World Der*e/oprner~r. Vol. 20. No. 2. pp. 303-30X. 1992 Printed in Great Britain. (1.305-750,X/Y $5.00 + (I.(10 0 lYY2 Pergamon Press plc Determinants of Health Care Expenditure in Africa: A Cross-Sectional Study KWAME P. GBESEMETE and ULF-G. GERDTHAM University of Linkiiping. Sweden Summary. -The paper prebents some quantitative evidence on the relationship hetwcen certain socioeconomic and demographic factors and per capita health care cxpcnditure in Alrica. First. GNP per capita. percentage of births attended by health staff and foreign aid received per capita together explained 78.31; of the variance in health care expenditure. Second. per capita GNP is the most significant factor explaining differences in health care cupenditurc. Contrary to what has been tound in comparisons of health care expenditure in the Organization for Economic Cooperation and Development (OECD) countries. we found that the income ekticity ih clot to unity. Furthermore. aid was consistently significant and positive in the three models estimated. Third. variables such as crude birth rate and the population under I5 year5 of age ;IS pcrcentagc of total population were not significant. 1. INTRODUCTION Health care expenditure per capita differs substantially among the African countries. For example. in 19X-l health care expenditure per capita was US%1 in Zaire. Burkina Faso and Uganda. It was US$7 in Kenya and Lesotho. US$Il in Zambia. US$17 in Zimbabwe while Gabon and Libya reported US$S 1 and US$105 respectively during the same period (Sivard. 1987). The purpose of this paper is to provide some quantitative evidence on the relationship between certain socioeconomic and demographic factors and per capita health care expenditure among African countries. In particular. we are interested in whether countries with higher fore- ign aid per capita devote more real resources to the health sector. We use cross-sectional data on 30 African countries to measure the extent to which the above-mentioned explanatory vari- ables are associated with higher health care ex- penditure. Crossnational studies of health care expenditure are commonly used in analyzing the Organization for Economic Cooperation and Development (OECD) countries (see Kleiman. 1074: Newhouse. 1977; Maxwell. 19X1; Leu. 1986; Culver, IYXX: Gerdtham et cd.. 19XX; Gerdtham, Siigaard. Andersson and Jiinsson (forthcoming); Parkin. McGuire and Yule. 19X7: Cullis and West. lY79). An outstanding result of these studies is that the GDP per capita is the most important determinant of health care ex- penditure and that health care expenditure in- creases proportionally more than per capita income (income elasticity above unity). In the works cited above. GDP per capita alone ex- plained about X(J-YO % of the variance in health care expenditure. To the best of our knowledge. there is a dearth of literature on the topic for Africa hence our choice of the African countries for the statistical analysis. The countries chosen have similar mortality levels. similar demographic distribu- tion. small differences in socioeconomic status and, above all. they are within the same geo- graphical region. A crossnational study of this kind has certain advantages. For example. it is possible to investi- gate the significance of health systems across countries or variables reflecting institutional. economic and demographic factors for health care expenditure. International comparisons. however. also have a number of problems. namely: (a) there are problems in defining what constitutes health care expenditure: (b) the use of exchange rate5 in converting the values of total health care expenditure into one monetary unit is misleading because exchange rates do not reflect the relative purchasing power across countries (Parkin. McGuire and Yule. LYXY: Gerdtham and Jiinsson. 