whicheveryonewouldhaveaccessand for whicheveryonewouldhaveto pay throughtaxes [Summers1989]. Exceptfor these services,there wouldbe no requirementfor the governmentto subsidizeinsurancefor anyone. If there wereno poverty,then, the role of the statein the health sectormightbe relativelylimited,and would--exceptfor the minimuminsurancerequirementjust described-- concentrateon the adequateprovisionof public goodsand the correctionof marketfailure in the domainof risk-sharing. Whetherin order to corrector compensatefor that failurethe stateshouldmandateor financeinsurancebeyondthat minimumof emergencycare, is a questionof the socialefficiencyof doingso, rather thanleavinginsuranceto the private market, as discussedbelowin Part 3. Thefact that insurancewouldnot needto be subsidized--theremovalof an equityjustificationfor interferingin the insurancemarket--does not meanthat governmentsshoulddo nothing,becausethe efficiencyfailingsof private voluntaryinsuranceare as importantas the inequitiesto which it givesrise. How Pbverty ComplicatesPublic Roles The existenceof poverty,of peopletoo poor to buy many "inexpensive"healthactivitiesor an "adequate"amountof insurance,complicates the questionof whatthe stateshouldand shouldnot do in severalways. Thesecomplications are not limitedto the domainsof privategoods,becauseproblemsof publichealth are often more severeamongthe poor. Theyare likelyto be at particularrisk fromcontaminatedair and waterand so to benefitmore thanthe non-poorfrom publichealthinterventions[World Bank 1992a,1993b],and theyoftensuffermore seriousconsequencesfromcommon illnesses. In general, imbalancebetweenneedand demandmaybe more importantwherethe poor are concerned,becausetheyhaveless knowledgeon whichto base their wantsfor health care as wellas less resourceswith whichto expressdemand. Medicalindigenceis inmost respectsno differentfrom povertywithrespect to food and otherbasic needs, and, as withthoseneeds, the rest of societymayagreeto subsidize the poor. The differenceis that povertyrelativeto predictable,low-costneeds such as food can be dealt witheither bytransfersor subsidiesin kind, or bysupplementingincome [Srinivasan1994]. Withhealth,the risk of needingvery costlycare generallymakesit more 26 efficientto deal withmedicalindigencebysubsidizinginsurancethan throughincome transfers. However,there is little experiencein mostpoor countriesin subsidizingprivate providersor insurersto meetthe healthneeds of the poor. This requiresgovernment administrativecapacityand appropriatepricingmechanisms,to preventexcessprovisionand evenoutrightfraud, whichhas been a majorproblem,for example,in Brazil [WorldBank 1993c;Medici and Czapski 1995]. A public subsidyto privateinsurance,as an alternativeto dealingdirectlywith largenumbersof providers,also requirespremiumsdifferentiatedby age, sex or other conditions,to reducethe scopefor risk selectionand make it feasibleto mandateuniversalcoveerage. In consequence,one of the mostimportanteffectsof poverty is that it makespublicprovision,with all its typicalproblems,lookattractiveor even necessaryin poor countries. Asidefromthe problemsof regulation,it is financiallydifficultto providethe poor withthe samelevelof servicesenjoyedbythosealreadycoveredbyprivateor social insurance. Evenextendingsocialsecuritycoverageto the poor, to replacethe more limited servicesofferedby Ministriesof Health, wouldbe verycostlyin many LatinAmerican countries[Mesa-Lago,1992]. Publicprovisionoftenmeanspoor healthcare for the poor, but public financingof privateserviceswouldnot easilysolvethe underlyingfinancial problemeven if it led to improvedqualityof care. Finally,in countrieswhere all publicmoneynowflowsthroughgovernmentor parastatalfacilities,shiftingto publicfinanceof privateprovidersrequiresthat public hospitalsand clinicsbe privatizedor at least givensufficientautonomyand capacityto managethemselvesand competeforpublic fundsagainstotherproviders. Suchchangesare potentiallyvery valuable,but theyare likelyto be particularlydifficult,sincepublic facilities need to be exposedto somefinancialrisk withoutthe dangerof collapsein publiclyfunded provision. The difficultyof incorporatingthe poor intothe sameinsuranceschemeswhichcover the non-poor,whetherby extendingsocialsecuritycoverageor by subsidizingthe purchase of privateinsurance,leadsto effortsto createinsurancespecificallyfor the poor, typicallyat the communitylevel. Anysuch schemeis intrinsicallylimitedby the low incomesof participants,so it cannotfinancevery costlyinterventionsand canonly yieldsubsidiesfrom the less poor to the more poor. It maynonethelessbe appropriatewhenthe insuranceis meant to pay for only suchhealth care as can alsobe providedlocallyand which thereforeis not verycostly--althoughperhapsstillcatastrophicfor a poor familyto finance. Unfortunately,the problemsof moralhazardand adverseselectionarise evenin these circumstances. Forexample,if insuranceis soldfor shortperiodsto accomodatefamilies' fluctuationsin income,then, as occurredin Burundi,peoplemaybuy the "healthcards" entitlingthemto servicesonlywhenthey are alreadysick or can anticipatea medicalneed [McPake,Hansonand Mills 1993]. That effectivelyeliminatesthe differencebetweenan 27 insurancepaymentand a fee, and reducesthe amountof moneythat can be raisedby the scheme. Theseproblemslead to complicationssuch as rewardingpeoplewho use less (curative)healthcare byreducingthe cost of their cardsfor the nextperiodor by charging an additionalfee or "fine"to thosewho payfor insuranceonlywhen ill [Chabot,Boaland da Silva 1991]. Suchincentiveswork againstmoralhazardand adverseselection,but if they are large enoughto havemuch effecttheymaygreatlyreducethe scheme'srevenueor the demandfor services. Andadministrativeexpensesmayabsorba largeshare of revenue. Povertyalso createsor strengthensreasonsfor the stateto intervenein low-cost, health-relatedactivities,whetherthese are inexpensivemedicationsand servicesor such non- medicalitemsas foodsupplements. Some interventionscanbe accommodatedbybroadening publichealth services,forexampleby includingmicronutrientsupplementsor treatmentfor intestinalparasites. Otherscan be coveredbyfinancingprivateproviders,such as the clinics which operateunder "covenants"withthe Braziliansocialsecuritysystem[WorldBank 1993c]. Howevertheseactivitiesare dealtwith, povertypushesgovernmentsto financea wider rangeof low-costinterventionsand to relyless on informationand regulation. If this werethe only forceat work, it wouldlead to a largerpublicshareof health expenditurein poor than in rich countries, simplybecausethere are morepoor peoplewho cannotpay for those interventions. In the sum of healthspending,however,this effectis overwhelmedby the tendency(andthe capacity)of governmentsto mandateor financemore insurancefor the non-poor,as incomerises. A third effectof povertyis to limitthe use of pricesto curtaildemandor control costs. Beingpoor alreadygreatlyconstrainsdemand,and poor peopleare necessarilymore sensitiveto prices forhealth care than the non-poor [Gertlerand vander Gaag 1992]. This meansboth thatuser fees can raise relativelylittlerevenuefrom the poor, and thatunless there are offsettingimprovementsin quality,utilizationmaybe sharplyreduced[Lavy1994, Litvackand Bodart 1993]. The experiencewithuser fees has been extensivelyanalyzed, notablyin Sub-SaharanAfricain connectionwiththe BamakoInitiative[Griffin1987,Creese 1991,WorldBank 1992,Vogel1993,Makinenand Raney1994,Nolanand Turbat 1994, Shawand Griffin1995]. Thereis evidencethat--asmightbe expected--utilizationdeclines, sometimessharply,if fees are raisedbut nothingelse changes. Thereis rather less informationon servicecharacteristics,suchas whetheruser fees improvethe availabilityof drugs. And almostnothingis knownabout the impacton healthoutcomesor on system efficiencyor cost-effectiveness.Fees are sometimesset arbitrarily or with inconsistent criteria; chargeslowenoughto haveno effecton the poor mayor maynot be worth collecting;and targetingby exemptingthe destitutefrom fees does not haveto be expensive [Grosh1992], but there is a risk of high administrativecostsand lownet revenues. The sameproblemsarise for collectinginsurancedeductiblesand copaymentsfromthe poor. Not onlydoes povertyincreasethe risk of ill health; sicknessand disabilitycan make or keep peoplepoor. Therelationbetweenhealth and povertyis sometimesregardedas anotherreason for the stateto investin health,in order to raise productivity. However,the fact that some health care increases incomes is not a separate objective for government 28 action. If health care madepeopleso muchmore productivethat the extra incomecould pay for the healthcare, then in perfectmarketspeoplecouldborrowagainsttheir future productivity. Whencapitalmarketfailurespreventsuch borrowing,and those failurescannot be correcteddirectly,then anypublic intervention--suchas financingthe health care or providingloansto consumers--thatsecuresthe healthgainswill also yieldthe increasein productivity. Summary: Justifications and Risks of State Intervention Asthe foregoinganalysisshows,there are three distinct, independentargumentsfor governmentsto intervenein healthcare rather thanleavingit entirelyto privatemarkets. One is to ensure the optimallevelof productionand consumptionof public goodsand goods whichhavea partlypubliccharacterbecauseof externalities. Thesecan be health care servicesthemselves,activitiesprotectiveof health,or informationthat helps peopletake better care of their healthand makebetteruse of services. A secondreason is to make insuranceworkmore efficientlyand more equitably,for thoseserviceswhichcan be producedin privatemarketsbut for whichrisk-sharingis requiredbecauseof highcostsand uncertaintyabout needs. Thethird reason is to subsidizethosetoo poor to buy insuranceor even, sometimes,those inexpensiveactivitesand serviceswhichthe non-poorcan affordout- of-pocket. Thesethree reasonsderivefromthe three domainsof health care definedby cost and bythe publicor privatenatureof services. Market failuresunderlietwo of these reasons,but in differentways. In the case of publicgoodsand externalities,the failurearises fromthe natureof the goodor service. In contrast,problemsin insurancemarketsarise from the waythe goodis financed. Noneof the three reasons is uniqueto the healthsector,but all are more importantin this sectorthan in much of the rest of the economy. Communicablediseasesgeneratemajorexternalitiesand evensomepure publicgoods; healthinsuranceis more complexthan insurancefor nonhuman assets; and povertyis both a cause and a consequenceof ill health, in the absenceof which the appropriatestate role in the sectormightbe smallerthan it is whenpovertyis widespread. Finally,agent-principalproblemsbetweenpatientsand healthcare providers keep marketsfromworkingperfectlywhere individualclinicalcare is concerned,whether or not the interventionsare costlyenoughto requireinsurance. Third-partypaymentsimply worsensthe inefficienciesgeneratedby incompleteand asymmetricinformation. The argumentsfor not leavinghealthcare and health insuranceto uncontrolledprivate marketsare all argumentsthat efficiencyor equitycan be improved,if the state intervenes appropriately. They are not argumentsthat anythingthe public sectordoes, will improve matters. Justas there is a well-definedset of marketfailurestypicalof the health sector, there are consistentgovernmentfailures,waysin whichgovernmentsact to createworse outcomesthan could be reached, and in somerespectsevenworseoutcomesthan markets wouldgenerate. The mostcommonand severecriticismof publicactionconcernsprovision [WorldBank 1980and 1987,Birdsalland James 1992]:especiallyin poor countries, governmentsoffermedicalcare whichis supposedto be freeto users, on equitygrounds,but 29 whichis centrally-controlled,under-financedand of poor qualityin both medicaland human terms. Becausethe budgetsof publicfacilitiesoftenare unrelatedto service output, and civil servicerules makeit difficultto fire, transferor disciplineunproductivestaff,the costsof health gainsmaybe veryhighevenif salariesand other inputcostsare low. And the pervasivelack of incentivesfor efficiencymeanthat capitalis also boughtin excess,not maintained,and under-utilized. Theresult is that evenratherpoor people, the supposed beneficiariesof the public system,oftenpay out of pocket forthoseprivateservicesthey can afford. This makesthempay twicefor someof their care, exacerbatingthe inequitiesarising from the tax systemand fromdifficultiesof accessdue to the geographiclocationof facilities. Governmentstypicallyfailwhere provisionis concerned,bytryingto do too much and by competingwithprivateprovidersonly in price terms--thatis, subsidizingprovision rather than competingonqualityand satisfaction. Withrespectto the other instrumentsof state action,failuresare more varied,and oftenresult fromdoing too little rather than too much. This is likelyto be the caseparticularlyfor regulationand for the disseminationof information. Mandatesshowa very mixedpattern: middle-incomecountriesin particular oftenmandateinsuranceforpart of the populationthroughsocialsecurityschemes,but do not effectivelymandateeither insuranceor care for everyone. Richer countries, in contrast, appearmuch lessprone to governmentfailurelargelybecausetheyrely much more heavily on regulationand mandates,and much less on publicprovision. Wherepublic facilitiesare important,as in some Europeancountries,they operateunder greaterautonomythan in poor countries, and this is balancedbygreaterregulationof privateproviders. The result is to concentratemore onthe right rolesfor publicaction, and less ondividingthe health sector into disjointprivateandpublic spheres. Thedistinctionis particularlyimportantbecausein many countries,the two sectorsoverlapgreatly:the sameprofessionalsworkpart-time in each, privateprovidersoftenuse publicfacilitiesto treatprivatepatients,and so on. Toprovidepublicgoodsand to subsidizehealth-relatedactivitiesforthe poor, twoof the three mainreasonsfor state action,both requirepublicfinance. In both these areas there is also room for the other instrumentsof stateaction;and the problemsassociatedwithrisk- sharingcan lead to variouscombinationsof interventions,whichmayor maynot include spendingpublicmoney. Societiesthereforehavemuch latitudein howmuch, and by what means, the governmentintervenesin healthcare markets,just as theyhavein decidinghow much to spendon health in relationto incomeand to their healthproblemsor needs. Part 3 takes up the empiricalquestionsof whatchoicesdifferentcountrieshavemade,why they seemto havedone so, and with whatconsequencesfor health. 30 PART3: EMPIRICALPATTERNSANDEXPLANATIONS To saythat governmentsmayor shouldintervenein health "tomakemarketswork better" is a reasondrawnfromeconomictheory,ratherthan frommedicineor publichealth. Eventhe argumentfor interventionto alleviateor reducepovertyis largelyeconomic:it refersonlyto somepeople'sincapacityto pay for services,or insurance,that are considered a desirableminimnumfor everyoneto have. Societies'choicesaboutwhat kindof health systemto have,how much to spendon it, and who shouldpay for it, in contrast, are not primarilychoicesabout marketfailureor aboutpoverty. They referto objectivesthat mix economic,medicaland socialconsiderations,and mayor maynot lead to the most appropriatekind and degreeof publicinterventionin health. Theseobjectivescan be characterizedin variousways[OECD 1995];one classificationis summarizedbrieflybelow, beforeproceedingto an empiricalexaminationof the leveland compositionof healthcare spendingand its relationto healthoutcomes. Part 3 concludeswitha discussionof the reasonsfor the observedpatternsinhealth expenditureand health systemorganization. The empiricaldata referredto inthe text, and fromwhichFigures3-9 are constructed,are reportedin AnnexTable1, unlessotherwiseindicated. The number of countriesfor whichdata are availablevariesgreatlyfromone variableor Figure to another, with 69 observationsfor the largestnumberof analyses. For someor all of the OECD countries, severalsourcesprovidebrief descriptionsof howthe healthsectoris organized [OECD 1992and 1995, Schieber,Pouillierand Greenwald1992,Dunlopand Martins 1995]. Institutionaldescriptionsfor someLatin Americanand Caribbeancountriesare found in [Pan AmericanHealth Organization1994],for Asiancountriesin [Griffin1992]and for Africain [Shawand Griffin1995]. EconomicIssues andHealthSystemObjectives Whatpeople mostobviouslywantfrom a health systemis goodhealth. Peopleseek care for the reliefof painor disability,to reducefear and shameconcerningactualor potentialmedicalproblems[Miller 1978,Chapter 1], for informationand for comfort--butall these reasonsare relatedto somedimensionof goodhealth. It does not matterthat people mayhavedifferentviewsof the intrinsicawfulnessof being sick or hurt in variousways; withoutthe goal of better health statusthere wouldbe no healthcare system. A secondobjectiveis low cost, whetherthat of medicalcare, protectiveand preventiveactivities,administrationof insuranceor any other resourceuse. Low cost has both a macroeconomicinterpretation--spendingno more than is reasonableon health care in total--anda microeconomicinterpretation--buyingeach interventionat the lowestcost consistentwith adequatequality. The objectiveof holdingdowncost in boththese sensesis clearlyimportantfor individualspayingfor care out-of-pocket.As insurancereducesthe importanceof costsfor individuals,this becomesmoreof a socialthan a personalgoal, with control of costsbecominga majorconcernof privateinsurers. And as the public share of health spendingincreases,holdingdownexpenditurebecomesan objectiveof governments. 