AIM-AP. Accurate Income Measurement for the Assessment of Public Policies. Citizens and Governance in a Knowledge-based Society

Transcription

1 Project no: AIM-AP Accurate Income Measurement for the Assessment of Public Policies Specific Targeted Research or Innovation Project Citizens and Governance in a Knowledge-based Society Deliverable 1.3a Public health in Greece Due date of deliverable: July 2007 Actual submission date: March 2008 Start date of project: 1 February 2006 Duration: 3 years Lead partner: CERES Revision [draft]

2 Health Services in Greece and the distributional impact of in-kind public transfers in Greece C. Koutsambelas Athens University of Economics and Business and CERES and P. Tsakloglou Athens University of Economics and Business, IZA and CERES Athens, June 2007 First draft, please do not quote

3 1.Introduction...3 PART The organizational structure of health care in Greece Financing mechanisms of health care in Greece Private Sector Assessment of Greek Health Care System: Efficiency and Equity The efficiency of health care system Vertical equity Horizontal Equity...15 Part Data and methodology Empirical results Size of public benefits Distributional effects Sensitivity analysis : Taking into account differences between sexes Sensitivity analysis: Taking into account private medical insurance Sensitivity analysis: Adopting an absolutist approach Limitations of the study Conclusion...35 APPENDIX A: Differentiating per age group health care costs between males and females...37 References

4 1.Introduction Empirical income studies usually rely on distributions of financial resources, thus disregarding non-financial elements such as the in kind benefits that the households derive from the consumption of public services. Since individuals derive utility from the consumption of goods and services irrespective of their origin (purchased or provided in-kind), an individual s monetary income may be considered as an insufficient approximation of his/her command over resources and, therefore, the estimates reported in the above empirical studies may be seriously biased. Consequently it seems important to establish how these non-cash incomes are distributed in the population of interest. This paper attempts to do so for one the most important in-kind public expenditures: the in kind public health transfers in Greece. From an international perspective, last decades there is a perennially growing concern on the improvement of health care in developed countries. As a fact most OECD countries spend an average of almost 9% of their GDP in health care. This trend is decisively enhanced by the state; about three quarters of the aforementioned financial resources come from state budgets. Greece is not an exception. It spends about 10% of its GDP in health care [5.1 % comes from public resources and the other 4.9 from private ones, OECD (2005)]. Regarding the health status of the population, it is considered satisfactory. In 2005, life expectancy at birth in Greece stood up at about 78.1 years, slightly higher than OECD average (77.8), life expectancy at 65 is also satisfactory [18.8 for females (OECD average 19.3), 16.7 for males (OECD average 15.9)], the infant mortality has fallen greatly during the last decades, as in other OECD countries as well, in 2004 it stood at well bellow OECD average (4.8 vs. 6.1), generally the same positive picture seems to hold, more or less, for a variety of health indicators (only obesity rates and proportion of daily smokers being an unfortunate exception). Nevertheless these encouraging statistics show a rather crude picture of health outcomes and more importantly it is very difficult to identify and measure their determinant 3

5 factors. Do medical infrastructures contribute more to public health than socioeconomic characteristics and lifestyle impacts? What are the relative merits of each factor? Would investment on the size and quality of public facilities exceed its cost? A variety of questions easily arise that challenge both academic ingenuity and political resourcefulness. Inescapably the way to move forward amounts on focusing on specific aspects of health care systems and addressing the issues one by one. Exactly this is the spirit of the present study, which, after a modest description of the Greek health, attempts to estimate the impact of in-kind public health transfers on inequality and poverty. The study consists of two distinct parts. Part 1 focuses on the institutional framework of health care in Greece, touching such issues as organizational structures, financing mechanisms, the role of the private sector etc. This assessment is accompanied with a short assessment of the system on grounds of both efficiency and equity. Part 2 diverges from the exclusively descriptive character of Part 1. In this section the Greek budget household survey and OECD data are combined in the methodological framework of insurance-based approach in order to derive estimates of the distributional effects of in-kind public transfers on aggregate inequality and poverty. PART 1 1.The organizational structure of health care in Greece Health care in Greece is provided by the National Health System (NHS), the medical units that belong to Social Insurance funds and the private sector. Almost 128 hospitals and 185 primary care units, dispersed across country belong to NHS, Social Insurance funds manage 5 hospitals and 300 health centers and finally there are also 218 private hospitals, over 400 private diagnostics centers and a large number of doctor that provide privately their 4

