Atl Econ J (2013) 41:89-91 DOI 10.1007/s11293-012-9342-2 ANTHOLOGY Social Capital and Income Inequality in the United States Rati Ram Published online: 17 October 2012 # International Atlantic Economic Society 2012 Many scholars have explored in recent years various correlates and consequences of social capital along with discussions of the concept. For example, relationship of social capital with population happiness, health, income, economic growth, and human development has been researched by several scholars. However, very few studies have considered the relationship between social capital and income inequality. One exception to that is the recent work by Robison et al. (Journal of SocioEconomics, 2011) which proposed a theoretical link between social capital and income distribution and conducted an empirical exploration for the U.S. states for the census years 1980, 1990, and 2000. Their key measure of social capital was somewhat narrowly focused on percent of households headed by a single female with children. Given the importance of the topic, it is of interest to work with a broader and more common proxy for social capital and also to use more recent inequality indexes. The theoretical framework suggested by Robison et al. linked greater social capital with increased trade across individuals or households, which raises average income but has an ambiguous effect on income distribution. It is possible to propose a simpler conceptual reasoning that generates a sharper implication about income inequality. It is reasonable to suggest that higher social capital or social trust is associated with a stronger sense of fairness and consideration for others, particularly relative to giving and receiving compensation for market work. Such a sense of fairness and consideration may be expected to mitigate wage and income inequalities. It is thus plausible to postulate increased social capital as an equalizer. This short paper pursues the foregoing theme by using fairly good social capital data and the most recent information on income inequality for the U.S. states. Social capital data are taken from the compilation by Bjornskov (Applied Research in Quality of Life, 2008). The numbers are for social trust, which is a primary indicator of social capital at the macro level, and are averaged over the period 1990-1998. The inequality measure is state-level household Gini index from R. Ram (*) Department of Economics, Illinois State University, Normal, IL 61790-4200, USA e-mail: rram@ilstu.edu 90 R. Ram American Community Survey (ACS) for the years 2006-2010 available at the Census Bureau's American Fact Finder. By way of a simple control variable, average household income from ACS for the year of 2005 is also taken from American Fact Finder. To provide a feel for the data, simple descriptive statistics for the variables are shown below. Mean SD Min. Max. Social trust (1990-1998) 31.37 3.23 24.69 37.97 Gini index (2006-2010) 0.45 0.02 0.41 0.54 Mean household income (2005, 000$) 61.76 8.77 45.98 87.47 The following regression estimates show the association between logarithms of social capital (LTRUST), income inequality (LGINI), and average household income (LY05), with robust t-statistics in parentheses. LGINI 0:594 0:227LTRUST 0:053LY05 1:80 5:76 1:55 LGINI 0:0850:207LTRUST 0:72 5:95 R2 : 0:31 R2 : 0:34 N 48 N 48 The estimates show that social capital is a highly significant equalizer. 1 % increase in social capital (trust) is expected to lower the Gini index by about 0.20 %. The following observations should also be of interest. 1. The role of income is marginal and lacks statistical significance. The sign on the term indicates a tendency for income inequality to increase with income, supporting several studies that have documented increasing inequality since the 1970s. 2. Although the estimates are reported here for logarithmic versions of the variables, an almost identical position emerges if the variables are entered linearly. Several other variants of the model yield a very similar scenario. 3. There is lack of indication of a significant quadratic relation between social capital and income inequality. If a quadratic term for trust is added, adjustedR2 goes down, neither LTRUST nor its square is significant at any meaningful level, and both t-statistics are below unity. 4. The parsimonious model and the possibility of a feedback from income inequality can lead to a reasonable concern about the quality of the estimates. However, several considerations should mitigate that concern. First, the social-capital variable is for the period 1990-1998 and is temporally predetermined relative to the Gini index, which is for 2006-2010. Second, Halbert White's well-known test, which has the joint null hypothesis of homoscedastic error and nomisspecification, is not rejected. If squares of residuals from the model (with income) are regressed on squares and cross-products of the regressors, chi-square statistic with six degrees of freedom is 5.05, which is not significant at any Social Capital and Income Inequality 91 sensible level. Third, the well-known RESET test also suggests lack of a significant specification error. In a simple version of the test, when the square of the predicted Gini is added to the regression, t-statistic for the squared term is 0.87. There is thus considerable evidence of social capital being a significant equalizer in the context studied. Policy measures that augment social capital might, besides yielding other benefits suggested in the literature, also be expected to mitigate the rising trend in inequality in the United States. Copyright of Atlantic Economic Journal is the property of Springer Science & Business Media B.V. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. RESEARCH AND PRACTICE Income Inequality in Health at All Ages: A Comparison of the United States and England Melissa L. Martinson, PhD Population health is worse in the United States than England, despite the much higher level of health care spending in the United States.1---3 Well-documented health differentials between the 2 countries exist for a wide variety of health measures at all ages.3 However, questions remain about the extent of cross-country differences in health disparities, in particular whether income inequalities in health are higher in the United States than in England across the life span. Health comparisons between the United States and England are interesting because, despite many societal similarities, there are differences in health care provision, social protection policies, and societal inequality between the 2 countries.4---6 In particular, previous studies have postulated that differences in health care systems between the United States and England (as well as other European countries) may account for the relatively poorer health in the United States as well as the greater health inequalities among Americans.2,7---9 Additionally, whereas both countries have liberal, residual welfare states, Britain has slightly lower income inequality and a greater focus on alleviating poverty, particularly among children, although it still lags behind many other European countries.5,10 A handful of studies have examined the magnitude of socioeconomic disparities in health in the United States and England, and the results are decidedly mixed. Banks et al. concluded that income- and education-based health gradients among older adults are steeper in the United States than in England,1 whereas Avendano et al. found that the wealth gradient for older adults is similar in the 2 countries.2 A series of articles examining the income gradient in health among children has produced conicting ndings.7,11,12 One comparative study of self-rated health by income, occupational, and poverty status in the United States and United Kingdom included middle-age adults; it found better health in the Objectives. I systematically examined income gradients in health in the United States and England across the life span (ages birth to 80 years), separately for females and males, for a number of health conditions. Methods. Using data from the National Health and Nutrition Examination Survey for the United States (n = 36 360) and the Health Survey for England (n = 55 783), I calculated weighted prevalence rates and risk ratios by income level for the following health risk factors or conditions: obesity, hypertension, diabetes, low high-density lipoprotein cholesterol, high cholesterol ratio, heart attack or angina, stroke, and asthma. Results. In the United States and England, the income gradients in health are very similar across age, gender, and numerous health conditions, and are robust to adjustments for race/ethnicity, health behaviors, body mass index, and health insurance. Conclusions. Health disparities by income are pervasive in England as well as in the United States, despite better overall health, universal health insurance, and more generous social protection spending in England. (Am J Public Health. 2012;102:2049-2056. doi:10.2105/AJPH.2012.300929) United Kingdom than in the United States, as well as a greater likelihood of health improving over time in the United Kingdom.13 However, that study did not examine the income gradient in health. No study to date has compared socioeconomic gradients in health throughout the life span in the United States and England. There is reason to believe that the income gradient in health is largest in middle to later adulthood, because the income gradient in health widens with age among children and narrows with age among the elderly because of increased mortality among low-income people.11,12,14,15 This widening of income disparities with age would likely be similar in the United States and England if the income gradient age pattern is being driven by a higher vulnerability to health shocks among low-income individuals than among high-income individuals. However, if an increase in income disparities with age is due to low-income individuals' lack of ability to respond to health shocks (e.g., through lack of insurance), one would expect to see a more rapid increase in income-based November 2012, Vol 102, No. 11 | American Journal of Public Health health inequalities in the United States than in England because of the highly variable US health care system. The extent to which income gradients in health and health trajectories differ in the 2 countries by age is not known, but it represents an important area of inquiry for understanding the processes leading to the well-documented cross-country differences in health. In this study, I describe and compare the extent of income-based socioeconomic gradients in health in the United States and England from birth to 80 years, for both females and males, using a large set of biological and self-reported health measures. This study provides a comprehensive description of the magnitude of income inequalities in health in the 2 countries. METHODS The National Health and Nutrition Examination Survey (NHANES) for the United States and the Health Survey for England (HSE) were used in this study. Both are large, nationally representative health surveys that have Martinson | Peer Reviewed | Research and Practice | 2049 RESEARCH AND PRACTICE comparable measures of health assessed through both physical examinations and interviews. The NHANES is a comprehensive survey conducted by the National Center for Health Statistics in the United States continuously since 1999.16 For the analyses presented here, I used data from years 1999 to 2006 of the continuous survey. Of the 41 474 participants from 1999 to 2006, individuals aged older than 80 years were removed. Additionally, about 8% of the sample was missing income data. The nal analytic sample was 36 360. Sample sizes varied by health measure because some conditions were assessed only for certain age groups. The HSE is an annual cross-sectional survey of private households in England conducted by the Joint Health Surveys Unit of the National