- Email: [email protected]
A profile-based approach for investigating the values-personality relationship
Personality and Individual Differences 131 (2018) 111–116
Contents lists available at ScienceDirect
Personality and Individual Differences journal homepage: www.elsevier.com/locate/paid
A profile-based approach for investigating the values-personality relationship
T
Georgi P. Yankov Department of Psychology, Bowling Green State University, 206 Psychology Building, 822 E. Merry St., Bowling Green, OH 43403-0232, United States
A R T I C LE I N FO
A B S T R A C T
Keywords: Values Personality Hogan Personality Inventory Profiling Clustering Latent classes Self Assessment
The relationship between values and personality was examined with latent class cluster analysis (LCCA). This innovative method combines data exploration with the confirmation of hypothesized latent structures. In this study it isolated respondents with similar value orientations and adjusted the clusters for their personality traits. This resulted in a descriptive profile of each cluster's values-personality structure. Such profiles are theoretically meaningful with regards to the individual Self, which is espoused in values, and moderated by personality. Personality traits are psychologically superordinate, but unlike values are less cognitively transparent and useful for self-attributions. The study used a test publisher's archival dataset of 1500 respondents to two established measures of values and personality. The LCCA uncovered five latent clusters which were characterized as: Traditionalists, Maximalists, Intellectuals, Climbers, and Followers. The study describes their value-personality profiles and interprets their personal strengths and weaknesses.
1. Introduction 1.1. Rationale Rokeach (1973) believed that values and personality are related, yet hierarchically different representations of the Self. To date, only few correlational studies have explored their relationship (Bilsky & Schwartz, 1994; Olver & Mooradian, 2003; Roccas, Sagiv, Schwartz, & Knafo, 2002). However, previous studies have not provided a theory of Self to integrate both constructs, nor have they used proper methodology to explore whether personality moderates the expression of values. This study fills in these gaps and suggests a non-correlational, latent profile-based investigation of the relationship between values and personality traits. In terms of the extensively used values taxonomy of Schwartz (1992), we know that agreeableness relates negatively to power and positively to benevolence; conscientiousness relates positively to achievement, security and tradition; openness to experience relates positively to stimulation, self-direction, and universalism; and extraversion relates positively to power and achievement (Parks, 2007). Also, it appears that the relationship between values and personality revolves around three clusters of values (Schwartz, 1992) which emphasize self-enhancement (power, achievement, and hedonism), selftranscendence (benevolence, universalism), and conservation (security and tradition). This study argues for reexamining the values-personality
relationship beyond the extant method of sample-based correlations. It conceives that values and personality position at different levels of a common source — one's Self (Hogan & Hogan, 1996). According to socioanalytic theory, when people respond to personality inventories they consciously think of how to represent their idealized Selves to others (Hogan, 1995). Values might provide a more direct access to the Self and its motivations, effectively recreating the personality traits in a system of conscious practical goals (Rokeach, 1973). In essence, responding to values items makes the values salient to the Self because the individual needs to order them in importance with regards to goalmotivated behavior. Conceiving values from the vantage point of the Self makes them strictly personal. Instead of suppressing the unique values-personality relationships of each person by averaging them out on a sample basis, the study examines the relationships within homogenous groups of similar individuals. As a result, more interpretative complexity results in the relationships because they can vary across these groups. For example, a group might be defined by complimentary pairs of values and personality traits such as the altruism-agreeableness one. Finding a group of altruistic, but disagreeable individuals is possible as well, and this might be a product of other values-traits pairs such as being ambitious and valuing hedonism. Isolating these groups and generating profiles of their values-personality relationships, however, is conditional on establishing an appropriate, non-correlational, methodological linkage between values and personality. Thus, based on the above arguments, the study has three objectives. First, to reexamine the
E-mail address: [email protected]. https://doi.org/10.1016/j.paid.2018.04.031 Received 17 November 2017; Received in revised form 12 April 2018; Accepted 21 April 2018 0191-8869/ © 2018 Elsevier Ltd. All rights reserved.
