ARTICLE IN PRESS
Social Science & Medicine 59 (2004) 2409–2419
Income inequality and self-rated health in US metropolitan areas: A multi-level analysis Russ Lopez* Department of Environmental Health, Boston University School of Public Health, 715 Albany St., Talbot 2E, Boston, MA 02118, USA Available online 25 May 2004
Abstract Income inequality has been found to affect health in a number of international and cross-national studies. Using data from a telephone survey of adults in the United States, this study analyzed the effect of metropolitan level income inequality on self-rated health. It combined individual data from the 2000 Behavioral Risk Factor Surveillance System with metropolitan level income data from the 2000 Census. After controlling for smoking, age, education, Black race, Hispanic ethnicity, sex, household income, and metropolitan area per capita income, this study found that for each 1 point rise in the GINI index (on a hundred point scale) the risk of reporting Fair or Poor self-rated health increased by 4.0% (95% confidence interval 1.6–6.5%). Given that self-rated health is a good predictor of morbidity and mortality, this suggests that metropolitan area income inequality is affecting the health of US adults. r 2004 Elsevier Ltd. All rights reserved. Keywords: Income inequality; Self-rated health; GINI,Metropolitan; USA
Introduction While there is general agreement that low income is a risk factor for disease and poor health (Syme, 1998), there is less consensus on the relationship between income inequality and health. Using data from the 2000 Behavioral Risk Factor Surveillance System (BRFSS) and the 2000 US Census, this multi-level analysis found a strong association between self-rated health and metropolitan area income inequality as measured by the GINI index after controlling for individual level risk factors previously associated with self-rated health.
Background There is a long history of research demonstrating that poverty status affects health. Individual and household *Tel.: +1-617-414-1439; fax: +1-617-638-4857. E-mail address:
[email protected] (R. Lopez).
poverty is a risk factor for asthma, cardiovascular disease, diabetes, many cancers, homicide, infectious diseases and all cause mortality (Ecob & Smith, 1999; Schalick, Hadden, Pamuk, Navarro, & Pappas, 2000; Sterling, Rosenbaum, & Weinkam, 1993). Neighborhood deprivation has also been found to affect health and social outcomes (Collins, & Margo, 2000; Cubbin, Hadden, & Winkleby, 2001; Guest, Almgren, & Hussey, 1998; Ren, Amick, & Williams, 1999). Others researchers have found that income inequality is correlated with harm to the environment (Ravallion, Heil, & Jalan, 2000). But does income inequality also affect health? Inequalities in society might have profound impacts on morbidity and mortality (Kawachi, Kennedy, & Wilkinson, 1999). Perhaps if the United States and other developed countries had continued their post World War II move toward greater income equality, this question and line of investigation might not have mattered. But starting in the 1970s, income inequality began to increase in industrialized countries. In the United States, inequality increased by 20% from its low
0277-9536/$ - see front matter r 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.socscimed.2004.03.033
ARTICLE IN PRESS 2410
R. Lopez / Social Science & Medicine 59 (2004) 2409–2419
point near 1970 to the end of the 1990s (Jones & Weinberg, 2000; Smeeding, 2000; Smeeding, Rainwater, & Burtless, 2000). Inequality here is used to describe the distribution of incomes across society. The reasons for this increase in inequality are not well identified but might include industrial restructuring, growth in numbers of the elderly and new immigrants, changes in welfare policy, reductions in the progressiveness of the income tax and other changes (Chevan & Stokes, 2000; Kasarda, 1995; Massey, 1996; Morrill, 2000; Nielsen & Alderson, 1997). The result of rising inequality has been to increase the share of income received by the wealthiest segments of society while reducing the share received by the poorest. Furthermore, many believe that rising inequality has resulted in the share of income received by middle-income households to have declined as well (Piketty & Saez, 2003, Daly, Crews, & Burkhauser, 1997; Krugman, 2002). International studies Concerns regarding the role of inequality in health date back many years. In 1979, a major international comparison demonstrated the contribution of income inequality to health and mortality (Rodgers, 1979). The current debate on the health–inequality connection began when a flattening out of improvements in mortality was found for countries with annual per capita income greater than $5000 (Wilkinson, 1994). Several international studies found differences in infant mortality to be related to income inequality (Hales, Howden-Chapman, Salmond, Woodward, & Mackenbach, 1999; Wildman, 2001). But the evidence for other causes of death is mixed (Lynch et al., 2003) and some deny an association between income inequality and health, and discount cross-country studies on methodological grounds because the relationship between individual income and health is itself non-linear (Ellison, 2002; Gravelle, Wildman, & Sutton, 2002). Still other work asserts that the relationship exists only in the United States (Mackenbach, 2002). United States—state level studies Looking at the United States, researchers have tried to determine if relative differences in income inequality explain the landscape of health and mortality differences. A series of studies found state-level income inequality to be associated with cardiovascular risk factors (Diez-Roux, Link, & Northridge, 1999), ageadjusted all-cause mortality (Kaplan, Pamuk, Lynch, Cohen, & Balfour, 1996), total mortality (Wolfson, Kaplan, Lynch, Ross, & Backlund, 1999), infant mortality, heart disease mortality, cancer mortality, cerebrovascular disease and homicide (Kennedy, Kawachi, & Prothow-Stith, 1996).