31 This begins as a goal of controlling public expenditure, but may quickly encompass the objective of holding down costs to employers and consumers as well. In fact, if the only goal were only to reduce or contain government spending, it might be met by shifting costs to the private sector--with the&possible result of worsening equity, satisfaction or health outcomes. This is equally true of the costs of other kinds of government intervention in health care markets, particularly regulation and mandates. A third general objective is satisfaction on the part of the consumer, patient or citizen. "Satisfaction" may mean slightly different things; in particular, there is a difference between being satisfied with how one is treated when sick, and how one feels about the system while well. The notion of satisfaction includes such intangibles as freedom to choose one's doctor or treatment, and all the elements of "caring" as distinct from "curing" in health interventions. This objective is often ignored under subsidized public provision, because people will use even unsatisfactory services if they are free, but dissatisfaction undermines public support and can reduce the effectiveness of public health efforts. The satisfaction of health care providers is also important, and sometimes conflicts with that of customers. People are often said to want "quality" health care, but this is not treated here as a separate objective. When the term refers to medical quality, to whether the care received is effective in improving health with the minimum risk or damage to the patient, then quality is essential to meeting the objective of good health. And when "quality" refers to all the non- medical aspects of care, it is a major determinant of patient or consumer satisfaction. There is similarly no need to consider either macroeconomic or microeconomic "efficiency" a separate goal, since it is defined by the relation between the health gains a system produces and what it costs to achieve them, in total or for individual interventions. A fourth overall objective is equity, which may refer to access to health care or to utilization of services. It is harder to define equity relative to health status, because of genetic and behavioral contributions having nothing to do with health services. Equity is also a question of fairness in how health care is paid for--how much subsidy (if any) from the rich to the poor, how much from the healthy to the sick, how much from the old to the young or vice-versa. All these comparisons are often made across income levels, leading to at least two income-related notions of equity: whether richer people use or benefit from health services more, and whether they pay proportionately to their incomes. To make it still more difficult to judge the overall equity of a health system, what is equitable with respect to health care may or may not correspond to equity in financing. And equity along any one of these dimensions can vary from one disease or treatment to another [Musgrove 1986]. These objectives are only some of the many criteria by which health systems might be judged [Cumper 1991], and it is not easy to say how well any system "works". The goal of better health is the most important, in the sense that what a systemcosts, how well it satisfies people and how equitable it is, matter only if it also produces health improvements. A system that worked well in those other senses but did nothing to better people's health status would have no reason to exist. 32 The various objectives are also commonly in confict [Weisbrod 1991], with particularly hard choices to make between equity and efficiency [Okun 1975]. Much of the variation among national health systems may reflect not only different degrees of performance but different choices about the relative importance of different goals. For example, both Canada and the United Kingdom finance most health care publicly [Abel- Smith 1995, Evans and Law 1995], relying on general revenues for nearly universal coverage. But Canada spends one of the world's highest shares of GDP on health (over 9 percent) and the United Kingdom spends a low share among high-income countries (just over 6 percent). Health care was consequently much more rationed in the United Kingdom, with long waiting lists for many procedures, at least until 1991 [Feldman 1994]. When health systems are evaluated across two or more dimensions at once, neither system appears to be clearly better than the other. Overall Health Spending and Public/Private Composition Governmentscan intervenein the health sectorvia fivedifferentinstruments,as indicatedin Part 1. Threeof these--informing,regulatingand mandating(otherthan financialmandates)--donot lendthemselvesto easysummaryor comparisonacrosscountries or systems. Even provisionof care is difficultto compare,exceptby crude measuressuch as totalconsultations,hospitalizations,hospitalbeds or professionalproviders,without adjustmentfor the type and severityof healthproblemstreated. Quantificationis easiest where financingis concerned,and the emphasishere is onthree aspects:the levelof expenditure,the distributionbetweenpublicand privatefinance,and the distributionbetween out-of-pocketspendingand insurance. Thismeanspayingless attentionto equityin the financingof healthcare, which hasbeen studiedin detail only forthe UnitedStatesand nine Europeancountries, and in less detail forfivepoor countries[VanDoorslaer,Wagstaffand Rutten 1993]. What is KnownaboutHealth Expenditure Informationon health-relatedspendingdoes not exactlymatchthe distinctionamonghealthcare domainsin Figure 1: there is no clear boundary to the out-of-pocket spending which affects health, nor is public expenditure always classifiedintopublic and privategoods. In countrieswithnationalhealth accountson the OECD model, the data approximatethe theoreticaldistinctionsmoreclosely,so far as financingis concerned. Publicgoodscorrespondtothe categoryof GovernmentPublic Health Activities. Truepublicgoodsusuallyabsorbonly a smallfractionof public and total moneyfor health,and do not presentmajor problemsor controversiesof classification. In the UnitedStates,for example,publichealthactivitiesaccountfor onlythree percent of total spending;personalservicesaccountfor 90 percent,and administrationfor the other seven percent [Lazenbyet. al. 1992]. Evenif three percentis too low,becausesomehighly- justified activitiesare under-financed,the optimumlevelin a richcountry is still probably onlya fewpercent. All other health care, however financed, is classified as Personal Health Care Services and correspondscloselyto privategoods--includingthosetreatmentswhichprovide 33 externalities and therefore are partly public. Program Administration, including the costs of administering private insurance as well as public programs, makes up the rest of the recurrent cost of health care. Total national health expenditure includes medical research and construction of health care facilities and laboratories as well as recurrent spending. It is possible to distinguish public from private health care spending in many countries, and at least for some of them, to make either or both of two further distinctions. One is between insurance and out-of-pocket costs in private spending. Since all public expenditure can be regarded as an implicit form of insurance, this allows a distinction between insurance of all types, public and private, and uninsured out-of-pocket expenditure. The other distinction is between direct financing and mandated social insurance in public spending. Health systems differ greatly in the relative importance of these two waysof paying for health care. HowExpenditureLevelandCompositionareRelatedto Income Like nearly all previous analyses of this question [McGuire et. al. 1993,OECD 1995]Figure 3 shows that health care is a "superior" or "luxury" good, which takes a steadily larger share of income as income rises. At the low incomes typical of Sub-Saharan Africa and poor countries of Asia, total health spending is about 3-4 percent of GDP. It rises to 6-9 percent of gross product in most OECD countries. This increasing importance relative to income is not surprising, given the large contribution health care can make to welfare and the very high costs of some interventions. What is surprising is that the public share of total health expenditure also rises with national income. As Figure 4 shows, the fraction that is either financed directly or mandated by the state increases from under 40 percent in most low income coutnries, to over 80 percent in some European countries. The data here and in subsequent analyses refer to expenditure and not to the public or private provision of services. However, a general pattern exists: in high-income countries, finance and provision are usually separated, with public expenditure going to private providers. In poorer countries, public money is mostly or entirely spent through public facilities. Both theory and the empirical evidence suggest that competition makes sense in the provision of care; it may improve health status, and it should certainly lower costs and increase consumer satisfaction compared to a situation in which consumers must choose between free or low-cost services with no choice of provider, and competitive services at higher out-of-pocket cost. Many of the problems of health services in poor countries appear to arise from the monopoly linkage of public money to public provision. None of the three general arguments for state intervention in finance of health services also justifies state delivery of those services, although, as indicated in Part 2, widespread poverty makes such public provision more attractive and may help account for its greater prevalence in poor countries. Both the share of income devoted to health and the division between private and public spending vary greatly around the trends in Figures 3 and 4, which makes it difficult to see any association between the two. Figure 5 therefore relates these two variables directly, Figure3 Shareof GDPSpentonHealth,Relatedto Income Share 0.14 0.12 - 0.1 - 0.04 0~~~~ 01.04 -Si S 0.02 - 0~~~~~~~~~ 0 5000 10000 15000 20000 25000 PerCapitaGDP (in1991internationaldollars) Pointsrepresentindividualcountries.The straightline showtheOLS regression, HealthExpenditureas a percentageof GDP =4.17(0.32)+0.1905(0.031)GDPpercapitainthousandsof 1991intemationaldollars (Standarderrorsin parentheses)N= 69Countries R-squared= 0.3505 Source:AlldatafromAnnexTable1. Figure 4 Public Expenditure on Healthas a Share of Total Health Expenditure, Relatedto Income Share 0.9 _ 0.8 0.7 -0.6 * 0.5 I* 0.4 - 0.3 _- % 0.2 _ * * % 0.1 0 1 0 5000 10000 15000 20000 25000 PerCapitaGDP(in1991internationaldollars) Pointsrepresentindividualcountries.Thestraightline showtheOLSregression, HealthExpenditureasa percentageof GDP = 45.44(3.00)+ 1.469(0.288)GDPpercapitainthousandsof 1991internationaldollars (Standarderrorsinparentheses)N =69Countries R-squared= 0.2762 Source:AlldatafromAnnexTable1. 36 omitting per capita income. The radial dimension (distance from the origin) shows the share of GDP devoted to health, while the angular dimension (from the horizontal toward the vertical axis) shows the distribution between public and private expenditure. The 69 countries are distinguished according to quartiles of income per capita. This way of presenting the data shows that there is no narrow path that all countries follow, but that there is nonetheless something of a pattern. First there is an expansion in the share of income devoted to health, driven by private expenditure; subsequently, private spending is displaced by public expenditure and a further increase occurs in the fraction of GDP that is spent on health care. Since these phenomena are related to increasing income, health systems can be said to evolve along this path as a country develops. However, this relation applies only for countries over all income levels taken together, not for smaller groups of countries at comparable incomes. A comparison of 24 OECD countries, for example, shows a random scatter, as do the data for poor countries separately. This is evident from the distribution of small circles corresponding to each quartile of the distribution of income per capita. Although particular subsystems such as social security or private insurance are said to "cover" certain population groups, in fact the same consumers often utilize a variety of providers and draw on different sources of finance. Where public provision is available at low cost but also low quality, people tend to use it for high-cost services such as hospitalization and to buy cheaper care, mostly ambulatory, from private providers [Baker and van der Gaag 1993, Bitran and Mclnness 1991]. Thus the composition of what is financed privately depends on what is offered publicly, and so does the total spending level. In contrast, public spending does not seem to take private expenditure into account [Murray, Govindaraj and Musgrove 1994]. Ministries of Health, in particular, seldom know what is being spent in the private sector and do not base their own programs on it. (To the extent that they plan their spending on the basis of the population's health problems, and those have been partly prevented or treated by the private sector, governments do of course respond indirectly to what is happening in the private sector.) Moreover, the tendency of consumers to use care that is financed in different waysmeans that there is no simple association between sources or mechanisms of finance, and different population groups or different health problems. Explaining Health Outcomes The patterns of spending shown in Figures 3-5 pose the question, how do the amount spent on health care, and the shares of public and private expenditure, affect the performance of the health system ? Given the difficulty of obtaining and summarizing comparable information on either satisfaction or equity, this is easiest to do with measures of health outcomes. Since health status is very strongly associated with a country's income [World Bank 1993b, Baker and van der Gaag 1993], the effect of income needs to be accounted for in any comparison of spending and results. If this is not done, it may falsely appear that more health expenditure is extremely effective in producing better health outcomes, and also, because the public share of spending rises with income, that more public expenditure in particular yields better health. Figure5 Percentageof GDPSpentonHealthCare,and Its DistributionbetweenPublicandPrivateSpending: CountriesClassifiedby Income Allpublic Shar0of \ Public-privatedistribution Shareof 0 0 health (30O expenditure - 4 *\ In GDP _ 0* \ All private Shareof healthexpenditureInGDP - 4 Key:quartileof GDPper capita * First(poorest)quartile * Secondquartile o Thirdquartile o Fourth(richest)quartile Source:AlldatafromAnnexTable1. 