This Article Abstract Full Text (PDF) Services Download to citation manager PubMed PubMed Citation

Negative Life Events and Age-Related Decline in Mastery: Are Older Adults More Vulnerable to the Control-Eroding Effect of Stress?

¹ Health Systems Research & Consulting Unit, Centre for Addiction & Mental Health, Departments of Psychiatry and Public Health Sciences, University of Toronto, Ontario, Canada.
² School of Public Health and Institute of Gerontology, The University of Michigan, Ann Arbor.

Address correspondence to John Cairney, Health Systems Research & Consulting Unit, Centre for Addiction & Mental Health, Toronto, Ontario, Canada, M5S 2S1. E-mail: john_cairney{at}camh.net

	Abstract

TOP Abstract Methods Results Discussion References

 
Objectives. The purpose of this study was to see if exposure to life events influences age-related decline in control.

Methods. The data came from a large, nationally representative sample of Canadians aged 18 and older (n = 17, 291). We examined the principal research question by testing for an interaction between age, life events, and mastery using linear regression, both cross-sectionally and over time.

Results. Similar to previous work, there was a nonlinear association between age and mastery. The data suggested that exposure to life events was associated with lower levels of perceived control at any age, but that the impact of stress exposure was stronger in older adults. This effect was also evident for change in mastery over time.

Discussion. The findings from this study suggest that exposure to life events is an important, yet overlooked, determinant of age-related decline in control. Loss of personal and social resources may be the reason older adults appear more vulnerable to the negative effects of stress.

Key Words: Mastery • Aging • Life events • Stress process

INDIVIDUALS'<--CO?1--> sense of personal control refers to the degree to which individuals feel that their actions direct or alter outcomes in their social worlds. Individuals possessing a strong sense of personal control report lower levels of psychological distress (Avison & Cairney, 2003; Pearlin, Lieberman, Menaghan, & Mullan, 1981), fewer physical health problems (Seeman & Lewis, 1995), and more positive subjective appraisals of health (Ross & Bird, 1994) when compared to individuals with a weak sense of personal control. In social gerontology, some have gone as far as to argue that sense of control is a fundamental marker for successful aging, suggesting that individuals who possess this orientation to the world are better able to adapt to the challenges of growing older (Rowe & Kahn, 1998).

Regrettably, given the importance of control to well-being, previous research has found an accelerated decline in perceived control after age 50 (Mirwosky, 1995, 1997; Wolinsky & Stump, 1996). Data from longitudinal studies have also shown a decline in control with age (Ross & Mirowsky, 2002; Wolinsky, Wyrwich, Babu, Kroenke, & Tierney, 2003). If researchers are to ensure that all older adults age successfully, they must first understand the reasons for this decline in control with age. At present, research has shown three factors—education, physical health, and subjective life expectancy—to be responsible for declining control across age groups (Mirowsky, 1995, 1997; Wolinsky & Stump, 1996; Wolinsky, Wyrwich, Babu, et al., 2003). Educational attainment is thought to strengthen feelings of control by improving problem-solving abilities and increasing access to other control-enhancing resources (e.g., money and better jobs; Ross and Mirowsky, 2002). Declining health status and increasing functional impairments exert a negative influence on sense of control by limiting the range of activities in which a person can exercise control over his or her environment (Schulz, Wrosch, & Heckhausen, 2003). With regard to the age decline in control, cohort changes in formal educational attainment favoring younger age groups (Mirowsky, 1995), and the increase in prevalence of physical health problems and functional limitations observed with age (Mirowsky, 1995; Wolinsky, Wyrwich, Babu et al., 2003), are explanations for why older age groups seem to report less control than younger groups. There is also some evidence that individuals who believe they will live longer (longer subjective life expectancy) maintain a stronger sense of control as they age (Mirowsky, 1997). Although the evidence for these three characteristics is compelling, not all relevant factors influencing age-related decline in control have been tested.

Rodin (1986) identified three explanations for age differences in personal control—increased prevalence of negative life events (especially those associated with social role transitions such as loss of spouse), increased health problems, and greater contact with the health care sector. Research has paid very little attention to either the role of health care contact or the impact of negative life events on declining perception of control with age. With regard to the latter, we could locate only two studies that examined the impact of stress on control (Wolinsky, Wyrwich, Babu, et al., 2003; Wolinsky, Wyrwich, Kroenke, et al., 2003); only one of these considered exposure to a major life event—9/11 (Wolinsky, Wyrwich, Kroenke, et al., 2003). Although 9/11 is most certainly a major event, it is very unique and of largely historical nature. Moreover, 9/11 by no means exhausts the universe of possible stressful events that may negatively impact on perceived control. In order to gauge the potential contribution of stress to the study of change in control, it is important to see if a wider range of events is associated with it. Perceived stress in this work was based on a two-item global daily stress scale, not a life events measure. Most investigators in the life event field prefer to measure events directly and explicitly, and this is most commonly done with a checklist of multiple stressful experiences (Turner & Wheaton, 1997). Because these checklists tend to be the most common and favored way of assessing stress, many questions about the interface between stress and age-related decline in control remain unanswered.

