L*r- PS POLICY RESEARCH WORKING PAPER 2952 The Effects of a Fee-Waiver Program on Health Care Utilization among the Poor Evidence from Armenia Nazmul Chaudhury Jeffrey Hammer Edmundo Murrugarra The World Bank Development Research Group Public Services and Europe and Central Asia Region Human Development Sector Unit January 2003 POLICY RESEARCH WORKING PAPER 2952 Abstract This study examines the impact of a fee-waiver program rounds utilization rates have indeed declined despite for basic medical services on health care utilization in comparable levels of income, and this decline has Armenia. Because of the reduction in public financing of occurred among both the poor and the rich, with average health services and decentralization and increased utilization falling by 12 percent between the two surveys. privatization of health care provision, private out-of- But families with four or more children, the largest pocket contributions are increasingly becoming a beneficiary group under the "vulnerable population" significant component of health costs in Armenia. To program, have decreased their use of health care services help poor families cope with this constraint, the in a disproportionate manner-21 percent reduction in Armenian government provided a free-of-charge basic use between the two survey rounds. This precipitous package service to eligible individuals in vulnerable drop in health care use by this vulnerable group, despite groups, such as the disabled and children from single being eligible for free medical services, suggests that the parent households. Drawing on the 1996 and 1998-99 program was inadequate in stemming the decline in the Armenia Integrated Survey of Living Standards (AISLS), use of health services. The authors further present which allows the identification of eligible individuals evidence to suggest that the free-of-charge eligibility under this program, the authors estimate the impact of program acts more like an income transfer mechanism, the fee-waiver program on utilization of health services, particularly to disabled individuals. particularly among the poor. Across the two survey This paper-a joint product of Public Services, Development Research Group, and the Human Development Sector Unit, Europe and Central Asia Region-is part of a larger effort in the Bank to understand the impact of health sector reform on health care utilization and poverty. Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Nazmul Chaudhury, room MC3-576, telephone 202-458-4230, fax 202-523- 0308, email address nchaudhury@worldbank.org. Policy Research Working Papers are also posted on the Web at http:/ /econ.worldbank.org. The other authors may be contacted at jhammer@worldbank.org or emurrugarra@worldbank.org. January 2003. (34 pages) The Policy Research Working Paper Series disseminates the fmidtngs of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than filly polished The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors They do not necessarily represent the view of the World Bank, its Executive Directors, or the countries they represent Produced by the Research Advisory Staff The Effects of a Fee-Waiver Program on Health Care Utilization among the Poor: Evidence from Armenia Nazmul Chaudhury Development Research Group, World Bank nchaudhury(iworldbank.org Jeffrey Hammer Development Research Group, World Bank jhammer( bworldbank.org Edmundo Murrugarra Human Development Department / Europe and Central Asia, World Bank emurrugarra0iworldbank.org This paper was written as part of the Armenia Poverty Update and Public Expenditure Review. The authors thank the comments from Kan Hurt, Susanna Hayrapetyan, Nick Haazen and Toomas Palu. Contents 1. Introduction ..............................................1 2. Health sector in Armenia and the fee-waiver program ..........................................1.... The eligibility to the vulnerable population program ..............................................2 3. Data sources and eligibility criteria ..............................................3 Comparability of the two cross sections ..............................................3 4. Major changes in eligibility, morbidity and utilization .............................................. 4 Changes in eligibility ..............................................4 Changes in morbidity profiles and health care utilization .........................................5 Service providers ..............................................9 Out-of-pocket payments and waivers ............................................. 11 Self-reported health status and eligibility categories ............................................. 14 5. The Model of Demand for Health Care and Results ............................................. 15 Regression results ............................................. 17 6. Conclusion ............................................. 19 References ............................................. . 20 Appendix Tables ............................................. . 21 1. Introduction The objective of this paper is to assess the effect of waiver instruments on health care utilization. Given the regressive effects of fees in health care utilization, governments implement waiver and exemptions to protect specific population groups or assure the delivery of specific services (Bitran and Giedion 2002). Waivers in health care are generally intended to ensure the subsidized service for specific population groups, which could be determined using a number of criteria - depending on the country -- such as geographic location, ethnicity, or even poverty indicators. Exemptions, on the other hand, intend to guarantee the free delivery of specific services that, for instance, entail significant externalities. In practice, individuals that are waiver beneficiaries could also receive exempted services, such as a disabled receiving TB care. Armenia, as many other former Soviet Union countries, has both types of interventions. This paper uses the evidence from a targeted waiver program in Armenia in order to assess its impact on health care utilization among the uncovered. The rest of the paper is organized as follows. Section 2 describes the evolution of the health care sector during the late nineties and the characteristics of the fee-waiver program in health care. Section 3 describes the data used and the caveats in comparison across different cross sections. Section 4 discusses the major findings regarding eligibility, morbidity, and health care utilization between 1996 and 1998/99. Section 5 presents the utilization model and discusses the results. Section 6 concludes the paper. 2. Health sector in Armenia and the fee-waiver program This section describes the evolution of the health sector during the nineties and the implications in terms of access to health care, especially among the poor and other vulnerable populations. The transition process in Armenia involved the health sector in at least two dimensions. First, the overall decrease in public expenditures in health care during the early nineties affected the number of personnel, quality of services and the maintenance of the existing infrastructure. Even though the fall of real spending in social areas after independence was reversed in the late nineties, important effects on the supply quality of health services and on the demand for health care were observed. The declining quality of services associated with lower wages, lack of drugs and deteriorated infrastructure was accompanied by a significant decline in the number of patients and increased informal payments. A second dimension is the market-oriented reform in the health sector, which involved a decentralization process and privatizations of some components of the system. Hospitals and polyclinics were converted into semi-private enterprises and the management of health care providers was decentralized allowing them to fix their health service prices, choose their mix between medical and administrative personnel, and allocate resources accordingly. In 1993 state health care institutions became state health enterprises, or semi-independent units that