- Email: [email protected]
Endowment and contrast: the role of positive and negative emotions on well-being appraisal
Personality and Individual Differences 37 (2004) 905–915 www.elsevier.com/locate/paid
Endowment and contrast: the role of positive and negative emotions on well-being appraisal Sheung-Tak Cheng
*
Department of Applied Social Studies, City University of Hong Kong, 83 Tat Chee Avenue, Kowloon, Hong Kong Received 29 April 2003; accepted 24 October 2003 Available online 11 December 2003
Abstract Linear additive models of affect and subjective well-being do not account for why people can uphold their sense of well-being during adversity. This paper presents a multiplicative model––the Affective Endowment-Contrast Theory, and reports a prospective study testing the theory. Endowment means that the accumulation of positive (PA) and negative affect (NA) has direct effects on well-being. Contrast, represented by the product term PA · NA, means that the effect of PA is stronger when NA is high than when NA is low. The effect of contrast thus partly compensates the main effect of NA during difficult times. Seventy-one university students completed affect diaries over a 4-week period. The results provided support for the well-being repair function of contrast. 2003 Elsevier Ltd. All rights reserved. Keywords: Happiness; Life satisfaction; Well-being; Affect; Endowment; Contrast
1. Introduction The affect balance model (Bradburn, 1969) assumes that subjective well-being (SWB) is a function of the difference between positive affect (PA) and negative affect (NA). It assumes that PA and NA contribute linearly and independently to SWB. This paper proposes a theory that extends the affect balance model to take into account the positive function of NA. Before I go on to talk about the theory, a clarification of terminology is necessary. The meaning of the term SWB is ambiguous in the literature. Sometimes, it is used to mean PA minus NA (Bradburn, 1969), sometimes life satisfaction alone (Neugarten, Havighurst, & Tobin, *
Tel.: +852-2788-8745; fax: +852-2788-8960. E-mail address: [email protected] (S.-T. Cheng).
0191-8869/$ - see front matter 2003 Elsevier Ltd. All rights reserved. doi:10.1016/j.paid.2003.10.013
906
S.-T. Cheng / Personality and Individual Differences 37 (2004) 905–915
1961), sometimes a combination of life satisfaction, PA and (lack of) NA (Diener, 2000; Lu et al., 2001), and sometimes broadened to include other positive cognitions, vigor and mental alertness (Hills & Argyle, 2002; Røysamb, Harris, Magnus, Vittersø, & Tambs, 2002). In this paper, I use the term well-being appraisal (WBA) to refer to an overall evaluation of life as a whole. Questions pertaining to global happiness and satisfaction (Andrews & Withey, 1976; Campbell, Converse, & Rodgers, 1976; Cantril, 1965; Diener, Emmons, Larsen, & Griffin, 1985) fall into this domain. 1.1. Limitations of a linear model of affect and well-being appraisal What is most remarkable about human beings is our ability to adapt under the most difficult circumstances (Seligman & Csikszentmihalyi, 2000). In this connection, a linear additive model that is commonly assumed in the field may be too simplistic. Such a model does not explain why individuals undergoing hardship can still maintain a reasonably high sense of well-being (e.g., Brickman, Coates, & Janoff-Bulman, 1978; Miller, Manne, Taylor, & Keates, 1996). Two psychological processes were suggested by Tversky and Griffin (1991) to explain the relationship between events and well-being (see also Schwarz & Strack, 2000). According to Tversky and Griffin, the same event has both an endowment and a contrast effect on well-being. As for endowment, a positive event raises oneÕs well-being over the long haul, and a negative event reduces it. As for contrast, an earlier positive event reduces oneÕs well-being due to the diminished utility of subsequent events, whereas an earlier negative event raises it. In one of the experiments reported in Tversky and Griffin (1991), university students read stories told by fictitious high school students. Two stories were provided, one about an event that had happened the week before, the other happened on the day of storytelling. The past event varied in terms of valence, whereas the present event, fixed neutral, varied in terms of whether it was related to the former or not, thus creating a 2 (valence of past event) · 2 (relatedness of the two events) between subjects design. They were asked to rate the happiness of the fictitious person on an 11-point scale. Results showed (a) a significant main effect for valence, meaning that the happiness rating was lower when the past event was negative than when it was positive, and vice versa (endowment effect), and (b) a significant interaction effect for valence · relatedness, meaning that a related negative event in the past produced higher happiness ratings, whereas a related positive event produced lower ratings (contrast effect). Tversky and Griffin (1991) further argued that other experiments designed to show moodÕs effect on WBA through mood-induction procedures could be interpreted within this endowment-contrast framework. Hence, endowment and contrast work in opposite directions, and while the affect balance model takes well into account the endowment effects, it ignores the possibility of contrasts. If the effect of a present event depends partially on its relation to past events, independent of its valence, then well-being cannot be represented by a simple subtraction equation. Suppose persons A and B each faces the same current events with the same outcomes. Suppose also that both had prior experiences with these events. Person AÕs past experiences were largely negative, whereas person BÕs were largely positive or neutral. According to the affect balance equation, everything else being the same, person B is predicted to have a higher sense of well-being than person A. However, according to the endowment-contrast paradigm, person B is not necessarily better off. If person BÕs prior experiences were largely positive, which one currently enjoys higher well-being depends on the relative size of the endowment effect (in person BÕs case) and the contrast effect (in person
S.-T. Cheng / Personality and Individual Differences 37 (2004) 905–915
907
AÕs case). If person BÕs prior experiences were largely neutral, then he/she is predicted to have a lower sense of well-being because of the contrast effect augmenting present well-being in person A. The presence of contrast effect has far-reaching implications for a psychology of well-being–– suffering can enhance happiness. One point requires clarification. Does contrast simply change the way we experience things, or does it also change the way judgments of well-being are made? There is no doubt that an earlier event would affect our experience with subsequent events, but this contrast effect would have been absorbed into, and hence completely redundant with, the endowment effects of all subsequent events. What Tversky and Griffin have shown is that contrast has an evaluative function with regards to overall well-being as well as an experiential function. It is this evaluative function that is the focus of this article. 1.2. The affective endowment-contrast theory Despite Tversky and GriffinÕs perceptive analysis, their framework has received very little attention in the literature. Data concerning the theory are limited to those obtained under experimental conditions in which participants were drawn to attend to the difference between two events, whereas naturalistic research is absent to the authorÕs knowledge. Can these laboratory findings be generalized to real life? Do people spontaneously engage in contrast in daily life? It should be noted that in naturalistic research, endowment and contrast effects in the same event are difficult to disentangle, since the effects of all events are mixed up together in oneÕs WBA. Nonetheless, the notions of endowment and contrast can be extended to naturalistic research with a theoretical twist. Since all events are coded in affective tone (Zajonc, 1980), the theory can be reformulated as follows: PA and NA have endowment and contrast effects on well-being. In this revised theory, the endowment effect means that the accumulation of PA and NA have direct impact on well-being. The contrast effect is predicated on the assumption that our total affective experience is larger than the sum of positive, negative and neutral moments. This means that the mix of PA and NA gives meaning to our quality of life that is beyond a simple affect balance equation. A powerful example is that a person whose PA and NA each occupies 50% of the time will probably report very different well-being than someone who spends 50% of the time in neutrality and the rest in equal amounts of PA and NA, even though their affect balance scores are the same. In periods when PA is prevalent and NA is rare, we tend to feel good. But when NA is prevalent, a modest amount of PA may be enough to restore us to a minimal sense of well-being, even though there may be less PA than NA. In fact, the contrast effect at hard times can offset the endowment effect of NA, thus helping to maintain our sense of well-being despite adversity. Finally, when both PA and NA are infrequent, life may become so uninteresting that ill-being would result, even though PA and NA are equally prevalent. What distinguishes this theory from the affect balance approach is that PAÕs effect on WBA is hypothesized to depend on the level of NA. It therefore specifies a multiplicative rather than a linear model. That is, the same amount of change in PA should result in more changes in WBA when NA is high than when NA is low. Because NA is typically low in our daily lives (Watson, Wiese, Vaidya, & Tellegen, 1999), another way of saying this is: A positive experience which has little or no effect ordinarily would go a long way at hard times. On the contrary, when life is smooth, the effect of adding more positive experiences may just be marginal.
