Introduction
“How to keep, how to gain, how to recover happiness is”, wrote William James , “for most men [and women] at all times the secret motive for all they do”. Lucas and Diener (Reference Lucas, Diener and Diener2009) commented a century later that “After decades of research on [happiness] researchers have often arrived at what to some seems a startling conclusion: the most important factor in determining a person’s [happiness] appears to be the personality with which he or she was born” (p. 83). Neither Albert Kozma nor Michael Stones thought so when they began to study happiness in older people more than 30 years ago. A decade later, when they wrote Psychological Well-Being in Later Life with Kevin McNeil, they were more convinced.
This article begins with an overview of issues that directed our research on happiness during that period, and what we consider to be groundbreaking after the publication of our book (Kozma, Stones & McNeil, Reference Kozma, Stones and McNeil1991). The following sections are in the same order as the chapters in the book. Each section begins by summarizing the chapter’s content and then discusses later developments. We do not pretend that this article is in any way exhaustive, because such an endeavor would require another book.
Issues and Milestones in Research on Happiness
The question that spurred our thinking about happiness seemed simple at the outset: What makes older people happy? We were not alone in asking this question. Other researchers anticipated influences on happiness in older people to differ from those in younger people because of biological age changes, social factors, and chronic conditions of disease and disability. The predominant psychosocial debate during the 1960–1980 period was whether disengagement or high activity provided a better prescription for happiness in later life – a controversy that lost momentum only after the concept of successful aging gained currency.
Along with many other researchers in the 1970s, we had doubts about whether measures developed for use with young people would be equally applicable for use with older people. Few measures were available that were specific to the happiness and activity pursuits of older people. Moreover, we had concerns about the factorial and content validities reported for those measures (e.g., the Life Satisfaction Indices; the Philadelphia Geriatric Centre Morale Scale). Consequently, we developed a battery of new measures and subjected them to extensive psychometric evaluation. The battery included the Memorial University of Newfoundland Scale of Happiness (MUNSH; Kozma & Stones, Reference Kozma and Stones1980), the Memorial University Activities Inventory (MUNAI; Stones & Kozma, Reference Stones and Kozma1986a), and the Memorial University Mood Scale (MUMS; McNeil, Stones, Kozma, & Andres, Reference McNeil, Stones, Kozma and Andres1994). All of these measures were featured in the book.
Research on happiness before 1980 was almost entirely cross-sectional. Longitudinal research lasting beyond a few months was innovative. The earliest of our findings over an 18-month longitudinal interval (Kozma & Stones, Reference Kozma and Stones1983) showed consistency with findings published the year before by Atkinson (Reference Atkinson1982). These findings indicate that happiness at baseline accounted for more of its variance in subsequent assessment than relationships with demographics, life domain satisfactions, and activity indicators, whether analyzed singly or in combination. This outcome surprised us given the widely accepted eighteenth- and nineteenth-century utilitarian philosophy that happiness is a consequence of life circumstance. Although the latter can have lasting effects, the smallness of the associated effect sizes relative to the stability of happiness made us realize the importance of life dispositions for theorizing about happiness. The problem was to explain why happiness is stable.
The answer offered in the book is that happiness is a propensity; an idea proposed half a decade earlier on the basis of relationships between happiness and activities (Stones & Kozma, Reference Stones and Kozma1986a). The findings from that study are consistent with a circumplex model of affect, with axes of hedonic evaluation and level of activity (Russell, Reference Russell1980). We suggested that happiness contains a stable long-term component around which short-term happiness (e.g., mood) fluctuates (Kozma, Stone, Stones, Hannah, & McNeil, Reference Kozma, Stone, Stones, Hannah and McNeil1990). Similar models use homeostatic terminology with the stable component termed a set point. Although set point theory retains popularity, it begs the question of what sets the set point (Heady, Reference Heady2006). Whereas personality traits seem obvious candidates, evidence at that time suggested low explanations of variance in happiness by personality traits (e.g., ≦ 10% for neuroticism; ≦ 4% for extraversion) compared with the temporal stability of happiness (e.g., .5–.7 over six months or longer).
A breakthrough occurred 15 years later. Lykken and Tellegen (Reference Lykken and Tellegen1996) used research on twins to partition variance in happiness into genetic and environmental components. Their study heralded a series of findings that attributed approximately 50 per cent of variability in one-time happiness and 80 per cent of the variability in stable happiness to heritability. Although their famous inference that “trying to be happier is as futile as trying to be taller” (p. 198) may be overreaching, and the underlying causal pathways linking genetic effects and happiness unclear, genetically relevant studies represent an empirical milestone in happiness research. Subsequent research suggested that the causal pathway includes genetic influence on personality traits, with the latter having stronger relationships with happiness than was evident when we wrote our book.
