Estimation of a health production function: evidence from East-European countries.
The countries of Eastern Europe have experienced extraordinary changes since the end of the 1980's when the socialist regimes were replaced by market-based economies. Many transformations have taken place during the last two decades and sizable improvements can be seen in different areas, but at the same time, these countries continue to face new challenges in the process of their transitions to the market economy and their efforts of integration into the western society. Even though Eastern European countries tend to have similar aspirations and problems, one cannot overlook the differences among them in terms of population size, level of income, level of development, and other social and economic characteristics.
The health care system is part of the overall reform agenda and this is no surprise considering its importance for the social wellbeing and its contribution to public health capital formation and economic growth. In one of its reports, the Commission on Macroeconomics and Health of the World Health Organization (WHO) estimates that a "10 percent increase in life expectancy at birth increases economic growth by at least 0.3-0.4 percent of gross domestic product per year" (Iliev and Suhrcke 2006). An important point to be noted here is that the changes and improvements of the health care system in the Eastern European countries are ongoing processes. During the communist regime, the health care system was centrally planned and administered; now, a more decentralized system is in place with more private providers and different forms of funding health care services (Rechel and McKee 2009).
Even though important progress has been made, most of the Eastern European countries still have a long way to reach the western countries' health care levels. The life expectancy at birth is rising in all countries at a different pace, but it is still below the level of western countries. For example, the lowest levels of life expectancy at birth in the European Union were in Romania (76.2 years) for women and in Lithuania (65.1 years) for men (OECD 2010). The infant mortality indicator is a mirror image of life expectancy, with higher rates for the newer members. On one side of the spectrum is Luxemburg with the lowest infant mortality rate of 1.8 per 1000 live births while on the other side are Romania and Bulgaria with 11 and 8.6 percent, respectively (OECD 2010). There is no surprise that the lowest level of health care expenditures as a share of gross domestic product is held by an Eastern European country. In 2008, Romania spent 6 percent of its GDP compared to Germany, Austria, Switzerland, and France which allocated more than 10 percent of their GDP (OECD 2010). The European Union has made efforts to implement policies that will help new members attain not only the economic status of the veteran states, but also the social, wealth, and national health levels.
Considering the importance of health care, this paper investigates the impact of the different economic, social, and environmental factors on the health status of member states in the Eastern European Bloc. Analyzing the health production function at the macro level can offer significant insights into determining the most efficient way of allocating resources for improving the overall health status of countries in the sample. Knowing the degree to which every factor contributes to the improvement of health status could help authorities to make decisions and design more appropriate policies with greater impact. The objective of this study is to estimate the impact of the macroeconomic, demographic, environmental, and life-style factors on health status (measured by infant mortality) for East European countries using panel data from 1997 to 2005.
The study is organized as follows. Section II provides a brief review of the literature. The model and data are discussed in Section III. The results and interpretations of the study are presented in Section IV. The last section summarizes the findings, draws conclusions, and makes some policy recommendations based on the results.
II. Literature Review
This paper is based on Grossman's (1972) seminal work of human capital model of demand for health which posits that health capital increases the market and non-market productive efficiency of an individual. Grossman's initial framework and its extended models help to explain a series of relations such as the health production function, an individual's demand for health, and an individual's health outcomes determined by various exogenous inputs (Schultz 2004). The importance of the model rests on two aspects, including the influence of health on labor productivity and the distinction between the demand for medical services and health (Grossman 1972; Jacobson 1999). According to Nixon and Ulmann (2006), there are two major approaches to explaining the effects of different inputs on health outcomes. First, health is considered to be a commodity and the individual maximizes its consumption subject to a budget constraint. It is also regarded as a capital good having an initial stock which is subject to accumulation and depreciation over time. In other words, health is simultaneously an investment good which yields satisfaction to consumers indirectly through increased productivity, fewer sick days, and higher wages and a consumption good that yields direct satisfaction or utility. Second, health represents an output that is determined by different inputs such as health care expenditures, demographic, life-style, and environmental factors, or other medical resources.
Even though there is a rich literature that looks at the relationship between the various explanatory variables and health status for many countries and regional economies of Western Europe and North America, there is a dearth of theoretical and empirical research that analyzes the impact of the economic, social, and environmental factors on the health status of Eastern-European countries using recent data. Studies which measure the conditions of the health capital can be of great interest. The Eastern European countries, many of which are just accepted into the European Union, are making efforts with limited resources to align their health systems to those of the Western European countries. Most of the previous studies use life expectancy at birth or mortality (age adjusted for infant, or adult) as dependent variables (Auster et al. 1969; Rosen and Taubman 1982; Berger and Leigh 1989). The array of independent variables ranges from health care expenditures and per capita gross domestic product to cigarette or alcohol consumption covering the economic, social, and environmental factors. Furthermore, Nixon and Ulmann (2006) provide a thorough review of the recent literature with detailed descriptions of existing research related to this paper.
III. Methodology and Data
This paper follows the footprints of a previous paper by Fayissa and Gutema (2005) which adopts Grossman's (1972) model and transposes it from the micro to the macro level. The health outcome measure (infant mortality rate) is specified as a function of the economic (GDP per capita, health care expenditures, education, food production index), social (marital status, population size, alcohol consumption), and environmental (urbanization, carbon dioxide emissions) factors. Thus, a log-linear Cobb-Douglas production function of the study can be written as:
ln [h.sub.i] = ln [OMEGA] + [summation] [[alpha].sub.i] ln [y.sub.i] + [summation][[beta].sub.i] ln [s.sub.i] + [summation] [[gamma].sub.i] ln [e.sub.i] + [[mu].sub.i], (1)
where [OMEGA] is the initial stock of health, [y.sub.i] are the economic factors, [s.sub.i] are the social factors, [e.sub.i] are the environmental factors, and [[mu].sub.i] is the disturbance term.
The analysis is based on country level data from the World Bank, the Eurostat (European Union data bank), the United Nations Development Programme (UNDP), and the World Health Organization (WHO) for 13 Eastern European countries spanning from 1997 to 2005. Other Eastern European countries such as Moldova and the rest of the former members of Yugoslavia have not been included because of data constraints. Most of the studies either use life expectancy at birth or mortality rates as a proxy for health status. According to Nixon and Ulmann (2006), infant mortality is a better measure because it is more strongly related to the health care system and medical procedures. Therefore, we use infant mortality rate as the dependent variable. Because infant mortality is used as a proxy for country's health status, a lower IMR can be interpreted as a better health status for the country. The data are drawn from the Eurostat and they represent the ratio of the number of deaths of children under one year of age per 1000 live-births during a particular year. Infant mortality (IMR) can be divided into two components: neonatal and post-neonatal. According to Rowland (1991), these two categories are determined by different factors. Prematurity, low birth weight, and birth defects can increase the risk of neonatal mortality, while infectious diseases are responsible for post-neonatal mortality. This paper also uses neonatal mortality as a dependent variable to provide additional insights regarding the factors that affect infant mortality. (2) Improvement in neonatal care using "simple, effective, low-cost appropriate technologies is a major investment for any health system" (Uxa et al. 2006). Infant mortality and neonatal mortality have decreased in Eastern European countries during the last two decades. The trends are presented in Chart 1. The reasons are twofold. First, the health care reform has lead to better medical services, and second, the social and economic conditions have considerably improved.