38 Severalpreviousstudies[Cochrane,St. Legerand Moore 1978,McGuireet. al. 1993,OECD 1995]haveanalyzedthe relationbetweenhealthcare spendingand health outcomes,includinga numberof other explanatoryvariables. Thesevariablesmay referto characteristicsof the healthcare system,or theymayrelatehealthoutcomesto the degree of literacy[Tresseraset. al. 1992]or schoolingof the populationas wellas its averageincome [WorldBank 1993b,Figure3.1], sincethere is abundantevidenceof an independenteffect of schoolingon health status. In the presentanalysis,regressionestimateswere made relatingtwomeasuresof healthstatus--lifeexpectancyat birth, and childmortality--to incomeand variouscombinationsof the shareof GPD spenton health,the share of health spendingthat is public, and interactionsamongthese variables. In all suchcombinations, only incomeper capitawasclearly significantin determininghealthstatus. Thereforein the analysesdiscussedbelow,both healthstatusand health expenditurewerepredictedon the basis of incomeonly. Countries' incomelevels,and the share of health spendingthat is public, were then introducedas classifyingvariables,to see if theyhelp explainthe observed patterns. Life Expectancy at Birth As incomeper capitaincreases,lifeexpectancytends towarda maximumof about 80 years. Withina country,the effectof extra incomein reducing mortalityis mostpronouncedat verylow incomes,evenafter takingaccountof the consumptionof medicalconsultations[Heysenand Musgrove1986]. Figures6a and 6b take the deviationsfrom a regressionrelatingincomeand lifeexpectancy,and comparethemto the deviationsfrom the regressionpicturedin Figure3, betweenincomeand the share of GDP devotedto health. (The income-lifeexpectancyrelationis describedin the Keyto the Figure.) The smallcirclerepresentingan individualcountry showsboth whetherit spends more or less than expectedon health,and whetherlifeexpectancyis greateror less than predicted,given its per capitaincome. A countryoccupiesthe sameposition in Figure 6b as in Figure 6a, so the patternof deviationsis the same. The differenceis in a third classifying or explanatoryvariable:countriesare classifiedin Figure6a by quartileof incomeleveland in Figure 6b byquartile of the shareof expenditurethat is public. It is clear from eitherfigurethat not onlydo somecountriesspendmore (or less) and buy more (or less)life expectancy,but the deviationsin theseindicatorsare not closely correlated. Countrieswhichspenda large shareof incomeon healthcare do not necessarily buy greaterlongevity,and somecountriesmanageto spendlessthan expectedand still enjoy longerlife. Whencountriesare distinguishedbyquartile of GDPper capita, in Figure 6a, it is evidentthat large deviationsin lifeexpectancy--peopledyingmuch youngeror much older than wouldbe predictedfrom a country'sincomelevel--areconcentratedin poorer countries. It also appearsthat amongthesecountries,higherspendingon health,relativeto income, does pay offin greaterlifeexpectancy.The dark circlesindicatingpoor countriesare distributedaroundan upward-slopingline, witha coefficientof 1.22 more yearsof life for each additionalone percentof GDPspenton healthcare (significantlydifferentfrom zero at the ten percentconfidencelevel). At highincomes,lifeexpectanciesconvergecloselyand becomelargelyindependentof bothincomeand health expenditure:the open circles indicatingupper-incomecountriesare scatteredalonga line witha slopeof 0.26 yearsof 39 extra life for each additional one percent of GDP devoted to health (significant only at the 20 percent confidence level). Increased expenditure at those income levels may be buying substantial improvements in the quality of life by reducing disability, but it no longer has much effect in extending life. This result is consistent with the finding that within the European Community, health care expenditure seems to have little impact even on mortality from amenable or preventable conditions [Mackenbach 19911. Health care spending does however appear to improve health status in six high-income countries when several life-style determinants of health are also taken into account [Wolfe 1986]. Figure 6b tests whether there is any association between these deviations and the share of public spending in total health expenditure. Countries with a relatively high public expenditure share (the open circles in the Figure) which spend a larger-than-predicted share of GDP on health, also show a higher-than-predicted life expectancy. Thus more health expenditure appears to pay off in greater longevity, when a large share of that expenditure is public. The slope of the relation between spending and life expectancy is 0.88 (significant at the six percent confidence level). That means it takes a increase of almost six percent of GDP to increase life expectancy by five years. No such relation is evident among countries with a relatively low share of public expenditure: spending more has no significant effect on mortality. These findings suggest that public money is usually concentrated somewhat more than private expenditure on those interventions which make a real difference to life expectancy, including those public health activities which are not provided in private markets. These results are partly consistent with analyses based directly on public expenditure alone, without considering private spending. In Latin American countries at mid- to high incomes, more public spending does extend life [Govindaraj, Murray and Chellaraj 1994, Appendix 17], although for a set of 58 countries at all income levels, there is no significant relation between public spending and longevity [Murray, Govindaraj and Musgrove 1993, Figure 2]. Since these studies do not take account of total health spending, it is not surprising that public expenditure does not show a systematic effect. In the present analysis, when life expectancy is related to income, the total health share in GDP and the actual level of health spending, the share of GDP that is spent on health is almost significant at the 20 percent confidence level. Childhood Mortality Since much of the difference among countries in life expectancy is due to differences in mortality at early ages, this same kind of analysis is repeated using mortality before age five as the measure of health status. Child mortality is strongly related to income per capita, with convergence at high incomes to a level of about five deaths per thousand. Poorer countries show more variation and larger deviations from the values predicted by a regression relating child mortality to income. Figure 7a shows these deviations and the regression from which they come, together with the deviations from expected health care spending: the dark circles represent low-income countries. The open circles representing richer countries, in contrast, show great variation in the share of GDP Figure6a LifeExpectancyandHealthExpenditure:CountriesClassifiedbyIncome 15 CD ~~~~~~~~~~~~~101 0 0~~~~~~~~~~~~~~ x -J~~~~~~~~~~~~~~~~~~C * -0.060-0.05 -0.04 -0.03 -0.026-0Ln 0.01 1 0.02 0.03 0.0400.05 E 2 0 DeviationfromPredictedPercentage ofGDPSpentonHealth Key:quartileof GDPpercapita DeviationsfromthepredictedshareofGDPspentonhealthare takenfromtheregressioninFigure3. Deviationsfrompredictedlife *First (poorest)quartile expectancyaretakenfromtheOLSregression(notshown), Secondquartile Ln(80- LifeExpectancy)3.1336(0.0671)- 0.000116(0.000006) = 0 GDPpercapitain 1991internationaldollars.(Standarderrorsin o Thirdquartile parentheses)N = 69countries R-squared= 0.8252 o Fourth(richest)quartile Source:AlldatafromAnnexTable1. Figure6b LifeExpectancyandHealthExpenditure: CountriesClassifiedbyShareof PublicExpenditure 15 CZ 0 C.X 0 0 0 OsO~~~ wx oow a) 0~~0 :0 22 -0.06 0-0.05 -0.04 -0.03 -0.02 (9.0.l01 002 *003 0.0*00 aCuD E 0 - 2 :5~~° 0 n c-. 00 l. 0~~~~~ -15 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ DeviationfromPredictedPercentageof GDPSpentonHealth Key:quartileof PublicShareinHealthExpenditure Deviationsfrom thepredictedshareof GDPspenton healthandfrom predictedlifeexpectancyarethesame * First(lowest)quartile asin Figure6a;onlytheclassificationof countries * Secondquartile is different(seeKey). O Thirdquartile o Fourth(highest)quartile Source:AlldatafromAnnexTable1. Figure7a ChildMortalityandHealthExpenditure:CountriesClassifiedby Income 7Vw 120 CD :L L 80 co CD ,2 40 ,w 60 ~~~~~~ Z' ~~~~S0 0~~~~~~~~~~~~~4 0~~~~~~~~~~ 0~~~~~~~~~~ 0~~~~~~~~~~~ CD* -0.06 -0.05 -0.04 -0.03 0.02 (-C.id L 00tol= 5 0.02 00.03 0.04 0.05 0~~~~~~~~~~ 0 Deito rmPeice ecnaeo GODP)SpenonHat .0 _ _ _ __40_ _ _ _ _ _ DeviationfromPredictedPercentage of GDPSpentonHealth Key:quartileof GDPpercapita DeviationsfromthepredictedshareofGDPspentonhealthare takenfromtheregression inFigure3.Deviationsfrompredictedlife * First(poorest)quartile expectancy aretakenfromtheOLSregression (notshown), Secondquartile Ln(ChildMortality)=4.6556 (0.1042)- 0.00015(0.00001) GDPpercapitain 1991internationaldollars.(Standarderrorsin 0 Thirdquartile parentheses)N = 69countries R-squared=0.7607 0 Fourth(richest)quartile Source:AlldatafromAnnexTable1. Figure7b ChildMortalityandHealthExpenditure: CountriesClassifiedbyShareof PublicExpenditure x1 120 0 U3 U) o .1000 0 *~~~~~~~~~08 co a) ~~~~~*60 40- 0 0 20~~~~ Q (D -06-.5 -. 4 0.030 -0.02 &-0.01 0 1 00.02 00.03 0.04 0.05 a 6 E 0 0~~~~~~~ Deviation fromPredictedPercentage ofGDPSpentonHealth Key:quartileof PublicShareinHealthExpenditure Deviations fromthepredictedshareofGDPspent onhealthandfrompredicted childmortaliyarethe * First(lowest)quartile sameasin Figure7a;onlytheclassificationof * Secondquartile countriesis different(seeKey). o Thirdquartile 0 Fourth(highest)quartile Source:AlldatafromAnnexTable1. 44 devoted to health, but almost no variation in child deaths. Contrary to the finding relative to longevity, there is no indication that spending a larger share of GDP, at a given income level, results in lower child mortality, for either rich or poor countries. This is consistent with an analysis of infant mortality which also finds little effect of health care expenditure and substantial effects of income, education and calorie intake [Kim and Moody 1992]. Figure 7b relates the same deviations in expenditure and child mortality as in Figure 7a, to the share of health spending that is public. It indicates that there is no relation between that share and child mortality, whether for all countries together or separately for those that spend either a high or a low public share. Multivariate estimates of the determinants of child mortality give much the same answer: income is always significant, but the health share in GDP, the public share in health spending and the share of public expenditure on health in GDP never are. These findings are consistent with two kinds of evidence: that child survival is strongly related to other factors such as adequate nutrition [Pelletier 1992, Beaton et. al. 1993], and that the crucial health care interventions for saving children's lives are relatively inexpensive [Bobadilla, Cowley, Musgrove and Saxenian 1994]. Thus a small part of public expenditure may be extremely effective at increasing child survival, but thereafter more health expenditure affects life expectancy more by preventing deaths at later ages than by saving children. This is probably true whether the expenditure in question is public or private. The result is that overall spending levels show little relation to child mortality despite the great importance of expenditure on a few interventions such as immunization. Since the data available on health spending for most countries do not permit disaggregation by type of intervention, it is impossible to judge the health impact of parts of that expenditure. This also makes it difficult or impossible to tell whether public spending follows the conceptually appropriate pattern discussed in Part 2. Summary: Expenditures and Results Since aging is a complex biological process which responds only very little to health care interventions [Ricklefs and Finch 1995], life expectancy appears to have a natural upper bound. It is not surprising that as populations approach that limit, additional spending on health care ceases to extend life. Nonetheless, when life expectancy is still well below any biological limit, health care does make a difference. What is more, public expenditure seems systematicallyto be more effective than private spending in this regard, presumably because it is devoted to a slightly different set of interventions. These probably include some of the interventions that are particularly effective at preventing child deaths--but because those commonly take up only a small fraction of expenditure, there is no discernible relation between total spending and child mortality. Without information on the composition of spending, little more can be said. Thus the evidence does not favor public over private spending irrespective of what is being financed, but does indicate the importance of enough public expenditure to ensure that the most valuable public health interventions are adequately provided. The crucial role of a few services may explain why China has been so successful in improving health over the last few decades, despite the fact that the government has acted more to provide than to finance health care, and people have had to pay for much of their 45 care out-of-pocket, even at public facilities [WorldBank 1992b]. This is in contrast to the general finding that expenditure seems less effective in countries where the public share is relatively low. It should also be noted that in other respects, such as coverage of insurance and appropriate incentives to technical and allocative efficiency in the production of health interventions, China since about 1980 should be considered more of a failure than a success. Health outcomes are the most important objective of a health system, but the other, harder to measure, goals of equitable treatment and client satisfaction at reasonable cost also matter. The preceeding analysis shows that the question of how effectively either kind of expenditure improves health status, depends on the outcome measure used. Mortality measures of course do not account for many of the less dramatic health gains which also explain why individuals and societies spend as they do on health care. And the patterns of early mortality or life expectancy at birth may be very different from those of life expectacy at higher ages; much health expenditure may go to extend life after age 60, for example. Comparisons of this sort using a more inclusive notion of health status, taking account at least of the major kinds of long-term disability from non-fatal illness and traumas, might reveal still other relations. Unfortunately, more complete estimates of health status or of the total burden of disease exist for very few countries and are not alwayscomparable because of differences in the coverage of causes of death or in the inclusion of disability [Bobadilla and Cowley 1995, Bobadilla 1996]. Out-of-Pocket Spending versus Private and Public Insurance Another reason why there is not more relation between health outcomes and the level and composition of expenditure may be that the private component includes two very different modes: explicit private insurance and out-of-pocket spending. The distribution of spending between these categories, like that between public and private expenditure, affects which interventions are purchased, and therefore the effect on health status. Unfortunately, it is extremely difficult to estimate how private expenditure is divided; data are available for very few countries. Figure 8 shows both the distinction between out-of-pocket spending and insurance (on the radial dimension), and the split between public and private financing of insurance (on the angular dimension--all public on the vertical axis and all private on the horizontal axis). Countries are classified by income level, from which it is evident that as income rises, insurance takes over from out-of-pocket financing. This is to be expected: security is a "normal" good, meaning that people buy more of it as their incomes increase. Insurance, which provides security, is also a normal good, and it is needed to finance the catastrophically expensive interventions that health systems more frequently provide at higher incomes. If it is supposed that in poor countries virtually all private expenditure is out-of- pocket, then the dark circles representing those countries would all lie along the vertical axis. The Income-Related Expansion of Insurance Out-of-pocket expenditure is half or more of health expenditure in very poor countries, but even in rich countries where the bulk of health expenditure is public it absorbs 4-20 percent of spending [Schieber, Pouillier and Greenwald 1992]. Outside the OECD countries, private insurance is still relatively rare (but growing) Figure8 DistributionofHealthExpenditureamongOut-of-Pocket Spending,PrivateInsuranceandSocialInsurance: CountriesClassifiedby Income Allpublicinsurance g 100 90 80 - 70 - Public-privatedistribution Shareof 60 - insurance (publicand 50 private) 40' 10 _ 01 | I | Allprivateinsurance 0 50 100 Allout-of-pocket Shareof insurance(publicandprivate) - Key:quartileof GDPpercapita * First(poorest)quartile * Secondquartile O Thirdquartile o Fourth(richest)quartile N= 17countries Source:AlldatafromAnnexTable1. 