سبک های خودارائه گری در خودسنجی: لینک کردن عوامل عمومی سبک های واکنش، ویژگی های شخصیتی و ارزش ها در یک مطالعه طولی

Abstract We investigated how response styles, personality traits, and values can be taken as manifestations of self-presentation styles in self-reports, and how self-presentation affects other self-report measures over time. Data on values and character traits at three time points across five years collected among a national representative sample in the Netherlands were utilized. A general response style factor consisting of extreme, socially desirable, and midpoint responding, a general factor of personality from the International Personality Inventory, and a general value factor from the Rokeach Value Survey were extracted, all of which showed scalar invariance across time. A latent self-presentation factor underlying the three general factors at each time point, and its stability and changes across time points was modeled. All three general factors loaded positively on the self-presentation factor. The latent mean of the self-presentation factor became smaller over time, yet effects of its impact on the relationships among various psychological variables remained small and stable over time. We conclude that survey respondents show a similar self-presentation style across domains and over time. Score corrections to deal with response styles are not recommended.

Introduction Most social interactions, including attitude expression and change, can be analyzed through the lens of self-presentation (Baumeister, 1982). Self-presentation is the use of behavior as a means of communicating information about (or an image of) oneself to others. Depending on personal dispositions and specific contexts, people exhibit preferred self-presentation styles, such as being assertive or defensive, acquisitive or protective, and active or passive. Response styles, defined as respondents’ systematic tendencies to respond to questionnaires on some basis other than the target constructs (Paulhus, 1991), can be considered an essential indicator of self-presentation styles in survey responses (e.g., Smith, 2004). We are interested in response styles that affect self-report Likert scales in different domains; more specifically, we set out to examine whether individuals show a similar self-presentation style across measures of personality (i.e., personal style) and values (i.e., normative function), and to what degree this style changes over time. The novelty of our study is that we use a variety of psychological measures to investigate self-presentation across psychological domains in a longitudinal study. 1.1. Response styles and their integration The most studied response styles include acquiescent response style (ARS), extreme response style (ERS), midpoint response style (MRS), and socially desirable responding (SDR). Traditionally, these response styles are viewed as sources of common method bias that should be controlled for. Recent evidence, however, suggests that response styles may have a substantive meaning, as they are found to share trait variance with personality and values. For example, ERS was positively associated with extroversion, conscientiousness, and individualistic values, SDR with extroversion and conscientiousness, ARS with agreeableness and compliance, and MRS with modesty (e.g., Austin et al., 2006 and Chen et al., 1995). To integrate these findings, He and van de Vijver (2013) confirmed that a general response style (GRS) can subsume these four response styles, with ERS and SDR as positive indicators and ARS and MRS as negative indicators. This integrated GRS was interpreted by the authors as a preferred communication style that represents the tendency of response amplification versus moderation. 1.2. Response styles, personality, and values Johnson (1981) suggested that response styles and the expressions of personality and values share some commonality, possibly all related to one’s self-presentation styles. In line with this suggestion, the general factor of personality derived from the Big Five personality traits, interpreted as a basic personality disposition integrating the most general non-cognitive dimensions of personality (Musek, 2007), was found to be positively correlated with GRS (He & van de Vijver, 2013). Bye et al. (2011) reported that personal values, associated with ARS, were related to intended impression management. If a general factor of values were to be extracted (e.g., Podsakoff & Organ, 1986), we expect this factor to be indicative of individuals’ self-presentation styles. 1.3. Stability and change in self-presentation styles Little has been done on the stability and changes of self-presentation styles. Yet, the stability of response styles over time has been demonstrated. Weijters, Geuens, and Schillewaert (2010) used different item sets at two time points to measure response styles over a one-year period. They modeled time-invariant and time-specific response style factors, and found that response styles showed considerable stability. We expect temporal stability of self-presentation styles. Furthermore, in this study items common to all data collection points are used as indicators of response styles, and expressions of personality and values, thus both the stability and the changes of their integration over time could be tested. 1.4. The present study The literature suggests that response styles and the expressions of personality and values may all be part of self-presentation styles. The stability and changes of self-presentation styles over time, in turn, may affect the associations among self-report measures. We used a general factor of response styles derived from socially desirable, extreme, and midpoint responding, a general factor of personality based on the Big Five personality traits, and a general factor of values based on six value dimensions across three time points (T1, T2, and T3), and we extracted a time-specific self-presentation factor from these three general factors at each time point to model its stability over time. The effects of changes in the self-presentation factor over time could be demonstrated through comparing correlations of this factor with external psychological measures and intercorrelations among these external psychological measures with and without this factor corrected for at each time point. Well-established measures, such as self-esteem, life satisfaction, and positive and negative affect, were used as external measures. Researchers found that self-esteem, life satisfaction, and positive affect are positively related to each other, whereas negative affect shows a negative association with self-esteem and life satisfaction (Preisendörfer and Wolter, 2014 and Robins et al., 2001). If individuals indeed prefer a similar style across domains and time, we expect that the correction of the self-presentation factor would result in similar changes (or lack of changes) in correlations with external variables and intercorrelations among external variables over time.

