Demographic and Social Factors

I. SOCIOECONOMIC STATUS

This summary draws heavily on the following review articles: Liberatos et al. (1988), Berkman & Macintrye (1997), Krieger et al. (1997), Lynch & Kaplan (2000), & Kawachi (2000).

Lynch & Kaplan (2000) argue that choice of SES measure should depend on how researcher believes SES is linked to health damaging exposures, health protective resources, & ultimately to health - e.g. is it exploitation, few tangible resources, lack of prestige, or a combination that causes poor health?

A. MEASURES

1. Individual-level measures

a. Education

Self-reports of total number of yrs of education, or, alternatively, attainment of particular educational milestones (e.g. completing high school)

Advantages.
o Education is easy to measure, generally available for both genders, excludes few members of the population, & is less likely to be influenced by disease in adulthood than income or occupation, although it is possible that childhood difficulties associated with low SES may adversely affect later educational attainment.
o Exposure to formal education involves gathering facts, learning concepts & how to access information. It provides a set of cognitive resources that has broad potential to influence health.
o In addition, higher levels of education generally are predictive of better jobs; higher incomes; & better housing, neighborhood, & working conditions.
o Educational attainment also has a socially symbolic value -- e.g., a degree from a prestigious university has a different social & symbolic value than one from a less prestigious institution.

Disadvantages.
o Education has different social meanings & consequences in different time periods & cultures.
o Economic returns on education may differ markedly across gender, racial, ethnic groups. Women & minorities realize lower returns for the same investment in education that do white men.
o SES does not rise monotonically with number of years of education.
o There is decreasing variability in educational level in industrialized nations over time
o Information on number of years or achieved level of education does not directly reveal much of what might be important about education in terms of its relationship with health. Without knowledge of cognitive, material, social, & psychological resources gained through education & accumulated over the life course, researchers cannot deepen understanding of education-health link nor address important intervention questions.

b. Income

Self-reported annual income at the personal, family, or household level

o Income allows access to material goods & services that may influence health - quality, type, & location of housing; healthier, more nutritious food; clothing; transportation; medical care; opportunities for cultural, recreational, & physical activities; child care; freedom from exposure to an array of environmental toxins.

However, income measures may not fully capture economic status or purchasing power b/c they:
o Do not measure wealth (accumulated assets) [see 'wealth,' below]
o Do not measure benefits such as health insurance coverage, disability benefits, etc.
o Do not measure income earned from informal transfers (many people exchange goods or services, and barter; recent immigrants & minorities work in an informal economy with no job security or benefits)
o Goods & services available to whites & residents of higher income neighborhoods tend to be better in quality & lower in price than those available to blacks & residents of lower-income neighborhoods.

Other disadvantages to income measures:
o Sensitive to changes in life circumstances
o Can be extremely volatile, fluctuating considerably
o Tend to increase with age up to age 65
o Not comparable across different family sizes or calendar years, unless adjusted
o High nonresponse rates in epidemiologic surveys in the US

o There is a graded relationship between income & health that is not limited to poverty. Each step up the income ladder may bring added "neo-material" benefits that can produce gains in health. Such neo-material conditions are intimately tied to psychological states, health behaviors, & social circumstances that influence health (Lynch & Kaplan 2000). Disposable income provides a buffer from many stresses of daily life. Indeed, most sources of social & environmental stress are not randomly allocated among the population. It is precisely those groups with the least disposable income who are subject to the largest cumulative burden of stressors (McLeod & Kessler 1990; Ross & Wu 1996).

o Lynch et al (1997a). Kuipio Ischemic Heart Disease Risk Factor Study. 2674 middle-aged men. Men who worked in low-paid employment were the most materially disadvantaged, had higher job & financial insecurity, & experienced more unemployment & work injury. These were the same men that smoked more, exercised less, ate less nutritious diets, got drunk more often, had cynically hostile outlook, & did not feel full of hope about the future.

o Lynch et al. (1997b). Alameda County Study. 29-yr study of economic hardship & functioning in 1124 men & women. Income measured in 1965, 1974, & 1983. The cumulative effect of sustained economic hardship on physical, psychological, cognitive, & social functioning in 1994 was examined. Psychological function was assessed in terms of depression, cynical hostility, & optimism. Cognitive function was assessed by 4 Likert-type questions about self-reported difficulties in remembering things, paying attention, word finding, & forgetting where things are placed. There were strong dose-response relationships between the number of periods of economic hardship & physical, psychological, & cognitive functioning.

c. Occupation

Occupation is the major structural link betw educ & income: education à occupation à income

o British Registrar General's Classification (Szreter 1984)

Perhaps the best-known occupational class measure. Developed in 1913 to assess combination of skill level & "standing within the community." It was explicitly not based simply on the average income of occupations. 5 categories: Social Class I (professional), Social Class II (intermediate), Social Class IIINM (skilled nonmanual), Social Class IIIM (Skilled Manual), Social Class IV (partly skilled), Social Class V (unskilled).