The explanatory variables selected for this paper can be categorized in three subgroups:
--economic factors: gross domestic product (GDP) per capita, total health care expenditures per capita, private health care expenditures, number of physicians per capita, number of hospital beds per capita, and food availability;
--social factors: education, adult alcohol consumption per capita, and population;
--environmental factors: urban population and C[O.sub.2] emission.
Appendix 1 provides the definitions and data sources for the variables in the study. The descriptive statistics are presented in Table 1.
GDP per capita is taken from the World Bank's World Development Indicators CD and it is calculated as the gross domestic product divided by midyear population. In order to maintain the uniformity of the data as much as possible, GDP per capita is used in constant 2000 US$. The causal relation between income and health is ambiguous since income can influence health through better food, better health services, and access to medical care, but better health can also lead to higher income through higher labor productivity, labor supply, or education. On the one side, income has a positive effect on health (Ettner 1996) and economic growth (Bhargava et al. 2001; Favaro et al. 2006). On the other side, the existing literature also posits another issue that money can buy better health up to a point; after that threshold is reached, however, health can be affected adversely by increasing income (Fayissa and Gutema 2005). The analysis of health status in the Eastern European countries does not have to address this problem due to the prevalence of poverty and income inequality in these countries. Thus, the expected sign of the coefficient of income is a priori ambiguous.
Total health expenditure is the sum of public and private expenditures on health. It is provided by the World Bank and includes the provision of health services (preventive and curative), family planning activities, nutrition activities, and emergency aid designated for health, but does not include provision of water and sanitation. According to the OECD (2010), the allocation among different health care services varies from one country to another depending on the availability of resources. Furthermore, the allocation of resources within a country is different depending on region. For example, Romania allocated 167 percent of the national average health care spending per capita to its capital city, Bucharest to the detriment of other regions (Vladescu et. al 2000; Rechel and McKee 2006). According to the European Communities and the WHO (2002), "total health care expenditures as a percentage of gross domestic product is the lowest in Romania, Latvia, and Bulgaria, while Estonia, Poland, Lithuania, Hungary, Slovenia, Slovakia, and the Czech Republic report levels equal to the EU minimum, though below the EU average." The effect of health care expenditures on the health outcome is not clearly defined; it can have a positive and significant effect as long as the allocation of resources to health care is not made to the detriment of investments on healthy living decisions such as better food, or health promoting activities (Fayissa and Gutema 2005 Nixon and Ulmann 2006).
This paper also employs private health expenditure as percent of GDP. Private health expenditure includes direct household (out-of-pocket) spending, private insurance, charitable donations, and direct service payments by private corporations. From the early 90s, the Eastern European countries have started the transition from a centrally-planed healthcare system to a mixed (public and private) system. Latvia and Bulgaria are countries which mostly rely on private healthcare representing 40 percent out of the total health care expenditure (WHO 2009). Waters (2008) provides a comprehensive review of the health insurance coverage for most of the countries included in our sample. For robustness checks, public health expenditure as a percent of GDP is used for the selected specification.
The literature regarding Eastern European health care systems raises another concern. In many of these countries, informal out of pocket payments are very common, especially where the doctors' salaries are low. However, this practice may lead to inequity in access to health care (Central and Eastern European Health Network 2002; WHO 2009). Even though informal payments would be an interesting component which may be added to the model, they were excluded because data are not available.
A study of 20 OECD countries shows that the mix between private and public health expenditures is important. A higher percent of public financed health expenditure is linked to higher mortality rates (Berger & Messer, 2002). The same relationship does not hold true for sub-Saharan African countries. There, an increase in health care expenditure, whether public or private, increases life expectancy at birth (Novignon et al. 2012). The difference in results can be explained by the difference in the economic status of the countries included in these studies.
Two other factors considered in this paper are number of physicians (per 1000 people) and hospital beds (per 1000 people). They are used as indicators of resources available to provide health care services. In Eastern Europe, the health care reforms have targeted to reduce the number of hospital beds in order to increase efficiency (European Communities and World Health Organization 2002). This effort can be seen from the declining trend confirmed by data. Even though many doctors try to immigrate to Western Europe (Golinowska 2006), the number of doctors increased in most of the countries. Previous research links a higher number of physicians with improved mortality rates (Nixon & Ulmann 2006).
Food production index is used as a proxy for food availability and it is defined by the World Bank as including food crops that are considered edible and that contain nutrients, excluding coffee and tea because they have no nutritive value. While the expected sign is negative, some problems may arise. Low income families may have difficulties in obtaining the essential food even though the food may be available. Mothers may neglect themselves in order to ensure their children receive enough food, resulting in poor nutrition and thus in higher infant mortality at birth. Another consideration is the effect of nutrition on the health status; the European Commission (2007), cited in WHO (2010), suggests that intake of fruit and vegetables is a good indicator of health. Because of lack of data, this aspect could not be included in the paper.
The gross enrollment (3) (primary, secondary, and tertiary) variable is used as a proxy for education or investment in human capital. It is the ratio of total enrollment to the population of the age group that officially corresponds to the level of education shown. The data are collected from the United Nations Development Programme's Human Development Index (HDI) which is a composite index of the rate of income growth, literacy, and life expectancy. Better education can decrease infant mortality, and thus the sign of the coefficient is expected to be negative. There is strong evidence in the literature that higher education for women leads to better health outcomes in general and in children's health status, in particular (Bozicevic 2006; OECD 2010). Education has a dual impact on health; the direct impact is determined by the health choices such as seeking medical care when needed or nutritional food intake, and the indirect effect is determined by the labor market outcomes (high wages). According to the OECD (2010), people who are unemployed over an extended period of time report bad or very bad health (Baert and De Norre 2009). In a recent study of Bulgaria, Estonia, Hungary, Poland, and Slovakia, the hypothesis that "persons with higher education are more likely to be in good health" failed to be rejected four out of five times (the exception being Bulgaria for which the result is not available) (Golinowska et al. 2006).