Personality and Individual Differences 131 (2018) 111–116
G.P. Yankov
The second requisite is to statistically operationalize the profilebased approach. When values scores are sample-centered (that is, averaged across all respondents) they are converted into ethno-cultural values and the value hierarchy of each respondent is lost. Samplecentering might be justified for cross-cultural and exploratory research, but might not be optimal for explaining individuals' values-personality relationships. Values and personality “meet” at the individual level because of their shared relationship to the Self and how it wants to represent itself to others. Therefore, unique personality traits might relate to unique value hierarchies, and the values-personality relationship should be examined through homogeneous groups of respondents with similar value orientations. An alternative to sample-centering could be a fairly new method gaining strength in the management and occupational health literatures where personalized development trajectories are of prime interest. It is called latent class cluster analysis (LCCA) and is a model-based clustering method which assumes that the latent/independent variable is nominal and the dependent/indicator data are generated by a mixture of underlying probability distributions (Vermunt & Magidson, 2002). In standard latent class analysis (LCA) class solutions are based on the locations/means of the observed dichotomous variables within each class, whereas in LCCA the cluster solutions are based on variances and covariances which allow for different homogeneity of responses within each cluster. Moreover, LCCA extends LCA for continuous data (also known as latent profile analysis) because it allows for mixed-mode indicators (nominal, ordinal, and continuous) and covariates in the model (thus making it similar to the MIMIC model in factor analysis; Vermunt, 2010). LCCA produces fewer clusters because it allows for local dependencies between pairs of indicators (e.g. power and achievement values forming a self-enhancement group of values). This dependency between indicators results in better classifications within clusters because the shared information between indicators is retained, which with regards to this study means that the values clusters will not be dominated by only one value. In essence, the proposed LCCA method might: 1) adjust for continuously measured values items, 2) derive discrete latent clusters of respondents with similar value hierarchies, 3) limit the possibility of one value dominating each cluster, and 4) allow for including personality as a regression-based covariate in the estimation of the cluster solutions. The inclusion of personality covariates recognizes the assumption that personality is the nomologically superordinate psychological factor that is reflected in one's espoused personal values. For example, a sample-centered design might show that the value of power correlates negatively with agreeableness and positively with assertiveness across all personality factors and facets (e.g., Roccas et al., 2002). This unsurprising finding, however, might be “what is left” of the individual values-personality relationships after they are sample-averaged. For example, with LCCA we might uncover a latent cluster characterized by respondents valuing high power, but also high tradition and affiliation. These individuals would disprove the correlational approach finding that people valuing power have elevated disagreeableness. Therefore, profiling with latent clusters is essentially a typological approach that aims at an alternative and more conceptually rich scheme for personality assessment.
values-personality relationship with regard to the Self. Second, to derive person-centered profiles of the relationship. Third, to examine these profiles in terms of extant values-personality research. 1.2. The values-personality relationship Values can regulate the expression of traits by being representations of abstract, context-free, and consciously selected terminal goals (Roccas et al., 2002). For example, regarding the value of honesty, we should espouse it equally in the companies of honest or dishonest people, today, and years from today. This is because during the childhood socialization individuals learn honesty as something we should do in an all-or-none manner, regardless of the circumstances (Rokeach, 1973). Values also represent group and cultural forces, and are learned in stages during early human socialization (Kohlberg, 1971). Because of this process of inculcation and the enduring societal pressure, values are more stable than attitudes or interests. In public