47 and nearlyall privatespendingis out-of-pocket,whetherit is a largeor a smallshare of total expenditure. And whilefor somecountriesthe only insuranceis public, there being essentiallyno privateinsuranceindustry,there are no healthsystemsfinancedmore than half by privateinsurance. Amonghigh-incomecountries,onlythe UnitedStatesand Switzerland rely heavilyon privateinsurance,and eventhere the shareis only about40 percent. It appearsfromFigure 8 thatpublic and privateinsuranceare substitutesrelativeto out-of-pocketexpenditure. That is, the levelof non-insurancespendingdoes not dependon the compositionof insurance. This is somewhatsurprising,sinceprivateinsurancegenerally requireshighercopaymentsand deductiblesthanpublic insurance,and thereforegenerates more out-of-pocketexpenditure. But the amountsmaybe smallenoughthat no connectionis apparentwhenall countriesare consideredtogether. It mayalso be that whenpublic funds go to financepublic provisionof poor quality,especiallyin low-incomecountries,consumers end up payingout-of-pocketto offsetthe deficienciesof thoseservices,althoughthere are no formalcopayments. Evenin countrieswithclose to universalpublicinsurance,there is often substantial privateinsurance. This is boughtprimarilyto supplementpubliccoverage(in Canada,Italy, Portugal,Spainand the UnitedKingdom)or to coverthe copaymentsrequiredunder public insurance(in France,Denmarkand the Medicareprogramin the UnitedStates). In Ireland, it is purchasedprimarilyto replacesocialinsurancealtogether;the sameis true in the Netherlands(particularlyfor the wealthy,whocan opt out of the public system)and in the UnitedStates(fora large shareof the populationwho are not entitledto public insurance). Suchinsuranceis boughtby a majorityof the populationin Franceand in Canada,but by only 15percentin the UnitedKingdomand 19percentin Germany. Private insurance accountsfor only 7 percentof spendingin Canada,despitethe highcoverage,becausethe public systemcoversnearlyall medicalneeds;the privateinsuranceshareis about 9 percent in France and 12-16percentin the UnitedKingdom,Japan and Germany [VanDoorslaer, Wagstaffand Rutten 1993;Schieber,Pouillierand Greenwald1992]. In the UnitedStates, about one-thirdof the beneficiariesof Medicare,the socialinsuranceprogramforthe elderly,also buy supplementalprivateinsuranceout of pocket[Sheaand Stewart19951. The tendencyacrosscountriesfor more and more of health care to be financedby insuranceat higherincomes,mirrorsthe historicaltendencyfor insuranceto expandrelative to out-of-pocketspendingas incomeincreasesand a country'shealthsectorbecomesmore complex. In the UnitedStates,forexample,patientspaid directlyfor an estimated88 percent of their healthcare in 1929,whenthe privateinsuranceindustrywasminusculeand government(national,stateand local)financedonlynine percentof care. By 1960,the out- of-pocketsharehad fallento 56percent, and privateinsurancehad expandedto 23 percent, slightlysurpassingthe governmentshare. The Medicaidand Medicareprogramsdid not exist yet, so there was no mandatedsocialinsurance. The shareof privateinsurance continuedto growuntil 1990,whenit wasabout 35 percent,but publicspendinggrew even faster,reaching41 percent. (Thisfigureincludesthe tax subsidyto privateinsurance.) Out- of-pocketexpenditurehas continuedto fall steadily,to about 23 percent[Bovbjerg,Griffin 48 and Carroll 1993]--whichis stillhighcomparedto mostOECDcountries,where insurance has evenmore thoroughlyreplaceddirectpaymentbyconsumers. In verypoor countries, part of privateout-of-pocketspendinggoesfor allopathicor traditionalmedicinewhich is neverfinancedpubliclyand seldomcoveredbyinsurance. The shift fromthis traditional healthcare to modemmedicalcare also helpsaccountfor the increasein insurance expenditure,althoughthe amountsspentontraditionalmedicineare usually quitesmallin absoluteterms since it catersto poor peopleand uses no modempurchasedinputs. The TwoFormsof Publicly-FinancedInsurance Anotherwayof classifyinghealth expenditureis to combineallprivatespending(insuranceplus out-of-pocket),and to separate public expenditureinto directfinanceandmandatedsocialinsurance. Informationis availablefor more countriesthanwhendistinguishingprivateexpendituresby whetherthey are paid by insuranceor not, becausethe twokindsof public financeusuallycorrespondto distinctinstitutions. In practice,direct financeis typicalof ministriesof health,whereas mandatedinsuranceusuallytakesthe formof a socialsecuritysystem. Bothemployeesand employersare requiredto contributea specifiedfractionof wages,usuallyup to some ceiling;employersare usuallynot requiredto provideprivateinsuranceor to self-insure. These systemscoveronlya fewpercentof the populationin manymiddle-and low- incomecountries,sizeablefractionsin Italy,the UnitedStates,Argentina,Mexico,China and Korea,and nearlythe entirepopulationin Japan, somecountriesof Europe(France, Germany,Spainand the Netherlands)and a fewmiddle-incomecountriessuch as Brazil and CostaRica. Nearlyuniversalcoverageis achievedonlyin high-incomecountries,or where socialsecuritybenefitsare extendedevento non-contributors(as in Brazil and CostaRica) ratherthan being limitedto contributors. In thosecasesthe socialsecuritysystemtakes over what is usually the functionof the Ministryof Healthto financecare for the "informal" sector,from whichwagetaxescannotbe collected. Althoughboth arrangementscoexistin manycountries, one or the otherusuallydominatespublicexpenditures,so that countriescan be classifiedas dependingchieflyon directfinanceor on socialinsurance. Mandatedinsurancesystemsdiffer fromdirectpublicfinanceprimarilyin using payrolltaxes ratherthan generalrevenues,althoughtheymayalso be financedpartly from such revenuesor from other, non-laborearmarkedtaxes. Thetwo formsof public expendituremaythereforehavedifferentmacroeconomiceffects. Theeffect of financing healthcare byone or the other routedependson such factorsas the incidenceof different taxes--payrolltaxes are usually slightlyregressivewhereasdirecttaxes, at least in high- incomecountries, are slightlyprogressive[VanDoorslearand Wagstaff1993]and on administrativecost and convenience.Mandatedinsurancegenerallyallowsfor more choice byconsumers,and maybe preferablefor that reason [Summers1989]. There is nothing uniqueto health aboutthe choiceof financingmechanism:the sameconsiderationsapply to pensions,forexample. Thesetwo waysof financinghealthcare also differpolitically,in that mandated expendituresare commonlyearmarked,whereasbudgetsdo not includeanyguaranteeof 49 spendingon health. And the workerswho payfor mandatedinsurancecommonlythink of the paymentas "theirmoney", whichgivesthema right to services,perhapsmorethan taxpayersthink of generalrevenuesas being their ownresourcesor as conferringspecific rights. Publicsubsidyof healthcare for the poor is generallyfinancedfrom general revenues,whichmeansthat in most systems,the poor haveless of a guaranteeof resources availablethan the populationcoveredbyemployment-relatedinsurance. Thesedifferencesare important,but someother featuresof a healthcare systemdo not vary accordingto how publicfinancingis organized. Witheithermandatedinsuranceor generalrevenuefinancing,whenthe statepaysfor but does not deliverservices,there is no reason forprovidersto be paid differentlythanthey wouldbe paid fromprivatesources. The incentivesand relativeadvantagesof the differentmechanisms--salaries,fee-for-service, capitation,diagnosticrelatedtreatmentgroups,etc.--dependonthe inherentincentivesunder each arrangementand on the type of service [Ellisand McGuire 1990,Barnumand Saxenian 1994]but are independentof who paysfor it. Similarly,whenservicesare financedby governmentthe right leveland type of user fees are independentof whetherthe provisionis public or private,dependingratheron the natureof the servicesand the incomesand other characteristicsof the users. Figure 9 classifiesexpendituresas private,directpublic financeand mandatedpublic insurance. The radialdimensioncorrespondsto the distinctionbetweentotalpublic and total privatespending:all privateexpenditureat the originand all public expenditureat the circumference. The angulardimensionthen showsthe proportionsin whicha countryuses generalrevenuefinancing(alongthe verticalaxis)or mandatedsocialinsurancepaid for largelybypayrolltaxes (alongthe horizontalaxis). Asin Figures6a, 7a, and 8, countries are classifiedby incomelevel. As the Figure shows,low-incomecountriesare more likely to rely on direct financethan to pay for care largelyor exclusivelyvia socialinsurance. Thisprobablyreflectsthe difficultyof collectingwagetaxesand achievingsubstantial coveragein an economywith little formalemployment. Nonethelesssomepoor countriesdo spendmuch more on socialinsurancethan throughgeneralrevenues,and a numberof other low-incomecountriesspendclose to equalamountsthroughthe twomechanisms,because while mandatedinsurancecoversfewerpeopleit oftenspendsmuch more per beneficiary than is spentthroughMinistriesof Health. Publicservants,forexample,maybe coveredby relativelygeneroussocialinsuranceevenin quitepoor countries. Thiskind of inequalityis also pronouncedin middle-incomeLatin Americancountrieswithonlypartial social insurancecoverage[Mesa-Lago1992]. Amonghigher incomecountriesthere seemsto be nopatternto the form of public financing. Some, like the UnitedKingdomand Canada,use direct financeexclusively, whereasothers, suchas Germany,rely largelyon employment-basedsocialinsurance. The UnitedStatesuses substantialsharesof bothgeneralrevenues(tofinancecare for the poor throughthe Medicaidprogram)and socialinsurance(tocoverthe elderlythroughthe Medicareprogram). Relianceon both formsof financealso characterizesItaly,Russia, China, Korea,and most LatinAmericancountries. Figure9 Distributionof HealthExpenditureamongPrivate Spending,MandatedSocialInsuranceandDirectly- FinancedPublicExpenditure:CountriesClassifiedbyIncome Alldirectfinance 100 90 70 0 Mandated/directly 70!r 0financed oo \ distribution Publicshare 60 O of health 0 expenditure 50 . ;\ 401 I * 0O 40 Q o 30 ~0 0 20~~~~~~~ 10 - ; @ 1 0. 0 0 CD i Allmandated insurance 0 50 100 Allprivate Publicshareof healthexpenditure > Key:quartileof GDPpercapita * First(poorest)quartile * Secondquartile 0 Thirdquartile 0 Fourth(richest)quartile N =58countries Source:AlldatafromAnnexTable1. 51 Does it makeany differenceto healthoutcomes,whetherpublicfinanceis direct or operatesthroughmandatedsocialinsurance? The majorityof OECDcountriescan be classifiedas usingchieflyone mechanismor the other; whenthis distinctionis addedto a relationbetweenlifeexpectancyand healthcare expenditure,it appearsthat countriesthat rely on direct financeachievesomewhatgreaterreductionsin mortalityas spending increases. That is, additionalexpenditureis moreeffectivein extendinglifewhen it occurs throughdirect publicfinance[Elola,Daponteand Navarro19951. Thisprobablyreflects more nearlyuniversalcoverageof those interventionswhichare most effectivein preventing early mortality,althoughthe distributionof incomewithina country and the amountand natureof privatehealth spendingmayalso matter. The samecomparisonbetweenhealthoutcomesand the formof publichealth care financecan be madefor a largernumberof countriesat all incomelevels,usingthe same regressiondeviationsas in Figures6a-7b. Atleast amongpoor countries,relianceon direct financingappearsto be associatedwitha tendencyto spendsomewhatmore on health care than expected. However,there does not seem to be any systematicrelationbetweenthe way that publichealth care expenditureis paidfor, and childmortality. This is consistentwith supposingthat whendirect financeis a smallshareof public expenditure,it is concentrated on the most effectivelife-savinginterventions,particularlyfor children. But as the direct financeshare increases,it comesthe coverthe same rangeof interventionsand the same populationas the alternative,mandatedsocialinsurance,so that there is no difference betweenthe two inhealth outcomes. State Intervention in the InsuranceDomain The availableevidencesuggeststhat there is no clearlybest wayto organizethe risk- sharingdomainof healthcare [Griffin1992],as betweenprivateand public insurance,or betweendirectpublicfinanceand relianceon mandatedsocialinsurance.This is partly becausemany questionsabout optimalpricesand provisionhavenot been answered,so the welfareeffectsof differentinstitutionaland financialarrangementsare not fullyunderstood [Diamond1992]. Nonethelessthe problemstypicalof privateinsurance,discussedin Part 2, are potentiallyso severeand so difficultto controlwithoutsubstantialpublic interventionby wayof mandatesand financing,that public insurancedoes appear superiorto unregulated privatevoluntaryinsurance[Barr1992]. Correspondingto this theoreticalconclusion,there is a cleartendencyfor publicly-financedinsuranceto expand,as incomerises--relativeto both privateinsuranceand out-of-pocketspending. Whyhavemostcountriesfollowedthe path towardpublic insurance? And whatsort of publicrole does this approachimply? Governmentscan fully solveone of the problemsof privateinsurance,lack of coverageforthe poor,only byat least partialsubsidy. Mandatingcoveragebyemployersis feasibleonly for thosepoor peoplewith formalemployment,who are never the poorest in society. Toaddress twoother problems--thelack of coveragefor high-riskconsumers,and highcosts--requiresregulation,at a minimum. However,it is rare for governmentsto do nothingmore than regulateprivateinsurance,and morecommonto mandateor financesocial 52 insurance. Part of the reasonfor this maybe an unwiseeffort onthe part of governmentsto do too much,particularlyin poor countries. But governmentsinpoor and middle-income countriesgenerallyhavenot "doneso much"as to eliminatethe needfor a large share of out-of-pocketexpenditure,and muchof the populationis too poor to buy health insurance voluntarily. WhySocialInsuranceis so 'WidespreadSeveralreasonsexplainmostof the concentration on socialinsurance,evenat incomeshighenoughthat privateinsuranceis feasiblefor much of the population. The first is historical,and refersto mandatesrather than direct governmentfinance. Sincemandatedinsurancebenefitsidentifiedgroups, the existenceof such protectionfor someworkersleadsto effortsbyother workersto obtainthe same benefits. Thiseffectshouldbe expectedto be stronger,the more labor is unionized[Navarro 1989]. There is a "ratcheteffect",bywhich socialinsurancetendsto grow oncestarted; it is easier to extendcoverageto newgroupsthanto reduce it for thosealreadyinsured. Direct finance,in contrast, is in principleavailablefor everyone'shealthcare and so its expansion shoulddependmore on a generalizedsocialdecisionto extendcoveragethan on competitive or emulativepressuresfromparticulargroupsin society. Also, mandatedinsuranceis a second-bestpolicyaccordingto eitheran extremeliberalor an extremeconservativeview, and so is easier to agreeon thaneither directlyfinanceduniversalcoverageor purely voluntaryprivateinsurance[Summers1989]. Non-mandatedprivateinsuranceprovidedvoluntarilyby employersand made obligatoryfor their employeesalso tendsto spread once introduced. However,sinceit is not requiredby law,it canmore easilybe reduced,or competitivebenefitscanbe providedin other formssuch as higherwages. And there is of courseno obligatorymechanismfor includingthe self-employed.Thusprivateemployer-relatedinsurancedoes not tend toward universalcoverage. This explanationof whysocialinsuranceexpandsappliesto Japan, manyEuropeancountries,and to a lesserextentto such countriesas Argentina,Brazil and Costa Rica. It does not explainthe (near)universalizationof socialinsurancebydirect governmentfinancein Canadaor the UnitedKingdom. Nor does it explainthe attemptto copy such coveragein low-and middle-incomecountries,includingmanyformer English colonies; historicalmodelsof what is appropriatefor a governmentmaygreatlyaffectthe public role in ex-colonialpoor countriestoday [vander Gaag1995]. Anotherfactorbehindthe expansionof socialinsurance--independentlyof how it is financed--isa senseof socialsolidarityor equity,whichsupportsa publicfinancingsystemto guaranteeuniversalaccess,with substantialsubsidiesfrom richto poor as well as from the healthyto the sick. The increasedequityin accessto care appearsto outweighthe inequity that ariseswhen socialinsuranceis financedbyslightlyregressivepayrolltaxes. The reality of such solidarity,as opposedto the rhetoric,has been widelyachievedonly in OECD countries;fewlow- and middle-incomecountrieshavemanagedto financeapproximately equalhealth care foralmosteveryone. Andthe degreeto whichsuch solidarityexists appearsto differ amongOECDcountries,being notablyless importantin the UnitedStates than in most of Europe. 