6 services, Tountas et al (2003). The NHS was established at 1983, satisfying a long-standing public demand, aiming at providing health care to all citizens (Universal coverage) free to the point of use despite of their economic and social status. NHS coexists with the approximately 35 social insurance funds that cover almost 95% of the population according to their occupational status (compulsory insurance). Nonetheless the high number and diversity of funds, three of them are the major one covering about 80% of the population (IKA is the scheme for employees of the private sector, OGA the scheme for farmers and OAEE for professionals). As far as benefit packages are concerned, NHS provides mainly hospital care and the Social Insurance funds a variety of services from primary health care to pharmaceuticals and dental care according to occupationally based insurance membership. Finally a special role is granted for the continuously expanding private sector, which provides hospital care, medical insurance, counseling, dental, care, diagnostic services etc. at a quality that naturally depends on the price paid. The ultimate responsible for the regulation, allocation of resources and strategic planning of the whole health care system belongs to the Ministry of Health and Social Cohesion. Ministry of Labour and Social Insurance is responsible for the supervision and the setting of the annual budget of insurance funds (the latter is a matter of negotiation between the Ministry and each fund). Finally the Ministry of Economics and Finance has a rather residual role in only financing the deficits of NHS and social insurance funds, Mossialos et al (2005). At the more decentralized level there are the PESYS (Regional Health Authorities). The role of the 17 PeSYs is to implement national goals at the regional level, coordinate activities within their responsibility areas and manage the delivery of health care. In other words, they act as intermediaries between the ministry and the hospitals. Nevertheless the decentralization process that was initially visualized at their establishment is impeded by the fact that PeSYs do not manage their own budget and therefore are accountable to the Ministry, Highlights on Health in Greece (2006). 5

7 2.Financing mechanisms of health care in Greece Barr (1988) defines three broad models of health care systems: A) The quasi-actuarial system. In this scheme insurance is either employer based or individuals purchase the medical insurance they choose from private markets. Also the factors of medical production are private. The ideal paradigm of a quasi-actuarial system is the USA health care. B) The earnings-related social-insurance contributions system. Central feature is the compulsory coverage of the working population, which is financed by earnings related contributions and possibly tax-subsidized.a well-known example of this Bismarkian type model is Germany. C) The universal medical care system. The archetypal paradigm is undoubtedly Great Britain, where the whole population enjoys full coverage financed by general taxation. Greece health care is an amalgam of all three. In a sense, Beveridgean and Bismarkian elements coexist with a still growing private market. The following diagram describes the direction of the monetary flows that finance health care in Greece. Chart 1 6

8 Taxes NHS Household Social Contribution Social Insurance Funds Private Payments (Formal and informal) Private Sector As it is seen from the chart, households finance health care through general taxation, compulsory social insurance and voluntary private payments (formal and informal). General taxation, which is the sum of direct and indirect taxes finances health care at a proportion of about 36%, social insurance by about 22% and the remaining proportion consists of private payments, Tountas et al (2005). General taxes are aimed mainly to NHS, however it is also financed by Social Insurance funds that buy services for their clients in a per diem basis. Unfortunately a significant amount of private payments, mainly informal ones, are directed to NHS (the most common form of informal payments according to anecdotal evidence is bribes to NHS stuff in order to avoid long-waiting lists). Social insurance funds buy also services from private sector in a capitation or fee for service basis. 7

9 3.Private Sector The health private sector in Greece constitutes a striking particularity of the Greek health care system, exactly because of the surprising way and intensity it evolutes. Kenneth Arrow (1963) in a very influential paper wrote; When the market fails to achieve an optimal state, society will recognize the gap and non market institutions will arise attempting to bridge the gap. The establishment of NHS at 1983 was exactly a case of a non market institution that was supposed to fill the gap, correct inefficiencies and substitute market forces in the social sensitive area of health care provision. Indeed one of the founding principles of NHS as it was stated in the first article of the law proclaimed that; Health is a social benefit, not subject to the laws of market profit 1. Moreover it was expected that the free provision of high quality public medical care to all citizens would inevitably crowd out private sector. Nevertheless, in reality it never happened. Instead of its gradual resolution, as it was forecasted, private sector experienced an impressive growth. Nonetheless it should be noted that the growth of health private sector was not exclusively a Greek phenomenon, but rather an international trend. The particularity of the Greek case lies on two facts. Firstly the driving forces behind this process are intrinsic, such as the inadequate and of low quality public medical services, Tountas et al (2005). And secondly private spending is already very high and still rising. The following table that utilizes OECD data illuminates of the situation. Among the seven AIM-countries, Greece takes the second place with respect total health expenditures as a percentage of GDP (about 10 %), note that the first position belongs to Germany (11%). However this high level of spending is not attributed to the size of the public sector, rather to the unusual high levels of private spending. Indeed, as it is apparent from table 1 private expenditures are the highest among the seven countries (4.8% for Greece, 3.7% for Netherlands, 2.4% for Germany etc) 1 Quoted from Tountas et al (2002) 8

10 Table 1: Health Expenditure as a share of GDP OECD (2005) Public Expenditure Private Expenditure Total Expenditure Germany Greece Netherlands Belgium Italy United Kingdom Ireland Regarding the composition of the private expenditures. The following chart-pie that is based on information derived from the 2004/2005 Greek Budget Household Survey gives a picture of how private expenditures are distributed across five relevant categories (dental care, hospital care, pharmaceuticals, private insurance, primary health care). Lion s share goes to primary health care (it is not a coincidence the fact that public primary health care is one of the most disorganized parts of the public health care) and dental (coverage for dental care is limited), high are also the share of pharmaceuticals (there is evidence of drug overconsumption), and last come hospital care and private insurance. The latter represents only the 4% of total health expenditures, however private medical insurance in Greece is a very promising market. 9