The absence of research on the impact of more commonly experienced life events on control is particularly surprising given the centrality of this construct in the stress process literature (Avison & Cairney, 2003; Pearlin & Pioli, 2003). In what is now regarded as a classic study in the field of stress and mental health, exposure to a negative life event, job loss, was associated with decreased mastery, which in turn led to negative secondary stress and eventually greater psychological distress (Pearlin et al., 1981). Stress, in Pearlin's conceptualization, "confront(s) people with dogged evidence of their own failures—or lack of success—and with the inescapable proof of their inability to alter the unwanted circumstances of their lives" (Pearlin et al., 1981, p. 340). It is not surprising to find that exposure to stress can often rob people of their sense of control. Several other studies have documented a similar effect of stress on mastery and subsequent distress (Aneshensel, Pearlin, Mullan, Zarit, & Whitlatch, 1995; Avison, 1995; Broman, Hamilton, & Hoffman, 1990). This erosion of perceived control by stress, while certainly not the only mechanism by which mastery may influence the stress process (Avison & Cairney, 2003; Pearlin, 1989), is nevertheless a pivotal process that helps explain the negative physical and mental health outcomes that arise from exposure to stress.

There are several reasons why life events are particularly important factors capable of eroding feelings of control. By definition negative life events are often sudden, dramatic experiences that have the potential to significantly alter one's social world (e.g., death of a spouse, unemployment; Wheaton, 1994). The sheer forcefulness of the changes that accompany life events is often enough to significantly challenge a person's belief in his or her ability to exert control in the world. Major life events can also compromise or overwhelm an individual's capacity for problem solving and reasoning (Caplan, 1981). The cognitive demands imposed by major negative events may be so overwhelming that the person is left feeling helpless. Finally, life events often give rise to secondary stressors that, in concert, further erode feelings of control (Pearlin et al., 1981; Wheaton, 1994).

Although life events are likely to erode feelings of control at any age, the impact of cumulative exposure to life events may be especially deleterious in old age because many of the personal and social factors thought to bolster feelings of control are themselves diminished. For example, income and cognitive function are associated with sense of control (Lachman & Weaver, 1998) and, for many adults, diminish with age (George, Landerman, Blazer, & Anthony, 1991). At a time when other personal and social resources are already compromised, the additional burden arising from stress exposure may be even more deleterious to feelings of control. Indeed, the accelerated rate of decline in perceived control after age 50 (Mirowsky, 1995) suggests that if stress does play a role in reducing feelings of mastery, the impact of stress exposure on mastery may be stronger in old age. However, age vulnerability to stress has not been tested in the extant literature. In this study, we used a large sample of individuals aged 18 and older to examine the association between age, stress, and control. In particular, we assessed the impact of one form of stress—life events—on mastery across age groups. As well, we examined whether cumulative exposure to multiple life events (exposure to more than one life event over the course of a year) is associated with decreasing control, and whether older adults are more vulnerable to the pernicious, control-eroding effects of stress.

	METHODS

TOP Abstract Methods Results Discussion References

 
Sample
In this study, we used the longitudinal biennial National Population Health Survey (NPHS) conducted by Statistics Canada. It is a telephone survey of a national probability sample of Canadian residents across all 10 provinces. Using a multistage, stratified random sampling procedure, Statistics Canada interviewers chose one person from each surveyed household to provide detailed personal information for the longitudinal component of the survey. Persons living on Native reserves, on military bases, in institutions, and in some remote areas in Ontario and Quebec were excluded. The first wave of the NPHS (Wave 1), collected in 1994, was a telephone survey of a national probability sample of Canadian residents across all 10 provinces. The household-level response rate was 88.7%. Of the 18,342 possible respondents aged 12 and older, 17,626 participated, resulting in a person-level response rate of 96.1%. In Wave 1, 16,291 respondents were aged 18 or older; after a listwise deletion of cases with missing values, the total sample was reduced to 15,410 (a loss of 5.2%). The longitudinal design of the NPHS required respondents to be reinterviewed every 2 years. However, we did not use data from Wave 2 (1996) or Wave 3 (1998) in this analysis because variables on mastery and recent life events were not included in these waves. These variables were, however, reintroduced in Wave 4, making it possible for us to assess change over these waves. Of the respondents with complete data in Wave 1, 840 died and 5,049 either could not be located, declined to participate, or provided incomplete data at Wave 4. This left 9,521 respondents for the longitudinal analysis.