Centre for Social Research and University College London.17---20 I used the 2003---2006 surveys for these analyses because, starting in 2003, weights became available making it possible to pool multiple years of data while maintaining the representativeness of the English population. The number of respondents in the 2003---2006 surveys was 71 717. The analysis sample excluded individuals older than 80 years. Approximately 19% of respondents had missing income data and were excluded from the analytic sample. The nal analytic sample was 55 783. It is worth noting that older adults were more likely to have missing data on income in both the US and English samples. Some biological measures were collected from representative subsamples and some questions were asked only of participants in certain age groups. including the United States and Great Britain.22 I dened hypertension as a mean systolic blood pressure of 140 millimeters of mercury or higher, mean diastolic blood pressure of 90 millimeters of mercury or higher, or reports of current treatment of hypertension with prescription medication.23 I assessed diabetes from glycosylated hemoglobin tests (HbA1c 6.5%).24 I categorized HDL as low (< 40 mg/dL), normal (40---59 mg/dL), or high (> 59 mg/dL); in addition, I used a binary measure of low vs normal or high HDL.25 In the absence of a lowdensity lipoprotein cholesterol measure, I used the total-cholesterol-to-HDL-cholesterol ratio.26 High cholesterol ratio was dened as a totalcholesterol-to-HDL-cholesterol ratio of 5:1 or above, although results were not sensitive to the ratio cutoff used. I used high C-reactive sensitivity protein, a biomarker for inammation, conditions were measured for individuals aged 12 years and older. An advantage of using the biological measures was the ability to capture health risk among individuals who were young and for whom illness was relatively rare. For adults, the categories of body mass index (BMI; dened as weight in kilograms divided by the square of height in meters) were based on the World Health Organization's standard.21 The categories were normal (BMI = 18.5---24.9 kg/m2), overweight (BMI = 25---29.9 kg/m2), obese (BMI 30 kg/m2), and underweight (BMI < 18.5 kg/m2). Obesity was specically examined as a health risk. For children (through age 17 years), age- and gender-specic thresholds were determined using the International Obesity Taskforce denition of the BMI categories (normal, overweight, and obese), which was based on BMI curves in 6 countries, TABLE 1Sample Characteristics of Survey Respondents in the United States and England, by Income Tercile: US National Health and Nutrition Examination Survey (1999-2006) and Health Survey for England (2003-2006) United States (n = 36 360) England (n = 55 783) Low Middle High Low Middle High 33.7 (0.4) 35.1 (0.4) 34.5 (0.4) 36.1 (0.3) 35.9 (0.3) 35.3 (0.3) 46.8 49.8 51.1 44.9 51.0 52.2 53.2 50.2 48.9 55.1 49.0 47.8 Non-Hispanic White 50.7 70.1 82.6 79.2 90.0 91.8 Hispanic 24.7 13.1 6.0 NA NA NA NA NA NA 10.5 4.7 4.8 18.7 11.0 6.4 8.2 3.9 2.4 Mean age (SE), y Gender, % Male Female Race/ethnicity, %a Asian Non-Hispanic Black 5.9 5.8 4.9 2.1 1.4 1.0 Measures of Health Cigarette smoking, % Other 31.2 23.7 15.8 36.9 23.3 16.9 There were several comparable health measures based on physical examinations or laboratory reports in the NHANES and HSE. I included the following risk factors or conditions in this study: obesity, hypertension, diabetes, low high-density lipoprotein (HDL) cholesterol, high cholesterol ratio, and high C-reactive protein. The NHANES and HSE documentation indicated that very similar protocols were used in the 2 countries. Obesity was calculated for respondents aged 4 to 80 years, C-reactive protein was measured for respondents aged 18 to 80 years, and the other Drinking 5 d/wk (age 20 y), % No health insurance, % 4.5 29.9 6.6 14.5 9.5 5.9 12.5 NA 18.7 NA 26.6 NA 0-12 y (US), 0-11 y (England) 64.6 45.1 24.7 50.4 29.9 14.4 13-15 y (US), 12-13 y (England) 27.0 33.3 30.4 23.1 26.3 17.5 8.4 21.6 44.9 26.6 43.8 68.1 2050 | Research and Practice | Peer Reviewed | Martinson Education, % 16 y (US), 14 y (England) Note. NA = not applicable. Because obesity was categorized differently for those younger than 18 years than for adults and because C-reactive protein was assessed only for those at least 18 years old, the adolescent group was categorized as 12-17 years and the young adult group as 18-34 years for measures of obesity and C-reactive protein. Unless otherwise noted, all gures pertain to individuals aged birth to 80 years. a Hispanic ethnicity was not available for England (individuals who are Hispanic could have classied themselves in any of the racial groups). Asian race was not available for the United States (individuals who are Asian are included in the \"other\" race/ ethnic category). American Journal of Public Health | November 2012, Vol 102, No. 11 RESEARCH AND PRACTICE TABLE 2Prevalence of Health Outcomes Among Female Respondents, by Income Tercile and Age Group: US National Health and Nutrition Examination Survey (1999-2006) and Health Survey for England (2003-2006) United States (n = 36 360), % Health Outcome and Age, Years England (n = 55 783), % Low Middle High Low Middle High 12-19 0.6 0.3 0.5 0.0 0.0 0.0 20-34 35-49 1.1 6.3** 2.2* 2.2 0.5 1.7 1.1 2.5** 1.2 1.2 1.1 0.5 50-64 15.0** 7.3 5.5 8.8** 3.3 2.1 65-80 14.3** 15.0** 7.4 13.5** 12.9** 5.7 4-11 15.3** 10.9 10.2 9.3 6.3 5.6 12-17 17.3** 18.0** 10.5 10.8 6.0 9.0 18-34 35.5** 31.6** 21.4 19.2** 13.0* 8.4 35-49 50-64 45.2** 47.4** 35.0 41.5* 30.8 31.9 26.6** 28.3 22.1** 25.5 15.8 23.7 65-80 38.4* 40.6** 29.4 24.8 27.2* 20.6 12-19 13.6** 11.9* 7.4 5.0 6.9 9.7 20-34 16.2** 11.4* 6.6 9.9** 5.7 3.1 35-49 14.6** 11.5* 6.6 8.1** 3.3 2.3 50-64 13.0** 7.0 5.1 6.3** 2.0 1.4 10.9** 6.7 5.4 5.3* 3.4 2.4 5.1** Diabetes Obesity Low HDL 65-80 High cholesterol ratio 12-19 2.4 3.6 3.3 4.8 20-34 12.0* 6.0** 12.3* 7.8 8.7** 4.9 2.3 35-49 21.8** 14.4 11.4 12.2** 7.3 5.3 50-64 21.8** 