The International Forum for Social Sciences and Health, World Congress:
"Health Challenges of the Third Millennium," Istanbul, August 21-26, 2005
Is Health Care Reform Sufficient to Raise Population Health in Turkey?
Authors: Zeynep Angin and Frederic C. Shorter
In Turkey, the health care system and insurance entitlements are seriously inequitable of access. Issues concerning the ownership, management, and conduct of health care services are high on the Turkish political agenda. Major changes are being proposed, but are not yet sorted for efficiency, consistency, and justice. There is a long history of attempts at reform, and whether new moves will actually take place and succeed is unclear. In this paper, improvements in the poorly functioning healthcare system are seen as necessary, but a warning is sounded. Reforms may not be able to raise the level of population health more rapidly than the slow trend of the past. Turkey's situation shows that policies affecting social environments, particularly in the domains of employment and education, have important consequences for population health.
The strategy of this paper is to look at past trends and current levels of population health with appropriate indicators of mortality. We find that mortality in the early years of life, including reproductive health, is a crucial determinant and indicator for the population's health generally. This leads us to consider how child health, with its mortality rates, is produced in households where children live. Initially, families must acquire resources including but not limited to healthcare entitlements that they can use in the production of health. They transform their resources into health inputs by their own homemade actions, by purchase (using entitlements when available), and by reaching out to markets and the community around them for what they perceive as needed to feed, clothe, and give healthful care to family members. While reaching out, they may encounter obstacles related to social inequalities and the poverty of a particular community environment.
We use data from the Turkish Demographic and Health Surveys to show how the continuous process of transforming family resources into health-inputs is accomplished in Turkey. These data lead us back to a consideration of the places where families live and acquire resources which have major determinative effects on their options. Thus, the process of improving health must inevitably address more of the society's structures than the health-care system itself.
Turkey entered its republican era (1923-) with serious shortages of adults, especially working age males, due to the Balkan wars, First World War, and Independence war just passed. Its population was a mere 13 million, down one-fourth from its earlier size. There was a strong desire, officially and privately, to rebuild families and raise the population as quickly as possible. Fertility climbed rapidly and was then sustained at high levels, but health conditions were very poor. The earliest measurements of general mortality show that life expectancy at birth was no higher than about 35 years during the 1930s. In 1945, the infant mortality rate was still 274 per thousand births (earliest robust measurement) and total fertility was above six children per woman nationally, though lower in the largest cities. A steady transition followed with improvements in mortality and declining fertility after 1950, so that Turkey is today a country with close to a replacement level of reproduction and a population of 70 million. [Note: Sources for the demography of Turkey include, most notably, Turkish State Institute of Statistics (SIS) 1995, produced by a team of demographers at SIS. Also see: Shorter & Macura (1983); Shorter (1985); SIS censuses 1927--2000 (various dates); and a series of five-yearly national household surveys, 1968 - 2003 (Hacettepe various dates). For the most recent estimates, we use the 2000 census (SIS 2003), and the 1998 & 2003 Demographic and Health Surveys (Hacettepe 1999, 2004), with interpolations to position data on the year 2000 throughout this paper. Vital registration data are not used due to incomplete coverage and inaccuracies in such matters as age and cause-of-death (Bulut et al. 1992).]
Projections suggest that population momentum from past trends and levels of fertility will work themselves through to a maximum of 90 million people, more or less, by 2050 depending on the future of fertility, mortality (especially early-age), and international immigration/emigration. Currently, Turkey is both a receiving country—especially of ethnic Turks and economic/asylum migrants from neighboring countries—and also a sending country mainly of economic migrants and families, with a net annual outflow (1995-2000) of no more than 54,000 per year (U.N. 2002). If admitted to the European Union, net emigration would undoubtedly increase but with limitations on the side of European demand impossible to predict. [Note: For international migration information , analyses of the demand side for labor in the European Union, and simulations for the Turkish response, see TAPV (2004: 105-202).]