11 Chart 2: Composition of private health spending BHS (2004) Private Insurance Hospital Care 4% 11% Pharmaceuticals 20% Dental Care 31% Primary Health Care 34% The growth of the private sector inescapably raises distributional concerns. This task is out of the scope of the present study, which focuses on the impact of public health transfers, however the following tables attempt to shed some light to the issue. Table 2 illustrates monthly mean health expenditures per category per quintile. Expenditures on pharmaceuticals are more or less the same across quintiles (this is rather natural because of the income elasticity of pharmaceuticals). In contrast expenditures patterns are different for all other categories, there expenditures increase as we move to higher quintiles. This may happen either because high-income households substitute public medical facilities with private one or they just consume more health care services. Finally the pattern is quite acute for private medical insurance; the bottom quintile spends on average 1.1 euros in contrast with the bottom quintile, which spends 14.4 euros per month. 10

12 Table 2: Mean health expenditure per quintile Quintile Pharmaceu ticals Dental care Hospital care Private Medical Insurance Primary Health Care Total 1 26,5 23,9 8,8 1,1 34,2 94,4 2 22,6 32,9 9,6 2,6 36,3 104,0 3 24,8 44,2 15,3 3,1 46,4 133,7 4 24,0 53,5 14,7 6,9 49,2 148,3 5 29,0 87,8 25,2 14,4 74,6 230,9 Of course the above table does not take into account the relative importance of each expenditure category to total equivalized household consumption. This is done in the next table, which illustrates the shares of private health expenditure per quintile. One could expect that the figures for the poor quintiles would be very low, rising steadily as we move up to higher income classes (in a country where health care is publicly provided, private services normally should be a luxury good consumed mainly by the rich). However total private health expenditure have the highest importance with respect to consumption for the bottom quintile (6.8%). Table 3: Shares of private health expenditure per quintile Quintile Pharmaceutica ls Dental Care Hospital Care Private Medical Insurance Primary Health Care Total 1 2,1% 1,6% 0,6% 0,1% 2,5% 6,8% 2 1,3% 1,7% 0,5% 0,1% 2,0% 5,6% 3 1,1% 1,9% 0,7% 0,1% 2,1% 5,9% 4 0,9% 1,9% 0,6% 0,2% 1,9% 5,4% 5 0,8% 2,2% 0,7% 0,4% 1,9% 6,0% 11

13 4.Assessment of Greek Health Care System: Efficiency and Equity Evaluating public services means touching both the issues of efficiency and equity. The demand for efficiency exists irrespective of the public nature of the good, for wasting of resources is in any case an unwanted economic misfortune. About the efficiency of a public health care system one can question many questions; whether the size of the public sector is optimal, the composition of health expenditures the proper one, the management of institutions effective, stuff productive etc. The issue of equity is also very important even if it is less straightforward and a bit more elusive than that of efficiency. Equity has two dimensions, the vertical equity (unequal treatment of unequals) and the horizontal equity (equal treatment of equals) or to put it more simply; is public provided health care disproportionately financed by high-income people? Do two individuals with the same medical needs receive equal treatment irrespective of their socioeconomic status? These are some equity-oriented concerns that raise naturally in any health related discussion about equity. In the two following sections, we attempt a very brief description of the inefficiency and inequity concerns of the Greek public health care, based on the respective literature. 4.1 The efficiency of health care system Ideally the assessment of the macro efficiency of the system would attempt to answer the questions of whether we could have the same health outcomes but at lower cost. The practical difficulty of the question can be avoided by focusing on the cost of the system in a comparable fashion. As we have already reported Greece spends a considerable amount of its GDP (about 10%) for medical services. Table 1 indicates that this share is higher than that of many other countries such as Belgium, UK, Netherlands, Italy, and Ireland. There can be two scenarios explaining this fact: first one that Greek population enjoys high standards of medical services and second that demand for medical services is higher in Greece than in the other countries. 12

14 Both of these scenarios can be rejected as implausible. Thus one can infer that the high-observed total volume of health expenditure is an evidence of macroinefficiency. This conclusion holds for total expenditures, which is the sum of private and public expenditures. Regarding public expenditures itself, public health care is rather underfunded and maybe this explains the public dissatisfaction for the quality of public medical services that triggers high private spending. Considering now the efficiency problems at the micro level, a major issue is what is known in the literature of health economics as third-party payment problem. In this case the provider of health services is reimbursed by a third party (public insurer) and not the patient. The problem arises just because both the provider (doctor) and the consumer (patient) face zero privates costs and both have the incentive to overconsume. That private costs are zero, of course does not mean that social costs are also zero, quite the contrary, often are quite large. This is the case in Greece, where private sector is contracted by the social insurance funds and fee for services retrospectively paid. Moreover there is not any coherent national regulation that specifies the standards of services of contracted providers and the budget constraints of reimbursement. Under these circumstances it is not surprising that cost abatement attempts failed and social insurance deficits bulged. Human resources are also misallocated. According to the most recent OECD data (2005), Greece has a ratio of 4.4 doctors per 1000 inhabitants, when the OECD average is only 2.9. This oversupply of doctors is accompanied with an undersupply of nurses (3.9 nurses per 1000 inhabitants). This yields a ratio of nurses to doctors of about 0.9, the lowest among the OECD countries. As a consequence the Greek Health care relies disproportionately more to relatively more expensive inputs (not only because doctors salaries are higher, but also the per capita costs of medical universities are very high). Furthermore, it seems that are interregional disparities in the distribution of human resources. Tountas et al (2002) report that in the greater Athens area in 2000 there were 8.8 doctors per 1000 people, 13

