WPS4371


Policy ReseaRch WoRking PaPeR                      4371




            Europe and Central Asia's Great
     Post-Communist Social Health Insurance
                           Experiment:

    Impacts on Health Sector and Labor Market Outcomes

                                Adam Wagstaff
                             Rodrigo Moreno-Serra




The World Bank
Development Research Group
Human Development and Public Services Team
October 2007

Policy ReseaRch WoRking PaPeR 4371


 Abstract

 The post-communist transition to social health insurance                             that, controlling for any concurrent provider payment
 in many of the Central and Eastern European and                                      reforms, adoption of social health insurance increased
 Central Asian countries provides a unique opportunity to                             national health spending and hospital activity rates, but
 try to answer some of the unresolved issues in the debate                            did not lead to better health outcomes. The authors also
 over the relative merits of social health insurance and tax-                         find that adoption of social health insurance reduced
 financed health systems. This paper employs a regression-                            employment in the economy as a whole and increased
 based generalization of the difference-in-differences                                unemployment, although it did not apparently increase
 method and instrumental variables on panel data from                                 the size of the informal economy.
 28 countries for the period 1990-2004. The authors find




 This paper--a product of the Human Development and Public Services Team, Development Economics Research
 Group--is part of a larger effort in the department to shed new light on health financing and delivery issues. Policy
 Research Working Papers are also posted on the Web at http://econ.worldbank.org. The corresponding author may be
 contacted at awagstaff@worldbank.org.




         The Policy Research Working Paper Series disseminates the findings 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 fully 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 views of the International Bank for Reconstruction and Development/World Bank and
         its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.


                                                         Produced by the Research Support Team

         Europe and Central Asia's Great Post-Communist
                    Social Health Insurance Experiment:
    Impacts on Health Sector and Labor Market Outcomes

                                                    by

                          Adam Wagstaffa and Rodrigo Moreno-Serrab

                     aDevelopment Research Group, The World Bank, Washington DC, USA
        bDepartment of Economics & Related Studies, University of York, York, YO10 5DD, United Kingdom




Contact author: Adam Wagstaff, World Bank, 1818 H Street NW, Washington, D.C. 20433, USA. Tel.
(202) 473-0566. Fax (202)-522 1153. Email: awagstaff@worldbank.org.

Keywords: Social health insurance; impact evaluation; health sector reform; Europe; Central Asia.

Acknowledgements: Our thanks to the World Bank's Research Support Budget for financial support, to
Armin Fidler, Ana Djordjevic, Peyvand Khaleghian, Jack Langenbrunner, Silvia Mauri and Pia Schneider for
information on social health transitions in Europe and Central Asia, and to the participants of seminars at
the World Bank, the University of York, and the Nordic Health Economists Study Group for helpful
suggestions. The findings, interpretations and conclusions expressed in this paper are entirely those of
the authors, and do not necessarily represent the views of the World Bank, its Executive Directors, or the
countries they represent.

                                                             1




1.       Introduction


         Ninety percent of the OECD countries finance the majority of their health expenditures

publicly (the exceptions are Mexico and the United States). The OECD is split roughly equally

into tax-financed systems and social health insurance (SHI) systems: in half of the 30 countries,

SHI contributions make up a majority of general government spending on health. In the rest of

the world, the fraction of countries financing the majority of their health spending publicly is

smaller (56% compared to 70% in the OECD), and only 21% finance the majority of their

government spending through SHI.1



         The relative merits of SHI and tax finance is an old debate, but one where the evidence is

surprisingly thin. There is fairly clear evidence that payments for health care are more

progressive or less regressive in tax-financed systems than in SHI systems.2 It is also clear that

tax-financed systems are more successful at ensuring universal coverage within a single health

system; SHI systems, by contrast, typically struggle to cover the informal sector and the poor.3

But the presumption that those not covered by a SHI scheme fare less well--in terms of the

quantity and quality of care they receive, the amount they pay for it out-of-pocket if they get it,

or both--is not always borne out by the limited evidence available.4




1 The figures in this paragraph are calculated from data in the World Health Report annexes.
2 Wagstaff et al. (1992; 1999) find that SHI is less progressive than tax-financed systems (in fact, is mostly
regressive) in the OECD countries. O'Donnell et al. (2007) find the same in Asia.
3 Carrin and James (2004) document the time it took several of the SHI countries to achieve universal coverage.
Baeza and Packard (2006) document coverage in several Latin American countries. Wagstaff (2007) includes
coverage statistics for several countries, as well as for different groups.
4 Being in a SHI scheme is typically found to be associated with higher rates of utilization, but not always lower
levels of out-of-pocket spending or a lower risk of especially large out-of-pocket payments--see e.g. Gertler and
Solon (2000), Waters, Anderson and Mays (2004), and Wagstaff and Lindelow (2005).

                                                       2


        Important as these issues relating to payments, coverage and utilization among

subpopulations are, they are not the focus of this paper. Rather the focus is on the relative merits

of SHI and tax-financed systems at the population level. Here too the evidence is thin. There is

little systematic evidence on whether SHI systems tend to spend more on health care, and if so

whether this translates into superior health outcomes. There are those who argue that SHI

enables higher levels of health spending. The population may, it is argued, be more willing to

pay SHI contributions than (other) taxes because the revenues are earmarked for health services

and contributions confer entitlements to use them. There are those too who argue that SHI

systems are more efficient at transforming money into health, because they more easily permit a

separation between the purchasing of health care (done by a SHI agency kept on its toes by the

contributors) and the provision of health care (which could be done by either government or

private providers).



        The evidence on these points is virtually nonexistent and there are counterarguments.

Contributions are linked to earnings through a formula and typically subject to ceilings that may

change infrequently, with the result that at times of rapid growth SHI revenues may not keep

pace with per capita incomes and a tax-financed system might produce higher revenue growth.5

Some governments have reduced their tax-financed health spending as SHI contributions have

increased, raising the suspicion that total government spending on health may not increase

following SHI adoption.6 Evasion in SHI schemes has proved a major problem7, so if the

treasury bases its tax-financed spending on theoretical contributions, total government spending

on health may actually fall following SHI adoption. Some SHI agencies have become corrupt

and unresponsive to their contributors, selecting providers through cronyism rather than through

5 See, for example, Lu and Hsiao (2003) on Taiwan's experience in this regard.
6 See, for example, Twigg (1999) on Russia's experience.
7 See, for example, Escobar and Panopolou (2003) on Colombia, and Twigg (1999) on Russia.

                                                            3


transparent and competitive contracts. 8 SHI agencies may be more prone to "capture" by

provider interest groups who may secure better terms (including higher wages) for providers.

One suspects that SHI systems are also likely to be administratively expensive, often requiring

revenue-collection efforts separate from those of the tax authorities, a contracting apparatus, and

where multiple SHI schemes exist alongside one another a loss of monopsony power. It is not

implausible that SHI systems spend less and are less successful at translating money into better

health. But even on these counterarguments, the evidence is at best anecdotal.



         It is not just with respect to the health sector that there is a lack of evidence on the merits

and demerits of SHI and tax-financed health systems. The same is true of their consequences for

labor markets. Some of the "old" SHI countries--notably France, Germany and the

Netherlands--have recently been reducing or are trying to reduce their reliance on payroll

financing9, in part out of a belief that it has contributed to unemployment and informalization of

the labor market. These issues have also been debated in Latin America.10 Yet there is really

very little direct evidence on this issue, most taking the form of indirect evidence from studies of

the employment effects of payroll taxes.11 This indirect evidence, however, refers to all payroll

taxes which is problematic as the impacts on labor supply may be different between, say,



8 A World Bank report in the late 1990s on health insurance in Argentina (World Bank 1997), commenting on the
purchasing capacity of health insurers, argued that "it is acknowledged by Argentines that personal connections and
corrupt practices, instead of quality and economy, weigh heavily in the award of capitated contracts and other
payments to medical providers and suppliers, and this adds substantially to the inefficiency and high cost of health
care in Argentina" (p.7).
9 France widened the tax base from earnings to include nonwage income. Germany is contemplating reducing the
emphasis on the payroll, while the Netherlands in 2005 introduced a reform where insurers receive only half their
income from payroll revenues (albeit channeled through a central fund), the rest coming from flat-rate direct
contributions from members (with offsetting income supplements for low income groups) (Gottret and Schieber
2006; International Network on Health Policy & Reform 2006).
10 See Baeza and Packard (2006) on this.
11 Kugler and Kugler (2003) have estimated the impact of payroll tax increases in Colombia that were implemented
in parallel with the health financing reforms. According to their estimates, the health financing reforms, which raised
the payroll tax rate by 5 percentage points, would have reduced wages by between 0.7% and 1.1%, and employment
by between 2% and 2.5%. By contrast, Bauer and Riphahn (2002) find that Germany's payroll tax has had very
limited employment effects.

                                                  4


pensions and health services, especially in poorer countries, since people may think it more

likely they will benefit from the health services they get to use from SHI contributions than the

pensions they may never enjoy resulting from pension contributions because they fear they will

not live long enough. Moreover, studies of the labor market consequences of payroll tax changes

do not answer the question of what happens to employment and informality when smaller (or

larger) payroll taxes are replaced by larger (or smaller) other taxes.



        Getting at relative merits and demerits of SHI and tax-financed health systems through a

cross-country econometric analysis where some systems are financed through SHI contributions

and others are financed through general revenues is clearly problematic because there are likely

to be unobservable variables that are correlated with both the type of financing system in place

and the outcomes of interest (i.e. SHI is potentially endogenous). A potentially more promising

strategy would be to look for changes in the way countries finance their health care, exploiting

the variations in changes across countries to eliminate (time-invariant) unobservable variables.

The difficulty with this approach is that in the group of countries that have the best data (the

OECD), there have been very few switches between the SHI and tax-financed camps (six "old"

OECD countries abandoned SHI in the 1970s and 1980s, notably Denmark, Greece, Iceland,

Italy, Portugal and Spain) and the transitions occurred some time ago, so the data available are

very limited.



        This paper looks instead to a (mostly) different group of countries where transitions have

occurred with greater frequency and more recently, namely the countries of (central and eastern)

Europe and Central Asia (ECA). Of the 28 countries that are part of the World Bank's ECA

region, 14 abandoned tax-finance and adopted SHI at some stage between 1990 and 2004 (and 4

other countries had adopted SHI prior to 1990). These countries provide not only an interesting

                                                             5


"experiment" from the point of view of health financing, but also are data-rich countries, having

inherited and largely maintained the communist tradition of extensive data-gathering, and falling

under the most data-rich regional office of the World Health Organization.12 The ECA health

financing experiment thus affords a valuable "laboratory" to try to answer some of the

unanswered questions listed above.



         To get at these issues, we use a regression-based generalization of the differences-in-

differences (DD) method on panel data from the 28 ECA countries for the period 1990-2004. We

explore different approaches to allowing for the possible endogeneity of SHI. The first is a

simple individual-specific effects model estimated along the lines of the DD approach. This

allows for the endogeneity of SHI insofar as the unobservables that are correlated with SHI

adoption and with our outcomes are time-invariant. We explore two more specifications. One

allows for a time-varying unobservable that can be correlated with SHI adoption and outcomes,

and whose growth rate is allowed to vary from one country to the next. The other approach is

Instrumental Variables (IV), implemented in the case of those outcomes for which the DD

generalizations do not seem to control adequately for the possible endogeneity of SHI adoption.



         The organization of the paper is as follows. Section 2 provides a brief history of the SHI

reforms in the post-communist ECA region. Section 3 outlines our methods. Section 4 describes

our data. The empirical results for the health and labor analyses are presented in sections 5 and 6,

respectively. Section 7 discusses the results and presents our conclusions.




12The European office of the World Health Organization developed and has maintained a huge database to track
progress towards its Health for All initiative. In addition, it is home to the European Observatory on Health Systems
and Policies, which has produced detailed overviews of the health systems of the member countries (known as
Health Systems in Transition (HiT) profiles), as well as a variety of volumes that discuss health systems and health
policies in the region.

                                                         6


2.      The SHI reforms of Europe and Central Asia: a brief history


        Under communism, health care in the ECA countries was financed out of general

revenues (and out-of-pocket payments) and delivered through a centrally-planned "Semashko"

model consisting of a tiered system of health providers, each allocated budgets according to

population-based norms, with health workers paid by salary.13 In the early 1990s, as most

countries shifted away from communism, several looked to SHI to solve several emerging

problems.



        The most pressing one was the dramatic decline in government revenues as a share of

GDP, caused by a variety of factors, including the growth of the private and informal sectors

where tax compliance was lower, a shrinking of traditional tax bases such as state-owned

enterprises, and pressures for tax cuts from a population experiencing declines in real income.

With falling GDP and revenues falling as a share of GDP, health sectors experienced substantial

cuts in government spending. SHI was seen, rightly or wrongly, as a way of protecting spending

levels in the health sector, the presumption apparently being that earmarking would help ensure

the health sector did not have to compete with other sectors in government spending allocation

decisions, and that earnings in the economy as a whole would fall less than government revenues

and be more stable. Providers were especially enthusiastic about SHI which they saw as a way to

increase their salaries.



        SHI was also perceived as having other advantages vis-à-vis both the financing and

delivery of health care. One was the perceived potential to reduce the grip of finance and health

ministries over the finance and delivery sides of the system, the vision being that payroll tax



13This section draws heavily on Langebrunner et al. (in press).

                                                 7


contributions could flow automatically to a SHI agency that would sit at arms' length from both

the finance and health ministry, and the agency could develop a purchasing capacity and make

government providers more accountable for their performance, through provider payment

reform, selective contracting, and competition between public and private sector providers for

SHI contracts. Autonomization of providers was seen as a logical part of this process, which was

seen as necessary for better performance and greater accountability.



       Of the 28 ECA countries, 14 introduced payroll taxes earmarked for health care at some

stage between 1990 and 2004, and four others had already done so prior to 1990. Early SHI

adopters in the 1990s included Estonia, Hungary, Lithuania, Macedonia, and Slovenia; all

adopted SHI in the period 1990-92. Some countries adopted much later: Bulgaria, for example,

adopted SHI as late as 1999. Often, both the employee and employer are liable, though of course

there may be wide difference between who is legally liable for what and who ends up bearing the

incidence of the payroll tax, the latter depending on conditions in the labor and product markets.

Contributions are mandatory, and in exchange for them the contributing employee is entitled to

receive health services under the terms of the SHI scheme. Groups other than formal sector

workers usually have some coverage. Contributions are required from the self-employed in all

SHI countries, and from pensioners in some. Other groups are financed out of general revenues,

but often the contributions are not specified and insufficient funds are provided in respect of

these groups, who sometimes have inferior de facto coverage.



       SHI does not always raise more than 50% of revenues, though in some countries its

importance has increased over time and has gradually grown to 50% or more. This is clear from

Figure 1, which also shows the timing of the introduction of earmarked payroll taxes in different

countries. In central and eastern Europe, SHI shares of total spending have tended to be higher,

                                                         8


and payroll tax rates have tended to be higher there as a result. In the first group of countries,

payroll tax rates are normally between 10% and 15% of earnings, while in the countries of the

former Soviet Union, they are less than 10%, often considerably so (Langenbrunner, Sheiman

and Kehler in press). It is worth noting that some countries (e.g. Latvia, Lithuania and Poland)

introduced earmarked taxes for health care, but the tax base is income not earnings, so from a

financing perspective these are not "pure" SHI systems.



        Some but not all of the SHI adopters--and, interestingly, some countries that have

earmarked taxes other than payroll taxes--have changed not only the sources of finance but the

way monies flow to providers.14 SHI countries now typically have a SHI agency, but so too do

Poland and Latvia that rely on income taxes or general revenues rather than payroll taxes. These

are typically independent of the ministry of health and have responsibility for administering the

SHI scheme or at least some functions, such as collecting contributions, setting or recommending

contribution rates and ceilings, pooling contributions, etc. Where it exists, the SHI agency pays

providers, but some funds still flow from the health ministry (allocations for capital spending, for

example, but also sometimes other items of spending too). Where there is a SHI agency, it

typically has explicit contracts with providers, though this has not always been the case, and has

been common only in recent years. The contracting, however, is not always selective, although

this too has become more common recently. Often there is no contracting with the private sector,

and where it does occur, it is typically in primary care.



        Most SHI countries have also shifted from budgets as a way of paying hospitals (the

biggest spenders in a health sector) to either fee-for-service (FFS) or a patient-based payment

method (PBP), such as diagnosis-related groups (DRGs). Figure 2 shows the timing of the


14This paragraph relies heavily on information provided in the HiT series.

                                               9


various hospital payment reforms, where we have used the HiTs series to classify a country's

predominant hospital payment method in a given year as falling into one of three categories: (i)

fixed budgets/block grants (the prevailing method under the communist Semashko system), (ii)

fee-for-service/payment by bed days, or (iii) patient-based systems (mainly DRG-based) (cf.

Ellis and Miller 2008). Of the 18 countries that adopted SHI, 12 switched from the use of

budgets, though in three cases the switch occurred with a lag. Some switched to FFS and stuck

with it, while others switched subsequently to a PBP. A few switched immediately to PBP.

Interestingly, some countries without payroll-based contributions also switched from budgets.

The lag in provider payment reform, the fact that different countries opted for different payment

methods and sometimes switched a second time after SHI adoption, and the fact that some non-

SHI countries also switched from budgets during our period all help to create an opportunity to

see how far any impact of SHI adoption is due to the shift to payroll finance and the setting up of

a SHI agency, rather than to provider payment reforms which could have occurred (and in some

cases did occur) even without the adoption of SHI.



3.      Methods


        Let yit be the outcome of interest in country i at time t, Xit be a vector of covariates

thought to potentially influence both the outcome and the SHI adoption decision, and SHIit be a

dummy variable taking on a value of 1 if country i has a SHI health financing system at time t.

An obvious model to estimate is:



(1)     yit =t + Xit + SHIit +i + uit ,

                                                     10


where t is a time-specific intercept, the coefficient  gives the impact of SHI on the outcome yit,

i is a country-specific effect which captures time-invariant unobservables potentially correlated

with SHI status, and uit is an idiosyncratic error term (iid over i and t). In the special case where

the Xit are omitted, eqn (1) collapses to the traditional difference-in-differences (DD) estimator.