Figure 2. Producers (20-54), youth (0-14), and elderly (65+): 1935-2025
[Notes: Data from Censuses to 2000; extended by a conservative projection to 2025 in which nrr falls to 1.0 by 2005, but no further, e0 (both sexes) rises only to 71.0 years, and no net population change occurs by international migration.]
By the year 2000 the population under age 15 stopped growing, births are constant at 1.4 million per year, the producer-age population is increasing rapidly, and the elderly over 65 are growing and picking up speed. After 2025, all the future growth of population will be among the elderly, since the producer population will stop growing—unless, of course, there is an upward shift in fertility which no one predicts at present.
From the standpoint of educating young people, policies can now to concentrate much more on quality and coverage of all places and social classes, rather than on chasing population growth. The great size of the potential producer population is seen as an "opportunity" for more production, but expansion in employment opportunities has not kept up; each economic recovery after a crises fails to add to visible employment. It seems that productivity growth, not employment, explains the remarkable capacity of the country to continue gaining in GDP per head. Corollaries in the Turkish case are that income distribution becomes progressively worse, the fiscal means to sustain health and education services have collapsed, and serious poverty is an increasing problem (Keyder 2004:75-77; Senses 2003).
As the number of elderly grows, there is, in principle, sufficient younger-age population to support them for the time being—serious imbalances will emerge later in the century; meanwhile, family systems that channel resources from young to old have weakened; and formal systems of social security offer poor coverage, as well as financial crises of their own. Since health maintenance and needs for treatment increase with age, inadequacies of health insurance and the health-care system are felt especially among the elderly and, of course, among the poor irrespective of age. These issues are traceable only in part to the demography of the country; they also reflect the history and current stance of social policies with regard to population health.
The measure we use for "population health" is expectation of life at birth. We also use the infant mortality rate as our preferred index of the life table; among all alternatives, it is the one most often measured successfully, directly or indirectly, with Turkish census and survey data. [Note: Technically, we note that the Turkish life table is constructed from observed values of mortality at young ages (positioned on 2000) and is extended to higher ages by the 'East' model life tables - proven in earlier studies to approximate observed patterns in Turkey of age-specific death rate rates (SIS 1995: 41-42).] Comparing Turkey to a "gold standard" represented by Sweden, we might argue that Turkey's deficit measures an arguable goal for long-term improvement.
Turkey's expectation of life in 2000 is 69.6 years for both sexes combined. For the period 1998-2002, Sweden has an expectation of life of 80 years (Statistics Sweden, 2004, Table 84). In Figure 3, we plot the decrease in population survivors that occurs as life is lived from birth to the end of life. These are survival curves (lx values of the life table). Visually the difference between the upper and lower curves measures the loss of life-years under the health regime in Turkey compared with that of Sweden. Numerically, it is an average of 10 person-years for individuals born in Turkey relative to those born in Sweden. Or, more graphically, we can estimate the difference in annual deaths under the two country's regimes. Following a suggestion by Tord Høivik (1977), we calculate hypothetical deaths in Turkey by the Swedish life table's death rates in 2000, and do the same for Turkey using Turkish life-table rates. The result is more than double the number of deaths in Turkey than there would be under a Swedish regime: 430,000 actual deaths in Turkey vs. a hypothetical potential of 208,000 deaths—an excess of 222,000 Turkish deaths in the year 2000.
Age-specific death rates in Sweden are much lower than in Turkey. As life progresses, by age 50-54 Turkish females have a death rate of 5.1 per thousand, Swedish females 2.6 per thousand; and at age 70-74 the comparison is 42.5 to 17.2 respectively.