15 whereas in other regions (Central Greece and the Aegean Islands) the corresponding ratio was less than 3.0. Finally the productivity of the NHS stuff has been many times questioned. Many studies report that hospital doctors and nurses have been accustomed to a public servant mentality, which is related to low productivity and corruption phenomena [Mossialos et al (2005), Mossialos and Davaki (2002), Tountas et al (2005), Souliotis and Kyriopoulos (2003), Davaki and Mossialos (2005), Tountas et al (2001) etc, Abel-Smith et al (1994)]. 4.2 Vertical equity Whether the health care system is vertically equitable depends on the distribution of its costs. So in this context, vertical equity is interpreted to mean financing according to the ability-to-pay principle, thus people who belong to the high-income strata should finance disproportionately more the health care system than the poor one. In principle it seems to be so, for NHS is financed by general taxation, which is supposed to be progressive by default. But in practice the literature have traced elements of regressivity that contaminate the taxation system and therefore its progressive nature seems doubtful, Liaropoulos and Tragakes (1998). First of all in Greece indirect taxation is a very significant source of tax revenue [ratio of indirect to direct taxes equals to 1.44 for 2002, Matsaganis & Tsakloglou (2004)], even if some luxury goods are taxed heavier than most other goods, indirect taxation doesn t seem to achieve any beneficial redistribution, Kaplanoglou and Newbery (2003). Another idiosyncrasy of the Greek taxation system and far more unpleasant than the heavy reliance on indirect taxation is the widespread tax evasion. The size of the hidden economy is indeed impressive, Schneider (2000) estimates it as 27% of official GDP for the year 1998 (the highest in its OECD selected sample). However from a distributional perspective it is not the size of the tax evasion that matters more, but rather how it is distributed across the income strata. Surely all 14

16 income groups more or less tax evade, but the gains from it probably are higher for the rich because they face higher marginal tax rates. Furthermore the evaluation of the progressivity of the tax system becomes even more complex if we consider tax-exempts and allowances. In any case it is improbable that this tax advantages enhance overall progressivity. Thus the progressivity of general taxation is ambiguous and the same also holds for social contributions. Social contributions, which are one of the main financing mechanisms of the health care system, are linked with earnings. But again the problem of tax evasion accompanied with the existence of upper contribution ceilings makes this source of financing seem also regressive. Finally not any extended analysis is needed for private expenditures. Out-of-pocket payments and private medical insurance are clearly regressive, market prices are the same for all, and consequently the relative burden is higher for the poor. 4.3 Horizontal Equity In our context the notion of horizontal equity refers to the fair distribution of benefits, thus a health care system is horizontally equitable if patients of equal need receive equal treatment, irrespective of their socioeconomic status. In this perspective a major inequity stems from the widespread informal payments that are directed to NHS. It is very common that many patients bribe the NHS stuff in order to jump the long waiting lists, have access to better facilities, or have treatment by doctors of their choice. Whichever the form of informal payments, it is straightforward that they create severe inequities in access, supply and quality of provided medical services. Geographical inequities are also one of the main problems of the system. There are wide discrepancies between the number of hospitals and number of hospital beds allocated in different regions. The statistics that report Tountas et al (2002) are indicative; In the greater Athens area in 2000 there were 6.4 hospital beds per 1000 population while the corresponding ratio in Central 15

17 Greece, was 1.2 beds per 1000 population. The same holds for the interregional allocation of human resources; in the greater Athens area in 2000 there were 8.8 doctors per 1000 inhabitants, whereas in other regions (Central Greece and the Aegean Islands) the corresponding ratio is less than 3.0. Naturally this results to high interregional patient flows from rural areas to urban areas. No matter that at the end of the day all these patients receive treatment, travel costs both in financial and psychological terms may be substantial. Last but no least, the fragmentation of social insurance funds has implications for horizontal equity. The quantity and quality of the benefit packages that are provided by the several social insurance funds to their claimants differ a lot. This differential treatment is strongly embedded in the Greek political system and has its roots in the political polarization that characterizes Greek society the last half of the century, Sotiropoulos (2003). Be that as it may, the fact is that the funds for civil servants, bank officers, public utility employees are not accidentally called noble funds in the colloquial, for they offer to their beneficiaries the most comprehensive benefit packages and wider freedom of choice of medical services and providers. Part 2 1.Data and methodology The data used in the paper are the micro-data of the 2004/5 Greek Household Budget Survey, which was carried out by the National Statistical Service of Greece. The survey covers all the private (non-institutional) households of the country and its sampling fraction is 2/1000 (around 6,500 households or 18,000 individuals). The baseline distribution is the distribution 16

