6
Divergent and Convergent Validity of Self-Esteem: A State Perspective
The vast majority of research on self-esteem has considered it as a trait, an individual difference that is more or less enduring across time and situations. As attested by the richness of the research and ideas described in the chapters in this book, research conceptualizing self-esteem as a trait has been quite valuable. Nevertheless, there is a growing body of research suggesting that it is also useful to consider self-esteem as a state, a construct that is presumed to change across time and situations, and this chapter will focus on the validity of state self-esteem. Evaluating the validity of a construct requires a clear definition of that construct, and for present purposes, self-esteem is defined primarily in terms of a global self-evaluation of one’s worth. Certainly, other definitions are possible, and the considerations discussed here can be applied to most any definition (or definitions) of self-esteem.
There are numerous ways to evaluate the validity of a construct, and this chapter focuses on convergent and divergent validity. Convergent validity exists when a measure of a construct is related to measures of other constructs to which the definitions of the constructs suggest it should be related. The convergent validity of state self-esteem was evaluated in terms of the strength of the relationships between measures of the state self-esteem and measures of other state constructs, such as depressive thinking, for which it was reasonable to expect self-esteem to be related. Divergent validity exists when a measure of a target construct is unrelated to measures of other constructs to which the definitions of constructs suggest it to be unrelated (the complement of convergent validity). Divergent validity also implies that a target construct predicts, or is related to, some outcome of interest above and beyond relationships between other measures and this outcome. The divergent validity of state self-esteem was evaluated in terms of both these criteria.
The choice of constructs used to evaluate the validity of a target construct or measure is critical, and choosing different constructs may lead to different conclusions about the validity of a measure. The validity of state self-esteem was evaluated in terms of the following constructs: psychological well-being (e.g., depression), affect, self-referential thinking, self-concept, and perceptions of control over the environment. These constructs were chosen on the basis of trait-level research that has consistently found relationships between trait measures of these constructs and the trait self-esteem. Although state- and trait-level relationships are mathematically independent and may represent different psychological processes (e.g., Nezlek, 2001; Tennen & Affleck, 1996), relationships among traits may represent good starting points for examining relationships among the same constructs considered as states.
The empirical basis for this examination was a series of studies in which participants provided data every day for 2–3 weeks. State-level constructs were measured with items taken from trait-level measures, reworded for daily administration. Daily self-esteem was measured, using an “uncharacteristic-characteristic of me” response scale, with four items from Rosenberg (1965), one of which was “Today, I had a positive attitude toward myself’. Daily depressogenic adjustment was measured with three items based on Beck’s (1967) triad, one of which was “How optimistic are you about how your life (in general) will be tomorrow?” Daily affect was measured using 20 items representing a circumplex model (e.g., Feldman Barrett & Russell, 1998). There were five items representing each of positive activated mood (e.g., happy), positive deactivated mood (e.g., relaxed), negative activated mood (e.g., nervous), and negative deactivated mood (e.g., sad). Self-referential thinking was defined in terms of the two dimensions of private self-consciousness suggested by Trapnell and Campbell (1999), reflection and rumination, and public self-consciousness (Fenigstein, Scheier, & Buss, 1975). For example, participants indicated “How much today did you …” think about your attitudes and feelings (reflection), ruminate or dwell on things that happened to you (rumination), and think about what other people thought of you (public self-consciousness). Self-concept was measured with four items taken from the Self-Concept Clarity scale (Campbell, Trapnell, Heine, Katz, Lavallee, & Lehman, 1996), and participants indicated how much each statement applied to them that day, e.g., “My beliefs about myself often conflict with one another.” Measures of perceived control over one’s environment were based on Deci and Ryan’s (1985) work on causality orientation. Based on analyses described in Nezlek and Gable (2001), perceived control was measured with four items, two items for control over the social and achievement domains. These items corresponded to Deci and Ryans impersonal and autonomy dimensions. For example, “Thinking back on your day today in terms of the social events that occurred and the relationships you have with others, how much choice did you really have about what you did?” All measures used 1–7 scales.