18.6** 12.1 17.0** 11.8 11.0 65-80 19.3* 15.4 13.4 15.9 11.9 14.4 18-34 35-49 41.3 51.7** 39.9 44.1 38.1 39.7 34.7 32.0** 29.2 27.7* 29.4 22.3 50-64 55.8** 50.6* 39.7 44.5** 33.6 29.9 65-80 54.9* 49.6 45.7 44.9* 45.6* 37.5 12-19 0.4 0.3 0.5 1.2 1.1 0.5 20-34 3.6* 2.6 1.3 4.1 3.3 3.6 35-49 23.8* 17.0 17.8 14.7 14.7 11.7 50-64 65-80 53.0** 78.9* 50.0** 76.1 38.6 70.1 48.0** 70.5** 33.8 65.5 30.0 61.2 Birth-3 11.7** 6.4 4.9 2.9 1.6 1.0 4-11 13.7* 9.7 8.6 8.1* 8.2* 3.3 12-19 17.4 16.4 19.5 7.2 4.8 5.2 20-34 17.4 13.4 15.8 7.5 5.9 6.2 35-49 15.8 15.6 14.3 9.7** 6.0 5.0 50-64 65-80 15.6** 15.1* 14.2* 12.3 9.7 10.7 9.4 8.7 7.2 9.8 6.6 5.8 High C-reactive protein Hypertension Asthma ever diagnosed Continued November 2012, Vol 102, No. 11 | American Journal of Public Health to classify individuals as low risk (< 1 mg/L), medium risk (1---3 mg/L), or high risk (> 3 mg/L) and to create a binary measure of high vs low or medium health risk.27---29 The self-reported health conditions were based on participants' responses to standard survey questions. These were chosen because of comparability between the 2 data sets and were used in previous research comparing health in the United States and England.3 Responses indicated whether the individual was ever told by a doctor that he or she had had a heart attack or angina, a stroke, or asthma (in England, the HSE simply asks whether the individual has asthma). Except asthma, all of these measures were available for individuals at least 20 years of age. Asthma was available for all ages. Age Groups and Income Measure I categorized age into broad groups that correspond to the Centers for Disease Control and Prevention's Stages of Life. The categories were as follows: infants (birth---3 years), children (4---11 years), adolescents (12---19 years), young adults (20---34 years), middle-age adults (35---49 and 50---64 years), and older-age adults (65---80 years). The primary independent variable of interest in this study was income-based socioeconomic status, which I constructed from the family income variable available in both the HSE and NHANES at present value, adjusted by the Organisation for Economic CoOperation and Development's (OECD) square root equivalence scale, and then divided into equal terciles by using the sample weights.30 The square root equivalence scale has been used in OECD publications on international income inequality and poverty since 2008. Additionally, because of the pooling of multiple years of data in the NHANES and HSE, I adjusted the measure for cost of living to the year 2006 using the Consumer Price Indexes for the United States and the United Kingdom.31,32 Use of terciles rather than absolute income adjusted for differences in average levels and the income distribution across the 2 countries; previous studies of older adults have used this method.1,2,33 I also adjusted the terciles by age group because of the uctuations in income throughout the life span. Martinson | Peer Reviewed | Research and Practice | 2051 RESEARCH AND PRACTICE TABLE 2Continued Heart attack 20-34 0.7 0.5 0.3 0.3 0.0 0.1 35-49 2.3 1.4 1.2 1.0* 1.5* 0.2 50-64 8.9** 4.4 2.0 3.1* 1.7 65-80 18.4** 10.4 7.6 11.2* 15.4** 7.1 0.1 Stroke 20-34 5.4** 0.7 0.5 0.2 0.3 0.0 35-49 2.5* 1.6 0.6 0.5 0.4 0.3 50-64 5.1** 3.2 1.3 3.9** 0.9 0.8 65-80 10.0** 8.3* 4.1 6.6 7.2 5.3 Note. HDL = high-density lipoprotein cholesterol. *P < .05; **P < .01 (for prevalence among low- and middle-income vs high-income individuals). TABLE 3Prevalence of Health Outcomes Among Male Respondents, by Income Tercile and Age Group: US National Health and Nutrition Examination Survey (1999-2006) and Health Survey for England (2003-2006) United States (n = 36 360), % Health Outcome and Age, Years England (n = 55 783), % Analysis Low Middle High Low Middle High 12-19 0.6 0.4 0.7 1.0 1.1 0.0 20-34 35-49 1.8 7.7** 0.8 4.5 1.0 3.5 1.1 5.7** 0.5 2.0 0.3 2.0 50-64 17.0** 10.2 8.3 9.0 7.4 6.9 65-80 19.6** 14.6 10.2 13.7 16.0** 9.5 4-11 14.1** 11.9** 6.2 9.2 7.2 5.9 12-17 19.3** 15.1 11.5 8.1 6.3 3.5 18-34 24.0 27.0* 21.6 13.2 13.8 14.8 35-49 50-64 32.5 38.2 35.8 36.8 31.2 34.7 27.6 30.6 19.9 25.1 25.0 25.2 65-80 33.1 35.8** 28.6 21.0 24.8* 17.8 12-19 22.4 22.5 23.4 21.0 20.3 16.9 20-34 33.0** 30.3* 23.4 18.6 14.4 13.6 35-49 33.1 32.7 30.0 17.9** 15.2* 11.5 50-64 31.6** 28.1 23.5 14.7 14.1 16.7 31.3* 27.6 23.8 18.1* 20.2** 12.3 Diabetes Obesity Low HDL 65-80 High cholesterol ratio 12-19 9.5 8.1 10.1 7.4 8.7 3.8 20-34 28.8 27.8 24.7 22.9 18.4 17.6 35-49 41.0 40.6 37.8 32.1 30.5 27.8 50-64 42.0** 36.1 32.4 28.5 25.7* 31.2 65-80 30.4** 28.8** 19.6 21.1 21.8 19.1 Continued 2052 | Research and Practice | Peer Reviewed | Martinson One set of sensitivity analyses contained education terciles, categorizied them as follows: low education (0---12 years of education in United States, 0---11 years of education in England), medium education (13---15 years of education in United States, 12---13 years of education in England), and high education ( 16 years of education in United States, 14 years of education in England). Although the education categories were not directly comparable, these categories are the same as those used in previous literature comparing the 2 countries.2 Other covariates included race/ethnicity (White, Black, Hispanic, and other for the United States; White, Black, Asian, and other for England), smoking, frequent alcohol drinking, and health insurance. I included these factors as they may inuence health differently in the 2 countries and have been used in previous research on the gradient among older adults.1 All are described in Table 1. In this study I used Stata statistical software version 11.0 SE (StataCorp LP, College Station, TX) to conduct all analyses. I used the SVY commands to adjust for complex sampling design in both studies and to produce robust standard errors and weighted all analyses to produce nationally representative results. I calculated weighted percentages for each health condition, separately for males and