could generate their own revenues parallel to state budget financing. In 1995 hospitals and polyclinics were permitted to provide private services in addition to state funded ones, 1 providing them additional autonomy with self-decision on staffing (World Bank 2002a). The separation between health care delivery and financing was established through the creation of the State Health Agency (SHA) in 1998, responsible for purchasing services to providers (hospitals).' In order to contract out services a Basic Benefit Package was established. The changes during the nineties represented the actual elimination of former free universal health care coverage since those allowed providers to generate their own revenues through OOP. As a result, the increased incidence of out-of-pocket expenditures -- and even worse, that of informal payments to medical and administrative stafY2 -- resulted in decreased health care utilization, especially among the poor. To respond, the government established a program that provided free of charge medical services based on two eligibility conditions: (i) the patient belongs to some vulnerable socio-economic categories; or (ii) the medical care is qualified as "urgent" by the medical staff.3 The definition of the vulnerable groups actually corresponded to the system of categorical social assistance benefits inherited from the former Soviet Union. All costs of services (not including medications) under the program for the "Vulnerable Population" are covered by the government and expected not to exceed 30 percent of the provider's total annual budget. All other interventions are expected to be cover by the resources generated by the providers. The eligibility to the vulnerable population program The primary focus of this study is to examine the effect of a fee-waiver program on health care utilization in Armenia. Vulnerable population groups were officially defined as those disabled persons (according to three degrees of disability), war veterans, children under the age of 18 with one parent, orphans under the age of 18, disabled children under the age of 16, families with four or more children under the age of 18, families of war victims, prisoners, children of disabled parents, victims of the Chemobyl disaster, and catastrophe workers. This paper uses the Armenia Integrated Survey of Living Standards (ISLS) for 1996 and 1998/99 in order to explore for factors that shape the demand for health care during this period of economic transition. In both surveys the eligibility criteria was expressed as belonging to one of the following five categories: (Cl) disabled; (C2) orphan; (C3) families with four or more children under the age of eighteen; (C4) children under the age of eighteen with one parent; and (C5) children of disabled parents. These criteria result from specific questions since the survey was designed to support the social assistance system and needed to precisely identify those individuals. ' Besides the increased responsibility of the provider's managerial team, the decentralization also increased the political dependence to the local authorities, such as the opinion of the local governor (marzpet). 2About 91 percent of patients reported mnaking informal payments in Armenia, the highest incidence among the countries surveyed in the Europe and Central Asia region (Lewis 2000). 3 Anecdotal evidence suggested that the subjective qualification of "urgency" affected the incidence of health care interventions and subsidies, providing free services to those not in urgency nor in vulnerable groups (Kurchiyan 1999; World Bank 2000). 2 3. Data sources and eligibility criteria This study uses two household surveys from Armenia - the 1996 Armenia Living Standard Survey and the 1998/99 Armenia Integrated Survey of Living Standards (AISLS). The 1996 survey was conducted during November and December of 1996 with a sample size of 20,076 individuals and was nationally representative. The 1998/99 AISLS was carried out throughout a full year (July 1998 to June 1999) with a sample size of 15,632 individuals and covering all regions. The two surveys are separate cross-sections, i.e., households/individuals cannot be tracked over time. The eligibility program that we focus upon in this study was already being phased in by late 1996/early 1997, and the eligibility program did not change substantially between November/December 1996 and July 1998/June 1999. Thus, we lack the base-line data to conduct a "before and after" program evaluation. However, given that program eligibility is exogenous, we can still examine the impact of the program on the two cross- sections (the eligibility program will the discussed more in details later on in this section). Also, while there were changes in the survey instrument between the two rounds, the morbidity module and the module from which we infer program eligibility, are consistent across the two rounds. Before examining health related issues, a discussion about the comparability of the two surveys is provided next. Comparability of the two cross sections A recent analysis of Real Private Consumption Armenia 1995-1999 poverty discussed the difficulties in comparing 240 poverty between two surveys 220 in Armenia since the surveys > 200 - differ in their sample design, 180 s survey period and 160 questionnaire format (World _14 Bank 2002a). For exarnple, a the measurement of poverty 120 indicators is affected by the 100 differences in the timing of 80 3 1 3 I 3 . 3 the two surveys due to 1995 1996 1997 1998 1999 seasonality in consumption and the problems to construct comparable consumption aggregates.4 The 1996 survey was conducted in the last quarter of 1996, coinciding with the peak of the consumption profile (see graph), compared to the 1998/99 AISLS. 4 The 1996 Survey asked only about expenditures, not consumption during the month previous to the survey date. Moreover, information on expenditures for 1996 was not collected using the same questionnaire for all households. Some households responded to a more aggregate expenditure questionnaire, than others. International evidence indicates that this type of differences in questionnaire design leads to sigmficant 3 To address the comparability problem, the mentioned report provided a limited comparison between the two Surveys under a number of restrictions. First, poverty indicators for 1998 were estimated only on information collected during the fourth quarter (October- November). Second, the 1996 poverty line with proper inflation adjustment was used, hence avoiding changes in the poverty line due to changes in its structure.5 These comparability problems also affect the analysis of this paper despite the identical health module because of the strong seasonality in morbidity observed in Armenia. The Armenian epidemiological profile indicates that respiratory diseases represent about half of the first diagnosis in Armenia (Ministry of Health 2000) despite the fact that mortality due to respiratory diseases has been declining and is only 5 percent of the total mortality.6 To avoid seasonality in health status and health care measures, the paper mainly examines the 1996 survey and the corresponding fourth quarter for 1998 (evidence for the full 1998/99 survey is also provided). 4. Major changes An eflig&blity, morbidity and ufflization This section describes the evolution of program eligibility (the crucial variable to identify the impact of the program), morbidity, utilization and expenditures. Changes in eligibility. In 1996, 14 percent of individuals were classified as eligible; in November-December 1998 (henceforth referred to as ND98) 15 percent of individuals were classified as eligible; and in the full 1998/99, almost 17 percent of individuals were classified as eligible. The almost negligible change in the fraction eligible between 1996 and the comparable period in 1998 confirms the quality of the survey since there were no significant changes in the eligibility to the fee-waiver program during the period. Even though the eligibility criteria are not explicitly poverty targeted, those socioeconomic criteria have strong correlations with poverty. The evidence on eligibility by consumption quintiles confirms this since the poor were more likely to be classified as eligible in both surveys (see IFRgure 1). Given a similar eligibility income gradient, what are the patterns in health care utilization? differences in consumption estimates between the two sub-samples and raises serious questions about their comparability (Olson and Lanjouw 2001). The 1998/99 AISLS on the other hand, collected information on both consumption and expenditures using the same format for all households. 