53 Universalsocialinsurancehas the advantagethat it reducesadverseselection,with the attendantinequities,by poolingrisks in relativelylargegroupsand includingevery individual in somegroup. (Riskselectionbythe insurersis eliminated,if insuranceis compulsoryfor all those in a pool.) Toeliminateadverseselectioncompletely,however,the insurancemust not onlycover allthe populationbut put themall in just one pool: whenthere are several mandatedschemes,contributionsand benefitscan still differgreatlyamongthem, and people'schoiceof where to workor livemaystill dependinpart on howtheyviewtheir healthrisks. An extremeform of this differentiationoccurs in Argentina,wherethere are some300 mandatedpools, but much of the populationis not includedin any of themand there are hugeincome-relateddifferentialsin the valueand coverageof insurance[World Bank 1993a]. Directfinancewith a single-payersystemis more effectivein controlling selectionproblemsthan socialsecurityschemeswithseveralpayersand risk groups. In principlethe problemsof adverseselectionmightbe controlledby regulationof privateinsurance,but that wouldbe difficultor impossiblewithoutthe leverageprovidedby public moneyto subsidizehigh-riskconsumers. Partlyas a result, socialinsurancereduces or avoidsthe problemof labor immobilitybecauseof employer-relatedinsurance. When coverageis close to universal,qualitycan be maintainedbecausealmosteveryoneuses the system. (Qualityis more of a problemwhenpublicfinanceis concentratedon the poor, who haveno alternativefor expensivecare.) Moreover,dominanceof the health marketby publicexpenditurestill allowsfor supplementaryprivateinsurance,so at least those consumerswho canaffordto, canmakemarginalchoicesabout more spendingand more healthcare. Theresult is that compulsorysocialfinancecan be describedas "socially efficient",comparedto the failingsof privatehealthcare markets[Barr 1992]. It does not followthat socialinsuranceis efficientat the levelof individualhealthcare decisions,so that marginalexpenditureis alwaysdirectedwhereit willhavethe greatesthealth impactor contributemostto consumers'welfare. But privatehealthcare marketsdo not achievethat kind of efficiencyeither. Anotherimportantfactorexplainingthe politicalacceptanceof socialinsuranceis that it is easier to hold downtotalhealthexpenditure,thanwithprivateinsurancewhich is subsidizedthroughthe tax system. The averageincreasein healthcare spendingof all OECDcountriestogetherbetween1980and 1990wasonly0.6 percentof gross product, witha further, almostequallylarge,increasein the nexttwoyears. Canada,whichalso relieson public spendingbut has less strictcontrolson it, has been less successfulin this regardthan mostWesternEuropeancountries,witha twopercentof GDP increaseduring the 1980s. The extremecase of failureto controlhealthcostsis the UnitedStates,one of only tworich countriesto relyextensivelyonprivateinsurance. Expenditurewentfrom 9.3 to 12.6 percentof GDPin ten years(1980-1990)and has continuedto expand. In Switzerland,where privateinsuranceis an evenlargershare of expenditurethan in the UnitedStates,health spendinggrewby one percentof GDP in the 1980sand thenby another one percentin the nexttwo years[Schieber,Pouillierand Greenwald1994]. 54 The OECDexperiencesuggeststhatpublicexpenditureis an efficientmechanismfor reaching,and enforcing,a socialdecisionto spendsomewhatless, evenif it meansaccepting somewhatpoorer health statusthanmightbe achievedwithmuch more spending. The willingnessto makethis choiceappearsto rise as healthconditionsimproveand as expenditureon healthbecomesa large shareof GDP. Absentany such mechanismto arrive at a socialdecision, it is difficultto reachconsensus,as the failedeffortsto reformthe UnitedStateshealth systemin 1993-94demonstrate[Skocpol1995;Blendon,Brodieand Benson 1995]. There is also someevidencethat administrativecostsrise with the number of public insurers[Pouillier1992,OECD 1995],being lowestwitha singlepayer(as in Canada or the UnitedKingdom),higherwhere there are a largenumberof them, as in Germany,with about 1200 "sicknessfunds", and Argentina,withsome300 employment-relatedinsurers, and intermediatein countrieswitha smallnumberof large socialinsurancefunds, as in France [Lachaudand Rochaix19931or Japan[Schieber,Pouillierand Greenwald1992]. However,these differenceshavelittleeffecton totalcosts. Non-FinancialStateInterventionIn principle,thereis nothinginevitableabout the dominanceof publicinsuranceat high incomes;and if it werenot for such state-financedor state-mandatedinsurance,the public shareof healthexpenditurecouldbe much smaller. A countrycould rely on privateinsurancefor everyoneexceptthe poor,with publicspending goingonlyto financepublicgoodsand to subsidizecoveragefor the poor. However,the statecould not retirefrom other formsof intervention,particularlyregulation,unless society werepreparedto acceptthe resultinginequitiesand the likelihoodof much highertotalcosts. Part of the problemof inequitywouldprobablyarise fromdifferencesin qualityof health care betweenthe poor and the non-poor,in a systemwherepublicfinancefor insurancewent only to those in poverty. In fact, no countryin the worldhas achieveda highlyequitablesystemwitha stable shareof health spendingrelativeto GDP,while relyingpredominantlyon privateinsurance. The combinationof finance,regulationand mandatesnecessaryfor such an outcomemaybe theoreticallyattractivebut is politicallyand operationallyvery difficultto reach. Any governmentwishingto achievethe samedegreeof equityand cost-controlas a typical OECD country,while holdingthe share of publicspendingin healthto much less than half, would haveno modelsto follow. And theremightbe substantialtransitioncostsin tryingto reach such a state,startingfromwhere any of the OECDcountriesare now. That is probablywhy evenmajor reformsin howpublicmoneyis allocated,as in the UnitedKingdomin recent years, havenot greatlyaffectedthe shareof healthcare expenditurethat is financedpublicly [Abel-Smith1995,OECD 1994]. In the OECDcountrieswhichrely on directpublic financeor mandatedinsurance, regulationis pervasive. Governmentsnegotiateor controlproviders'fees, establishpayment mechanismsthat create incentivesto holddowncosts,set globalbudgetsfor regions, subsystemsor individualhospitals,and specifyor influencethe servicescoveredby social 55 insurance[Abel-Smith1984,Culyer 1990]. Evenin the UnitedStates,where less than half of healthexpenditureis public, suchinterferencecharacterizesthe twolarge publicprograms (Medicareand Medicaid). Doctors'feespayableunder the programsare controlled,and hospitalsare reimbursedbya fixedprice fora conditionor treatment(DiagnosticRelated Groupsor DRGs)ratherthan byfee-for-service. Partlyin consequence,during the 1980s expendituresper person insuredunderthese programsrose less than privateinsurance spendingper insuredperson [Bovbjerg,Griffinand Carroll 1993]. Andprivateinsurersare also increasinglyinvolvedin influencingor controllingproviders'decisionseither directlyor by wayof paymentmechanisms. Pure,arm's-lengthmarketsin health insurancedo not exist anywhere. PART4: CONCLUSIONS The conceptualdiscussionin Part 2 establishesthat the activitieswhich comprise health care canbe classifiedintothree domains,each of which correspondsto a distinct reasonfor the stateto intervenein the marketfor such care. Thisleadsto the conclusion that governmentsoughtto undertakethree broad kindsof intervention: * They shouldfinancepublic goodsand healthserviceswith substantialexternalities,to ensure that theyare producedand consumedin the rightamounts. * They shouldregulateprivatehealthinsurance,or else financeinsurancepublicly,to reducethe consequencesof adverseselectionfor bothefficiencyand equity. * They shouldsubsidizethe poor byfinancingsomeminimumof care, whetherby providingit directlyor by financingprivateproviders. Althoughthese are validgeneralprinciples,theydo not constitutespecificadvice on what the publicrole shouldbe in the health sector--whichinstrumentsto use for which purposes,howmuch to spendand howto allocateit, and so on. The empiricalevidencein Part 3 offerssomeguidancein this regard. It showsthat governmentstypicallydo intervene considerably,in particularby financinghealthcare or bymandatinghealth insurance;and that the degreeof such interventiontendsto increaseas countriesbecomericher. The reasonsfor this includethe increasingimportanceof insuranceat higherincomes,and the greaterease of controllingexpenditureswhenthey are publiclyfinanced. There is also some evidencethat publicexpenditureis more effectivethanprivatespendingin improvinghealth, presumablybecauseof its greaterconcentrationon a fewextremelyeffectiveinterventions. Still, countriesvarygreatlyin the degreeand kind of stateinterference,and display only a few empiricalregularitieslinkingthe public roleto health outcomesor other results [Dunlopand Martins 1995;OECD 1994;Saltman1995]. Whatmore can be said by wayof recommendationsfor governmentpolicyor action? It is convenientto separatethis into two questions. The first is, whatgeneraladvicecan be givenabout whatgovernmentsshouldand 56 shouldnot do to assureequity and efficiencyin the healthcare market? The secondis, how shouldpublicmoneybe spentonhealth-relatedinterventions? The rest of this paper offers someconclusionsin each of these areas, drawingon the theoreticaland factualmaterial presentedabove. The Appropriate Public Rolein Health As a firstapproximation,it is easier to say whatgovernmentsshouldnot do in health thanto specifywhatthey shoulddo. That is, it is clear that certainactionsare likely or certain to violateone or moreof the objectivesof a healthcare systemto an important degree. In the topographicmetaphorused earlier,such actionscorrespondto fallingoff the plateauof satisfactoryoutcomesand intoone of the surroundingchasms. Four "don'ts"are discussedin what follows:theyrefer to the waythe publicis taxed,chargedor exemptedto pay for health care; the providersto whichgovernmentstransferpublicfunds; the way providersare paid; and the servicesthey are paid to provide. WhatGovernment Should Not Do Thefirstthing governmentsshouldnot do is to use the tax system,or any systemof feesat publicfacilities,to makethe poor subsidizethe health care of the rich. Conceptually,subsidiesarejustifiedonly forthe poor, and broader financingof care throughinsuranceis a questionof efficiencyratherthan of equity. This is not onlya matterof whetherthe richuse more of publicly-financedservicesthan the poor do, althoughgreat inequitiesoftenarise becausethe rich havemore accessto those services. Financialequity also dependson who paysthe taxes. More narrowly,governmentsshould not contributeto social securityfinancingfromgeneralrevenues,unlesscoverageis universal,becausewhenonly part of the populationis coveredit is usuallythe poor who are excluded. And governmentsshouldnot treatprivateinsurancecoverageas a cost to employersunless it is also treatedas incometo beneficiaries. Suchpracticesare not only inequitable;they are also inefficientto the extentthat they leadto excessivespendingon health care, or reducelabor mobility. Controllinginequitablesubsidiesdoes not meanthat socialsecuritysystemsshouldbe dismantledor must be madeuniversal. Evenwhen incomplete,suchmandatedinsurance includessubstantialprogressivesubsidiesfrom high-paidto low-paidworkers,and the mere fact that somepeople in societyreceivemore generoushealthcare is necessarilynot a problem, so long as they pay for it. Whatmattersis that governmentsnot makeeveryone pay for what someare excludedex ante fromreceiving. Perversesubsidiesnot onlycause immediateinefficienciesor inequities,theyalso createintereststhat oppose subsequenthealth systemreform. This is evidentin the UnitedStates[Skocpol1995]and Chile [Musgrove 1995d],and it is the reasonwhythe designof subsidiesis a crucialpart of any reformto extendor improvecoveragefor the poor. The secondthing governmentsshouldnot do is tie publicfinanceto publicprovision. The choice of whetherto providecare throughpublic or parastatalfacilitiesshouldbe treated as a "makeor buy" decision,subordinateto the largerdecisionabout whatto pay for. That 57 does not necessarilymeaneliminatingpublicprovision,whichwill sometimesbe the best solution. It meansratherthat competitionbetweenpublic and privateprovidersshouldbe basedon costsand on quality,and not on price to the consumer,as is commonlythe case. If the volumeof public provisionweredeterminedby whenit wasactuallysuperior,insteadof by artificialprice differences,therewouldalmostcertainlybe less of it. Consumerswould use moreprivateprovision,if theycould obtainit at the sameprice or benefitfrom the same subsidy,as whengoingto publicproviders. Thesurvivingpublic facilitieswouldalmost surely be moreefficientin the use of resourcesand more satisfactoryto users. Another valuableconsequencewouldbe to givethe poor access,at least for someservices,to the same providersused by the rich, thus reducinginequity. To achievethose goals,however,requiresotherchangesin howpublic institutions operate,changesthat probablycannotoccur so longas publicfunds go automaticallyand exclusivelyto governmentfacilities. Thisconclusionis most pertinentto poor countries where publicsystemsare mostlikelyto be inefficientand to be used only becausethey are free or nearly so. High-incomecountrieswitha large shareof publicprovision,particularly in hospitals,sufferless from theseproblems. A third thinggovernmentsshouldnot do is pay forhealth care by fee-for-service, unless other mechanismsare used to controlexpenditures. This is seldoma problemwith publiclyprovidedservices,but makesit hard to controlcostswhengovernmentsfinance privateproviders. Even negotiatingor controllingthe fees is not enough, as the Canadian experiencedemonstrates,sinceproviderscanrespondto fees theyconsidertoo low,by increasingoutput[Evans,Barer and Labelle 1988]. Thishelpsexplainwhy Canadianhealth expenditurehas risen fasterthan that in Europeancountrieswhichrely on other payment mechanisms. Europeancountrieswhichpay for healthcare by fee-for-servicealso rely on globalbudgets,utilizationreviewsor other instrumentsof cost containment[OECD 1994, 1995]. Sinceexpendituresequalpricestimesquantities,and sincequantitiesof services respondto prices and are difficultto controldirectly,countrieswhich pursueboth macreconomicand microeconomicefficiencyinhealth spendingusuallycontrolbothprices and expenditures. In contrastto Europeanand NorthAmericanexperience,Brazil applies this systemto block grants forfederalexpenditurein stateswithrelativefinancialautonomy; but in stateswherefederalmoneyis paid directlyto providers,the federalgovernmentalso controlsone crucialquantity,the numberof hospitalizationsthat federalmoneywill pay for [WorldBank 1994b]. In all cases,the governmentsetspricesfor services. It also helps, in controllingexpenditures,to define "services"as completetreatments for specificconditionsratherthan as all the individualcomponentsof such treatments,so they canbe financedby mechanismssuch as the DiagnosticRelatedGroup(DRG)payments used in the UnitedStatesand the providercanbe requiredto assumepart of the financial risk. Changingfrom a systembased on overallbudgetsand on salariesfor providers,to a fee-for-servicesystemwithoutoffsettingthe resultingincentivesto over-provisioncan lead to an explosionof costs,as in the CzechRepublic[Boland1995]. 