A demonstration in model development and testing

Politics of health care in Turkey: A demonstration in model development and testing

Atav, A Serdar

Based on alternative models of neighborhood service distributions in the United States, the primary purpose of this paper was to develop a conceptual model to determine the distributive structure of health services in developing countries. The propositions of the neighborhood service distribution literature were translated into testable hypotheses that included industrial development, agricultural development, need, and political interest. These hypotheses were integrated into a path analytical model that was tested for the distribution of general practitioner physicians among Turkish provinces. The findings supported the need and political interest hypotheses. The model developed and tested by this study promises to be widely applicable to other countries and policy indicators.

Key Words: Turkish Health Policy, Urban Policy, Turkey, Health in Developing Countries, Path Analysis

The World Health Organization (WHO) recently completed its first ever analysis of the world’s health systems. The main message from this analysis was that the health and well being of people around the world depend critically on the performance of the health systems that serve them. Yet, there is wide variation in performance within and among countries with similar levels of income and health expenditures. The WHO report emphasized the importance of understanding the underlying reasons for variations and inequalities within and among systems so that system performance, and hence the health of populations, can be improved (WHO, 2000).

While inequalities among countries receive a significant amount of attention, (WHO, 2000; Macfarlane, Racelis, & Muli-Musimime 2000; Roemer, 1993; Roemer, 1991), quantitative analyses of inequalities within countries are yet to be conducted. It is universally agreed upon that an equitable and rational geographic distribution of health resources would help improve health conditions in most developing countries. Still, international health policy studies have not touched upon the geographic distribution problem from an analytical perspective: other than anecdotal explanations, a comprehensive and conceptual approach with testable hypotheses does not exist to approach the question of why certain regions in a developing country receive more and better health care services than other regions.

In an entirely different context, this problem of geographic distribution of public services has long been within the domain of American municipal service studies. Cumulative policy studies in the United States regarding the distribution problem focused on the underlying criteria that affect the distributive decisions of local governments. A body of literature on neighborhood service distribution emerged in the seventies and the eighties proposing that through the identification of patterned inequalities and the changeable aspects of social, political and economic forces, the study of “who gets what and why” has the potential of assisting policy makers to reduce unjust distribution of public services and help maint,-n social stability (Lee, 1994).

Based on alternative models of neighborhood service distributions in the United States, the primary purpose of this paper was to develop a conceptual model to determine the distributive structure of health services in developing countries. The model was then tested within the Turkish political context.

Developing Country Political Context

Unlike the decentralized form of government in the United States, most developing countries have strong traditions in centralized national governments. In the absence of checks and balances, like ‘federalism’ or `the separation of powers’, any ruling party in government has the potential of reordering national priorities and translating its ideological concerns into actual policy, both at the national and the regional levels. The ideology of the ruling party or the leader can determine the level of health care services enjoyed by their citizens. For example, one of the most positive outcomes of the Cuban revolution was the reorganization of the national health services. Treating health as a basic human right and providing universal access, Cuban politicians were able to design a health care system with minimal resources that now ranks about the same as the United States in terms of overall system performance (WHO, 2000). Similarly, radical changes in Oman and Columbia transformed these countries’ health care systems. They now rank much higher than the United States in most measurements of health system performance (WHO, 2000).

In the American political system large scale policy shifts are difficult, if not impossible. In most parts of the developing world, on the other hand, elections or military takeovers are expected to bring ‘revolutionary’ changes in budget priorities and distributive decisions. Although one might think the outcome of these radical changes is obvious, international health policy studies have not dealt with the issue of service distributions. Rather, studies have concentrated on the evaluation of the ‘impact’ of small scale, internationally funded regional projects including water, sanitation, nutrition, and birth control (Knippenberg, 1997; Smith & Bryant, 1988).

Alternatively, at the aggregate or national level, studies have been more descriptive than analytical (Roemer, 1993; Roemer, 1991). These approaches lack the theoretical and methodological sophistication of policy studies conducted in the developed world. Recently, the World Bank and the World Health Organization began to focus on inequalities in distribution of health services (WHO, 2000; Makinen et al, 2000). However, the analyses in these studies only relate to access issues among different income groups without referring to geographic distribution of services.

