Remittances, Public Health Spending and Foreign Aid in the Access to Health Care Services in Developing Countries

Transcription

1 Remittance, Public Health Spending and Foreign Aid in the Acce to Health Care Service in Developing Countrie Alaane Drabo, Chritian Hubert Ebeke To cite thi verion: Alaane Drabo, Chritian Hubert Ebeke. Remittance, Public Health Spending and Foreign Aid in the Acce to Health Care Service in Developing Countrie..4.. <halh-55996> HAL d: halh Submitted on 6 Jan HAL i a multi-diciplinary open acce archive for the depoit and diemination of cientific reearch document, whether they are publihed or not. The document may come from teaching and reearch intitution in France or abroad, or from public or private reearch center. L archive ouverte pluridiciplinaire HAL, et detinée au dépôt et à la diffuion de document cientifique de niveau recherche, publié ou non, émanant de établiement d eneignement et de recherche françai ou étranger, de laboratoire public ou privé.

2 CERD, Etude et Document, E.4 Document de travail de la érie Etude et Document E.4 Remittance, Public Health Spending and Foreign Aid in the Acce to Health Care Service in Developing Countrie Alaane Drabo and Chritian Ebeke Centre d Etude et de Recherche ur le Développement nternational (CERD-CNRS), Univerité d Auvergne. 65 Bd Françoi itterrand 63 Clermont Ferrand, France. Tel: (+33) Fax: (+33) Alaane.Drabo@uclermont.fr (Alaane Drabo), Chritian_Hubert.Ebeke@u-clermont.fr (Chritian Ebeke) We would like to thank Bernadette Kamgnia Dia, Audibert artine, athonnat Jacky, Huang Xiaoxian, Hélène Ehrhart and Emilie Caldeira and all the participant to the th annual Conference of the Global Development Network in Prague for their helpful comment and dicuion. We are alo grateful to participant at the CERD doctoral eminar held in January.

3 CERD, Etude et Document, E.4 Abtract The aim of thi paper i to analyze the repective impact of remittance, health aid and public pending on the acce to health care ervice in developing countrie. The pecific objective are threefold. Firtly, we quantify the differential impact of remittance on the acce to public and private health care ervice. Secondly, we determine whether remittance and foreign health aid are complement or ubtitute in the acce to health care ervice. Latly, we evaluate the heterogeneity of the impact of remittance in the acce to public and private health care ervice by quintile of income. We provide a rigorou econometric analyi by controlling for the endogeneity of remittance, public pending and foreign aid. We find that remittance, health aid and public pending are important determinant of acce to health ervice in recipient countrie. Another intereting reult come from the fact that, remittance lead to a ectorial glide in the ue of health care ervice from the public to the private ector for the intermediate income cla. Thi reult hold alo for the richer quintile that are the major recipient of remittance in developing countrie. oreover, remittance and foreign health aid are complement for the acce to health care ervice in low income countrie. Finally, thee reult ugget that policie aiming at increaing remittance are appropriate for developing countrie but alo that, the optimal therapy for the low income countrie i the combination of remittance and foreign aid. Keyword: Remittance, health aid, public pending, acce to health care ervice, developing countrie, intrumental variable method JEL code: F4 O9

4 CERD, Etude et Document, E.4. NTRODUCTON Several argument can be evoked in explaining the interet on tudie undertaken on health care in developing countrie. Firt of all, the tatu of health of the population i a major preoccupation inofar a it determine the level of productivity of the labor force and contribute to growth a well a to poverty reduction (Bloom and al., 4; Cartenen & Gundlach, 6; Weil, 7). The key role of health a input for development ha been reaffirmed at the international level, a evidenced by the illennium Development Goal (DG). ndeed, three of the eight objective of the DG are centered directly on health. Thee are: the maternal health, the fight againt HV ADS and the reduction of child mortality. Secondly, the paradox to which developing countrie are confronted ret on the ize of their need in the acce to health care ervice in a context of evere financial contraint. Public aid to development that remain one of the main ource of the external financing of health care ervice in developing countrie in general and in Africa in particular i rather tagnating wherea the poibilitie of the internal ource of financing health care in thee countrie are further contrained by the lack of fical reource and by numerou intitutional problem. ndeed, a number of recent tudie on health care ytem in Africa pointed out the exitence of ubtantial gap between budget voted at the central level and dipoable reource that are actually received by the local communitie (Gauthier & Wayne, 9; Reinikka & Svenon, 4, 5). Beide the problem of corruption and bad governance (Rajkumar & Swaroop, 8), the concern for the progreivity of the advantage of public pending on health or more generally, the quetion of equity in the acce to health care ervice remain acute in the context of developing countrie in general, and of African countrie in particular (Catro - Leal and al., 999; Gwatkin, 3; Gwatkin and al., 7; Kamgnia, 8; Kamgnia and al., 8; Berthelemy & Seban, 9; Yazbeck, 9). f public pending and foreign aid on health tand a key ource of financing health and of inequality reduction in the acce to health care ervice in developing countrie, other important ource for financing heath care ervice do exit a well. ndeed, everal recent paper (Edward & Ureta, 3, Chauvet and al., 8, Amuedo Dorante and al., 7) pointed out the important role that remittance play in the financing of houehold health. Remittance contitute one of the mot important feature of the current globalization and Ratha & Zhimey (8) for the World Bank etimate to 4 billion of US dollar, the total volume of fund received by developing countrie through thee channel in 7. The poitive impact of remittance on development led to a number of tudie. The poitive effect of remittance on the accumulation of human capital (Borraz, 5; Hanon & Woodruff, 3; Calero and al., 8; Yang, 8; Banak & Chezum, 9), on the accumulation of phyical capital (Woodruff & Zenteno, 7), on the reduction of income inequalitie (Koechlin & Léon, 7; Chauvet & eplé-somp, 7), on conumption (Adam and al., 8), on economic growth (Giuliano & Ruiz - Arranz, 8; Catrinecu and al., 9) and on poverty reduction (Adam & Page, 5; Gupta and al., 9) have been well demontrated. The interet for tudie on remittance i jutified on at leat three ground. Firtly, remittance differ from the other type of fund received by developing countrie inofar a they directly go in the pocket of the houehold. Secondly, remittance exhibit a relative tability, or to ay a omehow countercyclical evolution, contrary to the other ource of fund received by developing countrie. Thirdly, following the international tatitical databae, the volume of remittance flowing into developing countrie exceed the flow of aid, what make remittance a non negligible ource of fund of which it become imperative to ize all the effect on receiving countrie. The current tudy expand on exiting empirical tudie that have analyzed the impact of remittance on the accumulation of human capital. However it differ from the other tudie on that it tree on the caual effect of remittance on the level of acce to health care ervice a well a on equity in the acce to health ervice. 3

