مقایسه شرایط خود-دیگران در ارزش های شخصی در مقابل پنج عامل بزرگ شخصیت
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|34271||2014||10 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Journal of Research in Personality, Volume 50, June 2014, Pages 1–10
Can we judge other people’s values accurately, or are values too subjective to assess? We compared self-other agreement in personal values with agreement in the Big Five personality traits. Self-other agreement in four higher-order values (median r = .47) and in six culture-specific value factors (median r = .50) was substantial and similar to that for the Big Five personality traits (median r = .51). When corrected for attenuation due to measurement error self-other agreement was high for all three scales (median rs > .65). The results suggest that people can assess values of others whom they know well with remarkable accuracy. Therefore, other-ratings of personal values can be used to validate and complement self-report value measures.
The limits of self-report methodology confront researchers in many areas of psychological science. People’s reports of their behaviour, attitudes, and personality may be affected by various response biases such as socially desirable, neutral, or extreme responding, and acquiescence (e.g., Mõttus et al., 2012, Paulhus, 1991 and Schwartz et al., 1997). Based only on people’s potentially biased self-reports, we cannot be certain whether a person truly endorses benevolence values highly or rejects power values, nor can we be sure a person is actually extraverted or agreeable. To get around self-report biases, it is necessary to collect data with an independent method of measurement. Judgments of other people (e.g., peers, spouses, siblings, parents, etc.) who know the person well can serve this purpose. The degree of agreement between self- and other-ratings—also called convergent validity (Campbell & Fiske, 1959) or consensual validity (McCrae, 1982 and McCrae et al., 2004)—can clarify the accuracy of self-reports. Self-other agreement, typically operationalized as a correlation between the two ratings, refers to the extent to which two perceivers (an informant and a target in our case) view the target in the same way (Kenny & West, 2010). Several studies have shown relatively strong self-other agreement in all the Big Five personality traits (e.g., Allik, Realo, Mõttus, Borkenau, et al., 2010, Allik, Realo, Mõttus, Esko, et al., 2010, Connolly et al., 2007 and Hall et al., 2008), in affective traits (Watson, Hubbard, & Wiese, 2000), and in subjective well-being (Dobewall et al., 2013 and Schneider and Schimmack, 2009). Surprisingly few studies, however, have investigated self-other agreement in personal values. How can we explain the relative lack of interest in self-other agreement by value researchers? One reason might be that personal values are considered “too individually subjective” (Hitlin & Piliavin, 2004, p. 359) to be judged by others. McAdams (1995), for instance, distinguished between individual differences in traits, which he described as so easily observable that even a stranger could judge them with some accuracy, and more privately held personal concerns like values, which are less accessible. This view is at odds with the Five Factor Theory (FFT) of personality, according to which, values, are so-called characteristic adaptations which are formed through the interaction of personality traits with the environment ( McCrae and Costa, 1999 and McCrae and Costa, 2008). As such, values can be assessed better by direct observation than Big Five personality trait domains can ( Allik & McCrae, 2002). Resolving these contradictory views requires an empirical assessment of whether self-other agreement is greater in personal values or in personality traits. The current study examines self-other agreement in personal values, both in four higher-order values and in six culture-specific value factors. In order to assess whether values show greater or lesser levels of self-other agreement than personality traits, we compared self- and other-ratings of values and of personality domains in the same sample. If the level of self-other agreement in values is comparable to the level in personality traits, it would suggest that other-ratings can also be used successfully in value research. Another way to think about self-other agreement refers to agreement about a person’s profile of values or traits. Does one person accurately perceive another’s hierarchy of values—the relative importance of different values to the other person? How accurate is an observer’s perception of the relative degree to which different traits characterize a person? With this aim in mind, we also assessed self-other agreement regarding individual’s profile of values and traits by computing—raw and distinctive—profile correlations (Furr, 2008). This approach can reveal how well an informant knows, for example, whether a target values self-transcendence highly, openness to change moderately but more than self-enhancement, and does not care at all about conservation. 