Residential segregation of socioeconomic variables and health indices in iran.

BACKGROUND: Measures of segregation are essential tools for evaluation of social equality. They describe complex structural patterns by single quantities and allow the comparison of inequalities over time or between residential places. In many countries, patterns of residential segregation are well described (e.g., South Africa, Great Britain, United States of America). In this study, for the first time in Iran, we measured residential segregation for some socioeconomic and health variables and described their pair wise correlation.
METHODS: We measured evenness dimension of segregation by generalized dissimilarity segregation index and information theory index and its ordinal equivalent for some determinants of socioeconomic status and health variables using data of last national census in Iran. Segregation indices were computed for 31 socioeconomic variables and four health indices.
RESULTS: All the provinces were in the category of low segregation for individual and family disability and death of at least one offspring of mother, but for infant mortality half of the provinces were moderately or highly segregated. For some of socioeconomic variables, many provinces were in the category of moderate, high, or extreme segregation. There was significant correlation between segregation of heath indices and some socioeconomic variables.
CONCLUSIONS: Correlation of segregation of determinants of socioeconomic status with segregation of health indices is an indicator of existence of hot zones of health problems across some provinces. Further studies using multilevel modeling and individual data in health outcomes at individual level and segregation measures at appropriate geographic levels are required to confirm these relations.

INTRODUCTION

Segregation is the degree to which groups are separated from each other. It describes the uneven distribution of population groups defined by a special characteristic among the subunits of a social space.[1] Population groups can be defined as subgroups of race/ethnicity, socioeconomic status, education, gender, age, etc., Social space can be a geographic social space like city, province, municipality district, or school, or an economic social space like labor market or other groups.

Segregation measurement has a longstanding history in epidemiology and social science. In 1928, Ernest Burgess Published his first paper on this subject. For many years, there were debates about the measurement of segregation. Since 1955, for 20 years, index of dissimilarity was the standard index of residential segregation until it was criticized by Charles Cortes and some new indices were invented.[2] In 1988, Massey and Denton did a systematic analysis on 19 segregation indices which they extracted from literature and claimed that segregation is not a one-dimensional construct. They described five dimensions for residential segregation named as evenness, exposure, clustering, centralization, and concentration. Evenness and exposure are aspatial dimensions, whereas clustering, concentration, and centralization are spatial dimensions of segregation and require information on the locations and areas of census tracts to compute.

Measures of segregation are the essential tools for evaluation of social equality. They describe complex structural patterns by single quantities. These measures allow us to compare inequalities over time or between residential places.

Segregation matters because it is related to different proximities of individuals to important resources. These resources might be institutional like health care facilities and hospitals, job opportunities, different types of schools, labor markets; social like social networks, cultural capital; contextual like different proportions of individuals with different educational levels, married versus single and divorced individuals, people of different socioeconomic status, and even environmental and social hazards like lead poisoning, pollution, higher violence, and crime rates.[134] Studying segregation is important mostly because of its consequences. Neighborhood residential segregation has been considered an important factor in health disparities.[5] Many recent studies have documented association between segregation and health outcomes like overall health,[6] infant mortality,[789] tuberculosis,[10] cardiovascular disease,[4] high body mass index,[11] and non-health outcomes like access to employment and education inequalities,[1213] early adolescent sexual activity,[14] Black homicidal rates,[1516] and health service use.[17]

In many developing and underdeveloped countries, the patterns of residential segregation are well described, e.g., South Africa,[18] Great Britain,[19] Singapore,[20] United stated.[21]

Previous literature on segregation has some limitations. First, most of the work in measuring segregation was limited to measuring racial segregation and segregation of other aspects of residential place like the demographic characteristics of residents, and the structural features of neighborhood or the socioeconomic characteristics of people living in a neighborhood have been neglected so far. Second, so far segregation has been measured mostly for dichotomous variables and few papers have used multigroup indices of segregation. Third, measures of segregation have been reported mostly for western contexts and few articles exist in the literature about these measures in eastern developing countries. Considering the above-mentioned limitations in the existing literature and with the help of recent advances in measuring segregation for multi-group and ordinal variables, we are going to answer the following questions. How segregated are the Iran provinces with regard to socioeconomic factors and some health indices, and is there any association between these indices?