Highly predictive of inequalities in morbidity & mortality, especially among employed men (e.g. Whitehall Studies, Marmot et al. 1995). However, because this classification scheme was developed in part to analyze infant mortality rates & modified to yield social class gradations in mortality, there is some circularity in using it to assess SES-health associations.

o Edwards - US Census Classification (US Census Bureau 1963; Haug 1977)

13-level classification scheme used by US Census Bureau. Classification is based on the distinction between manual & non-manual occupations, & reflects occupational skill & status. One disadvantage of this classification scheme is that each occupational category contains wide variations in education & income.

o Nam-Powers Occupational Scale Score (OSS; Nam & Powers 1983)

Continuous ranking of occupations based on median education & income of people in particular US occupations, updated to 1980 census. Scores range from 0-100. Each score is interpretable as a cumulative percentile. Data are available for male, female, black & total labor forces. The scale has not been sufficiently used to provide empirical evidence of its performance.

o Siegel's Prestige Scale (Siegel 1971)

Continuous scale based on occupational prestige rankings from US National Opinion Polls conducted 35+ yrs ago. Scores are available for male labor force only.

o Treiman's Standard International Occupational Prestige Scale (Treiman 1977)

Based on occupational prestige rankings collected 30+ years ago in both developed & developing countries. This is the only scale theoretically suitable for cross-country comparisons. Within the US, the scale is very similar to Siegel's scale; the correlation between the 2 scales is 0.93. Scores range from 0-100. The scale includes 509 occupations that can be grouped into 8 occupational levels. Scores are available for male labor force only.

o Wright's Social Class Typology (EO Wright 1985, 1996)

Categorization based on occupational hierarchy of managers, supervisors, & workers, plus information on supervision of other workers, & control over decision making. This typology distinguishes between 4 basic class categories: wage laborers, petty bourgeois (self-employed with no more than 1 employee); small employers (2-9 employees); & capitalists (10+ employees).

Wright's & similar measures of social class are beginning to be incorporated into public health research (e.g. Soderfeldt et al. 1987; Krieger 1991; Wohlfarth 1997; Muntaner et al. 1998).

Composite indices:

o Duncan Socioeconomic Index (SEI; Duncan 1961; Stevens & Featherman 1981; Stevens & Cho 1985).

Continuous scale based on occupational prestige rankings from US National Opinion Polls, plus income & education. Scores available for 1950, 1970, & 1980 census. Scores range from 0-99; often exhibit positively skewed distribution. Very frequently used in social science research.

o Hollingshead Index of Social Position (Hollingshead & Redlich 1958; Hollingshead 1975)

Prestige scale similar to Duncan SEI. Combines information on an individual's level of education & occupational rank, as based on Hollingshead's personal rating of people's relative social standing in New Haven CT in early 1960s. Continuous score, or scored categorically as "social classes." Available for 1970 census. Widely used during 1960s & early 1970s.

o Nam-Powers Socioeconomic Score (SES; Nam & Terrie 1986)

Continuous scale that combines Nam-Powers Occupational Status Score for a given individual's occupation (see above) with that person's educational level & family income. Available for 1960, 1970, & 1980 census. Scores range from 0-100. Each score interpretable as a cumulative percentile. Data available for male, female, black & total labor force. Scores are normally distributed. The scale has not been sufficiently used to provide empirical evidence of its performance. Potentially redundant if used in combination with individual's education & income.

Critique of existing scales:

o Many occupational scales are or are becoming outdated. Occupational rankings are relatively unstable over time. New requirements & economic needs change job standing in terms of prestige & income over time.

o Occupational classes are made up of heterogeneous occupations, & there is considerable variation within each class in education, income, & prestige. Studies that examine more homogeneous occupational groups within specific industries or employment settings tend to find much bigger differences between these groups in mortality than are found for the occupational classes in which they are normally classed. e.g. standardized mortality ratios for CHD among men in British army show a 6-fold difference between private soldiers & direct-entry officers, a difference that is much greater than the difference between all Registrar General's Social Class I & Social Class V men (Lynch & Oelman 1981). Conventional occupational class measures may underestimate SES-health link because of imprecision of measurement.

o Most measures were developed & validated on working men. Extrapolating to other populations can be problematic. There are no completely successful resolutions as to how to classify homemakers, retirees, minorities who may hold the same job as white males but do not gain the same benefits; persons working in informal or illegal sectors of the economy; unemployed adults; children, or working couples. Approaches to measuring occupational class of such groups usually rely upon proxy measures (e.g., last or main occupation for unemployed or retired workers, spouse's occupation for homemakers, parents' or father's occupation for children).