females, in each age group by income tercile. I used modied Poisson models to estimate risk ratios demonstrating the risk of morbidity faced by low- and medium-income individuals relative to the high-income individuals within each country. These models are well suited to the estimation of risk ratios, particularly in studies using complex sampling design, where generalized linear binomial regression models have documented convergence problems.34 RESULTS The US and England samples are described by income tercile in Table 1. There was variation in the representation of age in the 3 income groups. For this reason, all subsequent analyses in this article were either adjusted by age or stratied by age group. In both countries, there was a slight gender difference by American Journal of Public Health | November 2012, Vol 102, No. 11 RESEARCH AND PRACTICE TABLE 3Continued High C-reactive protein 18-34 22.4 23.8 19.5 20.5 16.8 15.0 35-49 34.3** 26.2 24.0 25.8** 19.0 18.2 50-64 44.1** 33.0 30.2 41.8** 27.7 27.7 65-80 46.7** 37.2 35.9 39.8** 45.0** 29.9 Hypertension 12-19 2.3 1.2 1.0 3.0 2.7 2.4 20-34 9.5 8.5 9.8 10.0 11.5 13.9 35-49 24.6* 24.5* 18.5 26.2* 21.2 21.2 50-64 49.9** 42.3 39.6 48.6 42.9 44.7 65-80 66.5* 64.2 59.5 63.7 66.0 61.8 Asthma ever diagnosed Birth-3 13.4* 10.8 8.4 9.2 4.4 3.7 4-11 12-19 17.8 17.7 20.7 18.1 17.8 17.1 13.2* 11.4 10.1 6.7 7.6 11.6 20-34 12.0 10.9** 16.3 6.0 6.1 5.9 35-49 10.9 10.8 9.3 6.4 5.0 4.2 50-64 12.0 9.5 10.8 6.7 4.2 4.9 65-80 7.9 8.2 7.8 6.2 4.8 6.3 20-34 0.5 0.7 0.0 0.0 0.2 0.0 35-49 50-64 3.4 13.0** 2.0 8.2 2.1 5.8 1.2** 13.7** 0.9 6.9 0.4 5.2 65-80 26.8** 19.6 16.1 27.2* 26.8* 17.4 20-34 0.4 0.5 0.4 0.4 0.0 0.0 35-49 1.0 1.1 0.3 0.7 0.0 0.3 50-64 4.7** 1.2 1.1 3.4** 0.7 1.0 9.4 6.1 10.3** 11.1** 5.4 Heart Attack Stroke 65-80 10.6* Note. HDL = high-density lipoprotein cholesterol. *P < .05; **P < .01 (for prevalence among low- and middle-income vs high-income individuals). income. Females were more likely to be low income in both countries, whereas males were more likely to be high income. For this reason, I stratied the analysis by gender. There was a marked pattern in the race distribution by income. Non-White race and ethnic groups were more likely than Whites to be in the lowest income tercile in both countries. Additionally, as expected, those with the lowest education were much more likely to be low income, whereas those with higher levels of education were much more likely to be high income. Health behaviors were also variable by income. Smoking (or being exposed to household smoke for those younger than 18 years) was most prevalent among low-income individuals and least prevalent among high-income people in both countries. There was also a clear gradient for drinking alcohol on 5 or more days per week for those aged 20 years and older, although this gradient went in the opposite direction, with high-income respondents drinking more. Finally, health insurance was distributed highly unevenly by income in the United States. Almost one third of low-income individuals were uninsured, whereas less than 6% of high-income individuals did not have health insurance. The entire English sample was provided health insurance through the National Health Service (NHS). November 2012, Vol 102, No. 11 | American Journal of Public Health The unadjusted prevalence percentages are presented for all relevant health conditions for females (Table 2) and males (Table 3). These tables show a signicant income gradient in health in both the United States and England and no systematic variation in this gradient by age in either country. Because there were no clear patterns in the income gradient by age, Tables 4 and 5 present results for all relevant ages, adjusted by age. This also facilitated the interpretation of the results. The risk ratios in Table 4 (females) and Table 5 (males) illustrate the relative levels of risk for health risk factors or disease among low- or medium-income versus high-income individuals within the same country. Models 1 and 4 represent the risk ratios adjusted only by age in the United States and England. The results demonstrate that the gradient in the 2 countries was quite similar. Despite the different health care systems and overall population health within each country, health inequality was pervasive in both the United States and England. These results were not sensitive to age group specication (available upon request). For females (Table 4: models 1 and 4), the gradient for heart attack and stroke appeared steeper in the United States than in England. For diabetes, low HDL cholesterol, high C-reactive protein, and asthma, the difference between those with high and low incomes was greater in England than in the United States. The gradients appeared identical for obesity, high cholesterol ratio, and hypertension. The difference in the gradient between the United States and England was not statistically significant for any outcome. The risk ratios demonstrated a steeper gradient for US males than for English males for obesity, diabetes, high cholesterol ratio, and hypertension (Table 5: models 1 and 4). There was no difference in the heart attack and stroke gradients, and the gradient was steeper in England for low HDL cholesterol, high C-reactive protein, and asthma. The only statistically signicant differences in the gradient in health between US and English males were for obesity among middle-income individuals and hypertension among low-income individuals. In these 2 cases, the gradient was steeper in the United States than in England. Importantly, the income gradient appeared to be more Martinson | Peer Reviewed | Research and Practice | 2053 RESEARCH AND PRACTICE TABLE 4Risk Ratios for Low- and Middle-Income Females Compared With High-Income Females: US National Health and Nutrition Examination Survey (1999-2006) and Health Survey for England (2003-2006) United States Health Outcome and Income Level Model 1 Model 2 England Model 3 Model 4 Model 5 Model 6 Obesity