5 CERD, Etude et Document, E.4 The general objective of the current tudy i to analyze the contribution of remittance, public pending and foreign aid in the acce to health care ervice in developing countrie. The pecific objective are: to quantify the differential impact of remittance in the acce of public and private health care ervice; to determine the interaction between remittance and foreign health aid in the overall acce to health care ervice; to evaluate the heterogeneity of the impact of remittance in the acce to public and private health care ervice by quintile of income. Thee hypothee are teted by uing intrumental variable approache, a method which allow u to deal with the potential endogeneity of our variable of interet. Uing a comprehenive and recent databae on intra-country acce to health care ervice at the international level (Health, Nutrition and Population (HNP) databae), our reult how that remittance, health aid and public pending are important factor of acce to health ervice in developing countrie. oreover, we have found that migrant remittance appear to be complementary to health aid in the low income countrie and not necearily in the other. We alo find that remittance lead to a ectorial glide in the ue of health care ervice from the public to the private ector for the intermediate income cla and for the richer quintile that are the major recipient of remittance in the developing world. The ret of the paper i organized a follow. Section i devoted to the literature review of the determinant of acce to health care ervice in developing countrie. Section 3 preent firtly a imple theoretical model of the relationhip between remittance and the demand for medical ervice in a contrained environment, and econdly the econometric equation, etimation method and data ued. We dicu our econometric reult in ection 5 while ection 6 conclude on policy implication.. LTERATURE REVEW.. Traditional determinant of acce to health ervice in developing countrie Bryant et al. () tudied the extent to which elected ocial and economic factor are influential in determining the participation in preventive cancer creening program (namely creening mammography and Pap tet) in Prince George, a large community in northern Britih Columbia. Uing the 994 National Population Health Survey, which contain a ample of 46 women, they how through a erie of logitic regreion analye that immigrant women, ingle women and women with le education in northern Britih Columbia experience low participation in health ervice, reulting in a higher rik for poor health and a poor quality of life. They alo found no aociation between ocial factor and previou mammography ue. Falkingham (4) ue the Tajikitan Living Standard Survey to examine the extent to which the level and the ditribution of out-of-pocket payment for health care payment act a barrier to health-care acce. He how that there are ignificant difference in health-care utilization rate acro ocio-economic group and that thee difference are related to ability to pay. Uing multilevel logitic regreion model and data from the Demographic and Health Survey of ali, Gage (7) explore factor that determine the utilization of maternal health care in rural ali. She how that the dearth of health facilitie, the lack of tranportation mean, ditance, houehold poverty and peronal problem are ome barrier to the acce of maternal health care ervice. n addition, living in area where women viit health ervice and living in cloe proximity to people with econdary or higher education increae the probability of women to ue maternal health ervice. 4