Conceptual Framework: American Neighborhood Service Distribution

In the decentralized tradition of American government structure, autonomous local governments have the responsibility for provision of a wide range of social goods essential for the preservation of ‘life’ (police, fire, sanitation), ‘liberty’ (police, courts, prosecutors), ‘property’ (zoning, planning, taxing), and `public enlightenment’ (schools, libraries). Local jurisdictions determine not only what and how much of municipal services will be provided, but also how these services will be distributed among competing groups. Any government that decides to provide certain services must also decide by what criteria to distribute public services among citizens or groups of citizens so that policy goals are advanced. Despite claims by governments to ‘optimal’ or ‘rational’ criteria by which services are distributed, striking differences exist in physical character and apparent quality of life in different areas. Rich argues that the differences in the conditions of streets and public facilities, the nature of housing stock, and so forth from one neighborhood to another are so noticeable that it is impossible to believe citizens enjoy even roughly equal public services (Rich 1982).

Although some of the differences in the physical characteristics of neighborhoods are attributable to differences in the private wealth of their inhabitants, even the narrow range of public services provided by local governments shape the quality of life (Rich 1982). Hence, municipal services have been a point of controversy in many cities. There has been increasing awareness that such services represent both real and potential wealth to citizens who suffer genuine losses if they are systematically deprived of their share of services (Antunes & Plumblee, 1977).

In the seventies and the eighties, a body of literature has focused on the underlying criteria effecting the distributive decisions of governments. These studies have resulted in the following hypotheses:

1. The electoral-interest hypothesis: The electoral interest explanation assumes a continuous feedback between voters and policy makers. Party officials reward jurisdictions that play an important role in their victory through allocating a higher share of existing resources. Area rewards also involve pay-offs to units that are seen by incumbents as potentially playing a role in future elections (Cingranelli, 1981).

2. The race preference hypothesis: It is argued that the quantity and/or the quality of urban services are positively related to the proportion of Anglos in a neighborhood population (Lineberry, 1977).

3. The class preference hypothesis: It is argued that the quantity and /or quality of urban services are positively related to the proportion of the neighborhood population that is of higher socio-economic status (Lineberry, 1977).

4. The power elite hypothesis: It is argued that the quantity and /or quality of urban services are positively related to the proportion of the neighborhood population occupying positions of power in urban government (Lineberry, 1977).

5. Need hypothesis: It is argued that the quantity and/or quality of urban services are functions of neighborhood conditions. Ecological aspects of urban neighborhoods, including but not limited to their age, density, geographical character, and residential-commercial mix determine the need for services that in turn affect service delivery (Sanger, 1982).

6. The incremental hypothesis: This hypothesis assumes that in the short-run it is not possible to make big shifts in the level of services. Changes in service levels are linear processes in which services for a certain year can be explained by a linear model that posits acceptance of a base and some incremental adjustments (generally positive) in subsequent periods.

A model integrating these competing hypotheses is presented in Figure 1.

Turkish Political and Health Care Context

Turkey is located at the juncture where Europe meets Asia forming a bridge and a link to each continent. Covering over 300,000 square miles, Turkey is about the size of Texas with a population of 65 million. A member of North Atlantic Treaty Organization (NATO) since 1952, Turkey provides land for numerous US military bases.

Although the political system in Turkey is a parliamentary democracy, the country has had three military takeovers since the early sixties in response to various cycles of political upheaval, economic problems, and domestic terrorism that paralyzed the civilian governments. Since 1983, the country has had a stable democracy despite the ongoing insurgency of the Kurdish minority. Economically, Turkey is placed at the high end of middle income countries with a gross national product per capita of $3,160 compared to $29,240 for the United States (World Bank, 2001).

Administratively Turkey is divided into 70 provinces. The provinces vary widely in the level of socio-economic development, in addition to ethnic structure. Given the strong unitary system of central government in Turkey, the provinces are merely administrative units, similar to neighborhoods or localities of American municipalities. Distributive decisions of the central government directly determine the quality and quantity of public services in each province. Thus, actions or non-actions of the central government have the potential of affecting the quality of life of residents in provinces.

The structure of health services in Turkey is very similar to the general structure of health services in other developing countries. The legal responsibility for the design and implementation of health policies in Turkey lies with the national Ministry of Health and Social Welfare (Akgun, Erdal, & Kisa, 2000). Most expenditures in health care are monopolized by the ministry. Consequently, a large proportion of health care facilities is owned by the government. The rapidly growing numbers of privately owned facilities are concentrated in urban areas (Atav, 1990).

