Scheier, M. F., Swanson, J. D., Barlow, M. A., Greenhouse, J. B., Wrosch, C., & Tindle, H. A. (2021). Optimism versus pessimism as predictors of physical health: A comprehensive reanalysis of dispositional optimism research. American Psychologist, 76(3), 529–548. Jun 2021. https://doi.org/10.1037/amp0000666
Abstract: Prior research has related dispositional optimism to physical health. Traditionally, dispositional optimism is treated as a bipolar construct, anchored at one end by optimism and the other by pessimism. Optimism and pessimism, however, may not be diametrically opposed, but rather may reflect 2 independent, but related dimensions. This article reports a reanalysis of data from previously published studies on dispositional optimism. The reanalysis was designed to evaluate whether the presence of optimism or the absence of pessimism predicted positive physical health more strongly. Relevant literatures were screened for studies relating dispositional optimism to physical health. Authors of relevant studies were asked to join a consortium, the purpose of which was to reanalyze previously published data sets separating optimism and pessimism into distinguishable components. Ultimately, data were received from 61 separate samples (N = 221,133). Meta-analytic analysis of data in which optimism and pessimism were combined into an overall index (the typical procedure) revealed a significant positive association with an aggregated measure of physical health outcomes (r = .026, p < .001), as did meta-analytic analyses with the absence of pessimism (r = .029, p < .001) and the presence of optimism (r = .011, p < .018) separately. The effect size for pessimism was significantly larger than the effect size for optimism (Z = −2.403, p < .02). Thus, the absence of pessimism was more strongly related to positive health outcomes than was the presence of optimism. Implications of the findings for future research and clinical interventions are discussed.
Discussion
The results of the present reanalyses confirm the findings from earlier quantitative and
qualitative reviews. The presence of optimism combined with the absence of pessimism (as
assessed by the overall/combined scale) is a reliable predictor of physical health. This was true
for an analysis that pooled all of the outcomes together and also true for the majority of analyses
that examined subgroups of outcomes separately. This replication of prior findings is noteworthy
inasmuch as over 80 percent of the studies included in the present reanalyses were not included
in the previous meta-analysis (Rasmussen et al., 2009).
The novel findings concern the relative strength of optimism and pessimism in contributing
to associations with health. Although each was a significant predictor of physical health, the Optimism, Pessimism, and Health
19
effect sizes associated with the absence of pessimism were generally greater in size than those
associated with the presence of optimism. The magnitude of these differences was great enough
to be significantly different for the analysis aggregating across outcomes, as well as for several
of the analyses that investigated subgroups of outcomes separately. Adjustment of the findings
for publication bias did little to alter the basic nature of the primary findings.
Moderator analyses were conducted on the effect sizes from the overall/combined scale, as
well as the two subscales. These analyses failed to identify any significant moderator. It is of
interest that there were no significant differences in effect sizes as a function of the type of study
employed. Cross-sectional studies are open to a number of methodological criticisms, most
notably the issue of reverse causality. Longitudinal studies examine associations across time, but
without provisions for equating the health of participants at baseline. As such, longitudinal
studies are subject to many of the same criticisms as are cross-sectional studies. Prospective
studies provide the gold standard, in that they offer an assessment of the change in the outcome
variable overtime (or otherwise start with participants who can be assumed to be equivalent in
health at baseline). Given these considerations, it is especially striking that the moderator
analyses revealed that study design did not significantly impact the magnitude of the effect sizes
that were obtained.
The foregoing discussion speaks to the statistical reliability of the effects that emerged. A
few words also need to be said about the magnitude of the effects that emerged. The effects sizes
reported here appear small. Several considerations should be borne in mind, however, when
evaluating the effect sizes obtained. First, as just noted, the effect sizes reported are adjusted for
a host of factors, including those related to demographics, study design, and other confounding
psychosocial factors. Thus, the effect sizes reported are unique to optimism and pessimism. It is
not surprising that the effect sizes are somewhat small, especially so inasmuch as shared variance
with related psychosocial factors had been removed.
The second point to make is that statistical effects, even small ones, can be quite
meaningful when applied to large numbers of people. Take for example, the effect size Optimism, Pessimism, and Health
20
characterizing the association between the pessimism subscale and mortality. The corresponding
adjusted odds ratio for this effect in the present reanalysis is 1.074 [95% CI (1.024, 1.126)]. In
terms of the number of people who lived and died in the United States in 2016 (the year the most
recent study in these reanalyses was published), this odds ratio implies that a 1-point change in
the pessimism direction of the pessimism subscale corresponds to an increase in 97,914 deaths
from all causes [95% CI (32,540, 162,641)].
Finally, it is worth mentioning that the size of the effects obtained using the present metaanalytic techniques are quite comparable to effects reported in other meta-analyses of
psychosocial factors and physical health when the studies are put on this same metric [see, e.g.,
Richardson et al. (2012) for a meta-analysis of perceived stress and incident coronary heart
disease and Kivimäki et al., 2012 for a meta-analysis of job strain and coronary heart disease].
Taken together, these considerations suggest that from a public health standpoint the magnitude
of the effects obtained in the present analysis are nontrivial and quite comparable to other
findings in the literature.
The present set of reanalyses has several potential limitations that should be highlighted.
First, search terms for the present analysis relied heavily on the framework used by Rasmussen et
al. (2009). The scheme used here is only one of many that could be adopted. Different search
terms could yield a different corpus of studies, and the findings obtained using those different
studies could be somewhat different.