3. Results We describe the results in three parts. Firstly, we report the longitudinal measurement invariance of the three general factors (GRS, the general personality factor, and the general value factor), a prerequisite for valid longitudinal comparisons. Secondly, we report the modeling of the self-presentation factor from these general factors, and its stability and changes over time. Lastly, we describe effects of correcting for the self-presentation factor on the relationships among external psychological variables at each time point. 3.1. Longitudinal measurement invariance of the general factors To demonstrate the measurement equivalence of GRS, the general personality factor, and the general value factor over time, confirmatory factor analyses were carried out in AMOS (Byrne, 2001). For GRS, we specified three time-specific GRS factors (so, one factor per time point), where ERS, MRS, and SDR at each time point were indicators; these time-specific GRS factors were correlated with each other, and uniqueness of each indicator was correlated across time points (e.g., ERS at T1, T2 and T3 were correlated) (configural invariance). We then added constraints to test invariance of measurement weights (i.e., factor loadings on GRS were equal across time points) and invariance of intercepts (i.e., indicators had the same intercepts across time points). The same model specifications were applied to the general personality factor and the general value factor. The model fit was evaluated by Chi-square tests, the Tucker–Lewis Index (acceptable above .90), the Comparative Fit Index (acceptable above .90), and the Root Mean Square Error of Approximation (acceptable below .06). The acceptance of a more restricted model was based on the change of CFI (less than .01) ( Cheung & Rensvold, 2002). Due to 31% missing values of all the items across all time points, we resorted to full information maximum likelihood estimation in confirmatory factor analyses (Schafer & Graham, 2002). For each general factor, invariance of measurement weights and intercepts was well supported by the fit indexes (Table 2), although the chi-square values were rather high, which might be caused by the large sample size. We concluded that the longitudinal measurement invariance for the three general factors was confirmed, thus these factors could be compared across time points. Table 2. Measurement invariance of the scales: Measurement weights and intercept invariance. Scale Invariance χ2 (df) TLI CFI RMSEA General response style1 Configural 222.89 (21)⁎⁎ .97 .99 .03 MW 248.51 (27)⁎⁎ .98 .99 .03 Intercepts 315.22 (33)⁎⁎ .98 .98 .03 General personality factor Configural 1587.43 (72)⁎⁎ .96 .97 .05 MW 1598.66 (80)⁎⁎ .96 .97 .04 Intercepts 1914.35 (90)⁎⁎ .96 .97 .05 General value factor2 Configural 37.8.37 (69)⁎⁎ .91 .95 .07 MW 3734.89 (77)⁎⁎ .92 .95 .07 Intercepts 4076.98 (87)⁎⁎ .92 .94 .07 TLI = Tucker–Lewis Index; CFI = Comparative Fit Index; RMSEA = Root Mean Square Error of Approximation; MW = Measurement Weights. 1 The variance of the error term for extreme response style was fixed to 0 to avoid negative variance. 2 The error terms of prosocial concern and stimulation/comfort were negatively correlated. ⁎⁎ p < .01 (two tailed). Table options 3.2. Stability and change in self-presentation style We tested the model depicted in Fig. 1 with full information maximum likelihood estimation. We used the factor scores of GRS, the general personality factor, and the general value factor as observed variables. Each of the general factors of the preceding time points predicted that of the following time points, and a latent self-presentation factor was defined by the three general factors at each time point; their loadings were constrained to be equal across time (e.g., same loading of GRS on the self-presentation factor at the three time points). The model fitted well, χ2(19, N = 9935) = 96.66, p < .01, TLI = .99, CFI = 1.00, and RMSEA = .02. The standardized regression weights