settings, what anthropologists have called “the social requiredness” and “the desirable” (Kluckhohn, 1951), makes people conform to dominant cultural values. In private life, values form and elaborate our internalized ideal Self. It sanctions our behavior by provoking guilt and shame, and thus ensures compliance with values. Being the most abstract layer of the conscious ideal Self (Wojciszke, 1989), values can activate behavior when the respective value is central to one's Self-concept (Verplanken & Holland, 2002). This process becomes apparent and verbally problematized in moral dilemmas (Lefkowitz, 2006). In comparison to values, traits do not have prescriptive nature; they are dispositions to show endogenous patterns/temperaments of thoughts, feelings, and actions (McCrae & Costa, 1990). Whereas values are measured in relation to their importance to the Self, traits can indicate positive or negative adjustment. It is probable that individuals attribute valence to traits through their value orientations, which might temper the behavioral expression of traits (Parks & Guay, 2009). This is because of two reasons. First, traits are frequently conceptualized and measured in terms of specific behaviors, and, therefore, they are not used as elaborated justifications for goal-directed behavior. Second, some personality traits and their relevant behaviors are undesirable and kept in check (e.g. hostility) because our cultural and personal values (e.g. benevolence) require us so. A hostile person can choose to suppress their hostility and be benevolent and respectful to the well-being of others. Alternatively, this person can justify hostile behavior because he or she highly endorses other two values — power and achievement. Thus, when salient to the Self, values provide for the conscious attribution and explanation of behavior, but in actuality the less salient to the Self personality traits might be driving automatic everyday behavior. 1.3. A profile-based approach for the values-personality relationship The desired profile-based approach depends on establishing reliable psychometric procedures that guarantee that responses reflect the values hierarchy of the Self and its relationship to personality. Specifically, the procedures have to guarantee not only that 1) reported values relate in various strength to the Self, but also that 2) a method exists for the statistical profiling of values and personality. Regarding the first requisite, Rokeach (1973) assumed that individuals respond to values items by comparing them for concordance with their Self-image. As a result, two measurement issues have risen regarding the responding to values items. First, some social desirability is expected but latest research indicates that it is not equivalent with error variance and can have substantive meaning depending on the examined value (Fisher & Katz, 2008). Second, ranking values on their importance in one's life, might provide better differentiation and validity over rating each value for its own importance in one's life (Krosnick & Alwin, 1988). However, current research is much less conclusive on the preferable approach (Maio, Roese, Seligman, & Katz, 1996).
2. Method 2.1. Sample Hogan Assessments allowed the author to use for research-purposes a dataset of 1500 US individuals who responded to two instruments: the Motives, Values, Preferences Inventory (MVPI; Hogan & Hogan, 1996) and the Hogan Personality Inventory (HPI; Hogan & Hogan, 1995). The sample was 30% female and 39.6 years (SD = 10.98) old on average. Schwartz and Rubel (2005) uncovered gender differences among values but they were of low effect size and varied across cultures. Given that 112
Personality and Individual Differences 131 (2018) 111–116
G.P. Yankov
Table 1 Descriptive statistics and correlations. Adjustment Aesthetics Affiliation Altruism Commercial Hedonism Power Recognition Scientific Security Tradition Mean SD ⁎ ⁎⁎
−0.01 0.21⁎⁎ 0.07⁎ 0.12⁎⁎ −0.14⁎⁎ 0.07⁎ −0.13⁎⁎ −0.07⁎ 0.00 0.13⁎⁎ 2.54 0.81
Ambition ⁎⁎
0.09 0.30⁎⁎ 0.16⁎⁎ 0.21⁎⁎ 0.01 0.23⁎⁎ 0.18⁎⁎ 0.11⁎⁎ −0.09⁎⁎ 0.06⁎ 2.66 0.63
Sociability ⁎⁎
0.12 0.56⁎⁎ 0.16⁎⁎ 0.18⁎⁎ 0.21⁎⁎ 0.20⁎⁎ 0.29⁎⁎ 0.07⁎ −0.04 0.06⁎ 1.64 1.02
Interpersonal 0.04 0.14⁎⁎ 0.13⁎⁎ −0.03 −0.16⁎⁎ −0.15⁎⁎ −0.19⁎⁎ −0.09⁎⁎ 0.02 0.03 2.08 0.80
Prudence
Inquisitive
−0.02 0.00 0.09⁎⁎ −0.01 −0.23⁎⁎ −0.13⁎⁎ −0.19⁎⁎ −0.09⁎⁎ 0.21⁎⁎ 0.11⁎⁎ 1.95 0.89
⁎⁎
0.17 0.07⁎ 0.16⁎⁎ 0.13⁎⁎ −0.03 0.11⁎⁎ 0.07⁎⁎ 0.40⁎⁎ −0.09⁎⁎ 0.07⁎ 2.36 0.75
Learning ⁎⁎
0.17 0.06⁎ 0.18⁎⁎ 0.05 −0.04 0.07⁎⁎ 0.00 0.21⁎⁎ 0.01 0.07⁎⁎ 2.00 0.89
Mean
SD
5.13 6.91 6.86 7.21 6.14 6.81 5.68 6.19 6.11 6.57
1.87 1.55 1.67 1.76 1.92 1.69 1.79 2.00 1.88 1.75
p < 0.05. p < 0.01.