Notice that the Turkish survival function drops at the beginning of life whereas the Swedish curve does not. This is due to an infant mortality rate in Sweden of only 3 deaths per thousand births, compared with 33 per thousand in Turkey. To achieve better population health in Turkey, it is necessary to raise the whole Turkish survival curve. Strategically, it may be argued that the most effective way to do this is to focus on the reproductive care of mothers and the achievement of healthy childhood. This is valuable not only at the beginning of the curve; it also has a powerful effect on risks later in life according to Hertzman (1999) and Mosley & Gray (1993). Thus, increases in early childhood survival lift both the position of the curve and postpone the inevitable decline of the curve later in life.
Improvement of reproductive health and child survival is signaled demographically by reductions in fertility, mortality, and the net reproduction rate. We chart this history for Turkey in Figure 4. Some (technical) understanding is necessary to use the chart. To measure the intensity of the reproductive process, we use the gross reproduction rate (GRR)—meaning births of girls to mothers, which is close to one-half the total fertility rate—to represent the trend in births. It is the upper curve in Figure 4. Then, we deduct the expected deaths of girls between their birth and the time of their own reproduction as mothers later in life. The result is the lower curve which shows how much reproduction is anticipated by the birth of these girls—only those girls who survive can be the mothers of the next generation. The lower curve is named the net reproduction rate (NRR). Over time, improvements in survival bring the two curves closer and closer together—think mainly of reductions in infant and child mortality, but also reductions in other causes of female mortality including maternal mortality (deaths directly associated with childbearing). In Turkey, 49 girls per thousand currently die between birth and their mean age of childbearing (27.2 years). The comparable figure for Sweden is 7 girls per thousand.
By the year 2000, the distance between the GRR and NRR narrowed, which indicates reduced mortality among children and women up to the age of reproduction. Visually, the gap in Figure 4 by 2000 may appear to be small but it is seven times as great as that for Sweden.
[Note: Our own interpolations to position data points comparably. Data from sources in Footnote 1. For the year 2000, we estimate tfr=2.36 and use the Turkish life table with an imr=33 per thousand births, from which we derive grr=1.15 and nrr=1.09 girls per woman respectively.]
The picture of historical change in reproduction shows that the commitment of mothers and fathers to bring forth and support children (reproduction) was very inefficient in past times—many births and many losses. Among the oldest generation still alive, the common saying was, 'One child, no child' (Bir çocuk, hiç çocuk), while today most parents expect their children to live. Thus, the bio-social effort required to gain a particular net reproductive result is much less than formerly. The worrisome point is that the rate of decline of both curves in Figure 4 has slowed down. While this might be interpreted as an asymptotic approach to a magical net reproduction rate of unity, we find in the data strong indications that there are impediments to further progress rooted in government social and economic policies, not limited to health care systems. All affect families and vary greatly across different regional and rural-urban jurisdictions.
Modeling the household production of health
We start with a household-production-of-health model and estimate the importance of some of the social resources that count as determinants of child mortality. Our statistical model is implemented with a national sample of data collected by the 1998 Turkish Demographic and Health Survey (Hacettepe 1999). We search for explanations of differences in mortality outcomes for children born to households across the entire national sample. We model these variables directly in relation to mortality, not restricting the analysis to determinants of individual-level health behaviors that might be intermediate determinants of child mortality.
Intermediate variables do, indeed, have effects on child mortality outcomes in specific individual contexts, and various studies have been conducted in Turkey to look at them. A good example would be Gürsoy's (1992, 1995) work on family dynamics and child survival in poor households of Istanbul. Behind these variables stand the social determinants of health which set in motion the overall process of producing population health, and it is these latter variables that we examine in our modeling.
A child mortality index is constructed for children born of women married less than 15 years at the date of the survey. [Note: The methodology was first seen in Farah and Preston (1982) and explained in general terms for individual-level studies of covariates of mortality in Trussell and Preston (1982).] By imposing the 15-year limitation, relatively recent mortality experience can be associated with the independent variables. The 1998 data set gives a mean level of the infant mortality rate of 47 per thousand centered on 1995, with a declining time trend in the sample population before and after that date. The earlier data set collected in 1993 gave a mean level of 59 per thousand centered on 1990 with similar trend.