5 An additional adjustment was that instead of using per adult equivalent consumption (used for 1998/99), the corparison uses per capita consumption similar to that used m 1996. 6 Other chronic diseases-such as cardiovascular, neurological, neoplasm, and kidney illnesses-represent a relatively small fraction of the first health care contacts. Cardiovascular diseases, however, represent 35 percent of the mortality for the population aged 0 to 64 m 1999. The incidence of infectious and parasitic diseases is less than 8 percent of cases, but evidenced a significant increase during the nineties (Ministry of Health 2000). 4 Figure1: Bigibility by Income 25 20 15 J 15 Q1 Q2 03 Q4 C5 Income Quintile | 1996 *1998N,ov,Dec 01998/99 Changes in morbidity profiles and health care utilization Both surveys contain informnation about self-reported morbidity. Individuals report whether they experience an illness within the last 30 days (preceding the survey). Self- reported morbidity measures tend to be associated with education, income and access to health care providers (Strauss and Thomas 1996), and in consequence better-off individuals tend to be more likely to report themselves as sick. However, while this relationship is linear in the first round, we observe a U-shaped relationship in the second round (i.e., the poorest and the richest individuals are more likely to report an illness - see Figure 2). We also find a small reduction in morbidity rates between the two samples: average morbidity in 1996 was 18.7 percent, while average morbidity in 1998/99 was 17.2 percent (average morbidity in ND98 was 14.5%).7 Self-Reported morbidity rates tend to be higher for Eligible individuals compared to their non-eligible counterparts (see Figures 3 and 4). While health care utilization (conditional upon being sick) tended to increase with wealth in the 1996 survey, the (simple graphical) relationship between utilization and wealth is not so clear-cut in the 1998/99 AISLS (see Figures 5 and 6). There was however, a significant reduction in health care utilization between the two survey rounds. Average utilization rate in 1996 was 50 percent, while the average utilization rate in 1998/99 was 36.6 percent (average utilization rate in ND98 was 36.8%). 7 This paper does not attempt to explain the slight reduction in self-reported morbidity, although some evidence from the U.S. and Indonesia points that self-reported morbidity is affected by the price of health care observed in the locality (see Dow and others 1997). 5 FIguro2: Changes In SelI-Rsported Morbidity Rates 2D a5 to .0 at 02 co CA co Figurze 3: Morbidity Rates 1993 30 25 20 Si 15 E }-( - 10 Q1 02 03 04 aS Incomo QntDubo --oEl- ht- 10 -t5onf-igte| Figure 4: WorbIdity Rates 1998/1939 30 25 ___ _2 20 hSi 10 5 0 Qt Q2 Q3 04 05 Incomo Qulntto -O---Eigiblo - > - Non-EtiI| 6 Figure 5: Health Care Utilization Rates 1996 70 60 d 50 w ' =~~~~~~l C 40 I0 30 J 20 VI N 10 0 01 02 03 Q4 0Q Income Quint ile -4--- *--igible - -U- -Non-Blgible Figure 6: Health Care Utilization Rates 1998/1999 60 450 30 -* 'A i a 20 .4 40 'tAA 01 02 03 Q4 02 Income Quintile -4---*-Si1gible -- -Non-SBigible Which socioeconomic groups experienced largest declines in health care utilization (conditional on being sick)? Reduction in utilization was however, most pronounced among the rich (see Figure 7), and that effect was amplified among rich eligible individuals (see Figure 8).8 If anything, in general, health care utilization by the poor seems to have increased. However, health care utilization rates fell for all eligible categories9 across the two rounds (see Table 1). While the reduction in health care utilization by the disabled (-13%) was similar in magnitude to average reduction in the sample (-12%), utilization by individuals 8 If the full 1998/99 AISLS is used to estimate this decline, a smaller drop is found because of the higher utilization durng the winter period that was not captured by the 1996 survey. 9 Orphans are left out due to extremely small sample size. 7 coming from families with four or more children (under 18) fell by 21 percent, despite the fact that this group was covered under the BBP prograin. We shall further explore this issue within a multivariate regression framework later on in this paper. Figure 7: Changes In Health Care U(tilization Rates 60 * 50 2 0 - 40~~~~~~~~~~~~~ '30 --~~~~~~~~~ IA 1 20 v Vt '' 10 0 Q1 02 03 04 05 Incomo Quintilo F--- 1996 --- -199Nov.Dec 1998/1999 Figure 8: Health Care Utilization Rates 1996 vs 1998 (Nov, Dec) 70 eo .- eo- _ a30 r . S20 O--~~~~~~~~~~~~~~~~~~~~~~~~- 10 - 0 01 02 03 04 05 tncomo Quletitl 4--- E99i5 e 1999 - - Eligibie 1998 (Nov, Dec) Non-Eliglble 1986 - X- Non-EUgtile 1998 (Nov, Dec) 8 Table 1: Health care utilization by eligibility category in 1996 and 1998/99 Utilization (%) Utilization (%) 1996 1998/99 Disabled 59.4 47.3 (N=276) (N=279) Belong to a Family with -4 children 48.5 27.6 (N=1 14) (N=199) Childfrom a single-parent HH 41.2 33.3 (N--1 7) (N=24) Child with a disabled parent 47 39.4 (N=100) (N=66) Service providers Despite the significant decline in health care utilization, the decomposition of utilization by providers shows few changes between 1996 and 1998/99, even across quintiles. During both survey rounds, among the sick that sought health care, the majority of the visits were to polyclinics and hospitals (see Figures 9 and 10). For comparable facilities across the survey rounds (i.e., polyclinics, hospitals and diagnostic centers), only hospital visitations exhibit any systematic relationship with income (see Figures 11 and 12). Overall, the proportion that sought health care in polyclinics experienced a small decline from 52 to 49 percent between 1996 and 1998/99.10 This is consistent with the larger decline in Primary Health care utilization rates compared to Hospital utilization (World Bank 2002b). The poor are more likely to seek care in polyclinics (more than 60 percent). The fraction of users going into polyclinics decreases among the better-off households probably reflecting poorer quality of health care and access to other facilities (hospitals). Hospital care, on the other hand, represented about 28 percent of those seeking health care in both years, and the rich were more likely to go there than the poor in both years (38 and 36 percent for 1996 and 1998/99). One of the few changes in utilization was the increase in Other sources of care. In 1996 Other (non specified) sources represented only 7 percent of those seeking care. In 1998/99 about 10 percent of patients went to other sources, including private doctors (particularly among the better off households). '° The share of Primary health care utilization in 1996 could be even larger is those covered by health posts (fyelcher) are mcluded. In total, polyclinics and yelchers represented 57 percent of health care according to the 1996 AILS. 9 Figure 9: Response to Morbidity - 1996 Pdydi,io 25% SW. 14% _ 1 ~~~4% ah- 4S L"- Figure 10: Response to Morbidity - 1998199 20% 04% 2S Figure 11: % Visits to a Facility by Income Quintile - 1996 40 35 30 25 _ _ _ I5 - 10 1s a1 02 03 04 05 10 Figure 12: % Visits to a Facility by Income Quintile - 1998/99 30. 