Types of Well-being
The first chapter of the book includes a nomenclature based on historical usage of terms. The oldest and umbrella term is mental health (used since the 11th century) with components that include state of mind and competence. Most pertinent here are states of mind that we divided into categories of (1) happiness, affect, and mood (terms introduced in the 13th–15th centuries), and (2) satisfactions (introduced in the 15th century). The reason for this division is because happiness is the term that has been used for centuries, both in philosophy and colloquially, to connote affective states of mind, whereas satisfaction of desires connotes a cognitive appraisal of attainment but not necessarily an affective state of mind (Michalos, Reference Michalos1980).
Not surprisingly, subsequent usage includes other than historical precedent to classify terms. Lucas and Diener (Reference Lucas, Diener and Diener2009) blurred the distinction between cognition and affect, defining subjective well-being (SWB) as “the extent to which people think and feel that their life is going well … colloquially [termed] as happiness” (p. 75; our italics). Keyes, Shmotkin, and Ryff (Reference Keyes, Shmotkin and Ryff2002) merged affect and satisfaction within their definition of SWB and defined psychological well-being (PWB) as “perception of engagement with existential challenges of life” (p. 1007).
Current usage of PWB is problematic for reasons of content, relevance, and causal attribution. The content (i.e., types of existential engagement) usually corresponds to six scales on the Ryff (Reference Ryff1989) PWB measure: (a) autonomy, (b) environmental mastery, (c) personal growth, (d) positive relations with others, (e) purpose in life, and (f) self-acceptance. These dimensions are far removed from the affective content of the PWB scales by Bradburn (Reference Bradburn1969), who more than anyone advanced the usage of PWB within the behavioral and social sciences. Waterman et al. (Reference Waterman, Schwartz, Zamboanga, Ravert, Williams and Agocha2010) criticized the Ryff (Reference Ryff1989) scales for lack of eudaimonic relevance, which was a purported reason for their development, describing them as “an array of psychological qualities indicative of mental health” (p. 43). Kashdan, Biswas-Diener, and King (Reference Kashdan, Biswas-Diener and King2008) commented that the Ryff (Reference Ryff1989) formulation reduces opportunities for causal attribution. An example is that positive relations with others can neither precede nor ensue from PWB because social relationships are part of Ryff’s definition of PWB. This criticism, in our opinion, is indeed damaging.
The preceding discussion indicates that the passage of time has failed to provide a consensual terminology. For present purposes, we bow toward current usage and adopt the term SWB as an umbrella for measures of happiness, affect, and life satisfaction. A justification for doing so is evidence that scales bearing these names have high intercorrelations. Satisfactions pertaining to specific domains of life (e.g., housing satisfaction, marital satisfaction, satisfaction with health) we term life domain satisfactions, with affect at any moment in time termed mood. Measures of existential engagement we simply term existential engagement measures, about which there can be no controversy.
Models and Measures
The second chapter of the book reviewed approaches to model development and measures of SWB used with older people. This section describes the models and measures reviewed in 1991 and those that found favour thereafter.
Models
Five prototypic approaches to model development discussed in the book make assumptions about happiness that we classified as bottom up, top down, personality, telic, and judgmental approaches. In the bottom-up approach, we considered happiness to be an outcome of levels on lower-order variables (e.g., having more money makes people happier). In the top-down approach, happiness is assumed to affect lower-order variables (e.g., happier people report higher financial satisfaction regardless of income). With personality approaches, it is suggested that personality traits help to stabilize levels of happiness over time. In telic approaches it is suggested that happiness results from goal attainment, with overall happiness determined by the proportion of goals attained or desires satisfied. Finally, in judgmental approaches, it is proposed that happiness is an outcome of comparison between present conditions and personally relevant standards, which may include needs and aspirations in the past and future (Michalos, Reference Michalos1985).
The models described in the book are conceptual in origin and may incorporate more than one approach towards development (e.g., the judgmental model of Michalos [Reference Michalos1985] assumed bottom-up directionality). We term a class of models that developed subsequent to the book as genetic models.
Genetic models refer to greater comparability of SWB between identical twins than between fraternal twins or siblings of different ages. Lykken and Tellegen (Reference Lykken and Tellegen1996) provided an influential example, with subsequent authors attributing some of the covariation of SWB with personality traits and certain situational measures (e.g., marital status, spousal support) to effects associated with genetic endowment (Schnittker, Reference Schnittker2008; Weiss, Bates, & Luciano, Reference Weiss, Bates and Luciano2008). In principle, findings from twin designs could occur for any of three reasons: (a) direct genetic effects on SWB, personality, and situational measures; (b) genetic effects on SWB, with the latter a determinant of personality and/or life situation; or (c) genetic effects on personality and/or life situation, with implications for SWB. Consequently, we consider findings from such designs as genetically related effects.