Dependent Variable (Mastery)
Researchers have operationalized the sense of control construct in a number of ways (e.g., fatalism, Wheaton, 1983; locus of control, Rotter, 1966). In this study, we used Pearlin and Schooler's (1978) seven-item measure of mastery (see Table 1). This measure is conceptually quite close to the sense of control measure used in previous work in this area (Mirowsky, 1995; Wolinsky & Stump, 1996; Wolinsky, Wyrwich, Babu et al., 2003; Wolinsky, Wyrwich, Kroenke et al., 2003). This index is scored so that higher scores indicate a greater sense of mastery (range = 0–28). Items 5 and 6 were reverse coded. The internal reliability of the scale (Cronbach's alpha) at Wave 1 was.94 and at Wave 4 was.77. Table 2 provides descriptive data for both Wave 1 and Wave 4 for this and all other measures in the study.

Recent life events
This index was based on the work of Wheaton (1994; see Table 1). It assessed the number of negative life events that the respondent or someone close to the respondent experienced in the 12 months prior to the survey; higher scores represent greater stress (range = 0–10).

Physician contact
Respondents were asked how many times they had seen or talked with a family doctor or general practitioner in the past 12 months, excluding any time that they might have spent as an overnight patient in the hospital.

Physical health measures
The first measure was an index of chronic conditions. In the NPHS, respondents were asked a series questions about the presence of chronic conditions, for example, "Do you have diabetes diagnosed by a health professional?" ("yes" or "no"). A list of 21 chronic conditions was selected from a larger list generated by Statistics Canada, which included asthma, arthritis, high blood pressure, chronic bronchitis, heart disease, stomach ulcers, effects of stroke, migraine headaches, and urinary incontinence. Following previous work, an index was created summing "yes" responses to each health condition (range = 0–21).

A measure of limitations in daily activities was also included. Respondents were asked six questions: "Do you need the help of another person with ...: (1) grocery shopping; (2) normal housework; (3) heavy housework; (4) personal care; (5) moving about in your home; (6) meal preparation?" "Yes" responses were summed to create an index. High scores indicate more limitations in daily activities.

Social support
An index of perceived social support was included. Respondents were asked to answer "yes" or "no" to four questions asking them whether they had someone (a) they could confide in, (b) they count on, (c) who could give them advice, and (d) who made them feel loved. High scores indicate greater perceived social support (Wave 1, =.67; Wave 4, =.64).

Socioeconomic measures
Socioeconomic status is measured by using two variables: education and income adequacy. Education, based on highest level of completed education, was coded into four categories: (1) less than high school, (2) high school, (3) some postsecondary, (4) postsecondary.

A five-level measure of income adequacy was used. This measure was derived by taking into account both total household income from all sources in the 12 months preceding the interview and the number of persons living in the household at the time of the survey. Individuals were then placed into one of five categories ranging from lowest to highest, based on low-income cutoff values derived by Statistics Canada. Low income adequacy, for example, was defined as less than $15,000 for a household of one or two persons, or less than $30,000 total household income when five or more persons are present. In this analysis, we combined the lowest two income categories and contrasted them with the higher ones.

Demographics
Age was coded in years (range = 18–102). Gender was coded 1 for women, 0 for men. Marital status included two binary variables for single and previously married (including widowed, divorced, and separated), with married or common-law as the reference category.

Analysis
We designed the analysis for this study to test for age differences in the effect of life events on mastery, controlling for other factors known to be correlated with sense of control. Because previous work has established that the association between age and control is nonlinear (e.g., Mirowsky, 1995, Shaw & Krause, 2001), it was important to test for a similar pattern in these data. We first tried to fit age to the data using the specification provided by Mirowsky (1995, 1997). Although statistically significant, the fit of the term derived from Mirowsky's formula was actually slightly lower than the simple quadratic term (r² = 0.007 compared to r² = 0.009). In light of these findings, and in the interest of parsimony, we included a quadratic (age-squared) term to model the relationship between age and mastery.