Alcohol consumption per capita is provided by the WHO (2010) and is computed as the sum of alcohol production and imports, less alcohol exports, divided by the adult population (aged 15 years or older). Alcohol consumption has generally been proven to have a negative impact on health. The WHO (2010) reports that approximately 25 percent of the differences in life expectancy among men 20-64 years old in Western and Eastern Europe can be explained by alcohol consumption. The consumption of alcohol has been associated with increased risks for several diseases and birth defects and hence positively related with mortality (OECD 2010). According to Gilmore et al. (2004), the effect of alcohol on health can be divided in two categories depending on the predominant consumption of spirit or wine. On the one hand, spirit consumption leads to death from acute intoxication, violence, and sudden cardiac attack while wine is responsible for liver cirrhosis. On the other hand, there are studies that suggest moderate consumption of alcohol may positively contribute to an improvement in health status at the margin by "decreasing the risk of heart failure among older persons" (Abramson et al. 2001).
Population is the last variable included as a social factor. It is not a determinant of health status per se, but it is included in the model in combination with food production index. As explained in Fayissa and Gutema (2005), food production index is calculated at the aggregate level. Thus, population is included in the model in order to switch from the macro perspective to the micro level. The food availability index has a valid effect on health when it is considered at per capita level, not at national level.
Among the environmental factors, urbanization rate represents the percent of the population living in urban areas and is made available by the World Bank. It has a deep and dual impact on health. First, access to medical care, availability of medical resources, and quality of these resources are considered to be higher in urban areas and, thus, positively impact health status. Second, it reflects the stress or the level of pollution that can have a negative consequence on health. For example, Golinowska et al. (2006) attribute the lower health status in rural Poland to inadequate medical resources, alcohol abuse, and accidents at work. Urbanization will also pick up the inequalities in the distribution of health resources between rural and urban areas. Zaborowski & Rebandel (2001) argue that 75 percent of medical staff is employed in urban areas serving 65 percent of the population (Zajac 2004). A similar problem of uneven distribution of specialists has been observed in many other Eastern European countries where doctors have moved from rural areas to urban areas (Rechel and McKee 2006).
The carbon dioxide emission variable is expressed in metric tons per capita and includes the emissions from the burning of fossil fuels and the manufacture of cement. This indicator is also available from the World Bank's data bank. The negative effect of environmental pollution (air, water) on public health has been well documented in previous studies. It could influence health "either directly by exposing people to harmful agents, or indirectly by disrupting life-sustaining ecosystems" (WHO 2009 cited in Remoundou & Koundouri 2009).We use carbon dioxide emission as a proxy for pollution to capture its negative effect on health. During the communist regimes, the Eastern European countries reached alarming levels of degradation of the natural environment (Jedrychowski 1995). Even though the integration into the European Union has imposed many restrictions intended to control and reduce pollution, these countries still face numerous challenges and higher risks than other members of the EU. For many countries in the sample, the numbers have improved during the study period and they are expected to improve in the future.
IV. The Empirical Model and Results
4.1 Pooled regression, fixed-effects, and random-effects models
The paper analyzes the determinants of health status in 13 Eastern European countries over a period of nine years from 1997 to 2005. The model to be estimated is the general form of Equation (1) and it is presented below:
[h.sub.it] = [[alpha].sub.0] + [summation] [[alpha].sub.i] [y.sub.it] + [summation] [[beta].sub.i] [s.sub.it] + [summation] [[gamma].sub.it] [[epsilon].sub.it], (2)
where [[alpha].sub.0] is a constant, [y.sub.it] are the economic factors, sit are the social factors and [e.sub.it] are the environmental factors of the ith country in period t, [[epsilon].sub.it] is a disturbance term with [epsilon]~N(0, [[sigma].sup.2.sub.[epsilon]]). For the estimation of the coefficients, a panel data analysis is used. Three different approaches for [[alpha].sub.0] are analyzed. The first approach assumes that [[alpha].sub.0] does not vary across countries or time. In this case, the model can be interpreted as a simple OLS regression. The second method specifies [[alpha].sub.0] to vary across countries, but remains constant for each country over time ([[alpha].sub.0] = [[alpha].sub.i] where i represents the country). This assumption yields the fixed-effect model. The last approach defines So different for each country as the previous method, but it also allows So to vary within each country. In this case, [[alpha].sub.0] is not a point estimate, but rather a disturbance term and can be defined as:
[[alpha].sub.i] = [alpha] + [[omega].sub.t], (3)
where [omega] ~ N(0, [[omega].sup.2.sub.[omega]]. In this case, a random-effects model is used to estimate the coefficients.
The results for each model, using infant mortality as dependent variable, are presented in Table 2.
In order to choose a reliable model for our analysis, we perform a series of tests. To decide between the first and second approach, the F-test is used to check if there are differences among intercepts ([H.sub.0]=there is no difference). The null hypothesis is rejected at p-value equal to 0.000. The conclusion is that the pooled regression model (OLS) would not capture differences among countries. A Lagrange multiplier test is also used to determine the significance of the country specific effects with the null hypothesis of [[sigma].sup.2.sub.[omega]] = 0. We are able to reject the null using a [chi square] of 90.02 at a p-value of 0.000. The importance of the country specific effects is revealed and the test of the suitability of a pooled regression model is rejected.
The problem that must be addressed at this juncture is the endogeneity between the observables and unobservables, i.e. the correlation between the independent variables and [[omega].sub.i] must be equal to 0; otherwise, the estimated parameters will be biased and suspect. The Hausman test is the appropriate method for testing for endogeneity with the null hypothesis of no difference in the parameter estimates using the fixed-effects and random-effects models. The null is rejected at the 5 percent significance level ([chi square] = 87.17; P-value = 0.00), suggesting that the [[omega].sub.i] is correlated with the observed variables and hence the fixed-effects model is more appropriate for the interpretation of the results. Analyzing the coefficients generated by the fixed effects model, there are five variables that have a statistically significant impact on the health status: income per capita, number of doctors, urbanization rate, carbon dioxide emission, and population. Because population was included in the model in combination with FPI, to switch from the macro to micro level, we do not discuss this estimate. The number of beds was excluded from the analysis because it has the same attribute as the number of doctors: to proxy the availability of medical care. These two variables are highly correlated, therefore, leading to multicollinearity problems. (4) The coefficients of the explanatory variables are elasticities since the estimates are based on the log of the variables. The coefficient of GDP per capita of 1.15 is negative and statistically significant, suggesting that a 1 percent increase in the GDP per capita decreases infant mortality rate by about 1.15 percent, other variables remaining unchanged. Taking into account the current trend with respect to GDP that most of the countries doubled if not more than doubled their numbers, it is expected that infant mortality will continue to decline in the future. Even though this would be the logical trend of thought, an increase in GDP is by no means a guaranty for the reduction of infant mortality.