Variation in local environments is particularly important in a national study of mortality. Communities differ in the availability of safe water and sanitation, clean air, and a host of other local conditions that affect health. These conditions are "givens" that can not be manipulated by individuals, and so they do not enter the household model directly. They do vary, however, by communities, for example along the rural-urban dimension and by different regions of the country. We include in the model two contextual variables standing for environment as controls; i.e., rural-urban and East region against the rest. For statistical work, the country is usually divided among five regions: West, South, Central, North, and East. Nineteen percent of the population lives in the East.
We estimate a regression model having the following variables: household asset index (proxy for income), mother's education, health insurance entitlement, whether the mother works or not, and two controls indicative of community of residence—rural-urban and East region against the rest of the country.
The household asset index is constructed from a list of assets (e.g. durables) and housing characteristics (e.g. sleeping density) in each household. Our asset index is based on whole-country weights from the World Bank data set for the 1998 Turkish Demographic and Health Survey (kindness of D. R. Gwatkin). We excluded assets related to water and sanitation because they are themselves intermediate determinants of mortality and would spuriously increase the correlation of the asset index with mortality.
Mother's education is included in the model, but not father's education. On the basis of preliminary work, we decided not to include the husband's education for two reasons: it correlated highly with the wife's education—undoubtedly because educational levels are part of the sorting that brings couples into a marriage—and it interacted strongly with the asset index. Mother's education is not meant to signify that the mother does everything or controls everything related to the child's well-being. Husbands are no doubt involved. For statistical work, however, mother's education is an excellent indicator of her social position and family resources for mothering in the family.
Whether the woman was working during the year of the survey—31 percent were—was included to learn whether she was contributing by her own earnings to child welfare. Although the results were positive, statistical significance was low. A variable for health insurance was also included—simply yes or no for some kind of coverage. We thought this would be an indicator of access to health care. Among all the families with young children, 57 percent had some type of coverage in 1998 (Table 1).
Table 1. Multiple Regression: Turkey Demographic and Health Survey, 1998
Dependent variable is Child Mortality Index: Mean = 1.000 ± 2.64
|Independent variables||Mean ± SD||ß||ß*||p|
|Household asset index||-.045 ±.81||-.251||-.077||.002†|
|Women's education in years||5.0 ±3.5||-.037||-.049||.038†|
|Woman works regularly or seasonally (0/1 variable)||.306 ±.46||-.080||-.013||.478|
|Health insurance: Family has some entitlement (0/1 variable)||.565 ±.49||+.038||+.007||.719|
|Urban (not rural)||.642 ±.48||-.227||-.041||.048†|
|East region (not elsewhere)||.205 ±.40||+.259||+.039||.043†|
N = 2969 ever-married women, marriage durations < 15 years, having 6395 children ever born.
* The symbol ß* is the standardized regression coefficient. It is independent of the units in which the variable is measured.
† Denotes acceptable statistical probabilities (p<.050).
The procedure weights women by the number of children, because children not women are at risk of mortality; and then uses number of women for tests of significance.
The child mortality index is computed as the proportion of an individual woman's ever-born children still surviving, divided by the expected proportion for all women in her marriage duration group. As a result, averaging over any duration group or over the entire sample, the mean will be 1.0. Women below 1.0 have better than average survival of children and women above 1.0 have below-average survival among their children.