1.1. Self-other agreement in values and other related constructs As noted above, considerable research has examined self-other agreement in such personality constructs as traits, emotional experience, and subjective well-being. For instance, across 36 studies of the Big Five personality traits, the average correlation between self- and other-ratings was r = .36 ( Connolly et al., 2007). In other studies, self-other agreement in personality traits has ranged from r = .40 to .70 ( Konstabel et al., 2012 and McCrae et al., 2004). Agreement correlations for affective traits are only slightly lower than for the personality domains ( Watson et al., 2000). For subjective well-being, a recent meta-analysis of 44 studies yielded an average self-other agreement correlation of r = .42 ( Schneider & Schimmack, 2009). Substantial cross-observer agreement has also been observed in such constructs as moral character ( Cohen, Panter, Turan, Morse, & Kim, 2013), sociopolitical ( Beer & Watson, 2008) and ideological attitudes and prejudice ( Cohrs, Kämpfe-Hargrave, & Riemann, 2012). As noted, the use of other-ratings in value research is relatively scarce. Rentfrow and Gosling (2006), for example, asked informants who knew only about their target’s top-ten music preferences to describe their values on an abbreviated version of the Rokeach Value Survey (Rokeach, 1973) and their personality traits on a 44-item Big Five Inventory (Benet-Martinez & John, 1998). In this zero-acquaintanceship study, the average agreement correlation across the specific value items was r = .15. Paryente and Orr (2010) reported agreement correlations between children’s perceptions of their parents’ values and their mother’s and father’s self-reports for tradition (r = .41/.39, respectively) and self-enhancement values (r = .56 /.52). Another study of a small student sample, yielded self-peer agreement correlations ranging from r = .33 (conservation) to r = .54 (self-transcendence), using 28 value items of the Schwartz Value Survey (SVS; Bernard et al., 2006 and Schwartz, 1992). Murray and colleagues (2002) asked dating and married couples to describe their own and their partners’ traits, feelings, and values, using a list of 18 values adapted from the Schwartz and Bilsky (1990) and Rokeach and Ball-Rokeach (1989) value measures. The similarity (i.e., the intraclass correlation 1) between men’s and women’s value profiles was .26 for those who were dating and .29 for married couples ( Murray et al., 2002). The strongest evidence for self-other agreement in personal values comes from a study by Lee and colleagues (2009). They examined both actual and assumed similarity of values using the full SVS. They reported self-other agreement correlations for the two major value dimensions of openness to change versus conservation (r = .42) and self-transcendence versus self-enhancement (r = .52). They also reported self-other agreement for the ten broad values, with correlations ranging from r = .18 for achievement to r = .49 for power. However, none of the above-mentioned studies took the examination of self-other agreement in personal values as their focus. 1.2. Comparing values and traits Should we expect different levels of self-other agreement in values as compared to personality traits? If so, why? In order to answer these questions, we must first examine how values and personality traits relate conceptually. Therefore, in the next sections we discuss the similarities and differences between these two constructs. 1.2.1. Conceptual similarities and differences between traits and values Values. Schwartz’s (1992) theory of basic human values defines values as desirable, trans-situational goals, varying in importance, that serve as guiding principles in people’s lives. Schwartz (2005a) summarized the features that are common to all values as follows: “(a) Values are beliefs. But they are beliefs tied inextricably to emotion, not objective, cold ideas. (b) Values are a motivational construct. They refer to the desirable goals people strive to attain. (c) Values transcend specific actions and situations. They are abstract goals. The abstract nature of values distinguishes them from concepts like norms and attitudes, which usually refer to specific actions, objects, or situations. (d) Values guide the selection or evaluation of actions, policies, people, and events. That is, values serve as standards or criteria. (e) Values are ordered by importance relative to one another. People’s values form an ordered system of value priorities that characterize them as individuals. This hierarchical feature of values also distinguishes them from norms and attitudes” (Chapter 1, Introduction). Values also differ from motives and needs, because “values are inherently desirable and must be represented cognitively in ways that enable people to communicate about them” ( Roccas, Sagiv, Schwartz, & Knafo, 2002, p. 789). Recent research ( Bilsky & Schwartz, 2008), however, suggests that different indicators of the same motive construct are correlated, independent of the assessment method (i.e., implicit versus explicit). Schwartz (1992) specified ten broad values according to the type of goal or motivational concern that they express: He grounded the ten values in one or more of three universal requirements of human existence: (1) needs of individuals as biological organisms, (2) requisites of coordinated social interaction between individuals, and (3) survival and welfare needs of groups. These motivationally distinct value orientations have been recognized and discriminated by people