Eqn (1) can be estimated as a fixed effects model, or in first differences. In the latter case, the

estimating equation can be expressed as


(2)     yit =t +Xit +SHIit +uit,


which can be consistently estimated by pooled OLS.



        Care needs to be taken to get accurate standard errors in this type of analysis. Bertrand et

al. (2004) have shown that many outcome variables used in published policy impact analyses

generate positive serial correlation in the uit. If ignored, and the model is estimated as a fixed-

effects specification, this positive serial correlation results in standard errors that are too small,

and t-statistics that are too large--possibly dramatically so. In such a case, first differences may

be preferred. Of course, if the uit in eqn (1) are serially uncorrelated, the error term in the first-

differenced version may well be subject to negative serial correlation, in which case the standard

errors would be overestimated. An obvious strategy is to report standard errors that are robust to

any type of serial correlation (and heteroskedasticity), whether one uses fixed effects or first

differences. This is what we do below in all our models. The Monte Carlo results reported by

Bertrand et al. (2004) suggest that with a sample of 28 countries the rate of rejection of the null

hypothesis of no impact ought to be close to the right one.15




15We also experimented with (block) bootstrapped standard errors, and obtained broadly similar results.

                                                         11


        Our basic framework of individual-specific effects represented by eqn (1) captures the

possible endogeneity of SHIit to the extent that the unobservables that are correlated with both

SHIit and yit are time-invariant. However, endogeneity of SHI adoption will still be a problem in

our basic framework if there are time-varying unobservables correlated with both SHIit and yit. In

this sense, an additional source of country heterogeneity can be introduced in a more general

specification, through a random trend model (cf. e.g. Wooldridge 2002):



(3)      yit =t + Xit + SHIit +i + git + uit ,


where gi is the country-specific trend of yit, and i and gi are allowed to be arbitrarily correlated

with the regressors. Eqn (3) thus allows for both time-invariant and time-varying unobservables

that may be correlated with yit and SHIit, albeit in a way that assumes the unobservable grows

over time in a linear fashion. One way of estimating this model is differencing eqn (3) to get



(4)      yit =t +Xit +SHIit + gi +uit,


and using a fixed effects estimator on this differenced equation.16 We can test eqn (3) against eqn

(1) by testing the joint significance of the gi. Friedberg (1998) uses such a model in her analysis

of divorce laws, and finds that allowing for state-specific trends is crucial to unearthing the

impacts of these laws.



        An alternative way to relaxing the assumption that SHI is endogenous only insofar as it is

correlated with the i and gi is to find instruments for the SHI adoption decision. Suppose we

have a set of instruments Zit and our basic model in levels:

16Alternatively we could use the first differences estimator once again, this time applied to eqn (4) so as to eliminate
gi, and estimate the resulting model by pooled OLS. However, this procedure would mean losing an additional
period of time for estimation purposes, which is why we have opted for the fixed effects estimator in the case of the
random trend model.

                                                          12


(5)      yit =t + Xit +SHIit + eit



Under the assumption that the instruments are weakly exogenous or predetermined, i.e.

E[Ziseit ]= 0 , s<t, t=1,...T, with T representing the number of time periods in the data, it is

possible to use lags of the potentially endogenous SHI variable as instruments, in addition to

"traditional" instruments obtained outside the model (Anderson and Hsiao 1981). If these

instruments are valid (i.e. exogenous and strong in the sense of being highly correlated with our

instrumented SHI dummy--assumptions which can be tested), they should control for any kind of

endogeneity including that arising from country-specific effects and trends. Eqn (5) can then be

consistently estimated by two-stage least squares (IV-2SLS) or using the more efficient two-step

generalized method-of-moments (IV-GMM) estimator (cf. e.g. Cameron and Trivedi 2005).



         We implement both IV approaches in the cases of those outcomes for which the potential

endogeneity of SHI adoption does not seem adequately controlled for by our DD-based

specifications (on how we determine this, see below), using as instruments for SHIit in eqn (5)

the first lag of the SHI dummy (SHIi,t-1) and an indicator for whether the country in question had

a SHI system prior to the communist takeover in the mid-late 1940s. Although under weak

exogeneity we could theoretically use more lags of the SHI variable as instruments, we include

only its first lag in the instrument set due to the lack of variation over time in our SHI dummy17,

leading to the redundancy of additional lags and loss of degrees of freedom for overidentifying

restrictions tests. The rationale for using the pre-communist presence of SHI arrangements as an

instrument is that the tradition of health systems based on the Bismarck model in some ECA

countries prior to communism may have increased the likelihood of SHI re-adoption in these

17As it can be seen in Figure 3, transitions between tax-funded health systems and SHI arrangements occur only
once in all but one of the countries in our sample, the only exception being Kazakhstan. The exact definition of what
constitutes a SHI system for our purposes is given later.

                                                      13


countries after the transition to market economies, but that pre-communist characteristics could

reasonably be thought to be uncorrelated with our health and labor outcomes (measured about 45

years later) except by affecting the probability of SHI adoption. The relevance of our two

instruments is assessed through a battery of tests, and Hansen tests are used to check the

exogeneity of our instruments in the estimated models. Given the relatively small number of

observations used to estimate some models, IV procedures (mainly GMM) potentially introduce

small sample bias in the estimations; this adds to the--often sizeable--increase in standard

errors expected when using IV methods relative to those obtained by OLS procedures. We

therefore draw our basic empirical conclusions from the results of the DD-based models, and

rely on the IV procedures mainly when SHI endogeneity seems to pose a problem even after

controlling for country-specific effects and trends.



4.       Data


         We use annual data on SHI status, health and labor market outcomes for the 28 ECA

countries, from 1990 to 2004. Our dataset has been constructed using a variety of sources; the

description in this section begins with our independent variable of interest, SHI status, and then

continues for the variables included in our health and labor models, respectively. In the health

sector models, data are generally available for most country-year combinations; fewer

observations are available for the labor market models.18




18In the case of Bosnia-Herzegovina, the period between 1992 and 1996 has been excluded from the analysis due to
the lack of data for many dependent variables and the complete disorganization of the health system--which
obviously included the SHI scheme--during the war period.

                                                         14


Social health insurance status

         We define our SHI dummy SHIit as taking a value of one if in country i at time t

earmarked payroll taxes for health care were collected from formal-sector workers and there was

a SHI agency in place. The required information was obtained mainly from the Health Systems in

Transition (HiT) document series published by the European Observatory on Health Systems and

Policies, a partnership between the European Office of the World Health Organization (WHO)

and governmental, national and international agencies. World Bank reports and consultations

with its staff working in the ECA region were also used in order to obtain data on countries for

which HiTs have not yet been published, and to double-check the information assembled through

the HiTs.19



         Our derivation of the SHI status indicator is shown in Table 1. Our strict definition means

that we end up classifying as non-SHI some country-year combinations that are often--we

believe, erroneously--classified as SHI (such as Latvia and Poland). Furthermore, we classify

Romania as SHI only after 1998; despite the fact that payroll taxes were used somewhat before

then, it was not until 1998 that SHI was fully set up with a SHI agency and with payroll

contributions making up the majority of health care revenues. We explore the sensitivity of our

results to not classifying these as SHI countries, by re-running our models with the three of them

classified as SHI for the years indicated in Table 1.




19Our classification is consistent with the evidence presented in Langenbrunner et al. (in press) which came to our
attention after the empirical work for the paper hade been completed.

                                                         15


Figure 3 depicts the pattern of SHI adoption over time in the countries of our sample. Our SHI

status dummy is equal to 1 in about half (218 observations) of the 442 country-year

combinations for which we have non-missing values of the indicator.


Health sector outcome variables

       Our health sector outcome measures include: per capita health spending (total, public and

private) and the share of spending going on salaries; population health status; hospital activity

rates and capacity utilization; and quality-of-care indicators. Our variable definitions and sources

are briefly described below and the descriptive statistics for them--for the full sample and

disaggregated by SHI status--are presented in Table 2.20



       We measure per capita health spending as total health care expenditures per capita

expressed in constant 2000 dollars adjusted for purchasing power parity (deflated using the

United States GDP deflator), to allow comparisons in real values between countries and over

time (Gerdtham and Jonsson 1992). The source for these figures is the World Bank's World

Development Indicators (WB-WDI) database. The WB-WDI database is the primary World Bank

database for development data, obtained from recognized international sources. It contains an

expanded set of the economic, health and other time series indicators published in the Bank's

World Development Reports. Average health spending for the period 1990-2004 was US$403

PPP per capita. The Czech Republic, Hungary and Slovenia are the countries with highest

spending levels (each with an average of at least US$857 PPP per capita between 1990-2004 and

at least US$1,225 PPP per capita in the last year), whereas Azerbaijan, Kyrgyz Republic,

Tajikistan and Uzbekistan have the lowest spending levels in our sample (at most US$132 PPP



20The complete list of definitions and sources for our health sector outcome variables can be found in Table 9 in the
Appendix.

                                                 16


per capita on average for 1990-2004 and US$163 PPP per capita in the last year). On average,

government health spending accounted for almost 70% of a country's total health spending over

the period of analysis and 60% in the year 2003; Armenia and Tajikistan exhibited the smallest

shares of government health spending in 2003--less than 21%--whilst in Czech Republic and

Slovak Republic that share was higher than 88% in the same year.



        We also include among our indicators health sector salaries as a percent of total health

spending. Data on this--like the data on many of the remaining health sector outcome

indicators--are taken from the World Health Organization's Health for All Database (WHO-

HFA). This database, maintained by the European Office (Copenhagen) of the WHO, contains

data for all European countries plus the former USSR republics in Central Asia on about 600

health indicators, including annual information on morbidity and disability; hospital discharges;

and health care resources, utilization and expenditure. The original sources of information are

mainly own WHO estimates, country statistical offices and other international organizations.

Given the general scarcity of available health information for developing countries, the

availability of this database is good news for researchers dealing with health-related issues

specifically in the former communist countries of Europe and Central Asia. In our attempt of

getting a comprehensive, general picture of the (potential) SHI impact on population health

conditions, we include dependent variables related to life expectancy, group-specific mortality

rates, disease-specific standardized death rates and incidence rates and measures of utilization of

services such as caesarean sections and immunization. We used the same database for obtaining

data on hospital indicators, which include measures of average length of stay, bed occupancy,

number of hospital beds (from the WB-WDI database), admissions and disease-specific

discharges. We also include in our analysis a few indicators of avoidable deaths--such as

                                                  17


standardized death rates for appendicitis and hernia and intestinal obstruction--as proxies for the

average quality of hospital care. Finally, alternative infant mortality and under-five mortality

rates were obtained from WB-WDI and the TransMONEE 2006 Database, a UNICEF IRC

(Florence) database which contains data for ECA countries except Turkey on 146 economic and

social indicators divided into ten different topics and ranging from 1989 to 2004.



        Simple comparisons of the average outcomes presented in Table 2 indicate that SHI

countries tend to spend more in health care, both in the public and private sectors, and a higher

fraction of the government health spending seems to be absorbed by salaries. On the other hand,

there is some indication that mortality and disease incidence rates are generally lower in SHI

countries, whilst no clear pattern emerges for immunization rates. As far as hospital indicators

are concerned, total length of stay, in-patient admissions and beds tend all to be lower in SHI

countries; most of our diagnosis-specific hospital discharges indicators are higher for SHI

countries, and there is no clear pattern concerning our quality-of-care proxy measures. Visual

comparisons of the evolution of SHI adoption in our sample vis-à-vis two health outcomes,

average total health expenditures per capita and WHO's average infant mortality rate (Figure 4

and Figure 5), show somewhat clear patterns: average health spending slightly decreased during

the first period of growing SHI adoption by ECA countries (1990-93) but experienced a

sustained increase during and after the second period of SHI growth (1995-98, when SHI

prevalence reached more than 50%), while the average infant mortality rate tended to remain

stable around 22 per thousand births during the first period but continuously decreased during

and after the second period, when SHI prevalence reached half of the countries. Determining

whether the differences and patterns described above are due to SHI adoption (that is, a causal

effect) or whether they merely reflect pre-existing differences--observable and/or

                                                         18


unobservable--between countries that eventually adopted SHI and those that did not (a selection

effect) is the main task of our empirical work.21


Labor market outcome indicators

         Our labor models are estimated with gross wages, unemployment, employment and

informality measures as dependent variables. We briefly describe their definitions and sources

below, and their descriptive statistics are presented in Table 3.22



         Total annual gross wages and salaries in local currency units were drawn from the WB-

WDI ECA regional database and transformed into yearly constant PPP averages for the

employed population aged 15-59 using purchasing power parity conversion factors and the US

deflator, both available at the WB-WDI database as well. We use annual data on total

unemployment and registered unemployment rates obtained from the Key Indicators of the Labor

Market (KILM) database, published by the International Labor Organization (ILO). The same

database is used for gathering data on agricultural employment and self-employment as shares of

total employment; these work categories serve as proxies for informal employment, as they have

been argued to be closely associated with last resort, low-productivity jobs in the region, thus

masking a situation of worsening underemployment and predominant subsistence farming

(Svejnar 1999; Alam et al. 2005). As yet another employment measure, we use two employment-

to-population ratios: the first is drawn from the ILO-KILM database, and employment-to-

population ratio for individuals aged 15-59 was obtained from the TransMONEE 2006 Database.




21Moreover, the comparisons of descriptive statistics between SHI and non-SHI countries presented in Table 2
cannot be strictly interpreted as a preliminary assessment of SHI effects, because eventual SHI adoptions occurred
on a staggered basis in our sample.
22Table 10 in the Appendix contains the full definitions and sources for all our labor outcome variables.

                                                   19


        Data on informal employment or the size of the informal economy are not readily

available for transition countries. Although there are rough estimates provided by ILO and some

national statistical offices, their reliability and comparability across countries is not guaranteed

and they are available for only a few countries in our sample. However, there have been some

attempts in the literature to estimate at least the size of the informal economy using indirect

approaches, focusing on various macroeconomic and institutional indicators which are argued to

be linked with the evolution of the informal sector. We follow the approach proposed by

Kaufmann and Kaliberda (1996) and Johnson et al. (1997) and measure the size of the informal

economy by comparing official measures of annual GDP growth with total electricity

consumption growth. The basic assumption is that electricity consumption is closely related to

overall economic activity, with a short-run electricity-GDP usually close to one; hence, the

annual growth in electricity consumption represents a good proxy for the total (formal plus

informal) GDP growth in a given year. If this is the case, rough estimates of the informal GDP

can be obtained simply by calculating the difference between the estimated total GDP and the

official GDP measure. This simple calculation method avoids important the endogeneity

problems likely to be found in estimates coming from modeling approaches which derive the size

of the informal economy from economic and institutional variables.



        The method is not free from criticisms, of course. The assumption of unitary electricity-

GDP elasticity has been criticized mainly on the grounds that (i) technical progress over time has

made electricity use more efficient than in the past, (ii) higher electricity prices reduce electricity

consumption per unit of output, and (iii) many informal activities are not electricity-intensive,

such as in the services sector. For these reasons, we also follow Kaufmann and Kaliberda and

Johnson et al. and depart from the unitary elasticity assumption, considering three different

                                                20


groups of transition countries identified in previous related research: "energy-efficient"

economies (Central and Eastern European countries) which are assumed to have an electricity-

GDP elasticity of 0.9 with a growing economy; "energy-neutral economies" (the Baltic

countries), assumed to have unitary elasticity; and "energy-inefficient" economies (the rest of the

former Soviet republics), assumed to have an elasticity of 1.15 when GDP is growing. We use

data on electricity output and official GDP for 17 ECA countries in order to obtain annual

estimates of the size of the informal economy for the period 1990-2003. As baseline data, we use

the same values as Johnson et al. (1997) for the initial (i.e. 1989) shares of the informal sector in

those countries.



        In our sample, average gross wages tended to be higher in SHI countries than in non-SHI

countries between 1990 and 2004; Armenia and Tajikistan exhibited the lowest averages in 2004

(less than US$150 PPP) while Croatia and Lithuania presented the highest averages in the same

year (more than US$2,500 PPP). Average total unemployment rates seem to have been very

similar for SHI and non-SHI countries during 1990-2004, with Hungary and Slovenia presenting

rates around 6% in 2004, and Macedonia and Poland reaching at least 19% in the same year.

Registered unemployment rates and employment-to-population ratios for individuals aged 15-59

indicate a somewhat worse employment situation for SHI adopters in such period. Self-

employment and agricultural employment were lower on average for SHI countries; and both

groups of countries presented very similar estimates for the average size of the informal

economy: the Czech Republic and Slovakia exhibited averages lower than 16% in 2003, in

contrast to averages higher than 41% for Azerbaijan and Georgia in that year. For two selected

labor variables, unemployment rate and size of the informal economy, there seems to be a close,

positive relationship between the rates of SHI adoption over time and their cross-country yearly

                                                        21


average values in the sample, as depicted in Figure 6 and Figure 7. As in the health sector case,

however, one needs to control for the observable and unobservable heterogeneity between non-

adopters and eventual SHI adopters in order to attribute the differences and patterns identified

above to SHI adoption.


Covariates in the estimating equation (the X-vector)

        We are not attempting to estimate a complete model of our health sector and labor market

outcomes, but rather to estimate the impact of SHI adoption. The criterion for including a

variable in our Xit vector is whether its omission would bias our estimate of , our SHI impact

parameter. We want to include in Xit therefore variables that are correlated with both our

outcomes and SHI adoption.



        Although evidence on the determinants of SHI adoption is scarce, it has been indicated

that SHI schemes emerged first in countries with higher initial (i.e. pre-transition) per capita

income levels, whilst tax-based funding prevailed in countries with lower initial per capita

income (Preker, Jakab and Schneider 2002). This positive correlation between income levels and

SHI status is also present in our data; thus, we include GDP per capita in our Xit vector23. This is

the only variable we include among the Xit in the labor market model; our specification is thus

similar to that of Gruber and Hanratty (1995) in their study of the employment and wage effects

of the introduction of tax-financed health insurance in Canada. In our main health sector

equations, we also include among the Xit the share of the population aged 65 or above, and the

urban population as a fraction of the total. We subsequently present additional results where we

include among the Xit dummies capturing the provider payment methods in force.