25L1 l,- fF al 02 03 04 5 bcn Q..bn5 Out-of-pocket payments and waivers The fees charged to patients and the ability to waive those costs may affect the decision to choose some providers. About 64 percent of the patients paid for health care in Armenia (Table 2). The poor are less likely to pay for health care (about 40 percent) and chances of paying are even lower in polyclinics (34 percent). This may reflect the effects of the free-of-charge Basic Benefit Package. The BBP, despite of not being poverty targeted, may have covered the individuals with lower consumption if the vulnerable categories are associated with poverty. On the other hand, a better off individual could also have used the BBP to obtain free-of-charge health services if she were eligible (either vulnerable category or urgency). As the BBP covers only basic services, individuals were likely to be subject to other payments, mainly through informal mechanisms. Table 2: Percent of patients that paid for services (Percent) Quintiles Polyclinic Diagnostic Hospital Private Other Total Center Doctor 1 34.0 75.0 44.4 80.0 33.3 39.4 2 50.4 18.2 68.3 64.3 40.0 52.5 3 58.5 71.4 66.7 100.0 14.3 59.6 4 75.8 100.0 77.4 76.2 40.0 76.1 5 69.7 100.0 87.5 69.0 6.7 75.3 Total 59.9 75.5 74.8 74.4 23.1 64.1 Polyclinics are the cheapest alternative for most patients. A sick individual from the poorest quintiles pays about 1,300 drains compared to 4,000 in a diagnostic center or 3,400 in a hospital. The expected cost (the cost weighed by the probability of being charged) is lower 11 in hospitals than in diagnostic centers or private doctors, explaining the choice of polyclinics and hospitals over other altematives (TabBe 3). 11ablie 3: Average cost for Patients that paidl for services (dramns) Quintiles IPoRyclinic Clengn@ster HlospitaR IFrivate IlDoctor Other Total 1 1,320 4,000 3,421 2,625 3,000 2,224 2 1,809 5,000 3,918 4,241 2,250 2,666 3 1,883 6,200 4,125 2,711 625 2,698 4 3,462 8,333 10,386 2,913 2,000 5,496 5 5,388 39,031 45,702 10,538 10,000 25,099 Total 3,384 24,937 26,027 5,697 2,646 12,175 The lower utilization and the choice of cheapest providers results in a regressive incidence of private expenditures. The poorest population quintile spends less than 2 percent of the private health care expenditures in Armenia. The richest quintile is responsible for almost 80 percent of the private expenditures. Inequality in private health expenditures, however, is not necessarily negative, since government intervention could be covering those poorest individuals. However, the BBP that is freely provided to some vulnerable groups does not cover drugs and pharmaceuticals. Share of expenditures in drugs represent about 20 percent for the overall population, but is almost 40 among the poorest quintiles (see Table 4). Table 4: D]istribntionm of private health expendtoures (Annenia 1998/99) Share of totaW private expenditures by quintule (%) Conmcenmtration E8ealth itemn I 2 3 4 0 index Dental 0.0 5.3 5.1 13.2 76.4 0.718 Diagnostic 2.1 1.2 6.7 14.8 75.2 0.675 Treatment 1.4 2.8 2.8 9.0 83.9 0.755 Other 0.5 1.3 6.6 6.8 84.7 0.777 Drugs 3.0 7.3 11.8 20.5 57.4 0.512 Total 1.7 3.7 4.9 11.6 78.0 0.703 Memo item: Total - Percent spent on drugs 36.5 39.6 48.1 35.6 14.9 20.2 - Private expenditures 222.3 492.5 652.2 1,532.8 10,298.7 13,198.4 (in thousands drarns) Source: ILCS 1998/99. Note: The table shows the sarple. What share of the government expenditure in health care is being captured by the poor? Individuals in the poorest quintile benefit only from 13 percent of total expenditures, compared to those in the richest quintile that capture almost 40 percent of the public resources 12 (see Table 5). Even though the individuals in the poorest quintiles are more likely to choose polyclinics as their major health care provider, most of the patients in the polyclinics are from the better-off households. This pattern is due to the differences in health care utilization across socioeconomic households, since individuals from better off households are more likely to seek health care once they are sick. In 1999, the government spent about 5 billion drains in polyclinics. Patients from the poorest quintile captured only 772 million compared to those in the richest quintiles that captured about twice that amount (1.4 billion). However, across different govermment health facilities, polyclinics represent the least regressive alternative. The concentration index for polyclinics (0.114) is less than one half than that of hospitals and other centers (0.276). Coincidentally, health care utilization in the poorest quintile (25.9 percent) is almost half of that of the richest (51.4 percent). Table 5: Distribution of public expenditures in health (excluding private expenditures) (mnllion drams) Received by quintfle Total Concentration 1 2 3 4 5 budget index Polyclinic 772.3 892.3 884.8 989.8 1,409.7 4,949.0 0.114 Diagnostic Center 190.7 524.5 333.8 286.1 1,001.4 2,336.5 0.276 Hospital 1,687.7 1,537.7 1,687.7 2,325.3 5,400.7 12,639.1 0.276 Total 2,650.8 2,954.5 2,906.3 3,601.2 7,811.8 19,924.6 0.236 Despite the regressive pattern of overall spending in the health sector, there is some evidence of a progressive pattern of public spending regarding the provision of the targeted program for the vulnerable. Evidence from the budget allocations during 1999 indicates that the allocation of free-of-charge services targeted to the vulnerable population has been positively associated with poverty incidence across regions (marz) in Armenia. It should be noted that despite being eligible to free-of-charge care, individuals are required to contribute informal payments, posing an additional burden both to the poor and non-poor. This stems from the fact that providers are Poverty and Free Health Care Expenditures not being paid the entire 80 -- cost of their expenses for 7 their coverage of vulnerable 7 groups, forcing them to 60 cross-subsidize by charging - higher fees to patients with DU50 higher incomes or directly G40) charging (via informal methods) the eligible 30 population (European 20 Observatory 2001). 20 ln(Health Expenditures) 13 Self-reported health status and eligibility categories While we find that health care utilization rates have declined between the two surveys, we do not have information on how this decline in utilization has actually effected adult or child health outcomes. We do, however, have information on self-reported health status. Respondents in both surveys were asked to rate their overall health status as "very well", "good", "normal", "not so good", or "bad". The use of this type of self-reported health status (often referred to as the Likert scale) has been shown to be a powerful predictor of subsequent of morbidity and mortality (Idler and Benjamini 1997). Our interest in this variable arises not as a predictor of future mortality, however, as a possible proxy for an individual's health status, and to examine how this self-reported measure has changed between the two surveys. Self-reported health status (SRHS) using the same five-point scale was collected in both rounds. We only draw upon the SRHS of adults for our analysis. In the 1996 survey, for 75 percent of the respondents, information on SRHS is missing, thus, 1996 results incorporating this additional information is not necessary comparable to results using the full adult sample. Given that caveat, comparisons of average SRHS across the two rounds do not indicate a decline in self-reported health status. Comparing sample averages, in both rounds, self-reported health of eligible individuals were worse than that of non-eligible, self-reported health of women were worse than that of men, self-reported health of urban residents were worse than that of rural residents, and the self-reported health of the poor was worse compared to the rich. Table 6, for example reports the distribution of SRHS responses by eligibility. In both rounds, a disproportionate number of eligible individuals classify themselves as being in "bad" health. Table 6: IDLstiributon of seRf-reported heaRth statu2s (Percent) Bad Not so good Normal Good Very well Eligible 96 16.02 15.59 37.63 26.75 4.01 Non-Eligible 96 8.66 21.25 40.62 25.5 3.97 Eligible 98/99 11.39 15.66 48.36 22.47 2.12 Non-Eligible 98/99 5.3 17.04 54.13 20.82 2.71 For all subsequent presentation, we will decompose eligibility into its separate components given that we had previously noted that utilization rates and changes in usage differed according to the eligibility category. As we see in Appeimsdl Table Al, families with four or more children, is the largest