in over 82 countries studied thus far (Schwartz, 2012). They form a quasi-circumplex structure, presented in Fig. 1, organized by the conflict (the more distant) and congruence (the closer) among the values (cf. Schwartz, 2005a). Full-size image (18 K) Fig. 1. The value circle of the Schwartz value theory: Ten broad values and four higher-order values. Adapted from Schwartz (1992). Universals in the content and structure of values: Theoretical advances and empirical tests in 20 countries. Advances in Experimental Social Psychology, 25, p. 45. Copyright 1992 by Elsevier. Adapted with permission. Hedonism shares elements of self-enhancement but is most frequently closer to openness to change, where we placed it in this study. Figure options Personality traits. According to the FFT ( McCrae and Costa, 1999 and McCrae and Costa, 2008), individual psychological differences can be divided into basic tendencies and characteristic adaptations. Personality traits are basic tendencies that “refer to more basic, abstract ways of living that are part of the human nature, and thus found in all cultures and at all times” ( McCrae, 2010, p. 58). More specifically, personality traits are enduring tendencies to behave, think, and feel in consistent ways ( McCrae and Costa, 1999 and McCrae and Costa, 2008). Cross-observer agreement is often taken as a major indication that personality traits are real, objective psychological attributes (Funder, 1995). According to McCrae and colleagues (2004), self-other agreement in personality traits “played a major role in establishing the Five-Factor Model (FFM) of personality as a widely accepted taxonomy of traits” (p. 180). Big Five personality traits are latent and therefore not directly observable; they must be inferred on the basis of thoughts, feelings, and behaviours. Furthermore, the FFT postulates that there is little, if any, influence of the environment (e.g., culture or life experiences) on the basic tendencies (Allik, 2002 and McCrae and Costa, 2008). The evidence for this comes from studies that show that the five personality factors are universal across cultures (McCrae and Costa, 1997 and McCrae et al., 2005), personality traits change little throughout life (Costa et al., 2000), and that parenting or child-rearing practices have little effect on child’s personality (e.g., Plomin, Corley, Caspi, Fulker, & DeFries, 1998). FFT views values as characteristic adaptations that help the individual to adapt to changes in the social and physical environment. Values are influenced by traits and learned in a particular times and contexts. Thus, the main conceptual difference between traits and values in the FFT is that traits are endogenous latent tendencies, shielded from the direct effects of the environment, whereas values are formed through socialization. As Olver and Mooradian (2003) put it, traits address nature and values address the interaction between nature and nurture. Recent evidence suggests, however, that values are partially heritable (Knafo and Spinath, 2011 and Schermer et al., 2008) and that common genetic factors influence values and traits (Schermer, Vernon, Maio, & Jang, 2011). Schwartz (2013) argues that genes and personality interact with the environment “to generate the substantial variance in value priorities across individuals …. [Values] serve as filters that transform the same social experience into different subjective experiences for each individual” (p. 5). This literature rejects the FFT view of values as completely lacking a direct genetic base. Moreover, there is an ongoing debate among value researchers about the causal relations between values and traits (see Fischer and Boer, submitted for publication, Nilsson, 2014 and Roccas et al., 2002). The extent to which traits and values are stable across time and contexts might contribute to observers’ success in perceiving them accurately. According to FFT theory, traits are relatively stable, and empirical research has supported this claim (Costa et al., 2000). If stability in values is lower, this would lead us to expect lower self-other agreement in values. However, test–retest correlations of values in student and adult samples, tested from 1 month to 2 years apart, were remarkably high (Schwartz, 2005b). Further, Bardi, Lee, Hofmann-Towfigh, and Soutar (2009) found only small changes in value priorities in four longitudinal studies that varied in life contexts, time gaps, and populations. Moreover, Bardi, Buchanan, Goodwin, Slabu, and Robinson (2013) found substantial stability of values over nine months among police recruits undergoing training, over three years of university among psychology and business students, and over 18 months among Polish immigrants following their immigration to Great Britain. Even among the latter, who went through a major life transition, the mean stability correlation for the 10 Schwartz values was .61 (SD = .07). Thus, traits and values may not differ enough in stability to make it easier to perceive one than the other. Values are goals that motivate behaviour, but they are not behaviours themselves. As values refer to motivation, not to action, observers must infer them indirectly (Bardi & Schwartz, 2003). Rokeach (1973) and Schwartz (1992) both suggested that it is difficult for others to infer a person’s values because a value may be expressed in a variety of relevant behaviours and because any single behaviour may express multiple values. However, Funder (1995, p.659) made the same point about traits: “…many different cues might be diagnostic of the same trait, whereas the same cues might be simultaneously diagnostic of different traits” (Funder, 1995, p. 659). Taken together, the theorizing and empirical findings provide no strong arguments for expecting a difference between traits and values in the level of self-other agreement. 