Content area
Full Text
Does employment protect and foster health? The cross-sectional correlation suggests that it does. People who work for pay report better physical well-being than others, most of whom are unemployed, retired, or keeping house (Kessler, House, and Turner 1987; Moser, Fox, and Jones 1986). Is it reasonable to interpret this correlation as an effect of employment? Some researchers think so. Employment increases status, power, and economic independence, as well as noneconomic rewards such as social support and recognition from others (Bird and Fremont 1991; Bird and Ross 1993; Gove and Tudor 1973; Gove and Geerken 1977;
Repetti, Matthews, and Waldron 1989; Ross and Bird 1994). Those benefits may translate directly or indirectly into better health. On the other hand, social selection provides a reasonable explanation too. Perhaps good health improves the chances of getting or keeping a paid job--the "healthy worker" hypothesis. Some people without jobs cannot work because of a disability. Others seem less attractive to employers because of a disease or disability. Thus, the correlation could represent an effect of health, rather than an effect of employment.
In this study we explore the hypothesis that employment protects and fosters health. For analytic purposes we consider evidence for the causal hypothesis and for the selection alternative. This provides a useful framework. It does not mean that the different views are inconsistent. For example, jobs may foster and protect the health required for continued employment. As well, unemployment may provoke health problems that hinder subsequent employment.
Evidence that Employment Improves Health
Researchers who study social differences in health know well the evidence that suggests a positive effect of employment. The following points summarize the facts as they relate to a causal interpretation.
1. Substantial, Reproducible Association. The positive correlation between employment and health appears consistently across studies that use health measures ranging from self-perceived health (e.g., Ross and Bird 1994) to reported symptoms or impairments (e.g., Herold and Waldron 1985; Kessler et al. 1987; Linn, Sandifer, and Stein 1985) to rates of serious disease, hospitalization, or death (Passannante and Nathanson 1985; Romelsjo et al. 1992; Verbrugge 1983, 1989; Waldron 1991a). Generally, the size of the correlation and compares with that of sociodemographic factors such as sex, race, education, and age (for adults up to 65 years).