908
S.-T. Cheng / Personality and Individual Differences 37 (2004) 905–915
Given the theoryÕs ‘‘affective loading,’’ I have named it the affective endowment-contrast theory. Statistically, endowment can be tested as main effects, and contrast as interaction effects in a multiple regression. Specifically, endowment effects predict a positive and significant regression coefficient for PA, and a negative and significant regression coefficient for NA. As for the contrast effect, an interaction term created by multiplying PA by NA should be significant after controlling for the main effects, and the direction of association should be positive, meaning that well-being is higher when both PA and NA are frequent, and lower when both are infrequent. The theory can thus be written as follows: WBA ¼ (PA ) NA) + (PA · NA). It can be seen from this formula that when the value of either PA or NA is small, the contrast term is negligible and WBA becomes primarily a function of endowment. However, when both PA and NA are high, the contrast effect will act to compensate for NAÕs endowment, theoretically by augmenting the effect of PA. This theory differs from Tversky and GriffinÕs in three fundamental ways. First, whereas for Tversky and Griffin the object of analysis is the event, in the new theory it is the affective response to the event. More importantly, affect is treated as a currency to experiences as money is a currency to commodities. The PA generated from different experiences is seen as qualitatively the same but quantitatively different, same for NA. Second, whereas in Tversky and GriffinÕs analysis the standard for comparison is always events in the past, in the new theory it can be both present and past events. For instance, a negative current event may help one re-appreciate past fortunes, and as a result enhances the overall sense of well-being. The renowned Archbishop Tutu, recently commenting on his cancerous state, said that the terminal cancer helped him rediscover the various nice things in life, including the wifeÕs love, BeethovenÕs music, rosesÕ dew, and grandchildrenÕs smiles (Apple Daily, April 16, 2001). In a special feature program, New York City Firemen (TVB Pearl, November 24, 2001), a fireman recounted, after 911, how ‘‘happy it was to come to work everyday (before the incident)’’ (italics mine). In fact, events like the 911 provide a standard against which both past and future events are compared. Hence, no assumption is made concerning the direction of comparison. Third, for Tversky and Griffin contrast can lower or raise well-being. In the new theory, contrast is biased toward raising well-being. There are three reasons for this view: (a) it is closer to commonsense observations, (b) it fits a broad social psychological literature on positive bias (e.g., Taylor & Brown, 1988), and (c) it allows specification of statistical models for testing. Although in reality both positive and negative contrasts exist (see Brickman et al., 1978), the net contrast effect is hypothesized to be positive. If the net contrast effect is allowed to go either way, any finding concerning the interaction term (e.g., significant or nonsignificant, positive or negative) would be consistent with the theory, and hence no sensible interpretation of the data is possible. This study is an empirical test of the affective endowment-contrast theory based on the diary reports of affect of university students.
2. Method Seventy-five undergraduates were recruited to fill out affect diaries over a 4-week interval for a nominal pay. At the conclusion of the 4-week period, they provided rating on the Well-being
S.-T. Cheng / Personality and Individual Differences 37 (2004) 905–915
909
Appraisal Index (see below). 1 Four participants failed to complete the study, leaving 71 participants (17 males and 54 females; mean age ¼ 22.0, SD ¼ 1.40) for the data analysis. The diaries were as follows. Twice each day the participants rated their affect in the past hour by filling out the Chinese Affect Scale which contains 10 PA and 10 NA items (Hamid & Cheng, 1996). The PA adjectives are: content, exuberant, excited, agreeable, happy, meaningful, joyful, comfortable, relaxed, and peaceful, whereas the NA adjectives are: sad, helpless, frightened, disappointed, bitter, tense, insecure, exhausted, annoyed, and depressed. The questions were printed in four booklets, one for each week, to be carried by the participants. Each affect was rated on a 5-point scale from 1 Ônone or littleÕ to 5 Ôa lot.Õ A time for responding, either on the hour or on the half-hour, was randomly sampled from each of the following time zones each day–– 11:00–16:30 and 17:30–22:00 h––in order to take into account the circadian rhythm of mood (Caminada & de Bruijn, 1992; Watson et al., 1999) and the daily routines of the students (classes in morning and afternoon, and social activities/studying in the evening). Data collection always started on a Monday for any participant and the completed booklets were returned on either the Monday or the Tuesday of the following week, depending on traveling schedules. All participants received a call during the first week of data collection to make sure that they had no problems following the procedure. The outcome of this phone call as well as subsequent dialogues when the weekly booklets were returned suggested that the procedure was followed rigorously. One week after the data collection was over, participants were phoned and were asked to rate themselves on the Well-being Appraisal Index. Much like the ones used by Campbell et al. (1976) and Strack, Schwarz, and Gschneidinger (1985), the Index is a composite of two questions: ‘‘All things considered, do you live in happiness?’’ and ‘‘All things considered, are you satisfied with your life?’’ Both questions were answered on a 7-point scale, from 1 Ôvery unhappy/dissatisfiedÕ to 7 Ôvery happy/satisfied,Õ with 4 being Ômixed.Õ Participants were not given prior notice of this interview so that they would not link their diaries to another anticipated assessment. The 1-week interval was set so that participants would not have a fresh memory of the information they had provided in the diaries, and thus avoided correlational inflation between WBA and affect. In a pilot study of 104 university students, this Index had an average correlation of 0.82 with the Satisfaction with Life Scale (Diener et al., 1985) over three administrations, and had 1-week, 2week and 3-week retest correlations of 0.82, 0.77 and 0.75 respectively. The CronbachÕs alpha in the present sample was 0.84. Three measures were generated from the affect diaries over the 4-week period: (a) average PA over four weeks, (b) average NA over four weeks, and (c) affect balance, by subtracting NA from PA. The measure of WBA was the average score of the two constituent items of the Index. A higher score always meant a stronger tendency of what the construct was intended to measure. In addition, the diary affect ratings were split into two halves so as to create sequential variables for PA and NA. Specifically, ratings of the first two weeks were aggregated and averaged to form
1
Originally, the study also examined the moderating role of optimism (Scheier & Carver, 1992) on the contrast-WBA relationship; that is, it was hypothesized that the contrast effect on WBA would be higher for more optimistic than for less optimistic people. This moderating relationship was not supported. Moreover, the endowment and contrast effects of affect on WBA were not changed whether optimism was included or not. Thus optimism was subsequently removed from the analysis so that the effects of affect on WBA would be more fully revealed.