2.3. Procedure
the sample was US-based and fairly large, such differences might be minimal.
The program Latent GOLD 5.1 (Vermunt & Magidson, 2016) generated the cluster solutions. The HPI scale scores served as covariates in the LCCA model. The study followed a 3-step analysis because it is more intuitive to first build a latent class model, and then relate it to predictors and covariates (Vermunt & Magidson, 2016). The step 3 submodule in Latent GOLD implements a bias adjustment which eliminates the threat of underestimating parameter estimates in this step (Bakk, Tekle, & Vermunt, 2013). In essence, the 3-step analysis starts by estimating a latent class model. Then cases are assigned to latent classes, and this classification information is saved. In the third step the latent classification scores saved in step 2 are regressed on the covariate variables correcting for the misclassification in the second step error to prevent bias (Asparouhov & Muthén, 2014).
2.2. Measures 2.2.1. Values The Schwartz Value Survey (Schwartz, 1992) is a widely used instrument for measuring personal values in cross-cultural psychology. It was developed with multidimensional scaling of international samples of students and teachers. Its exploratory nature and sample specificity might not guarantee that the measured values have sufficient construct validity. In contrast, the MVPI is a theoretically-grounded and empirically validated instrument which is also extensively used for values assessment. The MVPI is more amenable to the proposed LCCA methodology because it has reliable values scales for the maximum-likelihood clustering algorithms. The study used MVPI's ten beliefs scales measuring “shoulds”, ideals, and life goals. Their names follow the ten broad MVPI factors: Aesthetics, Affiliation, Altruism, Commerce, Hedonism, Power, Recognition, Science, Security, and Tradition.
3. Results 3.1. Descriptives Table 1 presents the descriptive statistics of the study's variables. Notable is the stable pattern of correlations between Ambition and Sociability and all values with the exclusion of the insignificant Hedonism-Ambition and Security-Sociability pairs. As expected, the selfenhancement values of Hedonism, Power, and Recognition relate negatively to Interpersonal Sensitivity, but even more so to Prudence. The values of Aesthetics and Science have stable relationships to the Inquisitive and Learning Approach. In general, two clusters of values have the highest and most consistent patterns of relationships to personality — the self-transcendence (Affiliation and Altruism) and the self-enhancement values. The conservation values (Security and Tradition) appear to relate most substantially only with Prudence.
2.2.2. Personality The HPI is extensively used by the HR industry because of its wellresearched validity for various jobs, and because it generally conforms to the authoritative Five-Factor Model of personality (FFM; John & Srivastava, 1999). HPI has seven personality factors: Adjustment, Ambition, Sociability, Interpersonal Sensitivity, Prudence, Inquisitive, and Learning Approach. In broad FFM language, Ambition and Sociability form Extraversion, Adjustment is Neuroticism, Prudence is Conscientiousness, Interpersonal Sensitivity is Agreeableness, and Learning Approach is an additional factor not captured by Openness to Experience and its HPI equivalent of Inquisitive. However, Smith, Hanges and Dickson (2001) found that HPI maps better for some FFM factors, and the desirable equivalence should reflect the HPI facets, called homogenous item clusters (HICs). For example, Adjustment's Trusting and Empathy HICs might relate better to Agreeableness, whereas Interpersonal Sensitivity's No Hostility HIC to Neuroticism. The archival dataset contained three items from each scale, with sufficient reliability and validity to include in research as markers for full-scale constructs. The corrected correlations (Levy, 1967) between the original long scales and the short, research-use scales used in the study are: Aesthetics (0.83), Affiliation (0.79), Altruism (0.82), Commerce (0.75), Hedonism (0.80), Power (0.75), Recognition (0.81), Science (0.83), Security (0.80), Tradition (0.75), Adjustment (0.45), Ambition (0.56), Sociability (0.78), Interpersonal Sensitivity (0.59), Prudence (0.66), Inquisitive (0.67), and Learning Approach (0.78). Also, because HPI uses a dichotomous (True/False) response scale, and the MVPI a trichotomous (Agree/Disagree/Neutral) scale, HPI scale scores varied between 0 and 3, and MVPI scale scores between 3 and 9.