There is some debate in the literature on the appropriateness of excluding lower order terms in regression models estimating nonlinear associations. Although convention dictates including lower and higher order terms in the same model, it has been argued that, when the level of measurement of the variable is interval-ratio, it is sufficient to include only the higher order (quadratic) term (Allison, 1977). We followed this approach to be consistent with previous work in this area (e.g., Mirowsky, 1995, 1997; Shaw & Krause, 2001).

Because our approach differed somewhat from previous work (e.g., Mirowsky, 1995, 1997; Wolinsky & Stump, 1996; Wolinsky, Wyrich, Babu et al., 2003), a comment on our strategy is warranted. Previous work in this area has tended to begin by first establishing the association between age (and, commonly, a transformation of age) and sense of control, followed by serial adjustments (stepwise entry of variables) for other factors such as education or functional ability to assess what impact, if any, the introduction of these measures has on the association between age and control. In other words, the approach establishes whether factors such as health status and/or education mediate the relationship between age and control. Although this is a reasonable approach, a different strategy is required when stress is brought into the foreground. For one thing, there is at best a weak relationship between age and stress. Moreover, where there is a relationship to be found, typically older people report fewer stressors than younger adults (Hughes, Blazer, & George, 1988). This may be due to factors such as role loss (e.g., work-related and marital stressors are simply not applicable for many older adults). So, a mediating model is likely to shed little light on age-related decline in control. Instead, we propose using a moderating model, which is built on the assumption that people become increasingly vulnerable to the effects of stress with advancing age. In technical terms, this hypothesis is supported if a statistical interaction is found between age-squared and stress on control. As such, we estimated the following equations:

where a is the intercept, b represents unstandardized regression coefficients, and Predictors denotes marital status, income adequacy, education, chronic health problems, limitations in activities of daily living, number of general physician visits in the past year, and social support. The first two models were both to test for a nonlinear association between age and mastery (Model 1) and to confirm that life events has little, if any, mediating effect on the association between age-squared and mastery (Model 2). Models 3 and 4 tested for an age vulnerability to stress before and after the introduction of other factors known to be associated with perceived control. In order to interpret the nature of the interaction between age, life events, and mastery, we used formulas provided by Aiken and West (1991) to graph the association between age, life events, and mastery.

We were also interested in whether the age differences in the impact of life events influence change in mastery over time. As noted earlier, mastery and life events were not included in Waves 2 and 3 of the NPHS but were reintroduced in Wave 4. It was possible, therefore, to examine change in mastery and life events as assessed at two intervals (6-year time gap). We estimated the following equation:

By including interactions that assessed both Age x Wave 1 Life Events and Age x Wave 4 Life Events, we provided a preliminary test that change in age-related reactions to life events shapes change in mastery over time. It is not uncommon, for example, for investigators to conduct longitudinal analyses by focusing solely on the interaction between age and Wave 1 life events on some follow-up measure such as mastery or sense of control. However, if the effects of age-related reactions to life events emerge fairly quickly (proximal to the exposure), then this may be more likely to be found cross-sectionally, and hence, the effect of Age x Life Events measured at Time 1 would not predict change in mastery. By including an interaction with life events measured after baseline measures of mastery are taken, but before the follow-up measures of mastery (Wave 4 life events), we can see if events that arise after mastery has been assessed at baseline exert a pernicious effect on mastery at follow-up, thereby fully exploiting the longitudinal nature of the data.

We weighted all data to adjust for nonresponse and bootstrapped all estimates using the WesVar program (Westat, 2002) to take into account the complex design of the NPHS.

	RESULTS

TOP Abstract Methods Results Discussion References

 
We present the results from this study in three sections. First we present findings from an analysis that was designed to see if loss of participants over time may have biased the longitudinal study findings. Next we present the results of the cross-sectional analyses using Wave 1 data to test for an Age x Stress (recent life events) interaction on mastery. Finally, we examine the effect of age reactivity to stress over time using both Waves 1 and 4 of the study.

Exploring the Effects of Sample Attrition
One of the challenges with longitudinal data analysis is sample loss (attrition) over time. Although it is difficult to determine whether sample attrition biases study findings, it is possible to ascertain whether select variables from Wave 1 predict loss over time (Wave 4). This provides some indication of departures from randomness in sample attrition and therefore highlights potential biases. To examine this, we first created a nominal-level variable contrasting those who remained in the study (reference category), from those who were presumed to be alive but had dropped out (refused to participate or could not be located; scored 2), from those who died (scored 3). Using multinomial logistic regression (not shown here), we examined the effect of Wave 1 measures on this nominal measure.

