The recession provides a natural experiment for understanding more about economic influences on health and health disparities. There is an extensive literature that discusses the impact of economic recession on health, and much of this literature suggests that the effects may be harmful. For example, a macroeconomic study by Brenner (2005) found that, although gross domestic product per capita is occasionally associated with increased mortality rates over the very short term, it is strongly inversely related to mortality rates over the medium to long term. Strully (2009), using data from the PSID, found that respondents who lost jobs but were then reemployed were more likely to develop health conditions. Gallo et al. (2006) found that displaced workers over age 50 were at higher risk for subsequent myocardial infarction and stroke.

The literature on this subject also contains some contrary views. Laporte (2004), for example, found that unemployment was associated with a decline in mortality in the short and long run and suggested that many studies failed to account for certain time-series properties of macro-level data. Tapia Granados (2005) found that economic expansions are associated with higher rates of mortality for all groups and causes of death studied (except for suicide). In Germany, Neumayer (2004) found that recessions lowered mortality rates for some, but not all, causes of death.

In order to determine the effects of recession on health, such factors as time lags, specific causes of death, lifestyle, environmental risks, and health care utilization need to be considered. Longer term trends in the economy, such as globalization, economic restructuring, and income inequality, can also interact in complex ways with economic recession and growth. Some indicators of economic stress, moreover, may be more relevant than others.

Dawn Alley provided an overview of conflicting perspectives on the effects of recession on health. According to Alley, some of the research on business cycles and mortality indicates that the population is healthier when unemployment is higher. One study, for example, found that a variety of physical health measures improve with rising unemployment (Ruhm, 2000): a 1 percent increase in state unemployment was associated with a 0.5 percent decrease in all-cause mortality, a 0.9 percent decrease in hospitalization, a 0.4 percent decrease in doctor visits, a 0.5 percent decrease in heart disease (self-reported), and a 1.9 percent decrease in cancer (self-reported). The fact that the negative association between unemployment and mortality is stronger among older rather than working-age adults tends to rule out causal explanations that are directly related to the labor force (Miller et al., 2009). Research on the individual-level effects of unemployment, financial strain, and material deficits, however, tends to suggest that people are healthier when incomes are higher (see, e.g., Sullivan and von Wachter, 2007). Alley said that not enough is known about the specific mechanisms through which the economic environment may affect health. As an example of such a pathway, one participant referred to a study that found that babies conceived in times of high unemployment tended to be healthier because of improvements in health behavior during recessions, and also because of changes in the types of mothers (in terms of educational background) who conceived during recessions (Dehejia and Lleras-Muney, 2004).

In response to a question about time lags, Alley noted that most models begin by looking at the contemporaneous effect of changes in the economic environment on health outcomes. Although there has been considerable debate over what constitutes an appropriate lag, there is not strong evidence of lagged effects of recession or of results changing with the use of different lags. She also suggested that further research in this area would be appropriate.

Alley proposed six sets of questions raised by the recession. First, how much of what is reported during economic downturns as better health is due to reporting error? How would results differ if outcomes included measured health or incident health conditions? (Alley indicated that these questions did not pertain to declines in mortality, which could not readily be ascribed to measurement error.) Epidemiologists have suggested that apparent declines in chronic conditions could be due to people not going to their doctors and not knowing—and not reporting—that they have a particular condition. (One participant noted that in many developing countries, people with higher socioeconomic status had worse self-assessed health—because they were able to learn more about their health from their doctor—even though they had higher life expectancy.) Possible data sources to explore these questions include the Medicare Current Beneficiary Survey and the National Health and Nutrition Examination Survey.

Second, if recessions are in fact associated with better health among older people (see Miller et al., 2009), why is that so? Previous explanations have emphasized unemployment-related factors (such as cutting back on hours and having more time with family), but these are unlikely to account for effects at the oldest ages. Alley also indicated that although researchers have examined how the mortality effects of recession vary across different age groups, more work needs to be done on how nonmortality effects might vary by age.

Several participants noted that the literature finding a positive relationship between recession and health outcomes may be reminiscent of the “ecological fallacy” (see Robinson, 1950), in which the causes of individual outcomes are attributed to characteristics of groups correlated with geographic areas; the ecological fallacy was especially common in earlier periods when people worked with data from census tracts rather than surveys. The ecological fallacy is an issue when aggregate-level data are being used on both the predictor and outcome ends, and it becomes less of a problem once individual characteristics and the specific risks faced by individuals are incorporated into the analysis.

Participants also suggested that rather than looking at general health measures (and not finding any effects), it might be useful to measure the biological effects of stressful events, looking at such indicators as cortisol levels, blood pressure, and inflammation markers. Men ages 55 to 65 may have become more susceptible to heart attacks because of the recession. Research indicates that self-reported financial strain is related to changes in cortisol levels over time (Steptoe, Brydon, and Kunz-Ebrecht, 2005), and it could be useful for studies like the HRS to collect data in ways that allow for such relationships to be explored, especially given technological advances that facilitate the collection of biomarkers. Bringing together different data sources could also be productive. The HRS, for example, does not have detailed diet or physical activity data, whereas some other data sets do have such measures.