The positive effect of number of doctors on infant mortality is expected and in accordance with previous research. When the number of physicians increases by 1 percent, infant mortality decreases by 0.43 percent. The density of doctors is one of the best proxies for the health care resources (Subramanian & Canning 2009) and the availably of trained personnel for prenatal and postnatal affects infant mortality. Another aspect of access to medical services is the urbanization rate. With most and best medical facilities located in urban area, the ease of access to health care is higher. Infant mortality decreases by three percent when the urbanization rate increases by one percent. This effect is statistically significant at the five percent level, and it also has economic significance. Considering that the average infant mortality for the considered sample is 90 for every 10,000 live births, the three percent decrease translates in three less infants per 10000 people dying before reaching one year of age. Even though one cannot say that the increase in urbanization is the solution to reduce infant mortality, an increased focus to improve access to medical facilities in rural area for countries in Eastern Europe can help.
The last significant variable is part of the environmental factors. An increase in carbon dioxide emissions by 1 percent leads to an increase of 0.54 percent in infant mortality. The positive relationship between health problems and air pollution in Eastern Europe has been widely documented, and many countries are taking the necessary steps to reduce pollution (WHO 2005). Even though some improvements have been made, there is still potential for further reduction in the level of dangerous pollutants (European Environment Agency, 2011).
Table 3 presents the estimates for neonatal mortality.
Performing the same battery of tests, the fixed effect model is the preferred specification. (5) The GDP, number of doctors, pollution as measured by carbon dioxide emissions maintain their significance and sign. The percent of private health care expenditures now becomes significant. Therefore, if it increases by one percent, the neonatal mortality decreases by 0.8 percent. Taking into consideration previously discussed causes of neonatal mortality, one can assume that the availability of better equipped intensive care units in private clinics can improve the chances of survival of infants.
Another variable that has a statistically significant impact on neonatal mortality is alcohol consumption. The estimate shows that a 1 percent increase in the alcohol consumption would lead to a decrease in infant mortality by 0.06 percent. This relation stands against the general belief that alcohol has an adverse effect on health status. An improvement of the model may be achieved by using the lagged values for alcohol consumption, taking into account the fact that the intake of alcohol does not have an immediate serious effect on health and it usually is a deferred outcome or consequence. Overall, while it has been established that the negative effect of alcohol consumption outweighs its positive effect, there are some scientific studies which suggest that the association between drinking and mortality during a 20-year period (which controlled for confounding factors such as previous problem of drinking) confirms a positive association of moderate drinking and reduced mortality among older adults (Abramson et al. 2001). Since we are measuring the alcohol consumption impact on neonatal mortality, the positive association is obviously not plausible.
4.2 Arellano--Bond Estimator
The concern we have with the previous specification is endogeneity. It is hard to believe that there is a feedback relationship between infant mortality and GDP, but if we consider that infant mortality is used as a proxy for health status, the feedback relationship is possible. Health status is affected by GDP and the GDP itself may be affected by the health status of the labor force. Since low infant mortality could affect the GDP in the future, but not the current value, it is necessary to perform further analysis. To address this issue, we use the Arellano-Bond (1991) dynamic estimator. The model is specified as follows:
[DELTA][y.sub.it] = [alpha][DELTA][y.sub.it-1] + [beta][DELTA][x.sub.it] + [DELTA][[epsilon].sub.it], (4)
where [DELTA][y.sub.it] is the first difference of the log of infant mortality, [[DELTA].sub.yit-1] is the lagged first differenced dependent variable, [[DELTA].sub.it] is a vector of first differenced explanatory variables and [DELTA][[epsilon].sub.it] is the first differenced error. The model relies on the assumption that the errors are serially uncorrelated. The explanatory variables can be exogenous, predetermined, or endogenous. The exogenous variables are not correlated with the error term while the predetermined variables are correlated with the past values, but not with the future errors of the error term, the endogenous variables are correlated with both past and future errors. A thorough description of the Arellano-Bond estimator can be found in Cameron & Trivedi (2009).
The advantages of using the Arellano and Bond estimator are multiple. First, it uses lags [y.sub.it-2] as instruments for the dependent variable and also allows any endogenous variable to be instrumented by using higher-order (t-2 or higher) lags. The advantage of using higher order lags is that it prevents correlation with the error term. In this paper, we treat GDP and enrolment as endogenous variables and create instruments for them, using their own lags. The remaining variables are considered exogenous in this paper. The second advantage of using the Arellano and Bond estimator is that it addresses the stationarity problem by using the first-differenced variables. Third, it eliminates the autocorrelation problem, and fourth, the time invariant unobserved characteristics are eliminated ([[alpha].sub.i] specified in the fixed-effects model).
The results are presented in Table 4. (6)
Comparing Arellano-Bond estimates to random effects results, most of the variables maintain their expected signs and significance. GDP per capita is statistically significant, having a positive impact on infant mortality. A 1 percent rise in GDP per capita reduces infant mortality by 1.15 percent. The coefficients for urbanization and pollution are also significant and maintain their expected signs. Urbanization contributes to the reduction of infant mortality and the effect is not trivial; a 1 percent increase in urbanization leads to a 3.56 percent decrease in infant mortality rate. The effect may have been triggered by the better and easier access to medical care and education. The living conditions in the rural areas are worse than those in urban areas for most of the Eastern European countries and many villages do not have access to more advanced medical services. The C[O.sub.2] emission variable that represents the damaging effects of environment causes, as expected, increases infant mortality. The coefficient for the number of doctors is negative and statistically significant, showing that 1 percent increase leads to a reduction of 0.43 percent in infant mortality.
Enrollment also has a positive and statistically significant effect on infant mortality. Its coefficient value of 0.73 implies that a 1 percent increase in the enrollment ratio will result in a 0.73 percent reduction in the mortality rate. This result suggests that increased investment in education may provide future mothers with the necessary knowledge for a healthy pregnancy and better awareness of healthy life-style choices. Because infant mortality reflects deaths caused during the first year of life, medical resources in general, and medical staff in particular, have a considerable impact in reducing infant mortality. A higher rate of enrollment may reflect better qualified people (e.g. medical doctors, nurses, technicians) who can provide better health services. The level of education is pretty much homogenous among the Eastern European countries and it is one of the few aspects which has not dramatically changed after the fall of the communist regimes. For robustness checks, Table 4 also presents the estimates of the same model but using the public health care expenditure instead of private healthcare expenditure. All coefficients maintain their signs and significance.
Table 5 presents the estimates for neonatal mortality using the Arellano and Bond estimator. (7)
This time, only the coefficients for GDP, private health care expenditure, and urbanization rate continue to be statistically significant. The robustness check (using public health care expenditure) supports these findings.