3.2. Cluster solution The number of clusters was determined through the complex models-penalizing Bayesian Information Criterion (BIC), the sample size-penalizing Consistent Akaike Information Criterion (CAIC), the parsimony of the model, its error terms, and its substantive explanatory value (Celeux, Biernacki, & Govaert, 1997). Table 2 presents the fit statistics of six initial exploratory models. The 5-cluster model shows the greatest decrement in BIC and CAIC. However, the chi-square distributed L2 statistic, which indicates the unexplained amount of association among the variables after the estimation, needs to be as low as possible, and it started tapering off from the 4-cluster model. Also, parameters-wise, the 4-cluster model was more parsimonious. Both models' p-values were significantly lower than 0.001 whereas the best model should have the largest and insignificant 113
Personality and Individual Differences 131 (2018) 111–116
G.P. Yankov
Table 2 Fits of the initial cluster solutions.
Model Model Model Model Model Model
1 2 3 4 5 6
No. clusters
BIC(LL)
CAIC(LL)
No. par
L2
df
p-Value
Class. err.
1-Cluster 2-Cluster 3-Cluster 4-Cluster 5-Cluster 6-Cluster
51,328.68 50,831.96 50,748.99 50,686.06 50,651.41 50,660.03
51,388.68 50,902.96 50,830.99 50,779.06 50,755.41 50,775.03
60.00 71.00 82.00 93.00 104.00 115.00
29,961.20 29,384.48 29,221.50 29,078.57 28,963.91 28,892.53
1381.00 1370.00 1359.00 1348.00 1337.00 1326.00
3.3e−5287 4.1e−5175 1.1e−5148 2.0e−5126 5.0e−5110 1.4e−5102
0.00 0.14 0.22 0.24 0.26 0.26
Fig. 1. Profile plots of the five clusters based on values.
p-value. Bootstrapping the L2's p-value provides better estimation because the assumption of its chi-squared distribution is relaxed. Bootstrapping with 500 iterations resulted in p's of 0.34 (5-cluster model) and 0.37 (4-cluster model). It seemed that the 4-cluster model was superior, but its residuals had more values substantially larger than 1 suggesting that the model fell short of reproducing the association between the variables. The 5-cluster model's residuals were also inflated but presented a more balanced pattern especially for the value of Aesthetics whose residuals averaged 1.84 vs. 5.11 for the 4-cluster model. Thus, apart from being slightly less parsimonious (i.e. having more parameters to estimate) and misclassifying 2% more cases than the 4-cluster model (the last column in Table 2), the 5-cluster model was the better model in terms of general fit.1 After randomly splitting the sample, the 5-cluster solution was the most robust in the two subsamples.
tests, it seems that cluster 5 is very similar to cluster 3 in terms of the self-enhancement values, to cluster 4 in terms of conservation values, and to cluster 1 in terms of Aesthetics and Science. The lower half of Table 3 (re-expressed in Fig. 2) displays the covariates means after adjusting the clusters for the personality covariation. The LCCA's third step models the probabilities for each personality trait to characterize the clusters, not the personality traits as dependent variables. Therefore, no SEs are available for the covariates means. The covariates Wald tests were all significant but for ambition (p = 0.07). Thus, all other covariates significantly contributed to the model's ability to discriminate between the clusters. In terms of substantive interpretation, Adjustment did not discriminate well between clusters but for the low value for cluster 4. Cluster 1 is dominated by high Prudence, but low Sociability and Inquisitive; cluster 2 by very high Sociability, Inquisitive and Learning Approach; cluster 3 by very high Interpersonal Sensitivity and Inquisitive; cluster 4 by lower Adjustment, Interpersonal Sensitivity, and rather low Prudence, but high Sociability; cluster 5 by very low Sociability, Inquisitive and Learning Approach, but high Interpersonal Sensitivity. Notably, cluster 5 is alike cluster 1 but for its low Prudence. Also, cluster 3 is alike cluster 2, but has lower Sociability and higher Interpersonal Sensitivity. Below are the suggested profiles of each cluster based on conjoint analyses of their values and personality means. They are interpreted in the language of the FFM, but taking into account the specificity of HPI.