Third, do recessions increase health inequality? It is possible that health improvements at the population level mask health declines in population subgroups, leading to increased health disparities. How can this be best measured, and which population subgroups (e.g., those with limited financial assets, existing health conditions, or limited social resources) might be most at risk of health declines? The possibility was mentioned that substantial negative effects on “losers” may be outweighed by smaller positive effects on more numerous “stayers” or even “winners” in a crisis. Alley also suggested that rather than thinking about the protective effects of recessions, it may be useful to frame some of these issues in terms of the potential negative consequences of rapid economic expansion. For example, older people may become more socially isolated or less able to depend on their children for social support.

One participant indicated that it might be helpful to look at the truly “miserable” component of the population that has low levels of support and is initially at risk for some other reason, as much of the recession- related dysphoria may be temporary and have little long-term effect. Another wondered whether the direction in which the unemployment rate is moving might not be more consequential for well-being than the absolute level of unemployment. The probability of experiencing a spell of unemployment, the expected length of an unemployment spell, the extent of underemployment, and abandoning the job search altogether may also have different and distinct effects. It was further suggested that it might be useful to study the factors, both negative and positive, that may affect those who are still employed. Such people may be subject to higher levels of stress because of greater demands from their employers, longer working hours, and heightened economic insecurity.

Fourth, what are the long-term implications of the negative mental health effects that may be caused by economic downturns? Ruhm (2000) found that a variety of physical health measures improved with unemployment, but that increases in state unemployment were associated with an increase in suicide and mental health disorders. Similarly, pilot data from an Internet consumer panel in Arizona, California, Florida, and Nevada, in summer 2008 suggest that serious psychological distress is more prevalent among those who experience greater strain in their housing situation (Pollack et al., 2010). Contradictory findings for physical and mental health need to be further investigated.

Fifth, not enough is known about the unique effects of different exposures that increase in prevalence during recessions, such as unemployment, reductions in income, reductions in wealth, food insecurity, reduced spending on health care, and unaffordable housing, including mortgage default. Which types of disadvantage are particularly harmful to health? (See Alley et al., 2009, for a discussion of the later life health consequences of material disadvantage in health care, food, and housing.) One promising avenue of research is to find ways of using variation in such factors to determine their effect (distinct from income) on health. For example, state housing laws may be related to foreclosure rates in ways that are not related to income and could serve as good instruments for looking at the potential relationship between foreclosures and health. Other examples may exist for such factors as food insecurity and health care access.

Sixth, how do the health effects of individual-level disadvantage interact with community-level economic factors? Does the health effect of becoming unemployed, for example, depend on whether there is a high level of unemployment in the community? On one hand, if an experience is normative, there may be less stigma and easier access to supportive resources. On the other hand, a recession might have negative effects on the community environment (e.g., with respect to crime) and reduce the local resource base for services.

Looking more specifically at the current recession, David Weir reported that the Gallup Well-Being Index (WBI) registered significant declines in fall 2008 but recovered substantially by May 2009. Changes in the overall WBI during this period were driven by changes in the “life satisfaction” component—that is, “Cantril’s ladder.”¹ The health behavior component (but not the physical or emotional health components) also tracked the overall WBI relatively closely. The movement in the overall WBI during this period, though noticeable, was small relative to the gradient between the top and bottom income quartiles.

Weir went on to explain that core HRS data from 2008, 2010, and beyond will help identify the magnitude of the effects of the downturn on health; international sister studies will permit cross-national comparisons. The HRS also conducted a postcrash Internet survey in May and June 2009 and a mail survey on well-being in November and December 2009, providing further data on the effects of the crisis.

The 2009 HRS Internet survey included information on consumer sentiment; positive and negative affect; satisfaction and control across a variety of domains; assets, ownership, losses, and rebalancing of investment portfolios; mortgage issues for self and family; consumption; employment and retirement expectations; and health behaviors. Comparing the same people across 2008 and 2009, Weir reported several findings: (1) the average reported chances of working full-time past both age 62 and age 65 increased; (2) the percentage of people not satisfied with their financial situation increased, and the percentage very satisfied with their financial situation decreased; (3) the percentage with no depressive symptoms decreased, and the percentage with four or more depressive symptoms increased; and (4) the percentages experiencing both frequent pain and severe pain increased. Although the smoking rate went down, drinking (including binge drinking) and church attendance did not change much.

The 2009 HRS well-being mail survey primarily had to do with hedonic well-being,² although it also included several global well-being measures, such as Gallup’s Cantril’s Ladder, general life satisfaction, and domain satisfaction. All the measures of well-being are characterized by income and wealth gradients. Well-being also tends to increase until people are in their late 60s, after which it plateaus and then, around age 75, begins to decline. Weir suggested that the fact that these different measures of well-being track each other so well raises interesting questions about how people answer the survey questions and what the various measures reflect.

A number of questions are common to the HRS core (2008), Internet (May and June 2009), and mail (November and December 2009) surveys. According to Weir, the mean difficulty of paying monthly bills increased steadily through all three surveys and over the course of the recession. Similarly, the percentage of people having difficulty paying bills increased between the core and Internet surveys and, to a lesser extent, between the Internet and mail surveys. Conversely, life satisfaction fell considerably between the core and Internet surveys and, to a lesser extent, between the Internet and mail surveys.