1YY la. 1YY lb); (c) inadequate statistical inferences are likely to be drawn due to the few degrees of freedom; and (d) there is the risk of the so-called ecological fallacy. i.e.. the 30-1 WORLD DEVELOPMENT use of aggregate data to explain an individual phenomenon. Parkin, McGuire and Yule (1987) note that earlier studies were subject to the latter limitation when they interpreted a technical income elasticity coefficient greater than one as meaning that health care is a luxury good. This study builds on earlier work on developed countries on the determinants of health care expenditure. The African health care expendi- ture model is presented in Section 2. Section 3 discusses the data and the methods used in the statistical analysis. Section 4 presents the empiri- cal results. and the final section provides a brief conclusion. 2. THE HEALTH CARE EXPENDITURE MODEL FOR AFRICAN COUNTRIES The most general measure of resource availa- bility is per capita income. In our study income is measured in terms of GNP per captta. Micro studies by Grossman (1972). Newhouse and Phelps (1974). Muurinen (1982) and Wagstaff (1986) reveal a very slight correlation between income and the utilization of or expenditure on health care. Among the reasons given for the weak correlation is that at the individual level. a consumer does not have to pav the full resource cost of utilization due to subsidies (Newhouse. 1977: Gerdtham, Sogaard. Jonsson and Anders- son, 1YYlc). This is not true. however. for the nation as a whole. Therefore, we posited a positive relationship between per capita GNP and health care expenditure. The degree of urbanization is measured here as urban population as a percentage of total popula- tion. Studies by Gugler and Flanagan (1978). and Senn (1975) suggest that urbanization in develop- ing countries is followed by the emergence of shanty towns with inadequate sanitation and overcrowding. Urbanization also coincides with industrialization. Health hazards associated with industrial pollution are well documented. For example, in a Nigerian case study, Adegbola (1987) observed that pollution accounts for 12% of the variation in mortality rates. In light of the above, we argue that living in a large town may affect health negatively. Therefore we expect a greater demand for medical services ccreris pm-i- bus. Health care expenditure is therefore hypothesized to be positively related to the degree of urbanization. The choice of this vari- able poses a problem because the degree of urbanization is likely to be highly correlated with per capita GNP. The age structure of the population may be of prime importance in determining the level of health care expenditure. The demand for medical services fluctuates with age. Those under 15 years of age utilize medical services more than average. Hence the demand for health care could be expected to increase the higher the percentage of the population below the age of 15. A disadvantage with these data is that in Africa. most child mortality occurs before the age of five (WHO. 1987). Hence. a greater proportion of those below the age of 15 are unlikely to utilize medical services which could have led to in- creased health care expenditure. This argument holds true because those who survive beyond the age of fivre often develop immunity against the major killers such as malaria (Nkruman. lY73). There is justification. however. for testing this hypothesis. Those who are under 15 and happen to utilize medical care are often extremely sick and the cost of treating them is high because their parents have put off seeking medical care as long as possible. Appia-Kubi (19X1). Fosu (1981) and Staugard (198â)) observed that in Africa. most sick people do not seek medical care until their disease has reached such an advanced stage that it is almost impossible to treat. We use the percentage of births attended by health staff as an indirect measure of the extent to which health services have reached the people. Indirectly. this variable reflects government com- mitment to the provision of health services. Hence. we hypothesized that in countries where rates of hospttal deliveries are high. health care expenditure is likely to follow suit. We acknow- ledge. however. the methodological problems in the use of this variable because more hospital deliveries can also be attributed to differences in attitude in the utilization of health