United States—county and metropolitan studies Examining smaller US geographic areas has also had mixed results and several of the studies that have found a relationship have been challenged on methodological or other grounds. A national study linking income inequality and mortality using counties found that racial concentration and income inequality had effects on mortality after adjustment for age, sex and race (McLaughlin & Stokes, 2002). Combining 1990 Census data with 1989–1991 mortality data, income inequality effects were most evident for infant mortality and mortality between the ages of 15 and 64 (Lynch, Kaplan, Pamuk, Cohen, & Heck, 1998). An income inequality and all-cause age-adjusted mortality study among North Carolina counties using mortality data from 1985 to 1994 found a relationship both for all counties and non-metropolitan area counties (Brodish, Massing, & Tyroler, 2000). A study of self-rated health found a link with county-level income inequality but not with census tract income inequality, further highlighting that geographic scale may influence any potential relationship between inequality and health (Soobader & LeClere, 1999). In contrast, a study primarily comparing counties found no relationship between income inequality and health once individual household income was controlled for (Brunner, 1997) While some have postulated that the segregation of the poor may be responsible for income inequality effects (Jargowsky, 1996), a study of metropolitan areas found that economic segregation did not affect the association between income inequality and mortality (Lobmayer & Wilkinson, 2002). These findings have been challenged because they may be subject to confounding by the percentage of blacks in metropolitan areas (Deaton & Lubotsky, 2003). Controlling for race and income, mortality for both blacks and whites was related to income inequality in US metropolitan areas in the period 1989–1991 (Cooper et al., 2001). Others found an association between income inequality and all-cause mortality, black mortality and white mortality in 1990. But while the authors controlled for the percentage of people in a metropolitan area who were Black and for the ratio of primary care physicians to the total population, they did not control for per capita income of metropolitan areas or the number of people who lived in poverty (Shi & Starfield, 2001). These effects may be complicated by the fact that income inequality may not affect all subgroups equally. While one county-level study found that income inequality affected the mortality risk for both Blacks and Whites (Cooper et al., 2001), another found that the effects of income inequality were limited to middle-aged Whites (LeClerer & Soobader, 2000). The apparent relationship between income inequality and Black and White mortality needs further exploration.
ARTICLE IN PRESS R. Lopez / Social Science & Medicine 59 (2004) 2409–2419
Some evidence suggests that metropolitan level income inequality may not affect health. One study of metropolitan area income inequality and self-rated health, controlling for individual’s sex, age, race and household income found no relationship; however this study used an employment oriented survey (the US Bureau of the Census’s Current Population Survey), not a health research study, which may have affected the outcome (Blakely, Lochner, & Kawachi, 2002). A small multi-level study of 9585 respondents completed by telephone in 1997–1998 in 60 metropolitan or economic areas found that a relationship between income inequality and self-reported health disappeared once individual characteristics were controlled for (Sturm & Gresenz, 2002). Other issues Yet to be explored is the question of geographic scale. At which geographic scale—census tract, neighborhood, city, county, metropolitan area state or national—is income inequality most likely to have effects and most appropriately considered? On the one hand, the general structure of a society is often set on the national level through federal policy, a centralized media, shared values and migration between states and metropolitan areas. However, many of the factors that most affect health: stress, affordable housing, social capital, etc., are most likely to operate on a smaller geographic level (Pickett, 2001; Diez-Roux, 1998). At the local level, neighborhoods may tend to become homogenous, lowering the amount of inequality within a neighborhood (but increasing inequality between neighborhoods). Given that metropolitan areas are defined to reflect social and economic units, they are an appropriate level of study. Also, metropolitan level inequality may reflect national and state-level inequality with income inequality at one geographic scale related to the degree of income inequality at other geographic scales. Other metropolitan level factors such as sprawl and racial segregation have been found to affect health (Ewing, Schmid, Killingsworth, Zlot, & Raudenbush, 2003; Frumkin, 2002; Polednak, 1997; Lopez, 2002). Self-rated health An important issue is whether self-rated health is meaningful. Is there a relationship between how an individual subjectively perceives his or her health and any objective measure of health status or health outcome? Health status, however measured, is more than just the absence of disease. It includes a whole set of factors that relate to how an individual feels and how well that individual can function in society and the environment. Generally, surveys that include self-rated health ask respondents to describe
2411
their health as Excellent, Very Good, Good, Fair, and Poor. In the United States, self-rated health has been found to be very predictive of mortality and to generally correlate with other assessments of patients’ health. It has been postulated that self-rated health is a good predictor of overall health status because it is by nature a more valid assessment of health status than alternative measures, it is more likely to include undiagnosed or early stage illness that might be missed by a physician, it is based on an individual’s complex assessment of their total health status, it includes an individual’s understanding of their social and familial history, it does not rely on current health status but incorporates an individual’s assessment of the trajectory of their health status, individuals may base their health-related behaviors at least in part on how they perceive their health status, and it reflects the availability of resources and the quality of environmental factors that may ultimately affect health (Idler & Benyamini, 1997). Also important, this association seems to hold across population subgroups (Bosworth et al., 1999; Kennedy, Kasl, & Vaccarino, 2001), though it may be less predictive in some US Hispanic communities, particularly for newer immigrants. The reasons for this may include an optimistic psychological status for new immigrants or it may be affected by the ‘‘healthy immigrant’’ effect, the propensity for immigrants to be healthier than people who do not migrate (Finch, Hummer, Reindl, & Vega, 2002). The relationship between self-rated health and health outcomes also appears to be valid in other countries (Miilunpalo, Vuori, Oja, Pasanen, & Urponen, 1997). The GINI index There are multiple ways of measuring income inequality (Coulter, 1989). Popular methods include the GINI Index, based on the cumulative distribution function of households’ income, and various percentile indexes, which compare the portion of total income received by the top fraction with the portion received by a bottom fraction (Cowell, 1998). Still other measures incorporate judgments about the ideal level of income distribution (Amiel, 1998). Fortunately, the choice of income inequality measure does not appear to affect the relationship between income inequality and health (Kawachi & Kennedy, 1997a). The formula for the GINI index, the measure used in this study, is: N X
2ðXi YiÞDXi;
i¼1
where Xi=1/n, Yi=cumulative % of income by unit, DXi ¼ Xi Xi1 ; and N is the number of income categories.