23The precise definition and source are indicated, along with those for other variables included in the Xit vector, in
Table 11 in the Appendix.

                                                22


Preliminary data analysis: the `parallel trends' assumption

        For all the generalizations of the DD approach discussed in the methods section, we will

obtain unbiased estimates of SHI impacts only insofar the crucial "parallel trends" identifying

assumption holds in our data, that is, only if the change in outcomes in non-SHI "control"

countries constitutes an adequate representation of what the change in outcomes in SHI

"treatment" countries would have been in the absence of SHI adoption. We obtain evidence that

the (directly untestable) "parallel trends" assumption cannot be discarded to hold in our context

through two alternative procedures. Firstly, we performed t-tests for both (i) baseline differences

in average outcomes--year 1990--between eventual SHI countries and non-adopters (excluding

from the sample countries which adopted SHI in 1990 or before) and (ii) baseline differences in

average changes in outcomes (1990-1991) between the same groups of countries (excluding

countries which adopted SHI in 1991 or before). The first test shows statistically significant

differences between baseline outcomes only in a few cases, and trend differences (1990-91)

occurred in even fewer cases--only four instances, namely for overall and male life expectancy,

and death rates by all causes and female breast cancer.



        Secondly, we ran "placebo" DD as a checking strategy: we created a false treatment

dummy variable which took on the value of one for eventual SHI adopters in the sample and zero

for countries that never adopted SHI, and used only 1990-1991 data (excluding countries that

adopted SHI in 1991 or before) for performing a DD regression comparing years 1991 and 1990.

Since there was in fact no "treatment" (SHI adoption) in either year, rejecting the null hypothesis

of individual insignificance of the false treatment variable for changes in a given outcome--

conditional on covariates--casts doubt on the unbiasedness of estimates obtained by DD

methods for that variable. These "placebo" regressions showed significant effects for only 9

                                                         23


health outcomes and 1 labor outcome.              24  Taken together, the results obtained by the two

alternative exploratory procedures offer support to the assumption of "parallel trends" between

SHI and non-SHI countries for the vast majority of health and labor outcomes; however, these

tests are of indicative nature only, due to the small sample sizes used as a result of restricting the

analysis to years 1990 and 1991. An additional test for endogeneity in the context of our models

based on DD generalizations is performed in the next section.



5.      Results: health sector outcomes


        We first report the results for the impact of SHI on health sector outcomes. For both our

health and labor analyses, we first jointly present the estimation results for the SHI dummy

models with country-specific effects only (eqn (2) estimated by OLS) and under the random

trend specification, i.e. with both country-specific effects and trends (eqn (4) estimated by fixed

effects). Next, we test for the possible endogeneity of the SHI status variable in the context of

our more general DD-based specification--the random trend model--for the case of each

outcome; for those cases where SHI appears to be endogenous even after country-specific effects

and trends are taken into account, we then present the results obtained by implementing our IV

strategy on the model in levels given by eqn (5).


Results for the models based on generalizations of the DD approach

        Table 4 presents the results for both the SHI dummy specification with country-specific

effects only and for the random trend specification. For only 12 of the 71 outcome indicators



24The false treatment dummy had a significant positive effect (at a 10% level) on neonatal mortality rate and
measles incidence, and negative effects on the share of government health spending absorbed by salaries,
postneonatal mortality rate, death rates by diarrhoea, liver diseases and smoking-related causes, syphilis incidence
and mumps immunization rate. In the labor models, a 10% significant effect (negative) was found only for
TransMONEE's employment-to-population ratio.

                                                          24


included in Table 4 is there any evidence of a significant impact of SHI. This itself is

noteworthy--the impacts of SHI are limited to just a few health sector outcome variables.



         Overall, the results obtained by using our random trend model are similar to those

obtained through the specification with country-specific effects only, both in terms of the sizes

and significance levels of individual coefficients. However, country-specific trends seem to be

important in explaining the evolution of health outcomes. We estimated our random trend

specification using a least-squares dummy variables approach and tested the joint insignificance

of the gi terms in eqn (4) through F-tests; according to the results (not shown), the joint

insignificance of the gi terms is strongly rejected in all models, with p-values smaller than

0.0001. For this reason, we consider the more general random trend specification to be preferred

to our simplest model without country-specific trends, even though the results coming from both

models lead generally to the same conclusions.25



         Table 4 suggests that SHI has significantly raised per capita health spending, by about

12% or US$47 PPP in 2000 prices. The increase in government health expenditures per capita--

of about 15% or US$45 PPP--seems to have accounted for almost the entire raise in total

spending during the period, with no significant effects found on private health expenditures. The

impacts on total and government health spending are reduced to about 8% and 12%

(respectively) when we switch to the alternative "non-classical" definition of SHI for Latvia,

Poland and Romania, where we classify the first two countries as SHI countries despite the fact

they do not meet the strict definition of SHI, and Romania as SHI from 1992 onwards even

though it was not until 1998 that Romania set up a formal mandatory SHI system. This reduction


25Since the random trend specification is estimated using a fixed effects estimator, which gives consistent parameter
estimates even if the regressors are correlated with the individual effects, using this model allows us also to control
for the potential correlation between SHI adoption and country-specific (linear) trends in health outcomes.

                                                         25


of impact when the definition is changed provides additional evidence that SHI--interpreted

strictly--does indeed increase health spending.26



        Whether the additional spending resulting from a transition to SHI is a good or bad thing

cannot be said without seeing what the extra resources buy. A second result that emerges from

Table 4 is that SHI has significantly increased the share of health spending going on wages and

salaries, with the mean impact being of the order of around 16%, equivalent to a mean increase

in the share of spending going on wages and salaries of around 6 percentage points. We must

interpret the latter result with caution, however, due to the reduced number of changes in our SHI

dummy used to identify the parameters of the random trend-fixed effects specification in this

case.



        The transition to SHI has not been associated with statistically significant improvements

in health outcomes. The one exception is the 8% reduction in post-neonatal mortality, although

there is a contradictory positive and significant impact of about 9% in one of the infant mortality

rate measures. Nevertheless, according to the estimates for the vast majority of our indicators,

SHI adoption does not seem to have caused either general improvements or adverse effects on

population health status.



        Our results also point to SHI impacting on hospital activity rates, capacity utilization and

the quality of care. Table 4 suggests that SHI reduced average total length of stay and increased

the bed occupancy rate. However, the impacts are small in magnitude, between 2-3% over the

sample average. We also find significant increases on in-patient admissions both overall and in



26Similarly, for all the remaining models estimated in this paper, using the alternative definition of the SHI variable
does not alter our qualitative results and only marginally affects the size of our parameter estimates in some
instances.

                                                26


acute care hospitals of about 2% and 4%, respectively. There is evidence of a significant positive

impact on discharges for patients treated for cerebrovascular diseases (an increase of around

4%), but only weak evidence of an impact on discharges by infectious diseases (11% increase in

the model without random trend). Finally, we find evidence of a sharp reduction in the surgical

infection rates due to SHI adoption, but again the reduced number of switches of the SHI dummy

in the sub-sample used to estimate this model suggests caution in interpreting such result as a

strong indication of improved hospital care quality under SHI systems.


Endogeneity of SHI vis-à-vis health sector outcomes

       Our estimation method allows for time-invariant unobservables and time-varying

unobservables (albeit ones that follow a linear trend) to be correlated with both our health sector

outcomes and SHI adoption. It might, however, be argued that this does not properly capture the

possible endogeneity of SHI. It might be that SHI was adopted precisely in response to changes

in one or other of our health sector outcomes--falling health spending or rising mortality, for

example. An informal yet very informative test of reverse causality proposed by Gruber and

Hanratty (1995) in a similar modeling exercise is to include in the model a lead dummy variable

indicating whether SHI will be adopted the following year. If causality goes from SHI to the

outcome variable, the coefficient on the lead dummy will be zero. A nonzero coefficient would

point towards causality running the other way or some other type of endogeneity that cannot be

captured by eqn (3).



       We adopted this approach in the SHI dummy random trend model for all the health

outcomes. The coefficient on the lead dummy turned out to be significant at the 10% level in

only two of the 71 models, namely those for measles immunization rate and length of stay in

acute care hospitals, and none was significant at the 1% level. For most outcomes, the

                                                          27


insignificance of the lead SHI dummy could not be rejected with fairly high associated p-values.

The inclusion of country-specific effects and trends seems therefore to control for the potential

reverse causality between health outcomes and SHI adoption in our data.


Results for the instrumental variables (IV) approach

        Since our earlier estimated impacts do not seem to be attributable to endogeneity, and

thus we can be more confident in that we are estimating causal effects in our random trend

specifications, we present in Table 5 the results of our IV estimations only for the two outcomes

for which the SHI dummy appears to be endogenous in the random trend specifications--

measles immunization rate and length of stay in acute care hospitals. The model in levels given

by eqn (5) is estimated by two-stage least squares (2SLS) and by the two-step generalized

method-of-moments (GMM), in order to achieve a balance between the results less likely to be

affected by small sample bias, as it is the case for the 2SLS estimates, and the results obtained by

the more efficient GMM method. According to the results in Table 5, our two instruments--the

SHI dummy lagged one period and our indicator for SHI existence prior to communism--have

very good explanatory power for the values assumed by SHIit, with the joint insignificance of the

instruments set in the first stage regressions being strongly rejected by F tests. However, it is

only for the measles immunization rate model that Hansen's tests of overidentifying restrictions

cannot reject the joint null hypothesis that our two instruments are uncorrelated with the error

term and that such instruments are correctly excluded from the estimated equation.27 Thus, based

on our IV estimates, SHI adoption did not have any effect on this remaining immunization




27For length of stay in acute care hospitals, replacing our indicator of SHI prior to communism by the second lag of
the SHI dummy in the instruments set (thus using SHIi,t-1 and SHIi,t-2 as instruments, so we have two instruments
again and can implement Hansen's test) does not lead to non-rejection of the null hypothesis in a new test for
instruments' exogeneity. This suggests that, for this outcome, it is the use of lagged values of the SHI dummy as
instruments based on the weak exogeneity assumption that may not be valid after all.

                                                           28


measure either, and no reliable conclusions can be drawn as far as our measure of length of stay

in acute care hospitals is concerned.28


Investigating the robustness of our basic results: SHI adoption and
embedded institutional reforms

         As noted in section 2, SHI adoption is often (though not always) associated with a change

in the way hospitals are paid, from budgets to either FFS or PBP. From an empirical point of

view, it might be argued that our estimates of the impacts of SHI adoption are simply picking up

the effects of provider payment reforms rather than the impact of SHI adoption per se. For

example, our positive impact on health spending might be argued to derive from the fact that

many SHI adopters switched from budgets to FFS, often sticking with FFS even now. We

investigate the possibility our results reflect concurrent shifts in provider payment methods,

focusing on the timing of changes in hospital payment methods.29 To do so, we modify our

preferred random trend specification--eqn (3)--by adding to the Xit vector dummies for FFS and

PBP ("fixed budgets/block grants" is the reference category).30


28We estimated the IV models for all our outcomes as a way of assessing the robustness of the results obtained
through our preferred random trend specification. In broad terms, our two instruments have very good explanatory
power concerning the SHI variable and the null of exogeneity cannot be rejected--according to Hansen's tests--for
the vast majority of health outcomes. The IV results (available from the authors upon request) provide general
support to the main conclusions derived from the random trend models; for instance, positive and significant effects
are found on total health expenditures per capita (14% or US$57 PPP) and no effects are generally found on
mortality, incidence of diseases or immunization measures. However, as expected, fewer significant results are
found in the IV specifications than in the random trend models due to the substantial increase in standard errors even
when GMM is used (random trend standard errors are typically multiplied by a factor of four when IV is performed).
Taking this IV feature into account, point estimates suggest that SHI may have had an even larger positive effect on
the share of wages and salaries than previously estimated--of around 22%, with an associated p-value of 0.117. The
results are inconclusive concerning length of stay and hospital admissions measures due to the rejection of the null
hypothesis in Hansen's tests for the validity of the instruments set.
29We would like to have expanded the scope of this part of the analysis to cover other potentially relevant changes
that may have been associated with SHI adoption, such as changes to the way primary care providers were paid, the
introduction of a gate-keeping function for primary care providers, and so on. We were unable, however, to get the
relevant data, year by year. At best, we could obtain typically only snapshots of the initial (i.e. communist) and
current arrangements, with no information on the timing of these changes over the decade.
30Table 12 in the Appendix presents the detailed timing of changes in predominant hospital payment methods for
the countries in our sample. As it can be noticed, SHI adoption and a change in the predominant hospital payment
method have occurred in the same year for some countries, e.g. Bulgaria, Czech Republic, Estonia, Kyrgyz Republic
and Macedonia.

                                                 29


       Overall, the new results presented in Table 6 offer support to our previous finding that

SHI adoption increased per capita health spending. Even after including the payment methods

dummies, we find that SHI led to an increase in annual government health care expenditures per

capita of around 11% or US$33 PPP in the "adopter" countries, compared to what would have

occurred had these countries not switched to SHI. This estimated effect is only 4 percentage

points smaller than that obtained from our original random trend specification (i.e. without the

payment dummies) and is significant at the 5% level. Interestingly, while the coefficients on FFS

and PBP are the expected sign, neither is significant. The results suggest, in other words, that

SHI adoption of itself increases government health spending, while switching from budgets to

either FFS or PBP does not of itself change government health spending per capita. Interestingly,

our results suggest that private spending on health is sensitive to the way hospitals are paid, with

FFS being associated with significantly higher private spending. The magnitude is sizeable: an

increase of 34% or US$36 PPP per capita.



       As far as health outcomes are concerned, there was little evidence in our original

specification of SHI adoption having any impact, and this remains the case even after including

provider payment method dummies. The latter have a few significant coefficients but there is no

consistent pattern. As far as hospital indicators are concerned, the results in Table 6 confirm that

quality of care as measured by surgical infection rates has markedly improved as a result of SHI

adoption. The new estimated increase in the bed occupancy rate due to SHI is similar to the

original estimate (3.5%). By contrast, the impact of SHI on acute care hospital admissions is

somewhat smaller when the provider payment method dummies are included among the Xit

(2.5% compared to 3.7%). This reduction is due to FFS having an independent positive impact

on admissions, and the introduction of FFS being correlated with the adoption of SHI. In our new

                                                            30


results, SHI no longer has an impact on length of stay, though neither does the way hospitals are

paid.



         Overall, our basic empirical conclusions seem fairly robust to the inclusion of changes in

hospital payment methods as potential confounders of SHI impacts.31 In particular, SHI adoption

per se--i.e., without any change in payment methods--is still found to lead to higher

government health spending, bed occupancy rates and (acute care) hospital admissions, but no

improvements on population health indicators.32



6.       Results: labor market outcomes


         In this section we report our results for the impact of SHI on labor market outcomes. The

format follows that of the previous section.


Results for the models based on generalizations of the DD approach

         Among the eight outcomes studied, in only one case do we see a significant impact,

namely for the average gross wage rate (Table 7).33 The coefficient is positive and its point

estimate is remarkably similar in the models with and without random trend; it implies an



31We implemented the "lead SHI dummy" test for reverse causality in the context of this new specification as well;
the lead SHI dummy is found to be significant at the 10% level in only four instances--measles and mumps
immunization rates, hospital discharges by respiratory diseases and length of stay in acute care hospitals. For the
latter outcome, our IV models (results not shown) lead to inconclusive results due to the rejection of the null
hypothesis of instruments' exogeneity in Hansen's tests.
32The estimated coefficients on hospital payment methods are of course very interesting by themselves, not least
because in practice it has been mostly SHI countries that have abandoned budgets as the predominant arrangement
in our sample; however, since an analysis of the influence of provider payment methods on health indicators is not
the primary aim of this paper, we only briefly comment on these results here. As it can be seen in Table 6, in
addition to the substantial, positive effect of fee-for-service on total and private health care spending per capita, this
arrangement is also found to decrease the length of stay in acute care hospitals and increase the number of inpatient
admissions and hospital discharges, suggesting that providers are given the incentives for using resources more
intensely under such payment method. Also noteworthy is the finding that neither fee-for-service nor SHI per se
induce hospital downsizing in terms of the normally excessive beds supply inherited from the communist system,
though switching from budgets to patient-based payment methods seems to provide some incentive for it.
33The models for labor market outcomes are estimated using the natural logarithm of the dependent variable.

                                                      31


increase of around 20% in the average gross wages and salaries after SHI adoption in a given

country. We find no evidence in Table 7 of SHI significantly increasing unemployment or

reducing employment; nor do we find any effects of SHI adoption on either the informal

economy measure or our informal employment proxies (self-employment and agricultural

employment as shares of the total).


Endogeneity of SHI vis-à-vis labor market outcomes

         We undertook the same type of informal test for endogeneity discussed above for the

health sector outcome variables. The coefficient on the lead SHI dummy was significant at the

5% level for two of the outcomes: registered unemployment rate and TransMONEE's

employment-to-population ratio. Moreover, the coefficient for the lead SHI dummy was

significant at the 10% level for our measure of the size of the informal economy. These results

suggest that the statistical insignificance of at least some of the SHI impacts estimated in our

previous models might be due to endogeneity. Therefore, in contrast to what we found for the

health outcomes, endogeneity seems to be a reason for concern in our labor models, and we

tackle this issue by implementing our IV methods in what follows.


Results for the instrumental variables (IV) approach

         The results for the registered unemployment, TransMONEE's employment-to-population

ratio and informal economy models estimated by 2SLS and GMM are reported in Table 8.34 As

for the health outcomes, the two instruments used seem to be extremely relevant for all labor

models, with partial F statistics above 250. Additional tests (not shown) confirm that our

instruments are relevant and strong in the context of these labor models: for the three cases,



34As in the random trend specification, the IV models are estimated using the natural logarithm of the dependent
variable at the left-hand side.