services. improved education. higher income and so forth. Crude birth rate is expected to raise the cost of maintaining a given health level. Hence. we hypothesized a positive correlation between the crude birth rate and health care expenditure. We argue that where foreign exchange is extre- mely scarce - as it is in most of the countries in our sample - an inflow of foreign capital (aid) would accentuate the rate of investment in most sectors by providing the foreign exchange com- ponent of investment. Apart from the salaries of health personnel. the greater part of health care expenditure requires foreign exchange. For ex- ample, the purchase of medical supphes such as imported drugs. equipment and spare parts demand foreign exchange. Hence we hypothe- sized a positive correlation between foretgn aid and health care expenditure careris pribus. There are. however. numerous problems in relating this variable to health care expenditure. For example. foreign aid can be channeled to HEALTH CARE EXPENDITURE 305 other sectors depending on government priori- ties. From the foregoing discussion. our model can be written as follows: HEPC = f (PBA. PCI. POP < 15. URBAN, BIRTHS. AID) or in a multiplicative form: HEPC = I!,,, x PBAâ, x PCIâ: x POP < lSB1 x LJRBANâi x BIRTHSâ- x AfDCIv, with PI 132 I33 fl, 13s (jh > 0 where HEPC = Health care expenditure per capita in US$ PBA = Percentage of births attended by health staff, i.e.. hospital deliveries PCI = Gross national product per capita in USS POP < l5= Population under 15 years of age as percentage of total population URBAN = Urban population as percentage of total population BIRTHS = Crude birth rates, i.e.. annual num- AID ber of births per I.000 population = Foreign aid received per capita in USS. 3. DATA AND METHODOLOGY fn the following section, we specify the sources of the data and the methods used in the present analysis. Apart from the data on the population under 15 years of age - which are from lY8S - and the percentage of births attended by health staff - which cover lY8.%87 - the statistical results presented below are based on an international cross-sectional analysis for 1984. We have chosen this year because of data availability. The data were taken from the Smisricrd Yearbook (United Nations, 1988). The Srute of the Worldâs Childretl (UNICEF. 1989). World Mi/ircrr_v and Social Expendirure 1987 (Sivard. 1987) and the World Developnzenr Reporr (World Bank. 1986). The sample consists of 30 African countries which. however. represent various stages of economic development according to the World Bankâs classifications. In 20 of them, per capita national income amounted in 19XY to less than $400 (low-income economies). In eight it ex- ceeded $400 falling within the $-!2(~1630 range (middle-income economies). The remaining two belong to the category of countries classified as upper middle-income states. with incomes rang- ing from $1.700 to $5.670. The 30 countries in the sample are: Ethiopia, Zaire. Burkina Faso. Malawi. Uganda. Niger, Tanzania. Somalia. Togo. Benin. Liberia, Ghana. Sierra Leone. Kenya. Sudan. Lesotho. Mozambique. Mali. Zambia. Nigeria. Cote dâIvoire. Zimbabwe, Cameroon. Tunisia, Sene- gal, Congo, Morocco. Botswana. Gabon. and Libya. The model was estimated using a multiple regression analysis. Tests were conducted for misspecification of the model. i.e.. heteroskedas- ticity in the residuals (Breusch-Paganâs Chi- squared tests). and against functional misspecifi- cations (RESET-test). We restricted the estima- tions to a linear multiplicative functional form (or double-log) because this form has been shown to be the most adequate empirically in crossnational comparisons of health care expenditure (Gerd- tham et 111.. 1988, 1990). 