eligible group in the sample. In Tlalble IRI and TablRe R2, we present an ordered probit specification of the correlates of self-reported health status for the adult sample. Most studies in this literature find a negative association between SRHS and age and being female, and a positive association between SRHS and wealth (Case and Deaton 2002). Results from the 1996 round, indicate that SRHS in negatively associated with age, 14 gender and urban residence (there are also strong district effects), however, insignificant wealth effects. Results from the 1998/99 round are more in line with other finding in this literature - besides been negatively associated with age, being female and urban residence, SRHS is also positively associated with wealth. While only disability is significant (negative) is the 1996 specification, both disability (negative) and belonging to a family with four or more children (positive) is significant in the 1998/99 specification. It should not come as a surprise that people classified as disabled would report themselves as being in poor health. The interpretation on the non-disability based category is not so intuitive. Families with four or more children tend to be poorer (even in non per-capita terms), poorer individuals tend to underreport morbidity spells and seek less health care. Health care usage in turn, can affect self-reported health status. Given that people who use the health care system are more likely to be better informed than non-users, in situations where lower-income individuals are less likely to use health care, the measurement error in SRHS will be amplified and systematically related to income (Strauss and Thomas 1998). Thus, we present results with and without including this variable given that we realize that by trying to compensate for our lack of information about innate healthiness, we might in turn introduce a systematic source of measurement error. 5. The Model of Demand for Health Care and Results We first model the reduced form demand for health care (Grossmann 1972) as a function of program eligibility and individual, household and regional characteristics. Ideally, besides controlling for latent health status, we should also control for more detailed community infrastructure factors, and prices of medical services for particular services. We however, do not have such information available for this study" l. In addition, the (exogenous) program eligibility indicator is included. We can express the structural equation underlying the observed behavior as: P,* = A, E, +a 7, +B':H, + O,Rj + 6i(1 where Pi is the individual's net benefit from seeking health care, E, is a binary indicator variable which takes on the value of 1 if the individual is eligible for subsidized/free health care, I, is a vector of own characteristics, Hi is a vector of households characteristics, R, is a vector of regional characteristics, and Ci is a normally-distributed error term with mean zero and variance or. We don't observe the latent variable P,*. We see only the results of the individual's evaluation of (2), which is manifest in the choice made by the individual to seek health care or not to seek health care: Pi =1 ifWP > O (2a) l We do have prices paid for medical consultation, however, given that the illness is not specified, and we are not modelmg for any specific spells of illness, we do not include that mfornation in this study. 15 P1 = 0 ifP,=l 8) and children between ages 6 through 17).18 We also ran the regressions without including household income quintiles, given that health and income are potentially simultaneously determined. Exclusion of income did not have any bearing on any of the salient findings in this analysis (results not reported). As we see in Table R3, girls are 10 percent less likely to seek health care when sick compared to boys. Besides a weak urban effect, nothing else was significant in explaining the variation in utilization rates. In the adult specification, disabled individuals are 11 percent more likely to seek health care; less likely to do so if female (albeit at a 10% level of significance); health care usage increases with education and wealth; urban denizens are more likely to seek health care compared to their rural counterparts; and, there are strong district effects (Column 1, Table R4). We present results after inclusion of SRHS in Column 2 of Table R4, however, we do not give much credence to those results given information on SRHS is missing for most of the sample. Unlike 1996 results, we no longer find any significant negative bias against girls in 1998/99 (Table R5), while on the other hand we now find that adult females are more likely to use health care (albeit significant at the 10% level - see Table R6). Disability is the only significant eligibility criterion in the child specification, and is robust to inclusion of SRHS. Education of household head is weakly significant, while seasonal and regional factors appear to be important determinants of health care utilization. In the adult specification without inclusion of SRHS, both disability and large family size are significant, albeit in opposite directions. Inclusion of SRHS somewhat weakens the significance of family size, however, the results still hold that suggest utilization decreases with family size and increases with disability. Wealth, education and seasonal factors also arise as significant determinants of adult health care usage. 15 Health posts that may be reported as Diagnostic Centers or Polyclinics in 1998/99. 16 Unspecified health care provider - there was no information on private providers in the 1996 survey. 17 Variable mteractions were included in both the probit and the multinomial logistic specifications (e.g., eligibility interacted with income), however, none of the interaction terms were significant in any of the adult or child specifications for (results not reported). 18 We do not present regression results for children aged 5 and under. Besides seasonal and regional factors, nothing is significant in explaining differences in utilization rates. This reflects the existence of auxiliary health care programs/exemptions for children, and suggests that there is considerable regional variation in these programns. 17 Did the program work ? The primary goal of this study was to help answer one fundamental question - did the program for the vulnerable population increase health care utilization in Armenia? Well, the answer primary hinges upon temporal and category factors, and first of all lets rephrase the question - was the program well targeted? In Tlables 7 amd $, we present t-test results of whether or not there was a significant difference in mean reporting of illness, income, and self-reported health status across these groups. In 1998/99, disabled individuals were both more likely to be sick and more likely to be poorer. Hence, we can say that in 1998/99 the program was appropriately targeted to the disabled, health care utilization was positively associated with disability, thus, there was a progressive income transfer to the disabled. In 1996, disabled were more likely to be sick, however, were not more likely to be poorer. As pointed out earlier, it is difficult to compare incomes across the two surveys, and anyway, it appears that targeting improved over time regarding the disabled. Results from the multinomial regressions of provider choice in both survey rounds indicate that the disabled were more likely to seek health care in Diagnostic Centers and Hospitals (see Tables IR7 and 1O8). Belonging to a family with four or more children, was not significant in explaining any of the provider choice (These findings were robust to inclusion of self-reported health status - results not reported). On the other hand, in 1998/99, even though families with four or more children were appropriately targeted, they were still less likely to utilize the health care system - when we did not observe the latter phenomenon (i.e., significantly less health care utilization) in 1996. We need more detailed institution and household level information to find out why this group is opting out of the health care