58 Finally,if governmentsmeanto pursuesomecombinationof betterhealth and lower costs, theyshouldnot--infactthey cannot--simplyfinancewhateverpeopledemandwhen care is free to consumers. Thisdoes not meangovernmentsshouldnot "subsidizedemand" rather than "subsidizingsupply"or providingservices. It meansonlythat there must be limitationson whatwill be paid for publicly:this issueis takenup at the end of Part 4. Suchlimitationsare a particularlycontentiousmatter,becauseboth goodhealthand cost contaimnentmaybe opposedto consumeror providersatisfaction,which are also politically importantobjectives. Nonetheless,twoempiricalobservationsare germaneto this decision. One is that privateinsurancealwayscarriessomelimitations,either as to the services coveredor as to cost-sharing. Exceptwhere povertyis important,so that cost-sharingis more difficult,there is no reasonfor publicfinanceto be systematicallymore generousin this regardthan privaterisk-sharingarrangements. The secondobservationis that, as the discussionconcerningFigure2 in Part 2 indicates,privatemarketswill tendto supplywhatpeopledemand,and public intervention typicallyactsto emphasizeneedsinstead. In other formsof public subsidy,it is commonto distinguishbetweenwantsand needs, and concentratespendingon the latter. This is the case, for example,withfood subsidies,whichare also verymuch health-related. Price subsidiesare usuallylimitedto "basic"foodstuffs,and foodstampscannot legallybe used to buy alcoholor other non-necessities.This maybe consideredpartly a matter of efficiency, assuringmore healthgain than wouldotherwiseoccur. But it is also a matter of equity: in contrastto actuarialprivateinsurance,whereevery purchaserbuysthe expectedvalueof the health servicesneeded,publicfinanceis involuntary.It comesfromtaxpayerswho havea legitimateinterestin meetingneeds, and therebygettingvaluefor their money,but not necessarilyin payingfor wants. None of these "don'ts"is easyto implement,becausesomeconsumeror provider interestscan be expectedto opposeevery one of them, in everycountry. But theyare arguablythe most importantconclusionsabout where the public/privatefrontier shouldrun in the health sector. It is notablethat none of theseconclusionsdependssolelyon features uniqueto the health sector. Thepeculiaritiesof healthmattermostfor the difficultyof distinguishing"needs"fromwants,and--partlyas a consequence--forthe dangersof paying providersfor whatevertheychooseto provide,withoutincentivesto controlcosts. A system which avoidedthe problemsdescribedabovewouldstill presentcomplexand difficult questionsfor the proper roleof the state,but it wouldhavemore latitudeto pursue improvements. And it wouldnot matterso much, exactlywhatobjectivesthe government pursuedor whichcombinationof instrumentsit applied. WhatGovernmentShouldDo Beyondthe prescriptionsforhowthe stateshoulddeal with the problemsof each of the three domainsof healthcare, several"dos" appearto be generallyvalidfor governments.If the objectiveis to minimizedeadweightlossesfrom public interventionand leaveas much roomas possibleforprivatechoices,then the first thing governmentsshoulddo is to use each less-intrusiveinstrumentto the point where a more intrusiveinterventionisjustified, followingthe sequenceof increasinglygreater 59 interference--inform,regulate,mandate,financeand provideservices. That is, govemnments shouldregulateprivateactivitywhenmerelyimprovingpeople'sinformationis not enough, deliverservicesif it is infeasibleto financeprivateprovidersequitably,and so on. Public financeis inescapablefor someactions,but particularlyin low- and middle-incomecountries much can probablybe accomplishedby betteruse of informationand regulation. Failureto use theseother instrumentswell, can increasethe need for publicfinance. Sometimesthe problemis that governmentsexploitthese instrumentstoo little.They do not regulateprivateinsurancewhenit first beginsto expand,which makessubsequent regulationpoliticallymore difficult[WorldBank 1994b;Musgrove1995d];or they do not initiallyreact whenhealth-damagingbehaviorssuch as smokingbecomemore entrenched. Sometimesthe problemis inappropriateregulation,whichneedlesslyrestrictscompetition,or enforcesinefficiencyin the publicsectorbycentralizingnearlyall decisions. Andpoor countries in particularoftenget the worstof bothworldsbypayingfor activitieswhich shouldbe, but are not, regulated:governmentsubsidyof medicaleducationwithoutadequate control of qualityor relationto needsis a commonexample. In the worstof cases, governmentsuse all the availableinstrumentsin exactlythe reverseorder. They try to providemore health care than theycanpay for, withthe result that mostservicesare under- financedand of poor quality;theytry to financeservicessomeof whichmightbe mandated and paid forby consumersor employers;theymandatecare, as by socialsecuritysystems, withoutadequatelyregulatingit; andthey do too little to informthe public and providers either of dangersto health or of howthe healthcare systemis actuallyworking. Muchof the criticismof governmentfailurein the health sector,especiallyin poor countries, describesthe result of gettingthingsbackward. These ideasare representedin Figure lOA,which showsthe appropriaterelation amongthe instrumentsbywhichthe statecan intervenein healthcare, and Figure lOB, whichportraysthe kind of inappropriateor imbalancedrelationoftenfoundin poor nations. Figure 1OAindicatesthat whateveris mandated,financedor providedpubliclyshouldalso be regulated,and much else besides. The statecan financesomecare that is not anyone's mandatedresponsiblity,and can also mandatecare or insurancecoveragewhichis financed privately. And most if not all publicprovisionshouldbe fully financedby govermnent,but possiblywith much more scopefor financethan for provisionbythe state. The largest sphere pertainsto information,coveringactivitiesin which there is no other public interferencewiththe market. In contrast,Figure 1OBshowsa much smallereffortto inform or regulate,very littleuse of mandatesthat are not alsopubliclyfinanced,and a sphere of provisionas large as, or evenlargerthan, that of finance. Of all the instrumentsof public action,regulationmaybe the mostunder-utilized. Braziland Chilebothprovideexamplesof the resultingproblems. In Brazil, the state financesthree-quartersof medicalcare but directlyprovidesonly aboutone-quarterof it, so the instrumentsof financeand provisionare used in the appropriateorder. But there is very little regulationof the competenceof medicalprofessionals,of the qualityof care, or of the rapidly-growingprivateinsuranceindustry. The lack of regulationeveninterfereswith Figure 1OaAppropriate Useof the Instruments of Public Intervention in Health Care 'El Information toconsumers, ~~providers andinsurers R Regulation ofpersonalbehavior firms,providersandinsurers Mandates for health care or m | | insurance coverage F I I 111 Financing of health care (explicit or implicit public insurance) P Provision of health care Figure 10b Typically Inappropriate Use of the Instruments of Public Intervention in HealthCare L............J providersandinsus . /~~~~ Regulato of persnl beavo finformspidr an insurers, jps Mandatesforhealthcareor M insurancecoverage F Financingof healthcare(explicit o'rimplicitpublicinsurance) P Provisionofhealthcare 62 financingthe system,sinceprivateinsurerssometimessend their customersto publicly- financedfacilitieswithoutpayingforcare, anduntil recentlyfraud waswidespread[World Bank 1994b;Medici and Czapski 1995]. In Chile,the privateinsuranceindustrywascreated by publicaction, withessentiallyno regulation--butwitha mandateallowingpeople to spend on privateinsurance,the tax contributionsthat formerlywereused to financethe public system[WorldBank 1994a]. The lackof regulationmaynot haveaffectedmedicalquality, but it has worsenedthe financialsituationof publicfacilities,raisedadministrativecosts, and promotedrisk selection. Evenmore seriousfailuresto regulatehavearisenin Eastern and CentralEurope, as formerstatemonopoliesof health financeand provisionhavegiven way to competitiveprivateprovision. A secondthing governmentsshoulddo is to stimulatecompetitionin the provisionof health care. This is largelya matterof promotingpublic/privatecompetition,for the reasons describedabove,but it also includesremovingany unjustifiedbarriers to competitionwithin the privatesector,and betweenfor-profitand non-profitproviderssuch as non-governmental organizations. The lack of competitionis usuallyless of a problemin the domainof low- cost, health-relatedactivitiesthan in the domainof costlieractivitiesrequiringinsurance. This recommendation extends a fortion to non-medical components of health care such as the "hotel"servicesof hospitals. How far competitionshouldbe carried is not alwaysobvious. Forexample,whetherpublicfacilitiesshouldmaketheir ownpurchasingdecisionsfor such inputsas drugs,dependsonwhethercentralbulkpurchasingreducescostswhilemaintaining adequatesupplies. Even in the latter case, there shouldof coursebe competitonamong suppliersfor such purchases. Exceptfor the risk that providerswillcompetebyofferingmore servicesrather than by raisingqualityor reducingcosts--arisk that is greatestwhenpaymentis by fee-for- service--competitionappearsto be beneficialin health care provision,just as in other industries. Competitionis less desirablein health carefinancing,bothbecauseadministrative costsare likelyto be higherand becauseit is competitionamonginsurersthat leads to risk selection. Experiencein OECDcountriessuggeststhat goodresultscanbe obtainedwithone or withmany insurers,but only if theyare closelyregulated. Third, governmentsshouldput as much of the incentivefor cost containmentas possibleon the supplyside of the market, ratherthan on consumers. This is almosta necessitywhere poor consumersare concerned,since their povertyalreadysharplylimits what theycan spend. But the evidenceis that evennon-poorconsumersdo not respondto higherpricesby usinghealthservicesmore cost-effectively.It also appearsthat providers haveconsiderablescope forcontrollingexpenditurebylimitingvolumeas well as unit costs. And theoryindicatesthat an optimalpaymentsystemshoulduse supply-sidemeasuresto controlcosts; reimbursingprovidersfully accordingto costsis nleverthe best solution [Ellis and McGuire 1990, 1993]. As incomeincreasesand povertydeclines,of course, it becomes easierto pass the burdenof cost containmentto consumers. However,it does not become medicallyanymore effectiveor economicallyany more efficientto do so. Moreover,as 63 income increases the capacity for supply-side responses by providers also increases, so it continues to be preferable to keep cost control incentives on the supply side of the market. Finally, it is urgent to deal with the pervasive problem of government failure, and to improve the capacity to do whatever government ends up doing. This is especially important when market failures are so important that for the public sector to withdraw, on the ground that it also is subject to failure, would only make matters worse. Much of the criticism that governments, particularly in poor countries, try to do too much in health arises because of how badly they appear to operate, more than from any evidence that they have exceeded some optimum degree of state intervention. More skill and understanding, and fewer internal barriers to efficiency,make sense whether the state's role shrinks, as by leaving more provision to the private sector, or expands, to finance more coverage for the poor. One of the things governments generally need to do better, particularly in poor countries where much private medical practice may be of low quality, is to use regulation, mandates, training and other interventions to help the private sector function better. This is increasingly recognized as an essential component of almost any health sector reform, and the need for it will only increase as health systems become more expensive and complex. All five instruments of intervention--information, regulation, mandates, finance and provision--need to be used well, and using less of one instrument and more of another will not, in general, reduce the need for governments to perform capably. How to Spend Public Money on Health Care Governmentsthat promise, or try, to pay for all the health care that people want, find that they cannot do it. Either total expenditure expands more than the population is politicallywillingto pay,or non-pricerationingbecomesincreasinglysevere. Tomitigate these problems,it is crucialto set somecriteria for the use of publicexpenditureon health care. This is particularlyimportant,giventhe tendencyfor public financingto expandas countries become richer and the health sector becomes more complex. As the evidence in Part 3 indicates, it is difficult to avoid the need for public choices about expenditure by financing relatively little care, relying on other mechanisms such as regulation, and leaving such decisions to private markets. A first rule about how to spend public funds (including those mandated by earmnarked taxes such as social securitycontributions)is to givepriorityto highlycost-effectivepublic goods. Thisadvice is mostpertinentfor poor countries,where these activitiescould absorb (nearly)all of what the statenowspendsonhealth, and wherethere is the least doubtabout what shouldbe includedin a publicly-financedpackageof services. However,it is still relevantfor rich countries,despitethe very smallshareof spendingrepresented,if theyhave not achieveduniversalcoverageof suchservicesas immunizationor exploitedall the justifiablepossibilitesfor influencingbehaviorthroughinformationand regulation. Oncepublicgoodsand serviceswith substantialexternalitiesare adequatelyfinanced, the questionbecomeshowto spendpublicfunds on individual,clinicalserviceswhich are 64 largelyprivategoods. At lowincomelevels,the best answerstartswithdefiningan "essentialpackage"of services,witheverythingoutsidethe packageregardedas "discretionary"and to be paid forprivately[WorldBank 1993b,Chapter3]. If the criterion for what goes intothe packageis cost-effectiveness,this approachmaximizesthe potential health gain. A packagecan also includeonly interventionsknownto be highlyeffective, with less regardfor their cost, as in the revisedversionof the publicly-financedMedicaid Plan for the poor in the stateof Oregon[Hadorn1991]. Thishas the effectof givinglife- savinginterventionsfor a fewindividualspriorityoverlow-cost,low-gaincare for large numbersof people: in effectit treatsgainsin health non-linearlyacrossbeneficiaries. However,there is alwaysa trade-offbetweenfinancingsomeservicesfor the entire population,irrespectiveof people'scapacityand willingnessto pay for them, and financing more servicesfor the poor. Makingthis distinctionrequiresdiscriminatingamong beneficiariesin someway,suchas by geographicdifferencesin the servicesfinancedor differentialuser charges. Theevidenceon howeasilyand effectivelythat can be done is mixedand incomplete,as Part 3 indicates,so the onlygeneralrule that can be suggestedis not to use public funds to pay for a servicefor non-poorbeneficiariesunless the poor have accessto the sameservice on equalor more favorableterms [Bobadilla,Cowley,Musgrove and Saxenian1994]. Servicesthat fall in the domainof low-costactivitiesnot requiring insuranceshouldbe subsidizedonlyfor the poor; serviceswhichmaybe catastrophically costlybut are still cost-effectivebecausetheyyieldlargehealth gains,can be financed--but obviouslynot subsidized--foreveryone. As countriesbecomericher,there are three broadchoices to follow. One is to continueto use the samecriteria of costs,effectivenessand equity,but to covera wider range--amore generouspackage--ofpublicly-financedcare, maintainingthe distinction betweenwhat is essentialand what is discretionary.There are twochief limitationsto this approach. One is the difficultyof choosingamongthe thousandsof medicalprocedures available,includingthe need forrepeateddecisionsas newproceduresare invented. This approachis followedin Canada,and in the publicly-fundedprogramsof the UnitedStates. In practice,the rangeof interventionsfundedpubliclybecomesvery large. The other limitationis the necessityto regulatethe providersthat deliverthe packageand ensure that publicmoneyactuallygoes to the desiredinterventionsand is not spenton thoseconsidered discretionary. Unfortunately,this is easiestunder fee-for-servicepaymentto private providers,where the governmentcanchooseto pay for someinterventionsbut not for others. It is probablyhardestunder publicprovisionwithbudgetsthat do not dependon service output, where economicincentivesare weakand simplyexhortatingprovidersto deliverthe services in the packagemaybe ineffective[Musgrove1995a]. A secondapproachis to maintaina smalleressentialpackage,whichwill not exhaust the availablepublic funds,and leavemorediscretionto providersas to what to do with publicmoney--subjectto overallspendingcontrolssuchas globalbudgetsor other incentives for cost containment. Thisrequiresless explicit,centraldecisionand maytakebetter accountof the variationamongpatientsbyleavingmoredecisionsto doctors. One advantage 65 of this approachis to makeit easierto "include"a particularprocedurein the packageand still limitthe numberof peoplewho receiveit, whichmay reduceinappropriatecare and avoidthe politicaldifficultiesof explicitexclusion. SeveralEuropeancountries,including both directfinancesystemssuch as the UnitedKingdomand mandatedsocialinsuranceas in Germany,havefollowedthis approach. Whenproviderscandeliveralmostany service they choose,but are constrainedby