3.3. Personality covariation The upper half of Table 3 (re-expressed in Fig. 1) provides the clusters' values scores from the first estimation step. The Wald tests for each value across the five clusters indicate which clusters means significantly differ. Cluster 1 is dominated by the values of Commerce, Power, Security, and Tradition; cluster 2 by all values but for Aesthetics and to some extent Hedonism; cluster 3 by Altruism, Commerce, and Science; cluster 4 by Affiliation, Commerce, Hedonism, and Power; cluster 5 by Affiliation and Commerce. Given the Wald significance
• Cluster 1/Traditionalists value the conservation and self-enhance-
ment values of commerce, power, security, and tradition. They do not value aesthetics, recognition, nor science. They are the least extroverted and most self-controlled and moralistic cluster; they are
1
Personal communication with LATENTGOLD's creators Jeroen Vermunt and Jay Magidson (August 2017).
114
Personality and Individual Differences 131 (2018) 111–116
G.P. Yankov
Table 3 Parameters of the 5-cluster model. Cluster 1
Cluster 2
Cluster 3
Cluster 4
0.26
0.26
0.20
0.17
Cluster size
Cluster 5 0.11
First step
Mean
SE
Mean
SE
Mean
SE
Mean
SE
Mean
SE
Aesthetics Affiliation Altruism Commercial Hedonism Power Recognition Scientific Security Tradition
4.07 6.46 6.48 7.28 6.19 6.92 5.45 5.28 7.20 7.33
0.12 0.13 0.14 0.13 0.15 0.16 0.14 0.16 0.16 0.13
5.81 7.73 8.00 8.07 6.76 7.87 6.69 7.05 6.50 7.54
0.14 0.09 0.09 0.09 0.12 0.09 0.12 0.13 0.13 0.11
6.34 6.51 7.44 6.60 5.00 5.75 4.85 6.83 5.55 6.33
0.18 0.13 0.15 0.17 0.17 0.17 0.15 0.17 0.16 0.15
4.83 7.27 5.98 7.25 7.12 7.34 6.29 6.33 5.02 5.04
0.18 0.14 0.19 0.17 0.21 0.17 0.22 0.22 0.19 0.23
4.20 6.24 5.23 6.00 5.16 5.18 4.43 4.83 5.22 5.18
0.21 0.20 0.22 0.27 0.23 0.23 0.20 0.26 0.26 0.26
Third step
Mean
Mean
Mean
Mean
Mean
Adjustment Ambition Sociability Interpersonal sensitivity Prudence Inquisitive Learning approach
2.61 2.51 1.08 2.12 2.32 2.01 1.79
2.70 2.93 2.40 2.07 1.94 2.63 2.30
2.65 2.70 1.53 2.33 2.05 2.67 2.23
2.32 2.83 2.20 1.68 1.27 2.41 1.93
2.58 2.40 1.21 2.28 2.01 1.97 1.57
•
1–5 1–3/5; 3–5 1–3 1–4; 3–5 2–4; 3–5 1–4; 3–5 2–4; 3–5 1–5; 2–3; 3–4 3–5; 4–5 1–2; 3–4; 4–5
1–2/3/5; 2–3/5; 3–5 1–3/5; 2–4; 3–4/5 1–3/5; 2–4 1–2/3/5; 3–5 2–3; 3–5 1–5; 2–3 1–4/5; 2–3; 3–4
and temperamental too.
low on Openness to Experience.