Epidemiologic studies have explored the pathways through which work affects health.

o Physical environment: chemicals, radiation, biological hazards, physical stress, noise, heat, unsafe conditions, cold, dust & other pollutants

o Psychosocial environment: Karasek's demand-control model, Siegrist's extension to include non-work factors & consideration of the income & other rewards derived from work. [See 'Occupational environment' below]

d. Wealth

Continuous or categorical measures of accumulated assets - i.e. value of housing, investments, inheritance, pension, liquid vs. nonliquid assets.

Wealth is a source of economic security & power. Assets are an indicator of a household's ability to meet emergencies or absorb economic shocks.

o Smith & Kington (1996). Survey of Asset and Health Dynamics of the Oldest Old. Age 70+. Nonlinear associations between health & both income & wealth, with associations strongest at the bottom of the income & wealth distribution. Associations larger for wealth than income.

e. Material deprivation

Classify individuals according to household assets such as whether the home is owned or rented, and whether there is a car or garden, etc. It has been argued that such measures create a more fine-grained hierarchy of SES than the traditional SES measures of education, income, & occupation. However, material deprivation measures are the least well-characterized measures of SES, as little research has been conducted into their social meaning & the processes by which they influence health.

o Townsend Deprivation Index (Townsend et al. 1988)

Individual-based measure of material & social deprivation. 77 items. Material deprivation items concern "dietary, clothing, housing, home facilities, environment, location, and work (paid and unpaid)." Social deprivation items concern "rights to employment, family activities, integration into the community, formal participation in social institutions, recreation & education" (see Krieger et al 1997)

f. Perceived social status

o MacArthur Scale of Subjective Social Status (Adler 2000; Adler et al. in press).

This measure assesses individuals' sense of their place in the social ladder. In pictorial format, it presents a "social ladder" & asks individuals to place an X on the rung on which they feel they stand. There are 2 versions of the ladder, one linked to traditional SES indicators (SES ladder) & the second linked to standing in one's community (community ladder). The difference between these 2 ladders may be of particular interest in poorer communities in which individuals may not be high on the SES ladder in terms of income, occupation, or education but may have high standing within social groups such as a religious or local community.

o Adler (2000). Preliminary data suggest that individuals' perceptions of their place in the hierarchy as assessed by the ladders is strongly associated with physical & mental health. This association is stronger than associations of health with objective SES indicators. e.g. Whitehall II Study of British civil servants. There is a higher incidence of depression & self-reported poor health in low status respondents, both as classified by the ladder & by grade of employment. The ladder shows more of a gradient. Subjective social status was not only significantly associated with key health outcomes (e.g. general health, depression, respiratory disease), but the effects of occupational grade became nonsignificant once subjective status was entered. The association between subjective status & physical health remained significant even when depression was controlled.

2. Group Measures

a. Areas-based measures

Area-based measures are usually aggregate correlates of individual measures. It is important to distinguish whether a particular measure is meant as a proxy for individual characteristics or whether it is meant to actually characterize a certain quality of the area itself.

Advantages of area-based measures:
o Characterize aspects of living conditions not captured by individual & household measures; especially important in studies of people from diverse racial/ethnic groups, given the greater likelihood, at each socioeconomic level, of white individuals to live in more affluent, safer, & less polluted neighborhoods than minority groups.
o Available for persons of all ages & both genders.
o May provide more meaningful estimate of relevant economic circumstances than more volatile income measures or more static education measures.
o Allow for contextual-based analyses that provide insight into how SES shapes population health at multiple levels.

o US census-based measures . Neighborhood units defined & characterized by the US Census Bureau include census tract (~ 4000 residents), census block-group (~ 1000 residents), & census block (~ 85 residents). Census tract & block-group boundaries are intended to demarcate populations that are relatively homogeneous with regard to sociodemographic characteristics. Block-groups tend to be more homogenous than tracts & can reveal otherwise hidden pockets of poverty & affluence. Census block data are generally not used because of privacy concerns.
Commonly used measures include:
o Social class: % working class = % of employed persons in 8 of 13 census-defined occupational groups: administrative support; sales, private household service; other service (except protective); precision production, craft, repair; machine operators, assemblers, inspectors; transportation & material moving; handlers, equipment cleaners, laborers),
o Poverty: % persons below poverty line; poverty area = 20+% persons below poverty.
o Wealth: % households owning home, % households owning 1 or more cars, % households with annual incomes of $50,000+
o Education: % of adults aged 25+ with less than a high school degree; undereducated neighborhood = 25+% of adults with less than a high school degree.
o Crowding = % persons living in households with 1+ person/room
o Population density = # persons/square mile