18 of equivalized household disposable income net of taxes and social contributions. The equivalence scales used are the so-called modified OECD equivalence scales (Hagenaars et al, 1995) that assign weights of 1.00 to the household head, 0.50 to each of the remaining adults in the household and 0.30 to each child (person aged below 14) in the household. All monetary values were expressed in constant mid-2004 values in order to remove the impact of inflation. Since the estimates in the HBS are expressed in monthly figures, the cost estimates of public health services are adjusted accordingly. The aim of the study, as it was stated already in the introduction, is to estimate the distributional impact of in kind public transfers. Generally speaking, this kind of estimation consists of two stages; at the first stage the beneficiaries of the public benefit are identified in the sample, at the second stage an estimated monetary benefit is imputed to them. The augmented distributions of income are then available to compare with the baseline distribution, and draw conclusions on the distributional impact of public transfer. In our context the first stage is simply passed over. For, according to insurance-based approach every citizen is entitled to public health services benefit (in Greece 95% of the population has social insurance and the other 5% of the uninsured population, at least in principle, can receive treatment in a public hospital). To put it simply, everyone in the sample receives the health benefit whether he or she has used public facilities or not. The economic rationale of this approach is that people derive utility from insurance per se. As far as the estimation of the health benefit is concerned, it is assumed that the value of the transfer to the beneficiary is equal to the average cost of producing the public good, we furthermore assume that the benefit is shared by all household members (not only the direct beneficiary); in other words, we implicitly assume that in the absence of the public transfer the burden of financing the provision of health services would be born by the household. Similar assumptions are standard practice in the analysis of the distributional impact of publicly provided services. Health benefit is not the same across 17

19 beneficiaries, because certain groups enjoy higher benefits. In the context of public health, it is well known that public costs differ across the various age groups, exactly because needs also differ. Thus utilizing OECD data we created tables of public health costs per capita per age group. As it can be seen from the following table 2, the derived cost patterns are the expected one. Costs are a bit high for the infants, they decline until adulthood, they remain more or less steady for the age groups 18-49, and then they rise sharply for the elderly. Thus, the curve reflects the health needs of the population as they vary with respect the age. Furthermore we implemented some additional scenarios of benefit attribution in order to test the robustness of our results to alternative plausible assumptions. First of all the cost diversification may be in reality depend on other variables besides age. Thus we iterated our calculations taking into account the sex of the beneficiary. Unfortunately relevant cost data for Greece were not available, hence we used cost patterns from other countries assuming that there is not any specific reason that the cost patterns in Greek case should be substantially different. In another application we considered another characteristic of the Greek Health care system, which is the coexistence of private and public insurance. People that buy private insurance possibly are reluctant of using public services, thus they either utilize a fraction of the public benefit or none at all. In this case expenditures for private insurance are deducted from the health benefit. 2 A detailed description of how the per capita costs were estimated, can be found in the Appendix 18

20 Table 4: Health costs per capita per age groups Age groups 19 Health Costs

21 2.Empirical results 2.1 Size of public benefits As it was foresaid the distributional impact of a public benefit depends on the location of the beneficiaries in the income distribution and the size of the public benefit. In our case everybody in the population receives the benefit, however some age groups (in fact the elderly) enjoy a higher benefit. So what really matters from a distributional perspective is where the most advantaged beneficiaries are located in the income distribution and how generous is the public benefit. Table 5 attempts to address these questions. This table depicts estimates of the mean transfer per capita and of the proportional increases in the disposable incomes per quintile due to the inclusion of the public health benefit in the income distribution. These estimates algebraically are defined as follows: Income distribution is divided into i quintiles ( i = 1,2...5 ) Benefit is arbitrarily divided into j benefits (j=1,2,3 for benefits attributed to age groups 0-14,15-64,over 65 respectively) Bij is the total benefit j accrued to quintile i B i is the total benefits accrued to quintile i Bj is the total benefit j accrued across all quintiles nn are the total population and the population of quintile i, i respectively M ij is the mean transfer to quintile i due to the inclusion of benefit j M i is the mean transfer to quintile i M j is the mean transfer of benefit j accrued to the whole population 20

22 Δy ( ) y ij is the proportional increase to disposable income of quintile i due to the inclusion of benefit j (similar interpretations as above hold for Δy ( ) y j Δy and ( y ) i ) It is also straightforward that M = B / n, M = B / n, M = B / n and ij ij i i i i ni ni ni ni ni Δy ( ) = [( y+ B ) y]/ y= B / y= ( B / n )/( y/ n ) = M / y y, ij ij ij ij i i ij i k= 1 k= 1 k= 1 k= 1 k= 1 j j Table 5 % Increase in disposable income due to transfers to persons aged Mean transfer per capita All All Quintile C D E F G H I J 1 (bottom) 2,3 10,5 20,3 33,2 5,6 25,3 49,1 80,0 2 1,4 7,0 9,2 17,6 5,7 28,1 37,2 71,0 3 1,0 5,9 5,2 12,1 5,3 31,6 27,5 64,5 4 0,7 4,8 2,9 8,4 5,3 34,5 20,6 60,4 5 (top) 0,4 2,8 1,5 4,7 4,8 36,7 19,6 61,1 All 0,8 4,9 4,8 10,6 5,3 31,3 30,8 67,4 Mean transfers to the age group 0-14 are more or less evenly spread quintiles and relatively modest (from 4,8 to 5,7). In contrast transfers to middle aged group are of considerable size (from 25,3 to 36,7) and higher for the average member of the top three quintiles, finally transfers to the elderly (over 65) are especially high (49,1 and 37,2) for the two bottom quintiles and lower for the top ones. Of course these estimations are obfuscated by demographics factors and it is difficult to draw definite interpretations, nevertheless an important implication for our analysis is that the poor quintiles enjoy a higher average benefit than the other three, mainly because more elderly (who as it 21