910
S.-T. Cheng / Personality and Individual Differences 37 (2004) 905–915
time 1 measures for PA and NA, and those of the final two weeks formed time 2 measures of PA and NA. This allowed further probing of the specific nature of the contrast effect, as elaborated below. 3. Results 3.1. Descriptive statistics Descriptive statistics for the variables are displayed in Table 1. One notable pattern evidenced in Table 1 was the relative infrequency of NA compared with PA, and this was consistent with the observations of Watson et al. (1999). Because of this, the average affect balance score was in the positive range. 3.2. Intercorrelations of the variables The intercorrelations of the variables are presented in Table 2. As a principle, the correlations between the total and its components are not shown. From Table 2, one can see that PA and NA evidenced high stability over time, as shown by the correlations between the time 1 and the time 2 measures. Such stability is consistent with the observation that peopleÕs lives do not normally change drastically over short periods, and this is especially so with students. Table 1 Descriptive statistics WBA Mean SD
4.87 1.09
PA
NA
AB
T1
T2
Total
T1
T2
Total
2.36 0.48
2.17 0.47
2.27 0.45
1.70 0.49
1.73 0.51
1.72 0.48
0.55 0.71
Note: WBA ¼ well-being appraisal, AB ¼ affect balance; T1 ¼ weeks 1–2, T2 ¼ weeks 3–4.
Table 2 Intercorrelations WBA (1) 2 3 4 5 6 7 8
0.74 0.70 0.75 )0.45 )0.36 )0.42 0.76
PA
NA
AB (8)
T1 (2)
T2 (3)
Total (4)
T1 (5)
T2 (6)
Total (7)
0.85 – )0.19 )0.14 – –
– )0.12 )0.12 – –
– – )0.15 –
0.86 – –
– –
–
Note: WBA ¼ well-being appraisal, AB ¼ affect balance; T1 ¼ weeks 1–2, T2 ¼ weeks 3–4. Correlations at or above 0.19 were significant at the 0.05 level, one-tailed. (–) ¼ not shown.
S.-T. Cheng / Personality and Individual Differences 37 (2004) 905–915
911
Consistent with a two-factor model of affect (Watson et al., 1999), PA and NA, whether aggregated over two or four weeks, had weak correlations with each other. Furthermore, PA was more strongly related to WBA than NA. Moreover, using 4-week aggregated scores, affect balance and PA had almost the same correlation with WBA. Thus, using PA alone to predict WBA would be as good as a measure that incorporates both PA and NA, as far as linear prediction is concerned. 3.3. Multiple regression Well-being appraisal was regressed on PA and NA (aggregated over four weeks), followed by their product term (for assessing interaction effects). To reduce the problem of multicollinearity, PA and NA were standardized before forming the product term (Aiken & West, 1991). The results are shown in Table 3. One can see that PA and NA together accounted for a massive 64% of the variance in WBA, with PAÕs effect stronger than NAÕs. Furthermore, the regression coefficient for the interaction term was positive and significant, adding another 2.1% to the explained variance of WBA. This indicated that after controlling for the main effects of PA and NA, WBA was higher for those whose PA and NA were both high. Fig. 1 shows the nature of the interaction, by plotting the relationship between PA and WBA for high (those above median) and low (those below median) NA participants. It can be seen from Fig. 1 that PA had a stronger effect on WBA when NA was high than when NA was low. This pattern of results is consistent with the contrast hypothesis but not with the affect balance model. Although those high in NA never ‘‘caught up’’ with those low in NA in WBA scores, they were able to enjoy a reasonably high sense of well-being with just a moderate level of PA. The nature of the contrast effect was further probed by forming four interaction terms between the biweekly measures of PA and NA (see Table 4). According to the theory, all product terms should be significant because contrasts occur regardless of the serial order of NA and PA. Although the theory postulates that contrast is biased toward raising well-being, it does not mean that all product terms have to have a positive effect on WBA as long as the net contrast effect is positive. This can be examined by looking at the relative effect sizes (i.e., contribution to R2 ) of the product terms bearing a positive effect versus those bearing a negative effect. In order to reduce the problem of multicollinearity among the variables and to maintain reasonable statistical power
Table 3 Hierarchical regression of well-being appraisal on affect aggregated over four weeks Standardized regression coefficients NA PA NA · PA Adjusted R2 a
P < 0:001. b P < 0:05.
Step 1
Step 2
a
)0.32 0.69a
)0.27a 0.69a 0.13b
0.644
0.665
912
S.-T. Cheng / Personality and Individual Differences 37 (2004) 905–915 7
Well-being Appraisal Index
6
5
Lo NA 4
Hi NA
3
2
PA
Fig. 1. Relationship between PA and well-being appraisal in high (above median) and low (below median) NA conditions.
Table 4 Specific contrast effects over time: four regression models on well-being appraisal Predictors
Standardized regression coefficients by model 1
1. 2. 3. 4. 5. 6. 7. 8.
T1NA T1PA T2NA T2PA T1NA · T1PA T1NA · T2PA T1PA · T2NA T2NA · T2PA
Contribution to adjusted R2 by product term
2
3
4
–
)0.28 0.68b – – 0.08 – – –
)0.30 – – 0.66b – 0.13 – –
0.70b )0.25a – – – 0.14c –
– – )0.27a 0.66b – – – 0.19a
0.004
0.015
0.022
0.049
a
b
Note: T1 ¼ weeks 1–2, T2 ¼ weeks 3–4. (–) ¼ irrelevant for the specific model. a P < 0:01. b P < 0:001. c P < 0:05.
given the sample size, the four product terms were each tested in a separate regression model. All product terms were formed using standardized constituent variables. It can be seen from Table 4 that all product terms had positive effects on WBA, and two of which were statistically significant. This pattern of results lends further support to the theoryÕs
S.-T. Cheng / Personality and Individual Differences 37 (2004) 905–915
913
premise that affective contrast tends to enhance well-being and has the effect of compensating the negative main effect of NA. Both significant terms involved time-2 NA as a basis for contrast with either concurrent or restrospective PA. The term time-1 NA · time-2 PA, which on the surface is most suggestive of the contrast effect, was only marginally significant (P ¼ 0:057). It seems that contrast is most likely activated by recent negative experiences, which prompts the search for positive experiences in the present or the recent past so as to repair damage to the sense of wellbeing.