                                                  32


Anderson canonical correlations likelihood ratio tests firmly reject the null of model

underidentification--with associated p-values smaller than 0.001--and Cragg-Donald F-tests

(Stock and Yogo 2002) strongly reject the null of weak instruments (F statistics above 200).

Furthermore, according to the results of the Hansen tests for overidentifying restrictions, our

excluded instruments are found to be exogenous for the models with the three outcomes as

dependent variables (p-values larger than 0.47). Thus, for the three labor models where we found

evidence of endogeneity in our random trend specification, it seems safe to base our conclusions

entirely on the IV results, since there is evidence that the latter models provide reliable parameter

estimates for such cases.



        Again as expected, GMM estimates are more efficient than the 2SLS ones, but even

GMM standard errors turn out to be four times larger on average than the random trend

estimates. Despite such large inflation in standard errors, the IV estimates provide evidence that

the lack of significant SHI impacts found in the previous models on two employment indicators,

the registered unemployment rate and TransMONEE's employment-to-population ratio, was due

to endogeneity. Positive and strongly significant impacts on registered unemployment rates are

reported in both the 2SLS and GMM columns of Table 8, suggesting that SHI adoption doubles

this unemployment measure. Although no effect was found on the total unemployment rate in the

random trend models, 2SLS estimations show some evidence that SHI adoption has led to a

deterioration of employment levels as measured by TransMONEE's employment-to-population

ratio: the point estimate is significant at the 10% level and implies a decrease of around 10% or

6.7 percentage points in this indicator in response to SHI adoption. A similar point estimate, very

close to the 10% level of statistical significance, is found in the GMM estimations. Finally, there

is once more no evidence of a SHI impact on the size of the informal economy.

                                               33


7.      Discussion and conclusions


        The health system reforms the European and Central Asian (ECA) countries implemented

during their transition from socialist economies in the 1990s provide a unique opportunity to

assess the impacts of social health insurance (SHI) on the health and labor sectors. We took

advantage of this highly unusual "experiment" in which many ECA countries unequivocally

switched from general tax-funded to SHI systems in a relatively short period of time, and on a

staggered basis, so as to shed light on two broad sets of currently unanswered questions: firstly,

how does SHI affect national health spending, the way such resources are spent, and population

health outcomes? Secondly, how does SHI impact national employment levels and

informalization of the economy? In order to obtain empirical evidence on these issues, we have

used regression-based generalizations of the differences-in-differences approach and

instrumental variables (IV) methods on panel data from 28 ECA countries for the period 1990-

2004.



        Overall, there is reliable evidence that our empirical methods have allowed us to

successfully control for the possible endogeneity of SHI adoption, and therefore that we have

identified causal relationships between SHI adoption and outcomes. This supporting evidence

comes from a variety of sources. In the first place, our random trend specifications control for

any time-invariant country-specific unobservables that may be simultaneously correlated with

SHI adoption and outcomes, and also control for country-specific linear trends in these

unobservables over time. Second, diagnostic checks show that we cannot rule out the validity of

the "parallel trends" assumption for the majority of the outcomes analyzed in this paper, thus

conferring credibility on the parameters obtained from our random trend models (which use the

evolution of outcomes in non-SHI countries as the counterfactual for countries that adopted

                                                        34


SHI). Third, the results from an informal test for endogeneity in the random trend models

suggest that reverse causality--that is, countries switching to SHI because of the evolution of

their outcomes--is not present in the context of our health sector analysis. Finally, a battery of

tests show that our IV models are well suited to the task of controlling for the potential

endogeneity of SHI adoption identified in the random trend models for some important labor

outcomes, thus providing reliable parameter estimates for the causal effect of SHI adoption in

such cases.



        Our estimates suggest that SHI adoption per se increased government health expenditure

per capita. We also obtain some evidence that part of the extra financial resources available in

the health sector due to SHI adoption have served to increase the fraction of salaries as

percentage of government health spending in SHI countries. This result provides quantitative

evidence in support of claims about the process of transition to SHI in some ECA countries being

favored and accelerated by pressure from health professionals, who expected to have their

income levels driven up by the introduction of a SHI system.35 Although SHI has impacted on

how physical resources are used through a reduction in the average hospital length of stay and

increases in hospital admissions and bed occupancy rates after SHI adoption, only one

indicator--the surgical infection rate--suggests a (large) improvement in the quality of hospital

care for SHI systems. Even though SHI systems ended up spending more on health care than

their tax-funded counterparts, our analysis of several mortality and morbidity indicators showed

that transition to SHI has not caused general improvements in health outcomes for ECA

countries.




35See, for instance, the individual WHO-HiT reports for the Czech Republic and the Russian Federation.

                                               35


       These results are mostly robust to the inclusion of dummy variables capturing shifts in

provider payment methods alongside the SHI status dummy; they are therefore pure SHI effects.

For example, the higher government spending caused by a transition to SHI is not a spurious

result attributable to the fact that some countries switched to fee-for-service (FFS) when they

adopted SHI. We are able to estimate separate provider payment effects because SHI adoption

did not always lead to provider payment reform and even when it did sometimes did so with a

lag, because some non-SHI countries reformed the way they paid hospitals as well, and because

some SHI countries switched provider payment methods more than once (some, for example,

switched to FFS only to change to a patient-based payment method (PBP) later on). Typically,

where we find an impact of provider payments on our outcomes we do not find an impact of SHI

adoption, and vice versa. For example, private health spending is affected (positively) by a

switch to FFS but not by SHI adoption, while government spending is (positively) affected by

SHI adoption but not by how hospitals are paid. An exception is acute care admissions, which

respond positively to both SHI adoption (a 2.5% increase) and a switch to FFS (a 3% increase).

The impact of SHI adoption on acute care admissions was thus greater in countries where SHI

adoption was associated with a switch to FFS than in countries where SHI adoption did not

coincide with a provider payment reform.



       The question arises: Why did health outcomes not improve as a result of SHI adoption

even though it led of itself to higher government health spending and higher acute in-patient

admissions? One part of the explanation is that the percentage increase in admissions was much

smaller than the percentage increase in spending (2.5% compared to 11%). Much of the extra

spending therefore resulted in more costly admissions and/or extra spending elsewhere in the

health system. Part of the story seems to be the higher salary share of costs as a result of SHI

                                                 36


adoption. But it also seems likely that costs were incurred undertaking new activities (e.g.

collecting contributions, writing contracts with providers) or that existing activities became more

costly (e.g. more tests being administered on in-patients, more expensive drugs being given,

etc.). It is also possible that SHI adoption may have resulted in less comprehensive and less well

integrated public health and prevention programs (cf. e.g. Allin et al. 2004), and that the extra

admissions and extra costs caused by the transition to SHI were incurred in treating additional

patients who would not have otherwise become sick. The fact that SHI adoption appears to have

led to increased numbers of infectious disease hospital discharges (Table 6) is consistent with

this story. Gaps in coverage may also be part of the explanation. Some groups seem to have

fallen through the coverage net, such as the Roma population (cf. e.g. Rechel and McKee 2003),

and there is anecdotal evidence that some formal sector workers wait to enroll until they get sick.

Because of lack of coverage, these groups may use primary care less than they would have

otherwise done, increasing the likelihood illness is left untreated until serious enough to warrant

hospitalization. Some of the extra hospital caseload associated with SHI may therefore simply be

due to people waiting until they get so sick that they require hospitalization.



         As far as the relationship between SHI adoption and the labor market is concerned, our

work offers some empirical support to the alleged damaging effect of SHI on employment levels.

We find that registered unemployment rates doubled in SHI countries compared to what the rates

would have been if those countries had not adopted SHI. In spite of the sizeable magnitude of

this effect, this result is not corroborated by a commensurate increase in total unemployment

rates. We do, however, find corroborating evidence of a negative impact of SHI adoption on

employment from our estimated negative impact of SHI on the employment-to-population ratio

of people aged 15-59. SHI is also estimated to have strongly increased average gross wages and

                                                    37


salaries in the formal sector. On the other hand, transition to a SHI system did not cause an

increase in the size of the informal sector of the ECA countries according to our--admittedly

fallible--measures of informality. There is also no evidence that SHI adoption led to an increase

on informal employment in SHI countries of the region.



        Of course, our results do not necessarily imply that SHI adoption everywhere must

necessarily reduce employment, and raise health spending without improving health outcomes.

The latter pair of results in particular is likely to hinge in part on the fact that SHI was introduced

with costly institutional reforms but ones that did little to stimulate the performance of the health

system. Nonetheless, the largely negative results in the paper ought to serve as a warning to those

contemplating shifting from general revenue finance to SHI.

                                                     38



                                Table 1: Definitions of SHI status in the dataset

                        Year of SHI
Country                   adoption     SHI dummy                              Comments

Albania                    1995        =1 1995 onwards
                                       =0 beforehand
Armenia                    Never       =0 throughout
Azerbaijan                 Never       =0 throughout
Belarus                    Never       =0 throughout
Bosnia and Herzegovina  Prior to 1990  =1 for 1991                            SHI in place prior and (in theory)
                                       Missing 1992-96                        during the to 1992-95 war, but war
                                       =1 1997 onwards                        period excluded from analysis
Bulgaria                   1999        =1 1999 onwards
                                       =0 beforehand
Croatia                 Prior to 1990  =1 1990 onwards
Czech Republic             1993        =1 1993 onwards
                                       =0 beforehand
Estonia                    1992        =1 1992 onwards
                                       =0 beforehand
Georgia                    1995        =1 1995 onwards
                                       =0 beforehand
Hungary                    1990        =1 1990 onwards
Kazakhstan                 1996        =1 1996-1998                           Abandoned SHI in 1998
                                       =0 otherwise
Kyrgyz Republic            1997        =1 1997 onwards
                                       =0 beforehand
Latvia                     Never       Option 1: =0 throughout                Latvia set up SHI agency before
                                       Option 2: =1 1997 onwards, 0 otherwise 1997 but it was funded through
                                                                              general revenues. Since 1997, 28.4%
                                                                              of income taxes are earmarked for
                                                                              health
Lithuania                  1991        =1 1991 onwards
                                       =0 otherwise
Macedonia, FYR             1991        =1 1991 onwards
                                       =0 1990
Moldova                    Never       =0 throughout
Poland                     Never       Option 1: =0 throughout                Health system funded not through
                                       Option 2: =1 1999 onwards, 0 otherwise payroll tax but earmarked income tax
Romania                    1998        Option 1: =1 1998 onwards, 0 otherwise Earmarked payroll taxes were
                                       Option 2: missing 1990-91              established in 1992 (2% of payroll)
                                         =1 1992 onwards

Russian Federation         1993        =1 1993 onwards
                                       =0 beforehand
Serbia and Montenegro   Prior to 1990  =1 throughout
Slovak Republic            1995        =1 1995 onwards
                                       =0 beforehand
Slovenia                   1992        =1 1992 onwards
                                       =0 beforehand
Tajikistan                 Never       =0 throughout
Turkey                  Prior to 1990  =1 throughout                          Two out of the three main insurance
                                                                              funds collect payroll-based
                                                                              earmarked contributions for health
                                                                              and cover approximately 87% of the
                                                                              population
Turkmenistan               Never       =0 throughout                          WHR has social security spending
                                                                              ranging from 6.1-9.9%.
Ukraine                    Never       =0 throughout
Uzbekistan                 Never       =0 throughout

                                                    39


                   Table 2: Descriptive statistics for health sector outcome variables

                                      Full sample               SHI = 1                SHI = 0
                                 Mean      SD      Obs    Mean      SD    Obs    Mean    SD    Obs
Health expenditures - Total      402.83 296.92     359    536.16 325.16   186    259.47 172.88 173
Health expenditures - Government 295.87 249.61     324    404.22 281.52   167    180.62 136.89 157
Health expenditures - Private    101.26   71.77    324    123.91   72.54  167    77.16  62.69  157
Salaries (%)                     39.41    12.63    168    40.24    16.62  69     38.82  8.95   98
Physicians                        2.97    0.93     343     2.72    1.01   158     3.20  0.79   184
Life expectancy                  70.49    2.91     380    71.52    3.03   181    69.55  2.45   198
Life expectancy (male)           66.41    3.51     377    67.53    3.76   178    65.39  2.92   198
Life expectancy (female)         74.64    2.65     377    75.64    2.61   178    73.73  2.36   198
Under-5 MR (TransMONEE)          21.59    13.07    383    15.95    8.19   179    26.58  14.51  203
Under-5 MR (WHO)                 21.12    13.05    366    15.37    7.73   167    25.99  14.61  198
Infant MR (WB)                   20.22    19.66    231    13.25    11.10  127    28.95  24.25  102
Infant MR (TransMONEE)           17.10    9.48     399    13.24    6.57   181    20.40  10.42  212
Infant MR (WHO)                  16.95    9.57     379    14.31    9.29   179    19.29  9.25   198
Perinatal MR                     12.40    4.74     352    10.91    5.25   161    13.64  3.87   191
Neonatal MR                       7.78    3.00     296     7.35    3.34   154     8.24  2.51   141
Postneonatal MR                   7.31    6.29     295     4.89    3.18   154     9.99  7.66   140
Maternal MR                      28.46    21.78    383    23.05    23.01  178    33.27  19.51  204
Maternal MR (3-year)             28.65    18.88    349    21.94    17.12  156    34.16  18.55  192
Caesarean sections               92.77    50.10    331    118.53   43.52  154    70.35  44.43  177
SDR all causes                  1145.92 184.92     366   1081.83 182.81   167   1200.68 169.12 198
SDR infeccious diseases          17.16    15.31    363    11.81    10.25  167    21.77  17.37  195
SDR tuberculosis                  9.84    8.02     361     7.57    8.23   167    11.80  7.33   193
SDR diarrhoea (under 5)          31.38    67.31    354    11.82    22.38  166    48.92  86.59  187
SDR ARI (under 5)                105.48 145.38     342    46.61    66.83  167    162.47 175.26 174
SDR heart disease                302.88 129.88     363    239.05 109.83   167    358.58 119.68 195
SDR liver diseases               29.41    21.36    321    26.16    18.02  151    32.39  23.66  169
SDR diabetes                     14.92    8.59     363    14.35    6.91   167    15.39  9.81   195
SDR circulatory diseases         623.62 125.33     363    576.83 120.03   167    664.11 115.85 195
SDR cerebrovascular diseases     175.36   53.23    363    171.89   55.21  167    178.57 51.46  195
SDR neoplasms                    172.78   47.55    363    190.83   45.72  167    157.40 43.72  195
SDR female breast cancer         21.58    6.68     363    24.52    5.70   167    19.11  6.46   195
SDR respiratory diseases         68.31    34.93    363    54.13    27.91  167    80.59  35.80  195
SDR bronchitis                   30.99    19.57    350    25.36    19.93  161    35.85  17.99  188
SDR digestive diseases           48.09    22.90    363    44.73    20.10  167    51.03  24.76  195
SDR alcohol causes               134.83   57.24    321    123.72   57.61  155    145.75 54.78  165
SDR smoking causes               542.38 167.07     321    466.26 131.64   155    615.31 164.40 165
Tuberculosis incidence rate      52.88    31.90    416    50.41    33.97  201    54.98  29.76  213
Hepatitis incidence rate         141.31 170.91     320    67.66    83.37  148    205.37 199.84 171
Hepatitis B incidence rate       17.23    19.29    384    10.38    9.61   178    23.24  23.26  205
Measles incidence rate           13.31    26.96    415    10.33    23.45  200    16.17  29.76  213
Mumps incidence rate             54.72    76.49    390    37.95    58.73  182    69.63  86.83  207
Syphilis incidence rate          31.77    52.15    381    28.21    55.82  180    35.14  48.61  200
Congenital syph incidence rate    0.16    0.30     218     0.23    0.40   88      0.11  0.18   129
Pertussis incidence rate          4.05    5.62     413     4.33    6.78   199     3.78  4.27   213

                                                           40


                                            Full sample                 SHI = 1                   SHI = 0
                                      Mean       SD       Obs      Mean   SD     Obs      Mean       SD     Obs
Diphteria incidence rate               1.31     5.19      415      0.67   2.94    200      1.93     6.61    213
Tetanus incidence rate                 0.09     0.11      407      0.11   0.12    196      0.07     0.09    210
Cancer incidence rate                245.01 147.37        335     318.21 159.21   138     193.72 113.47     197
Tuberculosis immunization rate        92.70     11.32     420      93.10  10.39   201     93.24     9.93    213
DPT immunization rate                 91.59     9.75      420      91.90  7.67    201     91.87     10.31   213
Measles immunization rate             91.22     9.80      420      91.57  8.57    201     91.70     9.45    213
Polio immunization rate               92.06     8.81      420      91.94  7.76    201     92.82     8.69    213
Mumps immunization rate               82.25     22.69     227      88.63  16.39   127     74.59     26.50    99
Rubella immunization rate             88.00     19.23     190      90.05  15.52   124     85.29     22.78    65
Length of stay (total)                12.75     3.04      398      11.23  2.93    193     14.19     2.39    204
Length of stay (acute care)           10.87     2.71      267      9.40   2.37    150     12.78     1.79    116
Bed occupancy rate                    72.80     14.78     278      74.85  9.82    149     70.36     18.78   128
Hospital beds                          8.10     2.91      342      6.72   2.59    155      9.28     2.64    186
In-patient admissions                 16.18     5.99      397      15.51  6.02    194     16.86     5.90    202
Acute care admissions                 15.16     5.39      277      15.17   5.5    150     15.20     5.26    126
Hospital discharges - infectious     826.10 444.11        353     658.64 352.78   170     981.66 464.08     183
Hosp discharges - cancers            809.34 588.59        346    1068.85 643.08   163     578.19 417.81     183
Hosp discharges - heart              669.00 468.43        343     684.00 425.49   163     655.41 504.97     180
Hosp discharges - circulatory       1904.45 1099.73 354          2092.45 1152.72 170     1730.76 1021.10    184
Hosp discharges - cerebrov           339.11 240.51        350     394.82 243.94   168     287.68 226.04     182
Hosp discharges - respiratory       2088.68 1014.24 351          1737.75 778.42   170    2418.28 1098.08    181
Hosp discharges - digestive         1623.59 626.99        354    1544.75 579.86   170    1696.44 660.83     184
Hosp discharges - musculo            776.92 508.15        354     809.35 536.77   170     746.96 479.71     184
SDR appendicitis                       0.30     0.18      347      0.23   0.14    158      0.36     0.19    188
SDR hernia & intestinal                2.23     0.75      350      2.34   0.74    161      2.14     0.74    188
SDR adverse effects                    0.20     0.33      183      0.19   0.33    116      0.19     0.28     66
Surgical infection rate                1.09     1.22       74      0.92   0.82    42       1.30     1.59     32
Note: Mean, standard deviation (SD) and number of observations (Obs) for the full sample and for the sub-samples of
observations with the SHI dummy equals to one (SHI=1) and zero (SHI=0).