Low 1.51** 1.38** 1.41** 1.49** 1.46** 1.43** Middle Diabetes 1.30** 1.27** 1.27** 1.25** 1.24** 1.26** income gradients in health that could not be explained by the factors available here. Age-specic analyses yielded similar ndings, with little variation in the potential explanatory factors by age group. Additionally, I conducted sensitivity analyses using education rather than income as the socioeconomic measure of interest and restricting the sample to Whites only. The results were not sensitive to these alternate specications (available upon request). Low 2.50** 1.94** 1.75** 2.70** 2.30** 1.95** Middle 1.71** 1.57** 1.44** 1.87** 1.83** 1.66** DISCUSSION Inequality in health by income was quite similar within both the United States and England, despite the healthier population in England. Americans and the English were affected by the income gradient in health at all ages, from childhood through to later adulthood. Factors such as race/ethnicity, smoking, frequent alcohol consumption, BMI, and health insurance could not explain the magnitude of the income gradient in either country. Low HDL Low 2.24** 2.31** 1.67** 2.60** 2.41** 2.32** Middle 1.59** 1.60** 1.31 1.36 1.35 1.41 1.72** 1.39** 1.86** 1.43** 1.33** 1.17 1.70** 1.11 1.72** 1.12 1.34** 0.99 High cholesterol ratio Low Middle Hypertension Low 1.21** 1.15** 1.14** 1.23** 1.21** 1.14** Middle 1.13* 1.11 1.08 1.06 1.06 1.01 High C-reactive protein Low 1.24** 1.18** 1.07 1.32** 1.32** 1.17** Middle 1.13* 1.11* 1.02 1.14** 1.14** 1.08 1.21* 1.26** 1.30** 1.58** 1.56** 1.43** 1.04 1.06 1.07 1.25 1.25 1.15 Asthma Low Middle Heart attack Low 2.62** 2.67** 2.35** 2.06** 2.04** 1.88** Middle 1.52* 1.53* 1.42 2.12** 2.10** 1.93** Stroke Low 3.03** 3.09** 2.41** 1.91** 1.99** 1.78** Middle 2.29** 2.30** 2.04** 1.23 1.22 1.15 Note. HDL = high-density lipoprotein cholesterol. Columns 1 and 4 control for age only; models 2 and 5 control for age and race/ethnicity; columns 3 and 6 control for age, race/ethnicity, health behaviors, body mass index, and health insurance (US only). *P < .05; **P < .01 (risk ratio for low- and middle-income vs high-income individuals). pronounced for females than for males in both countries, particularly for obesity, diabetes, low HDL cholesterol, high total cholesterol ratio, hypertension, asthma, and heart attack or angina. In Tables 4 and 5, models 2 and 5 added controls for race/ethnicity and models 3 and 6 added additional controls for the following health behaviors: smoking (adults) or being exposed to household smoking (aged 18 years), frequent drinking ( 20 years), BMI ( 4 years), and health insurance (for the United States only; all ages). In the models controlling for race/ethnicity, it was clear that racial and ethnic differences explained little of the within-country income gradient. Diabetes was the only outcome for which there was an attenuation of the risk ratios in both countries for both genders when race/ethnicity was added to the models, although there was no change in statistical signicance. Likewise, controlling for health behaviors and insurance had little consistent impact on the income gradient (models 3 and 6), although high C-reactive protein for US females and low HDL cholesterol for English males were exceptions. Overall, both countries had large, signicant 2054 | Research and Practice | Peer Reviewed | Martinson Potential Explanations for Health Inequalities The comparability of the gradient in the United States and England2 countries with very different health care and social protection systemsis surprising and not easily explained. However, several potential mechanisms can be ruled out. It is clear that differences in race/ ethnicity did not account for the income gradient in the 2 countries. Additionally, the analysis excluded both frequent alcohol consumption and smoking behaviors as possible explanations for the gradient among the English and Americans. Obesity and overweight did not appear to explain income inequality in health in either country. Perhaps most crucially, it is difcult to ascertain the extent to which health care differences in the 2 countries also inuence the level of income inequalities within the United States and England. Despite the universal health care provided by the NHS in England, its income gradient in health appears similar to that of the United States, where health care access is very uneven.35 Additionally, since the gradient is equally steep for both the biological and self-reported measures based on a doctor's diagnosis, country-level differences American Journal of Public Health | November 2012, Vol 102, No. 11 RESEARCH AND PRACTICE Finally, there is no systematic variation in the magnitude of the gradient in health by age in either country. Depending on the health outcome of interest, the gradient may be steepest for children, young and middle-age adults, or the oldest adults. The gradient varies by gender and country as well. By young adulthood, the income gradient in health is well established in both of these countries. This suggests that income-based health inequalities affect both the young and old in societies and are probably not the result of stress accumulation and a compounding of disadvantage and health shocks throughout the life course. It would be fruitful for future research to examine gradients across the life span using longitudinal data to understand both the development of health disparities and the relationships between trends in social policy and health inequalities. TABLE 5Risk Ratios for Low- and Middle-Income Males Compared With High-Income Males: US National Health and Nutrition Examination Survey, 1999-2006, and Health Survey, 2003-2006 United States Health Outcome and Income Level England 1 2 3 Low 1.16** 1.15** 1.19** Middle Diabetes 1.19** 1.19** 1.19** Low 2.01** 1.74** Middle 1.25 1.18 Low 1.21** Middle 1.14* 4 5 6 1.13 1.16* 1.20* 0.97 0.97 0.97 1.69** 1.64** 1.45** 1.32* 1.10 1.32* 1.32* 1.17 1.27** 1.26** 1.29** 1.24** 1.12 1.16** 1.12* 1.16 