1. RESULTS The results from the estimation of the model are reported in Table 1. Table 1 shows the results of regressing health care expenditure per capita (HEPC) on six explanatory variables, namely percentage of births attended by health staff (PBA) - used here as a proxy for health care delivery - gross national product per capita (PCI). population under 15 years of age as a percentage of total population (POP < 15). urban population as a percentage of total population (URBAN). the crude birth rate (BIRTHS), and foreign aid received per capita (AID). The equations for the named variables are specified in log-linear (double-log) form, which means that the coeffi- cients can be interpreted directly as partial elasticities. With 16. 19. and 20 degrees of freedom, the reported Râs (adjusted) for the three models are 71.2%. 75.8% and 75.1% respectively. All three models seem to be hetero- skedastic. hence the models have been reesti- mated with an estimator which allows for hetero- skedasticity (white estimator). None of the models was found to be functionally misspecified according to the RESET-test. We have opted for the third equation because the incremental F-test between Hc;< and Ho? (where Hc,z was the null hypothesis) was found to be insignificant, i.e., the null hypothesis Ho3 was accepted. The percentage of births attended by health staff (PBA) is significant in all three regression equations. The coefficients carried the expected sign. indicating a positive relationship between the named variable and health care expenditure. Consistent with the results of earlier studies. per 306 WORLD DEVELOPMENT Model: HE = B,, x PBAlâ x P& x POP < I#â x URBANââ x BIRTHSââ x AIDââ. X F METHOD of estimation = Weighted least squares (white) H,; H -0.V-K) l-i,, 0. 16-1 d.f. I (1 Râ 7Y. I Râ (adj) 71.7 SSE 5.417 B-P (d.f.)l 74.5t;3(6) â RESET (d.f. )I1 I .6X5(3) F-teat (d.l.)n -0.24 Excluded I .hâi 0.363 Excluded I .-I(> EUClUdl2d Il.h(l Excluded EKlUdC!J 2..wt 0. I75 J.(I1â 0. IS7 -1.llâ 19 20 x0.0 7s.3 75.x 75.1 5.hlY (,.l)Y.3 15.JY4(4)â â5.577(i): 2.7V2(3) I .43J( 3) F(2.16) = 0.753 F(I.1â)) = I.hO3 *Rrjrcticm at 1ââ Icvcl. iRejection at 5% Icvel. ftlrjrction at 10â~ level. IB-P = Breusch-Pagan test for hetcroskedazticity in the residuals IjRESET = Residual Itâ\t ~/+2bt = Incremental F-te5t H (,: against Hr.. where H, ;: is the null hypâthcâGb. H,,: against H,,:. whrrr H,., is the null hypclthrsis. tâtc capita GNP is highly significant and positive in the three models presented. Ax is evident in Table I - by comparing the r-ratios - the impact of nonincome variables is not as strong as that of per capita GNP. The income elasticity in our third regression equation is barely one. We interpreted this to suggest that health care expenditure increases at approximately the same rate as income. The population under 15 as a percentage of total population was included only in the first equation. For reasons stated above. we expected a positive correlation between the named variable and health care expenditure. The negative coefficient reported for this variable in equation (I), however. is of interest despite its low level of significance. As stated by Kleiman (1974) - who also reported a negative sign for this variable -the finding negates the commonly held view that the main consumers of health services are those below 15 years of age. The urban population as a percentage of total population was included only in equations (1) and (2). The variable carried the expected sign but was not significant. The crude birth rate was included only in the first regression. It was not significant. neither did it carry the expected sign. Foreign aid received per capita was included in all the three models. The variable carried the right sign and was highly significant. 5. SUMMARY AND CONCLUDING REMARKS This paper considers the differences in health care expenditure among 30 African countries. Although we expected that six variables would be of prime importance in determining health care expenditure per capita in Africa. our best-fitting model indicated significance for three variables: percentage of hirths attended by health staff. gross national product per capita, and foreign aid received per capita. Out of the three variables. per capita GNP is the most correlated to health care expenditure. as indicated by the r-ratio. Specifically the results indicate that when the per capita GNP increases by 10% health care ex- penditure also rises by about 1%. This