system (e.g., are informal payments higher for some groups? is that in tum due to the fact that it is easier for the health facility to get reimbursed for some groups relative to others?). TablRe 7: 1996 T-test compnairmns More likely to self-report More likely have More likely to have yourself as sick ? less income ? lower SRIIS ? (full sanple) (full sample) (adult sample) Disabled nES _ s ',_ Families with 4 or more NO _ __ ns Children under 18 Children under 18 from a s ns Single Parent Household Children under 18 with a s ns Disabled Parent Note: YES, NO - represents statistically significant differences; ns - difference is not significant; t both in per-capita and non per-capita consumption and income terms 18 Table 8: 1998/99 T-test comparHsoims More likely to self-report More likely to have More likely to have yourself as sick ? less income ? lower SIRiHS ? _________________(full saniple) (full sample (adult sample) Disabled YES YLES Families with 4 or more NO y_s__ ,r | ES Children under 18 Children under 18 from a Ns YESl Single Parent Household Children under 18 with a NC lDisabled Parent Note: YES, NO - represents statistically significant differences ; ns - difference is not significant; * both in per-capita and non per-capita consumption and income terms 6. concl1usDon The primary goal of the program for the vulnerable population -- which allowed certain vulnerable groups to seek 'free of charge' medical services-- was to mitigate against the decrease in health care utilization, particularly among the poor. Unfortunately, utilization rates have continued to fall - for both the poor and the rich. Despite comparable income levels, average utilization fell by 12 percent between the two surveys. This was most probably due to a confluence of factors - increase in formal user fees, increases in informal payments, and decline in quality of service delivery. Families with four or more children, the largest beneficiary group under the program, have decreased their usage of health care services in a disproportionate manner - 21 percent reduction in usage between the two survey rounds. This precipitous drop in health care usage by this vulnerable group despite being eligible for free medical services, suggests that the program just by itself was inadequate in stemming the decline in the usage of health services. The evidence suggests that this program worked as an income transfer mechanism for the disabled and that a more credible budget reimbursement process is needed to ensure any impact among the poor (who may more price sensitive). There is no way of saying that would have happened to utilization rates among these groups if the program had not been in place. What we can say is that despite been poor, the fall in utilization rates among the disabled was similar to the sample average, and it is likely that this program helped to mitigate against fu-ther slippage in utilization rates among the disabled. There are suggestions (without empirical scrutiny) that recent policy changes in Armenia may actually have improved access to health care among the poor due to the expansion of the eligibility criteria. In 1999 the Government of Armenia implemented a major reform in their Social Assistance system moving from the categorical benefits as examined in this paper, to a poverty-targeted (means-tested) benefit program allocated at the household level (Poverty Family Benefit, PFB). The health sector in 2001 decided to include the PFB 19 expansion was accompanied by increased information about its use, proper institutional design that attenuates the incentive for informal payments, and improved quality standards, then coverage of the poor might actually have improved. This, however, remains as an unanswered empirical question.19 References The word "processed" describes informally reproduced works that may not be commonly available through library systems. Bitran, R., and U. Giedion (2002) "User fees for health care: Waivers, exemptions, and implementation issues." June 2002. World bank. Processed. Case, A., and A. Deaton (2002). "Consumption, health, gender, and poverty." Princeton University. Processed. Dow, W., P. Gertler, R. Schoeni, J. Strauss, and D. Thomas (1997). "Health care prices, health and labor outcomes: Experimental evidence." Labor and Population Program Working Paper Series 97-01, RAND, Santa Monica, Calif. European Observatory on Health Care Systems (2001) "Health Care Systems in Transition. Armenia 2001." Report N. EUR/0 1/5012669(ARM). [Retrieved on November 18, 2002 from http://www.who.dk/document1e73698.pdfl Grossmann, M. (1972) "On the Concept of Health Capital and the Demand for Health." Journal of Political Economy 80(2, April): 223-55. Idler, E. L., and Y. Benjamini (1997) "Self-rated health and mortality: a review of twenty-seven community studies." Journal of Health and Social Behavior 38(1): 21-37. Kurkchiyan, M. (1999) "Report on Health Care in Armenia." Prepared for the World Bank. Processed. Lewis, M. (2000) "Who is Paying for Health Care in Europe and Central Asia?" World Bank. Processed. Ministry of Health (2000) "Health and Health Protection. Armenia 1999." Republican Center for Information and Analysis. Ministry of Health, Yerevan. Olson, J., and P. Lanjouw (2001) "How to Compare Apples and Oranges: Poverty Measurement Based on Different Definitions of Consumption." Review of Income and Wealth, Series 47(1, March): 25-42. Strauss, J., and D. Thomas (1996) "Measurement and Mismeasurement of Social Indicators." American Economic Review 86(2, May): 30-34. (1998) "Health, Nutrition, and Economic Development." Journal of Economic Literature XXXVI(2, June):766-817. World Bank (2000) "Armenia: Institutional and Govemance Review." Report No. 20269, April 5. World Bank (2002a) "Armenia Poverty Update." Human Development Sector Unit. Europe and Central Asia Region. June. Processed. World Bank (2002b) "Armenia Public Expenditure Review." Poverty Reduction and Economic Management Unit. Europe and Central Asia Region. June. Processed. '9 We are planning to work with the latest round of the ISLS 2001 dataset to examine how changes in program design have effected health care utilization in Armenia. 20 Appendix Tables Table Al: Composition of eligibility categories (Percent) 1996 1998/99 Sample Eligible Sample Eligible Disabled 2.43 17.4 3.81 22.73 Belong to a Family 8.92 62.85 9.77 57.36 with >=4 Children Child from a Single 0.57 3.6 0.94 5.11 Parent HH Child with a Disabled 2.90 15.72 2.89 14.47 Parent Orphan 0.06 0.43 0.04 0.23 Table A2: Comparison of surveys 1996 and 1998/99 1996 1998/99 Sample size 4,260 households 3,600 households Field work * 2 months: November- * 12 months: July 1998- June December 1996 1999 * No significant inflation * Inflation adjustment during the period. needed: Food = 7%; energy and telephone prices = 20%. Major policy changes None * Elimination of energy subsidies. * Changes in social assistance programs Expenditure * 75% of the sample * All households completed a information responded aggregate diary with detailed monthly expenditures expenditures during the during the last 30 days last 30 days. * 25% filled a detailed diary * All households completed a on expenditures during the section on Annual last 30 days. Consumption for a very limited list of items. 21 Table Rl: Armenia 1996 Adult (>=18) Sample Ordered Probit Regression Dependent Variable Self-Reported Health Status 1 = bad; 2 = not so good; 3 = normal; 4 = good; 5 = very well Independent Variables Eligibility Criterion Disabled -1.145 (9.98)** Belong to a Family 0.024 With >W4 Children (0.33) Other Individual Characteristics Age 29-39 -0.323 (6.04)** Age 40-50 -0.590 (10.56)** Age 51-61 -0.937 (14.78)*" Age > 61 -1.188 (18.22)*" Female -0.078 (2.14)* Secondary 0.075 (0.95) Special-Secondary 0.060 (0.72) Post-Secondary 0.140 (1.63) Household Wealth Q2 0.058 (1.03) Q3 0.094 (1.58) Q4 0.127 (2.15)* Rich 0.052 (0.92) Comnmunity Characteristics Urban -0.262 (5.02)*y Observations 3644 F-Test for Joint Significance Income Quartiles[X2(4)] 5.23 Age Categories[x2(4)] 406.91*0 Education Categories[X2(3)] 3.83 District Fixed-Effects[X2(10)] 64.9** (not reported) Robust z statistics in parentheses * significant at 5%; ** significant at 1%. Note: Coefficients are marginal probabilities. 