The socialization of health services in the early sixties has remained mostly on paper and has failed to provide universal access to health services (Ministry of Health, 1998). Despite the centralized nature of the health care system, Turkish health sector has grown very complex in recent years. Many attempts to reform the health care system have resulted in a system that still fails to provide access to the disadvantaged citizens. For instance, the policy of the ministry to require health personnel for obligatory rural service has not worked. In the absence of any infrastructure or support systems, appointing physicians to remote rural areas has failed to generate expected positive health outcomes (Atav, 1990). The problem of access is best reflected in infant mortality rates. With an infant mortality rate of 38 per 1,000 live births, Turkey ranks the highest in its category of countries (World Bank, 2001). According to the World Health Organization, Turkey ranks 66th out of 191 countries in distribution and

Health policy studies in Turkey are rarely conducted. Rather, studies concentrate on epidemiological studies of subgroup of the Turkish population (Kalayci, Saraclar, Sekerel, Adalioglu, Kumcu, Egor, Bozer, & Tuncer, 1999; Koc, 2000; Sonmez, Basak, Camci, Baltaci, Karzyebek, Yazgan, Ertin, & Celik, 1999; Altuglu, Sayiner, Erensoy, Zeytinoglu, & Bilgic, 1998; Vicdan, Kukner, Dabakoglu, Ergin, Keles, & Gokmen, 1996), or on the descriptions of the Turkish health care services (Akgun, Erdal, , & Kisa, 2000; Ministry of Health, 1998; Ersoy & Sarp, 1998). A Conceptual Model to Analyze the Distribution of Turkish Health Services

The arguments of urban services distribution studies in the United States may provide a conceptual framework explaining the causes of differences in levels of health services among regions of a developing country. Theoretically, factors affecting decisions for the distribution of public services among Turkish provinces are analogous to those among the neighborhoods of an American municipality. Under its unitary system, policy making and administrative decisions in Turkey are monopolized by Turkish central governments. The governing party in Ankara has the potential power to allocate public services among provinces on the basis of any criteria it chooses, not unlike the power of local governments in America to distribute municipal services among neighborhoods. In this sense, the map of Turkey can be thought as the map of an American city, the borders of provinces representing the borders of neighborhoods of the city. Although the kind of services provided in these two contexts may be different, the theoretical distributive rules that apply to municipal governments of America apply also to the central government of Turkey. Therefore, taking the country of Turkey as a test case, the following hypotheses were extracted from the hypotheses of the urban service distribution scholars:

1. The socio-economic development hypothesis: This hypothesis is based upon the class preference and the power elite hypotheses. These hypotheses suggest that socio-economically advantaged groups receive higher shares of services than others. Since the unit of analysis is provinces in Turkey rather than neighborhoods of a city, this proposition is conceptualized in general socio-economic development terms.

The higher the socio-economic development of a province, the higher is the quantity of public services provided by the government.

2. The Political Reward Hypothesis: This hypothesis is based on the electoral interest hypothesis. It assumes that policy makers reward their voters or potential voters. Because policy makers in Turkey are the members of the incumbent governing party, the hypothesis becomes:

The higher the actual or potential electoral support provided from a province to the incumbent governing party, the higher is the quantity of public services provided by the government.

3. The Need Hypothesis: This hypothesis proposes that allocative decisions are made rationally based on the attributes of the provinces. Economic and technical concerns (effectiveness and efficiency) require that resources be allocated to the most needy provinces. Hence, when translated to the Turkish context:

The higher the need for public services in a province, the higher is the quantity of public services provided by the government.

4. Ethnic Preference Hypothesis: This hypothesis is similar to the race preference hypothesis. There are various minorities in Turkey including Kurds and Arabs concentrated in certain provinces. Theoretically, it could be70th in the overall performance of the health care system (WHO, 2000).assumed that the Turkish governments would tend to bring services to predominantly Turkish provinces, as would local governments to predominantly Anglo neighborhoods. A meaningful translation, therefore, of the race preference hypothesis would be:

The higher the proportion of ethnic Turks in a province, the higher is the quantity of public services provided by the government.

5. Incremental Hypothesis: This hypothesis suggests that in the short run, radical changes in the level of services do not occur and that the level of services in a province can be explained by the previous level of services in that province.

These propositions can be incorporated into a general conceptual model explaining the distribution of public services among provinces in Turkey as shown in Figure 2.


In order to demonstrate the testing of the model proposed in Figure 2, data were collected on all provinces primarily from the Statistical Yearbooks of Turkey for 1965 through 1983 during which Turkey had eight ideologically contrasting center right (3), center left (3), and military governments (2). Through a path analysis for each government period and a series of multiple regressions, direct, indirect, and total causal effects were calculated on selected variables. The dependent variable chosen for this demonstration was the change in the level of general practitioner physicians per 10,000 population (from one government to the other). Since change was already in the dependent variable, the incrementalism component of the model was not included in the analyses.