positive association between per capita GNP and health care expenditure accords with the results in earlier studies. The only difference is that com- pared to the OECD countries - where the HEALTH CARE reported income elasticity for health care ex- penditure is 1.5-2.0 - the reported income elasticity for the African states as seen in equa- tion (3) is barely 1.07. Our proxy variable on health care delivery. i.e., the percentage of births attended by health staff. was significant at the I% level. The coefficient indicates that when the percentage of births attended by health staff increased by 10%. health care expenditure rose by 0.28%. Also of substantial importancs to the level of health care expenditure is foreign aid received per capita. Health care expenditure went up by 0. ISâ!% when foreign aid per capita increased by 10%. The positive correlation between this variable and per capita health care expenditure is of interest. It reinforces the view that in countries with limited resources, increased foreign aid is essential for higher allocation of resources to. among other things. the health sector. Apparent- ly an increase in foreign aid served as a buffer against the deterioration in export earnings which most African economies suffered in the early 1980s. With increased foreign aid most countries probably were able to keep their commitment to the health care sector. which otherwise suffers from budget cuts in times of crisis. The theoretical status of the urbanization variable is unclear. According to Cumper (198-l). high population density such as is found in urban settlements may lead to greater availability of social sevices. This in turn may offset negative EXPENDITURE 3 17 features such as increased disease transmission and greater demand for healthcare expenditure. We have. however. found support - as evident in equation (I) - for the thesis that a high level of urbanization is positively correlated to health care expenditure. Gbesemete and Jonsson ( 1990) reported a positive and statistically significant correlation between population density and in- fant mortality and a negative correlation between infant mortality and health care expenditure. It should be emphasized that the empirical relationships stated above do not necesbaril? reflect causality running from the variables at issue to health care expenditure alone. For example. an increase in health care expenditure may contribute to economic growth and raise income. Furthermore. an increase in the size of the health sector may increase health care expenditure and the percentage of births attended by health staff. It is also worth mentioning that the amount of foreign aid received bv a country may depend on variables not included-in the regression -such ah measures of income inequality - that are related to health care expenditure. Finally. the coefficients presented should be interpreted as estimates of the average strength of the net influence of the independent variables on health care espenditure per capita at the national level taken over all countries in the sample. Hence. the coefficients should be treated with caution and should not be taken as estimates of such influences for specific country. REFERENCES Adegbola. 0.. âThe impact of urbanization and in- dustrialization on health conditiom: The case of Nigeriaâ World Heuldr Srari.vits. Vol. 40. No. Iâ (19X7). pp. 74-X?. Appia-Kubi. K., Marr Câwes. God Herds. Religiotl turd Medical Practice Amorrp rhr Akarls 01â Gham (lâoto- W;I. NJ: Allanheld. Osmun Publishers. IYXI). Cullis. J. G.. and P. A. West. The Ecorzornics o]âHealrh (Oxford: Martin Robertson and Company Ltd.. lY7Y). Culver. A. J.. âHealth care expenditures in Canada. Myth and reality: past and future.â Canadian Tax Paper No. X2 (Toronto: Canadian Tax Foundation. 198X). Cumper. G. E.. Derertnirrarrts of Hetzlth Levels in De~~e/opir~g Commriex. (Letchworth. UK: Research Studies Press Ltd.. 19X-l). Fosu. G. B.. âDisease classification in rural Ghana: Framework and implications for health behaviour.