22 Table R2: Armenia 1998/99 Adult (>=18) Sample Ordered Probit Regression Dependent Variable Self-Reported Health Status 1 = bad; 2 = not so good; 3 = normal; 4 = good; 5 = very well Independent Variables Eligibility Criterion Disabled -1.104 (18.84)** Belong to a Family 0.162 With >=4 Children (3.67)** Other Individual Characteristics Age 29-39 -0.416 (13.47)** Age 40-50 -0.714 (22.09)** Age 51-61 -0.955 (23.77)** Age > 61 -1.368 (33.49)** Female -0.158 (7.13)** Secondary -0.008 (0.15) Special-Secondary -0.032 (0.55) Post-Secondary 0.088 (1.47) Household Wealth Q2 0.110 (3.09)** Q3 0.142 (3.92)** Q4 0.187 (5.22)** Rich 0.235 (6.20)** Community Characteristics Urban -0.007 (0.23) Month Fixed-Effects August98 -0.065 (1.10) September98 -0.082 (1.37) October98 -0.067 (1.16) November98 0.053 (0.93) December98 -0.024 (0.42) January99 -0.087 23 (1.57) February99 -0.053 (0.92) March99 -0.118 (2.09)* April99 -0.029 (0.51) May99 0.001 (0.01) June99 -0.003 (0.06) Observations 10087 F-Testfor Joint Significance Income Quartiles[x2(4)] 43.6700 Age Categories[x2(4)] 1290" Education Categories[x2(3)] 14.10" Month Fixed-Effects[x2(1 1)] 17.45A District Fixed-Effects[X2(10)] 148.4100 (not reported) Robust z statistics in parentheses A significance at 10%; * significant at 5%; " significant at 1%. Note: Coefficients are marginal probabilities. 24 Table R3: Armenia 1996 Children (>5 & <18) Sample Probit Regressions Dependent Variable Sought Health Care if Sick Independent Variables Eligibility Criterion Disabled 0.184 (1.23) Belong to a Family -0.061 With >=4 Children (0.89) Child from a 0.082 Single Parent HH (0.55) Child with a -0.081 Disabled Parent (1.05) Other Individual Characteristics Age 11-14 -0.076 (1.55) Age >=15 -0.116 (1.41) Female -0.093 (2.16)* Secondary 0.010 (0.10) Special-Secondary 0.194 (1.09) Household Characteristics Q2 -0.069 (0.99) Q3 0.021 (0.31) Q4 -0.057 (0.78) Rich 0.036 (0.53) Community Characteristics Urban 0.106 (1.76)^ Observations 583 F-Test for Joint Significance Income Quartiles[x2(4)] 4.0 Age Categories[X2(2)) 3.61 Education Categories[X2(3)] 1.25 District Fixed-Effects[X2(10)] 11.85 (not reported) Robust z statistics in parentheses A significant at 10%; * significant at 5%; ** significant at 1%. Note: Coefficients are marginal probabilities. 25 Table R4: Armenia 1996 Adult (>=18) Sample Probit Regressions Dependent Variable Sought Health Care if Sick Independent Variables (1) (2) Eligibility Criterion Disabled 0.114 0.017 (3.43)-* (0.31) Belong to a Family 0.012 0.072 With >=4 Children (0.26) (0.83) Other Individual Characteristics Health Status -0.117 (5.23)** Age 29-39 -0.036 -0.061 (0.94) (0.89) Age 40-50 -0.014 -0.158 (0.38) (2.43)* Age 51-61 0.003 -0.104 (0.08) (1.56) Age >61 -0.021 -0.096 (0.59) (1.47) Female -0.035 -0.054 (1.72)^ (1.53) Secondary 0.063 0.148 (1.97)* (2.56)* Special-Secondary 0.066 0.076 (1.86)^ (1.20) Post-Secondary 0.089 0.204 (2.33)* (3.10)t* Household Characteristics Q2 -0.024 -0.006 (0.73) (0.11) Q3 0.060 0.077 (1.82)^ (1.39) Q4 0.113 0.195 (3.55)4* (3.55)** Rich 0.138 0.158 (4.42)** (2.94)** Community Characteristics Urban 0.114 0.084 (4.15)** (1.54) Observations 2684 930 F- Testfor Joint Sign ificance Income Quartiles[x2(4)] 39.06** 20.91** Age Categories[X2(4)] 1.78 6.70 Education Categories[X2(3)] 5.71 12.96** District Fixed-Effects[X2(10)] 40.68** 29.74** (not reported) Robust z statistics in parentheses Asignificant at 10%; * significant at 5%; ** significant at 1%. Note: Coefficients are marginal probabilities. 26 Table R5: Armenia 1998/99 Children (>5 & <18) Sample Probit Regressions Dependent Variable Sought Health Care if Sick Independent Variables (1) (2) Eligibility Criterion Disabled 0.561 0.302 (4.30)** (2.04)* Belong to a Family 0.068 0.038 With >=4 children (0.77) (0.43) Child from a 0.046 0.056 Single Parent HH (0.41) (0.46) Child with a -0.100 -0.108 Disabled Parent (1.14) (1.23) Other Individual Characteristics Health Status -0.161 (4.19)** Age 11-14 -0.070 -0.094 (1.22) (1.70) Age >= 15 -0.061 -0.094 (0.87) (1.40) Ferale -0.043 -0.053 (0.81) (1.02) (0.82) Education of HH-Head Secondary -0.144 -0.150 (1.49) (1.52) Special-Secondary -0.041 -0.036 (0.41) (0.35) Post-Secondary 0.024 0.048 (0.21) (0.41) Household Wealth Q2 0.025 0.039 (0.27) (0.41) Q3 0.151 0.171 (1.69) (1.91) Q4 0.149 0.136 (1.70) (1.58) Rich 0.201 0.261 (2.19)* (2.76)** Community Characteristics Urban 0.250 0.244 (3.25)** (3.02)** Month Fixed-Effects August98 0.337 0.398 (1.99)* (2.32)* September98 0.002 0.017 (0.02) (0.12) 27 October98 -0.024 -0.032 (0.18) (0.24) November98 0.165 0.264 (1.00) (1.53) December98 0.224 0.283 (1.48) (1.86) January99 -0.157 -0.064 (1.38) (0.52) February99 -0.017 0.067 (0.14) (0.53) March99 -0.218 -0.187 (1.93) (1.58) April99 0.013 0.078 (0.08) (0.49) May99 0.160 0.271 (0.89) (1.47) June99 0.193 0.193 (1.12) (1.13) Observations 389 389 F-Test for Joint Significance Income Quartiles[x2(4)] 7.18 9.77* Age Categories[X2(2)] 1.67 3.56 Education Categories[X2(3)] 7.424 8.91t Month Categories [X2(l 1)] 30.9** 28.8** District Fixed-Effects[X2(10)] 30.56t* 25.74** (not reported) Robust z statistics in parentheses * significant at 5%; ** significant at 1%. Note: Coefficients are marginal probabilities. 28 Table R6: Armenia 1998/99 Adult (>=18) Sample Probit Regressions Dependent Variable Sought Health Care if Sick Independent Variables (1) (2) Eligibility Criterion Disabled 0.123 0.085 (3.56)** (2.38) Belong to a Family -0.108 -0.081 With >=4 children (2.28)* (1.65)- Other Individual Characteristics Health Status -0.081 (5.06)** Age 29-39 -0.005 -0.041 (0.11) (0.98) Age 40-50 -0.019 -0.060 (0.46) (1.46) Age 51-61 -0.052 -0.104 (1.24) (2.46)* Age >61 -0.008 -0.072 (0.19) (1.73)- Female 0.040 0.037 (1.77)- (1.63)" Secondary 0.053 0.064 (1.28) (1.55) Special-Secondary 0.115 0.135 (2.45)* (2.87)** Post-Secondary 0.176 0.201 (3.48)** (3.98)** Household- Wealth Q2 0.024 0.029 (0.66) 0.79) Q3 0.120 0.125 (3.24)** 3.36)** Q4 0.087 0.091 (2.35)* 2.46)* Rich 0.096 0.109 (2.74)** 3.07)** Community Characteristics Urban 0.040 .037 (1.30) 1.20) Month Fixed-Effects August98 0.008 0.007 (0.15) 0.13) September98 0.039 0.040 (0.74) 0.75) October98 -0.072 0.082 (1.29) 1.49) November98 -0.033 0.036 29 (0.59) 0.64) December98 -0.111 0.114 (2.12)* 2.18)* January99 -0.093 0.086 (1.89) 1.74) February99 -0.139 0.126 (2.93)** 2.64)** March99 -0.160 0.158 (3.29)** 3.25)** April99 -0.076 0.075 (1.40) 1.37) May99 -0.004 0.012 (0.07) 0.20) June99 -0.039 0.036 (0.70) 0.64) Observations 2002 2002 F-Testfor Joint Significance Income Quartiles[X2(4)] 14.72** 6.33* Age Categories[X2(4)] 2.53 6.63 Education Categories[x2(3)] 18.02* 2.80* Month Categories [X2(l1)] 6.22** 3.01** District Fixed-Effects[X2(10)] 0.80 14.38 (not reported) Robust z statistics in parentheses ^ significant at 10%; * significant at 5%; ** significant at 1%. Note: Coefficients are marginal probabilities. 30 Table R7: Armenia 1996 Adult (>=18) Sample Multinomnial Logit Regression Dependent Variable: Type of Health Provider Sought if Sick Note: No health care sought is the comparison group Independent Variables (1) (2) (3) (4) (5) Diagnostic Polyclinic Hospital Fyelcher Other Eligible Category Disabled 0.736 0.346 0.750 -1.585 -0.218 (2.53)* (1.69) (7.24)** (1.36) (0.40) Belong to -0.144 -0.070 0.079 0.292 0.547 a Family (0.20) (0.26) (0.37) (0.64) (1.25) with >4 Chlldren Other Individual Characteristics Age 29-39 -0.081 0.036 -0.118 0.636 -0.859 (0.42) (0.11) (0.57) (2.04)* (1.95) Age 40-50 -0.289 0.209 -0.179 0.299 -0.253 (1.41) (0.66) (0.92) (0.77) (0.77) Age 51-61 -0.268 0.390 -0.154 -0.059 -1.035 (0.78) (1.60) (0.69) (0.10) (3.58)** Age > 61 -0.558 0.170 -0.165 0.507 -0.693 (1.51) (1.05) (0.72) (1.58) (1.69) Female -0.378 0.013 -0.381 -0.204 0.052 (2.16)* (0.12) (3.75)** (1.07) (0.26) Secondary 0.227 0.197 0.198 0.851 0.177 (0.68) (1.40) (0.89) (1.41) (0.49) S-Secondary 0.386 0.251 0.140 0.678 0.087 (0.74) (1.95) (0.52) (2.49)* (0.22) Household Wealth Q2 -0.446 -0.033 0.001 -1.086 -0.178 (1.51) (0.17) (0.01) (2.01)* (0.61) Q3 0.441 0.148 0.430 0.049 0.208 (1.24) (1.25) (1.35) (0.07) (0.46) Q4 0.923 0.269 0.704 0.375 0.623 (2.49)* (1.69) (2.79)** (0.95) (1.56) Rich 0.608 0.151 1.177 -0.007 0.820 (2.18)* (0.52) (3.80)** (0.02) (2.69)** Post-Sec 0.679 0.299 0.106 -0.322 -0.176 (1.49) (2.38)* (0.40) (0.54) (0.35) Community Characteristics Urban 0.106 0.344 0.357 -0.823 0.530 (0.32) (1.26) (1.25) (0.89) (1.25) Constant -2.891 -1.513 -1.856 -3.387 -2.830 (4.05)** (4.61)** (5.45)** (5.39)** (7.40)** Observations 2684 2684 2684 2684 2684 31 F-Tests for Joint Significance Income=0[x2(4)1 22.86** 4.71 34.87** 21.83** 13.57** Age=O[X2(4)] 5.14 18.19** 2.06 8.04* 40.45** Edu=O[X2(3)] 6.76A 6.71A 1.27 6.61^ 3.35 District=01x2(l0)I 17.20A 24.07** 37.06** 77.37** 34.68** Fixed-Effects (not reported) Absolute value of z statistics in parentheses A significant at 10%; * significant at 5%; ** significant at 1%. 