Elsevier

Social Science & Medicine

Volume 56, Issue 4, February 2003, Pages 769-784
Social Science & Medicine

The measurement of SES in health research: current practice and steps toward a new approach

https://doi.org/10.1016/S0277-9536(02)00073-4Get rights and content

Abstract

The resurgence of social epidemiology has yet to induce corresponding research into basic measurement issues. This paper aims to motivate investigators to refocus attention on the measurement of socioeconomic status (SES). With a primarily American focus, we document striking paucity of basic research in SES, review the history of SES measurement, highlight the central limitations of current measurement approaches, sketch a new theoretical perspective, present new pilot results, and outline areas for future research. We argue (1) that lack of conceptual clarity and the bypassing of standard psychometric techniques have retarded SES measurement. And (2) social epidemiologists should revisit the measurement of SES and consider whether a richer, psychometrically induced, approach would be more useful. Our pilot study suggests a great deal of uniformity between existing SES measures and that a new approach may be worthy of pursuit.

Introduction

Interest in social epidemiology has been increasing, as evidenced by recent editorials in scholarly publications (e.g., McKinlay & Marceau, 2000) and a US National Institutes of Health (NIH) conference on social science and health (June 27–28, 2000, Bethesda, MD). Several papers and chapters (Berkman & Kawachi, 2000; Krieger, 1994; Lantz et al., 1998; Macintyre, 1993) lend further support. More and more health researchers believe that a narrow focus on individuals outside of historical, social and biophysical contexts limits the understanding of disease etiology, health, and intervention modes.

Much of the attention to social factors in disease and health is given to socioeconomic status (SES), a central feature of the social structure of all complex societies. Accordingly, the number of studies of how disease and health relate to SES is growing. At least five reasons may explain this. First, science, at least ideally, is cumulative and there is a great deal of precedent. The strong relationship between SES and health has been documented for centuries, dating back to ancient Greece, Egypt, and China (Krieger, Willains, & Moss, 1997; Liberatos, Link, & Kelsey, 1988; Lynch, Kaplan, Cohen, Tuomilehto, & Salonen, 1996). Recent studies have consistently shown that SES, which endeavors to “encapsulate complex information about a person's life” (Blane, 1995), continues to be linked to disability and disease (House et al. (1992), House et al. (1994), 1994; Williams & Collins, 1995).

Second, SES is important to agencies interested in understanding and explaining the public's health. NIH's interest in the topic as a subject of basic research has been strong (see PA-98-098 and ES-00-004). There should be no question that funding structures influence the research.

Third, in light of the declining impact of acute infections, SES is relevant to social policy concerning public health. A better understanding of the relationship between SES and disease etiology may reveal important new points for medical intervention and epidemiological screening (Taylor, Repetti, & Seeman, 1997), as well as public policy (House et al., 1994). Unlike some correlates of disease, SES arguably can be changed by social policy, and this is arguably good health policy (Kaplan & Lynch, 2001). Advances in the understanding of genetic predispositions will enhance the importance of such policy.

Fourth, the socioeconomic structures in the US, and elsewhere, are rapidly changing (Chevan & Stokes, 2000). Economic inequality in the US has increased, shifting a greater proportion of income and wealth to the upper SES and a corresponding relative impoverishment of those on the lower SES levels especially the concentrated poverty among racial minorities (Massey & Denton, 1993; Wilson, 1987). The composition of social strata is also changing with increasing numbers of female and minority executives, newly minted high-tech millionaires, fewer manufacturing jobs, a decline in labor union membership, substantial numbers of temporary or part-time employees, many partial retirees, 401(k) investments, multi-organizational careers, and increases in social mobility over the life-course (DiPrete & Grusky, 1990).

Finally, SES may be important to social epidemiologists because absence of good SES data, racial/ethnic disparities in health may be construed as signs of genetic differences or behavioral choices rather than powerful clues about how forms of racial discrimination and structural constraints, past and present, harm health (Krieger et al., 1997; Williams, 1996; and the November 2000 volume of the American Journal of Public Health 90(11)).