4. Discussion The findings showed that the well-being enhancement effect of PA was stronger when NA was high than when NA was low––something that was not predicted by the affect balance model but was well-predicted by the affective endowment-contrast theory. A finding that was not accounted for by the theory was the disparate endowment effects of PA and NA, and this was not entirely unexpected from the literature (Lucas, Diener, & Suh, 1996). Overall, NA had a relatively small endowment on WBA, and this makes the contrast effect especially interesting because one does not need a large contrast to compensate for the negative endowment. The small endowment effect of NA, as well as the contrast effect, implies that WBA scores should be negatively skewed in the general population. Because the main purpose of contrast is to restore our sense of well-being in the face of adversities, it belongs to a class of regulatory strategies that Heckhausen and Schulz (1995) called secondary control. Future research employing cross-cultural samples will reveal if the contrast effect is generalizable across cultures. Further analysis of contrasts across time revealed that all contrasts between NA and PA tended to have a positive effect on WBA, and that contrast is most readily activated by recent negative experiences. However, such a conclusion should be regarded as tentative, given the small sample size, the nature of the sample (i.e., university students), and the time frame and the intercorrelations of the measures. We have seen that time-1 NA · time-2 PA reached marginal significance; hence more distant negative experiences might also carry contrast effects. For example, if the study was to focus on a special sample (e.g., recovered cancer patients), and the period of study was to span from the beginning of treatment to a year into remission, this more distant (negative) experience which probably leaves a permanent mark in oneÕs life would still form a basis for contrast. Moreover, as the measures of PA were highly correlated between time 1 and time 2, same for NA, there was a fair degree of redundancy among the four regression models in Table 4. Indeed, ignoring the issue of statistical power for the moment, when the four biweekly measures of affect were entered as a block in a hierarchical regression, followed by the four product terms entered on a stepwise basis, only T2NA · T2PA met the criterion for entry, adding 2.1% of explained variance to the equation (cf. Table 3). On hindsight, had the study focused on just the last two weeks of the diary period, the size of the contrast effect found would have been much larger (from 2.1% to 4.9%; see Tables 3 and 4). Notwithstanding the limitations, the results suggest that WBA is highly dependent on PA and less on NA. Moreover, NA also serves as a basis against which PA is evaluated so that the effect of PA on WBA is further augmented. This positive bias in the way we use positive and negative
914
S.-T. Cheng / Personality and Individual Differences 37 (2004) 905–915
experiences may be related to another bias that has been reported in the literature, namely our time perspective regarding positive and negative events. People commonly believe that positive events tend to recur whereas negative events tend not to (Pyszczynski, Holt, & Greenberg, 1987). It is therefore easy for us to regard positive events as more representative of our current state of affairs than negative events. According to Schwarz and Strack (2000), this should enhance the likelihood of positive events being assimilated into our sense of well-being, and negative events serving as the standard for comparison. In this connection, it is interesting to note that depressed people believe negative events to be equally likely, if not more likely, as positive events (Peterson & Seligman, 1984; Pyszczynski et al., 1987). People who are depressed might have temporarily lost this ability to repair well-being. There is a reason why the contrast function is carried by NA rather than PA. Watson et al. (1999) observed that NA is characteristically low over time but becomes highly elevated in problematic situations, leading to a positively skewed and leptokurtic distribution within an individual. Hence when NA is noticeably increased, something must be wrong and our coping mechanism is called upon to deal with its impact on well-being. However, low PA is not necessarily associated with crisis (e.g., we may work intensively for hours and days to get something done, during which there is very little experience of PA) and hence it bears no clear-cut relationship with contrast. A final caveat is warranted before coming to a close. It should not be mistaken that happiness is best achieved by subjecting oneself to suffering. The available data from Table 3 indicate that PA and NAÕs combined endowment effect is much larger than the effect of their contrast. Thus, it is highly unlikely that one can find happiness when PA is low. However, as the size of the contrast term is in direct proportion to the amount of NA, future studies employing populations with more intensive negative experiences (e.g., life-threatening illnesses) might reveal a larger contrast effect on well-being. In any case, the data suggest that positive experiences are more valued when life is rough than when it is easy. Acknowledgements I am grateful to Jacky Cheung and two anonymous reviewers for their helpful comments on an earlier version of the manuscript.
References Aiken, L. S., & West, S. G. (1991). Multiple regression: testing and interpreting interactions. Newbury Park, CA: Sage. Andrews, F. M., & Withey, S. B. (1976). Social indicators of well-being: AmericansÕ perception of life quality. New York: Plenum. Bradburn, N. M. (1969). The structure of psychological well-being. Chicago: Aldine. Brickman, P., Coates, D., & Janoff-Bulman, R. (1978). Lottery winners and accident victims: is happiness relative. Journal of Personality and Social Psychology, 36, 917–927. Caminada, H., & de Bruijn, F. (1992). Diurnal variation, morningness–eveningness, and momentary affect. European Journal of Personality, 6, 43–69. Campbell, A., Converse, P. E., & Rodgers, W. L. (1976). The quality of American life: perceptions, evaluations, and satisfactions. New York: Russell Sage Foundation.
S.-T. Cheng / Personality and Individual Differences 37 (2004) 905–915
915
Cantril, H. (1965). The pattern of human concerns. New Brunswick, NJ: Rutgers University Press. Diener, E. (2000). Subjective well-being: the science of happiness and a proposal for a national index. American Psychologist, 55, 34–43. Diener, E., Emmons, R. A., Larsen, R. J., & Griffin, S. (1985). The satisfaction with life scale. Journal of Personality Assessment, 49, 71–75. Hamid, P. N., & Cheng, S.-T. (1996). The development and validation of an index of emotional disposition and mood state: the Chinese Affect Scale. Educational and Psychological Measurement, 56, 995–1014. Heckhausen, J., & Schulz, R. (1995). A life-span theory of control. Psychological Review, 102, 284–304. Hills, P., & Argyle, M. (2002). The Oxford Happiness Questionnaire: a compact scale for the measurement of psychological well-being. Personality and Individual Differences, 33, 1073–1082. Lucas, R. E., Diener, E., & Suh, E. (1996). Discriminant validity of well-being measures. Journal of Personality and Social Psychology, 71, 616–628. Lu, L., Gilmour, R., Kao, S.-F., Weng, T.-H., Hu, C.-H., Chern, J.-G., Huang, S.-W., & Shih, J.-B. (2001). Two ways to achieve happiness: when the East meets the West. Personality and Individual Differences, 30, 1161–1174. Miller, D. L., Manne, S. L., Taylor, K., & Keates, J. (1996). Psychological distress and well-being in advanced cancer: the effects of optimism and coping. Journal of Clinical Psychology in Medical Settings, 3, 115–130. Neugarten, B. L., Havighurst, R. J., & Tobin, S. S. (1961). The measurement of life satisfaction. Journal of Gerontology, 16, 134–143. Peterson, C., & Seligman, M. E. (1984). Causal explanations are a risk factor for depression: theory and evidence. Psychological Review, 91, 347–374. Pyszczynski, T., Holt, K., & Greenberg, J. (1987). Depression, self-focused attention, and expectancies for positive and negative future life events for self and others. Journal of Personality and Social Psychology, 52, 994–1001. Røysamb, E., Harris, J. R., Magnus, P., Vittersø, J., & Tambs, K. (2002). Subjective well-being. Sex-specific effects of genetic and environmental factors. Personality and Individual Differences, 32, 211–223. Scheier, M. F., & Carver, C. S. (1992). Effects of optimism on psychological and physical well-being: theoretical overview and empirical update. Cognitive Therapy and Research, 16, 201–228. Schwarz, N., & Strack, F. (2000). Reports of subjective well-being: judgmental processes and their methodological implications. In D. Kahneman, E. Diener, & N. Schwarz (Eds.), Well-being: the foundations of hedonic psychology (pp. 61–84). New York: Russell Sage Foundation. Seligman, M. E. P., & Csikszentmihalyi, M. (2000). Positive psychology. American Psychologist, 55, 5–14. Strack, F., Schwarz, N., & Gschneidinger, E. (1985). Happiness and reminiscing: the role of time perspective, affect, and mode of thinking. Journal of Personality and Social Psychology, 49, 1460–1469. Taylor, S. E., & Brown, J. D. (1988). Illusion and well-being: a social psychological perspective on mental health. Psychological Bulletin, 103, 193–210. Tversky, A., & Griffin, D. (1991). Endowment and contrast in judgments of well-being. In F. Strack, M. Argyle, & N. Schwarz (Eds.), Subjective well-being: an interdisciplinary perspective (pp. 101–118). Oxford: Pergamon. Watson, D., Wiese, D., Vaidya, J., & Tellegen, A. (1999). The two general activation systems of affect: structural findings, evolutionary considerations, and psychobiological evidence. Journal of Personality and Social Psychology, 76, 820–838. Zajonc, R. B. (1980). Feeling and thinking: preferences need no inferences. American Psychologist, 35, 151–175.