Because the credibility of a model depends on the methodology used for its evaluation, we would be remiss not to mention advances in methodology. Much of the earlier research on happiness in older people was cross-sectional, with the use of cross-lagged panel designs a rarity (Stones & Kozma, Reference Stones and Kozma1986b). Causal modeling was in its infancy at that time, deployed mainly in a psychometric context, but subsequently used to test between models that assumed bottom-up or top-down directionality. In the following, we briefly summarize the basic assumptions.
Causation associated with the bottom-up or top-down approach is asymmetric (Diener, Reference Diener1984; Stones, Hadjistravopoulos, Tuukko, & Kozma, Reference Stones, Rattenbury, Kozma, Haight and Webster1995). Bottom-up models graph the direction of causality from variable X that correlates with SWB as, X → SWB, with the size of effect interpreted as the variance in SWB explained by X. Top-down models graph the direction from SWB to variable X (i.e., SWB → X), with the effect size interpreted as the variance in X explained by SWB. Assumptions in these models are that variable X is respectively an antecedent or outcome of SWB, with evidence of bi-directionality a complicating occurrence (e.g., a bi-directional relationship between positive affect and job performance; Lyubomirsky, King, and Diener, Reference Lyubomirsky, King and Diener2005).
Other causal models take account of higher-order confounding or lurking variables. These variables correlate with both SWB and variable X, such that variable L → SWB and L → X. The effect of variable L can render the level of correlation between SWB and variable X spurious if L has direct effects on both X and SWB. Alternatively, L might moderate the effects of X on SWB (and vice versa) or mediate their relationship. Although researchers have routinely controlled for confounding variables such as sex and other demographic factors, only recently has direct higher-order determination of X and SWB by L assumed theoretical prominence in the literature on SWB.
Measures
Measures reviewed in the book included single-item measures and multi-item scales such as the Life Satisfaction Scales (LSRS, LSI-A, LSI-B, LSI-Z), the Philadelphia Geriatric Center Morale Scale (PGC), the Affect Balance Scale (ABS) and the MUNSH. All the multi-item scales except the ABS used older people during development and validation. We discussed the reliability of the scales based on internal consistency and temporal stability, and their validities against self-ratings of happiness, proxy ratings by persons knowledgeable about the respondent, and contrasts between normative groups (e.g., persons living independently with no known psychiatric problems versus persons receiving psychiatric care). The overall findings showed psychometric statistics for the PGC and MUNSH that were more favourable than for the Life Satisfaction Scales and ABS, with the MUNSH having higher temporal stability than the PGC. We also reviewed evidence on threats to validity that included social desirability bias and response bias, with no significant threat reported (Kozma & Stones, Reference Kozma and Stones1985, Reference Kozma and Stones1987, Reference Kozma and Stones1988).
The two most widely cited measures in use today are the Satisfaction with Life Scale (SWLS; Diener, Emmons, Larsen, & Griffin, Reference Diener, Emmons, Larsen and Griffin1985) and the Ryff (Reference Ryff1989) scales. Both had adequate psychometric statistics in the validation studies, and both have had wide application in research with older people. We did not review them in the book because they lacked a sufficient pedigree for use with older people at that time. With hindsight, we realize that such an omission was overly cautious because psychometrically adequate SWB measures retain those properties irrespective of age and cultural demarcations (e.g., use of the MUNSH to measure happiness in young people in China).
Components of Well-being
The third chapter discussed structural relationships pertaining to SWB measures, including first-order and second-order factor structures, relationships between composite scores, and a distinction between short-term and long-term components of happiness.
Structural Relationships and Correlations of Composite Scores
Of the measures discussed in the book, the MUNSH has a one-dimensional structure whereas the Life Satisfaction Scales, PGC, and ABS have multidimensional structures when used with older samples. However, attempts to cross-validate the structures of the multidimensional scales were largely unsuccessful except in the case of the ABS. Other findings indicated that a second-order factor explains variance in the first-order factors of the LSI-A and PGC, thereby indicating that the primary factors have variance in common.
With respect to SWB measures not reviewed in our book, the SWLS has a one-dimensional structure in younger and older samples, whereas the Ryff (Reference Ryff1989) scales have multidimensional structures, with inter-scale correlations having a median of r = .52 and a range from .32–.76. However, subsequent studies generally failed to replicate the latent structure and factorial validity of Ryff’s scales (e.g., Abbott et al., Reference Abbott, Ploubidis, Huppert, Kuh, Wadsworth and Croudace2006; Kafka & Kozma, Reference Kafka and Kozma2002).