Using the theory developed by Grossman (1972), the framework provided by Fayissa and Gutema (2005) and panel data for 13 of the countries in the Eastern European Bloc over the 1997-2005 period, the study analyzes the impact of economic, social, and environmental factors on the health status of countries in the sample. Based on the Arellano-Bond model which addresses multiple econometric issues, the results indicate that economic growth as measured by GDP per capita, investment in human capital formation, higher number of physician, pollution reduction, and residence in urban areas have statistically significant effects in reducing infant mortality. The findings may serve as a starting point in developing a health policy that is directed toward improving the health of the Eastern European region to the level of Western European countries.
Acknowledging that these countries have different histories, backgrounds, and resources, the paper uses different models which are capable of controlling for the dissimilarities. (8) The current upward trend in GDP promises bright prospects in the improvement of the health status of the region in general and infant mortality in particular. The results also suggest that education plays a key role in improving public health by providing the necessary knowledge to make wise decisions about things that affect health. The study also reveals that urbanization rate and pollution reduction (represented by the C[O.sub.2] emission) should not be ignored by the policy makers; both factors can lead to better health. Urbanization allows better access to medical care and a higher living standard, while pollution reduction can diminish the risk of environmental illnesses. Considering the positive effect of private health care on neonatal mortality, the reorganization of health care system is an important and necessary step. Many Eastern European countries have already started this process. To make a policy recommendation based on reduced form evidence may be a precarious proposition; nevertheless, a good strategy for the Eastern European countries is to focus part of their limited resources on promoting education, cleaner environment, and better access to medical services (public and private), particularly in rural areas.
APPENDIX 1. Variables: definition and data source Abbreviation Definition Source Infant I.Mort The number of infants Eurostat Mortality dying before reaching one year of age, per 1,000 live births in a given year Neonatal N.Mort The number of neonates World Bank Mortality dying before reaching 28 days of age, per 1,000 live births in a given year GDP per capita GDP Gross domestic product World Bank divided by midyear population Gross EDU The number of students UNDP Enrollment enrolled in primary, in Education secondary and tertiary levels of education, regardless of age, as a percentage of the population of theoretical school age for the three levels. Total Health THE THE is the sum of World Bank Expenditure public and private as % of GDP health expenditure. It covers the provision of health services, family planning activities, nutrition activities, and emergency aid designated for health but does not include provision of water and sanitation. Health PRV PRV includes direct World Bank expenditure, household private (out-of-pocket) (% of GDP) spending, private insurance, charitable donations, and direct service payments by private corporations. Health PUB PUB consists of World Bank expenditure, recurrent and capital public spending from (% of GDP) government budgets, external borrowings and grants (including donations from international agencies and nongovernmental organizations), and social (or compulsory) health insurance funds. Physicians DOC Physicians include World Bank & (per 1,000 generalist and Eurostat people) specialist medical practitioners. Food FPI FPI covers food crops World Bank Production that are considered Index edible and that contain nutrients. Coffee and tea are excluded because (they have no nutritive value. Alcohol ALC Recorded adult (15+) WHO Consumption per capita alcohol consumption in liters Population POP Total population Urban URB Urban population World Bank Population refers to people (% of total) living in urban areas. C02 Emissions C02 C02 emissions are World Bank (metric tons those stemming from per capita) the burning of fossil fuels and the manufacture of cement. They include carbon dioxide produced during consumption of solid, liquid, and gas fuels and gas flaring. APPENDIX 2. Descriptive statistics by country LMORT N.MORT GDP EDU THE Belarus Mean 9.23 5.88 1,392 83.86 6.51 Std. Dev. 2.07 1.00 265 4.90 0.23 Min 6.90 4.40 1,062 77.00 6.15 Max 12.60 7.20 1,871 88.70 6.87 Bulgaria Mean 13.53 10.59 1,678 75.61 6.66 Std. Dev. 2.05 1.27 260 4.12 1.12 Min 10.40 8.60 1,352 70.00 4.89 Max 17.50 12.20 2,107 81.50 7.87 Croatia Mean 7.14 5.20 4,432 70.50 7.07 Std. Dev. 0.92 0.55 505 3.06 0.53 Min 5.70 4.40 3,857 67.00 6.29 Max 8.20 6.00 5,238 75.00 7.82 Czech Mean 4.32 3.81 5,759 76.21 6.91 Rep. Std. Dev. 0.79 0.66 491 4.64 0.37 Min 3.40 3.00 5,245 70.00 6.55 Max 5.90 5.00 6,676 82.90 7.44 Estonia Mean 7.86 4.93 4,605 88.93 5.33 Std. Dev. 1.76 1.27 917 4.60 0.47 Min 5.40 3.30 3,555 81.00 4.86 Max 10.00 7.00 6,211 96.00 6.21 Hungary Mean 8.07 6.39 4,927 82.70 7.49 Std. Dev. 1.34 1.03 621 5.63 0.59 Min 6.20 5.00 4,055 74.00 6.74 Max 9.90 8.00 5,870 