Endowment and contrast: the role of positive and negative emotions on well-being appraisal Sheung-Tak Cheng
*
Department of Applied Social Studies, City University of Hong Kong, 83 Tat Chee Avenue, Kowloon, Hong Kong Received 29 April 2003; accepted 24 October 2003 Available online 11 December 2003
Abstract Linear additive models of affect and subjective well-being do not account for why people can uphold their sense of well-being during adversity. This paper presents a multiplicative model––the Affective Endowment-Contrast Theory, and reports a prospective study testing the theory. Endowment means that the accumulation of positive (PA) and negative affect (NA) has direct effects on well-being. Contrast, represented by the product term PA · NA, means that the effect of PA is stronger when NA is high than when NA is low. The effect of contrast thus partly compensates the main effect of NA during difficult times. Seventy-one university students completed affect diaries over a 4-week period. The results provided support for the well-being repair function of contrast. 2003 Elsevier Ltd. All rights reserved. Keywords: Happiness; Life satisfaction; Well-being; Affect; Endowment; Contrast
1. Introduction The affect balance model (Bradburn, 1969) assumes that subjective well-being (SWB) is a function of the difference between positive affect (PA) and negative affect (NA). It assumes that PA and NA contribute linearly and independently to SWB. This paper proposes a theory that extends the affect balance model to take into account the positive function of NA. Before I go on to talk about the theory, a clarification of terminology is necessary. The meaning of the term SWB is ambiguous in the literature. Sometimes, it is used to mean PA minus NA (Bradburn, 1969), sometimes life satisfaction alone (Neugarten, Havighurst, & Tobin, *
Tel.: +852-2788-8745; fax: +852-2788-8960. E-mail address: [email protected] (S.-T. Cheng).
0191-8869/$ - see front matter 2003 Elsevier Ltd. All rights reserved. doi:10.1016/j.paid.2003.10.013
906
S.-T. Cheng / Personality and Individual Differences 37 (2004) 905–915
1961), sometimes a combination of life satisfaction, PA and (lack of) NA (Diener, 2000; Lu et al., 2001), and sometimes broadened to include other positive cognitions, vigor and mental alertness (Hills & Argyle, 2002; Røysamb, Harris, Magnus, Vittersø, & Tambs, 2002). In this paper, I use the term well-being appraisal (WBA) to refer to an overall evaluation of life as a whole. Questions pertaining to global happiness and satisfaction (Andrews & Withey, 1976; Campbell, Converse, & Rodgers, 1976; Cantril, 1965; Diener, Emmons, Larsen, & Griffin, 1985) fall into this domain. 1.1. Limitations of a linear model of affect and well-being appraisal What is most remarkable about human beings is our ability to adapt under the most difficult circumstances (Seligman & Csikszentmihalyi, 2000). In this connection, a linear additive model that is commonly assumed in the field may be too simplistic. Such a model does not explain why individuals undergoing hardship can still maintain a reasonably high sense of well-being (e.g., Brickman, Coates, & Janoff-Bulman, 1978; Miller, Manne, Taylor, & Keates, 1996). Two psychological processes were suggested by Tversky and Griffin (1991) to explain the relationship between events and well-being (see also Schwarz & Strack, 2000). According to Tversky and Griffin, the same event has both an endowment and a contrast effect on well-being. As for endowment, a positive event raises oneÕs well-being over the long haul, and a negative event reduces it. As for contrast, an earlier positive event reduces oneÕs well-being due to the diminished utility of subsequent events, whereas an earlier negative event raises it. In one of the experiments reported in Tversky and Griffin (1991), university students read stories told by fictitious high school students. Two stories were provided, one about an event that had happened the week before, the other happened on the day of storytelling. The past event varied in terms of valence, whereas the present event, fixed neutral, varied in terms of whether it was related to the former or not, thus creating a 2 (valence of past event) · 2 (relatedness of the two events) between subjects design. They were asked to rate the happiness of the fictitious person on an 11-point scale. Results showed (a) a significant main effect for valence, meaning that the happiness rating was lower when the past event was negative than when it was positive, and vice versa (endowment effect), and (b) a significant interaction effect for valence · relatedness, meaning that a related negative event in the past produced higher happiness ratings, whereas a related positive event produced lower ratings (contrast effect). Tversky and Griffin (1991) further argued that other experiments designed to show moodÕs effect on WBA through mood-induction procedures could be interpreted within this endowment-contrast framework. Hence, endowment and contrast work in opposite directions, and while the affect balance model takes well into account the endowment effects, it ignores the possibility of contrasts. If the effect of a present event depends partially on its relation to past events, independent of its valence, then well-being cannot be represented by a simple subtraction equation. Suppose persons A and B each faces the same current events with the same outcomes. Suppose also that both had prior experiences with these events. Person AÕs past experiences were largely negative, whereas person BÕs were largely positive or neutral. According to the affect balance equation, everything else being the same, person B is predicted to have a higher sense of well-being than person A. However, according to the endowment-contrast paradigm, person B is not necessarily better off. If person BÕs prior experiences were largely positive, which one currently enjoys higher well-being depends on the relative size of the endowment effect (in person BÕs case) and the contrast effect (in person
S.-T. Cheng / Personality and Individual Differences 37 (2004) 905–915
907
AÕs case). If person BÕs prior experiences were largely neutral, then he/she is predicted to have a lower sense of well-being because of the contrast effect augmenting present well-being in person A. The presence of contrast effect has far-reaching implications for a psychology of well-being–– suffering can enhance happiness. One point requires clarification. Does contrast simply change the way we experience things, or does it also change the way judgments of well-being are made? There is no doubt that an earlier event would affect our experience with subsequent events, but this contrast effect would have been absorbed into, and hence completely redundant with, the endowment effects of all subsequent events. What Tversky and Griffin have shown is that contrast has an evaluative function with regards to overall well-being as well as an experiential function. It is this evaluative function that is the focus of this article. 1.2. The affective endowment-contrast theory Despite Tversky and GriffinÕs perceptive analysis, their framework has received very little attention in the literature. Data concerning the theory are limited to those obtained under experimental conditions in which participants were drawn to attend to the difference between two events, whereas naturalistic research is absent to the authorÕs knowledge. Can these laboratory findings be generalized to real life? Do people spontaneously engage in contrast in daily life? It should be noted that in naturalistic research, endowment and contrast effects in the same event are difficult to disentangle, since the effects of all events are mixed up together in oneÕs WBA. Nonetheless, the notions of endowment and contrast can be extended to naturalistic research with a theoretical twist. Since all events are coded in affective tone (Zajonc, 1980), the theory can be reformulated as follows: PA and NA have endowment and contrast effects on well-being. In this revised theory, the endowment effect means that the accumulation of PA and NA have direct impact on well-being. The contrast effect is predicated on the assumption that our total affective experience is larger than the sum of positive, negative and neutral moments. This means that the mix of PA and NA gives meaning to our quality of life that is beyond a simple affect balance equation. A powerful example is that a person whose PA and NA each occupies 50% of the time will probably report very different well-being than someone who spends 50% of the time in neutrality and the rest in equal amounts of PA and NA, even though their affect balance scores are the same. In periods when PA is prevalent and NA is rare, we tend to feel good. But when NA is prevalent, a modest amount of PA may be enough to restore us to a minimal sense of well-being, even though there may be less PA than NA. In fact, the contrast effect at hard times can offset the endowment effect of NA, thus helping to maintain our sense of well-being despite adversity. Finally, when both PA and NA are infrequent, life may become so uninteresting that ill-being would result, even though PA and NA are equally prevalent. What distinguishes this theory from the affect balance approach is that PAÕs effect on WBA is hypothesized to depend on the level of NA. It therefore specifies a multiplicative rather than a linear model. That is, the same amount of change in PA should result in more changes in WBA when NA is high than when NA is low. Because NA is typically low in our daily lives (Watson, Wiese, Vaidya, & Tellegen, 1999), another way of saying this is: A positive experience which has little or no effect ordinarily would go a long way at hard times. On the contrary, when life is smooth, the effect of adding more positive experiences may just be marginal.