ARTICLE IN PRESS 2412
R. Lopez / Social Science & Medicine 59 (2004) 2409–2419
The range of potential values of GINI is 0 (complete equality), to 100 (complete inequality).
Methods This multi-level study combined data from a national survey of adults with the 2000 Census. The 2000 Behavioral Risk Factor Surveillance System data was obtained from the Centers for Disease Control website (CDC, 2000). It consists of over 188,000 telephone interviews from all 50 states, the District of Columbia, and certain US possessions. Characteristics from the BRFSS, race/ethnicity, gender, education, having ever smoked at least 100 cigarettes over a lifetime, educational attainment, and age, were included in the analysis because they have been previously identified as being related to self-rated health (Lantz, et al., 1998). Some variables were used to create new dummy variables for gender, Black race and Hispanic ethnicity. For example, respondents reporting they were Hispanic were coded as one in a Hispanic dummy variable, all other respondents were coded as 0. The BRFSS collects household income data from respondents in broad categories. The mean point of a respondent’s household income category was assigned to each respondent and the ratio of that income to the official poverty threshold for the respondent’s household size was calculated. This income ratio was then used in this study to control for household income adjusted for household size. For example, if a respondent reported their household income was between $20,000 and $25,000 and their household size was 4, the respondent was assigned an income ratio of 1.24 (22,500/18,104). Self-rated health, categorized as Excellent, Very Good, Good, Fair and Poor was included from the BRFSS. Excellent and Very Good were collapsed together as were Fair and Poor. This allowed for an examination of the effects of income inequality on the risk of reporting the worst health. GINI index measures of income inequality for every US metropolitan area were calculated based on household income long-form data from the 2000 Census. Metropolitan areas are defined by the US Office of Management and Budget and generally consisted of one or more central cities and their surrounding area. If a metropolitan area was subdivided into primary metropolitan areas, the primary areas were used. Metropolitan areas in New England had two alternative definitions, town-by-town based definitions and county-based definitions. The New England County Metropolitan area definitions were used here. Using a code common to both Census data and the BRFSS, the GINI Index were assigned to every respondent living in a metropolitan area. The log of each metropolitan area’s per capita income, derived from the Census from the 2000 long
form, was also assigned based on each respondent’s metropolitan area of residence. The resulting sample excluded respondents living outside metropolitan areas, respondents in Puerto Rico, and respondents whose metropolitan area status was missing or otherwise unable to be classified. The sample was analyzed with Stata (Stata, 1984– 2001). Descriptive statistics were calculated for the total sample and six variables including the quartile version of the GINI Index. Next, Pearson Correlation Coefficients were obtained in order to assess direct relationships between hypothesized independent variables and selfrated health status as well as to detect potential problems of collinearity. This analysis used hierarchical linear modeling in the regression analysis to control for the possibility that respondents living in a given metropolitan area might share common outcomes because they live in the same metropolitan area. It adjusts standard errors for clustering in a metropolitan area by using respondent’s metropolitan area identifier. The sample was weighted to reflect unequal probabilities of a respondent being included in the survey. The potential relationship between each independent variable and self-rated health was analyzed separately using multinomial logistical linear regression procedures. Excellent/Very Good vs. Fair/Poor self-rated health was the comparison. A full model was used to assess the relationship between self-reported health and income inequality while controlling for other factors. Control variables were age, sex, smoking status, education, household income (adjusted for household size), metropolitan per capita income (log) Hispanic ethnicity and Black race. The model was repeated using quartiles of the GINI score. Finally, a series of multiple regression models for certain subpopulations, Black, White, Hispanic, Male, and Female, were evaluated to determine if the effects of income inequality varied between groups.