The final operationalization of the independent variables had to take into consideration the problem of multicollinearity. Initial correlation analyses of the variables in the model strongly indicated that race (as operationalized by the percentage of Kurds and Arabs in provinces); need (as operationalized by female literacy rates and population density); and socio-economic development variables (as operationalized by per capita taxes and per capita agricultural production for industrial use) were highly correlated with each other (r > .45). In the absence of alternative variables, the multicollinearity problem required the restructuring of the general model suggested earlier and the reoperationalization of the variables. For example, socio-economic development and need were combined into two indices of industrial development (per capita taxes and population density) and agricultural development (agricultural output and female literacy rate). Political variables were separated into two: percent votes for the ruling party in the last elections and volatility, a measure of the fluctuations in voting patterns in provinces between two elections. The resulting model is presented in Figure 3.

Through multiple regressions, path coefficients were calculated for each separate government period to assess direct, indirect, and total causal effects. As a demonstration of the process, path coefficients for one government period, the center left party that ruled Turkey between 1961 and 1965, are presented in Figure 3. The summary of all eight path analytical models is in Table 1. The reported numbers are total causal effects.


The patterns of distribution for general practitioner physicians indicate that in general, industrially and agriculturally disadvantaged provinces were favored, supporting the need hypothesis. Although the socio-economic hypothesis was supported, the expected direction of the relationship was reversed. There are two exceptions, however: the two military governments and the 1974-1975 center left government. During the military government of 1971, a major policy change took place and practitioner physicians were relocated to industrially and agriculturally developed provinces (total effects of .22 and .50 respectively). During the next military period, agriculturally developed provinces were favored in the distribution of practitioner physicians (.32).

Although the center left government from 1971-1974 reversed the trend to the advantage of agriculturally less developed provinces (-.07), only during the succeeding center right period, once again both industrially and agriculturally developed provinces were favored (-.14 and .35). All the subsequent governments followed the same pattern of reallocating practitioner physicians toward less developed provinces (total effects range between -.14 and -.65).

Political rewards hypothesis was partially supported. It appears that during coalition governments, provinces that voted for the leading party in the government received more services than other provinces (Party support= .23, .12 and .21). Volatility of votes did not follow a particular pattern.

Summary and Conclusion

This study reported an analysis of patterns of health policy distribution with the promise that such an analysis would address the fundamental questions of policy studies concerning the relationships between “who gets what, when, how much” and political, economic, and social factors. The answers to such questions may, on the one hand, assist policy makers to reduce unjust distribution of public services policy makers, and on the other hand, may make policy makers accountable for their allocative decisions in democratic contexts.

The conceptual framework chosen for this research was based on studies of municipal services distribution in American cities. Propositions from the municipal services distribution in American cities were incorporated into a general analytical model to examine the effects of theoretically relevant factors onthe distribution of health services in developing countries. This new model is general enough so that it could be applied not only to neighborhoods of a city, but also to any state, province or district that is influenced by the distributive decisions of a central government. The model included the components of socio-economic development (agricultural and industrial), percent votes for the ruling party in the previous elections, volatility of votes, need, incrementalism, and ethnicity.

Turkey, as the country of analysis is a developing country with serious social, economic, and political problems. It has a highly centralized unitary administrative system that has remained intact over many years, despite three military interventions and other political difficulties. Policymaking and administrative decisions in Turkey are monopolized by the central government. The quantity and distribution of practically all public expenditures are determined by the central government, not unlike the power of local governments to distribute municipal services among neighborhoods.

Within this framework, the changes in the distribution of general practitioner physicians among provinces were analyzed with path analysis from 1961 to 1983. Socioeconomic development, need, and political rewards hypotheses were supported for the distribution of general practitioner physicians.

The findings of this study suggest that the methodologies implemented in this study need to be extended to other service indicators and other policy areas to further test the conceptual model. Similarly, inclusion of other developing countries in the analyses would refine the model and strengthen the overall generalizability of conclusions.


Akgun, S., Erdal, R., & Kisa, A. (2000) Turkish health care system. Abstract from 128th American Public Health Association Annual Meeting Abstracts, Abstract Number: 4268. Washington, DC: American Public Health Association Publications.