Relationships between self-esteem and depressogenic adjustment, affect, self-referential thinking, self-consciousness, and perceived control were estimated from data described in Nezlek (2005). Sample sizes for these analyses ranged between 558 and 735 participants and between 8079 and 11,153 days. Relationships between self-esteem and self-concept clarity were estimated from data described in Nezlek and Plesko (2001). These data were analyzed with a series of multilevel models in which days were treated as nested within persons. Conceptually, a within-person regression equation describing the relationships between self-esteem and other measures was estimated for each person, and then the mean of these relationships was calculated. [See Nezlek (2001) for a more detailed description of, and rationale for, these analyses.]
The relationships between state self-esteem and other state constructs were in directions consistent with the definitions of the constructs. State self-esteem was positively related to depressogenic adjustment, both measures of positive affect, self-concept clarity, and perceived control, and it was negatively related to both measures of negative affect, rumination, and public self-consciousness. On days when peoples self-esteem was higher, compared to days when their self-esteem was lower, people experienced more positive affect, their self-concept clarity was greater, etc., and they experienced less negative affect, ruminated less, etc.
Usually, validity coefficients are described in terms of correlations, and the estimated within-person correlations between self-esteem and other state-level constructs are presented in Table 6.1. Due to the nature of the multilevel analyses, these correlations were derived from shared variance estimates, and these shared variances are also presented in Table 6.1. This procedure is described in Nezlek (2001). These relationships are sufficiently strong (within-person correlations between .25 and .70) to demonstrate the convergent validity of the state self-esteem; however, they are not so strong to suggest that the state self-esteem cannot be differentiated from these constructs.
TABLE 6.1 Shared Variances and Within-Person Correlations between State Self-Esteem and Other State Constructs
Construct |
Shared Variance |
Correlation |
Depressogenic adjustment |
0.51 |
.71 |
Positive activated affect |
0.36 |
.60 |
Positive deactivated affect |
0.36 |
.60 |
Negative activated affect |
0.32 |
–.57 |
Negative deactivated affect |
0.33 |
–.57 |
Reflection |
0.09 |
.30 |
Rumination |
0.13 |
–.36 |
Public self-consciousness |
0.06 |
–.24 |
Self-concept clarity |
0.25 |
.50 |
Control over events |
0.11 |
.33 |
As can be seen from these data, these relationships varied in strength, with the relationship between self-esteem and depressogenic adjustment being the strongest. The fact that this relationship was the strongest of the group parallels research on trait-level relationships; however, it also needs to be considered in light of the fact that one of the depression items, “Overall, how positively did you feel about yourself today” is basically a measure of self-esteem. Most theories and explanations of depression posit that depression includes a diminished sense of self-worth, so the substantial overlap between the two constructs is not surprising.
At the trait level, some have suggested that self-esteem can be represented by combinations of variables, such as affect (e.g., Watson, Suls, & Haig, 2002). Such a possibility was considered by predicting self-esteem from the four affect measures, and in a separate analysis from reflection, rumination, and public self-consciousness. The estimated within-person multiple R (based on changes in the error variance in self-esteem) when all four affect measures were included as predictors was .75, and when all three measures of self-referential thinking were included the reduction was .46. It is important to note that these correlations correspond to 56% and 21% shared variance between self-esteem and the two combinations of variables.
Simply demonstrating that the correlation between two measures is not 1.0 does not de facto indicate that the measures have divergent validity. Additional corroboration can come from demonstrating that the relationship between a target measure (i.e., self-esteem) and another, criterion measure does not overlap with the relationship between the criterion measure and a comparison measure (e.g., positive effect). The selection of “criterion” measures is not straightforward. Nevertheless, much of the research on within-person variability in psychological states has concerned relationships between such states and daily events, and so daily events were used as the criteria to evaluate the divergent validity of the state self-esteem. Is daily (state) self-esteem related to daily events above and beyond the relationship between daily events and other daily (state) measures?
In all of the studies described thus far, participants also described the events that occurred to them each day using a variant of the Daily Events schedule (Butler, Hokanson, & Flynn, 1994). These events represented four categories, combinations of positive vs. negative and social vs. achievement. Each event was rated on a 0 (did not happen) to 4 (happened and was very important) scale. Event ratings were averaged within each category to create an impact score. Impact scores were used for the analyses instead of frequency counts (i.e., how many events in each category occurred) because impact scores had more desirable psychometric properties (e.g., homogeneity of variance). It should be noted that relationships between frequency counts and daily measures of psychological states were functionally equivalent to the results discussed here. [See Nezlek & Gable (2001) for a brief discussion of impact scores and frequency counts.]