METHODS

Segregation indices

Residential segregation indices measure distribution of different population dimensions across residential units within larger geographic areas. Thus, to measure segregation, one must first define the larger area and its subunits. Second, the population dimension along which one wishes to measure segregation must be defined since measuring segregation among two, multiple unordered, or multiple ordered categories of a variable needs a different set of measurement tools. Third, the dimension of segregation which is going to be measured should also be defined. In this paper, segregation was measured for each province of the country across its municipality spaces. According to country divisions, Islamic Republic of Iran in 2006 had 30 provinces, 1012 cities; but among these cities, there were metropolises like Tehran which contains about 10% of the population. It would not be appropriate to consider these cities as one subunit in the province. Therefore, like the work of Ethington for Los Angles city,[22] we divided each major city into its municipality districts and considered them as subunits of analysis in each province. In this way, Tehran province had 55 cities but 85 municipality spaces.

A large number of indices of residential segregation have been used in researches on segregation and its consequences. Among these indices, measures of evenness are more widely used in papers. Evenness measures how subgroups of a population are distributed evenly across a geographic area. The most commonly used index for measuring this dimension of segregation is dissimilarity index (denoted as D).[32324] It can be interpreted as the proportion of all people who should be transferred among residential units to have the same group proportions across residential units.[24] It can be written as:

where r refers to municipality spaces in each province, t is the population of municipality space r, π is the proportion of group m in municipality space r, π is the proportion of group m in the province, and T is the total population of the province.

In 2002, a revised index of dissimilarity named generalized dissimilarity index was proposed by Reardon et al., which is used for measuring segregation among multiple unordered categories of a variable (e.g., multiple age groups). This index reduces to binary dissimilarity index when there are two categories.[25] Another useful measure of evenness is information theory index (denoted as H). This index measures variation in diversity across subareas. It is interpreted as difference between diversity (entropy) of the system and the weighted average diversity of individual units, expressed as a fraction of total diversity of the system. Diversity (entropy) of the population (denoted as E) is defined as:

Information theory index can be used for measuring segregation for multiple categorical variables. Besides, this index satisfies the principles of exchange, transfer, additive group decomposability, and additive organizational decomposability, while the dissimilarity index does not satisfy these properties. We used seg command in Stata software to compute segregation indices and double checked the results in excel software.

These two measures were used for computation of segregation in this article. Since these two indices are not appropriate where the variable is an ordinal, for measuring segregation among ordered groups of ordinal variables (e.g., education attainment) using the work of Reardon in 2009, we used ordinal information theory index (denote as H0). This measure was computed according to two methods presented by Reardon.[27]

Database

We extracted data from the general population and housing census in Iran. This census is one of the biggest data gathering projects in Iran, which is conducted every 10 years and the latest one was in 2006. Different data about socioeconomic, demographic, and health variables from both individuals and family units were gathered through this census. A random sample containing 20% of data was presented in two databases by Iranian National Statistic Center: an individual database, which each record of it contained the data of one person and a family database which each record of it contained the aggregated data of one family. These databases were used for measuring segregation.

Variables

Generalized dissimilarity index and information theory index were measured for the demographic, socioeconomic, and health variables defined in Tables 1 and 2. These variables were the standard questions of census. One of the special features of the latest census was adding some questions about health outcomes including infant mortality, existence of disability in individuals, existence of disability in the family, and death of offspring of the mother, to the standard questions of the census. According to the definition mentioned in technical guideline of the census, any person with at least one of the following problems was considered a disabled person: Blindness, deafness, speech disorders, upper or lower extremities’ amputation at any level, any physical or functional disorder in upper extremities, lower extremities, or trunk, and any mental disorder.[28]

Table 1

Description of demographic, socioeconomic, and health variables used for measuring segregation (individual dataset)

Table 2

Description of demographic, socioeconomic, and health variables used for measuring segregation (family dataset)

Analysis

For interpreting the values of dissimilarity index and information theory index according to the sociology literature, we used the following cut-off points: Extreme segregation (40-100), high segregation (25-39), moderate segregation (10-24), and low segregation (0-9) for information theory index; and extreme segregation (70-100), high segregation (40-69), moderate segregation (30-39), and low segregation (0-29) for dissimilarity index.[29303132] If a measured index for a variable was classified in one level according to dissimilarity index and in another level according to information theory index, the segregation was classified in the higher level.