6 CERD, Etude et Document, E.4 Sepehri and al. (8) examine the impact of individual, houehold and commune-level characteritic on a woman deciion to eek prenatal care, on the number of prenatal viit and on the choice between giving birth at a health facility or at home. They ued data from the Vietnam latet National Houehold Survey data for and a random intercept logitic model pecification. They how that women deciion regarding the place of child birth and the utilization of prenatal care are jointly determined by the oberved individual characteritic (inurance tatu, level of education, marital tatu, birth order of the child, unwanted pregnancy), houehold characteritic (income cla, houehold ize, ethnicity, ditance to the nearet hopital) and commune-level characteritic (urban or rural, income per capita) a well a by the unoberved commune-level characteritic. urawki & Church (9) propoe a model that can be ued to maximize the acce to health facilitie by making elected improvement to the road ytem. They have applied their model to the Suhum Ditrict of Ghana and their application how that ignificant increae to all eaon health ervice acce can be obtained by a modet level of road invetment. Uing houehold urvey and minitry data in ome African countrie (South Africa, Cote d voire, Ghana, Guinea, Kenya, adagacar and Tanzania), Catro-Leal and al. () how that public pending in health favour motly the better-off rather than the poor. Seban & Berthelemy (9) i one of the firt paper uing macroeconomic data to addre the iue of the acce to health ervice in developing countrie. Uing Health, Nutrition and Population (HNP) data compiled by Gwatkin (7), they how that acce to health ervice depend on ome ocioeconomic characteritic of houehold uch a mother education level and wealth, but they find no evidence that public health expenditure i a robut determinant of acce to health ervice in developing countrie. However, one of the main drawback of their empirical analyi i they have not taken into account the potential endogeneity of public pending on health, what can be lead to evere bia in the parameter etimated... nternational migration and remittance in the literature on health in developing countrie The literature on the link between international migration and health in developing countrie i recent and can be plit into two broad reearch area. On the one hand, author have analyzed the impact of international migration and remittance directly on the health tatu in recipient countrie. On the other hand, the quetion turn on the impact of migration and remittance on the acce to health care. nternational migration, remittance and health tatu Few paper have addreed the impact of migration and remittance on health. The exception are motly focued on infant mortality. Brockerhoff (99) and Sengonzi and al. () invetigate the effect of female migration on the urvival chance of their children in Senegal and Uganda, repectively. They find that rural to urban migration ignificantly increae child urvival chance. With data on exican municipalitie, Lopez-Cordova (4) conclude that larger proportion of remittance and migrant houehold at the community level are aociated with lower infant mortality rate. Kanaiaupuni and Donato (999) analyzed the effect of village migration and remittance on infant urvival outcome in exico, and concluded that remittance reduce infant mortality. However, the author reach an oppoite concluion for the effect of migration: higher rate of infant mortality in communitie experiencing intene migration. Further evidence of the impact of migration on child health ha been provided by Hildebrandt and ckenzie (5). The author invetigate the impact of international migration on everal children health outcome in exico. Their reult how that migrant houehold have lower rate of infant mortality and higher birth weight. oreover, they find evidence that migration alo raie maternal health knowledge and the likelihood that the child wa delivered by a doctor. On the other hand, preventative health care (uch a 5

7 CERD, Etude et Document, E.4 breatfeeding, viit to doctor, and vaccination) eem to be le likely for children from migrant houehold. Fajnzylber and Lopez (7) analyze the effect of international migrant remittance on anthropometric child health indicator in two countrie, Guatemala and Nicaragua. They alo etimate the impact of remittance on the probability that the delivery of children born in the year preceding the urvey wa aited by a doctor, and on the probability that children aged to 5 received the complete et of required vaccination. The anthropometric meaure on which they focu are the Weight-for- Age (WAZ) and Heightfor-Age (HAZ) z-core for children aged to 5 year old. Their reult ugget that children from houehold that report receiving remittance tend to exhibit higher health outcome than thoe from nonrecipient houehold with imilar demographic and ocio-economic characteritic. The reult alo indicate that the impact of remittance on children health i concentrated on low income houehold located in the firt quintile of the income ditribution. Bhargava and Docquier (8) ue Country-level longitudinal data at three-year interval over 99 4 to analyze the factor affecting emigration of phyician from Sub-Saharan countrie and the effect of thi medical brain drain on life expectancy and number of death due to ADS. They etimate a triangular ytem of equation in a random effect pecification for medical brain drain rate, life expectancy, and number of death due to ADS. They found that in countrie in which the HV prevalence rate exceed 3 percent, a doubling of the medical brain drain rate i aociated with a percent increae in adult death from ADS. Chauvet and al. (8) analyze the repective impact of aid and remittance on human development a meaured by infant and child mortality rate with a panel data on a ample of 9 developing countrie, and cro-country quintile-level data on a ample of 47 developing countrie. Their reult ugget that remittance ignificantly improve child health. By contrat, medical brain drain, a meaured by the expatriation rate of phyician, i found to have a harmful impact on health outcome, o net impact of migration on human development i therefore mitigated. Latly, remittance eem to be much more effective in improving health outcome for children belonging to the richet houehold. igration, remittance and acce to health ervice ncreaed pending on health ervice by migrant houehold may improve the acce to private ector health ervice for everyone in the community a ervice provider move to the community in repone to thi increaed demand. n a tudy of the conequence of migration in Kerala (ndia), Zachariah and al. () find that the receipt of remittance i aociated with greater ue of hopital facilitie to treat illne, and a three-fold increae in the ue of private hopital for childbirth. Lindtrom & unoz-franco (6) ue data from the 995 Guatemalan Survey of Family Health (EGSF) conducted in four rural region of Guatemala to explore how migration experience and ocial tie to migrant influence the likelihood of uing formal maternal health-care ervice. They identify aimilation, diffuion, and remittance a three potential pathway through which migration can affect health-care ervice utilization in rural area. They find that urban migration experience and having relative abroad, are aociated with a greater likelihood of formal prenatal care utilization, after taking account of background characteritic and enabling reource. igration experience at all level i alo trongly aociated with formal delivery aitance; however, thi aociation operate primarily through the poitive aociation between migration and enabling reource. oreover, financial cot and geographic acce are the mot important barrier to formal delivery aitance, wherea awarene and acceptance remain a important barrier to the ue of formal prenatal care in rural Guatemala. Reanne et al. (9) ue data from a 6 urvey of two localitie in the municipal city of Tepoztlán and orelo (exico). They ued logitic regreion to determine whether houehold remittance expenditure on health care were conditional to the type of health inurance coverage. They found that individual who lacked inurance coverage or who were covered by the Seguro Popular program are ignificantly more likely 6