First, daily events (positive and negative, social and achievement) were regressed onto self-esteem. Self-esteem positively covaried with positive social and achievement events and negatively covaried with negative social and achievement events (all p values < .001). Moreover, separate analyses found that all of the validating constructs were also related to daily events (all p values < .001). These analyses are discussed in detail in Nezlek (2005) and in Nezlek and Plesko (2001). Next, validating constructs were entered as predictors, singly and in the groups described above (the four mood measures and the three self-referential thinking measures). In all cases, relationships between self-esteem and the event measures remained significant (all p values < .001) after taking into account within-person (daily) variability in the other measures.
Taken together, these two sets of results suggest that measures of global state self-esteem based on Rosenberg’s trait measure have convergent and divergent validity. That is, state self-esteem was related to other states in ways that were consistent with the definitions of the constructs being considered. Equally important, state self-esteem was related to daily events above and beyond relationships between state self-esteem and other state-level constructs and the relationships between daily events and these other constructs, suggesting that state self-esteem has divergent validity.
In addition, relationships between state self-esteem and other states were generally weaker than relationships between the same constructs measured at the trait level. This difference may have been due to the fewer items used to measure each construct, although some research [unpublished analyses from Nezlek & Gable (2001) and Nezlek & Plesko (2001)] has found reasonable (.65 to .9) item-level reliabilities for the state-level constructs discussed here. Moreover, the multilevel analyses reported here take into account different types of unreliability.
Assuming that state relationships between self-esteem and other constructs are weaker than corresponding trait relationships suggests that state- and trait-level self-esteem constructs differ. The high correlations some have found between trait self-esteem and other measures may reflect the fact that responses to trait measures reflect some sort of general assessment of one’s life more than state measures. Most trait measures either explicitly or implicitly ask people to make some sort of aggregate judgment. When people describe their typical mood (e.g., Watson, Suls, & Haig, 2002), they are implicitly aggregating across some unknown period of time and number of occasions, and they go through a similar process when describing their general self-worth. In contrast, and by definition, state measures reflect the immediate situation more than the past. State measures are time and/or situation bound.
Some think of trait measures as functionally equivalent to aggregates of state measures [see Fleeson (2001) for a discussion of this issue]. If this is the case, then aggregates of state measures of self-esteem may be more valid measures of self-esteem than trait measures because they may be less prone to selective recall and other biases and may have greater divergent validity. Although this either/or argument could be made, it is probably not the most productive to pursue. It is probably more productive to think of state and trait measures of self-esteem as measures of related, but distinct constructs. Such a position presents an interesting challenge of determining just what each of these measures truly represents.
Conceptualizing State Variability as a Trait
The preceding discussion has examined how changes in state self-esteem correspond to changes in other state measures; however, a large body of research, pioneered by Kernis and colleagues, has treated changes in the state self-esteem as a trait variable, typically labeled “instability”. Instability is defined in terms of within-person standard deviations, and this measure is treated like a trait. The logic of this approach can be extended to other states. For example, within-person standard deviations could be calculated for perceived control over the environment. Moreover, the within-person relationships between state self-esteem and other states suggest that the instability of state self-esteem may be related to the instability of other states. That is, other states covary with state self-esteem, and it is reasonable to expect that within-person changes in these measures would be positively related to within-person changes in self-esteem. For example, the more unstable a person is in terms of self-esteem, the more unstable he or she will be in terms of perceived control. If the relationships among a set of measures of instability are sufficiently strong, it is possible that they may form (or more accurately, reflect) the existence of a some sort of a general “g” factor of instability. If this is the case, then unstable self-esteem may be an “s” (specific) factor of this general factor.
The existence of “g” factor of instability was examined by Gable and Nezlek (1998), who collected daily measures (for 3 weeks) of self-esteem, depressogenic adjustment, perceived control over the environment (as described previously), daily anxiety, and causal uncertainty (Weary & Edwards, 1994). They calculated within-person means and standard deviations for each of these measures, and a confirmatory factor analysis found that all five means loaded on one factor and all five standard deviations loaded on another factor (an instability factor). Moreover, the correlation between the two factors was –.34, very similar to the correlations between mean self-esteem and self-esteem instability reported by Kernis and colleagues (e.g., Kernis, Grannemann, & Mathis, 1991).