a totalbudget,the result is oftenqueueingor waitinglists for non-emergencycare. This clearlyreducespatientsatisfaction,but it canbe medically efficient. And the totalwelfarelossto societymaybe no largerthanwhen such controlsare absentand some resourcesare wastedon excessuse of medicalcare [Feldman1994]. A third possibilityis to circumscribepublicfundingto a smallessentialpackageand alloweverythingelseto be financedprivately,subjectto somecombinationof mandatesand regulation. Underthis approach,the poor wouldbe coveredonly for the essentialpackage. No countryhas really succeededwiththis approachin controllingadverseselectionor in holdinghealth expendituresteadyas a shareof income. And forcountrieswhich have alreadyreacheda large share of expenditurethroughsocialinsurance,reducingthe public shareand expandingprivateinsurancewouldprobablyraise costsand increaseinequity. Supplementalprivateinsurancecan alwaysexistat the margin,evenin countriessuch as Canada withgenerouspublicfinance[Hicks1996];the choicefor publicpolicyis where to draw the margin. Withsomeexceptions,high-incomecountrieshavedrawnit to includea large public sharein finance. The questionis whetherlower-incomecountriescould copy the successof such arrangementswitha significantlysmallerdependenceon government finance. There is littleempiricalevidenceas to whetherthat wouldbe easyto do. The notionof a packageas a wayto determinehowto use publicresourcesfor health, followsthe distinctionamongdomainsof healthcare developedin Part 2. That is, it is based on the properties of the healthinterventionsthemselves. So far as any distinctionis made among population groups, it is limitedto favoring--orat least not discriminating against--thepoor. This followsfromthe distinctionbetweenservicesthat the non-poorcan affordout of pocket, and thosewhichwouldbe catastrophicallycostlyfor much of the population. In consequence,a packageof publicly-financedhealthcare shouldnot be based in the first instanceon the notionof "vulnerablegroups"in the population,definedbyany other criterionsuch as ageor sex. There is noethicalbasisfor givingpreferenceto certain populationgroups independentlyof their healthproblemsand of the scopefor state intervention. Moreover,as the discussionof "meritgoods"indicates,manyexamplesof "vulnerability"correspondto otherjustificationsfor a publicrole. This is the case for some publicgoodswhichpreferentiallybenefitcertaingroups,and for povertyin general. Neither the "dos" nor the "don'ts"suggestedhere specifyjust howmuch health care shouldbe publiclyfinanced(or provided),or which taxesto use, or whichpayment mechanisms. Stillless do theydescribethe idealmixtureof mandates,regulationsand informationgovernmentsshouldadopt. Countriesdisplaya varietyof choiceson these matters. Nonethelessit is clear that somestate interventionin the health sectoris highly justified, and that much could be gainedbyavoidingcertaincommonand costlymistakesof 66 inadequate,excessiveor inappropriateinterferencein health care markets. If the frontier betweenprivateand public activityis properlydrawn, the statecanperform better and therebyhelp ensurethat the health sectoras a wholeperformsbetter. 67 References Aaron, Henry J. 1994. Issues every plan to reform health care financing must confront. Journal of Economic Perspectives. Vol. 8, No. 3. Abel-Smith, Brian. 1995. Assessing the experience of health financing in the United Kingdom. In Dunlop and Martins 1995. Abel-Smith, Brian. 1984. Cost containment in 12 European countries. Wrld Health Statistics Quarterly. Vol. 37. Arrow, Kenneth J. 1963. Uncertainty and the welfare economics of medical care. American Economic Review. Vol. 53, No. 5. Arrow, Kenneth J. 1985. Theoretical issues in health insurance. In Collected Papers of Kenneth J. Arrow, Vol. 6, Applied Economics. Cambridge: Harvard University Press. Baker, Judy, and Jacques van der Gaag. 1993. Equity in health care and health care financing: evidence from five developing countries. In Van Doorslaer, Wagstaffand Rutten 1993. Barnum, Howard, Joseph Kutzin and Helen Saxenian. 1995. Incentives and provider payment methods. The World Bank. HRO Working Paper 51. Barr, Nicholas. 1992. Economic theory and the welfare state: a survey and interpretation. Journal of Economic Literature. Vol. 30, No. 2. Beales, Howard, Richard Craswell, and Steven C. Salop. 1981. The efficient regulation of consumer information. Journal of Law and Economics. Vol. 24, No. 3. Beaton, George H., R. Martorell, K. A. L'Abbe, B. Edmonston, G. McCabe, A. C. Ross and B. Harvey. 1993. Effectiveness of Vitamin A Supplementation in the Control of Young Child Morbidity and Mortality in Developing Countries: Summary Report. Toronto: University of Toronto. Behrman, Jere. 1990. The action of human resources and poverty on one another: what we have yet to learn. Washington, D. C.: World Living Standards Measurement Study, Working Paper No. 74. Bennett, Sara. 1991. The mystique of markets: public and private health care in developing countries. London: London School of Hygiene and Tropical Medicine, Public Health and Policy Departmental Publication No. 4. 68 Bennett, Sara, George Dakpallah, Paul Garner, Lucy Gilson, Sanguan Nittayaramphong, Beatriz Zurita and Anthony Zwi. 1994. Carrot and stick: state mechanisms to influence private provider behavior. Health Policy and Planning. Vol 9, No. 1. Birdsall, Nancy, and Estelle James. 1992. Health, government and the poor: the case for the private sector. Washington, D. C.: World Bank, Policy Research Working Paper No. 938. Bitran, Ricardo, and D. McInness. 1991. Health care demand in Latin America: lessons drawn from the Dominican Republic and El Salvador. Washington, D. C.: World Bank, Economic Development Institute (processed). Blendon, Robert J., Mollyann Brodie and John Benson. 1995. What happened to Americans' support of the Clinton Plan ? Health Affairs. Summer. Bobadilla, Jose Luis. 1996. Searching for Essential Health Packages in Low- and Middle- Income Countries. Washington, D. C.: World Bank (processed). Bobadilla, Jose Luis, Peter Cowley, Philip Musgrove and Helen Saxenian 1994. Design,content and financing of an essential national package of health services. In World Health Organization 1994. Boland, Vincent. 1995. Czech health minister sacked in costs crisis. Financial Times. 10 October. Bovbjerg, Randall R., Charles C. Griffin and Caitlin E. Carroll. 1993. U.S. health care coverage and costs: historical development and choices for the 1990s. Journal of Law, Medicineand Ethics. Vol. 21, No. 2. Campos, Ant6nio Correia, Carlos Costa, Francisco Ramos, Luciano Patrao, M. Fatima Cortes, Manuel Roque, M. Rosario Giraldes and Rogerio de Carvalho. 1987. A Combinagao Plablico-Privadaem Saude:Privilegios,Estigmase IneficiMncias. Lisbon: Escola Nacionalde SauidePublica. Chabot, Jarl, Manuel Boal and Augusto da Silva. 1991. National community health insurance at village level: the case from Guinea-Bissau. Health Policy and Planning. Vol. 6, No. 1. Chernick, Howard A., Martin R. Holmer and Daniel H. Weinberg. 1987. Tax policy toward health insurance and the demand for medical services. Journal of Health Economics. Vol. 6. Coase, Ronald. 1988. The Firm, the Market, and the Law. Chicago: University of Chicago Press. Cochrane, A. L., A. S. St. Leger and F. Moore. 1978. Health service "input" and mortality "output" in developed countries. Journal of Epidemiology and Community Health. Vol. 32. 69 CongressionalBudgetOffice,U. S. Congress. 1994.An analysisof the Administration's healthproposal. Creese, Andrew.1991.User chargesforhealth care: a reviewof recentexperience.Health Policy and Planning. Vol. 6, No. 4. Culyer,A. J. 1990.Cost containmentin Europe.In Organizationfor EconomicCooperation and Development. 1990. Health Care Systems in Transition: The Searchfor Efficiency. SocialPolicyStudiesNo. 7. Paris: OECD. Culyer,A. J. 1995.Need:the ideawon'tdo - but westill need it. SocialScienceand Medicine.Vol. 40., No. 6. Cumper, George E. 1991. The Evaluation of National Health Systems. New York: Oxford UniversityPress. Cutler, DavidM. 1994.A guide to healthcare reform.Journalof EconomicPerspectives. Vol. 8, No. 3. Departmentof PublicHealth& Policy.1993.HealthPolicyUnit. Reportof the Workshopon The Public/Private Mix for Health Care in Developing Countries. London: The London Schoolof Hygiene& TropicalMedicine. Diamond,Peter. 1992.Organizingthe healthinsurancemarket.Econometrica.Vol.660, No. 6. Dunlop, David W., and Jo. M. Martins, eds. 1995. An International Assessment of Health Care Financing: Lessons for Developing Countries. Washington, D. C.: World Bank, EconomicDevelopmentInstitute. Ellis, RandallP., and ThomasG. McGuire.1990.Optimalpaymentsystemsfor health services. Journal of Health Economics. Vol. 9. Ellis, RandallP., and ThomasG. McGuire.1993.Supply-sideand demand-sidecost sharing in health care. Journal of Economic Perspectives. Vol. 7, No. 4. Elola, Javier,AntonioDaponteand VicenteNavarro.1995.Healthindicatorsand the organizationof health care systemsinWesternEurope. AmericanJournalof PublicHealth. Vol.85, No. 10. ExpandedProgrammeon Immunization(EPI). 1993.Newsletter.Washington,D. C.: Pan AmericanHealthOrganization.December. 70 Expanded Programme on Immunization (EPI). 1995. Newsletter. Washington, D. C.: Pan American Health Organization. April. Evans, Robert G., Morris L. Barer, and Roberta J. Labelle. 1988. Fee controls and cost control: tales from the frozen north. Milbank Memorial Fund Quarterly. Vol. 66, No. 1. Evans, Robert G., and Maureen M. Law. 1995. The Canadian health care system: where are we and how did we get there? In Dunlop and Martins 1995. Feldman, Roger. 1994. The cost of rationing medical care by insurance coverage and by waiting. Health Economics. Vol. 3. Feldstein, Martin S. 1973. The welfare loss of excess health insurance. Journal of Political Econonmy.Vol. 81. Foldvary, Fred E. 1994. Public Goods and Private Communities. Cornwall, England: Hartnolls Limited. Frank, Richard G., and David S. Salkever. 1994. Nonproft organizations in the health sector.Journalof EconomicPerpectives.Vol. 8, No. 4. Fuenzalida-Puelma, Hernan L., and Susan Scholle Connor. 1989. TheRight to Health in the Americas: a Comparative Constitutional Study. Washington, D. C.: Pan American Health Organization. Geoffard, Pierre-Yves, and Tomas Philipson. 1994. Market structure and disease eradication: private versus public vaccination. Chicago: Department of Economics, University of Chicago (processed). Gertler, Paul, and Jacques van der Gaag. 1990. The Willingness to Payfor Medical Care: Evidence from TwoDeveloping Countries. Baltimore, Maryland: The Johns Hopkins University Press for the World Bank. Govindaraj, Ramesh, Christopher J. L. Murray and Gnanaraj Chellaraj. 1994. Health Expendituresin Latin America. Cambridge: Harvard Center for Population and Development. Griffin,CharlesC. 1987. UserChargesfor HealthCarein Principleand Practice. Washington, D. C.: World Bank Economic Development Institute. Seminar Paper No. 37. Griffin,CharlesC. 1992.Health CareinAsia: A ComparativeStudyof Cost and Financing. Washington, D. C.: World Bank. Grosh, Margaret. 1992.AdministeringTargetedSocial ProgramsinLatin America:From Platitudes to Practice. Washington, D. C.: World Bank. 71 Hadorn, D. C. 1991.Settinghealthcare prioritiesin Oregon:cost-effectivenessmeetsthe Ruleof Rescue.Journalof the AmericanMedicalAssociation.Vol.265. Haveman,Robert, and BarbaraWolfe.1993.Children'sprospectsand children'spolicy. Journalof EconomicPerspectives.Vol.7, No.4. Heysen,Socorro,and PhilipMusgrove.1986.Interdepartmentaldifferencesin life expectancyat birth in Peru as it relatesto income,householddrinkingwater,and provision of medicalconsultations.Bulletinof the PanAmericanHealthOrganization.Vol. 20. Hicks,Vernon.1996. Privatecommunicationto WilliamMcGreevey,25 January. Houseof Commons.1995.Scienceand TechnologyCommittee. HumanGenetics:Minutes of Evidence. Wednesday17May.London:Her Majesty'sStationeryOffice. Jamison,Dean, W.Henry Mosley,AnthonyR. Meashamand Jose-LuisBobadilla,eds. 1993.Disease ControlPrioritiesin DevelopingCountries. Oxford:OxfordUniversityPress. Johns, T., S. L. Boothand H. V. Kuhnlein.1992. FactorsinfluencingvitaminA intakeand programmesto improvevitaminA status.Foodand NutritionBulletinVol. 14No. 1. Kim, Kwanrgkee,and PhilipM. Moody. 1992.More resourcesbetter health ? a cross- nationalperspective.SocialScienceand Medicine.Vol. 34, No. 8. Lachaud,Claire,and LiseRochaix. 1993.France. In VanDoorslaer,Wagstaffand Rutten 1993. Lavy,Victor. 1994.Householdresponsesto publichealth services:cost and quality tradeoffs.Conferenceon socialsectorinvestmentsheld at the WorldBank, Washington,D. C., 18May (processed). Lazenby,HelenC., KatharineR. Levit, DanielR. Waldo,GeraldS. Adler,SuzanneLetsch and CathyA. Cowan.1992.Nationalhealth accounts:lessonsfromthe U. S. experience. Health CareFinancingReview.Vol. 13, No. 4. Litvack,Jennie,and ClaudeBodart. 1993.User fees and improvedqualityof health care equalsimprovedaccess:resultsof a fieldexperimentin Cameroon.SocialScienceand Medicine.Vol. 37, No. 3. Lohr, K. N., R. H. Brook, C. J. Kamberg,G. A. Goldberg,A. Leibowitz,J. Keesey,D. Reboussinand J. P. Newhouse.1986.Effectof cost-sharingon use of medicallyeffective and less effectivecare. MedicalCare.Vol.24. 72 Makenbach,J. P. 1991.Healthcare expenditureand mortalityfrom amenableconditionsin the EuropeanCommunity.HealthPolicy.Vol. 19. Makinen,Marty,and LauraRaney.1994.Role andDesirabilityof UserChargesfor Health Services.Bethesda,Maryland:Abt Associates,Inc. McGreevey,William.1990.SocialSecurityinLatinAmerica,Issues and Optionsfor the Bank. Washington,D. C.: WorldBankDiscussionPapers, 110. McGuire,Alistair,DavidParkin,DavidHughesand KarenGerard. 1993.Econometric analysesof nationalhealthexpenditures:canpositiveeconomicshelp to answernormative questions? HealthEconomics.Vol.2. McLachlan,Gordon,and AlanMaynard,eds. 1982.The Public/PrivateMixfor Health: the Relevanceand Effectsof Change.London:NuffieldProvincialHospitalsTrust. McPake,Barbara,KaraHansonand AnneMills. 1993.Experienceto date of Implementing the BamakoInitiative:a Reviewand Five CountryCaseStudies. London:LondonSchoolof Hygieneand TropicalMedicine. Medici, Andr6Cezar, and ClaudioAndreCzapski. 1995.Evolu,do e Perspectivasdos GastosPiublicosem Sauideno Brasil.Consultants'Report.Washington,D. C.: WorldBank. Mesa-Lago,Carmelo.1992.HealthCarefor the Poorin LatinAmericaand the Caribbean. Washington,D. C.: Pan-AmericanHealth Organizationand Inter-AmericanFoundation. Milgrom,Paul, and JohnRoberts. 1992.Economics,Organizationand Management. EnglewoodCliffs, NewJersey:PrenticeHall. Miller,Jonathan. 1978.TheBody in Question.NewYork:RandomHouse. Murray,ChristopherJ. L., RameshGavindarajand PhilipMusgrove.1994.Nationalhealth expenditures:a global analysis.In WorldHealthOrganization1994. Murray,ChristopherJ. L., Alan D. Lopezand Dean. T. Jamison. 1994.The globalburden of diseasein 1990:summaryresults,sensitivityanalysisand futuredirections.In World Health Organization1994. Musgrave,RichardA. 1959.The Theoryof PublicFinance,A Study in PublicEcononry. New York:McGraw-Hill. Musgrove,Philip. 1986.Measurementof equityin health. V,brldHealth StatisticsQuarterly. Vol, 39, No. 4. 73 Musgrove,Philip. 1995a.Cost-effectivenessand health sectorreform.Washington,D. C.: WorldBank, HCO WorkingPaperNo. 48. Musgrove,Philip. 1995b.Cost-effectivenessand the socializationof health care. Health Policy.Vol. 32. Musgrove,Philip. 1995c.Mismatchof need, demandand supplyof services:picturing differentwayshealth systemscan go wrong.Washington,D. C.: WorldBank, HCO WorkingPaper No. 59. Musgrove,Philip. 1995d.Reformasal sectorsaluden Chile: contexto,l6gicay posibles caminos.In Giaconi,Juan G., ed., La Saluden el SigloXXI: CambiosNecesarios.Santiago, Chile: Centrode EstudiosPublicos. Navarro,Vicente.1989.Whysomecountrieshavenationalhealth insurance,others have nationalhealth services,and the U. S. has neither.Social Scienceand Medicine.Vol. 28, No. 9. Newhouse,JosephP., et. al. 1993.Freefor All? Lessonsfrom the RANDHealthInsurance Experiment.Cambridge:HarvardUniversityPress. Nolan, Brian, and Vincentlbrbat. 1995. CostRecoveryin PublicHealth Servicesin Sub- SaharanAfrica. Washington,D. C.: The WorldBank,EconomicDevelopmentInstitute. Okun, Arthur M. 1975.Equalityand Efficiency:the Big Thrdeoff.Washington,D.C.: The BrookingsInstitution. Organizationfor EconomicCooperationand Development(OECD). 1992. TheReformof Health CareSystems:a ComparativeAnalysisof SevenOECDCountries.Paris: OECD. Organizationfor EconomicCooperationand Development(OECD). 