1.15 1.09 1.17** 1.1 1.21** 1.12* 1.20** 1.08 1.11 1.01 1.08 1.01 1.02 0.97 Low 1.20** 1.19** 1.26* 1.02 1.01 1.03 Middle 1.09 1.09 1.08 0.98 0.98 0.97 Low 1.35** 1.32** 1.25** 1.41** 1.39** 1.27** Middle 1.11 1.10 1.04 1.14* 1.14* 1.07 1.00 1.05 1.13 1.31* 1.32* 1.27 0.98 0.99 0.94 0.96 0.96 0.98 Low 1.81** 2.03** 1.94** 1.81** 1.92** 1.90** Middle 1.23 1.27 1.21 1.46* 1.53** 1.51** Low 2.22** 2.07** 2.05** 2.19** 2.47** 2.34** Middle 1.52* 1.49* 1.40 1.41 1.40 1.34 Obesity Low HDL High cholesterol ratio Low Middle Hypertension Limitations High C-reactive protein Asthma Low Middle Heart attack Stroke Note. HDL = high-density lipoprotein cholesterol. Columns 1 and 4 control for age only; columns 2 and 5 control for age and race/ethnicity; columns 3 and 6 control for age, race/ethnicity, health behaviors, body mass index, and health insurance (US only). *P < .05; **P < .01 (risk ratio for low- and middle-income vs high-income individuals). in access to health care also cannot be confounding the results. However, one cannot conclusively eliminate the possibility that differences in the type of health care provided in each country may be inuencing individuals differently at each point across the income distribution. Finally, it may be that the NHS improves health for the entire English population, but that overall income inequalities in both countries translate directly into health inequalities. If this is the case, the income gradient will be very difcult to reduce until overall income inequality is reduced. Previous studies have not conducted comparisons separately by gender. The nding that the income gradient is in fact steeper for women in both countries is intriguing and suggests that future comparative studies should investigate gender differences. Recent work by Martinson et al. also found that the health differential between individuals in the United States and England is greater for females. 3 Health inequalities appear to affect women more than men, and this is an area ripe for further research. November 2012, Vol 102, No. 11 | American Journal of Public Health There are some limitations to this study that suggest future directions for comparative studies in health disparities between the United States and England. The rst limitation is the questionable comparability of self-reported health in different countries. In this study, however, the high degree of comparability of the biological measures is a strength in both sets of data, as these measures are less susceptible to measurement error than selfreported survey measures.1,3,19 The results are similar among the self-reported and biological measures. Second, large health surveys such as the HSE and NHANES, while providing very highquality comparable health outcomes, have limited sociodemographic and behavioral measures to examine as potential mechanisms. Additionally, this study was unable to disentangle age and cohort effects in examining the income gradient by age group. To this end, one cannot denitively conclude that the relationship between income and health does not strengthen throughout the life coursemore evidence that future studies using longitudinal data are sorely needed to examine the health trajectories in the United States compared with other countries, such as England. Third, although using income terciles allows for the examination of relative health inequalities in each country, the difference between high- and low-income individuals is Martinson | Peer Reviewed | Research and Practice | 2055 RESEARCH AND PRACTICE greater in the United States than in England as reected by the Gini coefcient.5 Finally, to further elucidate the remarkably similar income inequalities in health found in these 2 countries, studies such as this should be extended to other countries. Although England has less income inequality than the United States, its citizens still experience a higher level of inequality than many other European countries. 3. Martinson ML, Teitler JO, Reichman NE. Health across the life span in the United States and England. Am J Epidemiol. 2011;173(8):858---865. Conclusions 7. Currie A, Shields MA, Price SW. The child health/ family income gradient: evidence from England. J Health Econ. 2007;26(2):213---232. The similarity of the income gradient in health in the United States and England is notable at all ages for a number of conditions. Although the English enjoy better overall health than Americans, both countries still grapple with large health inequalities. This comparison of the income gradient in health suggests that the policy discussion on reducing health disparities requires attention to broader social conditions, not simply health insurance and health care. Understanding disparities in an international context, especially by extending this comparison with other countries, will help shed light on the pervasiveness of health inequalities by income. j About the Author At the time of the study, Melissa L. Martinson was with the Ofce of Population Research, Princeton University, Princeton, NJ. Correspondence should be sent to Melissa L. Martinson, University of Washington, School of Social Work, Seattle, WA 98195 (e-mail: melmart@uw.edu). Reprints can be ordered at http://www.ajph.org by clicking the \"Reprints\" link. This article was accepted June 3, 2012. Acknowledgments This research was funded in part by the National Institutes of Health (grant T32HD001763). I thank Nancy Reichman and Julien Teitler for their helpful comments. Human Participant Protection Because this study employed only the analysis of de-identied secondary data, no protocol approval was needed. References 1. Banks J, Marmot M, Oldeld Z, Smith JP. Disease and disadvantage in the United States and in England. JAMA. 2006;295(17):2037---2045. 2. Avendano M, Mackenbach JP, Glymour MM, Banks J. Health disadvantage in US adults aged 50 to 74 years: a comparison of the health of rich and poor Americans with that of Europeans. Am J Public Health. 2009;99(3):540-- 548. 