Discussion of Results
From the standpoint of equity in paying for health care, insurance certainly is helpful. However, the regression model shows that having insurance has a very weak statistical effect on child mortality and a high probability of being due to chance. Other studies also find that health insurance per se may not do the trick, because "successful and effective utilization of [health] care may depend [instead] on class-related contextual factors of an individual's daily life" (Adler, Boyce, Chesne, Folkman & Syme, 1993,
In Turkey, quality of care does indeed depend on class-related factors. According to an Istanbul study in public hospitals, "when the patients are from low-education groups, with low skills of Turkish language, or from other ethnic groups (e.g., gypsy [Note: The ethnic group is roman (Roma), usually denoted in colloquial Turkish discourse as çingene (gypsy).]), they are treated badly to the extent of discrimination. (Turan, Ortayli, Nalbant & Bulut, 2003,
Insurance is tied mainly to formal employment. Since the late 1990s, the unemployed poor became eligible for a "Green Card" entitling them to health care at state institutions (excluding university hospitals), but the process to obtain a card is a struggle of personal contacts, influence, and persistence. Thus, the insured public obtains health care through an extremely inequitable insurance system, and must typically rely on making private payments for some portion or all of the services they receive. Moreover, even in 2002 about one-third of families had no health insurance, though estimates vary (Mollahalilogulu, Kartal, Eristi, & Özbay, 2004).
Studies by Çelik and Hotchkiss (2000) using the 1993 TDHS data set show that having health insurance may influence positively and significantly whether and where maternity care is sought, but they do not directly confront the issue of whether the insurance system, as it stands, has a beneficial effect on child survival.
Household asset index
The household asset index, which may be understood to be a proxy for income, has a strong and statistically significant relationship to child survival at the household level. Inequalities in the income distribution are at the root of this particular problem. Poverty is to a great extent spatially delineated. Research on the poverty zones of cities shows that the neighborhoods which house many newcomers (gecekondu)—over generations by now—were expected to be transitional, but instead many are now permanently depressed and segregated (varos). Social exclusion and the shame of being poor poses a particular problem (Isik & Pinarcioglu, 2001,
A recent study cites the Lorenz curve for household income in Istanbul showing Gini coefficients rising over time: 1978, 1987, 1994 showing coefficients of 0.38, 0.43, and 0.58 respectively. (Behar, Koksal, Güvenç, Isik & Ercan, 1999:
While most of the research on poverty zones is focused on cities, inequalities are also serious in large, predominantly rural areas of Anatolia and the East where long-standing social relations are still dominated in many instances by landed capital in few hands, and the unease is shown by continuous out-migration of surplus labor.
The demonstration that mothers' education is a significant determinant of child mortality justifies a closer look at the power of education prior to adjusting for other factors. We saw in Figure 4 that both declining fertility and reductions in early mortality produced a decrease in net reproduction, which we consider is healthier reproduction. However, that history was substantially different among mothers of differing educational backgrounds. Table 2 breaks out the information in Figure 4 by education of mothers. Lack of access by less educated women to whatever it takes to reach more efficient reproduction evidently has negative consequences. Thus, progress toward well-controlled and efficient reproduction is held back "from the bottom."
Table 2. Educational Inequalities Among Mothers Bearing
the Next Generation of Children, Turkey in 1997
|Education of Mothers||Fertility (GRR)||Net Reproduction (NRR)|
|No education / primary incomplete||1.90||1.71|
|Primary completed (5 years)||1.24||1.17|
|Secondary completed and over||0.79||0.75|
Infant mortality rates are strong indicators for the measurement of inequality, partly because they are more often measured credibly, and also because they are indicative of health across the whole spectrum of reproduction through childhood. We present Table 3 for Istanbul, using mothers' education levels as the indicator of inequality of family resources. Why Istanbul? The process of change over time can be seen well with available data, and it would not be greatly different for the country as a whole. For dates before 1990, detailed census samples are available for Istanbul but not the whole country (Census 1% sample 1970 & 3% sample 1985). They support an assessment of change over a period of 30 years up to 1997 (Census of 2000).