8 CERD, Etude et Document, E.4 to reide in houehold that pend remittance on health care than individual covered by an employer-baed inurance program. 3. THEORETCAL ODEL AND EPRCAL SPECFCATONS The three following hypothee are ucceively teted in thi paper: remittance increae the acce to private health care ervice more than that of public ervice in developing countrie while foreign aid and public pending increae more the utilization of public health care ervice; remittance and foreign aid are complement in poor countrie and ubtitute in rich countrie in the acce to health ervice. remittance lead to a ectorial glide in the ue of health care ervice from the public to the private ector for the intermediate and the richer income clae that are the major recipient of remittance in developing countrie. 3.. Theoretical background: A model of acce to medical ervice in preence of remittance The model preented here i a implified verion of Groman (97) model. We exploit the pecification of Zweifel et al. (997), but we augment the model with the poibility that the individual can receive remittance to finance health expenditure. Conider an individual with a planning horizon of two time period. During each period he or he experience a nonnegative amount of ick time t, which i the lower the larger health tock H. The individual derive a poitive utility from conumption good X while deriving diutility from ick time t ( H ). The utility function defined over thee argument i aumed to be time independent. Future utility i dicounted by a ubjective factor β. Thu, the individual maximize dicounted utility U, U ( t ( H ) X ) + βu t ( H ) ( ) U = U,, X with <, t U ( t ) U >, >, X U X <, t <. () H The crucial component of the Groman model i the equation that define the change in the health tock over time. On the one hand, health capital depreciate at a rate δ, cauing health to woren over time. On the other hand, the individual can increae health capital by inveting. Thi involve the purchae of medical ervice or pending t unit of time on preventive effort. n all, one ha: H = H δ +, with >, ( ) ( t ) t >, <, ( t ) <. () Equation () contitute a contraint that will enter the individual maximization problem. 7

9 CERD, Etude et Document, E.4 8 The econd contraint faced by the individual i the budget contraint. Our innovation here i that, abtracting from health inurance, health care expenditure amounting to p ha to be financed out of labor income, initial wealth A or remittance received from abroad L, with w denoting the wage rate of the initial period and p denoting the price of medical care. Remittance are targeted to buy health care, more preciely, to cover a fraction φ ( ) < φ of health care expenditure p. Conumption (at price c) mut be poitive in both period. Total time available i normalized at in both period. Altogether, the following budget contraint hold after dicounting to preent value with the return of aving R, ( ) ( ) ( ) ( ) cx R cx p H t w R t t w A + + = + + φ (3) To olve thi maximization problem, the Lagrangean method i ued, ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ),,,,,,, H t H X H t U X H t U X X t H = δ µ β l ( ) ( ) ( ) ( ) cx R cx p H t w R t t w A φ λ (4) Firt-order condition for an optimum are given by the derivative with repect to all deciional variable. We obtain: = = µ λ β H t w R H t t U H l (5) = = w t t λ µ l (6) ( ) = = φ λ µ p l (7) = = c X U X λ l (8). = = c R X U X λ β l (9)

10 CERD, Etude et Document, E.4 Firtly, dividing equation (6) by equation (7) yield t = p w ( φ) () Secondly, dividing equation (8) by equation (9) yield U X U X = βr () Finally, equation (5) may be rewritten by olving (9) for R λ, t w U U p β = µ = λ ( φ) H c X t U X p = c ( φ), uing (7) and (8) () Our principal objective i to derive an explicit equation of medical ervice demand. To achieve thi, we, t. t i aumed to be of the Cobb-Dougla type, need to pecify the invetment function ( ) α ( ) α α E = t e E, α <, > < α. (3) E Note that education E erve to magnify the effect of medical ervice and time pent in favour of health t. The tructural demand for medical ervice To derive the tructural demand curve for medical ervice, we firt logarithmize the invetment function (3). Thi lead to ln = α ln + ( α ) ln t + α E (4) E With w = w = w, equation () can thu be rewritten a: ( α ) w = (5) t α p ( φ) α ln t = ln p + ln φ + ln ln w + ln (6) Taking logarithm reult in ( ) ( ) ( ) ( ) α By ubtituting thi expreion into (4) yield the following equation for and thu, ln = ln H by equation (), ln in the cae where δ =, 9

11 CERD, Etude et Document, E.4 α ( ) ( ) ( ) ( ) ( ) ( ) ln = ln H + + α ln p α ln φ α ln w α E E ln (7) α Thi expreion i the equation of tructural demand for medical ervice. We can ee eaily that, the higher the amount of remittance received, the higher the quantity demanded. The reduced demand function for medical ervice The tructural demand function for medical ervice (7) depend unfortunately on the health capital H which i choen optimally by the individual. To olve thi problem, one need firt, to pecify how ick time t depend on health capital H. The following functional form i aumed, θ ( H ) = θ H, t θ >, θ >. (8) The derivative of t with repect to H give, U n the pure invetment model, =. Thu, equation () become, t t θ = θθ H H (9) U t w U X p β = ( φ) H c X c () β uing () and implifying yield, w = ( φ) Setting = R =, From equation (9), () can be rewritten in the logarithmic form a, t p H () ( ) = ln( w ) + ln( α ) + ( ln ln ) + ln( θ ) + ln( θ ) + ( θ ) ln( ) ln( φ) ln p H () Recall that δ =, and thu, ln = ln H, o () can be rewritten a: ln ( H ) ln( ) = ln( p) + ln( ) ln( w ) ln( α ) ln( θ ) ln( θ ) ( θ ) ( H ) φ (3) ln The reduced form of the demand for medical ervice can be obtained by the reolution of the following ytem of two equation:

12 CERD, Etude et Document, E.4 ( H ) ln( ) = ln( p) + ln( φ) ln( w ) ln( α ) ln( θ ) ln( θ ) ( θ ) ln( H ) ln ln = ln H ( α ) ln( p) ( α ) ln( φ) + ( α ) ln( w) α α + EE ln α (4) Recall that the econd equation of thi ytem i the tructural demand for medical ervice a in (7). The reduced form of the demand for medical ervice i given by: ( ) = Θ ( + α ( ε ) ) ln( φ) ( + α ( ε ) ) ln( p) + ( + α ( ε ) ) ln( w) ( ε ) α E ln (5) Where / ( θ + ) < ε and w w = w. n contrat to the tructural demand function for health in (7), thi function doe not depend on health tock. The poitive relationhip between remittance and the demand for medical ervice i reaffirmed. The + α ε +. theoretical impact i given by ( ) E 3.. Empirical framework A number of model are evaluated in an attempt to tet the pecified hypothee. Hence, a bae model would be modified to fit each one of the pecific hypothee. The bae model Specification The primarily concern i to determine the differential effect of remittance on acce to public and private health care ervice. We depart from an econometric model that relate remittance to the ue of thoe ervice while controlling for other potential determinant of the ue of health care ervice. Baed on ome recent tudie on the acce to health care ervice in developing countrie (Gwatkin et al., 7; Berthelemy & Seban, 9), the following model i pecified: ha d, ijt = X d, d, d, ijtβ + φ hf jt + δ rjt + µ i + ε d, ijt The variable are all expreed in logarithm term with, d ijt ha, repreenting the percentage of children of the i th quintile that had acce to a health care ervice of the type (public or private) in the treatment of illne d in the j th country in the t th year. X, hf and r repreent the vector of control variable (education of mother and per capita income), the other ource of financing health care (public pending and foreign health aid) and finally remittance flow in each country, repectively. µ i the quintile fixed effect and ε ijt i a white noie diturbance. n the model, we control for the education of the mother a it i a key determinant of the child health. We control a well for the level of development given that it logically exit a relationhip between a country degree of development and it population acce to health care ervice. Two coefficient are of a pecial interet in the current pecification: the coefficient aociated with the remittance( δ ) and the coefficient aociated with the other ource of financing( φ ). i

13 CERD, Etude et Document, E.4 A eriou problem in thee kind of pecification i the endogeneity of remittance, foreign aid and of public pending, a pointed out by Chauvet and al. (8). Endogeneity of remittance, foreign aid and public pending in health Source of endogeneity A naïve etimation of the parameter of r and of hf, may lead to biaed etimate if the iue of endogeneity i not taken into account. Three ource of endogeneity are generally oberved: a meaurement error on the conidered variable, a revere cauality bia and an omitted variable bia. A far a remittance are concerned, officially recorded value are known to be meaured with error. Etimate of unrecorded remittance range from to percent of official tatitic on remittance (Freund & Spatafora, 8). Balance of payment tatitic produced by developing countrie often neglect remittance received via money tranfer operator and almot alway exclude thoe tranferred via informal mean uch a hawala operator, friend, and family member. oreover, if remittance are targeted by the migrant to directly finance the health of their ibling, thi would reult in a revere and negative cauality between remittance and the acce to health care ervice in the ene that remittance would repond to a low level of houehold acce to health care ervice. Regarding the omitted variable bia it could exit a third variable, which would imultaneouly affect the level of remittance received by houehold and their level of acce to health care ervice. That i epecially the cae for the intenity a well a of the quality of emigration in the country. For intance, international migration of phyician repreent a migration of killed individual which in return ha a clear effect on the amount of remittance received by the conidered country (Adam, 9). However, the everity of the medical brain drain may reduce the acce of population to health tructure for their need for treatment. nternational migration ha another effect in the conidered model: the effect on fical revenue. ndeed, although migration of individual i poitively correlated to the level of remittance, it reduce the number of tax payer which in turn lead to a lower amount of tax revenue and thu to a decreae in the level of public pending on ocial ector. The endogeneity of public pending a well a of foreign aid to the health ector may be due to a meaurement error (which i more pronounced in the cae of diaggregated level of foreign aid) on the one hand, and to a imultaneity bia on the other hand. n effect, both public pending and foreign health aid are targeted to improve population acce to medical ervice and to enure better quality of treatment provided. n an econometric perpective, thi poitive revere cauality between the acce to health care ervice and the ource of funding lead to evere etimation problem in the parameter. Factoring in the endogeneity of remittance, public pending on health and health aid The endogeneity of the conidered variable can be dealt with in variou way. A firt approach i that of a election on obervable. Such an approach conit in adding a large number of control variable. Such variable are uppoed to be trongly correlated with the variable of interet a well a with the dependent variable. Although thi approach i quite eay to implement, the trategy i baed on the aumption that the reearcher ha controlled for all the variable uch that the hypothei of the orthogonality of the endogenou variable of interet with the error term i not rejected. However, it i impoible to think that with thi approach one can control for all the implied bia. oreover, the choice of the control variable i often conditioned by the availability of data on the control variable. A econd approach i baed on the election on unobervable intead. Thi trategy conit in the introduction in the model of fixed effect which in fact meaure all other (time-invariant) determinant of acce to health care ervice and for which a direct meaure i not poible. n the current paper, we purport