Despite the growing interest, Kaplan and Lynch (1997), among others, remind us of how little we know about the ways in which SES influences health. Although SES is not itself a causal factor, understanding the linkages between SES and health can provide clues to the actual mechanisms involved. Only by mapping such linkages will we learn how best to construct remedies.

It is anomalous that relatively little attention has been paid to either the conceptualization or measurement of SES, especially in America. Although the term “socioeconomic” was apparently coined by the American sociologist Lester Ward in 1883 (Jones & McMillan, 2001), there is still no consensus on a nominal definition of SES nor does a widely accepted SES measurement tool exist (Campbell, 1983; Rose, Pevalin, & Elias, 2001). Conceptualizing and measuring SES is among the more difficult and controversial subjects in social research. Prominent scholars have debated the theory, operationalization, and usefulness of SES constructs for about 125 years. Campbell and Parker (Campbell, 1983) conclude that the debate over SES will end when social research ends. Krieger et al. (1997) even suggest that we eliminate the notion altogether, and instead focus on manifest variables. We tend to disagree, and think that SES is a conceptually useful proxy for describing access to resources and constructing remedies.

The importance of SES measurement is, ironically, well stated by Krieger et al. (1997) who write “…developing consistent and broadly comparable measures of [SES] that can be incorporated into a wide variety of federally and privately sponsored data sets is essential.” This paper pursues those objectives. We (1) assess the relative attention paid to SES measurement, (2) review briefly the history of SES measurement in social science, (3) identify the central limitations of current measurement approaches, (4) sketch a new theoretical perspective, (5) present pilot results from the new perspective, and (6) draw some conclusions and outline areas for future research. Even if our proposed approach does not meet with favor, we will be satisfied if we motivate more basic research into the conceptualization and measurement of constructs fundamental to social epidemiology.

No single book, much less an article, could summarize the vast amount of research on SES. This paper's focus on American efforts is sufficiently daunting. We attempt to provide references as appropriate.

Section snippets

SES and Health

The use of SES as an explanatory variable in health research typically involves measuring the extent to which SES is related to health and/or using SES as a control in looking at other correlates of health. While these uses have increased dramatically since 1960, over the same period research into how to measure SES itself has remained relatively rare and fraught with both theoretical and methodological problems.

The gap between “SES Measurement” and “SES and Health” studies is quite large.

Historical background

The primary reason why measuring SES has been so retarded is the lack of conceptual clarity about the essential nature of social stratification. Although the concepts of social structure, social class, and socioeconomic status, are central to the social sciences, theorists have not agreed on definitions. As a result, methodologists have rarely relied on explicit theory to operationalize the notions. While most agree that Marx's treatment of class-relations was discounted in the US,3

Steps toward a new approach

Criticism without an alternative is usually of little use. This section sketches some ideas for a new composite measure of SES. Although separate variables, such as education and income, are preferred when statistically controlling for SES in a regression model, we believe a single composite measure remains best for stratified analyses, graphical presentations, and explanations to lay audiences–tasks common in health-related research. With respect to our proposed approach: we rely on

Conclusions

Constructing a valid measure of SES may be one of the most controversial topics in social and behavioral science. With few exceptions, this methodological issue has been ignored by health researchers and social epidemiologists. Yet if we wish to know about the relationship between SES and health outcomes, and understand the mechanisms through which SES affects health, we must critically evaluate our measurement of SES. To do otherwise may yield spurious relationships and undermine the

Acknowledgements

This paper was supported by HL61573 from the US National Heart, Lung and Blood Institute. The comments and criticisms of several colleagues and friends improved this paper; deserving special note are Andre Araujo, Ichiro Kawachi, Kevin Smith, and members of the Social Epi Workgroup at the University of Minnesota. Scholarly comments from an anonymous reviewer were especially helpful. This paper was begun when the first author was with the New England Research Institutes, Watertown, MA. The usual

References (106)

  • L.F Berkman et al.

    Social epidemiology

    (2000)
  • D.B Bills et al.