• Cluster 2/Maximalists has the highest values loadings from all five, •
Cluster pairs with non-sig. (p > 0.05) Wald tests for means differences
• Cluster 5/Followers are a mix of the strengths and weaknesses of the
and its respondents are “high-valuers.” Maximalists are the most extraverted cluster, and are very open to experience and academically capable. Cluster 3/Intellectuals value altruism, and science, but not the values of hedonism, power, or recognition. Being so humble, they are the easiest to live with, sensitive, non-aggressive, and intellectual cluster. They might be the less extraverted and self-centered type of Maximalists. Cluster 4/Climbers are alike Maximalists in their quest for self-enhancement, and very different than the Traditionalists' emphasis on conservation. Climbers are the least conscientious cluster, they cooperate and understand others with difficulty, and might be anxious
other clusters. They are as humble as the Intellectuals, but do not share their openness; are emotionally stable and introverted as the Traditionalists, but not nearly as self-controlled and moralistic as them; like the Climbers disregard security and tradition, but unlike them cooperate easily.
4. Discussion Only recently have psychologists started to profile with LCCA (e.g. Lawrence and Zyphur, 2011). In this study, it produced a result that is descriptively and interpretatively richer than the sample-based correlations between values and personality traits. With LCCA values and
Fig. 2. Profile plots of the five clusters with the personality covariates. Note: no scale scores were below 1, hence the scaling axis starts from 1. 115
Personality and Individual Differences 131 (2018) 111–116
G.P. Yankov
Acknowledgements
personality are dynamically integrated within the ideal Self. The latter safeguards the suggested profiles from dustbowl empiricism. This study claimed that values organize and “express” the Self, but personality, being psychologically superordinate and thus less transparent to the Self, moderates the expression of values. Thus, we can interpret the five profiles from a value and from a personality-based perspective, but insight is conditional on a conjoint analysis. Values, being more face-valid and cognitively assessable, might be central to the Self, but personality characteristics would probably sway behavior in social contexts where values might conflict (i.e. dilemmas) and individuals need to act. Thus, personality becomes the superordinate factor which often decides the course of our “valued” actions. For example, the Traditionalists endorse tangible rewards, rules, and status. Knowing that they are not very sociable, but are very controlled and rule-bound, they might not thrive in very innovative and peoplerelated jobs. Maximalists are high valuers, but might not thrive in workplaces that do not match their sociability and intellect. The Intellectuals value science and harmony. However, these quiet individuals might become unproductive among very extroverted and competitive coworkers. The Climbers want to be powerful, rich, and liked. Yet, to be productive at work and private life, they should be more cooperative and sensitive, less impulsive, obey the rules and follow society's norms. The Followers, although limited in size, seem to mostly want to be in people's good books. Personality-wise, this “silent class” should focus on building their assertiveness, and start learning, organizing, and working on goals that would grow their self-efficacy and expertise. Given the cluster sizes, almost three fourths of respondents are either Traditionalists, Maximalists, or Intellectuals. Climbers, and especially Followers, are limited in size and show more unfavorable valuespersonality profiles, which is in tune with the assumption that variation in the value-personality profiles can become more and more idiosyncratic when the number of isolated clusters increases. This has implications for extant values-personality research. Instead of replicating Schwartz's (1992) values hierarchy this study indicated that individuals not only have clear-cut value preferences like the Traditionalists (conservation values), but also can espouse values from conflicting groups like the Maximalists (self-enhancement and self-transcendence values). This is a further proof that the values hierarchy is personal, and appraising it with zero-order correlations on a sample basis might conceal substantial inter-individual idiosyncrasies. This study poses several caveats. Statistically, cross-validation with larger population-representative samples is required to verify the five profiles. This is because clustering algorithms overfit the sample and do not hypothesize latent structures that generalize to the population. Also, other profiling methods such as grade-of-membership analysis (Woodbury & Manton, 1982) can be used for triangulation. Exploring the profiles with facet-level personality covariates would also convey a richer, more discriminative interpretation. Knowing this information would be helpful for understanding the consequences of hiring individuals in the suggested profiles. Practically, by profiling individuals conjointly on values and personality, selection practitioners might assess more fully the person-organization fit of applicants. That is, both candidate personality and value fit would be considered. Also, talent managers might use the profiles to design specific developmental trajectories and trainings. The last two could target the potential personality derailers of each profile. By exploring the Self from a values-personality perspective we might help individuals to better align what they value and want to be with who they are.