We reported in our book that analysis of interrelationships between composite scale scores showed an overarching single factor. Stones and Kozma (Reference Stones and Kozma1985) carried out eight structural analyses of composite scores on more than 20 SWB measures. All analyses provided single-factor solutions. Subsequent findings indicated substantial correlations of composite scores on different types of well-being measure. Examples of these correlations include SWB with the Ryff (Reference Ryff1989) composite measure (r = .59), SWB with the Questionnaire for Eudaimonic Well-Being (QEWB; r = .43), and the Ryff (Reference Ryff1989) measure with the QEWB (r = .63) (Keyes et al., Reference Keyes, Shmotkin and Ryff2002; Waterman et al., Reference Waterman, Schwartz, Zamboanga, Ravert, Williams and Agocha2010). Exploratory and confirmatory factor analyses of the Ryff (Reference Ryff1989) scales and different measures of SWB showed two oblique factors with loadings that roughly correspond with two types of well-being scale (Keyes et al., Reference Keyes, Shmotkin and Ryff2002). Because the factors are highly correlated (i.e., coefficients of .45 and .84 in different analyses), inclusion of a second-order factor would likely improve the fit of the model.
An implication of the preceding discussion is that a single overarching dimension contributes substantially to variance in scales termed SWB, existential engagements, and existential well-being. The conclusion appears inescapable that well-being has a common core of meaning irrespective of the content measured.
Short-Term and Long-Term Affective States
The book reviewed evidence on short-term and long-term affective states in people of different ages. Short-term states include positive affect, negative affect, and vigor, which are consistent with the internal structure of the MUMS. Long-term states include affect and satisfaction over prolonged durations, as measured by items on the MUNSH. Findings reported in the book and verified in subsequent research indicated that measures of short-term states are reactive to mood induction procedures whereas measures of long-term states show minimal reactivity (Kozma et al., Reference Kozma, Stone, Stones, Hannah and McNeil1990; Kozma, di Fazio, Stones, & Hannah, Reference Kozma, di Fazio, Stones and Hannah1992). The effects on short-term states were greater following negative than positive mood induction. Age differences emerged only when a positive mood induction followed a negative induction, such that the changes in short-term affect reflected that sequence more consistently in younger than older people.
Research subsequent to the book also examined the relationship between levels of happiness and the intensity of experiencing affect (Stones & Kozma, Reference Stones and Kozma1994). The findings suggested that (a) very happy people experience emotion as an induced arousal rather than a change in hedonic tone, (b) moderately happy people experience emotion more as a change in hedonic tone than arousal, and (c) very unhappy people experience emotion as a change in hedonic tone accompanied by increased arousal.
Relationships with SWB
The fourth chapter discussed relationships with variables classified as demographics, life domain satisfactions, activities, stress, environmental context, and features of personality. While correlation does not imply causation, even psychologists are prone to attribute causation in the absence of evidence from experimental or longitudinal studies. Lyubomirsky, King, et al. (Reference Lyubomirsky, King and Diener2005) noted a pervasive bias among researchers to attribute causation in a bottom-up rather than top-down direction. Here, we review conclusions in the book and findings from subsequent research.
Demographics
The book concluded that demographic variables showed only weak correlations with SWB. Subsequent reviews are consistent with that conclusion. Myers and Diener (Reference Myers and Diener1995) wrote about happiness and satisfaction that “knowing a person’s age, sex, race, and income (assuming the person has enough to afford life’s necessities) hardly gives a clue” (p. 17). Lyubomirsky, Sheldon, and Schkade (Reference Lyubomirsky, Sheldon and Schkade2005) estimated that life circumstance explains about 10 per cent of the variance in SWB. Waterman et al. (Reference Waterman, Schwartz, Zamboanga, Ravert, Williams and Agocha2010) reported low relationships of eudemonic well-being with demographic indicators.
Researchers almost uniformly interpret demographic relationships as examples of bottom-up causation. However, Lyubomirsky, King, et al. (Reference Lyubomirsky, Sheldon and Schkade2005) described longitudinal evidence for top-down causation that includes retained employment, income advancement, marriage, and satisfaction with marriage. Based on comparison of unrelated people, ordinary siblings, and identical twins, Schnittker (Reference Schnittker2008) concluded that genetically related effects contribute to the relationship of SWB with marital status and spousal support, whereas belonging to the same family contributes to the relationship of SWB with socioeconomic indicators such as years of education and income.