Table3. Infant Mortality by Level of Mother's Educational Attainment
ISTANBUL, 1967, 1982, 1997 (15-year intervals)
|Infant Mortality Rate 1q0
(Deaths per 1000 Births)
|Children Born to Mothers
(Percent of All Children)
|Level of Mother's Education||Reference Dates||Census Dates|
|Below primary (some schooling possible, but mostly illiterate)||142||116||59||51||26||11|
|Primary school (5 yrs) completed||81||83||38||39||59||67|
|Middle (orta) completed||59||58||32||5||6||8|
|Secondary (lise) completed or higher education||37||34||30||5||9||14|
This table gives a great deal of information about the considerable decline in infant mortality during the last 30 years, from 110 per thousand to 39 per thousand overall, almost one-fourth of what it was earlier. For those women with relatively higher education—secondary and above—the improvement in child survival was modest—only 7 points on 37 points (18 percent)—which tempts one to conclude that women in the "upper" education groups were knowledgeable about caring for their children—and had the resources and contacts—long before such success was enjoyed by lower classes of women.
In the lowest education group—mostly illiterate—there were large gains over time, with infant mortality declining from 142 to 59, a decrease of 58 percent. The overall gain through time was additionally due to the changing educational ranks of mothers. The right-hand panel of the table shows that there was a very substantial shift from children being born of mostly illiterate women, to being born of primary and higher education women. Thus, their life chances improved dramatically. That shift must be credited to the increase of education—and the underlying improvement in family resources that this implies—at the lower ranks for women. Proportions of women in the lowest rank bearing children fell to 5 percent over the 30 years (not shown in table). Even though these mothers have more children than more educated women, the proportions of children at risk in this lowest educational group fell to 11 percent of all children. Thus, improvements in the educational resources of women—standing for all education and life opportunities of her family—are a primary source of better health for children.
Rural urban contexts
The ecological factors that affect household production of health vary by the place where the household resides. If the supply of all that is needed for the production of health were equal everywhere in the country, then differences in health outcomes would be attributable to family demand factors such as education, income, and health insurance (entitlements)—factors explored by the model. However, the environments of households are patently unequal. One statistic that signals this problem is the rural-urban control variable, because it says that households of similar resources are less successful in producing health in rural than urban environments. The relevant environmental conditions include water and sanitation, transport access to markets and public institutions, employment opportunities, schooling, and healthcare services.
Keeping track of development in the villages of Turkey, where 35 percent of the population live, the Turkey State Planning Organization (2001, ch. 12-7) states that by now there is road access to every village and an electricity connection; however, only 7.5 percent of villages have a sewage system, 31 percent piped water, and overall only 75 percent have some form of healthy drinking water. Such environmental disadvantages make the process of producing health in children more difficult for households, confirmed by the negative effects of rural residence in the model.
Aiming to increase the supply side of health services, the government subsidizes private hospitals and private high-tech services (since about 1990), but the uneven spatial distribution has gotten worse. Thomson and Saka (2003,
Regional contexts: East region against the rest
The strong showing that living in the East region is a negative influence on child mortality stands out separately from the basic social determinants. While the East is more rural, has lower income, and lower educational attainments than other regions, those factors are already in the model to account for variance in child mortality. The question is why it is additionally disadvantageous to be in the East. We estimated the same model with the 1993 survey data and found no East effect ( p = .855). For other independent variables, results were similar, not different, for 1993 and 1998.