23 was explained early enjoy higher health benefits than the other age groups) are concentrated to them. Regarding proportional increases in the income of the various population quintiles resulting from public health transfers, patterns are clear. In all cases, the increase in the quintile income diminishes as we move up the income distribution. All transfers taken together account for over a third of the income of the bottom decile, the corresponding share declining gradually as we proceed to higher deciles, reaching 4.7% in the case of the top decile. These figures, undoubtfully, suggest the strong progressive nature of the public health benefit. 2.2 Distributional effects This sub-section is primarily devoted to the examination of the impact of public transfers to aggregate inequality and poverty. Before moving to the analysis of the aggregate distributional effects of public health transfers. Table 6 examines the impact of public health transfers on aggregate inequality; that is, it reports the proportional change in a number of inequality indices when we move from the distribution of disposable income to the distribution of disposable income augmented by the public transfers. As inequality indices we chose the widely used Gini index and two members of the parametric family of Atkinson (1970) indices. The value of the inequality aversion parameter in the latter is set at (e=0.5 and e=1.5). Both indices satisfy the desirable properties for an inequality index (anonymity, mean independence, population independence, transfer sensitivity). Higher values of e make the Atkinson index relatively more sensitive to changes closer to the bottom of the distribution while, in practice, the Gini index is relatively more sensitive to changes around the median of the distribution [Cowell (2000), Lambert, (2001)]. Regarding relative poverty table 6 also reports the changes in the values of a number of poverty indices when we move from the distribution of disposable 22

24 income to the augmented distribution of resources. Caution is required here, as it is clear that these in-kind services are not precisely equivalent to cash disposable income. Nevertheless, it is instructive to consider how relative income poverty measures change with a shift in the definition of resources. The table depicts changes in relative poverty; i.e. the poverty line is moving with the median of the distribution. More specifically, we adopted the approach of Eurostat and set the poverty line equal to 60% of the median of the corresponding distribution. The poverty indices selected belong to the parametric family of Foster et al (1984) (FGT). When the value of the poverty aversion parameter is set at a=0, the index becomes the widely used head count poverty rate, that is the share of the population falling below the poverty line. When a=1, the index becomes the normalized income gap ratio, while when a=2 the index satisfies the axioms proposed by Sen (1976) (anonymity, focus, monotonicity and transfer sensitivity) and is sensitive not only to the population share of the poor and their average poverty gap, but also to the inequality in the distribution of resources among the poor. When moving from the distribution of disposable income to the augmented distribution of resources that includes both disposable income and the value of health transfers, the Gini index declines by 10.22%, while the two Atkinson indices decline by around 20%. The bulk of the strong equalizing effect of public health benefits can be attributed to the over 65 transfers, even if the partial effects for the other age groups are also strong. Even stronger are the effects regarding relative poverty, The results reported in Table 6 suggest that poverty is reduced according to all versions of FGT index after the inclusion of health transfers in the concept of resources. The poverty reducing effect of health transfers is enhanced for higher values of the poverty sensitivity parameter alpha; when α=2 (highest value of the parameter) recorded poverty is reduced by an impressive 56%. Once again the bulk of the poverty reduction can be attributed to the health benefits of the elderly. The significant recorded decline in both relative inequality and poverty should not be surprising given 23

25 that in our setting the benefit is universal 3 (i.e. everyone is entitled to it) and the elderly (who enjoy a higher benefit) are located relatively more in the bottom part of the income distribution. Table 6 Proportional change due to transfers to persons aged Inequality and Value of the index All poverty indices A B C D E F Gini 0,3260 0,2927-1,1-3,6-5,5-10,2 Atkinson0.5 0,0867 0,0697-2,3-7,5-10,1-19,7 Atkinson1.5 0,2432 0,1901-3,1-8,6-9,8-21,8 Poverty Rate (FGT0) 0,1972 0,1225-9,7-14,4-25,8-37,9 Normalised Poverty Gap (FGT1) 0,0537 0, ,7-21,0-29,9-48,8 Poverty Intensity (FGT2) 0,0226 0, ,2-26,0-30,0-56,3 2.3 Decomposition Analysis This section is devoted to decompositions of aggregate inequality and poverty and attempts to answer the question how does the inclusion of public education transfers in the broader concept of resources affect the structure of inequality and poverty?. This is primarily accomplished by the decompositions of inequality and poverty by population sub-groups that are presented in Tables 7 and 8. Table 7 reports the results of inequality decomposition analysis by population subgroup using as index of inequality the mean logarithmic deviation (second Theil index Theil (1967), Shorrocks (1984)) that is strictly additively decomposable, when the population is 3 Technically speaking if we add to everyone in the population an equal amount, relative inequality by definition would decline. In our case the health benefit is not equal across population, nevertheless there is not very important differences among the benefits the various age groups enjoy. 24