Epidemiologic studies of the relationship between neighborhood socioeconomic characteristics and health generally rely upon census-based measures. Krieger et al (1997) review findings & discuss the validity of such measures. They also issue a cautionary note about the use of (non-census) zip codes. Unlike census tracts & block-groups, zip codes encompass relatively large geographic areas containing 30,000 or more people who are generally not sociodemographically homogeneous. Zip codes are not ideal units upon which to construct area-based SES measures.

o Mayer-Jencks' Material Hardship measure (Mayer & Jencks 1989).

US area-based measure of material deprivation. Assesses unmet needs for food, housing, & medical care in the past year.

o Has been used in combination with social network & social capital measures to predict infant health outcomes in urban settings (see Krieger et al. 1997)

o Carstairs-Morris index (Carstairs & Morris 1991).

British area-based (postal codes) index of material deprivation. Combines information on % of unemployed people, overcrowded households, households with no car, people in social classes IV & V. Widely used in Great Britain.

o Townsend index (Phillimore, Beattie, & Townsend 1994).

British area-based index of material deprivation. Combines information on % people with no car, in overcrowded housing, in non-owner occupied housing, & unemployed. Widely used in Great Britain.

o Ben-Shlomo et al. (1996). Examined relationship between material & social deprivation (as measured by Townsend Index) in 369 local authorities (geographic areas administered by local gov't) & average mortality rates. Mortality was strongly associated with deprivation. The average trend in death rates was 26 per 100,000 per quartile of deprivation (p<0.001). Mortality was also positively associated with degree of socioeconomic variability in the area units that made up the local authorities. The more variable the extent of deprivation within an area (as indicated by interquartile range of Townsend scores), the higher the death rate, with an average trend of 7 per 100,000 (p<0.001) across quartiles of variation.

b. Income inequality measures

Measures of income distribution are generally calculated from census data.

o Proportion of total income earned by bottom X percent of households in a given region. If incomes were perfectly equally shared, proportion of total income earned by the bottom 50% (for example) of households should account for exactly half of the aggregate income.

o Gini coefficient . This measure is derived from the Lorenz curve, which is a graphical device for displaying the cumulative share of total income accruing to successive income intervals. Th curve shows the shares of income earned by successive deciles of households arrayed in order from the bottom 10% upwards. If incomes were equality distributed, the Lorenz curve would follow the 45 degree diagonal. As the degree of inequality increases, so does the curvature of the Lorenz curve, & thus the area between the curve & the 45 degree line gets larger. The Gini is calculated as the ratio of the area between the Lorenz curve & the 45 degree line, to the whole area below the 45 degree line. Theoretically ranges from 0 (perfect equality) to 1 (perfect inequality). It is the most widely used measure of income distribution (Kawachi & Kennedy 1996; Kennedy et al. 1996).

o Robin Hood Index (Atkinson & Micklewright 1992). Conceptually, this measure is the proportion of aggregate income that must be redistributed from rich to poor households to attain perfect equality of incomes. Mathematically, it is equivalent to the maximum vertical distance between the Lorenz curve & the line of equal incomes.

o Atkinson Deprivation Index (Atkinson 1970; see also Cowell 1995).

o Theil Entropy Index (Theil 1967; see also Cowell 1995).

All measures listed above are strongly correlated with health/mortality in ecologic/multi-level analysis:

o Kaplan et al. (1996). Unit of analysis: 50 US states. Measure of income distribution: share of total income earned by bottom 50% of households in each state. The states ranged from a minimum of 17.5% to a maximum of 23.6%. A strong correlation (r=-0.62, p<0.001) was found between this measure of income inequality and age-standardized mortality rates. The association was observed in both men & women, & in both whites & blacks.

o Kennedy et al. (1996). Unit of analysis: 50 US states. Income inequality, as measured by the Gini coefficient & Robin Hood Index, was strongly correlated with age-adjusted total & cause-specific mortality rates. For example, adjusting for poverty rates & median income, a 1% increase in the Robin Hood Index was associated with excess mortality of 21.7 per 100,000, suggesting that even a modest reduction in inequality could have an important public health impact.

o Lynch et al. (1998). Unit of analysis: 282 greater metropolitan areas in US. Measures of income inequality included Gini coefficient, Atkinson Deprivation Index, Theil Entropy index, proportion of income earned by the bottom 50% of households. Metropolitan areas with high income inequality & low per capita income has excess mortality of 149.8 deaths per 100,000 compared to areas with low inequality & high per capita income.

o Diez-Roux et al. (2000). 1990 Behavioral Risk Factor Surveillance System (BRFSS) study & US Census. Associations between state income inequality & 3 of 4 CVD risk factors (high BMI, history of high BP, sedentarism, but not smoking) - were observed after controlling for absolute level of individual income. Associations statistically significant in women but not in men.