What changed in the East between the two surveys? As Kiris ci & Winrow (1997, pp.122-136), tell the story, between 1992 and 1994 (only 2 years), 5,210 schools were closed. By 1995, security forces acknowledged that 2,253 villages had been evacuated while the actual numbers were much higher. Medical doctors registered as on duty were 4 per 1000 population in 1990 (as compared with 12 in the West) and declined thereafter. There was extensive migration to, and crowding in city centers of the region. Thus, factors of civil violence and disruption during and following the civil war (1984-1998) were sufficient to have powerful effects on mortality. Furthermore, it is notable that the reference period for mortality in the 1998 survey data was the 15 years prior to 1998, which happens to correspond with the period of greatest trouble in the East. This correlates with the negative events just enumerated, thereby showing a temporal link. Disadvantage in the East is also measured eloquently by the successive Turkish Demographic and Health Surveys (Hacettepe University Institute of Population Studies, 1994,
We conclude that the East community effects on mortality and the rural residence effects, both seen in the statistical analysis, are important consequences of inequalities in Turkey. Politico-administrative policies that take the form of strong environmental differences across communities are a pathway from unequal income to unequal population health. As Lynch and Kaplan (1997, p.212) explain, ". . . inequitable income distributions may be associated with policies and social processes that systematically under-invest in human, physical, health and social infrastructure; and this underinvestment may have health consequences." Of course, causation also runs the other way creating a vicious cycle—under-investment limits the ability of population in deprived localities to equip themselves for employment and obtaining incomes. For the most part, the issue is not one of failure of the "victims" to use their resources well, but of their inability to obtain resources in the first place. These are structural and community-wide supply-side constraints.
"Fixing" the health-care system and so making it available for everybody effectively and efficiently is but one part of the solution. Variations in health among communities, not limited to the simple rural-urban divide, but also by social and economic classes, are well documented both by statistics and in-depth qualitative studies (Turkey State Planning Organization, 2001, Ch. 7; Akder, 2002; Chawla, 2003). Policies that create more egalitarian economic and social conditions are also necessary. Commenting on political systems, Coburn (2004) shows that countries with social democratic forms of welfare regimes have better population health than the ones relying heavily on market forces—a challenging thought when we note that Turkey has been moving in the latter direction.
Turkey needs structures to take responsibility for impoverished segments of the population. What the society can do to spread and enhance both employment and educational opportunity is a good place to begin. Furthermore, the social services and social transfers could be increased and made available to everyone according to appropriate criteria. Charity is not the answer. As Bugra and Keyder (2003:50) say, citizen rights that respect all will do better for the society.
Adler, N.E., Boyce, W.T., Chesne, M.A., Folkman, S., & Syme, L. (1993). Socioeconomic Inequalities in Health: No Easy Solution. JAMA, 269(24), 3140-3145.
Akder, H. (2002). Human Development Report: Turkey 2001
Behar, C., Koksal, S., Güvenç, M., Isik, O., & Ercan, H. (1999). Turkiye'nin Firsat Penceresi: Demografik Donusum ve Izdusumleri,
Bulut, A. G. Gökçay, O. Neyzi, and F. Shorter (1992), "Perinatal, Neonatal, and Under-five Mortality in Istanbul Based on Representative Samples of Burial Records," in Ties Boerma, ed., Measurement of Maternal and Child Mortality, Morbidity and Health Care: Interdisciplinary Approaches. Liege Belgium: International Union for the Scientific Study of Population, Ordina Press.
Bugra, A., & Keyder, C. (2003). New Poverty and the Changing Welfare Regime of Turkey
Çelik, Y., & Hotchkiss, D.R. (2000). The Socio-economic Determinants of Maternal Health Care Utilization in Turkey, Social Science & Medicine, 50, 1797-1806.
Chawla, M. (2003). Turkey: Reforming the Health Sector for Improved Access and Efficiency, Vol.1: Main Report
Dervis, K., Gros, D. Öztrak, F., Bayar, F., Isik, Y. (2004). Relative Income Growth and Convergence, Turkish Policy Quarterly, 3(4), 1-24. URL:www.esiweb.org/turkey/tpq2004.php
Frantz, D. (2001). Private Money Enlivens University Education in Turkey, New York Times March 13 2001. New York.
Gürsoy, A. (1992). Infant Mortality: a Turkish Puzzle? Health Transition Review, 2(2), 131-149.