Several significant associations emerged. Compared to those who remained in the study, women (b = –0.85, p <.001), individuals with missing values on income (b = –0.61, p <.05), and individuals with some postsecondary-level education (b = –0.60, p <.01) were less likely than men, those from the high-income group, and those with high school level education, respectively, to have died by Wave 4. Individuals who were older (b = 7.62 x 10^–4, p <.01), those who had higher levels of physical disability (b = 0.43, p <.01), and those who had a higher number of chronic health problems (b = 0.11, p <.01) were also more likely to have died by Wave 4. When comparing those who had dropped out or could not be located versus those who remained, women (b = –0.16, p <.01) and those with higher mastery scores (b = –0.01, p <.05) were less likely to have dropped out by Wave 4. Single individuals (b = 0.28, p <.01), those with low income adequacy (b = 0.28, p <.01), those with higher levels of physical disability (b = 0.14, p <.01), and those with higher levels of social support (b = 0.07, p <.05) were all more likely to have dropped out of the study. As we proceed in reviewing the substantive findings of this study, readers should consider the potentially biasing influence of this nonrandom sample loss.

Findings From the Cross-Sectional Analysis
In the first part of the analysis, we present simple descriptive statistics for key study variables in both waves (see Table 2).

In the first set of substantive analyses we assessed whether an age-related vulnerability to stress influences the perception of mastery. As outlined in "Analysis," we tested four models. In Model 1 of Table 3, we tested for the association between age and mastery using a quadratic term for age. As others have observed, the relationship between age and mastery is nonlinear (b = –2.42 x 10^–4; β = –0.09; p <.001). In Model 2, we entered the life events measure into the equation with age-squared. We found support for the hypothesis that exposure to life events erodes feelings of personal control; exposure to a greater number of life events was associated with lower mastery (b = –0.88; β = –0.21; p <.001). However, as hypothesized above, there was no evidence that exposure to stress mediates the association between age and mastery. In fact, the coefficient representing age-squared increased slightly after we introduced life events into the model. In Model 3, we tested for an interaction between age and life events. The results indicated that the effect of life events on mastery was not the same across age groups (b = –1.91 x 10^–4; p <.001). This relationship was maintained after we adjusted for several other known predictors of mastery, including physical health status and perceived social support (see Model 5).

View larger version (9K): [in this window] [in a new window]   Figure 1. Net effect of life events on mastery with age. Note that 95% of observed values for life events stress fell below 3. Dotted lines represent model projections beyond this level

Findings From the Longitudinal Analysis
We also used a similar regression-based approach to examine the relationship between change in age-related reaction to life events and change in feelings of control over time. We were particularly interested in whether exposure to new life events in the year preceding the fourth wave and age predict change in control over time. Table 4 provides the results of this longitudinal analysis. Analyses (not shown) revealed that the linear term for age, not the quadratic, provided the best fit with change in control as the dependent measure. The results in Table 4 show that the interaction between age and exposure to life events in the year prior to the baseline survey was not significantly associated with change in mastery over a 6-year period. However, Age x Exposure to Life Events in the year preceding Wave 4 was significantly associated with change in perceived mastery over time (b = –1.03 x 10^–2; p <.05). We undertook similar steps to clarify the nature of this interaction using the formulas provided Aiken and West (1991). The additional computations (not shown) suggested a similar effect to the one found in the cross-sectional analysis; the effect of exposure to recent life events on change in mastery was strongest at age 65 (ß = –0.22; b = –0.96), slightly less at age 45 (ß = –0.17; b = –0.75), and weakest at age 25 (ß = –0.12; b = –0.54). The difference between coefficients was noteworthy, with the coefficient for the unstandardized coefficient increasing by 44% as we moved from the youngest to the oldest age groups: (0.96 – 0.54) / 0.96.

	DISCUSSION

TOP Abstract Methods Results Discussion References

 
Although research shows that sense of control declines with age (Mirowsky, 1995; Shaw & Krause, 2001; Wolinsky & Stump, 1996; Wolinsky, Wyrich, Babu et al., 2003), previous work has paid little attention to the role of life events in this relationship. Our results show that life events do exert a negative effect on mastery and that the effect of stressful life events is not the same at different points of the life course. Rather, the effect of life events on mastery in old age appears to have an especially erosive effect. In other words, although exposure to negative life events can erode feelings of control in younger and middle-aged adults (Pearlin et al., 1981), it appears to have an even more pronounced effect in later life. This resonates with what researchers know about both the impact of life events on psychosocial resources, and the situation of many older adults with regard to social and personal resources. Indeed, the differential age vulnerability observed here may have arisen from the loss of resources that are thought to offset the pernicious effects of stress. Old age has been characterized as a time of loss, both personally and socially (Baltes & Smith, 1999). One may therefore regard negative life events as a catalyst by which, in the face of already compromised social and personal resources, sense of control is eroded. This may be especially true with regard to cognitive functioning. The social gerontological literature has already established the importance of cognitive function as a resource for coping with stress (Krause, 1996; Krause & Thompson, 1998). This literature may be particularly useful for understanding the age-related vulnerability to stress in relation to control. Unfortunately, our data did not include a measure of cognitive function, thereby precluding an empirical examination of the role of this factor in perceived control.