4. Esping-Andersen G. The Three Worlds of Welfare Capitalism. Cambridge, UK: Polity Press; 1990. 5. Smeeding TM. Public policy, economic inequality, and poverty: the United States in comparative perspective. Soc Sci Q. 2005;86:955---983. 6. Organisation for Economic Co-Operation and Development. OECD health data 2010: statistics and indicators. 2009. Available at: http://www.oecd.org/ document/30/0, 3746, en_2649_37407_12968734_ 1_1_1_37407,00.html. Accessed May 15, 2011. 8. Muennig PA, Glied SA. What changes in survival rates tell us about US health care. Health Aff (Millwood). 2010;29(11):2105---2113. 9. Mainous AG, Diaz VA, Saxena S, et al. Diabetes management in the USA and England: comparative analysis of national surveys. J R Soc Med. 2006;99(9):463---469. 10. Waldfogel J. Britain's War on Poverty. New York, NY: Russell Sage Foundation; 2010. 11. Case A, Lee D, Paxson C. The income gradient in children's health: a comment on Currie, Shields and Wheatley Price. J Health Econ. 2008;27(3):801---807. 12. Case A, Lubotsky D, Paxson C. Economic status and health in childhood: the origins of the gradient. Am Econ Rev. 2002;92(5):1308---1334. 13. Sacker A, Wiggins RD, Bartley M, McDonough P. Self-rated health trajectories in the United States and the United Kingdom: a comparative study. Am J Public Health. 2007;97(5):812---818. 14. Currie J, Stabile M. Socioeconomic status and child health: why is the relationship stronger for older children? Am Econ Rev. 2003;93(5):1813---1823. 15. Condliffe S, Link CR. The relationship between economic status and child health: evidence from the United States. Am Econ Rev. 2008;98(4):1605---1618. 16. National Center for Health Statistics. NHANES analytic and reporting guidelines: September 2006 version. 2006. Available at: http://www.cdc.gov/nchs/data/ nhanes/nhanes_03_04/nhanes_analytic_guidelines_dec_ 2005.pdf. Accessed May 15, 2011. 17. Sproston K, Mindell J. Health Survey for England 2004: methodology and documentation. 2006. Available at: http://www.ic.nhs.uk/webles/publications/ healthsurvey2004ethnicfull/HealthSurveyforEngland Vol2_210406_PDF.pdf. Accessed May 15, 2011. 18. Sproston K, Primatesta P. Health Survey for England 2003: methodology and documentation. 2004. Available at: http://www.dh.gov.uk/prod_consum_dh/ groups/dh_digitalassets/@dh/@en/documents/ digitalasset/dh_4098912.pdf. Accessed May 15, 2011. 19. Craig R, Mindell J. Health Survey for England 2005: methodology and documentation. 2007. Available at: http://www.ic.nhs.uk/webles/publications/hseolder/ vol5mad.pdf. Accessed May 15, 2011. 20. Craig R, Mindell J. Health Survey for England 2006: methodology and documentation. 2008. Available at: http://www.ic.nhs.uk/webles/publications/HSE06/ HSE06_VOL3.pdf. Accessed May 15, 2011. 2056 | Research and Practice | Peer Reviewed | Martinson 21. World Health Organization Technical Report Series. Physical status: the use and interpretation of anthropometry. 1995. Available at: http://whqlibdoc.who.int/ trs/WHO_TRS_854.pdf. Accessed May 15, 2011. 22. Cole TJ, Bellizzi MC, Flegal KM, Dietz WH. Establishing a standard denition for child overweight and obesity worldwide: international survey. BMJ. 2000;320 (7244):1240---1243. 23. National Heart, Lung, and Blood Institute. Sixth Report of the Joint National Committee on the Prevention, Detection, Evaluation and Treatment of High Blood Pressure. 1997. NIH publication 98---04080. Available at: http://www.ncbi.nlm.nih.gov/books/NBK8632/pdf/ TOC.pdf. Accessed May 15, 2011. 24. Saudek CD, Herman WH, Sacks DB, Bergenstal RM, Edelman D, Davidson MB. A new look at screening and diagnosing diabetes mellitus. J Clin Endocrinol Metab. 2008;93(7):2447---2453. 25. National Cholesterol Education Program. Third Report of the National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III) Final Report. Circulation. 2002;106 (25):3143---3421. 26. Kinosian B, Glick H, Garland G. Cholesterol and coronary heart disease: predicting risks by levels and ratios. Ann Intern Med. 1994;121(9):641---647. 27. Ridker PM. C-reactive protein and the prediction of cardiovascular events among those at intermediate risk: moving an inammatory hypothesis toward consensus. J Am Coll Cardiol. 2007;49(21):2129---2138. 28. Licastro F, Candore G, Lio D, et al. Innate immunity and inammation in ageing: a key for understanding age-related diseases. Immun Ageing. 2005;2(1):8. 29. Pearson TA, Mensah GA, Alexander RW, et al. Markers of inammation and cardiovascular disease: application to clinical and public health practice: a statement for healthcare professionals from the Centers for Disease Control and Prevention and the American Heart Association. Circulation. 2003;107(3):499---511. 30. Organisation for Economic Co-Operation and Development. What are equivalence scales? 2008. Available at: http://www.oecd.org/dataoecd/61/52/35411111. pdf. Accessed May 15, 2011. 31. US Dept of Labor, Bureau of Labor Statistics. Consumer Price Index, all urban consumers (CPI-U), US city average. 2011. Available at: ftp://ftp.bls.gov/pub/ special.requests/cpi/cpiai.txt. Accessed May 15, 2011. 32. UK Ofce for National Statistics. Tables of CPI for UK. 2009. Available at: http://www.statistics.gov.uk/ StatBase/Product.asp?vlnk=15089. Accessed May 15, 2011. 33. Banks J, Marmot M, Oldeld Z, Smith JP. The SES health gradient on both sides of the Atlantic. NBER Working Paper 12674. November 2006. Available at: http://www.nber.org/papers/w12674. Accessed May 15, 2011. 34. Zou G. A modied poisson regression approach to prospective studies with binary data. Am J Epidemiol. 2004;159(7):702---706. 35. US Census Bureau. Income, poverty, and health insurance coverage in the United States: 2008. 2009. Available at: http://www.census.gov/prod/2009pubs/ p60-236.pdf. Accessed May 15, 2011. American Journal of Public Health | November 2012, Vol 102, No. 11 Copyright of American Journal of Public Health is the property of American Public Health Association and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use