Potential pathways between income inequality & health: o Underinvestment in human capital (Kaplan et al. 1996)

o Income disparities disrupt social cohesion & lead to disinvestments in social capital (Kawachi & Kennedy 1997; Kawachi et al 1997)

o Direct psychological pathways, e.g. frustration engendered by invidious social comparisons (the "relative deprivation" hypothesis) (Kawachi et al. 1994). Epidemiologic studies that directly connect frustrated expectations to health outcomes are sparse. Dressler (1996, 1998) uses the term "cultural consonance in lifestyle" to refer to degree to which individuals succeed in achieving the lifestyle that is considered customary for their community. To the extent that individuals strive & fail to meet cultural ideals, negative health consequences follow. The degree of departure in cultural consonance is a strong predictor of systolic BP, even after adjustment for established clinical risk factors for hypertension (Dressler 1996). Adverse consequences of relative deprivation are not confined to psychological realm. As societies become more prosperous, material needs increase not just because people think they need more when neighbors have more but also for practical reasons. Many consumer goods introduced as luxuries gradually become necessities.

B. Epidemiologic evidence on the SES-health link

1. SES as a predictor of health outcomes

o SES is a critical predictor of health outcomes -- including mortality (e.g. Bassuk et al., under review), cardiovascular disease (Kaplan & Keil 1993), & cognitive decline (see below) -- throughout the life course. While some studies suggest that SES wanes in importance as a determinant of health outcomes in elderly populations, many other studies find no effect modification by age.

2. SES as a predictor of dementia & cognitive decline

o Low educational/occupational attainment is associated with risk of dementia/Alzheimer's disease (AD) & with cognitive decline more generally. Studies showing positive associations:

Education & dementia/AD:
o Zhang MY, Katzman R et al. (1990). Shanghai study
o Canadian Study of Health & Aging (1994). Canadian Study of Health & Aging
o Cobb JL, Wolf PA et al. (1995). Framingham Study
o Stern Y, Gurland B et al. (1994, 1995). N Manhattan Study
o Callahan CM et al (1996). Community-based sample of black Americans
o Ott A, Breteler MMB et al. (1995). Rotterdam Study
o Fratiglioni L et al. Kungsholmen Project in Sweden

Education & cognitive decline:
o White L et al. (1994). EPESE
o Farmer ME et al. (1995). Epidemiologic Catchment Area Study

Occupation & dementia/AD:
o Dartigues et al. (1992). PAQUID Study.
o Fratiglioni L et al. (1991). Kungsholmen Project in Sweden.

Occupation & cognitive performance:
o Fuhrer R, Head J, Marmot MG (1999). Whitehall II Study.

Studies with null results:

Education & AD:
o Hagnell O et al. (1992). Lundby Study.

Occupation & dementia/cognitive decline:
o Hagnell O et al. (1992). Lundby Study.
o Jorm AF et al. (1998). Canberra/Queanbeyan Australia study. o Albert (1995) reviews possible explanations for observed associations between education & cognitive function in aging populations.

3. Pathways between SES & health Psychosocial pathways:

o Taylor & Seeman (1999) review evidence on the role of psychosocial resources (including personal control, optimism, coping style, social support, & reactive responding) in the SES-health relationship.

Biologic pathways:
o Kaplan & Keil (1993) review evidence on the role of biologic factors in the SES-health relationship, with an emphasis on cardiovascular disease outcomes.
o Lynch J et al. (1998). Kuipio Ischemic Heart Disease Risk Factor Study. Low SES potentiates effects of heightened cardiovascular response to stress on 4-yr progr
ession of carotid atherosclerosis in 882 middle-aged men.
o Pickering T (1999). Cardiovascular pathways: Socioeconomic status and stress effects on hypertension and cardiovascular function. In Adler NE, Marmot M, McEwen BS, Stewart J (eds). Socioeconomic status & health in industrial nations. Annals of the New York Academy of Sciences. 1999:262-277.