1994. 7TheReformof Health CareSystems:a Reviewof SeventeenOECDCountries.Paris: OECD. Organizationfor EconomicCooperationand Development(OECD). 1995.NewDirectionsin Health CarePolicy.Health PolicyStudiesNo. 7. Paris: OECD. PanAmericanHealth Organization(PAHO).1991.ThePublicPrvate Mix in National Health Systemsand the Role ofMinistriesofHealth. Washington,D. C.: PAHO. PanAmericanHealth Organization(PAHO).1994.Health Conditionsin the Americas. Washington,D. C.: PAHO. Pauly,Mark V. 1968.The economicsof moralhazard.AmericanEconomicReview.Vol.58. 74 Pauly,Mark V. 1974.Overinsuranceand publicprovisionof insurance:the roles of moral hazardand adverseselection.QuarterlyJournalof Economics.Vol. 88. Pauly,Mark V. 1986.Taxation,health insurance,and marketfailurein the medical economy.Journalof EconomicLiterature.Vol.24. Pauly,Mark V. et. al. 1988.Issuesrelatedto volumeand intensityof physicianservices. Reportpreparedfor the U. S. HealthCare FinancingAdministration.Baltimore. Pelletier,David,L. 1994.Therelationshipbetweenchildanthropometryand mortalityin developingcountries:implicationsforpolicy,programsand futureresearch.Journalof Nutrition. Vol. 124, No. 10S. Pouillier,Jean-Pierre.1992.Administrativecostsin selectedindustrializedcountries.Health CareFinancingReview.Vol. 13, No.4. Reinhardt,UweE. 1992.Organizingthe health-caresector: regulatoryvs. market approaches.Princeton:WoodrowWilsonSchool,PrincetonUniversity(processed). Ricklefs,RobertE., and CalebE. Finch. 1995.Aging:a NaturalHistory.New York:W. H. Freeman, ScientificAmericanLibrary. Russell,Louise B. 1986.Is PreventionBetter Than Cure?Washington, D. C.: The BrookingsInstitution. Saltman, Richard. 1995.ApplyingPlannedMarketLogic toDevelopingCountries'Health Systems:an InitialExploration.DiscussionPaperNo. 4, Forum on HealthSectorReform. Geneva:WorldHealthOrganization. Schieber,George,Jean-PierrePouillier,and LeslieM. Greenwald.1992. U. S. health expenditureperformance:An internationalcomparisonand data update. Health CareFinancingReview,InternationalComparisonofHealth Systems.Vol. 13, No. 4. Schieber,George,Jean-PierrePouillier,and LeslieM. Greenwald.1994.Health System Performancein OECDCountries,1980-1992.HealthAffairs. Fall. Skocpol,Theda. 1995.Therise and resoundingdemiseof the ClintonPlan. HealthAffairs. Spring. Shaw,R. Paul, and CharlesC. Griffin.1995.FinancingHealthCarein Sub-SaharanAfrica throughUserFeesand Insurance.Washington,D. C.: WorldBank, Directionsin Development. 75 Shea, DennisG., and R. PatrickStewart.1995.Demandfor insuranceby elderlypersons: privatepurchasesand employerprovision.HealthEconomics.Vol.4. Srinivasan,T. N. 1994.Destitution:A Discourse.Journalof EconomicLiterature.Vol. 32. Stiglitz,Joseph. 1989.Principaland agent.In Eatwell,John, MurrayMilgateand Peter Newman,eds. TheNew Palgrave:Allocation,Informationand Markets. NewYork:W. W. Norton. Summers,LawrenceH. 1989.Somesimpleeconomicsof mandatedbenefits.American EconomicsAssociationPapersand Proceedings.Vol. 79, No. 2. Tresseras,R., J. Canela,J. Alvarez,J. Sentisand L. Salleras.1992. Infantmortality,per capitaincome,and adultilliteracy:an ecologicalapproach.AmericanJournalof Public Health. Vol.82, No. 3. U. S. Departmentof Healthand HumanServices.1989Reducingthe Health Consequencesof Smoking:25 years of Progress.A Reportof the SurgeonGeneral.Rockville,Maryland: PrepublicationVersion,DHHSPublicationNumber(CDC)89-8411. U. S. Departmentof Healthand HumanServices.1992.Smokingand Healthin the Americas.Atlanta,Georgia:DHHSPublicationNumber(CDC)92-8419. Vander Gaag, Jacques. 1995.Public and PrivateInitiatives:W,rking Togetherfor Health andEducation.Washington,D. C.: WorldBank, Directionsin Development. VanDoorslaer,Eddy,AdamWagstaffand Frans Rutten. 1993.Equity in the Financeand Deliveryof Health Care:an InternationalPerspective.Oxford:OxfordUniversityPress. Vogel,RonaldJ. 1993.FinancingHealthCare in Sub-SaharanAfrica: A PolicyStudy.Silver Spring, Maryland:Basic HealthManagement,Inc. for the AfricanDevelopmentBank. Weisbrod,BurtonA. 1991.The HealthCare Quadrilemma:An Essayon Technological Change,Insurance,Qualityof Care, and Cost Contaimnent.Journalof EconomicLiteratwe. Vol. 29, No. 2. pp. 523-552. Wolfe,BarbaraL. 1986.Healthstatusand medicalexpenditures:is there a link? Social Scienceand Medicine.Vol. 22, No. 10. WorldBank. 1980.Health:SectorPolicyPaper.Washington,D. C.: WorldBank. WorldBank. 1987.FinancingHealthServicesin DevelopingCountries:An Agendafor Reform.A WorldBankPolicyStudy.Washington,D. C.: WorldBank. 76 WorldBank. 1991. VWrldDevelopmentReport:The ChallengeofDevelopment.Washington, D. C.: WorldBank. WorldBank. 1992a.VKrldDevelopmentReport:Developmentand the Environment. Washington,D. C.: WorldBank. WorldBank. 1992b.China:Long-TermIssuesand Optionsin the Health Transition. Washington,D. C.: WorldBank. WorldBank. 1993a.Argentina:PublicFinanceReview--fromInsolvencyto Growth. Washington,D. C.: WorldBank. ReportNo. 10827-AR. WorldBank. 1993b. VWorldDevelopmentReport:Investingin Health. Washington,D. C.: WorldBank. WorldBank. 1993c.The Organization,DeliveryandFinancingof Health Carein Brazil. Washington,D. C.: WorldBank. ReportNo. 12655-BR. WorldBank. 1994a.Chile:TheAdultHealthPolicyChallenge.Washington,D. C.: World Bank. ReportNo. 12681-CH. WorldBank. 1994b.PrivateSectorand SocialServicesin Brazil:WhoDelivers,Who Pays, WhoRegulates.Washington,D. C.: WorldBank. ReportNo. 13205-BR. WorldBank. 1995.The WorldBankin the HealthSector.Washington,D.C. WorldHealth Organization.1994. GlobalComparativeAssessmentsin the HealthSector: DiseaseBurden,ExpendituresandInterventionPackages.Geneva:WorldHealth Organization. Zeckhauser,Richard. 1994.Publicfinanceprinciplesand nationalhealthcare reform. Journalof EconomicPerspectives.Vol. 8, No. 3. 77 Data Sources--notcitedin the text World Bank. 1992. Financing Health Services - Zimbabwe. TechnicalWorking Paper. WorldBank. 1992.PharmaceuticalExpendituresand Cost-RecoverySchemesin Sub- Saharan Africa. Washington, D. C.: World Bank, TechnicalWorking Paper No. 4, Africa Technical Department. WorldBank. 1992. GovernmentFinancialStatisticsYearbook. World Bank. 1991. Health Care Cost, Financing and Utilization- Nigeria. World Bank. 1992. Health Insurance Issues in the 1990s - Indonesia. Sector Report. World Bank. 1993. Health Sector Review - Venezuela. World Bank. 1994. Health Sector Review: Present Status and Future Options - Jamaica World Bank. 1991. Health Sector Study: Key Concerns and Solutions - Pakistan. WorldBank. 1993.IslamicRepublicof PakistanHealthSectorStudyKeyConcernsand Solutions. Population and Human Resources Division, Country Department 3, South Asia Region. WorldBank. 1994.PolicyandFinanceStrategiesfor StrengtheningPrimaryHealth Care Services- India. Sector Report. World Bank. 1990. Population, Health, and Nutrition Sector Report. Mozambique. World Bank 1995 PovertyAssessment - Tunisia. World Bank. 1993. Review of Social Sector Issues - Nicaragua. 78 STATISTICAL ANNEX Each observationin Figures3 through9 refersto a country;observationsare limitedto those countriesfor whichboth publicand privateexpenditureare estimateddirectly (not from regressions). Incomeestimatesare for 1991,expenditureand demographicestimatesfor 1990or 1991;estimatesare sometimesbased on data for a year as closeto 1990as possible, betwteen1983and 1987,withadjustmentproportionalto changesin GDP for expenditure data. The variablesused to generateFigures3 through9 are shownin AnnexTable1 and are describedbelow. The tablelists all countriesforwhichdata were used in any of the Figures;the numberof countriesforwhichdata were availablevariesfromone Figure to another. Income (italicizedfiguresarederivedfrom regressionestimates) 1. Incomeper capita, purchasingpowerparity (PPP)dollars 2. Incomequartile, amongallcountries(notjust those for whichhealthexpenditure is estimated). Values= 1 (poorest),2, 3, 4 (richest) HealthStatus 3. Life expectancyat birth (years) 4. Child mortalityunder agefive(per thousandchildrenin agegroup) Healthexpenditureand public/privateshares 5. Totalhealthexpenditureas shareof GDP 6. Publicexpenditureon health as shareof totalhealth expenditure 7. Publicexpenditureshare quartile,amongall countriesfor whichvar (6) is available. Values= 1 (lowest),2, 3, 4 (highest) Compositionof finance 8. Directly-financedpublicexpenditure(expenditureby Ministryof Health and all other publicagenciesexceptsocialsecurityorganizations)as shareof totalpublic expenditure 9. Insurance(non out-of-pocket)expenditureas share of totalhealthexpenditure, 10. Publicexpenditureas shareof totalinsurance(non out-of-pocket)expenditure, Variablesused to generateeach Figure Figure 3 relatesvariables1 and 5. Figure4 relates variables1 and 6. Figure 5 relatesvariables5 and 6, withcountriesclassifiedby variable2. Figure6a relatesdeviationsfrom a regressionrelatingvariables1 and 5 to 79 deviationsfroma regressionrelatingvariables1 and 3, withcountriesclassifiedby variable2. Figure 6b is identicalto Figure6a, exceptthat countriesare classifiedby variable7. Figure 7a relates deviationsfroma regressionrelatingvariables1 and 5 to deviationsfroma regressionrelatingvariables1 and 4, with countriesclassifiedby variable2. Figure 7b is identicalto Figure 7a, exceptthat countriesare classifiedbyvariable7. Figure 8 relates variables9 and 10, withcountriesclassifiedbyvariable2. Figure 9 relates variables6 and 8, withcountriesclassifiedby variable2. Displayof variablesin Figures Figures3 and 4 showthe variablesdirectly. Figures5, 8 and 9 showthe informationin polar coordinates:one variableis displayed radially(distancefrom the origin)and the other in angularform (shareof the anglebetween the horizontaland verticalaxes). Figures6a, 6b, 7a and 7b showdeviationsfromestimatedvaluesratherthan the valuesof the variables. Sourcesof data Variables1-4weretakenfor allcountriesfromWorldBank [1993b]and the WorldBank's Economicand SocialData Base. Variables5-7 weretakenfrom Govindaraj,Murray and Chellaraj[1994]for Latin Americanand Caribbeancountriesand fromMurray,Govindaraj and Musgrove[1994]for all other countries. Variables8-10 wereestimatedfroma variety of sources, includingthe GovernmentFinancialStatisticsYearbook;Mesa-Lago[1992]for LatinAmericancountriesArgentina,Brazil,Colombia,CostaRica, Ecuador,Mexico,Peru, Uruguayand Venezuela);Griffin[1992]for Asia(Bangladesh,Bhutan,China, Malaysia, Myanmar,Nepal, Papua NewGuinea);Shawand Griffin[1995]for Ethiopia,Nigeria and Tanzania;Vogel[1993]for ten Africancountries(BurkinaFaso, Burundi,Cote d'Ivoire, Ethiopia,Ghana, Guinea, Kenya,Lesotho,Madagascarand Malawi);OECD [1992]and Schieber,Pouillierand Greenwald[1992]for OECDcountries(Canada,France, Germany, Japan, the UnitedKingdomand the UnitedStates);and WorldBankreportson individual countriesas indicatedat the end of the list of references(India,Jamaica,Mozambique, Nicaragua,Nigeria,Pakistan,Tunisia,Venezuelaand Zimbabwe). AnnexTable1 VARIABLENUMBER COUNTRY 1 2 3 4 5 6 7 8 9 10 SUB-SAHARAN AFRICA Mozambiqwe 600 1 47 280 0.06 0.75 4.00 1.00 0.75 Tanzarur 570 1 51 165 0.05 0.68 3.00 1.00 0.82 0.83 Ethiopia 370 1 48 197 0.04 0.61 3.00 1.00 Uganda 1120 1 46 185 0.03 0.47 2.00 0.62 0.76 Burundi 720 1 48 180 0.03 0.52 2.00 Madaga9cr 710 1 51 170 0.03 0.50 2.00 0.76 SierraLeone 800 1 42 360 0.02 0.71 4.00 0.52 Malawi 800 1 45 201 0.05 0.58 3.00 0.62 Rwanda 680 1 46 222 0.04 0.54 2.00 Mali 480 1 48 200 0.05 0.54 2.00 BurkinaFaso 750 1 48 159 0.09 0.82 4.00 Niger 790 1 46 320 0.05 0.68 3.00 0.81 0.83 Nigera 1360 1 52 191 0.03 0.44 1.00 Konya 1.00 0.89 0.50 1350 1 59 83 0.04 0.63 3.00 0.56 Benin 1800 1 51 170 0.04 0.65 3.00 Togo 1310 1 54 143 0.04 0.61 3.00 Cote d'lvore 1510 1 52 90 0.03 0.52 2.00 Senegal 1680 1 48 156 0.04 0.62 3.00 Ghana 2000 2 55 170 0.04 0.49 2.00 Zimbabwe 2160 2 60 58 0.06 0.52 2.00 0.65 Ca4nroon 2400 2 55 125 0.03 0.38 1.00 0.68 Guinea 1 44 268 0.04 0.59 3.00 South Africa 3 63 91 0.06 0.57 2.00 ASIA Inda 1150 1 60 127 0.06 0.22 1.00 0.21 026 Chiha 1680 1 69 43 0.04 0.60 3.00 0.00 1.00 Nepal 1130 1 53 135 0.05 0.49 2.00 0.44 0.49 1.00 Banglesh 1160 1 51 137 0.03 0.44 1.00 1.00 0.44 1.00 Bhutan 620 1 0.02 1.00 MyannMr 1 0.03 1.00 Pakistn 1970 2 59 139 0.03 0.53 2.00 1.00 0.53 Laws 1930 2 s0 171 0.03 0.40 1.00 SriLanka 2650 2 71 22 0.04 0.49 2.00 0.53 Indonesia 2730 2 60 111 0.02 0.35 1.00 0.67 Philippines 2440 2 65 62 0.02 0.50 2.00 0.80 PapumNew Guinea 1830 2 56 169 0.04 0.64 3.00 1.00 0.64 Thalland 5270 3 69 36 0.05 0.22 1.00 0.90 Malaysia 7400 3 71 20 0.03 0.43 1.00 1.00 0.43 Korea. Rep. 8320 4 70 10 0.07 0.41 1.00 0.12 LATINAMERICAANDCARIBBEAN Haii 1220 1 55 156 0.07 0.26 1.00 Honduras 1820 1 es 62 0.06 0.45 1.00 Bolvis 2170 2 59 125 0.06 0.29 1.00 0.17 Dominican RepubIc 3080 2 67 U 0.06 0.34 1.00 0.77 Ecuador 4140 2 66 42 0.04 0.63 3.00 0.62 El Salvador 2110 2 6s 52 0.06 0.30 1.00 0.32 Guatemala 3180 2 64 84 0.05 0.33 1.00 0.61 Jamaica 3670 2 73 26 0.09 0.35 1.00 1.00 0.44 0.80 Nicangua 255 2 6s 106 0.08 0.62 3.00 1.00 Paraguay 3420 2 67 37 0.04 0.25 1.00 0.31 Poru 3110 2 64 73 0.03 0.34 1.00 0.59 0.34 1.00 Argenlina 5120 3 71 26 0.10 0.61 3.00 0.45 Brazil 5240 3 66 69 0.06 0.43 1.00 0.50 0.49 Chile 7060 3 72 20 0.07 0.48 2.00 Colombia 5460 3 69 21 0.05 0.57 2.00 0.43 Costa Rica 5100 3 76 0.09 0.82 4.00 0.19 Mexco 7170 3 70 38 0.05 o.66 2.00 0.15 Panama 4910 3 73 0.09 0.60 3.00 0.41 AnnexTable1 VARIABLENUMBER COUNTRY 1 2 3 4 5 6 7 8 9 10 Unrguay 6670 3 73 23 0.08 0.76 4.00 0.81 0.90 0.85 Trinidad&Tobago 8380 4 71 0.05 0.59 3.00 Venezueda 8120 4 70 26 0.04 0.47 2.00 0.75 Antigua& Barbuda 3 74 0.05 0.54 2.00 Bahamas 4 69 0.05 0.54 2.00 Barbados 4 75 0.06 0.52 2.00 Besze 2 68 0.06 0.48 2.00 VirginIslands 4 74 0.05 0.57 2.00 Dominica 3 72 0.08 0.64 3.00 Grenada 3 70 0.06 0.69 3.00 Guyana 1 65 0.10 0.44 1.00 SLKitts&Nevis 3 70 0.07 0.51 2.00 SLLucia 3 72 0.08 0.72 4.00 St.Vincent 2 71 0.06 0.63 3.00 Surname 3 6s 0.04 026 1.00 EASTERNEUROPE,MIDDLEEASTANDNORTHAFRICA Kazakhstan 4490 2 Poland 4500 2 71 20 0.05 0.80 4.00 Egypt 3600 2 61 S6 0.03 0.38 1.00 Morocco 3340 2 63 71 0.03 0.35 1.00 Bulgaria 4980 3 72 21 0.05 0.81 4.00 Czechoslovakia 6250 3 72 13 0.06 0.85 4.00 0.05 Hungary 6060 3 70 20 0.06 0.63 4.00 0.84 Ro;ania 90.0 3 70 31 0.04 0.62 3.00 Turkey 4840 3 67 94 0.04 0.36 1.00 0.58 Aigeia 5640 3 6o 62 0.07 0.77 4.00 Iran 4670 3 65 64 0.03 0.56 3.00 Jordan 4670 3 69 34 0.04 0.47 2.00 Syria 5220 3 67 44 0.02 0.19 1.00 Tunisia 4690 3 67 45 0.05 0.67 3.00 0.63 0.92 0.73 SaudiArabia 10850 4 6s 81 0.05 0.65 3.00 Yemen 2 52 163 0.03 0.47 2.00 OECDANDOTHERWESTERNEUROPE Greece 7680 3 77 13 0.06 0.76 4.00 Portugal 9450 4 74 13 0.07 0.61 3.00 Ireland 11430 4 75 10 0.07 0.62 4.00 0.94 Israel 13460 4 76 10 0.04 0.50 2.00 0.03 NewZealand 13970 4 76 11 0.07 0.62 4.00 0.85 Spain 12670 4 77 10 0.07 0.79 4.00 0.69 HongKong 18520 4 78 7 0.08 0.19 1.00 Singapore 15760 4 74 8 0.02 0.58 3.00 0.96 UnitedKhgdom 1W40 4 78 9 0.08 0.55 4.00 0.97 0.99 0.86 Australia 160 4 77 9 0.08 0.70 4.00 0.54 Htdy 17040 4 77 11 0.08 0.77 4.00 Nefterands 1620 4 77 9 0.08 0.72 4.00 0.00 BegIum 17510 4 76 11 0.08 0.83 4.00 Austria 17690 4 76 10 0.06 0.66 3.00 0.90 France 18430 4 77 9 0.09 0.74 4.00 0.80 0.96 0.78 Canada 19320 4 77 9 0.09 0.75 4.00 0.79 0.95 0.79 UnitedStates 22130 4 76 11 0.13 0.44 1.00 0.58 0.89 0.80 Germany 19770 4 76 9 0.06 0.73 4.00 0.72 0.97 0.75 Denmark 1760 4 75 10 0.06 0.64 4.00 0.07 Finland 16130 4 76 6 0.07 0.84 4.00 0.54 Norway 17170 4 77 10 0.07 0.95 4.00 0.67 Sweden 17490 4 78 8 0.09 0.90 4.00 0.04 Japan 19390 4 79 6 0.07 0.74 4.00 0.97 0.76 Switzerland 21780 4 78 9 0.08 0.68 3.00 About this series... This series is produced by the Health, Nutrition, and Population Family (HNP) of the World Bank's Human Development Network. The papers in this series aim to provide a vehicle for publishing preliminary and unpolished results on HNP topics to encourage discussion and debate. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliated organizations or to members of its Board of Executive Directors or the countries they represent. Citation and the use of material presented in this series should take into account this provisional character. For free copies of papers in this series please contact the individual authors whose name appears on the paper. Enquiries about the series and submissions should be made directly to the Editor in Chief Alexander S. Preker (apreker@worldbank.org) or HNP Advisory Service (healthpop@worldbank.org, tel 202 473-2256, fax 202 522-3234). For more information, see also www.worldbank.org/hnppublications. THE WORLD BANK 1818 H Street, NW Washington, DC USA 20433 Telephone: 202 477 1234 Facsimile: 202 477 6391 Internet: www.worldbank.org E-mail: feedback@worldbank.org