26 partitioned into mutually exclusive and exhaustive groups according to household type, socioeconomic group and educational level of the household head and age of the population member. After the inclusion of public health benefit, aggregate inequality declines by 22.1%. Irrespective of the partitioning of the population, inequality within particular population groups declines almost always, as does the share of the within groups and between groups component in aggregate inequality. When examining the contribution of particular population groups to aggregate inequality, we observe that the contribution of groups with elderly people declines (older single persons or couples, pensioners, persons aged above 65) while that of groups with children rises (couple with children up to 18, persons aged bellow 25). Table 8 tries to identify population group at high of poverty and how this risk and their contribution to aggregate poverty change after the change in the concept of resources. The task is accomplished thanks to the property of additive decomposability that FGT index possess. The differentiation of population was done by the same criteria as in the preceding inequality decomposition analysis. In general, after the inclusion of public health benefit in the concept of resources the poverty risk declines in all population sub-groups. Again we observe that the contribution of groups with elderly people to aggregate poverty sharply declines, especially for high values of the sensitivity parameter. 25

27 Table 7 Characteristic of household or household head A Β C D E F Household type Older single persons or couples (at least one 65+) , ,3 0,06 0,06 Younger single persons or couples (none 65+) , ,9 0,24 0,22 Couple with children up to 18 (no other HH members) , ,4 0,34 0,35 Mono-parental household , ,7 0,02 0,02 Other household types , ,5 0,32 0,33 % Within groups inequality , ,3 97,6 98,1 % Between groups inequality , ,9 2,4 1,9 Socioeconomic group of HH head Blue collar worker ,13 0,14 White collar worker ,09 0,10 Self-employed ,33 0,33 Unemployed ,02 0,01 Pensioner ,27 0,25 Other ,08 0,08 % Within groups inequality % Between groups inequality Educational level of HH head Tertiary education ,16 0,17 Upper secondary education ,22 0,23 Lower secondary education ,11 0,11 Primary education or less ,35 0,32 % Within groups inequality % Between groups inequality Age of HH member Below ,25 0, ,51 0,52 Over ,20 0,18 % Within groups inequality % Between groups inequality A: Population Share B: Mean Log Deviation (Disposable Income) C: Mean Log Deviation (Disposable Income + Health Benefit) D: % Change in Inequality E: % Contribution to Aggregate Income Inequality (Disposable Income) F: % Contribution to Aggregate Income Inequality (Disposable Income + Health Benefit) 26

28 Characteristic of household or household head Household type Table 8 Popul. FGT0 FGT1 FGT2 Share A B C D A B C D A B C D Older single persons or couples (at least one 65+) 7.8 0, ,9 0,15 0,07 0, ,0 0,14 0,03 0, ,7 0,11 0,02 Younger single persons or couples (none 65+) , ,4 0,21 0,22 0, ,9 0,24 0,23 0, ,7 0,27 0,24 Couple with children up to 18 (no other HH members) , ,5 0,34 0,40 0, ,2 0,34 0,42 0, ,1 0,35 0,44 Mono-parental household 1.5 0, ,9 0,02 0,03 0, ,9 0,02 0,03 0, ,2 0,03 0,04 Other household types , ,9 0,28 0,29 0, ,8 0,26 0,28 0, ,0 0,25 0,27 Socioeconomic group of HH head Blue collar worker ,1600-9,6 0,19 0,22 0, ,7 0,15 0,19 0, ,3 0,13 0,15 White collar worker , ,3 0,03 0,03 0, ,0 0,01 0,02 0, ,5 0,01 0,01 Self-employed , ,7 0,28 0,32 0, ,8 0,32 0,40 0, ,0 0,37 0,47 Unemployed 2.3 0,3337-1,5 0,04 0,05 0, ,5 0,04 0,05 0, ,0 0,03 0,05 Pensioner , ,4 0,35 0,26 0, ,3 0,35 0,19 0, ,5 0,32 0,15 Other 8.4 0, ,8 0,11 0,13 0, ,1 0,13 0,15 0, ,3 0,14 0,17 Educational level of HH head Tertiary education , ,2 0,04 0,05 0, ,2 0,04 0,04 0, ,9 0,03 0,04 Upper secondary education ,1532-8,4 0,21 0,25 0, ,9 0,21 0,28 0, ,5 0,22 0,29 Lower secondary education , ,7 0,14 0,15 0, ,2 0,13 0,16 0, ,8 0,14 0,16 Primary education or less , ,0 0,61 0,55 0, ,9 0,62 0,52 0, ,1 0,60 0,54 Age of HH member Below ,2096-8,6 0,29 0,34 0, ,0 0,30 0,38 0, ,3 0,31 0, , ,2 0,40 0,45 0, ,3 0,39 0,46 0, ,0 0,40 0,48 Over , ,1 0,32 0,21 0, ,6 0,32 0,15 0, ,6 0,29 0,11 All 0, ,0 100,0 100,0 0, ,1 100,0 100,0 0, ,1 100,00 100,00 A: Value of the Index (Distribution of Disposable Income) B: % Change in Poverty (after the inclusion of Health Benefit) C: % Contribution to Aggregate Poverty (Disposable Income) D: % Contribution to Aggregate Poverty (Disposable Income + Health Benefit) 27