908
S.-T. Cheng / Personality and Individual Differences 37 (2004) 905–915
Given the theoryÕs ‘‘affective loading,’’ I have named it the affective endowment-contrast theory. Statistically, endowment can be tested as main effects, and contrast as interaction effects in a multiple regression. Specifically, endowment effects predict a positive and significant regression coefficient for PA, and a negative and significant regression coefficient for NA. As for the contrast effect, an interaction term created by multiplying PA by NA should be significant after controlling for the main effects, and the direction of association should be positive, meaning that well-being is higher when both PA and NA are frequent, and lower when both are infrequent. The theory can thus be written as follows: WBA ¼ (PA ) NA) + (PA · NA). It can be seen from this formula that when the value of either PA or NA is small, the contrast term is negligible and WBA becomes primarily a function of endowment. However, when both PA and NA are high, the contrast effect will act to compensate for NAÕs endowment, theoretically by augmenting the effect of PA. This theory differs from Tversky and GriffinÕs in three fundamental ways. First, whereas for Tversky and Griffin the object of analysis is the event, in the new theory it is the affective response to the event. More importantly, affect is treated as a currency to experiences as money is a currency to commodities. The PA generated from different experiences is seen as qualitatively the same but quantitatively different, same for NA. Second, whereas in Tversky and GriffinÕs analysis the standard for comparison is always events in the past, in the new theory it can be both present and past events. For instance, a negative current event may help one re-appreciate past fortunes, and as a result enhances the overall sense of well-being. The renowned Archbishop Tutu, recently commenting on his cancerous state, said that the terminal cancer helped him rediscover the various nice things in life, including the wifeÕs love, BeethovenÕs music, rosesÕ dew, and grandchildrenÕs smiles (Apple Daily, April 16, 2001). In a special feature program, New York City Firemen (TVB Pearl, November 24, 2001), a fireman recounted, after 911, how ‘‘happy it was to come to work everyday (before the incident)’’ (italics mine). In fact, events like the 911 provide a standard against which both past and future events are compared. Hence, no assumption is made concerning the direction of comparison. Third, for Tversky and Griffin contrast can lower or raise well-being. In the new theory, contrast is biased toward raising well-being. There are three reasons for this view: (a) it is closer to commonsense observations, (b) it fits a broad social psychological literature on positive bias (e.g., Taylor & Brown, 1988), and (c) it allows specification of statistical models for testing. Although in reality both positive and negative contrasts exist (see Brickman et al., 1978), the net contrast effect is hypothesized to be positive. If the net contrast effect is allowed to go either way, any finding concerning the interaction term (e.g., significant or nonsignificant, positive or negative) would be consistent with the theory, and hence no sensible interpretation of the data is possible. This study is an empirical test of the affective endowment-contrast theory based on the diary reports of affect of university students.
2. Method Seventy-five undergraduates were recruited to fill out affect diaries over a 4-week interval for a nominal pay. At the conclusion of the 4-week period, they provided rating on the Well-being
S.-T. Cheng / Personality and Individual Differences 37 (2004) 905–915
909
Appraisal Index (see below). 1 Four participants failed to complete the study, leaving 71 participants (17 males and 54 females; mean age ¼ 22.0, SD ¼ 1.40) for the data analysis. The diaries were as follows. Twice each day the participants rated their affect in the past hour by filling out the Chinese Affect Scale which contains 10 PA and 10 NA items (Hamid & Cheng, 1996). The PA adjectives are: content, exuberant, excited, agreeable, happy, meaningful, joyful, comfortable, relaxed, and peaceful, whereas the NA adjectives are: sad, helpless, frightened, disappointed, bitter, tense, insecure, exhausted, annoyed, and depressed. The questions were printed in four booklets, one for each week, to be carried by the participants. Each affect was rated on a 5-point scale from 1 Ônone or littleÕ to 5 Ôa lot.Õ A time for responding, either on the hour or on the half-hour, was randomly sampled from each of the following time zones each day–– 11:00–16:30 and 17:30–22:00 h––in order to take into account the circadian rhythm of mood (Caminada & de Bruijn, 1992; Watson et al., 1999) and the daily routines of the students (classes in morning and afternoon, and social activities/studying in the evening). Data collection always started on a Monday for any participant and the completed booklets were returned on either the Monday or the Tuesday of the following week, depending on traveling schedules. All participants received a call during the first week of data collection to make sure that they had no problems following the procedure. The outcome of this phone call as well as subsequent dialogues when the weekly booklets were returned suggested that the procedure was followed rigorously. One week after the data collection was over, participants were phoned and were asked to rate themselves on the Well-being Appraisal Index. Much like the ones used by Campbell et al. (1976) and Strack, Schwarz, and Gschneidinger (1985), the Index is a composite of two questions: ‘‘All things considered, do you live in happiness?’’ and ‘‘All things considered, are you satisfied with your life?’’ Both questions were answered on a 7-point scale, from 1 Ôvery unhappy/dissatisfiedÕ to 7 Ôvery happy/satisfied,Õ with 4 being Ômixed.Õ Participants were not given prior notice of this interview so that they would not link their diaries to another anticipated assessment. The 1-week interval was set so that participants would not have a fresh memory of the information they had provided in the diaries, and thus avoided correlational inflation between WBA and affect. In a pilot study of 104 university students, this Index had an average correlation of 0.82 with the Satisfaction with Life Scale (Diener et al., 1985) over three administrations, and had 1-week, 2week and 3-week retest correlations of 0.82, 0.77 and 0.75 respectively. The CronbachÕs alpha in the present sample was 0.84. Three measures were generated from the affect diaries over the 4-week period: (a) average PA over four weeks, (b) average NA over four weeks, and (c) affect balance, by subtracting NA from PA. The measure of WBA was the average score of the two constituent items of the Index. A higher score always meant a stronger tendency of what the construct was intended to measure. In addition, the diary affect ratings were split into two halves so as to create sequential variables for PA and NA. Specifically, ratings of the first two weeks were aggregated and averaged to form
1
Originally, the study also examined the moderating role of optimism (Scheier & Carver, 1992) on the contrast-WBA relationship; that is, it was hypothesized that the contrast effect on WBA would be higher for more optimistic than for less optimistic people. This moderating relationship was not supported. Moreover, the endowment and contrast effects of affect on WBA were not changed whether optimism was included or not. Thus optimism was subsequently removed from the analysis so that the effects of affect on WBA would be more fully revealed.