89.30 8.33 Latvia Mean 11.04 10.01 3,687 83.69 6.20 Std. Dev. 2.54 1.83 783 6.90 0.14 Min 7.80 7.60 2,727 71.00 5.96 Max 15.30 12.80 5,047 90.20 6.43 Lithuania Mean 8.23 5.88 3,689 84.93 6.19 Std. Dev. 1.15 1.02 682 7.16 0.28 Min 6.70 4.50 2,910 75.00 5.70 Max 10.30 7.60 4,873 94.00 6.52 Poland Mean 8.01 5.82 4,536 85.02 5.96 Std. Dev. 1.29 1.04 439 4.57 0.30 Min 6.40 4.60 3,875 77.00 5.52 Max 10.20 7.60 5,225 90.00 6.34 Romania Mean 18.21 13.01 1,854 70.64 5.15 Std. Dev. 2.10 1.14 239 3.27 0.34 Min 15.00 11.20 1,616 68.00 4.55 Max 22.00 14.60 2,259 76.80 5.66 Slovak Mean 7.79 6.72 4,034 75.48 6.63 Rep. Std. Dev. 0.91 0.80 378 1.57 0.71 Min 6.20 5.70 3,613 73.00 5.66 Max 8.80 8.10 4,733 78.30 7.53 Slovenia Mean 4.40 3.04 9,965 86.70 8.29 Std. Dev. 0.58 0.37 961 7.02 0.35 Min 3.70 2.50 8,530 76.00 7.81 Max 5.20 3.60 11,432 95.00 8.66 Ukraine Mean 11.41 7.70 731 81.28 6.29 Std. Dev. 1.72 0.74 145 4.13 0.45 Min 9.40 6.50 591 77.00 5.64 Max 14.20 8.50 962 86.50 6.95 PRV PUB DOC FPI ALC Belarus Mean 1.58 76.07 4.53 107.44 10.54 Std. Dev. 0.24 3.34 0.09 9.79 1.37 Min 1.17 71.10 4.37 97.00 8.17 Max 1.91 82.74 4.68 126.00 12.23 Bulgaria Mean 2.48 62.30 3.49 93.22 10.20 Std. Dev. 0.63 4.90 0.09 10.89 1.57 Min 1.45 56.27 3.37 75.00 7.87 Max 3.07 70.31 3.64 104.00 11.72 Croatia Mean 1.10 84.22 2.38 97.44 12.96 Std. Dev. 0.12 2.63 0.09 9.28 0.56 Min 0.93 80.05 2.27 85.00 11.98 Max 1.27 87.34 2.50 111.00 13.75 Czech Mean 0.79 88.58 3.35 98.11 14.41 Rep. Std. Dev. 0.22 2.65 0.21 5.25 0.98 Min 0.62 83.73 3.03 87.00 13.04 Max 1.18 90.48 3.55 106.00 15.32 Estonia Mean 1.03 80.24 3.20 105.00 10.41 Std. Dev. 0.18 4.50 0.05 6.08 4.09 Min 0.67 77.06 3.14 97.00 4.74 Max 1.19 89.17 3.27 116.00 16.24 Hungary Mean 2.12 71.91 3.04 100.56 12.48 Std. Dev. 0.41 3.96 0.22 9.94 0.27 Min 1.26 68.98 2.68 86.00 11.97 Max 2.53 81.28 3.34 119.00 12.80 Latvia Mean 2.79 55.14 2.79 109.89 9.13 Std. Dev. 0.14 2.74 0.08 9.98 0.54 Min 2.56 51.21 2.67 98.00 8.40 Max 2.98 59.24 2.88 126.00 10.20 Lithuania Mean 1.73 71.76 3.66 110.56 9.44 Std. Dev. 0.23 4.52 0.06 10.65 2.43 Min 1.46 64.57 3.57 91.00 6.35 Max 2.07 76.01 3.76 124.00 12.50 Poland Mean 1.88 68.57 2.29 102.00 8.70 Std. Dev. 0.26 3.24 0.08 3.67 0.70 Min 1.57 64.65 2.14 98.00 7.71 Max 2.20 71.96 2.43 110.00 9.84 Romania Mean 1.55 70.46 1.96 104.56 10.54 Std. Dev. 0.36 7.02 0.12 10.92 1.10 Min 0.92 62.13 1.79 89.00 8.99 Max 1.89 82.68 2.17 127.00 12.57 Slovak Mean 1.16 83.16 3.19 105.44 10.51 Rep. Std. Dev. 0.61 7.72 0.04 8.49 0.45 Min 0.48 72.08 3.15 92.00 9.90 Max 1.97 91.68 3.23 118.00 11.06 Slovenia Mean 2.28 72.60 2.21 102.00 10.72 Std. Dev. 0.32 2.96 0.07 2.74 2.22 Min 1.91 68.18 2.13 99.00 7.84 Max 2.75 75.66 2.35 106.00 13.85 Ukraine Mean 2.81 55.03 2.98 104.44 6.53 Std. Dev. 0.19 2.70 0.02 8.72 2.04 Min 2.57 50.22 2.95 91.00 3.74 Max 3.10 58.40 2.99 118.00 8.50 POP URB CO2 Belarus Mean 9,955,032 70.40 6.14 Std. Dev. 114,600 1.21 0.31 Min 9,775,591 68.70 5.82 Max 10,100,000 72.20 6.63 Bulgaria Mean 7,995,563 69.19 5.79 Std. Dev. 218,596 0.68 0.36 Min 7,740,000 68.24 5.25 Max 8,312,068 70.20 6.30 Croatia Mean 4,472,667 55.81 4.90 Std. Dev. 55,608 0.46 0.39 Min 4,426,000 55.18 4.32 Max 4,572,000 56.50 5.41 Czech Mean 10,200,000 73.91 12.12 Rep. Std. Dev. 39,462 0.29 0.50 Min 10,200,000 73.50 11.00 Max 10,300,000 74.36 12.87 Estonia Mean 1,366,969 69.48 12.72 Std. Dev. 17,887 0.13 0.97 Min 1,346,100 69.40 11.62 Max 1,400,000 69.76 13.88 Hungary Mean 10,200,000 65.25 5.81 Std. Dev. 71,270 0.59 0.19 Min 10,100,000 64.60 5.60 Max 10,300,000 66.30 6.15 Latvia Mean 2,361,959 68.16 2.97 Std. Dev. 48,817 0.18 0.25 Min 2,300,500 68.00 2.57 Max 2,450,000 68.52 3.36 Lithuania Mean 3,491,116 66.91 3.94 Std. Dev. 55,417 0.20 0.34 Min 3,414,300 66.60 3.48 Max 3,580,000 67.18 4.55 Poland Mean 38,400,000 61.61 8.13 Std. Dev. 221,657 0.06 0.41 Min 38,200,000 61.50 7.76 Max 38,700,000 61.70 9.08 Romania Mean 22,100,000 53.63 4.29 Std. Dev. 390,586 0.09 0.39 Min 21,600,000 53.50 3.84 Max 22,600,000 53.80 5.14 Slovak Mean 5,384,983 56.29 7.37 Rep. Std. Dev. 5,729 0.07 0.35 Min 5,378,900 56.20 6.77 Max 5,395,115 56.42 8.01 Slovenia Mean 1,991,362 50.34 7.49 Std. Dev. 5,994 0.48 0.23 Min 1,982,600 49.50 7.26 Max 2,000,500 50.80 8.00 Ukraine Mean 48,800,000 67.32 6.54 Std. Dev. 1,222,207 0.28 0.34 Min 47,100,000 67.04 6.20 Max 50,600,000 67.80 7.06
Abramson, Williams, Harlan Krumholz, and Viola Vaccarino. 2001. "Moderate alcohol consumption and risk of heart failure among older persons." Journal of American Medical Association 285(15): 1971-1977.
Arellano, Manuel and Stephen Bond. 1991. "Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations." Review of Economic Studies 58: 277-297.
Auster, Richard, Irving Levenson, and Deborah Sarachek. 1969. "The Production of Health, an Exploratory Study." Journal of Human Resources 4: 411-436.
Baert, Katrien and Bart De Norre. 2009. Perception of health and access to health care in the EU-25 in 2007. Eurostat Statistics in Focus, 24, Luxembourg: European Commission.
Berger, Mark and Paul Leigh. 1989. "Schooling, self-selection, and health." Journal of Human Resources 24: 433-455.
Berger, Mark and Jodi Messer. 2002. "Public financing of health expenditures, insurance, and health outcomes." Applied Economics 34: 2105-2113.
Bhargava, Aloc, Dean Jamison, and Christopher Murray. 2001. "Modeling the effects of health on economic growth." Journal of Health Economics 20: 423-440.