32 Table R8: Armenia 1998/99 Adult (>18) Sample Multinomial Logit Regression Dependent Variable: Type of Health Provider Sought if Sick Note: No health care sought is the comparison group Independent Variables (1) (2) (3) (4) (5) Diagnostic Polyclinic Hospital Fyelcher Other Eligible Category Disabled 0.671 0.306 0.509 0.057 -0.359 (3.78)** (0.52) (2.32)* (0.11) (0.47) Belong to -0.329 -34.175 -0.746 0.628 -33.402 a family (1.09) (0.00) (1.91) (1.17) (0.00) with >= Children Other Individual Characteristics Age 29-39 0.008 0.866 -0.281 0.116 0.288 (0.03) (1.25) (1.10) (0.20) (0.33) Age 40-50 0.048 -0.402 -0.289 -0.041 0.243 (0.21) (0.51) (1.18) (0.07) (0.28) Age 51-61 0.098 0.276 -0.936 -0.515 0.730 (0.41) (0.38) (3.22)** (0,78) (0.86) Age > 61 0.274 -0.398 -0.548 -0.187 0.976 (1.22) (0.52) (2.19)* (0.33) (1.22) Female 0.330 -0.105 0.060 0.005 0.186 (2.56)* (0.28) (0.39) (0.02) (0.47) Secondary 0.278 0.193 0.264 -0.223 0.342 (1.18) (0.27) (0.88) (0.40) (0.50) Special-Sec 0.623 -0.394 0.526 -0.435 1.015 (2.42)* (0.46) (1.61) (0.67) (1.39) Post-Sec 0.873 0.792 0.846 -0.145 -0.169 (3.26)** (0.95) (2.48)* (0.21) (0.19) Household Wealth Q2 0.296 0.988 -0.429 0.070 1.012 (1.42) (1.34) (1.71) (0.13) (1.42) Q3 0.673 0.956 0.218 0.414 0.834 (3.27)** (1.32) (0.95) (0.78) (1.10) Q4 0.503 0.892 0.109 0.187 0.627 (2.46)* (1.22) (0.47) (0.34) (0.82) Rich 0.580 0.727 -0.056 0.736 1.399 (3.01)** (1.00) (0.25) (1.49) (2.02)* Community Characteristics Urban 0.472 -0.690 -0.132 0.161 0.250 (2.67)** (1.31) (0.64) (0.36) (0.46) Month Fixed-Effects August98 -0.138 0.044 0.185 0.910 0.298 (0.52) (0.06) (0.49) (1.08) (0.31) September98 0.178 -0.712 0.393 0.736 -0.678 (0.66) (0.77) (1. o) (0.85) (0.54) October98 -0.863 -0.174 0.360 0.646 -0.515 33 (2.58)** (0.21) (0.97) (0.71) (0.41) November98 -0.199 -1.252 0.127 0.782 -33.501 (0.68) (1.08) (0.32) (0.87) (0.00) December98 -0.634 -0.916 -0.088 0.094 -33.372 (2.18)* (1.12) (0.22) (0.10) (0.00) January99 -0.583 -1.285 0.024 -33.557 0.642 (2.13)* (1.41) (0.07) (0.00) (0.75) February99 -0.768 -2.168 -0.628 0.036 0.963 (2.84)** (1.86) (1.65) (0.04) (1.14) March99 -1.272 -1.181 -0.043 0.084 -0.999 (4.14)** (1.41) (0.12) (0.09) (0.79) April99 -1.049 -0.954 0.464 0.743 0.458 (3.19)** (1.16) (1.26) (0.86) (0.44) May99 -0.222 -1.252 0.141 0.876 1.514 (0.73) (1.06) (0.35) (0.95) (1.75) June99 -0.299 -0.561 0.184 -33.385 0.296 (1.00) (0.60) (0.48) (0.00) (0.29) Constant -2.207 -3.873 -1.296 -4.574 -5.340 (4.48)** (2.43)* (2.39)* (3.10)** (3.48)** Observations 2002 2002 2002 2002 2002 F-Tests for Joint Significance Income=0[X2(4)] 13.63** 2.21 6.88 3.06 4.80 Age=0[X2(4)] 2.88 7.13 12.10** 1.29 3.03 Edu=O[X2(3)] 17.09** 3.20 9.80** 0.57 4.57 Month=0[x2(11)] 44.70** 9.25 13.24 4.56 11.63 District-0[X2(10)] 19.45** 4.38 19.56** 5.70 3.47 Fixed-Effects (not reported) Absolute value of z statistics in parentheses A significant at 10%; * significant at 5%; ** significant at 1%. 34 Policy Research Working Paper Series Contact Title Author Date for paper WPS2923 Does Foreign Direct Investment Beata K Smarzynska October 2002 P Flewitt Increase the Productivity of Domestic 32724 Firms7 In Search of Spillovers through Backward Linkages WPS2924 Financial Development, Property Stijn Claessens November 2002 R Vo Rights, and Growth Luc Laeven 33722 WPS2925 Crime and Local Inequality in Gabriel Demombynes November 2002 P Sader South Africa Berk Ozler 33902 WPS2926 Distinguishing between Rashmi Shankar November 2002 P Holt Observationally Equivalent Theories 37707 of Crises WPS2927 Military Expenditure Threats, Aid, Paul Collier November 2002 A Kitson-Walters and Arms Races Anke Hoeffler 33712 WPS2928 Growth without Governance Daniel Kaufmann November 2002 K Morgan Aart Kraay 37798 WPS2929 Assessing the Impact of Carsten Fink November 2002 P Flewitt Communication Costs on Aaditya Mattoo 32724 International Trade Ileana Cristina Neagu WPS2930 Land Rental Markets as an Klaus Deininger November 2002 M Fernandez Alternative to Government Songqing Jin 33766 Reallocation? Equity and Efficiency Considerations in the Chinese Land Tenure System WPS2931 The Impact of Property Rights on Klaus Deininger November 2002 M Fernandez Households' Investment, Risk Songqing Jin 33766 Coping, and Policy Preferences: Evidence from China WPS2932 China's Accession to the World Aaditya Mattoo December 2002 P Flewitt Trade Organization The Services 32724 Dimension WPS2933 Small- and Medium-Size Enterprise Leora F Klapper December 2002 A Yaptenco Financing in Eastern Europe Virginia Sarria-Allende 31823 Victor Sulla WPS2934 After the Big Bang? Obstacles to the Karla Hoff December 2002 A Bonfield Emergence of the Rule of Law in Joseph E Stiglitz 31248 Post-Communist Societies WPS2935 Missed Opportunities Innovation and William F Maloney December 2002 P Soto Resource-Based Growth in Latin 37892 America WPS2936 Industrial Ownership and Hua Wang December 2002 Y D'Souza Environmental Performance Yanhong Jin 31449 Evidence from China WPS2937 The Determinants of Government Hua Wang December 2002 Y D'Souza Environmental Performance. An Wenhua Di 31449 Townships Empirical Analysis of Chinese Policy Research Working Paper Series Contact Title Author Date for paper WPS2938 Recurrent Expenditure Requirements Ron Hood December 2002 M Galatis of Capital Projects: Estimation for David Husband 31177 Budget Purposes Fei Yu WPS2939 School Attendance and Child Labor Gladys L6pez-Acevedo December 2002 M. Geller in Ecuador 85155 WPS2940 The Potential Demand for an HIV/ Hillegonda Maria Dutilh December 2002 H. Sladovich AIDS Vaccine in Brazil Novaes 37698 Expedito J. A. Luna Mois6s Goldbaum Samuel Kilsztajn Anaclaudia Rossbach Jose de la Rocha Carvalheiro WPS2941 Income Convergence during the Branko Milanovic January 2003 P. Sader Disintegration of the World 33902 Economy, 1919-39 WPS2942 Why is Son Preference so Persistent Monica Das Gupta January 2003 M Das Gupta in East and South Asia" A Cross- Jiang Zhenghua 31983 Country Study of China, India, and the Li Bohua Republic of Korea Xie Zhenming Woojin Chung Bae Hwa-Ok WPS2943 Capital Flows, Country Risk, Norbert Fiess January 2003 R. lzquierdo and Contagion 84161 WPS2944 Regulation, Productivity, and Giuseppe Nicoletti January 2003 Social Protection Growth OECD Evidence Stefano Scarpetta Advisory Service 85267 WPS2945 Micro-Finance and Poverty Evidence Shahidur R Khandker January 2003 D. Afzal Using Panel Data from Bangladesh 36335 WPS2946 Rapid Labor Reallocation with a Jan Rutkowski January 2003 J Rutkowski Stagnant Unemployment Pool: The 84569 Puzzle of the Labor Market in Lithuania WPS2947 Tax Systems in Transition Pradeep Mitra January 2003 S Tassew Nicholas Stern 88212 WPS2948 The Impact of Contractual Savings Gregorio Impavido January 2003 P. Braxton Institutions on Securities Markets Alberto R Musalem 32720 Thierry Tressel WPS2949 Intersectoral Migration in Southeast Rita Butzer January 2003 P. Kokila Asia. Evidence from Indonesia, Yair Mundlak 33716 Thailand, and the Philippines Donald F Larson WPS2950 Is the Emerging Nonfarm Market Dominique van de Walle January 2003 H. Sladovich Economy the Route Out of Poverty Dorothyjean Cratty 37698 in Vietnam9 WPS2951 Land Allocation in Vietnam's Martin Ravallion January 2003 H. Sladovich Agrarian Transition Dominique van de Walle 37698