To assess the association of segregation of health variables with segregation of other socioeconomic and demographic variables, matrix of correlations was applied and significant correlations between 0.3 and 0.7 were reported as weak correlation and correlations more than 0.7 were reported as strong correlation.

RESULTS

According to the results of Iranian national census in 2006, the population in the country was 70,495,782; 50.54% of the population was males and 49.46% was females. 25.14% of the population was under 15 years old, 69.62% was between 15 and 65 years, and 5.24% were more than 65 years. Literacy rate of the population of more 6 years of age was 84.2%. About 59.5% of the population was living in their birthplace, and 16.15% of the population had history of migration in the 10 years preceding the census. In the population older than 10 years, 39.9% was active and 60.1% was inactive, 5.26% did not have a job, 21.66% were students, 27.64% were housewives, 5.33% had income without job, 34.61% had job, and 5.5% were in other categories. Regarding marital status, 38.8% of the population older than 10 years was single, 56.8% was married, and 3.4% was widow or divorced. Considering the number of marriages in the population older than 10 years, the number of marriages was zero in 40.17%, one in 57.89%, and two or more in 1.94%.

The proportion of disability according to the medical definition presented by scientific committee of census in the total population was 1.56% (1.44-1.69), in the population less than 10 years old was 0.65 (0.59-0.71), and in the population older than 10 years was 1.75% (1.61-1.89). According to the census data, 23.5% (20.9-26) of mothers who had given birth to at least one child had lost at least one of their offspring and 0.048% (0.40-0.055) of mothers who had given birth to a live infant in the 365-day period preceding the census had lost their infant.

Generalized dissimilarity segregation index and information theory index measured for all the individual socioeconomic and health indices are presented in Table 1.

All the provinces were in the category of low segregation for age groups, sex, literacy, education level, type of activity, activity status, employment type, marital status, existence of disability in individuals, death of at least one offspring of mother, and number of remarriages. For birthplace, Tehran, Qazvin, Isfahan, and Chahar Mahal and Bakhtiari provinces were moderately segregated. For immigration history in the previous 10 years, just Tehran province was moderately segregated. For infant mortality, 13 provinces including Qazvin, Khorasan Razavi, Khuzestan, W. Azarbaijan, Lorestan, Markazi, Kurdistan, Zanjan, Hamedan, Kohgiluyeh and Buyer Ahmad, Bushehr, Chahar Mahal and Bakhtiari, and Fars were moderately segregated (in order of increasing segregation) and Semnan and Mazandaran were highly segregated [Figure 1].

Figure 1

Classification of provinces according to segregation of infant mortality

Generalized dissimilarity segregation index and information theory index were also measured for family socioeconomic and health indices which are defined in Table 2.

All the provinces were in the category of low segregation for existence of disability in the family, number of family members, number of literate individuals in the family, number of illiterate individuals in the family, number of individuals with job in the family, number of individuals without job in the family, number of students in the family, existence of vehicle in the family, owning a house, and numbers of rooms per capita. For household size, just Tehran province was moderately segregated. For existence of motorcycle in the family, 10 provinces including Hamedan, S. Khorasan, Gilan, Chahar Mahal and Bakhtiari, Hormozgan, E. Azarbaijan, Kurdistan, Khuzestan, Ilam, and Qazvin (in order of increasing segregation) were in the category of moderate segregation and two provinces including Fars and Kermanshah were in the category of high segregation. For existence of computer in the family, two provinces including Khorasan Razavi and Tehran were moderately segregated. For access to internet by the family, three provinces including Sistan va Baluchestan, Khorasan Razavi, and Tehran were moderately segregated.