Results The final metropolitan-dwelling sample, once purged of rural residents, Puerto Rico dwelling respondents and others without an identified metropolitan area, consisted of 108,661 interviews. This was smaller than but very reflective of the full BRFSS. It is smaller in part because rural areas are oversampled in the full BRFSS. The final sample had slightly more females than males, had a high number of people who had ever smoked cigarettes and were more likely to have graduated from college than not to have graduated from high school. Less than 10% of the sample lived in households with a total annual income less than $15,000. (Table 1) The range of the GINI index was 35.4–49.04. The mean was 40.61 with a standard deviation of 2.1. While
ARTICLE IN PRESS R. Lopez / Social Science & Medicine 59 (2004) 2409–2419
2413
Table 1 Selected descriptive statstics Number
Total BRFSS Total metropolitan sample Sex Male Female Race/ethnicity White non-hispanic Black non-hispanic Hispanic Ever smoked 100 cigarettes Yes No Education None or only kindergarden 1–8 9–11 12 or GED College 1–3 College 4 yrs or more (College graduate) Income Less than 15,000 Greater than 15,000, refused or did not know GINI index quartiles First (35.4–39.4) Second (39.41–40.8) Third (40.81–42.0) Fourth (42.01–49.04)
Weighted percent
108661 108661
Self-rated health Excellent/Very good (%)
Self rated health Fair/Poor (%)
54.75 51.45
15.46 14.42
43992 64669
48.28 51.72
57.88 54.96
13.23 15.54
82713 11187 8809
70.71 10.53 13.38
60.58 48.84 39.96
12.00 18.50 24.86
50883 57482
46.27 53.48
51.62 60.49
17.23 11.99
188 3413 7251 31483 30415 35654
0.20 4.43 7.43 29.09 27.82 30.81
36.91 19.18 33.15 49.42 59.72 71.07
35.74 44.61 30.79 16.78 12.74 6.12
9770 98891
8.89 91.11
31.04 58.83
35.81 12.34
28522 26002 27258 26879
20.2 24.8 26.2 28.8
60.83 58.12 56.37 52.12
12.75 13.54 14.21 17.25
Note: Full BRFSS data from 2000 BRFSS Codebook, April 17, 2001. http://www.cdc.gov/brfss/surveydata/2000/codebook 00.rtf.
not perfectly normally distributed, the departure from the normal distribution was not great enough to compromise regression results. (skewness=0.64, Kurtosis=3.64). In the final sample, women reported overall poorer self-rated health than men. Whites tended to have better self-rated health status than any other racial/ethnic group and Blacks were about as likely as Hispanics to rate their health fair or poor but more likely than Hispanics to rate their health as very good or excellent. Respondents reporting they had smoked at least 100 cigarettes in their lifetime tended to have worse selfrated health. Very few respondents reported only kindergarten or less education but respondents with more education tended to have better self-rated health. Respondents reporting total household income $15,000 or less were more likely to have lower self-rated health than other respondents. Dividing metropolitan areas into quartiles of the GINI index, there was a definite trend in health status with respondents living in metropolitan areas with higher GINI Index scores (the higher levels of income inequality) tending to report poorer self-rated health. Overall, respondents living in
metropolitan areas with the lowest GINI Index score (lowest income inequality) had the best self-rated health. Considered individually, none of the independent variables had a large correlation with self-rated health status though none of these relationships were likely to have resulted by chance. In analyzing the correlation between individual independent variables, the correlation coefficients were not large. There did not appear to be any potential problem of collinearity between the variables (Table 2). In the univariate analysis, the individual independent variables tended to perform as predicted. The GINI index was strongly and negatively related with self-rated health status. Not smoking, better education, younger age, male gender, being of White race and higher income were all associated with better self-rated health status. The log of the median per capita income of a respondent’s metropolitan area was not associated with the level of self-rated health (Table 3). The multiple regression model used gender, age, education, an individual’s household income, metropolitan per capita income, race and the GINI index score of an individual’s metropolitan area as
ARTICLE IN PRESS R. Lopez / Social Science & Medicine 59 (2004) 2409–2419
2414 Table 2 Correlation coefficients Self-rated GINI health Self-rated health GINI
Never smoked
Education Age
Black
Female
Low income
Hispanic
Per-capita income
1
0.0565 0 Never smoked 0.0984 0 Education 0.2876 0 Age 0.2376 0 Black 0.0533 0 Female 0.031 0 Low income 0.2001 0 Hispanic 0.0709 0 Per-capita 0.0505 income 0
1 0.0023 0.4395 0.0741 0 0.0108 0.0004 0.0667 0 0.0187 0 0.073 0 0.1046 0 0.2294 0
1 0.0945 0 0.0642 0 0.0565 0 0.0754 0 0.0247 0 0.052 0 0.0039 0.1965
1 0.1231 0 0.0833 0 0.0471 0 0.2204 0 0.1837 0 0.1099
1 0.0627 0 0.045 0 0.0685 0 0.1179 0 0.0179
0
0
1 0.0447 0 0.0761 0 0.1006 0 0.0477 0
1 0.0777 0 0.0031 0.3015 0.0025 0.4171
1 0.0737 0 0.0671 0
1 0.005
1
0.101
Pearson product moment correlation coefficients with P values.
Table 3 Univariate regression Independent variable GINI Never smoked Education Age Black Female Income (ratio to poverty threshold) Hispanic Per-capita income (Log)
Self-rated health Good
Fair/Poor
1.044 (1.033, 1.055) 0.8 (0.767, 0.834) 0.685 (0.67, 0.70) 1.07 (1.067, 1.082) 1.33 (1.24, 1.42) 1.074 (1.028, 1.123) 0.073
1.092 (1.078, 1.107) 0.655 (0.613, 0.699) 0.476 (462, 491) 1.19 (1.18, 1.20) 1.62 (1.48, 1.77) 1.24 (1.16, 1.32) 0.733
(0.861, 0.886) 1.84 (1.69, 1.998) 0.808
(0.713, 0.753) 2.86 (2.59, 3.16) 0.827
(0.595, 1.098)
(0.466, 1.467)
Significant at the 0.05 level. Significant at the 0.01 level.