14 CERD, Etude et Document, E.4 to introduce quintile fixed effect to control for the heterogeneity between income quintile. However, we cannot introduce country fixed-effect given that about 6% of our databae i made up of countrie with at mot one year of obervation on the dependent variable. A third approach for treating for the endogeneity of the included variable i to reort to intrumental variable. A valid intrument for the variable of interet would have to atify three condition. Firt, it would need to be correlated with the uppoed endogenou variable. Second, the meaurement error in the intrument would need to be uncorrelated with the meaurement error in endogenou variable. Third, the intrument hould not be correlated with the acce to health care ervice, except through the endogenou variable or through the effect on the other variable that have been already controlled for. The choice of pertinent intrument for each one of the endogenou variable of our interet (remittance, public pending and foreign aid) i driven by the recent literature. A in Chauvet and al. (8), remittance would be intrumented by the level of financial development (the ratio of broad money upply () to GDP). ndeed, the greater the level of financial development, the higher the amount of remittance received, given the reduction of the tranaction cot induced by the financial widening (Freund & Spatafora, 8). Following Filmer & Prichett (999) and Bokhari and al. (7), we ue a intrument for public pending on health, the per capita public pending on the health ector, and the per capita pending on defene of a country' geographic neighbor. The higher i pending on defene in neighboring countrie, the higher the hare of military pending in the conidered country budget and thu, the lower the level of public pending on ocial ector. However, it tend to be a poitive correlation between pending on health in neighboring countrie and pending on health in the conidered country. Concerning foreign aid on health ector, we follow the intrumentation procedure initiated by Tavarez (3) and recently reviited by Brun and al. (6) and Chauvet and al. (8). The idea behind thi procedure i that the level of foreign aid received by a given country from each one of the major donor i related to the variou apect of the proximity (geographical and cultural proximitie) between the conidered developing country and the donor. We retain major OECD countrie a principal ource of foreign aid to developing countrie. To addre the iue of geographical proximity, we ue the invere ditance between a country and each of the donor. Total aid end by each donor i then, weighted by the invere geographical ditance. To take into account cultural proximity, we ue a dummy variable which take if the country hare the ame language with the donor. We alo ue a dummy variable for the religiou proximity between the donor and the recipient. The common language or common dominant religion dummie are ued a weight for the amount of aid given by each donor. Altogether, foreign aid received by each country i intrumented by three variable. The quality of the conidered intrument hould be confirmed by the performance of the pecified model. On that repect, we would rely on the Stock and al. () weak-intrument tet and the tandard over identification tet to empirically jutify their ue. Are remittance complement or ubtitute of foreign health aid? An econometric model of the interaction between external ource of funding n order to tet the exitence of a complementary or ubtitutability relationhip between external ource of funding in developing countrie (remittance and foreign health aid), we modify the bae model by adding an interaction term of remittance and health aid per capita. The following model i pecified: ha d ijt = X * d ( r hf ) + µ ε d d d d ijtβ + φ hf jt + δ rjt + δ jt jt i + ijt Two ditinct reult may derive from the etimation of thi model. On the one hand, there could exit a complementary effect between remittance and health aid. n thi cae, the following condition hould hold: 3

15 CERD, Etude et Document, E.4 d, > d, > d, φ >, δ and δ. On the other hand, there could be a ubtitution effect. n that cae, we d, > d, < d, hould oberve that φ >, δ and δ. However, we mut be cautiou when interpreting the reult of the model. ndeed, a low level of health aid per capita might meaure either an increae of the financial contraint for a recipient country or an increae in the financial autonomy of the country. Thu, we have to take into account thi duality in the econometric model. To achieve thi objective, we make the aumption that a diminution in the amount of health aid in a low income country reflect the tightening of financial contraint faced by thi poor country wherea it not necearily the ame interpretation for an emerging country. From an economic ene, we are teting the hypothei that the nature of the interaction between remittance and foreign aid might be conditioned by the level of economic development (income per capita). Our preferred econometric model i then a model which allow the incluion of a double interaction between remittance, health aid and income per capita. The following pecification hold: ha d, ijt = X d, d, ( r * Aid ) + δ ( r * Aid * y ) δ ( r * y ) d, d, d, d, ijtβ + φ Aid jt + δ rjt + δ jt jt 3 jt jt jt + 4 jt jt d, ( Aid y ) + µ d, 5 jt * jt i ε ijt + δ + We are particularly intereted by the cro derivative of the acce to health care ervice with repect to d, ha remittance and foreign health aid ijt d d = δ + δ 3 y jt. rjtaid jt Thi calculu i a way to meaure empirically the ubtitution or the complementary effect between d d remittance and aid given different level of income per capita. When δ and δ 3 have oppoite ign d d ( δ > and δ 3 < ), a threhold effect of income per capita expreed in per capita dollar term d = δ d δ3 y* e arie. Under thee condition, we can therefore conclude that remittance are a complement of health aid in poor countrie and thi complementary effect diappear progreively when the level of per capita income rie. The duality behind the nature of the interaction between remittance and health depending on the level of per capita income i indeed traightforward. The hypothei that remittance and foreign aid are complement in poor countrie and ubtitute in relative rich countrie imply reflect the idea that poor countrie need both remittance and health aid to achieve high rate of utilization of modern health care ervice while relative rich one do not need abolutely the two flow. Heterogeneity in the remittance impact on the acce to health ervice by income quintile f median and richer clae are the primarily and the mot concerned by international migration (given the cot to migrate abroad), then remittance in return are received in a larger amount by thee clae in a conidered country. One would expect that the impact of remittance on the acce to health will be the 4