For existence of telephone in the family, Qom, Kerman, and Tehran provinces were moderately segregated. For use of natural gas for cooking, Hamedan, Qazvin, Gilan, and E. Azarbaijan were moderately segregated, Isfahan, W. Azarbaijan, Kohgiluyeh and Buyer Ahmad, Lorestan, Semnan, Golestan, and Mazandaran provinces were highly segregated, and Fars, Ardebil, Khuzestan, Yazd, N. Khorasan, Tehran, Chahar Mahal and Bakhtiari, Markazi, Kerman, Khorasan Razavi Kermanshah, Kurdistan, Zanjan, and Bushehr provinces were extremely segregated. For use of natural gas for heating, Markazi, Qazvin, Khuzestan, Hormozgan, and Tehran provinces were moderately segregated and Bushehr was highly segregated. For main source of drinking water, Yazd, Markazi, S. Khorasan, Gilan, Ardebil, Bushehr Sistan va Baluchestan, Khorasan Razavi, W. Azarbaijan were moderately segregated, Mazandaran, E. Azarbaijan, Ch. and Bakhtiari, Tehran, Kermanshah, Semnan, Isfahan, Kurdistan, Golestan, and Hamedan provinces were highly segregated, and Kerman, Hormozgan, Khuzestan, Fars, and Kohgiluyeh and Buyer Ahmad were extremely segregated.

For existence of bathroom in the household, Khorasan Razavi and Chahar Mahal and Bakhtiari provinces were moderately segregated and Kohgiluyeh and Buyer Ahmad and Qazvin provinces were highly segregated. For type of bathroom effluent disposal, just Kerman province had low segregation, Ilam, Mazandaran, Kohgiluyeh and Buyer Ahmad, Zanjan, and Yazd were moderately segregated, Golestan, N. Khorasan, Tehran, Fars, S. Khorasan, Hormozgan, W. Azarbaijan, E. Azarbaijan, Qom, Markazi, Ardebil, Kermanshah, Kurdistan, Bushehr, and Semnan were highly segregated, and Gilan, Khuzestan, Qazvin, Khorasan Razavi, Sistan va Baluchestan, Chahar Mahal and Bakhtiari, Isfahan, Lorestan, and Hamedan were extremely segregated. For existence of separate kitchen in the household, Golestan, Sistan va Baluchestan, Kurdistan, Ilam, E. Azarbaijan, Mazandaran, and W. Azarbaijan were moderately segregated and Hormozgan, Kermanshah, S. Khorasan, Khuzestan, Khorasan Razavi, Bushehr, Hamedan, Isfahan, Kerman, Lorestan, Qazvin, Yazd, Fars, Chahar Mahal and Bakhtiari, Semnan, Gilan, Kohgiluyeh and Buyer Ahmad, and Markazi were highly segregated.

To answer the question whether segregation of health indices is correlated with segregation of other socioeconomic variables, we used pair wise correlation of existence of disability in individual, existence of disability in family, infant mortality, and death of at least one offspring of mother with other socioeconomic variables. The statistically significant correlations are presented in Table 3.

Table 3

Correlation of segregation indices measured for health indices and socioeconomic variables

DISCUSSION

Segregation is a pervasive social issue. The segregation of social factors can explain current inequities in societies. For example, sex segregation across different occupations can explain the gender gap in earnings.[33] Residential segregation has been thought to contribute to Black poverty and higher Black mortality among Whites. Racial segregation among schools has been blamed for low educational achievement among minority groups; besides, it is considered as the prime suspect in explaining the gap in human capital between racial groups in western countries.[34] In some other contexts, segregation might be useful. A segregated place for a social characteristic means accumulation of people with that characteristic in smaller places. It is debated that this formation of homogenous living residential areas might be a solution for highly polarized conflicts in the Middle East.[35]

In many countries, measures of segregation are being reported according to census data. In the United States, residential racial segregation indices are available since 1980 and many articles in the literature have investigated different effects of segregation on social and health outcomes. In this study, for the first time in Iran, we computed the measures of segregation for socioeconomic and health variables using the national census data of 2006.