Odds ratios and 95% confidence intervals. Excellent/Very good health is the comparison group.
independent variables and self-rated health as the dependent variable. The multinomial logistic regression analysis compared respondents with Good and Fair/ Poor self-rated health with the Excellent/Very Good self-rated health group. Almost all the independent variables performed as predicted. Having never smoked, having more education, and higher income tended to result in respondents being more likely to rate their health as Excellent/Very Good. Older age, being Black, Hispanic or female tended to make respondents more likely to rate their health as less than Excellent/Very Good. Again, the log of the median per capita income of a respondent’s metropolitan area had no impact on a self-rated health (Table 4). The impact of GINI index scores on self-rated health was dramatic, becoming increasingly important as selfrated health declined. The association of the GINI index score with the likelihood of a respondent rating their health as Fair/Poor was very large and constant whether the raw or quartile GINI score was used. The modeled difference in the risk of Fair/Poor self rated health between the median of the highest quartile of GINI in the sample to the median of the lowest quartile of GINI in the sample would be 29.25% (regression model not shown). The effects of GINI on self-rated health did not show a large variation between population subgroups. The effects of income inequality were greater for Blacks than
ARTICLE IN PRESS R. Lopez / Social Science & Medicine 59 (2004) 2409–2419 Table 4 Multivariate regression
Discussion Full model
Fair/Poor health
Odds ratio 95% confidence interval
GINI
1.040 (1.016, 0.636 (0.557, 0.655 (0.623, 1.038 (1.035, 1.760 (1.565, 0.982 (0.855,
Never smoked Education Age Black Female
Income (ratio to poverty threshold) Hispanic Per-capita income (Log)
1.065) 0.726) 0.689) 1.041) 1.979) 1.129)
0.809 (0.791, 0.828) 2.376 (1.964, 2.875) 1.118 (0.823, 1.518)
Significant at the 0.05 level. Significant at the 0.01 level.
Multinomial logistic regression model. Comparison group is Excellent/Very good health.
Table 5 GINI index odds ratios for selected subpopulations GINI index odds ratio (95% confidence interval) Men Women Black White Hispanic
2415
1.0555 (1.0281, 1.0249 (0.9931, 1.0428 (1.0104, 1.0295 (1.0061, 1.0245 (0.9591,
1.0836) 1.0577) 1.0763) 1.0534) 1.0943)
Multinomial logistic regression model. Comparison group is Excellent/Very good health. Significant at the 0.05 level. Signifficant at the 0.01 level.
Whites (and not statistically significant for Hispanics). Income inequality did not have a statistically significant effect on Females but had a large and significant effect on Males (see Table 5).
This multi-level study found an association between metropolitan area income inequality and self-reported health status in over 100,000 metropolitan dwelling adults in the 2000 BRFSS telephone sample. This association continued after controlling for individual level factors such as income, sex, age, race, metropolitan area per capita income and education. The association was greatest for Female and Black respondents. Association is not necessarily causation, especially in a study utilizing surrogate and group level measures. This study should be interpreted cautiously and in the context of other, similar studies. It is important to consider that the BRFSS excludes people living in group quarters and people living in households without telephones, which may result in some of the poorest and most vulnerable populations not being included in the survey. The effects of their exclusion depend on how income inequality might affect the poorest segments of society. If inequality highly affects poor people, then this study under-reports the overall affects on income inequality. If the absolute burdens of poverty are so great that inequality effects are lessened on this group, this study would over-report the effects of income inequality and would most likely be more reflective of inequality’s effects on the near poor and those with higher incomes. There may be several reasons why this study found a relationship while the earlier, Blakely et al. study, did not (Blakely et al., 2002). The other study used data from the employment-related Current Population Survey (CPS) rather the health-related BRFSS. Perhaps respondents in the CPS were more focused on their employment status and the requirements for participating in the work place. Workers may be more reluctant to report poorer health status to an employment-related survey. Or the effect of household income might have been better evaluated because the employment survey might have had more detailed and accurate income data. Also, growing numbers of workers with low incomes or workers without health insurance might also have affected the survey in other, unknown ways. The earlier study combined income inequality data from 1990 with survey data from 1996 and 1998. While the authors postulated a time lag of effect between inequality and health effects, there may not be such a lag for self-rated health or the lag might smaller. Alternatively, the time lag between inequality and health might be longer and stronger and health outcomes in 2000 might reflect several decades of rising inequality. Perhaps income inequality worsened at the end of 1990s and made its association with health effects stronger by the year 2000. The end of the 1990s economic boom may have had one time, special effects on health as well with respondents reacting to the height of the late 1990s stock market
ARTICLE IN PRESS 2416
R. Lopez / Social Science & Medicine 59 (2004) 2409–2419
bubble. There could have been changed in welfare or other benefits that have resulted in a net worsening of the effects of inequality. After one study found a relationship between metropolitan income inequality and mortality in the United States but not Canada, the authors postulated that a higher level of social services provided by government could mitigate the impact of inequality (Ross, 2000). The earlier study did not control for respondents who were Hispanic or smoked cigarettes, two factors that were found in this study to affect self-rated health. Given the relative infrequency of the US Census, the earlier study used the best available data at that time. However, this study using the newer, more coordinated and perhaps more appropriate data from 2000 found a relationship where the earlier study, relying on an employment survey did not. The smaller multi-level study of 60 metropolitan areas may have found different results because its subset of metropolitan areas may not be representative of all US metropolitan areas (Sturm & Gresenz, 2002). The distribution of GINI values across US metropolitan areas is large, almost as great as the differences found between developed countries. For example, Sweden, the developed country with the greatest equality, had a 1995 GINI index score of 0.221 compared to a US 1997 GINI index score of 0.372 (Smeeding, 2002). The 14 point spread of GINI index scores between US metropolitan areas is