These results suggest that the instability of state self-esteem is a specific manifestation of a more general factor of instability. This does not mean that the instability of self-esteem is not a meaningful or worthwhile construct; rather, it suggests that when thinking about the instability of self-esteem we need to be mindful of the possible operation of a broader factor. Of course, simply demonstrating that the instability of self-esteem shares variance with the instability of other states does not mean that the instability of self-esteem does not have divergent validity. The instability of self-esteem may be related to other measures above and beyond relationships between these other measures and the instability of other states. That is, it might be useful to think more in terms of s-factors than a g-factor of instability, much like the improvements that measuring facets of the Five Factor Model (FFM) sometimes provides over measures of the broader factors. Clearly, resolving this issue requires further research.
References
Beck, A. T. (1967). Depression: Clinical, experimental, and theoretical aspects. New York: Harper & Row.
Butler, A. C., Hokanson, J. E., & Flynn, H. A. (1994). A comparison of self-esteem lability and low trait self-esteem as vulnerability factors for depression. Journal of Personality and Social Psychology, 66, 166–177.
Campbell, J. D., Trapnell, P. D., Heine, S. J., Katz, I. M., Lavallee, L. F., & Lehman, D. R. (1996). Self-concept clarity: Measurement, personality correlates, and cultural boundaries. Journal of Personality and Social Psychology, 70, 141–156.
Deci, E. L., & Ryan, R. M. (1985). The General Causality Orientations Scale: Self-determination in personality. Journal of Research in Personality, 19, 109–134.
Feldman Barrett, L., & Russell, J. A. (1998). Independence and bipolarity in the structure of current affect. Journal of Personality and Social Psychology, 74, 967–984.
Feningstein, A., Scheier, M. F., & Buss, A. (1975). Public and private self-consciousness. Journal of Consulting and Clinical Psychology, 43, 522–527.
Fleeson, W. (2001). Toward a structure- and process-integrated view of personality: States as density distributions of traits. Journal of Personality and Social Psychology, 80, 1011–1027.
Gable, S. L., & Nezlek, J. B. (1998). Level and instability of day-to-day psychological wellbeing and risk for depression. Journal of Personality and Social Psychology, 74, 129–138.
Kernis, M. H., Grannemann, B. D., & Mathis, L. C. (1991). Stability of self-esteem as a moderator of the relation between level of self-esteem and depression. Journal of Personality and Social Psychology, 61, 80–84.
Nezlek, J. B. (2001). Multilevel random coefficient analyses of event and interval contingent data in social and personality psychology research. Personality and Social Psychology Bulletin, 27, 771–785.
Nezlek, J. B. (2005). Distinguishing affective and non-affective reactions to daily events. Journal of Personality, 73, 1539–1568.
Nezlek, J. B., & Gable, S. L. (2001). Depression as a moderator of relationships between positive daily events and day-to-day psychological adjustment. Personality and Social Psychology Bulletin, 27, 1692–11704.
Nezlek, J. B., & Plesko, R. M. (2001). Day-to-day relationships among self-concept clarity, self-esteem, daily events, and mood. Personality and Social Psychology Bulletin, 27, 201–211.
Rosenberg, M. (1965). Society and the adolescent self-image. Princeton, NJ: Princeton University Press.
Tennen, H., & Affleck, G. (1996). Daily processes in coping with chronic pain: Methods and analytic strategies. In M. Zeidner & N. Endler (Eds.), Handbook of coping (pp. 151–180). New York: Wiley.
Trapnell, P. D., & Campbell, J. D. (1999). Private self-consciousness and the five factor model of personality: Distinguishing rumination from reflection. Journal of Personality and Social Psychology, 76, 284–304.
Watson, D., Suls, J., & Haig, J. (2002). Global self-esteem in relation to structural models of personality and affectivity. Journal of Personality and Social Psychology, 83, 185–197.
Weary, G., & Edwards, J. A. (1994). Individual differences in causal uncertainty. Journal of Personality and Social Psychology, 67, 308–318.