Although a broad range of stress coping resources and strategies are potentially available to the older adult, problem-focused coping is among the most studied (Krause & Thompson, 1998). Problem-based coping requires reasonably sound cognitive and problem-solving skills, abilities that become compromised with age, even in community-dwelling older adults (George et al., 1991). In the context of the stress process, the diminishing capacity for problem-based coping due to increasing cognitive impairment with age may explain why control is especially compromised when people are confronted with stressful events in later life. Indeed, previous work has shown a greater vulnerability to stress in cognitively impaired older adults (Krause & Thompson, 1998). Although speculative, cognitive impairment may be an important mechanism linking stress vulnerability to perceived control. Further work is clearly required to understand why life events appear to have a greater impact on mastery in later life.

Our analyses also reveal that the impact of age-related reactions to life events occur relatively soon after exposure to stress. We found that the age differences in reaction to life events in the year preceding the last wave of our analysis were associated with negative changes in mastery. Conceptually, this makes sense; it is difficult to imagine why mastery would be adversely affected by exposure to negative life events years after the initial stressors have occurred. Unlike trauma, which produces lasting, pervasive negative effects (Krause, Shaw, & Cairney, 2004), negative life events still fall within the realm of normal experiences. Moreover, many older people eventually recover from the stressors that confront them (Turner & Avison, 1992). For example, Wolinsky, Wyrwich, Kroenke, and colleagues (2003) reported that the impact of 9/11 on sense of control was no longer evident 6 to 9 months after this catastrophe occurred. When viewed in conjunction with the findings from our study, these results point to a fluid process that is best captured with data that have been gathered at more than one point in time.

There are several limitations in these data that should be considered when further pursuing this line of inquiry. We cannot conclusively determine causality with these data, primarily because we can never be sure that all known factors composing both life events and mastery were included in the analyses. Nevertheless, by using two waves of data, we have addressed one of the classic criteria for establishing causality (temporality). Our results reveal that change in age-related reactions to life events are associated with negative changes in mastery over time. As our findings reveal, this is best observed when life events are assessed after baseline measures of mastery are taken but before the follow-up measures of mastery are obtained. Even so, a more fine-grained series of analyses is needed to identity the precise point when age and stress exert a maximum effect on mastery. We assessed life events in the year prior to the interviews, but it is possible, for example, that events arising in the 6-month period prior to the interview have the greater effect on mastery. We were unable to evaluate this option because data on the precise timing of each event were not available.

Second, there is considerable debate in the literature about the utility of global measures of control for understanding age-related change (Krause, 2003; Pearlin & Pioli, 2003). Although virtually all research in this area has relied on these kinds of measures (Mirowsky, 1995, 1997; Shaw & Krause, 2001; Wolinsky & Stump, 1996; Wolinsky, Wyrwich, Babu et al., 2003; Wolinsky, Wyrwich, Kroenke et al., 2003), it is important to extend this work to examine whether specific domains of social life may be more or less sensitive to change with age. Krause (2007) recently examined whether feelings of control over highly valued social roles (role identity hypothesis) also show decline with age in later life. His work confirmed that control in such roles shows a similar pattern to more global measures and that social support, in particular anticipated social support, is a control-enhancing factor at least until age 75. It is not known whether stress can also help to explain age-related decline in highly valued social roles. A fruitful approach in this regard would be to examine specific kinds of stress that may be more damaging to certain roles. For example, older adults who perceive their most important social role to be provider for the household may be especially vulnerable to stressors that compromise control in this domain (e.g., financial stress or worry). Because financial resources tend to decline with age, older adults who continue to value this particular role over others are likely to experience loss of control in this domain over time. Independent of domain-specific analyses, other more direct measures of control exist (see Mirowsky, 1997) and should be included in future work.