comparable to the 20% growth of US income inequality over the past three decades. This suggests that the contribution of income inequality to differences in health status between metropolitan areas might be important. Income inequality may affect people’s self-reported health by reducing access to the best-quality housing or health care or it may be part of a chain of causality that ultimately affects the environment of all people. While the differences in the effects of the GINI index were not large between subgroups, perhaps they reflect some of the underlying social patterns in the distribution of the health effects of income inequality. The GINI index’s odds ratio was greater for Blacks than Whites, which may be indicative of the social status of non-Whites in the United States and suggesting that vulnerable populations bear a disproportionate burden of the impacts of income inequality. In contrast, the GINI coefficient was greater for males than females, perhaps suggesting that the overall status of women might be a greater determinant of health than income inequality. Income inequality may increase stress that then impacts health, or people in areas with greater income inequality unfavorably compare themselves to others. A number of explanatory pathways connecting income inequality and health have been proposed including individual income, psychosocial environment and neomaterial interpretations (rising overall income results in increased perceived need for material goods such as cars,
telephones, internet access, etc.) (Lynch, Smith, Kaplan, & House, 2000; Marmot, 2001; Marmot & Wilkinson, 2001). In modeling the potential effects of bias on individual and population level health, it was found that increased income inequality should have an effect on both levels (Wildman, 2001). Alternatively, it has been proposed inequality in educational attainment is the key factor driving the link between income inequality and mortality (Muller, 2002). This proposed causal pathway has been disputed, however (Blakely & Kawachi, 2002). Looking in more detail at explanations of the association between income inequality and health, a decline in social capital has been identified as key to the association (Costa & Kahn, 2001; Gold, Kennedy, Connell, & Kawachi, 2002; Kawachi & Kennedy, 1997b; Kawachi, Kennedy, Lochner, & Prothow-Stith, 1997). Others point to stress as the mechanism that connects the two (Brunner, 1997). These explanations may not be mutually exclusive. This study found that the degree of metropolitan area income inequality, as measured by the GINI index, is associated with self-rated health in the United States. To the extent that self-rated health reflects actual health status and may ultimately be associated with an individual’s quality of life, mortality risk and personal finances, and may affect society’s overall quality of life, overall health status and total health care expenditures, this relationship is important. High levels of metropolitan area income inequality appear to have significant health impacts on individuals and the United States as a whole. This potential relationship between income inequality and health is cause for concern about the economic structure of the US as well.
Acknowledgements The author wishes to thank Leslie Boden, Richard Clapp and H. Patricia Hynes of the Boston University School of Public Health for their comments on the draft of this article. This publication was made possible by grant number 5F31E11219-01 from the National Institute of Environmental Health Sciences (NIEHS), NIH. Its contents are solely the responsibility of the author and do not necessarily represent the official views of NIEHS, NIH.
References Amiel, Y. (1998). The subjective approach to the measurement of income inequality. London: Suntory and Toyota International Centres for Economics and Related Disciplines. Blakely, T., & Kawachi, I. (2002). Education does not explain association between income inequality and health. British Medical Journal, 324, 1336–1337.
ARTICLE IN PRESS R. Lopez / Social Science & Medicine 59 (2004) 2409–2419 Blakely, T., Lochner, K., & Kawachi, I. (2002). Metropolitan area income inequality and self-rated health—a multi-level study. Social Science and Medicine, 54, 65–77. Bosworth, H., Siegler, I., Brummett, B., Barefoot, J., Williams, R., Clapp-Channing, N., Lytle, B., & Mark, D. (1999). The association between self-rated health and mortality in a well-characterized sample of coronary artery disease patients. Medical Care, 37, 1226–1236. Brodish, P., Massing, M., & Tyroler, H. (2000). Income inequality and all-cause mortality in the 100 counties of North Carolina. Southern Medical Journal, 93, 386–391. Brunner, E. (1997). Socioeconomic determinants of health: Stress and the biology of inequality. British Medical Journal, 314, 1472–1476. CDC. (2000). Behavioral risk factor surveillance system survey data. Atlanta, GA; US Department of Health and Human Services, Centers for Disease Control and Prevention. Chevan, A., & Stokes, R. (2000). Growth in family income inequality, 1970–1990: Industrial restructuring and demographic change. Demography, 37, 365–380. Collins, W., & Margo, R. (2000). Residential segregation and socioeconomic outcomes: When did ghettos go bad? Economics Letters, 69, 239–243. Cooper, R., Kennely, J., Durazo-Arvizu, R., Oh, H., Kaplan, G., & Lynch, J. (2001). Relationship between premature mortality and socioeconomic factors in black and white populations of US metropolitan areas. Public Health Reports, 116, 464–473. Costa, D., & Kahn, M. (2001). Understanding the decline in social capital, 1952–1998. Cambridge, MA: National Bureau of Economic Research. Coulter, P. (1989). Measuring inequality. A methodological handbook. Boulder, CO: Westview Press. Cowell, F. (1998). Measurement of inequality. London: Suntory and Toyota International Centres for Economics and Related Disciplines. Cubbin, C., Hadden, W., & Winkleby, M. (2001). Neighborhood context and cardiovascular disease risk factors: The contribution of material deprivation. Ethnicity and Disease, 11, 687–700. Daly, M., Crews, A., & Burkhauser, R. (1997). A new look at the distributional effects of economic growth during the 1980s: A comparative study of the united states and germany. Federal Reserve Bank of San Francisco Economic Review, 2, 18–37. Deaton, A., & Lubotsky, D. (2003). Mortality, inequality and race in American cities and states. Social Science & Medicine, 56, 1139–1153. Diez-Roux, A. (1998). Bringing context back into epidemiology: Variables and fallacies in multilevel analysis. American Journal of Public Health, 88, 216–222. Diez-Roux, A., Link, B., & Northridge, M. (1999). Income inequality and cardiovascular disease risk factors. Abstract of the 39th annual conference on cardiovascular disease epidemiology and prevention. Circulation, 99, 1112. Ecob, R., & Smith, G. (1999). Income and health: What is the nature of the relationship? Social Science and Medicine, 48, 693–705. Ellison, G. (2002). Letting the GINI out of the bottle? Challenges facing the relative income hypothesis. Social Science and Medicine, 54, 561–576.