29 2.3 Sensitivity analysis 2.3.1: Taking into account differences between sexes Health needs may differ between sexes and especially if one approaches the issue from a life cycle perspective. In order to capture such effects in our analysis we used the age profiles for public expenditure for males and females for Sweden, from which we derived a cost pattern that we applied to the Greek data assuming that more or less these patterns are the same across countries 4 in the Appendix there is a detailed description of the calculations. As it can be seen from Table 9 the differences of the distributional indices between this scenario and the baseline one are not significant. Any clear patterns are very difficult to discern, thus we can conclude that the results of the baseline scenario are insensitive to the assumption that health care costs differ between males and females. Table 9 Proportional change due to transfers to persons aged Inequality and Value of the index All poverty indices A B C D E F Gini 0,3260 0,2927-1,1-3,7-5,4-10,2 Atkinson0.5 0,0867 0,0697-2,3-7,5-10,0-19,6 Atkinson1.5 0,2432 0,1902-3,1-8,6-9,7-21,8 Poverty Rate (FGT0) 0,1972 0,1213-9,5-14,7-26,0-38,5 Normalised Poverty Gap (FGT1) 0,0537 0, ,7-20,8-29,8-48,4 Poverty Intensity (FGT2) 0,0226 0, ,1-25,9-30,0-56,0 4 Once again in the Appendix there is a detailed description of the calculations made.

30 2.3.2 Sensitivity analysis: Taking into account private medical insurance As it was stated earlier one of the striking characteristics of the health care in Greece is the significant role of the private sector. A still growing percentage of people buy private medical insurance (this category of private expenditure is the most relevant to our approach.). It is quite plausible that people that have private contracts substitute wholly or partly public services with private ones. To capture this effect we decided to estimate net of private insurance expenditure health benefits. Thus: NB i =Β apmi, i i NB i is the public health benefit net of private insurance expenditures for person i, B and i PMI i are the public health benefit and private medical insurance expenditure for person i, parameter a measures the intensity of substitution between public and private insurance, so the higher the value of the parameter the more people substitute public with private facilities. The parameter was set arbitrary to 0.5 and 1 in our estimations. A natural objection against this methodology is that people with private contracts still receive the public insurance and thus they enjoy the whole public health benefit. Even if this is a legitimate counterargument, the truth is that in Greece many people choose private medical insurance not only because they opt for high quality services but also because they are totally dissatisfied with public services. Possibly these people would prefer not to have public insurance at all, were it noncompulsory. If this is true, the imputation of the public benefit overestimates their utility. In table 10, parameter a is set to one, again the decline in aggregate inequality and poverty is significant, Gini declines about 10.5%, while the two Atkinson indices decline by about 20% and 22% respectively. Also FGT indices decline by about 39% for the Poverty Rate up to 56% for the FGT2. It is important to notice that in this setting the decrease in distributive indices is even more pronounced than in the baseline scenario. This is natural; people that have private contracts belong to the higher quintiles, when their benefits are reduced because of the deduction of the private expenditures for 29

31 private medical insurance, the overall progressivity of the public benefit is enhanced. Table 10 Proportional change due to transfers to persons aged Value of the index Inequality and All poverty indices A B C D E F Gini 0,3260 0,2919-1,09-3,90-5,49-10,46 Atkinson0.5 0,0867 0,0694-2,34-7,91-10,14-19,98 Atkinson1.5 0,2432 0,1896-3,12-8,93-9,81-22,03 Poverty Rate (FGT0) 0,1972 0,1198-9,71-17,08-25,90-39,24 Normalised Poverty Gap (FGT1) 0,0537 0, ,58-22,03-30,15-49,42 Poverty Intensity (FGT2) 0,0226 0, ,01-26,83-30,28-56,49 In table 11 the same exercise is repeated but now the sensitivity parameter is set to 0.5. The above conclusions still hold, the only difference is that now the progressivity enhancement due to the deduction of private medical insurance expenditures is a bit smaller. 30

32 Table 11 Proportional change due to transfers to persons aged Inequality and Value of the index All poverty indices A B C D E F Gini 0,3260 0,2921-1,09-3,79-5,48-10,41 Atkinson0.5 0,0867 0,0694-2,33-7,73-10,13-19,91 Atkinson1.5 0,2432 0,1896-3,10-8,78-9,79-22,01 Poverty Rate (FGT0) 0,1972 0,1210-9,75-15,34-25,76-38,65 Normalised Poverty Gap (FGT1) 0,0537 0, ,62-21,68-29,84-48,87 Poverty Intensity (FGT2) 0,0226 0, ,10-26,59-29,99-56, Sensitivity analysis: Adopting an absolutist approach In line with the standard analysis of inequality, the above analysis is based on relativities since it is based on the mean independence axiom. This axiom is used in the framework of inequality analysis in order to avoid getting different estimates of particular inequality indices when the income distribution is measured in different metric units (dollars, euros, pounds, etc.). However, in the framework of the present analysis it can have a perverse effect, since in order to keep the level of inequality constant, the beneficiaries should receive transfers proportional to their (equivalized) disposable income. This is a rather unusual treatment that contravenes the very rationale behind of public transfers. At least according to the Greek constitution, two beneficiaries with equal needs should be entitled to an equal amount of public health transfers, irrespective of their income. Under these circumstances, it may be preferable to base our analysis on absolute rather than relative inequality indices [Kolm (1976), Blackorby C, and Donaldson D. (1980)]. This is done initially for the entire distribution in Table 12. The index used is the extended Gini index 31