Life Domain Satisfactions
We reported in the book that correlations of SWB and life domain satisfactions are higher than between SWB and objective status on those life domains. With respect to health, housing, financial, marital relationship, and employment, the variance in SWB accounted for by life domain satisfactions was approximately three times higher than explained by corresponding objective indicators. Models at that time generally interpreted these relationships as examples of bottom-up directionality (Michalos, Reference Michalos1985) – for instance, relationships pertaining to economic indicators are from income → financial satisfaction → SWB. An exception was a study by Stones and Kozma (Reference Stones and Kozma1986b) that suggested top-down directionality between SWB and life domain satisfactions in a cross-lagged panel design with older people.
Subsequent studies confirmed earlier findings of higher correlations of SWB with life domain satisfactions than with objective indicators corresponding to the same domains. An example from South Africa showed such findings in racial groups living under first-world and third-world conditions (Møller & Saris, Reference Møller and Saris2001). However, studies that examined directionality in the relationship between SWB and life domain satisfactions generally provided support for top-down rather than bottom-up directionality (i.e., SWB → life domain satisfaction; Kozma, Stone, & Stones, Reference Kozma, Stone and Stones1999; Lance, Mallard, & Michalos, Reference Lance, Mallard and Michalos1995; Møller & Saris, Reference Møller and Saris2001).
An exception is a study by Clyburn, Stones, Hadjistavropoulos, and Tuokko (Reference Clyburn, Stones, Hadjistavropoulos and Tuokko2000) that compared bottom-up with top-down directionality in caregivers of patients with Alzheimer’s disease. The measures included objective stressors (e.g., disturbing symptoms by Alzheimer patients; tangible support for the caregiver), and inverse measures of life domain satisfaction (caregiver burden) and SWB (self-rated depression). The best-fitting model had a causal chain from stressors → burden → depression. Because giving care to a family member with Alzheimer’s disease is highly stressful, these findings suggest that top-down directionality may not apply near the negative extreme of an objective distribution associated with high stress.
Intentional Activities
The conclusion in the book was that higher levels of social and physical activity confer moderate benefit to the SWB of older people. Subsequent research was supportive of relationships between intentional activities and SWB.
Positive psychology studies have provided such evidence with respect to positive and negative cognitions, acts of kindness, and expressions of gratitude (Lyubomirsky, Sousa, & Dickerhoof, Reference Lyubomirsky, Sousa and Dickerhoof2006; Sheldon & Lyubomirsky, Reference Sheldon and Lyubomirsky2006). Lyubomirsky et al. (Reference Lyubomirsky, Sousa and Dickerhoof2006) reviewed evidence from longitudinal and experimental studies showing that happier people are more likely to exhibit positive intentional behaviors related to altruism that affect social relationships. Although these findings support top-down directionality, Sin and Lyubomirsky (Reference Sin and Lyubomirsky2009) provided evidence for bottom-up directionality in a meta-analysis of 51 intervention studies of intentional activities intended to promote positive feelings, positive behaviors, or positive cognitions.
The median effect sizes in the Sin and Lyubomirsky (Reference Sin and Lyubomirsky2009) study were r = .24 for gains in SWB and r = .26 for alleviation of depression. The effect sizes were higher for individuals who (a) were more depressed at baseline, (b) elected to participate in intervention, (c) were older, and (d) received individualized intervention. With respect to age, the median effect size was near zero in children and adolescents, .25 in young adults, .40 in middle-aged people, and .51 in older adults. These finding suggest strong bottom-up effects of intervention particularly in older people who were more depressed at baseline.
Stress
Evidence reviewed in the book suggested that major life events, whether favourable or stressful, have effects on the enduring stability of SWB that are surprisingly weak (Atkinson, Reference Atkinson1982; Costa, McCrae, & Zonderman, Reference Costa, McCrae and Zonderman1987). People adjust to misfortune, and even positive stress, like winning a lottery, does not have long-lasting effects on happiness (Brickman, Coates, & Janoff-Bulman, Reference Brickman, Coates and Janoff-Bulman1978).
Subsequent evidence suggested that everyday hassles have a stronger relationship with happiness than major life change (Chamberlain & Zika, Reference Chamberlain and Zika1992). However, a review by Lucas and Diener (Reference Lucas, Diener and Diener2009) provided evidence that some life changes, such as acquired disability, have long-lasting effects on SWB.
The Environment
Evidence reviewed in the book indicated low correlations between SWB and environmental variables (e.g., community size, age concentration, perceived safety). An exception is residence in long-term care homes where SWB was much lower than for individuals living outside of institutions. However, residents in institutions have higher levels of disease and disability that may exacerbate the restrictive effects of an institutional environment.