    A scale to measure the socioeconomic status of occupations in Brazil

    Rural Sociology

    (1985)
  • Backlund, E., Sorlie, P. D., & Johnson, N. J. (1996). The shape of the relationship between income and mortality in the...
  • D Blane

    Social determinants of health—socioeconomic status, social class, and ethnicity

    American Journal of Public Health

    (1995)
  • P.M Blau et al.

    The American occupational structure

    (1967)
  • K.A Bollen

    Structural equations with latent variables

    (1989)
  • G.W Borhnstedt

    Use of the multiple indicators—multiple causes (MIMIC) model

    American Sociological Review

    (1977)
  • P Bourdieu

    The Forms of Capital

  • Bowles, S., & Gintis, H. (2000). Walrasian economics in retrospect. Quarterly Journal of Economics,...
  • R.T Campbell

    Substantive and statistical considerations in the interpretation of multiple measures of SES

    Social Forces

    (1983)
  • F.S Chapin

    Contemporary American institutions

    (1935)
  • A Chevan et al.

    Growth in family income inequality, 1970–1990Industrial restructuring and demographic change

    Demography

    (2000)
  • J.S Coleman

    Social capital in the creation of human capital

    American Journal of Sociology

    (1988)
  • J.S Coleman

    The foundations of social theory

    (1990)
  • K Davis et al.

    Some principles of stratification

    American Sociological Review

    (1944)
  • R.F DeVellis

    Scale developmentTheory and applications

    (1991)
  • T.A DiPrete et al.

    Structure and trend in the process of stratification for American men and women

    American Journal of Sociology

    (1990)
  • O.D Duncan

    A socioeconomic index for all occupations

  • A.M Edwards

    Socio-economic grouping of the gainful worker of the United States, 1930

    (1938)
  • D.R Entwistle et al.

    Some practical guidelines for measuring youth's race/ethnicity and socioeconomic status

    Child Development

    (1994)
  • G Evans

    On tests of validity and social classWhy prandy and blackburn are wrong

    Sociology

    (1998)
  • R Farley et al.

    The color line and the quality of life in America

    (1987)
  • D.L Featherman et al.

    Prestige or socioeconomic scales in the study of occupational achievement?

    Sociological Methods and Research

    (1976)
  • F.F Furstenberg et al.

    Social capital and successful development of among at-risk youth

    Journal of Marriage and the Family

    (1995)
  • J.H Goldthorpe

    The social grading of occupations

    (1974)
  • L.A Goodman

    On the measurement of social mobilityAn index of status persistence

    American Sociological Review

    (1969)
  • M Gordon

    The logic of socio-economic scales

    Sociometry

    (1952)
  • L.W Green

    Manual for soring socioeconomic status for research in health behavior

    Public Health Reporter

    (1970)
  • D.B Grusky et al.

    Social stratification

  • Haug, M. R. (1977). Measurement in social stratification. Annual. Review.of Sociology....
  • M.R Haug et al.

    The indiscriminate state of social class measurement

    Social Forces

    (1971)
  • R.M Hauser et al.

    Socioeconomic index of occupational mobilityA review, update and critique

  • R.W Hodge et al.

    Occupational prestige in the United States1925–1963

    American Journal of Sociology

    (1964)
  • A.B Hollingshead

    Commentary on the indiscriminate state of social class measurement

    Social Forces

    (1971)
  • Hollingshead, A. B. (1975). Four factor index of social status. Yale University, Department. of Sociology, Working...
  • A.B Hollingshead et al.

    Social class and mental illness; A community study

    (1958)
  • House, J. S., Kessler, R. C., Herzog, A. R., Mero, R. P., Kinney, A. M., Breslow, M. J. (1992). Social stratification,...
  • J.S House et al.

    The social stratification of aging and health

    Journal of Health and Social Behavior

    (1994)
  • I.G Jones et al.

    Social class analysis—an embarrassment to epidemiology

    Community Medicine

    (1984)
  • F.L Jones et al.

    Scoring occupational categories for social researchA review of current practice, with Australian examples

    Work, Employment and Society

    (2001)
  • Cited by (692)

    View all citing articles on Scopus
    View full text