Finally, our measure of life events was also limited in relation to other similar measures, as it measured only eight life events. Moreover, it was not possible to ascertain whether the event occurred to the participant or to other family members, making it impossible to determine whether the impact of stress on personal control is more or less pronounced when negative events occur directly to the individual or are experienced vicariously through others. If the impact of events is differentially determined by the relationship with the person who experienced them, then by not weighting these events we are likely to have underestimated the effects of stress. As well, life events that may be especially relevant to older adults (e.g., moving) were excluded from the index altogether. Therefore, once again, we may have underestimated the impact of cumulative exposure to life events, which in turn may have influenced the age-related patterns found here. Moreover, many kinds of stressors, including chronic strains, daily hassles, role strains, and exposure to traumas such as violence, were not included in the NPHS but may also impact perceptions of control. It will be important in future work to tap into the broader universe of stressors, including age-specific stressors, to examine the control-eroding impact of stress.

Although, as with any study, limitations exist, perhaps the most significant contribution of this work is the introduction of stress, specifically life events, into the study of age-related decline in control. Indeed, it is interesting that research has not explored the role of stress given the influence of stress process research to this area. Our results show that cumulative exposure to life events can now be added to educational attainment, functional disability, and subjective life expectancy as a key determinant of age differences in perceptions of control. It is only through a better understanding of the factors that shape perceptions of control with age that researchers may hope to intervene to improve the quality of life for older adults.

	Acknowledgments

 
This research was supported by a career award (Canada Research Chair) from the Canadian Institutes of Health Research to Dr. John Cairney. Dr. Neal Krause was supported by Grant RO1 AG009221 from the National Institute on Aging. We wish to acknowledge Scott Veldhuizen for his assistance with the data analysis for this project.

J. Cairney planned the study, supervised the statistical analyses, and participated in the writing of the paper. N. Krause contributed to the plan of the study, supervised the statistical analyses, and also participated in the writing of the paper.

	Footnotes

 
Decision Editor: Kenneth F. Ferraro, PhD

Received for publication February 5, 2008. Accepted for publication February 7, 2008.

	References

TOP Abstract Methods Results Discussion References

 