2417
Ewing, R., Schmid, T., Killingsworth, R., Zlot, A., & Raudenbush, S. (2003). Relationship between urban sprawl and physical activity, obesity. and morbidity. American Journal of Health Promotion, 18, 47–57. Finch, B., Hummer, R., Reindl, M., & Vega, W. (2002). Validity of self rated health among Latino(a)s. American Journal of Epidemiology, 155, 755–759. Frumkin, H. (2002). Urban sprawl and public health. Public Health Reports, 117, 201–217. Gold, R., Kennedy, B., Connell, F., & Kawachi, I. (2002). Teen births, income inequality, and social capital: Developing an understanding of the causal pathway. Health and Place, 8, 77–83. Gravelle, H., Wildman, J., & Sutton, M. (2002). Income, income inequality and health: What can we learn from aggregate data? Social Science and Medicine, 54, 577–589. Guest, A., Almgren, G., & Hussey, J. (1998). The ecology of race and socioeconomic distress: Infants and smoking. Demography, 35, 23–34. Hales, S., Howden-Chapman, P., Salmond, C., Woodward, A., & Mackenbach, J. (1999). National infant mortality rates in relation to gross national product and distribution of income. The Lancet, 354, 2047. Idler, E., & Benyamini, Y. (1997). Self-rated health and mortality: A review of twenty-seven community studies. Journal of Health and Social Behavior, 38, 21–37. Jargowsky, P. (1996). Take the money and run: Economic segregation in US metropolitan areas. American Sociological Review, 61, 984–998. Jones, A., & Weinberg, D. (2000). The changing shape of the nation’s income distribution 1947–1998. Washington, DC: US Department of Commerce, Economics and Statistics Administration, US Bureau of the Census. Kaplan, G., Pamuk, E., Lynch, J., Cohen, R., & Balfour, J. (1996). Income inequality and mortality in the United States: Analysis of mortality and potential pathways. British Medical Journal, 312, 999–1003. Kasarda, J. (1995). Industrial restructuring and the changing location of jobs. In State of the union: America in the 1990s, Vol. 1. Economics trends (pp. 215–267). New York: Russell Sage Foundation. Kawachi, I., & Kennedy, B. (1997a). The relationship of income inequality to mortality: Does the choice of indicator matter? Social Science and Medicine, 45, 121–127. Kawachi, I., & Kennedy, B. (1997b). Socioeconomic determinants of health: Health and social cohesion: Why care about income inequality? British Medical Journal, 314, 1037–1040. Kawachi, I., Kennedy, B., Lochner, K., & Prothow-Stith, D. (1997). Social capital, income inequality, and mortality. American Journal of Public Health, 87, 1491–1498. Kawachi, I., Kennedy, B., & Wilkinson, R. (1999). The society and population health reader, Vol. 1. Income inequality and health. New York: The New Press. Kennedy, B., Kasl, S., & Vaccarino, V. (2001). Repeated hospitalizations and self-rated health among the elderly: A multivariate failure time analysis. American Journal of Epidemiology, 153, 232–241. Kennedy, B., Kawachi, I., & Prothow-Stith, D. (1996). Income distribution and mortality: Cross sectional ecological study of the robin hood index in the united states. British Medical Journal, 312, 1004–1007.