Research after publication of the book included comparisons of SWB in different countries. Much of that interest revolved around economic issues. Veenhoven and Hagerty (Reference Veenhoven and Hagerty2006) argued that national SWB levels increased during the past half-century as a result of evolving economies. However, the evidence they provided suggests relatively minor changes. In the United States, for example, they reported that it would take 167 years to obtain a 1-point gain on a 10-point happiness scale. In what they term “developing” nations during the period 1981–2000/2001, Japan, Mexico, and South Africa showed gains on a comparable measure of .53–.96 points. Although South Korea showed no such gain during this period, SWB increased during subsequent decades. However, social and political changes concomitant with economic growth in those countries have made it debatable as to whether the changes in SWB owe simply to increases in income.
Characteristics Related to Personality
The book reviewed evidence that traits such as extraversion and neuroticism respectively account for approximately four per cent and 10 per cent of variance in SWB scores. Subsequent research using well-established measures of personality obtained stronger relationships with measures of SWB differentiated into components of positive and negative affect. A meta-analysis of such studies found correlations of .25–.44 between positive affect and measures of extraversion and .46–.54 between negative affect and measures of neuroticism (Steel, Schmidt, & Shultz, Reference Steel, Schmidt and Shultz2008). These findings suggest that such traits have stronger relationships to SWB than reported in the book.
Two bodies of literature that developed subsequent to publication of the book emphasize eudemonic principles of flourishing and structural relationships among personality traits. Park, Peterson, and Seligman (Reference Park, Peterson and Seligman2004) illustrated the former in an examination of relationships between SWB and character strengths. The authors defined character strength as a disposition that affects behaviours, desires, and feelings in ways likely to promote excellence and flourishing. Their findings regarding 24 character strengths showed correlations with life satisfaction at levels from near zero to approximately r = .6. Some of these strengths, they admitted, have meaning that overlaps with SWB (e.g., zest). They also found that control for effects of the “Big Five” personality traits (i.e., neuroticism, extraversion, conscientiousness, agreeableness, and openness to experience) lowered but did not eliminate covariance between SWB and character strengths, thereby suggesting that the reported relationships are not fully attributable to personality. Park et al. (Reference Park, Peterson and Seligman2004) did not speculate about direction of causality, because of their cross-sectional design, but gently reminded readers of Aristotle’s proposition The Nicomachean ethics that happiness is not a consequence of virtuous action but “inherent” in such action.
With respect to the structure of personality, Musek (Reference Musek2007) reported a higher-order factor claimed to subsume not only the Big Five personality traits but also measures related to happiness. Rushton, Bons, and Hur (Reference Rushton, Bons and Hur2008) termed this factor the “General Factor of Personality” (GFP). Rushton and Irwing (Reference Rushton, Irwing, Chamorro-Premuzic, von Stumm and Furnham2011) reviewed subsequent studies that link the GFP with SWB and related measures (e.g., self-esteem, depression). However, the status of the GPF is controversial within personality research, with many concerns expressed in a publication by Ferguson, Chamorro-Premuzic, Pickering, and Weiss (Reference Ferguson, Chamorro-Premuzic, Pickering, Weiss, Chamorro-Premuzic, von Stumm and Furnham2011).
Genetically Related Effects
No research on genetic influences on SWB was available when the book was published. Evidence is now available from at least nine studies of monozygotic (MZ) and dizygotic (DZ) twins. These finding suggest that genetically related effects do the following: (a) account for approximately 50 per cent of variance in one-time assessment of SWB; and (b) contribute directly or indirectly to covariance of SWB with personality traits, life domain satisfactions, existential engagements, and demographic measures. The age range encompassed by the studies is 2–92 years.
A number of studies have shown genetically related effects on SWB. Lykken and Tellegen (Reference Lykken and Tellegen1996) studied 2,310 individuals, finding that genetic effects explained approximately 50 per cent of total variability in happiness scores and 80 per cent of the variability that is stable over time. Johnson, McGue, and Krueger (Reference Johnson, McGue and Krueger2005) studied 833 twins with mean ages of 59.4 and 64.4 years at two times of measurement. At the first and second measurements, they reported genetically related effects of 27 per cent and 31 per cent on a well-being measure, 42 per cent and 41 per cent on a higher-order measure of positive emotion, and 33 per cent and 45 per cent on a higher-order measure of negative emotion. Stubbe, Posthuma, Boomsma, & De Geus (Reference Stubbe, Posthuma, Boomsma and De Geus2005) reported genetically related effects that explain 38 per cent of the variance in life satisfaction in a study of 5,668 individuals.