                                                           41


                      Table 3: Labor market outcome variables: descriptive statistics

                                         Full sample                    SHI = 1                   SHI = 0
                                 Mean          SD      Obs         Mean    SD    Obs      Mean      SD      Obs
Gross wage                      1115.49 876.72         257       1555.22 921.52 119       736.31 626.16     138
Unemployment                     11.68        5.91     229         11.66  6.16   162      11.73     5.29     67
Registered unemployment           9.27        8.26     295         12.98  9.11   141       5.87     5.56    154
Empl-to-pop - ILO                51.39        6.31     150         51.22  5.96   110      51.86     7.25     40
Empl-to-pop - TransMONEE         67.31        9.00     360         64.54  9.12   158      69.47     8.30    202
Informal economy                 21.82       21.07     238         22.35  19.94   93      21.49    21.82    145
Self-employment                  17.24        9.26     190         16.13  8.85   132      19.77     9.73     58
Employment in agriculture        25.26       16.64     268         23.07  18.93  162      28.61    11.64    106
Note: Mean, standard deviation (SD) and number of observations (Obs) for the full sample and for the sub-samples of
observations with the SHI dummy equals to one (SHI=1) and zero (SHI=0).

                                                                          42


                                                   Table 4: Basic results: health sector outcomes


                                                                     Basic model                   Random trend model
                                                                Eqn (2) estimated by OLS      Eqn (4) estimated by fixed effects
                                                                                       SHI                                SHI
                                                                               p-    impact                      p-     impact    #
                         Dependent variable                 Coef     SE      value     (%)    Coef     SE      value      (%)    shifts
                          Health expenditures - Total       45.87   28.07    0.114    0.113   47.47   27.70    0.098     0.117    11

lth                 gn    Health expenditures - Government  45.57 20.66 0.036         0.150   45.20 20.53 0.037          0.149    11
                      ndi Health expenditures - Private      0.61   13.77    0.965    0.006   2.31    12.85    0.859     0.022    11
   Hea                pes Salaries (%)                       4.35   1.55     0.012    0.111   6.34    2.17     0.009     0.162    3
                          Physicians                         0.04   0.04     0.322    0.013   0.03    0.04     0.410     0.010    13
                          Life expectancy                   -0.21   0.23     0.370    -0.003  -0.24   0.22     0.297    -0.003    13
                          Life expectancy (male)            -0.24   0.27     0.371    -0.004  -0.27   0.26     0.316    -0.004    13
                          Life expectancy (female)          -0.15   0.16     0.374    -0.002  -0.17   0.16     0.286    -0.002    13
                          Under-5 MR (TransMONEE)            0.10   0.87     0.911    0.005   -0.12   0.89     0.896    -0.006    14
                          Under-5 MR (WHO)                   0.92   0.69     0.195    0.045   0.81    0.67     0.241     0.039    13
                          Infant MR (WB)                     1.03   0.50     0.068    0.092   0.99    0.50     0.074     0.089    7
                          Infant MR (TransMONEE)             0.25   0.57     0.667    0.015   0.10    0.57     0.859     0.006    14
                          Infant MR (WHO)                    0.40   0.49     0.429    0.024   0.37    0.48     0.446     0.023    14
                          Perinatal MR                       0.39   0.32     0.244    0.032   0.19    0.36     0.598     0.016    13
                          Neonatal MR                        0.56   0.39     0.158    0.075   0.43    0.45     0.348     0.057    10
                          Postneonatal MR                   -0.46   0.27     0.099    -0.065  -0.55   0.27     0.056    -0.078    10
                          Maternal MR                        2.59   2.08     0.225    0.091   2.80    2.15     0.206     0.099    13
                          Maternal MR (3-year)               1.31   0.99     0.198    0.045   1.46    1.11     0.199     0.050    13
                          Caesarean sections                -0.27   0.87     0.761    -0.003  -0.90   0.98     0.370    -0.009    10
                          SDR all causes                     5.96   20.82    0.777    0.005   9.37    20.95    0.659     0.008    13
                          SDR infeccious diseases            1.38   1.79     0.447    0.078   1.45    1.87     0.446     0.082    13
                          SDR tuberculosis                   1.63   1.48     0.282    0.157   1.83    1.53     0.245     0.177    13
      tcomes              SDR diarrhoea (under 5)            4.09   5.24     0.442    0.134   2.74    5.52     0.625     0.089    13
            ou            SDR ARI (under 5)                  1.63   5.00     0.747    0.016   2.25    5.43     0.683     0.022    12
              lth         SDR heart disease                 -0.36   6.01     0.953    -0.001  1.93    5.66     0.735     0.006    13
                 Hea      SDR liver diseases                -0.53   1.19     0.659    -0.017  -0.25   1.14     0.828    -0.008    10
                          SDR diabetes                       0.66   1.29     0.615    0.044   0.89    1.33     0.509     0.060    13
                          SDR circulatory diseases          -2.00   14.24    0.889    -0.003  1.79    14.49    0.903     0.003    13
                          SDR cerebrovascular diseases       0.88   4.80     0.856    0.005   1.16    4.94     0.816     0.007    13
                          SDR neoplasms                      2.23   2.20     0.320    0.013   2.00    2.24     0.380     0.012    13
                          SDR female breast cancer           0.06   0.43     0.883    0.003   0.09    0.43     0.844     0.004    13
                          SDR respiratory diseases           2.03   1.85     0.283    0.029   2.23    2.00     0.276     0.032    13
                          SDR bronchitis                     3.68   3.47     0.300    0.118   3.62    3.67     0.334     0.116    13
                          SDR digestive diseases            -0.32   1.08     0.773    -0.006  -0.06   1.03     0.953    -0.001    13
                          SDR alcohol causes                -1.27   3.64     0.731    -0.009  -0.87   3.43     0.802    -0.006    11
                          SDR smoking causes                 1.65   13.69    0.905    0.003   1.02    14.20    0.943     0.002    11
                          Tuberculosis incidence rate       -2.63   2.71     0.342    -0.048  -2.29   2.86     0.430    -0.042    14
                          Hepatitis incidence rate          34.08   26.07    0.203    0.264   29.43   26.71    0.281     0.228    14
                          Hepatitis B incidence rate         1.69   1.07     0.127    0.104   0.85    1.09     0.440     0.052    12
                          Measles incidence rate            -5.09   7.56     0.507    -0.431  -6.53   7.55     0.395    -0.553    14
                          Mumps incidence rate               4.68   8.90     0.603    0.081   4.26    9.67     0.663     0.074    14
                          Syphilis incidence rate            7.38   9.77     0.457    0.201   7.69    10.58    0.474     0.209    14

                                                                43


                                                            Basic model                         Random trend model
                                                    Eqn (2) estimated by OLS              Eqn (4) estimated by fixed effects
                                                                             SHI                                      SHI
                                                                   p-       impact                           p-     impact       #
            Dependent variable                  Coef        SE    value      (%)          Coef      SE      value     (%)      shifts
            Congenital syph incidence rate      -0.01     0.02    0.704     -0.053        -0.02    0.03    0.482    -0.109       7
            Pertussis incidence rate            1.05      1.06    0.333     0.258         0.95     1.11    0.402     0.234      14
            Diphteria incidence rate            -0.05     0.68    0.936     -0.037        -0.13    0.70    0.850    -0.089      14
            Tetanus incidence rate              0.02      0.02    0.287     0.208         0.02     0.02    0.264     0.219      14
            Cancer incidence rate               2.37      3.06    0.447     0.010         -0.02    3.10    0.994     0.000      14
            Tuberculosis immunization rate      0.92      1.41    0.520     0.010         0.84     1.42    0.559     0.009      14
            DPT immunization rate               -0.40     1.17    0.738     -0.004        -0.25    1.27    0.846    -0.003      14
            Measles immunization rate           -0.24     0.55    0.661     -0.003        -0.20    0.58    0.737    -0.002      14
            Polio immunization rate             1.25      1.40    0.381     0.013         1.36     1.46    0.360     0.015      14
            Mumps immunization rate             9.69      6.55    0.155     0.116         8.32     6.98    0.248     0.100      10
            Rubella immunization rate           13.90    10.73    0.212     0.153         8.98     7.28    0.233     0.099       6
            Length of stay (total)              -0.32     0.17    0.063     -0.026        -0.30    0.17    0.081    -0.024      14
            Length of stay (acute care)         -0.16     0.22    0.463     -0.015        -0.14    0.22    0.524    -0.013      10
            Bed occupancy rate                  1.91      1.06    0.085     0.026         2.19     0.99    0.039     0.030       9
            Hospital beds                       -0.17     0.19    0.371     -0.021        -0.23    0.19    0.238    -0.028      13
            In-patient admissions               0.44      0.17    0.015     0.027         0.37     0.19    0.061     0.023      14
            Acute care admissions               0.63      0.20    0.004     0.042         0.57     0.21    0.014     0.037      10
            Hospital discharges - infectious    90.63    45.92    0.060     0.110         81.85    49.06   0.108     0.099      13
  ls        Hosp discharges ­ cancers           25.18    17.38    0.160     0.030         21.68    17.95   0.238     0.026      13
    ita     Hosp discharges ­ heart             11.65    10.53    0.279     0.017         16.43    10.93   0.145     0.024      13
       sp   Hosp discharges ­ circulatory       37.40    23.54    0.125     0.019         32.16    26.94   0.244     0.017      13
         Ho
            Hosp discharges ­ cerebrov          12.45     6.65    0.073     0.036         12.65    6.75    0.073     0.036      13
            Hosp discharges ­ respiratory       96.49    60.19    0.121     0.047         86.83    69.61   0.224     0.042      13
            Hosp discharges ­ digestive         20.08    17.52    0.263     0.012         7.36     18.81   0.699     0.005      13
            Hosp discharges ­ musculo           17.74    10.60    0.107     0.022         12.93    10.43    0.227    0.016      13
            SDR appendicitis                    -0.04     0.05    0.436     -0.141        -0.05    0.05    0.405    -0.156      13
            SDR hernia & intestinal             -0.16     0.11    0.172     -0.071        -0.18    0.11    0.111    -0.083      13
            SDR adverse effects                 0.00      0.05    0.931     0.021         0.00     0.06    0.983    -0.006       6
            Surgical infection rate             -1.32     0.41    0.013     -1.427        -1.38    0.37    0.006    -1.488       3
Notes: Results refer to the coefficient (Coef) and standard-error (SE, cluster-adjusted) of the SHI dummy variable. Significant at
5% and 10%. P-values from two-sided t-tests. SHI impact (%) calculated over the average outcome in the corresponding estimating
sub-sample. In the last column, number of shifts refers to the number of transitions between tax-funded and SHI systems in the sub-
sample used to estimate the corresponding coefficients.

                                                                                                                                                                                                                                                                                                                                                                        eth                                                                                                                                               age                                                                                                                                                                                                                                                                                                                             ear
                                                                                                                 en                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                      stage
                                                                                                                                                                                                    J                   200.
                                                                                                                   Hans        statistic                                                                                                                                                                                 135. eref                                         tor                                                                                                                                               avereth                                                                                                                                                          rstifeth                                                                                                                                                       ntse


                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                    over                                                                                                                                                              in
                                                                                                                                                                                                                                                                                                                                                                              Results                                                                                                                                                   ed                                                                                                                                                                                                                                                                                                                       instrumeth

                                                                                                                                                                                                     F                                                                                     945.                          790.                                                                                                                                                                                                             lat
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                             lcu                                                                                                                                                                                                                                                                                                                           atth
                                                                                                                       Partial          statistic                                                                                                                                              55                            26                                                                                                                                                                                                                 ca                                                                                                                                                                                                                                                                                                                             is
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                        instruments
                                                                                                                                                                                                                                                                                                                                                                                     communism.                                                                                                                                                   (%)

                                                                                                                                                  (%)                                                                                                                                                                                                                                          to                                                                                                                                                    ct
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                       pa                                                                                                                                                          cluded

                                                                                                                                                     pact                                                                   12.0                                                                                               36.3                                                              priory
                                                                                                                                                         im                                                                     -0                                                                                                 -0
    es                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                   imIHS.ts                                                                                                                                                        ex                                                                                                                                                      thesisopyhl
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                           of                                                                                                                                                               nult

                                                                                                                                                           SHI                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                               subset


      outcomr                                                          esatmitse                                                                                                                                                                                                                                                                                                                       ountrceht                                                                                                                                                 -test
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                   the                                                                                                                                                          joinehte
                                                                                                                                                                                                                                  6                                                                                                  1                                                                          in                                                                                                                                                                                                                                                                                                                    of

                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                        test                                                                                                                                                            wher,
                                                                                M                                                                             p-value                                                              310.                                                                                               000.                                                                                                                                                                                                                            two-sided
                                                                                                                                                                                                                                                                                                                                                                                                                  existed                                                                                                                                                                                                                                                                                                                   ce
             secto
                                                                                 GM                                                                                                                                                                                                                                                                                                                                      SHI                                                                                                                                                   from
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                   es                                                                                                                                                         ifican
                  ealthh                                                                                                                                             SE                                                                901.                                                                                               970.                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                               portederear                                                                                                     .
                                                                                                                                                                                                                                                                                                                                                                                                                            ether
                                                                                                                                                                                                                                                                                                                                                                                                                                 wh                                                                                                                                                  P-valu.                                                                                                                                                        insignt                                                                                                                                                                                                                                                                   tion
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                        tions
                                                                                                                                                                                                                                           5.9                                                                                                2.9                                                                                  for                                                                                                                                                      5%                                                                                                                                                             join                                                                                                                                                              rict                                                                                                 equa

                                                                                                                                                                       Coef                                                                   -1                                                                                                 -3
                        selected                                                                                                                                                                                                                                                                                                                                                                                                      toraci                                                                                                                                                  att                                                                                                                                                              the                                                                                                                                                               res

                                s:                                                                                                                                                                                                                                                                                                                                                                                                          ind                                                                                                                                                  cani                                                                                                                                                             tore                                                                                                                                                              ngiy                                                                                              inameth

44                                                                                                                                                                         (%)                                                                                                                                                                                                                                                                 an                                                                                                                                                                                                                                                                                                                     ref                                                                                                                                                               ntife                                                                                                from
                                  result
                                                                                                                                                                              pact                                                              32.0                                                                                               10.3                                                                                                                                                                                                                                              Signif


                                                                                                                                                                                  im                                                                -0                                                                                                 -0                                                                                        andym                                                                                                                                                                                                                                                                                                                   -robust)                                                                                                                                                            overid
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                   of
                                                                                   estimates                                                                                                                                                                                                                                                                                                                                                          dum                                                                                                                                                  variable.

                                        variables                                                                                                                                   SHI
                                                                                                                                                                                                                                                                                                                                                                                                                                                         SHI                                                                                                                                                                                                                                                                                                                     (cluster                                                                                                                                                            testeth                                                                                     cludedxeylct

                                                                                                                                                                                                                                                      4                                                                                                  4
                                                 ntale                                      squarest                                                                                                                                                                                                                                                                                                                                                        the                                                                                                                                                     dummy                                                                                                                                                                                                                                                                                                                                                                                                                                    corre

                                                                                                                                                                                       p-value                                                         280.                                                                                               000.                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                              for

                                                                                                    leaseg                                                                                                                                                                                                                                                                                                                                                     ofgal                                                                                                                                                     SHIeth                                                                                                                                                          statistics
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                   F                                                                                                                                                                                                                                                              aresnte

                                                      Instrum                                                                                                                                 SE                                                           951.                                                                                               011.
                                                             5:
                                                                                                          Two-sta                                                                                                                                                                                                                                                                                                                                                   rstifethear                                                                                                                                                of                                                                                                                                                                                                                                                                                                                              statisticsJ                                                                                               rumt
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                    Partial                                                                                                                                                                                                                                                                  ins

                                                                                                                                                                                                                                                               3.1                                                                                                5.2                                                                                                                                                                                                                                                            djusted)a-
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                           ter                                                                                                                                                                                                                                                                                                                            Hansen's                                                                                              udedcl
                                                               Table                                                                                                                            Coef                                                              -2                                                                                                 -3                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                               ex
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                              clus                                                                                                                                                         sub-sample.