Gürsoy, A. (1995). Child Mortality and the Changing Discourse on Childhood in Turkey. In E.W. Fernea (Ed.), Children in the Muslim Middle East
Hacettepe University Institute of Population Studies (1994). Turkey Demographic and Health Survey 1993, Associated 1993 data set downloaded from www.measuredhs.com with permission. Ankara: HIPS.
Hacettepe University Institute of Population Studies (1999). Turkey Demographic and Health Survey 1998, Associated 1998 data set downloaded from www.measuredhs.com with permission. Ankara: HIPS.
Hacettepe University Institute of Population Studies (2004). Turkey Demographic and Health Survey 2003, Associated 2003 data set scheduled for release in late 2006. Ankara: HIPS.
Hertzman, C. (1999). Population Health and Human Development. In
Høivik, T. (1977), "The Demography of Structural Violence," Journal of Peace Research,
Isik, O., & Pinarcioglu, M.M. (2001). Nöbetlese Yoksulluk: Sultanbeyli Örnegi, Istanbul: Iletisim Yayinlari
Kirisci, K., & Winrow, G.M. (1997). The Kurdish Question and Turkey: An Example of a Trans-state Ethnic Conflict London: Frank Cass
Keyder, Ç. (2004). The Turkish Bell Jar. New Left Review, 28, 65-84
Lynch, J.W., & Kaplan, G.A. (1997). Understanding How Inequality in the Distribution of Income Affects Health. In I. Kawachi, Kennedy, B.P., & Wilkinson, R.G. (1999). The Society and Population Health Reader: Volume I, Income Inequality and Health . New York: The New Press.
Mollahaliloglu, S., Kartal, M., Eristi, H.E., & Özbay, H. (2004). Türkiye Saglik Hesaplari Çalismasi, Türkiye Iktisat Kongresi 2004
Mosley, W.H., & Gray, R. (1993). Childhood Precursors of Adult Morbidity and Mortality in Developing Countries: Implications for Health Programs. In J.N. Gribble, & S. Preston (Eds.), The Epidemiological Transition: Policy and Planning Implications for Developing Countries. Washington: National Academy Press.
Senses, F. (2003). Economic Crisis as an Instigator of Distributional Conflict: The Turkish Case in 2001. In Z. Önis, & B. Rubin (Eds.), The Turkish Economy in Crisis. London: Frank Cass.
Shorter, F.C., & Macura, M. (1983). Trends in Fertility and Mortality in Turkey 1935-1975, Committee on Population and Demography, National Research Council
Shorter, F.C. (1985). The Population of Turkey after the War of Independence. International Journal of Middle East Studies, 17, 417-441.
Statistics Sweden (2004). Statistical Yearbook of Sweden 2004 Stockholm, Sweden
TAPV: Turkish Family Health and Planning Foundation. (2004). Population Challenges, International Migration and Reproductive Health in Turkey and the European Union: Issues and Policy Recommendations. Istanbul: TAPV.
Thomson, S. & Saka, Ö. (2003). Health and Health Care in Turkey. Euro Observer,
Turan, J.M., Ortayli, N., Nalbant, H., & Bulut, A. (2003). Istanbul'da Üç Hastanede Gebelik ve Dogum Hizmetlerine Yakindan Bakis, Istanbul: Kadin ve Çocuk Sagligi Egitim Arastirma Birimi, Istanbul Tip Fakültesi.
Turkey State Institute of Statistics (1995). The Population of Turkey, 1923-1994: Demographic Structure and Development with Projections to the Mid-21st Century
Turkey State Institute of Statistics (2003). 2000 Census of Population: Social and Economic Characteristics of Population. Ankara: One national volume and 81 province volumes.
Turkey State Planning Organization (2001). Eighth Five-year Development Plan. Ankara.
United Nations Population Division (2002) International Migration 2002. New York: UN Publication.
A. Bulut, S. Erder, J. M. Turan, and the Turkish Study Group of the Jackson School of International Studies, University of Washington, provided valuable comments.