Altuglu, L, Sayiner, A., Erensoy, S., Zeytinoglu, A., & Bilgic, A. (1998). Screening for human immunodeficiency virus type I and 2 in a Turkish blood donor population. International Journal of Infectious Disease, 4, 202-204.

Antunes, G. & Plumlee, J. (1977). The distribution of an urban public service: ethnicity socioeconomic status and bureaucracy as determinants of the quality of neighborhood streets. Urban Affairs Quarterly, 12, 312-332.

Atav, A. S. (1990). Politics of Health Care Policy in Turkey. Ann Arbor: University Microfilms.

Cingranelli, D. (1981). Race, politics, and elites: testing alternative models of municipal service distribution. American Journal of Political Science, 25: 664692.

Ersoy, F & Sarp, N. (1998). Restructuring the primary health care services and changing profile of family physicians in Turkey. Family Practice, 15, 576-578.

Kalayci, O., Saraclar, Y, Sekerel, B., Adalioglu, G., Kuyucu, S., Egor, G., Bozer, H., & Tuncer, A. (1999). Prevalance of asthma symptoms among Turkish Cypriot schoolchildren. Turkish Journal of Pediatrics, 41, 413-420.

Knippenberg, R., Alihonou, E., Soucat, A.; Oyegbite, K. Calivis, M. Hopwood, I., Niimi, R., Diallo, M. Conde, M., Ofosu-Amaah, S.(1997). Implementation of the Bamako initiative: Strategies in Benin and Guinea. International Journal of Health Planning and Management (suppl. 1), 12, S29-S47.

Koc, I. (2000). Determinants of contraceptive use and method choice in Turkey. Journal of Biosocial Science, 32, 329-42.

Lee, J. (1994). Policy type, bureaucracy, and urban policies: integrating models of urban service distribution. Policy Studies Journal, 22, 87-109.

Lineberry, Robert L. (1977). Equality and Urban Policy – The Distribution of Municipal Public Services. Beverly Hills: Sage.

Macfarlane, S., Racelis, M., & Muli-Musiime, F, (2000). Public health in developing countries. The Lancet, 356, 841-846.

Makinen, M., Waters, H., Rauch, M., Almagambetova, N., Bitran, R., Gilson, L., McIntyre, D., Pannarunothai, S., Prieto, A.L., Ubilla, G., & Ram, S. (2000). Inequalities in health care use and expenditures: empirical data from eight developOing countries and countries in transition. Bulletin of the World Health Organzation, 78, 55-65.

Ministry of Health (1998). Public Health Profile In Turkey: “Health Sector” Preliminary Report. Ankara: Saglik Bakanligi Yayinlari.

Rich, R. (1982). The political economy of urban service distribution. In The Politics of Urban Public Services, Ed. Richard Rich. Lexington: Lexington Books.

Roemer, M. (1991) National Health Systems of the World, The Countries (vol 1). New York: Oxford University Press.

Roemer, M. (1993) National Health Systems of the World, Volume 2, The Issues. New York: Oxford University Press.

Sanger, M. (1982). Academic models and public policy: the distribution of city services in New York. In The Politics of Urban Public Services, Ed. Richard Rich. Lexington: Lexington Books.

Smith, D., & Bryant, J. (1988). Building the infrastructure for primary health care: An overview of vertical and integrated approaches”. Social Science and Medicine, 26, 909-917.

Sonmez, H., Basak, 0, Camci, C., Baltaci, R., Karazeybek, H., Yazgan, F, Ertin, I., & Celik, S. (1999). Journal of Hum Hypertension, 13, 399404.

Statistical Yearbook of Turkey, 1985, 1981, 1980, 1975, 1970, 1965, 1960. Ankara: Basbakanlik Devlet Istatistik Enstitusu.

Vicdan, K, Kukner, S., Dabakoglu, T, Ergin, T, Keles, G., & Gokmen, 0. (1996). Demographic and epidemiological features of female adolescents in Turkey. Journal of Adolescent Health, 18, 54-58.

WHO (2000). The World Health Report 2000 Health systems: Improving performance. Geneva, Switzerland: The World Health Organization.

World Bank. (2001). Turkey at a glance [On-line]. Available: http://www.worldbank.org/data/countrydata/aag/tur_aag.pdf [Accessed 2001, January 16].

A. Serdar Atav, Ph.D., Assistant Professor, Decker School of Nursing, Binghamton University, Binghamton, New York.

Copyright Riley Publications, Inc. Center for the Study of Multiculturalism and Health Winter/Spring 2001

Provided by ProQuest Information and Learning Company. All rights Reserved