Downward social comparison and subjective well-being in late life: The moderating role of perceived control

Abstract

Objective: Drawing from Heckhausen and Schulz's Motivational Theory of Life-span Development, this study examined perceived control as a moderator of the protective relationship between downward social comparison and subjective well-being among older adults.

Methods: Community-dwelling older adults (N = 97, 63% female, ages 79–97) were interviewed in their own homes at three time-points over a nine-year period. Interviews assessed older adults’ perceived control over daily tasks, their use of downward social comparison in response to task restriction, and their subjective well-being.

Results: Regression analyses yielded a significant interaction between downward social comparison and perceived control for three subjective well-being outcomes: life satisfaction, perceived stress, and depressive symptoms. Follow-up analyses revealed that downward social comparison was associated with greater subjective well-being at low levels of perceived control; but was unrelated to subjective well-being at high levels of perceived control.

Conclusion: These findings corroborate Heckhausen and Schulz's theorized goal-opportunity congruence premise and have implications for quality-of-life interventions to assist community-dwelling older adults.

Keywords:

• downward social comparison
• older adults
• subjective well-being
• secondary control
• perceived control

Acknowledgments

This research was supported by a Social Sciences and Humanities Research Council of Canada (SSHRC) postdoctoral fellowship to the first author (#756-2009-0304 04); as well as a SSHRC standard grant and a Canadian Institutes of Health Research (CIHR) operating grant to the second author.

Notes

Notes

1. It is important to note that past research has demonstrated both positive and negative effects of downward social comparison on subjective well-being (for detailed discussion, see Taylor & Lobel, 1989). Whether downward social comparison is beneficial or detrimental seems to depend upon whether the comparer feels a sense of assimilation or contrast to the target. Assimilation occurs when the comparer believes ‘I could end up in the same circumstances’ as the target; contrast occurs when the comparer believes ‘I am very different’ from the target. Thus, another person's poor-standing may increase the comparer's well-being (contrast), but it could also reduce well-being if it suggests that the comparer's own status could deteriorate to the same low level (assimilation). Whether assimilation or contrast occurs depends upon: (1) The comparer's psychological connection or sense of identification with the target. Contrast is more likely when the comparer does not feel close to, or does not identify with, the target; and (2) Relevance of the comparison dimension: Contrast is more likely when the comparison dimension has personal relevance to the comparer. Our measure of downward social comparison was designed to produce contrast (not assimilation), and thereby promote subjective well-being among the comparer. We phrased our items as a comparison with a general or vague comparison target (‘others your age’) to promote low identification with the target; and, we ensured that the comparison was made in a relevant domain by selecting only those participants who were struggling with task restriction (see Section ‘Method’).

2. Group comparisons revealed no significant differences between the N = 97 participants in our study and the excluded N = 135 participants in terms of age (M = 85.04, SD = 4.24; M = 85.01, SD = 4.36 respectively; t ₂₃₀ = 0.04, p > 0.05), gender (37.1% male; 37.0% male respectively; χ ²= 0.001, df = 1, ns.), or education (M = 10.18, SD = 2.51; M = 10.65, SD = 2.69 respectively; t ₂₃₀ = 1.35, p > 0.05). In addition, attrition analyses revealed no differences between those who did (n = 60) and those who did not (n = 37) participate in 2005 in terms of all study variables, with one exception: The groups differed in terms of chronological age. The n = 37 participants who did not participate in the 2005 follow-up interview were significantly older than the n = 60 who did complete the follow-up ((M = 87 years old, SD = 4.2 versus M = 84 years old, SD = 3.8, respectively; t ₉₅ = 3.51, p < 0.001). This age difference may have contributed to the reasons for participant unavailability in 2005, which included movement from the community to personal care home, general unwillingness, scheduling conflicts, inability to contact, and death.

3. The distributions of both predictor variables (perceived control and downward social comparison) were negatively skewed (skew = −1.02, SE = 0.246; skew = −1.14, SE = 0.245 respectively). Non-normality among predictor variables is particularly problematic when the variables will be used to create a multiplicative interaction term: Non-normality can result in high correlation values between the centered predictor variables and the multiplicative interaction term (Aiken & West, 1991; Jaccard & Turrisi, 2003). To reduce the negative skew and normalize the distributions, we conducted square root variable transformations (reflected) for both predictor variables (Osborne, 2002). Unfortunately, the variable transformations did very little to reduce the skew of the predictor variables (perceived control skew reduced from −1.02 to −1.00; downward social comparison skew reduced from −1.14 to −1.12); and did not significantly reduce the correlation between the centered predictor variables and the multiplicative interaction term. As such, we chose to trim extreme values at the tail-end of the distribution (Osborne & Overbay, 2004). Rather than exclude participants with extreme low values (and incur a loss of power) we opted for truncation: The extreme values (i.e., those beyond two standard deviations below the mean) were re-coded into the nearest ‘acceptable’ value. For perceived control this meant that six scores ranging from 1 to 3 were re-coded into a value of ‘4’; for downward social comparison this meant that four scores ranging from 0.00 to 1.25 were re-coded into a value of ‘1.75’. This truncation procedure reduced the distribution skew from −1.01 to −0.65 for perceived control, and −1.14 to −0.03 for downward social comparison, and reduced the correlation between the centered predictor variables and the multiplicative interaction term.

4. A degree of inconsistency in the appearance of the interaction for perceived stress is worth noting: Participants with high perceived control reported somewhat greater levels of stress than we would have expected; relative to their low perceived control counterparts. That said, the direct effect of perceived control on perceived stress was not significant; nor was the slope of perceived stress on downward social comparison statistically significant at high levels of perceived control.