                                                                                                                                                                                                                                                                                                                                                                                                                                                                               instruments                                                                                                                                                        E,                                                                                                                                                                                                                                                                                                                                                                                                                                    the
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                    (S                                                                                                                                                                                                                                                                                                                                                                                                                                     atth
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                      orrre                                                                                                                                                           timating
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                              es                                                                                                                                                  cluster-robust                                                                                               and
                                                                                                                                                                                                                                                                    e                                                                                                                                                                                                                     excluded
                                                                                                                                                                                                                                                                     rat                          re)                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                             term
                                                                                                                                                                                                                                                                                                     ca                                                                                                                                                                                           the                                                                                                                                                                                                                                                                                                                                                                                                                                                                                           lumn,

                                                                                                                                                                                                      e                                                                 oni                                                                                                                                                                                                                                                                                                                                                                                andard-ts                                                                                                                                                                                                                                                                                                                                 co
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                      error

                                                                                                                                                                                                       ablri
                                                                                                                                                                                                            va                                                             zatnium                     (acute                                                                                                                                                                                        models,                                                                                                                                                        and)fe                                                                                                                                                                                                                                                                                                                             last                                                                                                the
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                corresponding                                                                                                                                                              the
                                                                                                                                                                                                                                                                                                             stay                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                             with

                                                                                                                                                                                                              entdn                                                               imse                                                                                                                                                                                                                      both
                                                                                                                                                                                                                                                                                                                 of                                                                                                                                                                                                                                                                                                                                                       (Cot                                                                                                                                                               the                                                                                                                                                              In
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                in
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                For                                                                                                                                                           cien
                                                                                                                                                                                                                                                                                      easl
                                                                                                                                                                                                                   Depe                                                                   M                        Length                                                                                                                                                                                          Notes:                                                                                                                                                         coeffi                                                                                                                                                          outcome                                                                                                                                                       regressions.                                                                                      uncorrelated

                                                                       45


      Table 6: Random trend results: health outcomes, SHI adoption and hospital payment methods


                                                                                        Hospital Payment Methods
                                                                SHI dummy         Fee-for-service        Patient-based
                                                                        Impact               Impact                 Impact
                          Dependent variable                Coef    SE   (%)   Coef    SE      (%)   Coef    SE      (%)
                           Health expenditures ­ Total      19.39 22.73 0.048  68.99 29.63    0.170  10.99 20.02    0.027
                                                                               34.55 22.59    0.115  -7.89  30.06 -0.026
lth                 gn     Health expenditures ­ Government 33.49 16.22 0.111
                                                                                              0.342  19.99 18.88    0.189
   Hea                ndie Health expenditures ­ Private   -14.93 15.67 -0.141 36.22 14.55
                      sp Salaries (%)                       7.19   5.71 0.180  1.47   2.32    0.037  -2.48   6.12   -0.062
                           Physicians                       0.03   0.04 0.011  -0.02  0.06    -0.006  0.02   0.04   0.006
                           Life expectancy                  -0.35  0.23 -0.005 0.30   0.26    0.004   0.19   0.25   0.003
                           Life expectancy (male)           -0.40  0.28 -0.006 0.38   0.35    0.006   0.18   0.29   0.003
                           Life expectancy (female)         -0.25  0.16 -0.003 0.16   0.16    0.002   0.19   0.20   0.002
                           Under-5 MR (TransMONEE)          -0.15  0.93 -0.007 0.04   0.90    0.002   0.08   0.65   0.004
                           Under-5 MR (WHO)                 0.65   0.74 0.031  0.30   0.88    0.014   0.44   0.93   0.021
                           Infant MR (WB)                   1.00   0.72 0.089  -0.01  0.52    -0.001  0.00   0.68   0.000
                           Infant MR (TransMONEE)           0.14   0.66 0.008  -0.19  0.57    -0.011  0.12   0.53   0.007
                           Infant MR (WHO)                  0.26   0.68 0.016  0.04   0.65    0.003   0.54   0.92   0.033
                           Perinatal MR                     0.16   0.49 0.013  0.05   0.46    0.004   0.09   0.72   0.007
                           Neonatal MR                      0.17   0.62 0.022  0.53   0.53    0.069   0.54   0.70   0.071
                           Postneonatal MR                  -0.46  0.28 -0.065 -0.49  0.28    -0.069  0.03   0.29   0.004
                           Maternal MR                      2.85   2.71 0.099  -0.61  3.71    -0.021  0.32   3.17   0.011
                           Maternal MR (3-year)             1.61   1.37 0.055  -1.23  1.41    -0.042  0.56   1.81   0.019
                           Caesarean sections               -1.72  1.19 -0.018 -1.91  2.18    -0.020  4.36   3.28   0.046
                           SDR all causes                   19.89 21.08 0.017 -20.60 18.87 -0.018 -23.74 21.79 -0.021
                           SDR infeccious diseases          1.13   1.99 0.063  0.18   1.45    0.010   1.12   1.32   0.062
                           SDR tuberculosis                 1.81   1.57 0.172  -0.34  0.76    -0.033  0.40   0.73   0.038
                           SDR diarrhoea (under 5)          1.86   4.48 0.058  4.29   6.40    0.135  -0.88   3.44   -0.028
      tcomes               SDR ARI (under 5)                2.49   6.53 0.023  -5.46  7.39    -0.051  4.13  10.39   0.039
            ou             SDR heart disease                6.21   6.71 0.020  -6.22  9.79    -0.020 -11.47 11.94 -0.038
              lth          SDR liver diseases               -0.27  1.24 -0.009 -1.63  1.02    -0.051  1.77   1.45   0.056
                 Hea       SDR diabetes                     1.57   1.05 0.103  -3.46  0.78    -0.228  0.70   1.64   0.046
                           SDR circulatory diseases        11.42  13.83 0.018  -6.76  12.61 -0.011 -35.53 17.70 -0.057
                           SDR cerebrovascular diseases     4.62   4.96 0.026  -2.67  5.64    -0.015 -12.58  6.34   -0.071
                           SDR neoplasms                    2.24   2.54 0.013  0.90   2.45    0.005  -1.59   1.43   -0.009
                           SDR female breast cancer         0.26   0.44 0.012  -0.28  0.37    -0.013 -0.46   0.40   -0.021
                           SDR respiratory diseases         3.42   2.49 0.049  -4.22  2.70    -0.061 -0.78   2.41   -0.011
                           SDR bronchitis                   5.27   4.62 0.172  -4.90  2.86    -0.160 -2.07   2.73   -0.067
                           SDR digestive diseases           -0.30  1.02 -0.006 -0.94  1.11    -0.019  1.88   1.41   0.038
                           SDR alcohol causes               -0.17  5.36 -0.001 -5.28  8.24    -0.039  3.35   7.11   0.024
                           SDR smoking causes               9.09  16.82 0.017  -7.00  14.22 -0.013 -23.13 13.60 -0.042
                           Tuberculosis incidence rate      -5.04  4.01 -0.092 4.40   2.47    0.081   7.30   5.17   0.134
                           Hepatitis incidence rate        15.90  25.59 0.121  20.91 23.52    0.159  30.44 19.17    0.231
                           Hepatitis B incidence rate       0.54   1.27 0.033  0.85   0.84    0.051   0.62   0.98   0.038
                           Measles incidence rate           -4.93  8.51 -0.412 -3.18  5.26    -0.266 -1.83   5.10   -0.152
                           Mumps incidence rate             14.19 20.65 0.257  0.50   17.37   0.009  -41.51 49.69 -0.751
                           Syphilis incidence rate          5.83   9.18 0.170  -2.41  6.26    -0.070 15.49 10.21    0.452
                           Congenital syph incidence rate   -0.03  0.03 -0.173 0.07   0.04    0.430  -0.08   0.06   -0.486

                                                              46


                                                                                      Hospital Payment Methods
                                                    SHI dummy                  Fee-for-service            Patient-based
                                                                Impact                       Impact                    Impact
            Dependent variable                 Coef      SE       (%)      Coef      SE       (%)    Coef      SE        (%)
            Pertussis incidence rate           1.61     1.39    0.395      -1.01    1.45     -0.248  -1.76    0.82     -0.430
            Diphteria incidence rate           -0.23    0.97    -0.152     -0.16    0.88     -0.105   0.54    0.99     0.351
            Tetanus incidence rate             0.01     0.02    0.107      0.03     0.01     0.300    0.00    0.02     0.050
            Cancer incidence rate              -2.42    3.29    -0.010     3.45      3.87    0.014    6.35    4.65     0.026
            Tuberculosis immunization rate     -1.26    1.82    -0.013     2.83      2.22    0.030    6.13    4.26     0.065
            DPT immunization rate              -1.82    2.27    -0.020     2.19     2.60     0.024    4.47    4.67     0.048
            Measles immunization rate          -0.25    0.84    -0.003     0.03      0.95    0.000    0.01    0.84     0.000
            Polio immunization rate            1.09     1.67    0.012      1.14     1.78     0.012   -0.19    1.35     -0.002
            Mumps immunization rate            4.88     6.21    0.059      7.31      6.13    0.088    2.44    7.67     0.030
            Rubella immunization rate          6.88     7.17     0.076     3.83      3.84    0.042    3.15    6.21     0.035
            Length of stay (total)             -0.17    0.18    -0.014     -0.24    0.22     -0.019  -0.27    0.28     -0.021
            Length of stay (acute care)        0.06     0.16    0.005      -0.33    0.14     -0.031  -0.36    0.30     -0.033
            Bed occupancy rate                 2.51     1.12    0.035      0.37     1.43     0.005   -2.23    1.36     -0.031
            Hospital beds                      -0.11    0.15    -0.014     -0.11    0.18     -0.014  -0.37    0.17     -0.047
            In-patient admissions              0.19     0.16    0.012      0.63     0.17     0.040    0.03    0.25     0.002
            Acute care admissions              0.38     0.18    0.025      0.48     0.21     0.031   -0.02    0.31     -0.001
            Hospital discharges - infectious   67.22   39.04    0.083     40.35 53.67        0.050   26.03 25.37       0.032

   ls       Hosp discharges - cancers          25.09   23.21    0.031     31.03 36.28        0.039   -46.50 33.94 -0.058
     ita    Hosp discharges - heart            3.17    10.82    0.005     32.32 22.66        0.051   23.81 15.62       0.038
        sp  Hosp discharges - circulatory     -42.09 62.89 -0.023 141.61 66.66               0.076  178.95 151.10 0.097
          HoHosp discharges - cerebrov         7.94     5.28    0.024     19.05 11.59        0.058   -0.10    7.72     0.000
            Hosp discharges - respiratory      44.61   61.82    0.022     170.54 85.36       0.086   27.25 79.85       0.014
            Hosp discharges - digestive       -17.45 16.98 -0.011         96.63 53.35        0.061   11.42 26.48       0.007
            Hosp discharges - musculo          4.48    10.96    0.006     48.61 25.08        0.064   -12.44 22.35 -0.016
            SDR appendicitis                   -0.05    0.06    -0.164     -0.04    0.05     -0.126   0.06    0.07     0.202
            SDR hernia & intestinal            -0.16    0.13    -0.071     -0.19    0.11     -0.087   0.10    0.15     0.046
            SDR adverse effects                -0.02    0.07    -0.103     -0.09    0.07     -0.436   0.10    0.07     0.459
            Surgical infection rate            -1.34    0.46    -1.446     -0.04    0.16     -0.043
Notes: Results refer to the coefficient (Coef) and standard-error (SE, cluster-adjusted) of the dummies for SHI status, fee-for-
service and patient-based payment methods. Significant at 5% and 10%. The estimated impact (%) of each dummy on a given
health outcome has been calculated over the average outcome in the corresponding estimating sub-sample.

                                                               47


                                   Table 7: Basic results: labor market outcomes


                                                Basic model                          Random trend model
                                          Eqn (2) estimated by OLS             Eqn (4) estimated by fixed effects

Dependent variable (log)              Coef           SE        p-value          Coef             SE         p-value      # shifts
Gross wage                            0.199        0.076        0.015           0.197          0.067         0.008           6
Unemployment                          0.050        0.050        0.333           0.022          0.056         0.692           7
Registered unemployment               0.182        0.198        0.368           0.166          0.182         0.373          11
Empl-to-pop ­ ILO                    -0.007        0.013        0.596          -0.010          0.017         0.581           3
Empl-to-pop - TransMONEE              0.020        0.014        0.170           0.020          0.014         0.172          13
Informal economy                      0.036        0.099        0.725          -0.002          0.106         0.985          10
Self-employment                       0.040        0.089        0.658           0.018          0.066         0.787           4
Employment in agriculture            -0.021        0.016        0.220          -0.015          0.017         0.410           9
Notes: In all models, the natural logarithm of the dependent variable is used. Results refer to the coefficient (Coef) and standard-
error (SE, cluster-adjusted) of the SHI dummy variable. Significant at 5% and 10%. P-values from two-sided t-tests. In the last
column, number of shifts refers to the number of transitions between tax-funded and SHI systems in the sub-sample used to
estimate the corresponding coefficients.

                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                      .

                                                              en                                                                                                                                                                                                                                                                                                                   SHI                                                                                                                                                 %
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                        10
                                                                                                                                        J                                                  260.                                                           520.                                                             210.                                                                                                                                                                                                                                                                                                                                                                lasteth                                                                                                                                                              eth


                                                                Hans                                                 statistic                                                                                                                                                                                                                                                                                                                                                                                                            and                                                                                                                                                                                                                                                                                                                          with
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                      In
                                                                                                                                                                                                                                                                                                                                                                                      whetherr                                                                                                                                                                                                                                                                                                                                                                                                                                                                             ted
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                             5%
                                                                                                                                                                                                                                                                                                                                                                                              fo                                                                                                                                               att                                                                                                                                                                                                                                                                                                                            ela


                                                                                                                                         F                                                                           519.                                                                           269.                                                                    280.                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                 corr
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                  cani                                                                                                                                                  gressions.er
                                                                    Partial                                                   statistic                                                                                  34                                                                             53                                                                      25              indicator                                                                                                                                                                                                                                                                                                                                                                                                                                                                            unear

                                                                                                                                                                                                                                                                                                                                                                                                         and                                                                                                                                          Signif                                                                                                                                                        stagets
                                                                                                                                                                                                                                                                                                                                                                                                            an                                                                                                                                              able.i                                                                                                                                                         fir                                                                                                                                                            ntsem

                                                                                                                                          eu                                                                                                                                                                                                                                                                                                                                                                                                                      var                                                                                                                                                         the                                                                                                                                                              instru
                                                                                                                                                                                                                           9                                  7                                                                0                                                                                                                                                                                                                                                                                                                                                                                                 in
                                                                                                                                            val
                                                                                                                                               p-                                                                           000.                               100.                                                             950.                                                                          dummy                                                                                                                                                                                                                                                                                                                                                                                                                                                                                  thet

                                                                                                                                                                                                                                                                                                                                                                                                                   SHI                                                                                                                                               dummy
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                   uments                                                                                                                                                                tha
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                            is
                                                                           stea                                                                                                                                                                                                                                                                                                                                       the                                                                                                                                                 SHI                                                                                                                                                            instr
                                                                                                                                                                                                                                                                                                                                                                                                                         of                                                                                                                                                  the
                                                                                                                                                                                                                                2                                  0                                                                2
    outcomes                                                                   estim                                                             SE                                                                              360.                               050.                                                             350.                                                                                  lagt                                                                                                                                                 of                                                                                                                                                                                                                                                                                                                            pothesisyh

            rketam                                                                  GMM                                                                                                                                                                                                                                                                                                                                        firseth                                                                                                                                                                                                                                                                                                        excluded
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                      of                                                                                                                                                                null
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                  adjusted)
                                                                                                                                                                                                                                                                                                                                                                                                                                      are                                                                                                                                                                                                                                                                                                                                                                                                                                                                                   intjo


                  labor                                                                                                                            efoC                                                                              3
                                                                                                                                                                                                                                      041.                              48.0                                                             3                                                                                               ntse                                                                                                                                              cluster-                                                                                                                                                     subseteth                                                                                                                                                                the

                                                                                                                                                                                                                                                                            -0                                                            020.                                                                                               rumtsin                                                                                                                                                                                                                                                                                                             of
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                   (SE,                                                                                                                                                                                                                                                                                                                             where

                                                                                                                                                                                                                                                                                                                                                                                                                                                    ed                                                                                                                                                 orrre                                                                                                                                                       test
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                       ce

                       selected
48                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                       ported,er
                                                                                                                                                       eu                                                                                                                                                                                                                                                                                             exclud
                                                                                                                                                                                                                                          4                                                               2                                   3                                                                                                                                                                                                                                                             standard-                                                                                                                                                                                                                                                                                                                             are                                                                                on.i

                                                                                       mates                                                             val
                                                                                                                                                            p-                                                                             010.                                                            080.                                980.                                                                                                         the                                                                                                                                                                                                                                                                                                          insignifican
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                     and                                                                                                                                                                                                                                                                                                                                                                                                                 equat
                                                                                                                                                                                                                                                                                                                                                                                                                                                               and
                               variables:                                                   esti                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                        f)e                                                                                                                                                          joint                                                                                                                                                           ctionsi
                                                                                                                                                                                                                                                                                                                                                                                                                                                                  ed,                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                         inam
                                                                                                resa                                                                                                                                                                                                                                                                                                                                                                 us                                                                                                                                                    (Co                                                                                                                                                            the                                                                                                                                                               restr
                                         ntale                                                                                                                                                                                                                                                                                                                                                                                                                         is                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                         the
                                                                                                    squ                                                                                                                                        4                                                               6                                   4                                                                                                                                                                                                                                                                                                                                                                                                                                         tor                                                                                                                                                                 ngyi
                                                                                                                                                              SE                                                                                370.                                                            050.                                350.
                                                                                                       leaste                                                                                                                                                                                                                                                                                                                                                            ableriav                                                                                                                                             efficient                                                                                                                                                         refe                                                                                                                                                                 ntif
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                       co                                                                                                                                                                                                                                                                                                                                dei                                                                         fromde

                                              Instrum                                                                                                                                                                                                                                                                                                                                                                                                                            ent                                                                                                                                                     the                                                                                                                                                                                                                                                                                                                                over
                                                     8:                                                      stag-                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                              of

                                                                                                                  Two                                           efoC                                                                                7
                                                                                                                                                                                                                                                     990.                                                           20.1                                8                                                                                                                                                                                                                                                                                   tore                                                                                                                                                                                                                                                                                                                                                                                                           excludyl

                                                                                                                                                                                                                                                                                                                        -0                               000.
                                                       Table                                                                                                                                                                                                                                                                                                                                                                                                                        dependeth                                                                                                                                                   refst                                                                                                                                               (cluster-robust)                                                                                                                                                              testeth                                                                          rrectoc
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                             of                                                                                                                                                      Resul.                                                                                                                                                                                                                                                                                                                              for                                                                              are
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                    statistics
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                               ithm                                                                                                                                                                                                                                                                                                                           F                                                                                                                                                             csi                                                                              ntse