1.2.2. Moderating effects of acquaintance and socially desirable responding Personality researchers (Funder, 1995 and McCrae and Costa, 2008) assume that people are usually capable of reporting on their own and other people's characteristics. McAdams (1995; McAdams & Pals, 2006) suggests that the process of getting to know someone starts with personality traits. Additional time is needed before interaction partners exchange information/cues about other psychological attributes such as attitudes or value hierarchies. Therefore, in order to allow for a fair comparison of self-other agreement in traits and values, we used well-acquainted informants in our study. Socially desirable responding may lead to inaccuracy in both self-reports and other-reports (Konstabel, Aavik, & Allik, 2006). This is especially true when reporting on characteristics with an evaluative component such as personal values and such traits as conscientiousness (Funder, 1995 and Kenny and West, 2010). Social desirability may affect responses to questionnaires both consciously (impression management) and unconsciously (self-deceptive enhancement; Paulhus, 1991). If informants and targets are close, the motivation to present the target in a more positive light than warranted by the facts may bias other-reports (see Allik, Realo, Mõttus, Borkenau, et al., 2010 and Allik, Realo, Mõttus, Esko, et al., 2010). Thus, socially desirable responding, by either targets or informants, potentially confounds self-other agreement correlations (Funder, 1995) because it says more about social norms than about the target’s personality (Konstabel et al., 2006). We return to this issue when discussing profile similarity (Section 1.3). 1.2.3. Correction for measurement error Previous research in the field of personality has identified several moderators of self-other agreement in addition to acquaintance and evaluativeness. Visibility, (assumed) similarity, and more technical aspects, such as wording or format of the questions, scale variance, and measurement error may significantly impact the magnitude of self-observer agreement (e.g., Allik, Realo, Mõttus, Borkenau, et al., 2010, Allik, Realo, Mõttus, Esko, et al., 2010 and Watson et al., 2000). We controlled one of these, random error in measurement, in order to obtain more accurate comparisons between self-other agreement in values and in personality traits. Any observation or response (e.g., a self- or other-rating) may include a large random component. This random error of measurement can be reduced by obtaining multiple responses or observations of the same construct, each with its own random error, and averaging them to generate a score for the construct (Schmidt & Hunter, 1996). Internal consistency coefficients that measure the reliability of test scores (e.g., coefficient alpha; Cronbach, 1951) can be used to estimate the proportion of true score variance in the measurement of a variable. These coefficients are a function of two parameters: (1) the average inter-correlation among the set of items in the scale that measures the construct and (2) the number of items in the scale (Simms & Watson, 2007). The study by Connolly and colleagues (2007) illustrates the impact of random error on self-other agreement in traits. When they adjusted the observed correlations for the (un)reliability of measurement, the weakest self-other agreement (agreeableness) increased from r = .30 to .46 and the strongest (extraversion) from r = .45 to .62. This suggested that the “true” associations between self- and other-ratings were considerably stronger than the observed correlations implied. It is especially important to correct for measurement error2 when comparing self-other agreement in this study because the proportion of error in the measurement of values and of personality traits is likely to differ. Previous research with the value scale we adopted from the European Social Survey (ESS), the PVQ21 (Schwartz, 2007), yielded lower alphas than the scale we use for our personality measure, the Short Five (S5; Konstabel et al., 2012). For example, the internal consistency coefficients of the ten specific values in Schwartz (2007) ranged from .36 to .70, and the Cronbach alphas of the four higher-order values ranged from .69 to .75. In contrast, the Cronbach alphas of the domain scales of the S5 typically range from .74 to .89 (Konstabel et al., 2012). The low reliabilities of the ten PVQ21 values reflects the facts that each scale has only two items (three for universalism) and that these items were constructed to cover the broad conceptual components of each value rather than to express the same narrowly defined content (Schwartz, 2005a). We report both the attenuated correlation coefficients to compare the observed levels of agreement with earlier research and the disattenuated coefficients to provide a fair comparison of traits and values. 