and the explained variance for each endogenous variable are presented in Fig. 1. All three observed general factors loaded positively on the self-presentation factor at each time point, indicating that participants had a similar style across personality and value domains, and that this self-presentation factor here represented amplifying versus moderating responses. The factor loadings of the general factors at T1 (.43–.63) were larger than these of T2 and T3 (.25–.40), which suggests a reduced impact of the self-presentation style across measurement occasions. Longitudinal model of the self-presentation style. Note: Standardized regression ... Fig. 1. Longitudinal model of the self-presentation style. Note: Standardized regression weights and factor loadings (all significant at p < .01) are presented next to the arrows. Numbers in bold represent proportions of variance explained. Figure options Due to model identification issues, the changes of the latent mean of the self-presentation factor over time could not be estimated in the model shown in Fig. 1. We tested it instead in a longitudinal measurement invariance model, in which the three observed general factors loaded on the self-presentation factor at each time point. The measurement weights invariance model (χ2(28, N = 9935) = 1572, p < .01, TLI = .90, CFI = .94, RMSEA = .08) and intercept invariance model (χ2(34, N = 9935) = 1,948, p < .01, TLI = .90, CFI = .92, RMSEA = .08) were largely supported. The latent means of the self-presentation factor were compared in the measurement intercept model with the mean of T1 fixed to zero. The model showed an acceptable fit, χ2(32, N = 9935) = 1,884, p < .01, TLI = .90, CFI = .93, RMSEA = .08. All three indicators loaded positively on the latent factor (ranging from .22 to 1.00). Compared with T1, the self-presentation factor at T2 had a lower mean (M = −.05, SE = .01, p < .01), and that at T3 even lower (M = −.10, SE = .01, p < .01), pointing to the decrease of the self-presentation factor over time. 3.3. Correction for the self-presentation factor Scores on self-esteem, life satisfaction, and positive and negative affects were correlated with the self-presentation factor (i.e., factor scores estimated in the model depicted in Fig. 1 in AMOS) and with each other before and after the self-presentation factor was partialled out at each time point (Table 3). Only respondents for whom data at all three points were available were included in the correlation analyses (n = 3879). The correlations of these external variables with the self-presentation factor were consistent across time points. Before correction, self-esteem and life satisfaction showed positive correlations with positive affect, and they had negative correlations with negative affect at all three time points. After partialling out the self-presentation factor, the patterning of all correlations remained the same but the values became slightly weaker. The average absolute correlation dropped from .30 to .21 at T1, from .31 to .25 at T2, and from .32 to .26 at T3. All in all, the correction for the self-presentation factor across time does not seem to strongly affect the associations among these external variables. Table 3. Correlations with the self-presentation factor and among external measures: Zero-order correlations (below diagonal) and with the self-presentation factor partialled out (above diagonal) at each time point. Self-presentation factor 1 2 3 4 T1 1. Self-Esteem .55 – .32 .10 −.29 2. Life Satisfaction .37 .45 – .13 −.18 3. Positive Affect .42 .31 .27 – .26 4. Negative Affect −.48 −.48 −.32 .01 – T2 1. Self-Esteem .49 – .41 .17 −.35 2. Life Satisfaction .29 .49 – .15 −.24 3. Positive Affect .35 .30 .23 – .20 4. Negative Affect −.42 −.48 −.33 .03 – T3 1. Self-Esteem .48 – .40 .18 −.37 2. Life Satisfaction .29 .48 – .17 −.26 3. Positive Affect .27 .29 .24 – .19 4. Negative Affect −.43 −.50 −.36 .04 – All correlations are significant at p < .01, except the ones italicized.