• Aiken, L. S., West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Newbury Park, CA: Sage.
• Allison, P. D. (1977). Testing for interaction in multiple regression. American Journal of Sociology, 83,144-153.
• Aneshensel, C. A., Pearlin, L. I., Mullan, J. T., Zarit, S. H., Whitlatch, C. (1995). Profiles in caregiving: The unexpected career. San Diego, CA: Academic Press.
• Avison, W. R. (1995). Roles and resources: The effects of work and family context on women's psychosocial resources and psychological distress. In J. R. Greenly (Ed.), Research in community and mental health (Vol. 8, pp. 233–256). Greenwich, CT: JAI Press.
• Avison, W. R., Cairney, J. (2003). Social structure, stress and personal control. In S. H. Zarit, L. I. Pearlin, & K. W. Schaie (Eds.), Personal control in social and life course contexts (pp. 127–164). New York: Springer.
• Baltes, P. B., Smith J. (1999). Mutilevel and systemic analysis of old age: Theoretical and empirical evidence for a Fourth Age. In V. L. Bengston & K. W. Schaie (Eds.), Handbook of theories of aging (pp. 153–173). New York: Springer.
• Broman, C. L., Hamilton, V. L., Hoffman, W. S. (1990). Unemployment and its effect on families: Evidence from a plant closing study. American Journal of Community Psychology, 18,643-659.
• Caplan, G. (1981). Mastery of stress: Psychosocial aspects. American Journal of Psychiatry, 138,413-420.[Abstract/Free Full Text]
• George, L. K., Landerman, R., Blazer, D. G., Anthony, J. C. (1991). Cognitive impairment. In L. N. Robins & D. A. Reiger (Eds.), Psychiatric disorders in America: The Epidemiology Catchment Area Study (pp. 291–327). New York: Free Press.
• Heckman, J. (1979). Sample selection bias as a specification error. Econometrica, 47,153-161.
• Hughes, D. C., Blazer, D. G., George, L. K. (1988). Age differences in life events: A multivariate controlled analysis. International Journal of Aging and Human Development, 27,207-220.[Medline]
• Krause, N. (1996). Stress, gender cognitive impairment, and outpatient physician use in later life. Journal of Gerontology: Psychological Sciences, 51B,P15-P23.[Abstract]
• Krause, N. (2003). The social foundations of personal control in late life. In S. H. Zarit, L. I. Pearlin, & K. W. Schaie (Eds.), Personal control in social and life course contexts (pp. 45–70). New York: Springer.
• Krause, N. (2007). Age and decline in role-specific feelings of control. Journal of Gerontology: Social Sciences, 62B,S28-S35.[Abstract/Free Full Text]
• Krause, N., Shaw, B., Cairney, J. (2004). A descriptive epidemiology of lifetime trauma and the physical health status of older adults. Psychology and Aging, 19,637-648.[Medline]
• Krause, N., Thompson, E. (1998). Cognitive functioning, stress, and psychological well-being in later life. Journal of Mental Health and Aging, 4,(1), 155-170.
• Lachman, M. E., Weaver, S. L. (1998). The sense of control as a moderator of social class differences in health and well-being. Journal of Personality and Social Psychology, 74,763-773.[Medline]
• Mirowsky, J. (1995). Age and the sense of control. Social Psychology Quarterly, 58,31-43.
• Mirowsky, J. (1997). Age, subjective life expectancy, and the sense of control: The horizon hypothesis. Journal of Gerontology: Social Sciences, 52B,S125-S134.[Abstract]
• Pearlin, L. I. (1989). The sociological study of stress. Journal of Health and Social Behavior, 30,241-256.[Medline]
• Pearlin, L. I., Lieberman, M. A., Menaghan, E. G., Mullan, J. T. (1981). The stress process. Journal of Health and Social Behavior, 22,337-356.[Medline]
• Pearlin, L. I., Pioli, M. (2003). Personal control: Some conceptual turf and future directions. In S. Zarit, L. I. Pearlin, & K. Schaie (Eds.), Personal control in social and life course contexts (pp. 1–19). New York: Springer.
• Pearlin, L. I., Schooler, C. (1978). The structure of coping. Journal of Health and Social Behavior, 19,2-21.[Medline]
• Rodin, J. (1986, September 19). Aging and health: Effects of the sense of control. Science, 233,1271-1276.
• Ross, C. E., Bird, C. E. (1994). Sex stratification and health lifestyle: Consequences for men's and women's health. Journal of Health and Social Behavior, 35,161-178.[Medline]
• Ross, C. E., Mirowsky, J. (2002). Age and the gender gap in the sense of personal control. Social Psychology Quarterly, 65,125-145.
• Rotter, J. B. (1966). Generalized expectancies for internal versus external control of reinforcement. Washington, DC: American Psychological Association.
• Rowe, J. W., Kahn, R. L. (1998). Successful aging. New York: Random House.
• Schulz, R., Wrosch, C., Heckhausen, J. (2003). The life-span theory of control: Issues and evidence. In S. H. Zarit, L. I. Pearlin, & K. W. Schaie (Eds.), Personal control in social and life course contexts (pp. 233–262). New York: Springer.
• Seeman, M., Lewis, S. (1995). Powerlessness, health and mortality: A longitudinal study of older men and mature women. Social Science & Medicine, 41,517-525.
• Shaw, B. A., Krause, N. (2001). Exploring race variations in aging and personal control. Journal of Gerontology: Social Sciences, 56B,S119-S124.[Abstract/Free Full Text]
• Turner, R. J., Avison, W. R. (1992). Innovations in the measurement of life stress: Crisis theory and the significance of event resolution. Journal of Health and Social Behavior, 33,36-51.[Medline]
• Turner, R. J., Wheaton, B. (1997). Checklist measurement of life events. In S. Cohen, R. Kessler, & L. U. Gordon (Eds.), Measuring stress: A guide for health and social scientists (pp. 29–58). New York: Oxford University Press.
• Westat. (2002). WesVar 4.2 user's guide. Washington, DC: Author.
• Wheaton, B. (1983). Stress, personal coping resources, and psychiatric symptoms: An investigation of interactive models. Journal of Health and Social Behavior, 24,208-229.[Medline]
• Wheaton, B. (1994). Sampling the stress universe. In W. R. Avison & I. H. Gotlib (Eds.), Stress and mental health: Contemporary issues and prospects for the future (pp. 77–114). New York: Plenum Press.
• Wolinsky, F. D., Stump, T. E. (1996). Age and the sense of control among older adults. Journal of Gerontology: Social Sciences, 51B,S217-S220.[Abstract]
• Wolinsky, F. D., Wyrwich, K. W., Babu, A. N., Kroenke, K., Tierney, W. M. (2003). Age, aging, and the sense of control among older adults: A longitudinal reconsideration. Journal of Gerontology: Social Sciences, 58B,S212-S220.[Abstract/Free Full Text]
• Wolinsky, F. D., Wyrwich, K. W., Kroenke, K., Babu, A. N., Tierney, W. M. (2003). 9-11, personal stress, mental health and sense of control among older adults. Journal of Gerontology: Social Sciences, 58B,S146-S150.[Abstract/Free Full Text]

This Article Abstract Full Text (PDF) Services Download to citation manager PubMed PubMed Citation