1.3. Aim of the study The main aim of the current study is to compare self-other agreement in two aspects of personality—personal values and personality traits—using the same sample of respondents. Our study goes beyond earlier research (e.g., Lee et al., 2009) in several important respects. First, when examining self-other agreement in values and personality, we correct for the biases caused by measurement error in both value and personality scales, as proposed by Schmidt and Hunter (1996). Second, we obtain other-ratings of both values and personality traits from two observers of each target. After estimating the degree of consensus between two judges, we use the mean score of the two other-ratings for each value and personality scale in order to reduce the common method bias due to individual response biases. As Chang, Connelly, and Geeza (2012) point out, “[w]hen personality ratings are averaged across multiple raters, the variance in ratings shared across raters (i.e., trait factors) increases, whereas the variance idiosyncratic to individual raters (i.e., method factors) declines” (p. 423). This has been demonstrated in the case of personality traits: Self-other agreement was lower when self-reports were correlated with a single other-report than when they were correlated with the average of two or more other-reports (McCrae et al., 2004). However, we are aware that it is not always possible to obtain ratings from two or more informants. To allow comparisons with the results of other studies and in order to show that the obtained self-other agreement is not “simply a result of the psychometric benefits of aggregation” (Vazire & Mehl, 2008, p. 1209), we also report the self-other agreement correlations between a target and a single informant. Thirdly, the self-other agreement correlations described above reveal how accurately a rater perceives the target’s self-rating of each single value or trait. It does not reveal whether raters accurately perceive a target’s value or trait hierarchy, that is, which values or traits are more and less highly rated by the target. For this purpose, we employ profile correlations (Furr, 2008) that allow us to examine similarity in shape of the two (self-reports versus average of the two other-reports) profiles. First, we calculate correlations between all pairs of raw, unadjusted profiles of self- and other-ratings. This is called overall similarity as measured by the raw profile correlation. Accurate perception of targets’ profiles may yield high overall similarity between self- and other-reports of values and traits, but profile normativeness (Cronbach, 1955) could also cause this. In other words, any two profiles may be similar because they both reflect an average profile. Moreover, social desirability, which is also associated with normativeness, may spuriously enhance accuracy (Furr, 2008). In order to eliminate the effects of the average profile and of social desirability on profile accuracy, we z-standardize the answers of both targets and informants within the sample (means = 0; SDs = 1) and calculate correlations between any pairs of distinctive profiles. This is called distinctive similarity as it “reflects the similarity between the unique aspects of the two profiles within a pair—the degree to which one distinctive profile matches another distinctive profile” ( Furr, 2008, pp. 1277–1278). When rating values, individuals differ in the mean rating they give to the set of all values. Based on the validated assumption that the ten broad values in his theory are a reasonably comprehensive representation of the different values people hold, Schwartz (2007) suggested that the true importance individuals attribute to all values on average is similar across people. Consequently, observed differences in individuals’ average value ratings largely reflect differences in response style, including the tendency to respond in a socially desirable manner. Applications of the Schwartz value instruments, comparing individual or group value priorities or relating value priorities to other variables, typically calculate each individual’s mean rating of all value items and use it to control for differences in response style. We adopt this procedure. Mean scores across all value items also have some substantive meaning (Lönnqvist et al., 2007 and Schwartz et al., 1997). However, regardless of the extent to which individuals’ mean value ratings signify style or substance, they too are relatively stable individual characteristics. Can informants assess targets’ mean value ratings with any accuracy? Informants are unlikely to know how targets use a response scale, whether they spread their answers across the scale or concentrate them toward the upper or lower ends (see Schwartz, 2005a). It is plausible, however, that informants might have a sense of the extent to which targets describe themselves as strongly endorsing few or many values, or to which they describe themselves in a socially desirable way (Lönnqvist et al., 2007 and Schwartz et al., 1997). It is therefore interesting to assess the correlation between the average self-rating of values and the average other-rating of the target’s values. This is the forth aspect by which we differ from previous research.