910
S.-T. Cheng / Personality and Individual Differences 37 (2004) 905–915
time 1 measures for PA and NA, and those of the final two weeks formed time 2 measures of PA and NA. This allowed further probing of the specific nature of the contrast effect, as elaborated below. 3. Results 3.1. Descriptive statistics Descriptive statistics for the variables are displayed in Table 1. One notable pattern evidenced in Table 1 was the relative infrequency of NA compared with PA, and this was consistent with the observations of Watson et al. (1999). Because of this, the average affect balance score was in the positive range. 3.2. Intercorrelations of the variables The intercorrelations of the variables are presented in Table 2. As a principle, the correlations between the total and its components are not shown. From Table 2, one can see that PA and NA evidenced high stability over time, as shown by the correlations between the time 1 and the time 2 measures. Such stability is consistent with the observation that peopleÕs lives do not normally change drastically over short periods, and this is especially so with students. Table 1 Descriptive statistics WBA Mean SD
4.87 1.09
PA
NA
AB
T1
T2
Total
T1
T2
Total
2.36 0.48
2.17 0.47
2.27 0.45
1.70 0.49
1.73 0.51
1.72 0.48
0.55 0.71
Note: WBA ¼ well-being appraisal, AB ¼ affect balance; T1 ¼ weeks 1–2, T2 ¼ weeks 3–4.
Table 2 Intercorrelations WBA (1) 2 3 4 5 6 7 8
0.74 0.70 0.75 )0.45 )0.36 )0.42 0.76
PA
NA
AB (8)
T1 (2)
T2 (3)
Total (4)
T1 (5)
T2 (6)
Total (7)
0.85 – )0.19 )0.14 – –
– )0.12 )0.12 – –
– – )0.15 –
0.86 – –
– –
–
Note: WBA ¼ well-being appraisal, AB ¼ affect balance; T1 ¼ weeks 1–2, T2 ¼ weeks 3–4. Correlations at or above 0.19 were significant at the 0.05 level, one-tailed. (–) ¼ not shown.
S.-T. Cheng / Personality and Individual Differences 37 (2004) 905–915
911
Consistent with a two-factor model of affect (Watson et al., 1999), PA and NA, whether aggregated over two or four weeks, had weak correlations with each other. Furthermore, PA was more strongly related to WBA than NA. Moreover, using 4-week aggregated scores, affect balance and PA had almost the same correlation with WBA. Thus, using PA alone to predict WBA would be as good as a measure that incorporates both PA and NA, as far as linear prediction is concerned. 3.3. Multiple regression Well-being appraisal was regressed on PA and NA (aggregated over four weeks), followed by their product term (for assessing interaction effects). To reduce the problem of multicollinearity, PA and NA were standardized before forming the product term (Aiken & West, 1991). The results are shown in Table 3. One can see that PA and NA together accounted for a massive 64% of the variance in WBA, with PAÕs effect stronger than NAÕs. Furthermore, the regression coefficient for the interaction term was positive and significant, adding another 2.1% to the explained variance of WBA. This indicated that after controlling for the main effects of PA and NA, WBA was higher for those whose PA and NA were both high. Fig. 1 shows the nature of the interaction, by plotting the relationship between PA and WBA for high (those above median) and low (those below median) NA participants. It can be seen from Fig. 1 that PA had a stronger effect on WBA when NA was high than when NA was low. This pattern of results is consistent with the contrast hypothesis but not with the affect balance model. Although those high in NA never ‘‘caught up’’ with those low in NA in WBA scores, they were able to enjoy a reasonably high sense of well-being with just a moderate level of PA. The nature of the contrast effect was further probed by forming four interaction terms between the biweekly measures of PA and NA (see Table 4). According to the theory, all product terms should be significant because contrasts occur regardless of the serial order of NA and PA. Although the theory postulates that contrast is biased toward raising well-being, it does not mean that all product terms have to have a positive effect on WBA as long as the net contrast effect is positive. This can be examined by looking at the relative effect sizes (i.e., contribution to R2 ) of the product terms bearing a positive effect versus those bearing a negative effect. In order to reduce the problem of multicollinearity among the variables and to maintain reasonable statistical power
Table 3 Hierarchical regression of well-being appraisal on affect aggregated over four weeks Standardized regression coefficients NA PA NA · PA Adjusted R2 a
P < 0:001. b P < 0:05.
Step 1
Step 2
a
)0.32 0.69a
)0.27a 0.69a 0.13b
0.644
0.665
912
S.-T. Cheng / Personality and Individual Differences 37 (2004) 905–915 7
Well-being Appraisal Index
6
5
Lo NA 4
Hi NA
3
2
PA
Fig. 1. Relationship between PA and well-being appraisal in high (above median) and low (below median) NA conditions.
Table 4 Specific contrast effects over time: four regression models on well-being appraisal Predictors
Standardized regression coefficients by model 1
1. 2. 3. 4. 5. 6. 7. 8.
T1NA T1PA T2NA T2PA T1NA · T1PA T1NA · T2PA T1PA · T2NA T2NA · T2PA
Contribution to adjusted R2 by product term
2
3
4
–
)0.28 0.68b – – 0.08 – – –
)0.30 – – 0.66b – 0.13 – –
0.70b )0.25a – – – 0.14c –
– – )0.27a 0.66b – – – 0.19a
0.004
0.015
0.022
0.049
a
b
Note: T1 ¼ weeks 1–2, T2 ¼ weeks 3–4. (–) ¼ irrelevant for the specific model. a P < 0:01. b P < 0:001. c P < 0:05.
given the sample size, the four product terms were each tested in a separate regression model. All product terms were formed using standardized constituent variables. It can be seen from Table 4 that all product terms had positive effects on WBA, and two of which were statistically significant. This pattern of results lends further support to the theoryÕs
S.-T. Cheng / Personality and Individual Differences 37 (2004) 905–915
913
premise that affective contrast tends to enhance well-being and has the effect of compensating the negative main effect of NA. Both significant terms involved time-2 NA as a basis for contrast with either concurrent or restrospective PA. The term time-1 NA · time-2 PA, which on the surface is most suggestive of the contrast effect, was only marginally significant (P ¼ 0:057). It seems that contrast is most likely activated by recent negative experiences, which prompts the search for positive experiences in the present or the recent past so as to repair damage to the sense of wellbeing.