Bozicevic, Ivana. 2006. Health in South-Eastern Europe: Recent Trends, Patterns and Socioeconomic Inequalities in Health. Health and economic development in South-Eastern Europe. Paris: WHO Regional Office for Europe.
Cameron, Colin and Trivedi Pravin. 2009. Micro-econometrics using Stata. College Station, TX: Stata Press.
Central and Eastern European Health Network. 2002. "Formal and informal household spending on health: a multi-country study in Central and Eastern Europe. Retrieved January 13, 2011 (http://www.hsph.harvard.edu/ihsg/publications/pdf/lac/MerckReportFINAL.pdf)
Ettner, Susan. 1996. "New evidence on the relationship between income and health." Journal of Health Economics 15: 67-85.
European Commission. 2007. White paper on a strategy for Europe on nutrition overweight and obesity related health issues. Brussels: Commission of European Communities.
European Communities and World Health Organization. 2002, Health status overview for countries of Central and Eastern Europe that are candidates for accession to the European Union. Copenhagen: WHO Regional Office for Europe.
European Environmental Agency. 2011. "Has policy improved Europe's air quality?". Retrieved January 20, 2011 (http://www.eea.europa.eu/ highlights/has-policy-improved-europe2019s-air-quality)
Favaro, Donata and Marc Suhrcke. 2006. Demographic and economic trends in South-Eastern Europe. Health and economic development in South-Eastern Europe. Paris: WHO Regional Office for Europe.
Fayissa, Bichaka and Paulos Gutema. 2005. "Estimating a health production function for Sub-Saharan Africa." Applied Economics 37: 155-164.
Gilmore, Anna, Esa Osterberg, Antero Heloma, Wiltold Zatonski, Evgenia Delcheva, and Martin McKee. 2004. Health policy and European Union enlargement--Free Trade versus the Protection of Health: The Examples of Alcohol and Tobacco. European Observatory on Health Systems and Policies Series. New York, NY: Open University Press.
Golinowska, Stanislawa, Sowa Agnieszka, and Madry Roman. 2006. Health status and health care systems in Central and Eastern European countries. Brussels: European Network of Economic Policy Research Institute.
Grossman, Michael. 1972. The demand for health: A theoretical and Empirical Investigation. New York: NBER.
Iliev, Dimo and Marc Suhrcke. 2006. Introduction Health and economic development in South-Eastern Europe. Paris: WHO Regional Office for Europe.
Jacobson, Lena. 1999. "The family as producer of health--an extended Grossman model." Journal of Health Economics 19: 611-637.
Jedrychowski, Wieslaw. 1995. "Review of recent studies from Central and Eastern Europe associating respiratory health effects with high levels of exposure to "traditional" air pollutants." Environmental Health Perspectives 103(2): 15-21.
Nixon, John and Philippe Ulmann. 2006. "The relationship between health care expenditures and health outcomes." European Journal of Health Economics 7: 7-18.
Novignon, Jacob. 2012. "The effects of public and private health care expenditure on health status in sub-Saharan Africa: new evidence from panel data analysis." Health Economics Review, 2:22.
OECD. 2010. Health at a glance: Europe 2010. OECD Publishing. (Also available at: http://dx.doi.org/10.1787/health_glance-2010-en.).
Rechel, Bernd and Martin Mckee. 2006. Health systems and policies in South-Eastern Europe. Health and economic development in South-Eastern Europe. Paris: WHO Regional Office for Europe.
Rechel, Bernd and Martin Mckee. 2009. " Health reform in central and Eastern Europe and the former Soviet Union." Lancet 374:1186-1195.
Remoundou, Kyriaki and Phoebe Koundouri. 2009. "Environmental Effects on Public Health: An Economic Perspective." International Journal of Environmental Research and Public Health 6(8): 2160-2178.
Rosen, Sherwin and Paul Taubman. 1982. Some socioeconomic determinants of mortality Economics of Health Care. New York, NY: Preager Publishers.
Rowland, Diane. 1991. "Health status in East European countries." Health Affairs 10(3): 202-215.
Schultz, Paul. 2004. "Health Economics and applications in developing countries." Journal of Health Economics 23: 637-641.
Subramanian Mansour and David Canning. 2009. "Short and long term relationship between physician density on infant mortality: a longitudinal econometric analysis." Working Paper No. 49, Program on the Global Demography of Aging, Boston, MA.
Uxa, Fabio et al. 2006. "Essential newborn care training activities: 8 years of experience in Eastern European, Caucasian and Central Asian countries." Seminars in Fetal and Neonatal Medicine 11: 58-64.
Vladescu, Cristian et al. 2000. Health care systems in transition: Romania. Copenhagen: European Observatory on Health Care Systems.
Zaborowski, Piotr and Henryk Rebandel. 2001. "The profession of physician in Poland: Demographic analysis approach." Gazeta Lekarska 4.
Zajac, Monica. 2004. Free movement of health professionals: The Polish experience. Health policy and European Union enlargement. New York, NY: Open University Press.
Waters, Hugh. 2008. "Health Insurance Coverage in Central and Eastern Europe--Trends and Challenges." Health Affairs forthcoming.
World Health Organization. 2005. Health basis for air quality management in Eastern Europe, Caucasus and Central Asia. Copenhagen: WHO Regional Office for Europe.
World Health Organization. 2009. Preventing disease through healthy environments: Towards an estimate of the environmental burden of disease. Geneva: World Health Organization. (Also available at: http://www.who.int/quantifying_ehimpacts/publications/preventingdisease.pdf.)
World Health Organization. 2010. European status report on alcohol and health 2010. Copenhagen: WHO Regional Office for Europe.
(1.) The 13 Eastern European countries in our sample are: Belarus, Bulgaria, Croatia, Czech-Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovak-Republic, Slovenia, and Ukraine.
(2.) Data are not available for post-neonatal mortality.
(3.) We also considered only secondary enrollment, but the results are not qualitatively, different.
(4.) A model that includes the number of beds was considered, but the coefficients do not change qualitatively. Therefore, we decided to present the straightforward version that excludes the number of beds.
(5.) Breusch and Pagan Lagrangian multiplier test for random effects [chi.sup.2](1) = 164.47 Prob > [chi.sup.2] = 0.00; Test of overidentifying restrictions: fixed vs random effects: Sargan-Hansen statistic 65.035 [Chi.sup.2] (10) P-value = 0.00.
(6.) Sargan test of overidentifying restrictions H0: overidentifying restrictions are valid [chi.sup.2](38)= 42.79 Prob>[chi.sup.2]=0.273. Test for zero auto-correlation in first-differenced errors H0: no autocorrelation: AR(1):z=-l.722 Prob>z=0.085; AR(2):z=-l.515 Prob>z=0.130; AR(3):z= 0.887 Prob>z=0.375.