For infant mortality, Mazandaran and Semnan provinces were highly segregated; in other words, the incidence of infant death was not evenly distributed across the municipality units in these provinces. In 12 provinces, the segregation wad moderate. This finding is in accordance with the results of the study on socioeconomic inequality in infant mortality across the provinces in Iran, which was performed on data of infant mortality in Iran from 1995 to 2000. In that study, Mazandaran province was one of the five provinces in Iran which had the highest inequality in infant mortality in Iran. Of the 12 provinces which were in the category of moderate segregation in our study, Khuzestan and Zanjan had high inequality, and Khorasan Razavi, W. Azarbaijan, Markazi, Kurdistan, Kohgiluyeh and Buyer Ahmad, Bushehr, Fars, and Qazvin had moderate inequality in infant mortality in that study. Lorestan, Hamedan, and Ch. and Bakhtiari were moderately segregated for infant mortality in 2006, but in that study they had low inequality.[36]

Tehran was the only province which was moderately segregated for immigration history. This province was also segregated for birthplace. Considering that Tehran is the capital city of the country and the rate of immigration to this province is high, this means that immigrant people are not evenly distributed across residential areas in this province.

Of the socioeconomic variables, segregation indices measured for housing conditions like main source of drinking water and existence of separate kitchen in the household were correlated with segregation of all four health indices in this study.

In the provinces where access to tap water and housing conditions of the families are not evenly distributed across the municipality units, there is more probability of finding hot zones for infant mortality, disability, and offspring death. Segregation of activity level and job characteristic of individuals, economic situation of family presented by house ownership and existence of motorcycle in family, number of literate individuals in the family, and number of students in the family were also correlated with segregation of two or more health indices in this study.

Sex segregation was correlated with segregation of infant mortality, existence of at least one disabled person in the family, and death of at least one offspring, and literacy segregation was correlated with existence of one disabled person in the family and death of one offspring, but all the provinces were in the category of low segregation for these tow characteristics. It can be concluded the even small changes in segregation of sex and literacy proportions in provinces might result in uneven distribution of health outcomes and creates hot zones for these health outcomes.

These findings are consistent with the results of some other studies in different parts of the world. In a study of inequality in nine developing countries, levels of inequality in consumption were associated with inequality in child mortality.[36] In a health survey conducted in India, levels of socioeconomic status composed of income, education, housing condition, and house ownership were associated negatively with child death rates.[37] In a study in a Brazilian city, geo-economic classification of city was correlated negatively with infant mortality.[38]

The critical question about the importance of segregation is on the causal pathways that segregation acts through them. Four interconnected mechanisms are suggested. A leading hypothesis is that residential segregation affects individual socioeconomic status. The second pathway is through creating unhealthy neighborhood environments. The third and fourth hypothesized mechanisms are modification of social capital of a residential place and changing individual risk behaviors and exposures to stressful situations. Degree of social trust between individuals in a neighborhood, extent of social networks, and tendency for mutual aid and support are the determining factors of social capital.[39] Variations in social capital might explain the differences in populations’ health.[40] Some studies suggest that segregation affects social capital.[41]

Limitations

This study is subjected to limitations associated with the use of census data. Information theory index and generalized dissimilarity index were used in the study because of their unique statistical characteristics and high use in other studies of segregation, but in some studies, isolation index has been used. Further studies are necessary to determine the best segregation index for measuring segregation of markers of socioeconomic status and health.

Selection of variables was limited to the questions of the Iranian national census. This approach makes comparison of measured indices across periods of census and checking the trend of changes feasible.

In this study, segregation was measured in each province across municipality spaces, and provinces were compared according to these measures. It is possible to measure segregation across smaller geographic units for each city in every province to compare the degree of segregation of major cities inside each province. This method might show better correlation of segregation of heath indices and determinants of socioeconomic status. The choice of health variables in this study was very limited because of limited health variables measured in the national census. It is recommended that by using data of national health surveys, segregation of other health variables is measured.

Aside from these limitations, this study is of high value since it is the first study in Iran which measures segregation and one of the first studies in the literature to measure segregation of socioeconomic status variables in national level.

CONCLUSIONS

In this study, segregation of Iranian provinces for major determinants of socioeconomic status and also some health indices were reviewed. For infant mortality and some of the socioeconomic factors, many of the provinces were moderately or highly segregated. Considering the effect of segregation on health outcomes which have been supported by large body of evidence in the literature, investigating why segregation is higher in some provinces deserves special attention.