Nes, Røysamb, Tambs, Harris, and Reichborn-Kjennerud (Reference Nes, Røysamb, Tambs, Harris and Reichborn-Kjennerud2006) studied genetically related effects on six-year cross-time correlations in SWB. They reported that additive genetic effects explain approximately 80 per cent of the stable variance in SWB. Weiss et al. (Reference Weiss, Bates and Luciano2008) examined life satisfaction in 973 twin pairs. They reported a heritability estimate of 54 per cent. Bartels and Boomsma (Reference Bartels and Boomsma2009) reported overall heritability estimates of 40 to 50 per cent for measures of life satisfaction, happiness, and quality of life in a study of 5,024 individuals. In a study on 3,023 unrelated individuals, 2,330 siblings, and 634 identical twins, Schnittker (Reference Schnittker2008) reported that genetically related effects account for 48 per cent of variance in happiness. Caprara et al. (Reference Caprara, Fagnani, Alessandri, Steca, Gigantesco and Sforza2009) used data on 692 twins and reported a 59 per cent contribution by heritability to life satisfaction. Finally, Rushton et al. (Reference Rushton, Bons, Ando, Hur, Irwing and Vernon2009) reported a 41per cent contribution by genetic effects to variance in well-being in 316 pairs of twins aged 17–74. Overall, the preceding findings suggest that genetically related effects explain approximately 40 to 60 per cent of the variance in SWB at a single time of measurement.
The following studies found that genetically related effects contribute to covariance between SWB and other variables (e.g., personality traits, specific satisfactions, existential engagement, and demographic measures). Bartels and Boomsma (Reference Bartels and Boomsma2009) found that covariance among four measures of quality of life, life satisfaction, and happiness was explained in part by additive and non-additive genetic factors. Caprara et al. (Reference Caprara, Fagnani, Alessandri, Steca, Gigantesco and Sforza2009) reported that genes influencing self-esteem, life satisfaction, and optimism largely overlap. Weiss et al. (Reference Weiss, Bates and Luciano2008) found common genetically related effects on SWB and the Big Five personality traits, with unique genetic influences on SWB from neuroticism, extraversion, and, to a lesser extent, conscientiousness. The third study reported by Rushton et al. (Reference Rushton, Bons, Ando, Hur, Irwing and Vernon2009) included the Big Five personality traits and existential-engagement measures related to humor and emotional intelligence. Their conclusions were that the same genetic effects that influence personality traits also contribute to variance in well-being and existential engagement measures. Schnittker (Reference Schnittker2008) found that genetic effects related to happiness also contributed to the covariance of happiness with marriage, spousal support, and wealth. An inference from the preceding findings is that heritability accounts for covariance between SWB and other variables.
A Model and its Applications
The book’s final chapter presented a model of SWB and discussed applications of that model. The model included three components: (a) a frequency distribution of positive and negative affective experiences; (b) the intensity of affective episodes; and (c) the fading of affect intensity over time. Although the former components are open to environmental influence, the rate of fading in a person parameter. The model addressed six factors: (a) the relationship between mood and longer-term SWB, (b) the reactivity of mood to external influences, (c) temporal stability of SWB, (d) relationships of SWB with life domain satisfactions, (e) low SWB in residents of long-term care homes, and (f) the effectiveness of interventions to promote SWB in older people.
Since publication of the book, theorizing has drifted from micro-level models toward macro-level formulations that favour eudaimonic and existential engagement notions about well-being. The following example of intervention to promote SWB in long-term care home residents illustrates this theoretical divide.
How to remediate low SWB in long-term care home residents was a prime motivation for our research. Our theoretical model anticipated low SWB because an institutional environment provides residents with relatively few happy moments. Eudaimonic theorists alternatively suggested that the lifestyle of residents tends to be unfulfilling because of loss of autonomy (Kasser & Ryan, Reference Kasser and Ryan1999). Social facilitation is one means to raise an affective distribution and promote autonomy. Consequently, findings from controlled studies that reminiscence groups showed significant gains in happiness are unsurprising from either perspective (Rattenbury & Stones, Reference Rattenbury and Stones1989; Stones, Rattenbury, & Kozma, Reference Stones, Rattenbury, Kozma, Haight and Webster1995). Moreover, meta-analyses of reminiscence studies found effect sizes comparable to those obtained with pharmacotherapy and psychological interventions aimed at alleviating depression (Bohlmeijer, Smit, & Cuijpers, Reference Bohlmeijer, Smit and Cuijpers2003). The mean effect size of .54 in such studies (Bohlmeijer, Roemer, Cuijpers, & Smit, Reference Bohlmeijer, Roemer, Cuijpers and Smit2007) is substantially higher than Sin and Lyubomirsky (Reference Sin and Lyubomirsky2009) reported for interventions intended to promote the well-being of younger people.