                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                   logarl                                                                                                                                                  unismmm                                                                                                                                                             Partial                                                                                                                                                         atiststJ                                                                          rumtsni
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                         tura                                                                                                                                                     co
                                                                                                                                                                    )                          ent
                                                                                                                                                                     og(le                                                                                                                                                                                                                                                                                                                                   naeth                                                                                                                                                  to
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                      ior                                                                                                                                                             ests.-tt                                                                                                                                                                                                                                          udedl
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                       Hansen's                                                                              exc

                                                                                                                                                                          ablri
                                                                                                                                                                               va                 oymplmenu                                                                   TransMONEE-                                                                    ym                                                                                                                                                                                                                                                                                                          pry                                                                                                                                                                                                                                                                                                                                                                                                                    the
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                  odels,m                                                                                                                                                   untr                                                                                                                                                                                                                                                                                                                               robust

                                                                                                                                                                                                                                                                                                                                                               econo                                                                                                                                                     all                                                                                                                                                    coeth                                                                                                                                                         two-sided                                                                                                                                                                                                                                            atth

                                                                                                                                                                                 entdn                                                                                                                                                                              al                                                                                                                                                                                                                                                                                                                                                                                                                                                                                 from                                                                                                                                                                                                                                            and



                                                                                                                                                                                      Depe                 eredstgieR                                                                                                                                                                                                                                                                                                       For                                                                                                                                                      in                                                                                                                                                                                                                                                                                                                              cluster-

                                                                                                                                                                                                                                                                                         pl-to-pop                                                                    rm                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                  rmet
                                                                                                                                                                                                                                                                                                                                                                        fo
                                                                                                                                                                                                                                                                                                  Em                                                                      In                                                                                                                                                   Notes:                                                                                                                                                  existed                                                                                                                                                             P-values                                                                                                                                                          column,                                                                          error

    1990-2003                                             earsy

                                                               ousi

             spending,


                      health                                       var,srtopeRht
49                          total
                                 of                                             Healdrl


                                   sharea                                              Wodna
                                                                                            Ts
                                         asIHS                                                Hi

                                                                                                urce:
                                              1:                                                     So


                                                Figure

     1990-2004


              thods,em

50                    ent                        HiTs


                         paym                        Source:



                             Hospital
                                     2:


                                       Figure

    1990-2003


                                           HiTs
51

             classification,IHS                Source:


                               3:


                                 Figure

                                                                                                                     ps gnidneps htlaeH

                                                                                                     0          0       0           0         0       0
    (1990-2004)                                                                                       60         50      40          30        20      10     0

                                                                                                                                                                  04

               ountriesc                                                                                                                                            03                                 00)
                        ECA                                                                                                                                                                               20
                                                                                                                                                                      02
                           in                                                                                                                                                                               PPP,
                                                                                                                                                                        01
                             capitar
                                    pe                                                                                                                                    00                                    (US$g

                                                                                                                                                                            99                                       pendins


                                      expenditure                                                                                                                             98                                            Health

52                                                                                                                                                                              97
                                                 health                                                                                                                                          Year
                                                                                                                                                                                                                                  d)e
                                                                                                                                                                                  96

                                                       average                                                                                                                      95
                                                              and                                                                                                                                                                    accumulat

                                                                 on                                                                                                                   49                                                      %(

                                                                                                                                                                                                                                                ries
                                                                   adopti                                                                                                               93
                                                                                                                                                                                                                                                    count
                                                                         SHI                                                                                                              92                                                             HIS
                                                                            of
                                                                                                                                                                                            91


                                                                              evolution                                                                                                       90

                                                                                       The
                                                                                          4:            70%  60%     50%s seir40% tnuoc 30%IHS   20%     10%   0%



                                                                                            Figure

                                                                                                                  m ytilatrom tnafnI

                                                                                                 25       20         15               10       5       0


                                                                                                                                                           04
    (1990-2004)
                                                                                                                                                             03                                 )OH
                                                                                                                                                                                                   W-

               countries                                                                                                                                       02                                    hstrbi
                        ECA                                                                                                                                      01                                        000
                           in                                                                                                                                                                                 1,r
                             rate                                                                                                                                  00                                            pe(


                                                                                                                                                                     99

                                 mortality
                                                                                                                                                                       98                                           etarytilatro
                                                                                                                                                                                                                                m

                                          infant                                                                                                                                                                                 nta
53                                                                                                                                                                       97               Year                                      nfI


                                                average                                                                                                                    96                                                          )d
                                                       and
                                                                                                                                                                             95                                                          tealu



                                                          adoption                                                                                                             94                                                             mucca


                                                                  SHI                                                                                                            93
                                                                     of

                                                                                                                                                                                   92                                                              %(seirtnuocI
                                                                                                                                                                                                                                                               SH
                                                                       evolution                                                                                                     91

                                                                                The
                                                                                   5:                                                                                                  90



                                                                                     Figure   70%     60%     50% ni se40%irtnuoc IHS
                                                                                                                                   30%    20%    10%    0%

                                                                                                                    tnemyolpmenU

                                                                                           %        %        %
                                                                                            14       12       10        8%      6%      4%      2%      0%


                                                                                                                                                             04

    (1990-2004)                                                                                                                                                30



               countries                                                                                                                                         02
                                                                                                                                                                                                  )
                                                                                                                                                                                                   %(
                        ECA                                                                                                                                        01                                e
                           in
                             tera                                                                                                                                    00                               atrtne
                                                                                                                                                                                                            m

                                 ent                                                                                                                                                                         oypl
                                                                                                                                                                       99
                                    loymp                                                                                                                                                                        emnU
                                                                                                                                                                         89

                                         unem                                                                                                                                               ra                       )d
54                                                                                                                                                                         97                 Ye                       tealu
                                             average                                                                                                                         96
                                                    and                                                                                                                                                                     mucca

                                                                                                                                                                               59                                                %(


                                                       doptiona                                                                                                                  94

                                                               SHI
                                                                  of                                                                                                               39                                              seitrnuocIHS



                                                                                                                                                                                     92

                                                                    evolution
                                                                                                                                                                                       91
                                                                             The
                                                                                6:                                                                                                       09


                                                                                  Figure      %        %        %         %       %       %       %
                                                                                               70       60       50        40      30      20      10     0%
                                                                                                                     seirtnuoc IHS

                                                                                                                    ymonoce lamrofnI

                                                                                              %       %       %          %       %       %
                                                                                               35      30      25         20      15      10     5%     0%
    (1990-2003)
                                                                                                                                                             30


               countries                                                                                                                                       02                               )PDG
                        ECA                                                                                                                                      01
                           in                                                                                                                                                                       altot

                             sizey                                                                                                                                 00                                    %(
                                                                                                                                                                                                           y
                                                                                                                                                                                                            om
                                                                                                                                                                     99                                       on
                                                                                                                                                                                                                ec
                                  econom                                                                                                                                                                          al
                                        al                                                                                                                             98                                           m
                                                                                                                                                                                                                     ornfI
                                                                                                                                                                         97
55                                        inform                                                                                                                                          ra

                                                                                                                                                                           69               Ye                            )d

                                                average                                                                                                                                                                     tealu
                                                                                                                                                                             59
                                                       and                                                                                                                                                                       mucca
                                                                                                                                                                               94                                                     %(

                                                          doptiona                                                                                                               93

                                                                  SHI
                                                                     of                                                                                                            29                                                   seitrnuocIHS

                                                                                                                                                                                     19

                                                                       evolution                                                                                                       90

                                                                                The
                                                                                   7:            %       %       %          %       %       %      %
                                                                                                  70      60      50         40      30      20     10    0%
                                                                                                                     seirtnuoc IHS

                                                                                     Figure

                                                      56


                                                  APPENDIX

               Table 9: Health sector outcome variables: definitions and sources

Variable                        Definition                                         Source
Total health expenditure per Sum of General Government and of Private              Own calculations, based
capita (constant 2000           Expenditure on Health. Estimates for this indicatoron the following sources:
international US$)              were produced by WHO. Data are in constant         1990-1997: WDI 2002
                                2000 international dollars (deflated using the US database (Serbia and
                                GDP deflator).                                     Montenegro =
                                                                                   Yugoslavia, Fed.). From
                                                                                   1998 onwards: WHO
                                                                                   estimates, HFA-DB.
Government health               General Government Expenditure on Health.          See above.
expenditure per capita          Estimates for this indicator were produced by
(constant 2000 international    WHO. Data are in constant 2000 international
US$)                            dollars (deflated using the US GDP deflator).
Private health expenditure per Private Expenditure on Health. Estimates for this See above.
capita (constant 2000           indicator were produced by WHO. Data are in
international US$)              constant 2000 international dollars (deflated using
                                the US GDP deflator).
Salaries as percentage of total Includes salaries, bonuses to fixed rate wages and WHO, HFA-DB.
government health               salaries, and overtime payments to employees in
expenditure                     the publicly financed health sector.
Physicians (per 1,000 people) Physicians are defined as graduates of any facility WDI Database - DDP,
                               or school of medicine who are working in the        based on data from
                               country in any medical field (practice, teaching,   World Health
                               research).                                          Organization, OECD,
                                                                                   TransMONEE,
                                                                                   supplemented by country
                                                                                   data.
Life expectancy at birth, in   Calculated by WHO/EURO for all countries            WHO, HFA-DB.
years                          which report detailed mortality data to WHO,
                               using Wiesler's method.
Life expectancy at birth, in   Calculated by WHO/EURO for all countries            WHO, HFA-DB.
years, male                    which report detailed mortality data to WHO,
                               using Wiesler's method.
Life expectancy at birth, in   Calculated by WHO/EURO for all countries            WHO, HFA-DB.
years, female                  which report detailed mortality data to WHO,
                               using Wiesler's method.
Mortality rate, under-5 (per   Probability of dying before age 5 years per 1000    TransMONEE 2006
1,000) ­ TransMONEE            live births, calculated as the number of deaths per Database, UNICEF IRC,
                               1000 live births until 5 years of age.              Florence.
Mortality rate, under-5 (per   See above.                                          WHO, HFA-DB.
1,000) ­ WHO
Mortality rate, infant (per    Number of infants dying before reaching one year WDI-DDP database.
1,000 live births) - World     of age, per 1,000 live births in a given year.
Bank                           Harmonized estimates of the World Health
                               Organization, UNICEF, and the World Bank,
                               based mainly on household surveys, censuses, and
                               vital registration, supplemented by World Bank
                               estimates based on household surveys and vital
                               registration.

                                                       57


Variable                        Definition                                           Source
Mortality rate, infant (per     Number of infants dying before reaching one year TransMONEE 2006
1,000 live births) ­            of age, per 1,000 live births in a given year.       Database, UNICEF IRC,
TransMONEE                                                                           Florence.
Mortality rate, infant (per     See above.                                           WHO, HFA-DB.
1,000 live births) - WHO
Perinatal mortality rate (per   Weight specific (1000 g +) fetal deaths and early WHO, HFA-DB.
1,000 births)                   neonatal deaths per 1000 births (live
                                births+stillbirths)
Neonatal mortality rate (per    Number of deaths in infants under 28 days of age WHO, HFA-DB.
1,000 live births)              in a year, per 1000 live births in that year (ICD-
                                10).
Postneonatal mortality rate     Number of deaths in infants between 4 weeks and WHO, HFA-DB.
(per 1,000 live births)         a year of age in a year, per 1000 live births in that
                                year (ICD-10).
Maternal mortality rate (per    Maternal deaths per 100,000 live births. A           WHO, HFA-DB.
100,000 live births)            maternal death is death of a woman while
                                pregnant or within 42 days of termination of
                                pregnancy, irrespective of the duration and site of
                                the pregnancy, from any cause related to or
                                aggravated by the pregnancy or its management,
                                but not from accidental or incidental causes.
Maternal mortality rate (per    Moving average (3 years) of maternal deaths per      WHO, HFA-DB.
100,000 live births) - moving 100,000 live births.
average (3 years)
Caesarean sections (per 1,000 Number of caesarean sections per 1000 live births. WHO, HFA-DB.
live births)
Death rate, all causes (per     Standardized death rate, all causes, all ages, per   WHO, HFA-DB.
100,000)                        100,000.
Death rate, infectious diseases Standardized death rate, infectious and parasitic    WHO, HFA-DB.
(per 100,000)                   diseases, all ages, per 100,000.
Death rate, tuberculosis (per Standardized death rate, tuberculosis, all ages, per WHO, HFA-DB.
100,000)                        100,000.
Death rate, diarrhoeal          Standardized death rate, diarrhoeal diseases, under WHO, HFA-DB.
diseases, under five years (per 5 years per 100,000.
100,000)
Death rate, ARI, under five     Standardized death rate, acute respiratory           WHO, HFA-DB.
years (per 100,000)             infections, under 5 years per 100,000.
Death rate, ischaemic heart     Standardized death rate, ischaemic heart disease, WHO, HFA-DB.
disease (per 100,000)           all ages per 100,000.
Death rate, liver diseases (per Standardized death rate, chronic liver disease and WHO, HFA-DB.
100,000)                        cirrhosis, all ages per 100,000.
Death rate, diabetes (per      Standardized death rate, diabetes, all ages, per      WHO, HFA-DB.
100,000)                       100,000.
Death rate, circulatory        Standardized death rate, diseases of circulatory      WHO, HFA-DB.
diseases (per 100,000)         system, all ages, per 100,000.
Death rate, cerebrovascular    Standardized death rate, cerebrovascular diseases, WHO, HFA-DB.
diseases (per 100,000)         all ages, per 100,000.
Death rate, neoplasms (per     Standardized death rate, malignant neoplasms, all WHO, HFA-DB.
100,000)                       ages, per 100,000.

                                                     58


Variable                       Definition                                          Source
Death rate, breast cancer (per Standardized death rate, malignant neoplasm          WHO, HFA-DB.
100,000)                       female breast, all ages, per 100,000.
Death rate, respiratory        Standardized death rate, diseases of the respiratory WHO, HFA-DB.
diseases (per 100,000)         system, all ages, per 100,000.
Death rate,                    Standardized death rate,                            WHO, HFA-DB.
bronchitis/emphysema/asthmabronchitis/emphysema/asthma, all ages, per
(per 100,000)                  100,000.
Death rate, digestive diseases Standardized death rate, diseases of the digestive WHO, HFA-DB.
(per 100,000)                  system, all ages, per 100,000.
Death rate, alcohol causes    Standardized death rate, selected alcohol related    WHO, HFA-DB.
(per 100,000)                 causes, per 100,000. The mortality from
                              combined, selected causes of death which are
                              known from literature to be related to alcohol
                              consumption. Includes: cancer of oesophagus and
                              larynx; alcohol dependence syndrom; chronic liver
                              disease and cirrhosis; all external causes.
Death rate, smoking causes    Standardized death rate, selected smoking related WHO, HFA-DB.
(per 100,000)                 causes, per 100,000. The mortality from
                              combined, selected causes of death which are
                              known from literature to be related to smoking.
                              Includes: cancers of mouth and pharynx, larynx,
                              traxea, bronchus, lung and oesophagus; ischaemic
                              heart disease; cerebrovascular diseases; chronic
                              obstructive pulmonary disease.
Tuberculosis incidence rate   Tuberculosis incidence per 100,000. Number of        WHO, HFA-DB.
(per 100,000)                 newly diagnosed tuberculosis cases, all forms
                              (ICD-9:010-
                               018; ICD-10: A15-A19) during the given
                              calendar year.
Hepatitis incidence rate (per Viral hepatitis incidence per 100,000.               WHO, HFA-DB.
100,000)
Hepatitis B incidence rate    Viral hepatitis B incidence per 100,000.             WHO, HFA-DB.
(per 100,000)
Measles incidence rate (per   Measles incidence per 100,000.                       WHO, HFA-DB.
100,000)
Mumps incidence rate (per     Mumps incidence per 100,000.                         WHO, HFA-DB.
100,000)
Syphilis incidence rate (per  Syphilis incidence per 100,000.                       WHO, HFA-DB.
100,000)
Congenital syphilis incidence Congenital syphilis incidence per 100,000.            WHO, HFA-DB.
rate (per 100,000)
Pertussis incidence rate (per Pertussis incidence per 100,000.                     WHO, HFA-DB.
100,000)
Diphteria incidence rate (per Diphteria incidence per 100,000.                     WHO, HFA-DB.
100,000)
Tetanus incidence rate (per   Tetanus incidence per 100,000.                        WHO, HFA-DB.
100,000)
Cancer incidence rate (per    Cancer incidence per 100,000. Number of patients WHO, HFA-DB.
100,000)                      with newly diagnosed cancer during given
                              calendar year.

                                                    59


Variable                      Definition                                         Source
Immunization rate,            Percentage of infants reaching their first birthday WHO, HFA-DB.
tuberculosis, infants (%)     in the given calendar year who have been fully
                              vaccinated against tuberculosis (BCG, 1 dose).
Immunization rate, DPT,       Percentage of children under 2 immunized against TransMONEE 2006
infants (%)                   diphtheria, pertussis and tetanus.                  Database, UNICEF IRC,
                                                                                  Florence.
Immunization rate, measles, Percentage of children reaching their second          WHO, HFA-DB.
infants (%)                   birthday who have been fully vaccinated against
                              measles (1 dose).
Immunization rate,            Percentage of infants reaching their first birthday WHO, HFA-DB.
poliomyelitis, infants (%)    in the given calendar year who have been fully
                              vaccinated against poliomyelitis (3 doses).
Immunization rate, mumps,     Percentage of infants reaching their second        WHO, HFA-DB.
infants (%)                   birthday in the given calendar year who have been
                              fully vaccinated against mumps.
Immunization rate, rubella,   Percentage of infants reaching their second        WHO, HFA-DB.
infants (%)                   birthday in the given calendar year who have been
                              fully vaccinated against rubella.
Average length of stay, all   Total number of occupied hospital bed-days of all WHO, HFA-DB.
hospitals                     hospitals divided by the total number of
                              admissions or discharges in those hospitals.
                              Length of stay (LOS) of one patient = date of
                              discharge - date of admission. If these are the
                              same dates, then LOS is set to one day. Bed-days
                              of newborns are excluded in the calculation.
Average length of stay, acute Total number of occupied hospital bed-days of      WHO, HFA-DB.
care hospitals only           short-stay hospitals divided by the total number of
                              admissions or discharges in those hospitals.
Bed occupancy rate (%),      Average number of days when hospital bed was        WHO, HFA-DB.
acute care hospitals only    occupied as percentage of available 365 days.
                             Calculation: utilized bed-days x 100/available
                             bed-days during the calendar year.