ARTICLE IN PRESS 2418
R. Lopez / Social Science & Medicine 59 (2004) 2409–2419
Krugman, P. (2002). For Richer. New York times sunday magazine. October 20 (pp. 62–84). Lantz, P., House, J., Lepkowski, J., Williams, D., Mero, R., & Chen, J. (1998). Socioeconomic factors, health behaviors, and mortality. Journal of the American Medical Association, 279, 1703–1708. LeClerer, F., & Soobader, M. (2000). The effect of income inequality on the health of selected US demographic groups. American Journal of Public Health, 90, 1892–1897. Lobmayer, P., & Wilkinson, R. (2002). Inequality, residential segregation by income, and mortality in US cities. Journal of Epidemiology and Community Health, 56, 183–187. Lopez, R. (2002). Segregation and Black/White differences in exposure to air toxics in 1990. Environmental Health Perspectives, 110(2), 289–295. Lynch, J., Davey-Smith, G., Hillemeier, M., Shaw, M., Rahunathan, T., & Kaplan, G. (2003). Income inequality, the psychosocial environment, and health: Comparisons of wealthy nations. The Lancet, 358, 194–200. Lynch, J., Kaplan, G., Pamuk, E., Cohen, R., & Heck, K. (1998). Income inequality and mortality in metropolitan areas of the United States. American Journal of Public Health, 88, 1074–1080. Lynch, J., Smith, G., Kaplan, G., & House, J. (2000). Income inequality and mortality: Importance to health of individual income, pyschosocial environment, or material conditions. British Medical Journal, 320, 1200–1204. Mackenbach, J. (2002). Income inequality and population health: Evidence favouring a negative correlation between income inequality and life expectancy has disappeared. British Medical Journal, 324, 1–2. Marmot, M. (2001). Income inequality, social environment and inequities in health. Journal of Policy Analysis and Management, 20, 156–159. Marmot, M., & Wilkinson, R. (2001). Psychosocial and material pathways in the relation between income and health: A response to Lynch et al. British Medical Journal, 322, 1233–1236. Massey, D. (1996). The age of extremes: Concentrated affluence and poverty in the twenty-first century. Demography, 33, 395–412. McLaughlin, D., & Stokes, S. (2002). Income inequality and mortality in US counties: Does minority racial concentration matter? American Journal of Public Health, 92, 99–104. Miilunpalo, S., Vuori, I., Oja, P., Pasanen, M., & Urponen, H. (1997). Self-rated health status as a health measure: The predictive value of self-reported health status on the use of physician services and on mortality in the working-age population. Journal of Clinical Epidemiology, 50, 517–528. Morrill, R. (2000). Geographic variation in change in income inequality among US states, 1970-1990. The Annals of Regional Science, 34, 109–130. Muller, A. (2002). Education, income inequality, and mortality: A multiple regression analysis. British Medical Journal, 324, 23–25. Nielsen, F., & Alderson, A. (1997). The Kuznets curve and the great u-turn: Income inequality in US counties, 1970–1990. American Sociological Review, 62, 12–33. Pickett, K., & Pearl, M. (2001). Multi-level analysis of neighborhood Socioeconomic context and health outcomes:
A critical review. Journal of Epidemiology and Community Health, 55, 111–122. Piketty, T., & Saez, E. (2003). Income Inequality in the United States, 1914–1998. Quarterly Journal of Economics, 118, 1–39. Polednak, A. (1997). Segregation, poverty, and mortality in urban african americans. New York: Oxford University Press. Ravallion, M., Heil, M., & Jalan, J. (2000). Carbon emissions and income inequality. Oxford Economic Papers, 52, 651–669. Ren, X., Amick, B., & Williams, D. (1999). Racial/ethnic disparities in health: The interplay between discrimination and socioeconomic status. Ethnicity and Disease, 9, 151–155. Rodgers, G. (1979). Income and inequality as determinants of mortality: An international cross-section analysis. Population Studies, 33, 243–251. Ross, C. (2000). Walking, exercising, and smoking: Does neighborhood matter? Social Science & Medicine, 51, 265–274. Schalick, L., Hadden, W., Pamuk, E., Navarro, V., & Pappas, G. (2000). The widening gap in death rates among income groups in the United States from 1967–1986. International Journal of Health Services, 30, 13–26. Shi, L., & Starfield, B. (2001). The effect of primary care physician supply and income inequality on mortality among Blacks and Whites in US metropolitan areas. American Journal of Public Health, 91, 1246–1250. Smeeding, T. (2000). Changing income inequality in oecd countries: Updated results from the luxembourg income study. Differdange, Luxembourg: Luxembourg Income Study. Smeeding, T. (2002). Globalization, inequality, and the rich countries of the G-20: Evidence from the Luxembourg Income Study (LIS). Center for Policy Research, Maxwell School of Citizenship and Public Affairs, Syracuse University, Syracuse, NY. Smeeding, T., Rainwater, L., & Burtless, G. (2000). United states poverty in a cross-national context. Differdange, Luxembourg: Luxembourg Income Study. Soobader, M., & LeClere, F. (1999). Aggregation and the measurement of income inequality: Effects on morbidity. Social Science and Medicine, 48, 733–744. Stata. (1984-2001). Version 7.0 (7.0 ed.). College Station, TX: Stata Corporation. Sterling, T., Rosenbaum, W., & Weinkam, J. (1993). Income, race and mortality. Journal of the National Medical Association, 85, 906–911. Sturm, R., & Gresenz, C. (2002). Relations of income inequality and family income to chronic medical conditions and mental health disorders: National survey. British Medical Journal, 324, 20–22. Syme, L. (1998). Social and economic disparities in health: Thoughts about intervention. The Milbank Quarterly, 76, 493–505. Wildman, J. (2001). The impact of income inequality on individual and societal health: Absolute income, relative income and statistical artifacts. Health Economics, 10, 357–361. Wilkinson, R. (1994). The epidmiological transition: From material scarcity to social disadvantage? Daedalus, 123, 61–77.
ARTICLE IN PRESS R. Lopez / Social Science & Medicine 59 (2004) 2409–2419 Wolfson, M., Kaplan, G., Lynch, J., Ross, N., & Backlund, E. (1999). Relation between income inequality and mortality: Empirical demonstration. British Medical Journal, 319, 953–955. Russ Lopez received his Master of City and Regional Planning from the Kennedy School of Government at Harvard
2419
University and his Doctor of Science in Environmental Health from the Boston University School of Public Health. A native of California, his current research focuses on the health effects of metropolitan area social and physical environments including the impacts of racial segregation, income inequality and urban sprawl. He currently holds positions at both Boston University and Brown University.