The explanatory models differ, however, when attempting to predict profiles of change in SWB following intervention. The prediction from our model is for gains in SWB that are independent of its baseline level. The eudaimonic model predicts higher intervention effects for residents more depressed at baseline because of low autonomy. Analysis or reanalysis of data from successful reminiscence interventions provided evidence to support the former prediction (Stones, Rattenbury, et al., Reference Stones, Rattenbury, Kozma, Haight and Webster1995; Stones, Hadjistravopoulos, et al., Reference Stones, Rattenbury, Kozma, Haight and Webster1995).
The purpose of the preceding example is not to garner support for a particular model but, rather, to reiterate a plea for precise models that encompass parameters related to individual differences. In their reconsideration of the happiness literature, Kashdan et al. (Reference Kashdan, Biswas-Diener and King2008) reasoned that concepts cast at a high level of abstraction may hinder rather than help scientific progress. We agree, having made the same argument 20 years before, but note that many researchers on happiness appear to have taken little heed.
Discussion
The major advances since we wrote the book concern personality and genetically related effects. Weiss et al. (Reference Weiss, Bates and Luciano2008) concluded that the most important finding in their study was that “subjective well-being was genetically indistinct from personality traits” (p. 209). They went on to suggest that “the relationship between subjective well-being and a range of health and social-relationship factors may also be mediated by common genetic effects” (p. 209). We now have preliminary evidence that they were correct, as evidenced by studies that controlled for personality when examining relationships of SWB with subjective satisfactions, existential engagement, expectations of reward and punishment, behavioral activation and inhibition, and demographics (Rushton & Irwing, Reference Rushton, Irwing, Chamorro-Premuzic, von Stumm and Furnham2011).
The preceding findings supplement earlier evidence of asymmetric causation between SWB and its correlates. The causal model advocated by Weiss et al. (Reference Weiss, Bates and Luciano2008) proceeds from genetically related effects on personality to correlations between SWB and other variables (i.e., genetically related effects → personality → SWB and its correlates). The studies that contribute to this depiction encompass the full human life span: Rushton et al. (Reference Rushton, Bons and Hur2008) examined genetic effects on personality from as early as 2–9 years; Johnson et al. (Reference Johnson, McGue and Krueger2005) studied the same issue using a design that included twins aged as old as 92.
We should be careful, however, not to overgeneralize from this depiction. Not all bivariate relationships pertaining to SWB are due to genetically related effects. Examples include studies that found no genetic influence on relationships of SWB with occupational mastery and support from friends (Schnittker, Reference Schnittker2008). Similarly, relationships of SWB with variables such as character strengths remain significant after control for personality traits (Park et al., Reference Park, Peterson and Seligman2004). Moreover, numerous controlled studies show positive outcomes after intervention to promote SWB, thereby illustrating bottom-up causation. Although estimates that attribute “up to 40 per cent” of the variance in SWB to intentional activities (Lyubomirsky, Sheldon, et al., Reference Lyubomirsky, King and Diener2005, p. 116) we consider to be overly generous outside the context of institutional residence, they have no doubt that non-genetic effects do make a substantial contribution to SWB.
An inference appears reasonable, however, that many relationships usually interpreted as examples of bottom-up causation have some grounding in personality. Although counselors may advocate volunteering as a means to promote well-being, people scoring higher on the trait of agreeableness more frequently volunteer (Carlo, Okun, Knight, & Guzman, Reference Carlo, Okun, Knight and de Guzman2005). People with a more neurotic temperament are less likely to find happiness in marriage, given evidence on frequencies of divorce (Rodrigues, Hall, & Fincham, Reference Rodrigues, Hall, Fincham, Fine and Harvey2009). Even controlled studies of intervention effectiveness may not be immune to the effects of personality on adherence and compliance (Cohen, Ross, Bagby, Farvolden, & Kennedy, Reference Cohen, Ross, Bagby, Farvolden and Kennedy2004). Such findings have helped to change our thinking about happiness since 1991.
At a time just before the Big Five Factor Inventory was accessible in a fully developed form, we thought that the stable component (as measured by happiness scales like the MUNSH) was a propensity weakly associated with other personality dispositions. Subsequent evidence suggests that happiness is more strongly integrated within genetically influenced personality structures. As to the future, we consider research on genetic-environment correlations – how heritable dispositions influence the ways in which different people shape their environments and react to the changes engendered – to be a recommended avenue to advance our understanding of happiness over the lifespan.