â Socicd Scier~cr and Medicirrr. Vol. 1SB (October 19X1). pp. 471-4X2. Gbcsemrte. K. P.. and D. Jonsson. âA comparison of empirical models on determinants of infant mortal- it)â: A cross-sectional btudv on Africa.â mimco (Linkiiping. Sweden: Unireisity of Linkiiping Dc- partmcnt of Health and Society. December IYYO). Gerdtham, U-G.. and B. J(insson. .âconversion factor instabilitv in international comparisons of health care expendit&.â. Jourrrcrl ojâHe[rlrtr Ecorwrnrcs. Vol. IO (19910). pp. â27-234. Gerdtham. U-G.. and B. JBnsson. âPrice and quantity in international comparisons of health care expendi- tures.â .-lpp/i~ll Ecor1onric5, Vol. â3 (IYYlb). pp. l3l%l~âX. Gerdtham. U-G.. F. Andersson. J. Si)gaard. and B. jonbson. âEconometric analysis of health care râ~- penditureb: A cross-section studv of the OECD countries.â CMT Rapport (Linkiiping. Sweden: Uni- versity of Linkdping. September IYXX). Grrdtham. U-G.. J. Siigaard. B. Jiinsson and F. Andersson. âA pooled cross bectional analysis of health care expenditures of the OECD countries.â in P. Zwcifel and H. Frech (Eds.). Hen/f/r Ecorwmict Workdvride (Dordrecht. Netherlands: Kluaer Acade- mic Puhlihhers. IYYlc. pp. 2X7-310). Gerdtham. U-G.. J. Stigaard. F. Andersson and B. 31)X WORLD DEVELOPMENT Jdnsson. âEconometric analysi:, ot health expendi- malaria in children.â Glicrtin Mrdicnl Jourrial. Vol. tures - A cross-section study ot the OECD coun- 12. No. 1 (March lY73). pp. 17-21. tries.â Jomxnl of Heal//r Ecorwnk (forthcoming). Parkin. D.. A. McCiuire and B. Yule. ..What do Grossman. M.. T/w Dcrmuld jbr He&h: A Thzorrric~~rl international comparisons of health expenditure md Gq~iricul ft~~vsfigatiorr (NBER Occasional really showâ!â Câorrlfnllrg! Medicinr. Vol. 11 ( IYXY). Paper IIY (New York: Colombia University Press. pp. I lblâ3. IY7â). Parkin. D.. A. hlcGuirc. B. Yule. âAggrcgatr health Guplcr. J.. and W. G. Flanagan. Urhmiztrrirur rurd care expenditures and national income: Is health care Social Câhmge in \Ve.st AfrCtr (Cambridge: Cam- bridge University Press. lY7X). Kleiman. E.. âThe determinants oi national outlay on healthâ in M. Perlman. (Ed. ). The Ecwromics of Heullh utld Medical Cure (London and Basingstukc: Macmillan Press Ltd.. 1Y7-1). Leu. R. E.. âPublic/private provision and health expenditures: A cross-country comparisonâ in A. J. Culyer and B. Jdnsson (Eds.). Public mtl Pri\wrta Prm*isior~ of Hm11i1 Câtrre (Oxford: Basil Blackwell. 1YXh). Maxwell. R. J.. Hccddr nrrd Wâcwlrlt (Lexington. MA: Lexington Books. 19X1). Muurinen. J. M.. .âDrmand tâor health: A grneralixcd Grossman model.â Jowxal of Henlrir Ecmomics Vol. I. No. I (May 10X2). pp. j-3. Newhouse. J.. ..Medical care expcnditurc: A crosb- national survey.â Jowx~d of Hwm~tt Rrtowm. Vol. I2 (1077). pp. 115-12s. Newhouse. J. P.. and câ. E. Phelps. .âPricr and income elasticities for medical care servicesâ in M. Perlman (Ed. ). The Ecorwrnics of Mrdiccd Câcw ( London and Basingstoke: The Macmillan Press Ltd. 197-l). pp. I?Y-161. Nkrumah. F. K.. âSevere anaemia associated with a luxur?! good?â Jowmul ofâ Hrulrh Ecorwmics. Vol. 6. No. 1 (June lYX7). pp. lOY-127. Scnn. C. L.. âHealth aspects of community develop- ment and housingâ in W. H&sun (Ed.). T/w Theory utul Pmcrica of fâuhlk Hen/A (Oxford: Oxford University Press. IY75). Sivard. R. L.. \%âorld Milirtrr~ arrd Socrnl Expwdirwrs. 12th edition (Washington DC: World Priorities Inc.. IYX7). Staugjrd. F.. Tmdiriwrul Msdicitw itr (I Trcmsitiorwl Socier! (Gaborone: Ipclcgeng Publishers. IYXY). United Nations. Sfrrrixficul Ywrhook (New York: United Nations. various years). UNICEF. Srnre o/â rlrs Worldâs Câhildre,r (New York: Oxford University Press. 1YXY ). Wagstaff. A.. ..The demand for health: Some new empirical evidence.â Joumol of Health Ecorromics. Vol. 5. No. 3 (September 19X6). pp. 195-233. World Bank. Mâorld De~v/oprnarr! Report (New York: Oxford University Press. various years). World Health Organization. E~mlucrriotl qf the S/rare,q for Hetrlrlr for All hx rhr Ymr 2000: .âk~etit11 Report m the Worltl flrulrlr Sirucrriotr. Vol. 2 (Brazzaville: WHO Regional Office for Africa. lYX7).
Comments
Report "Determinants of health care expenditure in Africa: A cross-sectional study"