4. Discussion The findings showed that the well-being enhancement effect of PA was stronger when NA was high than when NA was low––something that was not predicted by the affect balance model but was well-predicted by the affective endowment-contrast theory. A finding that was not accounted for by the theory was the disparate endowment effects of PA and NA, and this was not entirely unexpected from the literature (Lucas, Diener, & Suh, 1996). Overall, NA had a relatively small endowment on WBA, and this makes the contrast effect especially interesting because one does not need a large contrast to compensate for the negative endowment. The small endowment effect of NA, as well as the contrast effect, implies that WBA scores should be negatively skewed in the general population. Because the main purpose of contrast is to restore our sense of well-being in the face of adversities, it belongs to a class of regulatory strategies that Heckhausen and Schulz (1995) called secondary control. Future research employing cross-cultural samples will reveal if the contrast effect is generalizable across cultures. Further analysis of contrasts across time revealed that all contrasts between NA and PA tended to have a positive effect on WBA, and that contrast is most readily activated by recent negative experiences. However, such a conclusion should be regarded as tentative, given the small sample size, the nature of the sample (i.e., university students), and the time frame and the intercorrelations of the measures. We have seen that time-1 NA · time-2 PA reached marginal significance; hence more distant negative experiences might also carry contrast effects. For example, if the study was to focus on a special sample (e.g., recovered cancer patients), and the period of study was to span from the beginning of treatment to a year into remission, this more distant (negative) experience which probably leaves a permanent mark in oneÕs life would still form a basis for contrast. Moreover, as the measures of PA were highly correlated between time 1 and time 2, same for NA, there was a fair degree of redundancy among the four regression models in Table 4. Indeed, ignoring the issue of statistical power for the moment, when the four biweekly measures of affect were entered as a block in a hierarchical regression, followed by the four product terms entered on a stepwise basis, only T2NA · T2PA met the criterion for entry, adding 2.1% of explained variance to the equation (cf. Table 3). On hindsight, had the study focused on just the last two weeks of the diary period, the size of the contrast effect found would have been much larger (from 2.1% to 4.9%; see Tables 3 and 4). Notwithstanding the limitations, the results suggest that WBA is highly dependent on PA and less on NA. Moreover, NA also serves as a basis against which PA is evaluated so that the effect of PA on WBA is further augmented. This positive bias in the way we use positive and negative
914
S.-T. Cheng / Personality and Individual Differences 37 (2004) 905–915
experiences may be related to another bias that has been reported in the literature, namely our time perspective regarding positive and negative events. People commonly believe that positive events tend to recur whereas negative events tend not to (Pyszczynski, Holt, & Greenberg, 1987). It is therefore easy for us to regard positive events as more representative of our current state of affairs than negative events. According to Schwarz and Strack (2000), this should enhance the likelihood of positive events being assimilated into our sense of well-being, and negative events serving as the standard for comparison. In this connection, it is interesting to note that depressed people believe negative events to be equally likely, if not more likely, as positive events (Peterson & Seligman, 1984; Pyszczynski et al., 1987). People who are depressed might have temporarily lost this ability to repair well-being. There is a reason why the contrast function is carried by NA rather than PA. Watson et al. (1999) observed that NA is characteristically low over time but becomes highly elevated in problematic situations, leading to a positively skewed and leptokurtic distribution within an individual. Hence when NA is noticeably increased, something must be wrong and our coping mechanism is called upon to deal with its impact on well-being. However, low PA is not necessarily associated with crisis (e.g., we may work intensively for hours and days to get something done, during which there is very little experience of PA) and hence it bears no clear-cut relationship with contrast. A final caveat is warranted before coming to a close. It should not be mistaken that happiness is best achieved by subjecting oneself to suffering. The available data from Table 3 indicate that PA and NAÕs combined endowment effect is much larger than the effect of their contrast. Thus, it is highly unlikely that one can find happiness when PA is low. However, as the size of the contrast term is in direct proportion to the amount of NA, future studies employing populations with more intensive negative experiences (e.g., life-threatening illnesses) might reveal a larger contrast effect on well-being. In any case, the data suggest that positive experiences are more valued when life is rough than when it is easy. Acknowledgements I am grateful to Jacky Cheung and two anonymous reviewers for their helpful comments on an earlier version of the manuscript.
References Aiken, L. S., & West, S. G. (1991). Multiple regression: testing and interpreting interactions. Newbury Park, CA: Sage. Andrews, F. M., & Withey, S. B. (1976). Social indicators of well-being: AmericansÕ perception of life quality. New York: Plenum. Bradburn, N. M. (1969). The structure of psychological well-being. Chicago: Aldine. Brickman, P., Coates, D., & Janoff-Bulman, R. (1978). Lottery winners and accident victims: is happiness relative. Journal of Personality and Social Psychology, 36, 917–927. Caminada, H., & de Bruijn, F. (1992). Diurnal variation, morningness–eveningness, and momentary affect. European Journal of Personality, 6, 43–69. Campbell, A., Converse, P. E., & Rodgers, W. L. (1976). The quality of American life: perceptions, evaluations, and satisfactions. New York: Russell Sage Foundation.
S.-T. Cheng / Personality and Individual Differences 37 (2004) 905–915
915
Cantril, H. (1965). The pattern of human concerns. New Brunswick, NJ: Rutgers University Press. Diener, E. (2000). Subjective well-being: the science of happiness and a proposal for a national index. American Psychologist, 55, 34–43. Diener, E., Emmons, R. A., Larsen, R. J., & Griffin, S. (1985). The satisfaction with life scale. Journal of Personality Assessment, 49, 71–75. Hamid, P. N., & Cheng, S.-T. (1996). The development and validation of an index of emotional disposition and mood state: the Chinese Affect Scale. Educational and Psychological Measurement, 56, 995–1014. Heckhausen, J., & Schulz, R. (1995). A life-span theory of control. Psychological Review, 102, 284–304. Hills, P., & Argyle, M. (2002). The Oxford Happiness Questionnaire: a compact scale for the measurement of psychological well-being. Personality and Individual Differences, 33, 1073–1082. Lucas, R. E., Diener, E., & Suh, E. (1996). Discriminant validity of well-being measures. Journal of Personality and Social Psychology, 71, 616–628. Lu, L., Gilmour, R., Kao, S.-F., Weng, T.-H., Hu, C.-H., Chern, J.-G., Huang, S.-W., & Shih, J.-B. (2001). Two ways to achieve happiness: when the East meets the West. Personality and Individual Differences, 30, 1161–1174. Miller, D. L., Manne, S. L., Taylor, K., & Keates, J. (1996). Psychological distress and well-being in advanced cancer: the effects of optimism and coping. Journal of Clinical Psychology in Medical Settings, 3, 115–130. Neugarten, B. L., Havighurst, R. J., & Tobin, S. S. (1961). The measurement of life satisfaction. Journal of Gerontology, 16, 134–143. Peterson, C., & Seligman, M. E. (1984). Causal explanations are a risk factor for depression: theory and evidence. Psychological Review, 91, 347–374. Pyszczynski, T., Holt, K., & Greenberg, J. (1987). Depression, self-focused attention, and expectancies for positive and negative future life events for self and others. Journal of Personality and Social Psychology, 52, 994–1001. Røysamb, E., Harris, J. R., Magnus, P., Vittersø, J., & Tambs, K. (2002). Subjective well-being. Sex-specific effects of genetic and environmental factors. Personality and Individual Differences, 32, 211–223. Scheier, M. F., & Carver, C. S. (1992). Effects of optimism on psychological and physical well-being: theoretical overview and empirical update. Cognitive Therapy and Research, 16, 201–228. Schwarz, N., & Strack, F. (2000). Reports of subjective well-being: judgmental processes and their methodological implications. In D. Kahneman, E. Diener, & N. Schwarz (Eds.), Well-being: the foundations of hedonic psychology (pp. 61–84). New York: Russell Sage Foundation. Seligman, M. E. P., & Csikszentmihalyi, M. (2000). Positive psychology. American Psychologist, 55, 5–14. Strack, F., Schwarz, N., & Gschneidinger, E. (1985). Happiness and reminiscing: the role of time perspective, affect, and mode of thinking. Journal of Personality and Social Psychology, 49, 1460–1469. Taylor, S. E., & Brown, J. D. (1988). Illusion and well-being: a social psychological perspective on mental health. Psychological Bulletin, 103, 193–210. Tversky, A., & Griffin, D. (1991). Endowment and contrast in judgments of well-being. In F. Strack, M. Argyle, & N. Schwarz (Eds.), Subjective well-being: an interdisciplinary perspective (pp. 101–118). Oxford: Pergamon. Watson, D., Wiese, D., Vaidya, J., & Tellegen, A. (1999). The two general activation systems of affect: structural findings, evolutionary considerations, and psychobiological evidence. Journal of Personality and Social Psychology, 76, 820–838. Zajonc, R. B. (1980). Feeling and thinking: preferences need no inferences. American Psychologist, 35, 151–175.