16 CERD, Etude et Document, E.4 greatet for the third, the fourth and the fifth income quintile. f thi reult i confirmed, it will poit remittance a an imperfect ubtitute for the other type of financing health ector in developing countrie. oreover, if remittance induce a ectorial glide in the utilization of health tructure from the public to the private ector, thi reult hould hold for the median and richer quintile in the ditribution of income. A houehold receive remittance they would increae their earch for better quality in health care delivery. Such income effect induced by remittance on the behavior of receiving houehold need to be taken into account in developing country. Then, if remittance do not crowd-out public fund in health ector, one may expect that public ubidie in the health ector will be received by the pooret of the remittance receiving countrie. n order to tet the hypothei of heterogeneity of the impact of remittance on the acce to health care ervice over the income quintile, the following pecification i adopted: ha d, ijt 5 d, d, d, = X ijt + φ hf jt + δ rjt + i= d, i d, ( r * µ ) + µ ε β ϕ + jt i i ijt where µ i i the quintile dummy variable, i i the income quintile index. We expect that all the ϕ i=3,4, 5 parameter will be negative in the regreion for the public ector and poitive in the regreion for the private ector. Variable on the acce to health care ervice 4. DATA Data on variable on the acce to health care ervice are taken from the tudy by Gwatkin and al. (7) on Health, Nutrition and Population in 56 developing countrie, and all the data are diaggregated by income quintile. n thi databae, more than half of the countrie are African. The report of Gwatkin et al. (7) i baed on data drawn from everal demographic and health urvey (DHS) conducted in thee countrie. Thee urvey target epecially maternal and child health with a tandardized quetionnaire. Data alo include ocioeconomic variable like mother education for each quintile. The report include everal indicator of health tatu and utilization of health ervice. For our work, we are only intereted in the econd type of indicator (acce to health). Two variable on the acce to health care ervice are choen for our purpoe: Treatment of fever: percent of children with fever, with or without cough or rapid breathing, in the two week before the urvey who had ought medical advice for fever, at any medical facility or provider, whether public or private, Treatment of diarrhea : percent of children with diarrhea in the two week before the urvey who had been taken for treatment, at any medical facility or provider, whether public or private, n our analyi, we alo differentiate the treatment of each dieae in a public tructure from that in a private tructure. The geographical coverage of thee data i defined in the appendix Table. Per capita public pending on health ector Data on public pending are drawn from everal ource: World Development ndicator, OECD Health Databae, WHO World Health Statitic, OECD Health Databae and F Government Financial Statitic. Other tudie have ued thi databae in order to analyze the determinant of child-health outcome and acce to health in developing countrie (Fay and al., 5; Ravallion, 7; Berthelemy & Seban, 9). 5

17 CERD, Etude et Document, E.4 Public pending in health for each country are in $US and divided by the total population of each country. Data on population come from World Development ndicator. Development aitance on health (health aid) The exiting reearch on global health reource flow ha yielded ome important etimate and finding, but it doe not provide comprehenive and ytematic etimate of development aitance on health (DAH) over an extended period of time. A majority of tudie have relied on databae maintained by the Development Aitance Committee of the Organization for Economic Cooperation and Development (OECD-DAC). While thee databae are a valuable ource of information, they do not capture all the external aid on health ector. The bigget gap in coverage tem from the fact that the databae only reflect official development aitance (ODA) flowing from government and leave out key private actor in the health domain like the Bill & elinda Gate Foundation (BGF), other private foundation, and non-governmental organization (NGO). The ntitute for Health etric and Evaluation (HE) ha launched a multi-year program for tracking DAH, which ha addreed thee conceptual and meaurement challenge and developed a comprehenive ytem for global health reource tracking. The primary goal of the program i to develop conitent timeerie data on DAH, which will be updated annually. The beginning year in thi databae i 99. Development aitance for health i defined a financial and in-kind contribution made by channel of development aitance to improve health in developing countrie. t include all dieae-pecific contribution a well a general health ector upport, and exclude upport for allied ector. Data are in US dollar per capita term. nternational migrant remittance Remittance data are drawn from the World Bank World Development ndicator Table. Remittance are defined a the um of three component: worker remittance, compenation of employee and migrant tranfer. Worker remittance are claified a current private tranfer from migrant worker who are reident of the hot country to recipient in their country of origin. They include only tranfer made by worker who have been living in the hot country for more than a year, irrepective of their immigration tatu. Compenation of employee i the income of migrant who have lived in the hot country for le than a year. igrant tranfer are defined a the net worth of migrant who are expected to remain in the hot country for more than one year that i tranferred from one country to another at the time of migration. Thi i the mot reliable databae on remittance for cro-country analyi. Data in the tudy are in US dollar per capita term. ilitary expenditure Data are drawn from the Stockholm nternational Peace Reearch ntitute (SPR). Thi intitution publihe data on military pending for 7 countrie ince 988. We ue military pending in US dollar per capita term. ncome per capita GDP per capita in US dollar contant price i ued a a proxy for the level of development. The data come from the World Bank World Development ndicator. Table of decriptive tatitic of variable i preented in Table A in Appendix. 5. ECONOETRC RESULTS Now, we turn to the etimation and the dicuion of econometric reult. 6