Hospital beds (per 1,000     Hospital beds include in-patient beds available in WDI Database - DDP,
people)                      public, private, general, and specialized hospitals based on data from
                             and rehabilitation centers. In most cases beds for World Health
                             both acute and chronic care are included.           Organization, OECD,
                                                                                 TransMONEE,
                                                                                 supplemented by country
                                                                                 data.
In-patient care admissions   Admission is the hospitalization of a patient in an WHO, HFA-DB.
(per 100)                    in-patient facility normally involving a stay of at
                             least 24 hours. In the case of death or discharge to
                             another health establishment, the actual stay may
                             be shorter than 24 hours. These cases are
                             registered as a one-day hospitalization. The
                             number of admissions excludes: a transfer from
                             one department to another one at the same
                             hospital; day-cases of day patients; weekend leave
                             when the patient has been released temporarily
                             and the hospital bed is still reserved; cases where
                             treatment is provided by hospital personnel at the
                             patient's home. Newborns are not included.

                                                      60


Variable                       Definition                                          Source
Acute care hospital            Same as above, short stay hospitals only.           WHO, HFA-DB.
admissions (per 100)
Hospital discharges,           Total number of patients discharged from all        WHO, HFA-DB.
infectious and parasitic       hospitals during the given calendar year with the
diseases (per 100,000)         principal diagnosis falling into the group of
                               infectious and parasitic diseases (Chapter I of
                               ICD-9/10). Discharge is the conclusion of a period
                               of in-patient care, whether the patient returned to
                               his home, was transferred to another in-patient
                               facility or died.
Hospital discharges, all       Total number of patients discharged from all        WHO, HFA-DB.
cancers (per 100,000)          hospitals during the given calendar year with the
                               principal diagnosis falling into the group of
                               cancers (Chapter II of ICD-9/10).
Hospital discharges,           Total number of patients discharged from all        WHO, HFA-DB.
ischaemic heart disease (per   hospitals during the given calendar year with the
100,000)                       principal diagnosis falling into the group of
                               ischaemic heart diseases (ICD-10: I20-I25).

Hospital discharges,           Total number of patients discharged from all        WHO, HFA-DB.
circulatory system diseases    hospitals during the given calendar year with the
(per 100,000)                  principal diagnosis falling into the group of
                               circulatory system diseases (Chapter IX of ICD-
                               10).
Hospital discharges,           Total number of patients discharged from all        WHO, HFA-DB.
cerebrovascular diseases (per hospitals during the given calendar year with the
100,000)                       principal diagnosis falling into the group of
                               cerebrovascular diseases (ICD-10: I60-I69).

Hospital discharges,           Total number of patients discharged from all        WHO, HFA-DB.
respiratory diseases (per      hospitals during the given calendar year with the
100,000)                       principal diagnosis falling into the group of
                               respiratory diseases (Chapter X of ICD-10).

Hospital discharges, digestive Total number of patients discharged from all        WHO, HFA-DB.
system diseases (per 100,000) hospitals during the given calendar year with the
                               principal diagnosis falling into the group of
                               digestive system diseases (Chapter XI of ICD-10).

Hospital discharges,           Total number of patients discharged from all        WHO, HFA-DB.
musculoskeletal system and     hospitals during the given calendar year with the
connective tissue diseases     principal diagnosis falling into the group of
(per 100,000)                  musculoskeletal system and connective tissue
                               diseases (Chapter XIII of ICD-10).
Death rate, appendicitis (per Standardized death rate, appendicitis, all ages, per WHO, HFA-DB.
100,000)                       100,000.
Death rate, hernia and         Standardized death rate, hernia and intestinal     WHO, HFA-DB.
intestinal obstruction (per    obstruction, all ages, per 100,000.
100,000)
Death rate, adverse effects of Standardized death rate, adverse effects of        WHO, HFA-DB.
therapeutic agents (per        therapeutic agents, all ages, per 100,000.
100,000)
Surgical infection rate, all  Average rate of in-patient surgical operations in   WHO, HFA-DB.
operations (%)                all hospitals with postoperative surgical wound
                              infection during the given calendar year (ICD-10:

                         61


Variable    Definition         Source
            T81.4)

                                                     62


                Table 10: Labor sector outcome variables: definitions and sources

Variable                   Definition                                             Source
Average wage rate, gross Total annual gross wages and salaries in constant        Own calculations based on
(constant 2000             PPP averages for the employed population aged 15- data from WB-WDI and
international dollars)     59.                                                    TransMONEE databases.
Unemployment, total (% Share of the labor force that is without work but          ILO - LABORSTA, KILM
of total labor force)      available for and seeking employment. National         database.
                           rates based on labor force sample surveys.

Registered                 Number of persons looking for work who are             ILO - LABORSTA, KILM
unemployment, total (%     entered on the registers at the end of each month.     database.
of total labor force)

Employment-to-             Proportion of a country's working-age population       ILO - LABORSTA, KILM
population ratio (%) - ILO that is employed. Original sources are labor force     database.
                           surveys and official estimates available mainly at
                           LABORSTA and OECD.

Employment-to-             Number of employed as per cent of population aged TransMONEE 2006
population ratio (%) -     15-59. Original sources are labor force surveys, ILO Database, UNICEF IRC,
TransMONEE                 and CIS Stat.                                          Florence.

Informal economy as a      Informal GDP as a percentage of the total (formal + Own calculations based on
share of the total GDP (%) informal) GDP. Annual estimates are based on the       initial values for the size of
- Own calculations         growth rate of the national electricity output as a    the informal economy
                           proxy for total GDP growth, which are compared to provided by Johnson et al.
                           the growth rate of the official (measured) GDP to      (1997) and WDI-WB
                           give an estimate of the evolution of the size of the   database.
                           informal economy.

Self-employment (% of      Self-employed individuals as a share of total          ILO - LABORSTA, KILM
total employment)          employment. The self-employed category is defined database.
                           as those individuals who are either own-account
                           workers (a person who operates his or her own
                           economic enterprise, or engages independently in a
                           profession or trade, and hires no employees) or
                           members of producers' cooperatives (a person who
                           is an active member of a producers' cooperative,
                           regardless of the industry in which it is established).


Employment in             Number of employees working in the agricultural         ILO - LABORSTA, KILM
agriculture (% of total   sector as a share of total employment. Employees        database.
employment)               are people who work for a public or private
                          employer and receive remuneration in wages,
                          salary, commission, tips, piece rates, or pay in kind.

                                                      63


            Table 11: Covariates included in the X-vector: definitions and sources

Variable                     Definition                                                Source
GDP per capita (constant     GDP per capita based on purchasing power parity (PPP).    WB-WDI
2000 international           Data are in constant 2000 international dollars (deflated Database.
dollars)                     using the US GDP deflator).
Population ages 65 and       Percentage of the total population that is 65 or older.   WB-WDI
above (% of total)*                                                                    Database.
Urban population (% of       Urban population is the midyear population of areas       WB-WDI
total)*                      defined as urban in each country and reported to the      Database.
                             United Nations.
* Excluded from labor models.

                                                 64


              Table 12: Classification of hospital payment methods in the dataset

                                               Hospital Payment Methods        Values assumed by the
                                                                            corresponding dummies in the
Country                   SHI adoption      Predominant method    Years               dataset
Albania                       1995                 B                              B=1 throughout
Armenia                       Never                B                              B=1 throughout
Azerbaijan                    Never                B                              B=1 throughout
Belarus                       Never                 ?                               all missing
Bosnia and Herzegovina    Prior to 1990            B                              B=1 throughout
Bulgaria                      1999                 B            Until 1998        B=1 (until 1998)
                                                  PBP         1999 onwards     PBP=1 (1999 onwards)
Croatia                   Prior to 1990            B            Until 1992        B=1 (until 1992)
                                                  FFS         1993 onwards     FFS=1 (1993 onwards)
Czech Republic                1993                 B            Until 1992        B=1 (until 1992)
                                                  FFS           1993-1997        FFS=1 (1993-1997)
                                                   B          1998 onwards      B=1 (1998 onwards)
Estonia                       1992                 B            Until 1991        B=1 (until 1991)
                                                  FFS           Until 2003       FFS=1 (1992-2003)
                                                  PBP         2004 onwards     PBP=1 (2004 onwards)
Georgia                       1995                 B            Until 1995        B=1 (until 1995)
                                                  PBP         1996 onwards     PBP=1 (1996 onwards)
Hungary                       1990                 B            Until 1992        B=1 (until 1992)
                                                  PBP         1993 onwards     PBP=1 (1993 onwards)
Kazakhstan             1996 (scrapped 1998)        B            Until 1995        B=1 (until 1995)
                                                  PBP         1996 onwards     PBP=1 (1996 onwards)
Kyrgyz Republic               1997                 B            Until 1996        B=1 (until 1996)
                                                  PBP         1997 onwards     PBP=1 (1997 onwards)
Latvia                        Never                B            Until 1993        B=1 (until 1993)
                                                  FFS           1994-1997        FFS=1 (1994-1997)
                                                  PBP         1998 onwards     PBP=1 (1998 onwards)
Lithuania                     1991                 B            Until 1993        B=1 (until 1993)
                                                  PBP         1994 onwards     PBP=1 (1994 onwards)
Macedonia, FYR                1991                 B            Until 1990        B=1 (until 1990)
                                                  FFS         1991 onwards      FFS (1991 onwards)
Moldova                       Never                B                              B=1 throughout
Poland                        Never                B            Until 1998        B=1 (until 1998)
                                                  PBP         1999 onwards     PBP=1 (1999 onwards)
Romania                       1998                 B                              B=1 throughout
Russian Federation            1993                 B                              B=1 throughout
Serbia and Montenegro     Prior to 1990             ?                               all missing
Slovak Republic               1995                 B            Until 1993        B=1 (until 1993)
                                                  FFS           1994-1998        FFS=1 (1994-1998)
                                                   B            1999-2001         B=1 (1999-2001)
                                                  PBP         2002 onwards     PBP=1 (2002 onwards)
Slovenia                      1992                 B            Until 1991        B=1 (until 1991)
                                                  PBP          1992 onwards    PBP=1 (1992 onwards)
Tajikistan                    Never                B                              B=1 throughout
Turkey                    Prior to 1990            B                              B=1 throughout
Turkmenistan                  Never                B                              B=1 throughout
Ukraine                       Never                B                              B=1 throughout

                                                                    65


                                                                Hospital Payment Methods                 Values assumed by the
                                                                                                      corresponding dummies in the
Country                            SHI adoption             Predominant method         Years                     dataset
Uzbekistan                             Never                          B                                      B=1 throughout
Notes: B = fixed budget/block grants; FFS = fee-for-service/payment by bed days; PBP = patient-based payment method.

                                            66



References

Alam, A., M. Murthi, R. Yemtsov, E. Murrugarra, N. Dudwick, E. Hamilton and E.
       Tiongson (2005). Growth, poverty, and inequality : Eastern Europe and the
       former Soviet Union. Washington, DC, World Bank.
Allin, S., E. Mossialos, M. McKee and W. Holland (2004). Making decisions on public
       health: a review of eight countries. Copenhagen, World Health Organization.
Anderson, T. W. and C. Hsiao (1981). "Estimation of dynamic models with error
       components." Journal of American Statistical Association 76: 598-606.
Baeza, C. and T. Packard (2006). Beyond Survival: Protecting Households from Health
       Shocks in Latin America. Washington, DC, World Bank.
Bauer, T. and R. T. Riphahn (2002). "Employment Effects of Payroll Taxes--An
       Empirical Test for Germany." Applied Economics 34(7): 865-76.
Bertrand, M., E. Duflo and S. Mullainathan (2004). "How Much Should We Trust
       Differences-in-Differences Estimates?" Quarterly Journal of Economics 119 1:
       249-75.
Cameron, A. and R. Trivedi (2005). Microeconometrics: Methods and Applications.
Carrin, G. and C. James (2004). Reaching universal coverage via social health insurance.
       Geneva, WHO, Health System Financing, Expenditure and Resource Allocation
       Discussion Paper #2.
Ellis, R. and M. Miller (2008). Provider payment methods and incentives. Encyclopedia
       of Public Health. K. Heggenhougen, Blackwell.
Escobar, M.-L. and P. Panopolou (2003). Health. Colombia : the economic foundation of
       peace. M. Giugale, O. Lafourcade and C. Luff. Washington, DC, World Bank:
       653-707.
Friedberg, L. (1998). "Did Unilateral Divorce Raise Divorce Rates? Evidence from Panel
       Data." American Economic Review 88(3): 608-627.
Gerdtham, U.-G. and B. Jonsson (1992). "International Comparisons of Health Care
       Expenditure--Conversion Factor Instability, Heteroscedasticity, Outliers and
       Robust Estimators." Journal of Health Economics 11 2: 189-97.
Gertler, P. and O. Solon (2000). Who Benefits From Social Health Insurance in
       Developing Countries? Berkeley, CA, University of California at Berkeley
Gottret, P. E. and G. Schieber (2006). Health financing revisited : a practitioner's guide.
       Washington, DC, World Bank.
Gruber, J. and M. Hanratty (1995). "The Labor-Market Effects of Introducing National
       Health Insurance: Evidence from Canada." Journal of Business and Economic
       Statistics 13(2): 163-73.
International Network on Health Policy & Reform (2006). Health Policy Monitor,
       Bertelsmann Stiftung, Gütersloh.
Johnson, S., D. Kaufmann and A. Shleifer (1997). "The Unofficial Economy in
       Transition." Brookings Papers on Economic Activity 2: 159-221.
Kaufmann, D. and A. Kaliberda (1996). Integrating the unofficial economy into the
       dynamics of post-socialist economies : a framework of analysis and evidence.
       Economic Transition in the Newly Independent States. B. Kaminsky. New York,
       ME Sharpe Press.

                                            67


Kugler, A. and M. Kugler (2003). The Labour Market Effects of Payroll Taxes in a
       Middle-Income Country: Evidence from Colombia, C.E.P.R. Discussion Papers,
       CEPR Discussion Papers: 4046.
Langenbrunner, J. C., I. M. Sheiman and J. C. Kehler (in press). Sources of Health
       Financing and Revenue Collection: Reforms, Lessons, and Challenges. Lessons of
       Health Reforms in Eastern Europe and Central Asia. J. Kutzin and e. al. Brussels,
       European Observatory.
Lu, J. R. and W. C. Hsiao (2003). "Does Universal Insurance Make Health Care
       Unaffordable? Lessons from Taiwan." Health Affairs 22(3 (May/June 2003)): 77-
       88.
O'Donnell, O., E. Van Doorslaer, R. Rannan-Eliya, A. Somanathan, S. R. Adhikari, B.
       Akkazieva, D. Harbianto, C. G. Garg, P. Hanvoravongchai, A. N. Herrin, M. N.
       Huq, S. Ibragimova, A. Karan, S.-M. Kwon, G. M. Leung, J.-F. R. Lu, Y. Ohkusa,
       B. R. Pande, R. Racelis, K. Tin, L. Trisnantoro, C. Vasavid, Q. Wan, B.-M. Yang
       and Y. Zhao (2007). "Who pays for health care in Asia?" Journal of Health
       Economics in press.
Preker, A., M. Jakab and M. Schneider (2002). Health financing reforms in central and
       eastern Europe and the former Soviet Union. Funding health care: options for
       Europe. A. D. E. Mossialos, J. Figueras and J. Kutzin,. Buckingham, Open
       University Press.
Rechel, B. and M. McKee (2003). Healing the crisis. A prescription for public health
       action in South-Eastern Europe. London, London School of Hygiene and Tropical
       Medicine.
Stock, J. H. and M. Yogo (2002). Testing for weak instruments in linear IV regression.
       Cambridge, Mass., National Bureau of Economic Research, NBER, Technical
       Working Paper # T 0284.
Svejnar, J. (1999). Labor markets in the transitional Central and East European
       economies. Handbook of Labor Economics, Vol. 3. O. Ashenfelter and D. Card.
       Amsterdam, Elsevier.
Twigg, J. L. (1999). "Obligatory medical insurance in Russia: the participants'
       perspective." Social Science & Medicine 49(3): 382.
Wagstaff, A. (2007). Social health insurance reexamined. Washington DC, World Bank,
       Policy Research Working Paper # WPS 4111.
Wagstaff, A. and M. Lindelow (2005). Can insurance increase financial risk? : the
       curious case of health insurance in China. Washington, D.C., World Bank, Policy
       Research Working Paper #3741.
Wagstaff, A., E. van Doorslaer, S. Calonge, T. Christiansen, M. Gerfin, P. Gottschalk, R.
       Janssen, C. Lachaud, R. Leu and B. Nolan (1992). "Equity in the finance of health
       care: some international comparisons." Journal of Health Economics 11(4): 361-
       388.
Wagstaff, A., E. van Doorslaer, H. van der Burg, S. Calonge, T. Christiansen, G. Citoni,
       U. G. Gerdtham, M. Gerfin, L. Gross, U. Hakinnen, P. Johnson, J. John, J. Klavus,
       C. Lachaud, J. Lauritsen, R. Leu, B. Nolan, E. Peran, J. Pereira, C. Propper, F.
       Puffer, L. Rochaix, M. Rodriguez, M. Schellhorn, O. Winkelhake and et al.
       (1999). "Equity in the finance of health care: some further international
       comparisons." J Health Econ 18(3): 263-90.

                                           68


Waters, H. R., G. F. Anderson and J. Mays (2004). "Measuring financial protection in
      health in the United States." Health Policy 69(3): 339-49.
Wooldridge, J. M. (2002). Econometric analysis of cross section and panel data.
      Cambridge, Mass., MIT Press.
World Bank (1997). Argentina: Facing the Challenge of Health Insurance Reform.
      Washington DC, World Bank, Report No. 16402-AR.