(7.) Sargan test of overidentifying restrictions H0: overidentifying restrictions are valid chi2(38)= 34.653 Prob>chi2=0.625. Test for zero autocorrelation in first-differenced errors H0: no autocorrelation: AR(1):z=- 1.923 Prob> z=0.054; AR(2):z=l.352 Prob>z=0.175; AR(3):z= -1.208 Prob>z=0.227.
(8.) Additional dummy variables we considered are: former members of USSR, religious affiliation, or influence of Western European countries. Because there is no variance in these variables, they are omitted by the Arellano and Bond specification. For consistency, we dropped them from all models.
Bichaka Fayissa, Department of Economics and Finance, Jennings A. Jones College of Business, Middle Tennessee State University, Box 189, Murfreesboro, TN 37132 Email: Bichaka.Fayissa@mtsu.edu
Anca Traian, Department of Economics and Finance, Jennings A. Jones College of Business, Middle Tennessee State University, Box 27, Muffreesboro, TN 37132 Email: firstname.lastname@example.org
TABLE 1. Descriptive statistics Mean Std. Dev. Min Max LMORT 9.17 3.90 3.40 22.00 N.MORT 6.85 2.90 2.50 14.60 GDP 3,945.31 2,360.50 590.96 11,432.24 EDU 80.43 7.52 67.00 96.00 THE 6.51 0.94 4.55 8.66 PRV 1.79 0.73 0.48 3.10 PUB 72.31 10.91 50.22 91.68 DOC 3.00 0.70 1.79 4.68 FPI 103.13 9.47 75.00 127.00 ALC 10.51 2.53 3.74 16.24 POP 12,800,000 14,300,000 1,346,100 50,600,000 URB 63.71 7.17 49.50 74.36 CO2 6.79 2.83 2.57 13.88 TABLE 2. Estimates for Infant Mortality Pooled Fixed Random Regression Effects Effects GDP -0.62 ** -1.15 ** -0.62 ** (0.082) (0.160) (0.082) ENR -0.73 ** -0.21 -0.73 ** (0.217) (0.241) (0.217) THE -0.10 -0.01 -0.10 (0.133) (0.125) (0.133) PRV 0.00 0.03 0.00 (0.053) (0.049) (0.053) DOC -0.49 ** -0.43 ** -0.49 ** (0.196) (0.202) (0.196) FPI 0.06 0.16 0.06 (0.111) (0.100) (0.111) ALC -0.06 -0.06 -0.06 (0.054) (0.052) (0.054) POP -0.13 ** -3.11 ** -0.13 ** (0.051) (1.177) (0.051) URB -0.25 -3.00 ** -0.25 (0.453) (1.195) (0.453) CO2 -0.02 0.54 ** -0.02 (0.102) (0.142) (0.102) No of Obs. 109 109 109 [R.sup.2] 0.80 0.84 0.80 Note: Standard errors are presented in parentheses. ** denotes significance at the 0.05 level. * denotes significance at the 0.10 level. TABLE 3. Estimates for Neonatal Mortality Pooled Fixed Random Regression Effects Effects GDP -0.66 ** -1.17 ** -0.66 ** (0.073) (0.099) (0.073) ENR -0.53 ** -0.05 -0.53 ** (0.163) (0.149) (0.163) THE 0.09 0.01 0.09 (0.090) (0.077) (0.090) PRV -0.11 ** -0.08 ** -0.11 ** (0.036) (0.030) (0.036) DOC -0.37 ** -0.22 * -0.37 ** (0.146) (0.125) (0.146) FPI -0.04 0.08 -0.04 (0.074) (0.062) (0.074) ALC -0.06 -0.06 * -0.06 (0.037) (0.032) (0.037) POP -0.14 ** -2.88 ** -0.14 ** (0.060) (0.727) (0.060) URB -0.07 0.39 -0.07 (0.453) (0.738) (0.453) CO2 0.01 0.26 ** 0.01 (0.092) (0.088) (0.092) No of Obs. 109 109 109 [R.sup.2] 0.90 0.93 0.90 Note: Standard errors are presented in parentheses. ** denotes significance at the 0.05 level. * denotes significance at the 0.10 level. TABLE 4. Arellano and Bond Estimates for Infant Mortality Model 1 Model 2 Infant Mortality L1 -0.29 ** -0.31 ** (0.122) (0.122) GDP -1.28 ** -1.24 ** (0.217) (0.213) ENR -0.73 ** -0.75 ** (0.299) (0.295) THE -0.16 -0.16 (0.183) (0.191) PR V 0.03 (0.055) PUB 0.07 (0.144) DOC -0.42 * -0.47 * (0.237) (0.242) FPI 0.14 0.14 (0.113) (0.113) ALC -0.07 -0.05 (0.065) (0.066) POP -4.03 ** -3.50 ** (1.551) (1.496) URB -3.56 ** -3.80 ** (1.532) (1.545) CO2 0.64 ** 0.64 ** (0.173) (0.172) Note: Standard errors are presented in parentheses. ** denotes significance at the 0.05 level. * denotes significance at the 0.10 level. Instruments used for GDP are GDP(t-2), GDP(t-3), GDP(t-4) Instruments used for Enrollment are Enrol Iment(t-2), Enrol lment(t-3), Enrollment(t-4) Instrument used for Infant Mortality is Infant Mortality (t-2) TABLE 5. Arellano and Bond Estimates for Neonatal Mortality Model 1 Model 2 Neonatal Mortality L1 0.83 ** 0.81 ** (0.069) (0.071) GDP -0.15 * -0.18 * (0.090) (0.092) ENR -0.01 -0.02 (0.059) (0.060) THE 0.03 -0.03 (0.036) (0.041) PRV -0.02 * (0.011) PUB 0.05 (0.030) DOC -0.03 -0.04 (0.047) (0.049) FPI 0.00 0.00 (0.023) (0.023) ALC -0.02 -0.02 (0.012) (0.012) POP 0.01 -0.14 (0.376) (0.376) URB -0.75 ** -0.73 ** (0.304) (0.305) CO2 0.02 0.02 (0.036) (0.037) Note: Standard errors are presented in parentheses. ** denotes significance at the 0.05 level. * denotes significance at the 0.10 level. Instruments used for GDP are GDP(t-2), GDP(t-3), GDP(t-4) Instruments used for Enrollment are Enrollment(t-2), Enrollment(t-3), Enrollment(t-4) Instrument used for Infant Mortality is Infant Mortality (t-2)
|Printer friendly Cite/link Email Feedback|
|Author:||Fayissa, Bichaka; Traian, Anca|
|Date:||Sep 22, 2013|
|Previous Article:||The decreasing excludability of digital music: implications for copyright law.|
|Next Article:||What really happens in the Solow model: technological progress versus population growth?|