| Vol. 12 No. 3 | ISSN: 0834-1729 |
| Income Dynamics and Mortality
by Peggy McDonough
Department of Sociology, York University |
| Introduction |
| Income has been characterized as exerting "one of the most profound influences on mortality" (Wilkinson, 1990). In fact, numerous studies have documented an inverse relationship between income and mortality, but one of the shortcomings of much of this wor k is its reliance on cross-sectional measures of income. The use of such measures assumes, implicitly at least, that income does not change much over time. Among workers in the United States, however, there is evidence of an increasing volatility of earni ngs. Household income may exhibit considerable variability, even during periods of economic stability, with changes in income linked to health and family and employment-related events such as divorce, widowhood, marriage or unemployment. A surprising degr ee of volatility is evident even among the elderly whose incomes are often presumed to be stable (Duncan, 1988).
Observations of income fluctuation raise questions about the extent to which the relationship between income and mortality is adequately represented by cross-sectional measures of income. Moreover, little is known about the extent to which income change i nfluences health beyond the more well-documented effects of income level. |
| The Volatility of Household Income over the Life Course |
 Using data from the Panel Study of Income Dynamics (PSID)2, Duncan (1988) examined the incidence of adverse economic changes characterized by a decline in income of 5 0% or greater between 1969 and 1979 (Figure 1). Roughly one-third of the population is estimated to have experienced this decline at least once over the 11 years. Not surprisingly, losses were observed most frequently among those who retired during the ob servation period. However, defying the stereotype of economic stability, these data also show that more than one-quarter of the elderly experienced at least one large drop in their incomes after retirement. At younger ages, the risk of income loss was sub stantially higher for women. Income losses that occurred during the 'prime' working years were not very predictable and were frequently the result of involuntary, rather than voluntary, events.
 Patterns of poverty show similar volatility (Figure 2) and indicate its largely temporary nature. Depending on the life-cycle, between 20% and 35% of women in the PSID sample e xperienced poverty at least once during the 11-year period. Poverty experiences were most frequent among older women, although they were by no means uncommon for women in their prime years. Persistent poverty, defined as living in poverty for more than ha lf of the 11-year period, characterized fewer than one-ninth of any of the subgroups. Like transient poverty, women were more likely to experience long-term poverty.
In spite of the conventional wisdom that average incomes in the United States rise until close to retirement and fall subsequently, longitudinal data show that it is a mistake to treat this path of average incomes as the typical income course of individua ls as they age. Family incomes are quite volatile at nearly every point in the life cycle, and the risk of sharp and unexpected declines in living standards is still significant at virtually every stage of life. |
| Income Dynamics and Mortality |
| To what extent do patterns of income fluctuation affect mortality? Are those who experience persistent low incomes at greater risk of death than those whose low incomes are more transitory in nature? Does a sharp decline in income increase the likelihood of dying?
 We examined these questions using data from the PSID from interview years 1968 through 1991 for individuals aged 45 years and older. The analyses suggest that the stability of economic resources is a powerful predictor of mortality risk (Figure 3). A household income of less than $20,000 for 4 to 5 years raised the risk of death relative to that of the stable middle-income reference group, while high-income for 4 to 5 years low ered the odds of dying by 30%. In contrast, the effects of transient income states (low or high income lasting 1 to 3 years) did not differ from that associated with stable middle-incomes. Hence, while income level plays a role, it may be the persistence of various income levels over time that is more important in predicting longevity. Measuring income cross-sectionally would fail to reflect this distinction and could attenuate its effects on mortality risk.
Table 1 presents the effects of year-to-year income losses of 50% or more on the likelihood of mortality. Those who experienced at least one sharp income loss over five years were 30% more likely to die than those not exposed to a decline of this nature. This effect persisted when a number of family, health and labour market-related events that could affect both household income and mortality were added to analytic models. In addition, income loss had an impact on mortality risk that was independen t of that exerted by income level and underscores the importance of considering income volatility in analytic models.
| Table 1: Adjusted Odds Ratios of All-Cause Mortality, by Income Loss and Income Level, PSID, 1972 through 1991, U.S. |
| . |
OR |
95% |
| . |
(n=45, 165) |
C.I. |
| At Least One Income Loss of >50% (t-5--t-1) |
| No |
1.00 |
- |
| Yes |
1.32 |
1.08-1.61 |
| Household Income (1993$) |
| <10,000 |
1.68 |
1.19-2.35 |
| 10-20,000 |
1.54 |
1.17-2.03 |
| 20-30,000 |
1.30 |
0.96-1.75 |
| 30-45,000 |
1.06 |
0.78-1.45 |
| 45-65,000 |
0.97 |
0.71-1.31 |
| >65,000 |
1.00 |
- |
| - 2 Log Likelihood |
10205.59 |
. |
| Note: OR adjusted for age, sex, race, family size, marital status change, health status change and major reduction in work hours. |
|
| Conclusion |
| The observations that emerge in the context of dynamic measures of income suggest that researchers cannot assume that cross-sectional measures of income accurately reflect income experiences over time, or the income-mortality relationship. A significant m inority of the PSID sample experienced transient poverty and sharp income loss. Moreover, these dynamic measures of income exerted effects on longevity that went beyond those related to the more typically-available measures of income level. These results suggest that although we may not always have the luxury of working with high-quality longitudinal data, at the very least, we should acknowledge the biases that may emerge in the cross-sectional context. |
References
Duncan G. The volatility of family income over the life course. In: Bates P, Featherman D, Lerner R, eds. Life span development and behavior. Volume 9. Hillsdale, NJ: Lawrence Erlbaum Associates; 1988:31-58.
Wilkinson, R. Income distribution and mortality: a 'natural' experiment. Sociology of Health and Illness. 1990;12:391-412. |
Endnotes
1This discussion is based on joint work conducted by Peggy McDonough, Greg Duncan (Institute for Policy Research, Northwestern University), David Williams and James House (Institute for Social Research, University of Michigan) and is supported by the Robert Wood Johnson Foundation.
2The PSID, initiated in 1968, is an on-going longitudinal study of a representative sample of women, men and children living in the U.S. |
|