18 CERD, Etude et Document, E.4 mpact of remittance, health aid and public pending on acce to health care ervice Thi ubection i devoted to the comparion of the repective impact of remittance, health aid and public pending on the acce to health ervice in developing countrie. Table 3 and 4 ummarize thee reult. Table 3 preent the reult of the impact of remittance and health aid on acce to health ervice for each dieae (fever and diarrhea). For each dieae, we ue a dependent variable, the total acce to health ervice, the acce to public and the acce to private ervice. We are alo intereted in the validity of our intrumental variable trategy. Two tet are performed in order to check the quality of our intrument. A we can ee in Table 3, the F-tet for the firt-tage regreion and the Hanen uridentification tet (OD) confirm the good quality of our intrumentation 3. Regarding the impact of our variable of interet, we find that the elaticity of acce to health ervice with repect to remittance i poitive and tatitically ignificant for the total and private health ervice acce. n contrat, remittance appear to have a non ignificant effect on the acce to public health care ervice (column ) and a negative effect in column (5). Table 3: Effect of Remittance and Health aid on acce to health care ervice Dependent variable : Acce to health ervice by dieae and ector Fever Fever_pub Fever_pri Diarrhea Diarr_pub Diarr_pri () () (3) (4) (5) (6) Remittance per capita (log).33***.9.78***.5** -.53**.66*** (.74) (.4) (3.55) (.6) (-.7) (4.379) Health aid per capita (log).59***.997*** (.99) (3.58) (.3) (.546) (.79) (-.78) GDP per capita (log) (-.837) (-.86) (-.33) (-.5) (-.9) (-.474) (log(gdp per capita))² (.688) (.58) (-.346) (-.459) (.353) (.37) other education.3***.3***.43***.7***.7***.49*** (5.63) (4.67) (4.845) (7.9) (5.7) (6.348) Quintile_ (.38) (.38) (.86) (.973) (-.346) (.379) Quintile_ (.354) (.48) (.63) (.53) (-.55) (.4) Quintile_ (.58) (.863) (.4) (-.44) (-.56) (.66) Quintile_ ** **.43** (.6) (-.6) (.365) (.89) (-.386) (.4) Contant (.9) (.9) (.8) (.34) (.46) (.666) Obervation Remittance intrumentation F-tat Health aid intrumentation F-tat Hanen OD p-value Note : All variable in the model are in log term except income quintile dummie. Fever_pub, Fever_priv, Diarrhea_pub, Diarrhea_priv expre the acce to health ervice for the treatment of fever and diarrhea in public and private facilitie. other education i available by quintile. Remittance are intrumented by the level of financial development (ratio of broad money to GDP) and health aid i intrumented by the total aid budget of the main donor weighted by dummie variable which take the value if donor and recipient countrie hare the ame language or the ame religion and otherwie. Robut t-tatitic are in parenthee. Standard error clutered at country level. * ignificant at %; ** ignificant at 5%; *** ignificant at %. Furthermore, health aid contribute to increae the total level of acce to health a well a the acce to public health ervice. Thee reult are in conformity with the hypothei formulated previouly. ndeed, remittance improve overall acce to health ervice in developing countrie, but alo induce a ectorial 3 The intrumentation equation are preented in Table A- in Appendix. 7

19 CERD, Etude et Document, E.4 glide of frequentation from the public ector to the private ector. We can interpret thi reult with the hypothei of a earch by recipient houehold, for higher quality of health ervice when their income increae. n contrat, the ectorial impact of health aid i in favour of the public ervice acce becaue aid i mainly devoted to thi ector. We can alo notice that the effect of health aid on the overall acce to health care ervice i larger than the impact of remittance. Thi i not a urpriing reult given that remittance are not received by the whole population in a country while foreign aid might theoretically be conumed by all fringe of the population. ntereting reult are alo found regarding control variable. For example, mother education i a major determinant of child acce to health care ervice. Table 4 preent the reult of etimation when health aid i replaced by public pending in the model. We cannot include imultaneouly the two variable becaue of colinearity. ndeed, health aid i financing a non negligible part of public pending in developing countrie. Reult are quai imilar to thoe of the previou model. Remittance induce a ectorial glide in the frequentation from public to private ector while public pending increae overall acce to health ervice a well a acce to public ervice. oreover, the impact of remittance although large and ignificant on the overall acce i lower than that of public pending on health. Following the reult of diagnotic tet performed, our intrumentation i valid only for diarrhea equation. Table 4: Effect of Remittance and Public Health Spending on acce to health care ervice Dependent variable : Acce to health ervice by dieae and ector Fever Fever_pub Fever_pri Diarrhea Diarrhea_pub Diarrhea_pri () () (3) (4) (5) (6) Remittance per capita (log).7**.3.5***.75***.63.64*** (.35) (.3) (3.44) (.844) (.53) (4.99) Public health pending per capita (log).365***.745*** -.98***.38***.88*** -.834*** (3.86) (5.69) (-5.53) (3.8) (6.59) (-4.337) GDP per capita (log).7***.78** (3.338) (.4) (.496) (.93) (.74) (-.377) (log(gdp per capita))² -.54*** -.5*** * -.96*.73 (-4.96) (-3.37) (-.96) (-.89) (-.779) (.687) other education.***.4*.39***.7***.4***.53*** (4.68) (.79) (3.75) (5.58) (3.435) (5.35) Quintile_ (.646) (.64) (.34) (.7) (-.6) (-.9) Quintile_ (.997) (.5) (.6) (.465) (-.43) (.47) Quintile_ (.34) (.767) (.5) (.5) (-.58) (.8) Quintile_5.33** *** * (.535) (.69) (.7) (.998) (-.445) (.79) Contant ** * (-.7) (-.5) (-.658) (.445) (.48) (.8) Obervation Remittance intrumentation F-tat Public pending intrumentation F-tat Hanen OD p-value Note : All variable in the model are in log term except income quintile dummie. Fever_pub, Fever_priv, Diarrhea_pub, Diarrhea_priv expre the acce to health ervice for the treatment of fever and diarrhea in public and private facilitie. other education i available by quintile. Remittance are intrumented by the level of financial development (ratio of broad money to GDP) and public ector health pending i intrumented by the average public ector health pending per capita, and the average defene pending per capita of a country geographic neighbor. Robut t-tatitic are in parenthee. Standard error clutered at country level. * ignificant at %; ** ignificant at 5%; *** ignificant at %. 8