Second, the yield rate for relevant studies was 32%. It is difficult to evaluate this yield rate
compared to other meta-analytic studies. This is the case because the data required for the
present study could not be extracted from published studies. Rather, the analysis was contingent
on authors of those published studies reanalyzing their data and forwarding on the results of
those re-analyses. It is likely that this extra requirement lowered the yield rate to some extent.
The third limitation concerns the homogeneous nature of the gender and racial composition
of the participants. Although these factors differed somewhat from study to study, over 90% of
the overall sample were white and women. Additionally, over 90% of the studies were conducted Optimism, Pessimism, and Health
21
in the United States. More studies are clearly needed to determine if the effects reported here are
replicable in more diverse populations.
Fourth, the conduct of the present research was a group effort. The analyses could not have
been done if consortium members had not conducted the needed analyses and forwarded their
findings to the primary authors for further meta-analytic processing. On the positive side, the
project represents one of the best examples of collaborative science in the truest sense of the
term. On the negative side, the more people involved, the more potential there is for error. This
concern is mitigated by the fact that the researchers involved had already published peer
reviewed papers with these same data, and as such had already demonstrated significant
capability with these analyses.
Finally, the outcomes examined in the present study all involved physical health. It is
unclear if similar findings would obtain if mental health outcomes were examined. Perhaps
optimism and pessimism would be equally robust as predictors of psychological well-being.
Perhaps optimism would be stronger. It is important not to extrapolate the findings obtained with
the present set of outcomes to possible findings involving other outcomes. Future research on
psychological well-being should report results for the optimism and pessimism subscales
separately, in order to evaluate the relative strength of the two dimensions in predicting
outcomes in that domain.
There is a more nuanced point to be made here than simply to acknowledge that the
differential impact of optimism and pessimism on psychological well-being needs to be
explored. That is, stress has been identified as one potentially important factor that might
mediate the impact of optimism (and pessimism) on physical health (Scheier & Carver, 2018).
How? The idea is that stress (and stress-related emotions) might modulate downstream biological
systems that underlie health and disease.
Optimists cope with and psychologically react to adversity in a different way than do
pessimists (Segerstrom et al., 2017). It would be interesting to see within this context if the
presence or absence of optimism and the presence or absence of pessimism relate differentially Optimism, Pessimism, and Health
22
to the various emotions that arise in reaction to stressful circumstances. It would further be
interesting to see if these potentially different emotions (that characterize the reactions of
optimists and pessimists to stress) might themselves be more or less strongly related to physical
health outcomes. Answering questions such as these could further in a significant way our
understanding of why it might be that the absence of pessimism is more strongly related to
physical health outcomes than is the presence of optimism.
Limitations aside, the present findings have at least three implications. First, future research
should, as a matter of course, provide effect size information for the overall/combined scale and
the two subscales separately—a suggestion that has been made previously (Scheier et al., 1994).
Such a practice is even more important now that quantitative data exist documenting the
differential associations of the two subscales with physical health. With the complete
complement of effect sizes reported, future research could continue to evaluate the importance of
the separate contributions of optimism versus pessimism without the need to establish
consortiums.
The present findings also hold important implications for positive psychology (Peterson &
Park, 2003; Seligman & Csikszentmihalyi, 2000). Positive psychology emphasizes those
characteristics that enable people to experience full, industrious, and resilient lives. As such, it
stands in contrast to traditional views that tend to focus on negative attributes, such as
depression, anxiety, and other characteristics which undermine successful living. Dispositional
optimism is often described as a good example of a variable falling within the positive
psychology domain (e.g., Dunn, 2018). As the present data make clear, however, the presence of
optimism does not provide the whole story. Optimism is important, but it does not appear to be
as important as the absence of pessimism in predicting physical health.
In the future, researchers in positive psychology might benefit from taking these findings
into account when planning and conducting research. Researchers should examine more closely
the predictor variables they are using to see if negative and positive characteristics might be
intermingled in the measures employed. If so, an effort should be made to tease apart the positive Optimism, Pessimism, and Health
23
and negative components of the measures to determine what is in fact responsible for doing the
predicting. Ultimately, it may turn out that it is the positive aspects of the measures that are
important, but it also possible that the negative features are the ones driving the observed
associations. Only by explicitly evaluating these possibilities will we know for sure.
The final implication concerns interventions. Future efforts to design and adapt
interventions to promote better health should keep in mind the differential links between
optimism, pessimism, and physical health. In this regard, it is interesting that some cognitive
behavior therapies seem to put a greater emphasis on lessening pessimism than they do on
promoting optimism. One example of such an intervention concerns cognitive restructuring
(Leahy & Rego, 2012), in which participants are trained to challenge the automatic thoughts,
beliefs, and expectancies underlying negative feelings. Participants confront their automatic,
negative thinking by systematically, and explicitly monitoring their moods and assessing in a
more objective fashion the information in the ongoing context that either supports or challenges
their negative thoughts. Perhaps existing interventions that focus more on lessening pessimism
such as those involving cognitive restructuring will be more successful in promoting better
health than will those that place a greater weight on promoting optimism, or even those that
place an equal weight on both components. Note that it is not a matter of causing harm, but more
a matter of targeting the component that offers the most gain.
It is also possible, however, that things are more complicated. Perhaps what works best will
depend on the nature of the outcome of interest (e.g., health behaviors versus biological
pathways). Intervention efforts with respect to optimism, pessimism, and physical health are still
in their infancy. As research in the intervention domain continues to evolve, it would seem
prudent to keep the distinction between optimism and pessimism in mind. Doing so may prove
profitable both practically and theoretically.
Posts