I would like to thank Mike Zickar and Jeff Foster for their help and advice. References Asparouhov, T., & Muthén, B. (2014). Auxiliary variables in mixture modeling: Three-step approaches using M plus. Structural Equation Modeling, 21(3), 329–341. Bakk, Z., Tekle, F., & Vermunt, J. (2013). Estimating the association between latent class membership and external variables using bias-adjusted three-step approaches. Sociological Methodology, 43(1), 272–311. Bilsky, W., & Schwartz, S. (1994). Values and personality. European Journal of Personality, 8(3), 163–181. Celeux, G., Biernacki, C., & Govaert, G. (1997). Choosing models in model-based clustering and discriminant analysis. Technical report. Rhone-Alpes: INRIA. Fisher, R., & Katz, J. (2008). Social-desirability bias and the validity of self-reported values. Psychology & Marketing, 17(2), 105–120. Hogan, R. (1995). A socioanalytic perspective on the Five-Factor Model. In J. S. Wiggins (Ed.). Theories of the Five-Factor Model. New York: Guilford. Hogan, J., & Hogan, R. (1995). Hogan personality inventory manual (2nd ed.). Tulsa, OK: HAS. Hogan, R., & Hogan, J. (1996). Motives, values, and preferences inventory manual. Tulsa, OK: HAS. John, O., & Srivastava, S. (1999). The big five trait taxonomy: History, measurement, and theoretical perspectives. In L. Pervin, & O. John (Eds.). Handbook of personality: Theory and research (pp. 102–138). (2nd ed.). New York: Guilford. Kluckhohn, C. (1951). Values and value-orientations in the theory of action. In T. Parsons, & E. Shils (Eds.). Toward a general theory of action. Cambridge, MA: Harvard University Press. Kohlberg, L. (1971). Stages of moral development. Moral Education, 1, 23–92. Krosnick, J., & Alwin, D. (1988). A test of the form-resistant correlation hypothesis: Ratings, rankings, and the measurement of values. Public Opinion Quarterly, 52(4), 526–538. Lawrence, B., & Zyphur, M. (2011). Identifying organizational faultlines with latent class cluster analysis. Organizational Research Methods, 14(1), 32–57. Lefkowitz, J. (2006). The constancy of ethics amidst the changing world of work. Human Resource Management Review, 16, 245–268. Levy, P. L. (1967). The correction for spurious correlation in the evaluation of short-form tests. Journal of Clinical Psychology, 23(1), 84–86. Maio, G., Roese, N., Seligman, C., & Katz, A. (1996). Rankings, ratings, and the measurement of values: Evidence for the superior validity of ratings. Basic and Applied Social Psychology, 18(2), 171–181. McCrae, R., & Costa, P. (1990). Personality in adulthood. New York: Guilford Press. Olver, J., & Mooradian, T. (2003). Personality traits and personal values: A conceptual and empirical integration. Personality and Individual Differences, 35(1), 109–125. Parks, L. (2007). Personality and values: A meta-analysis. Paper presented at the annual conference for the Society of Industrial and Organizational Psychology. New York: New York. Parks, L., & Guay, R. (2009). Personality, values, and motivation. Personality and Individual Differences, 47(7), 675–684. Roccas, S., Sagiv, L., Schwartz, S., & Knafo, A. (2002). The big five personality factors and personal values. Personality and Social Psychology Bulletin, 28(6), 789–801. Rokeach, M. (1973). The nature of human values. New York: Free Press. Schwartz, S. (1992). Universals in the content and structure of values: Theoretical advances and empirical tests in 20 countries. Advances in Experimental Social Psychology, 25, 1–65. Schwartz, S., & Rubel, T. (2005). Sex differences in value priorities: Cross-cultural and multimethod studies. Journal of Personality and Social Psychology, 89(6), 1010–1028. Smith, B., Hanges, J., & Dickson, W. (2001). Personnel selection and the five-factor model: Reexamining the effects of applicant's frame of reference. Journal of Applied Psychology, 86(2), 304–315. Vermunt, J. (2010). Latent class modeling with covariates: Two improved three-step approaches. Political Analysis, 18(4), 450–469. Vermunt, J., & Magidson, J. (2002). Latent class cluster analysis. Applied Latent Class Analysis, 11, 89–106. Vermunt, J., & Magidson, J. (2016). Technical guide for latent Gold 5.1: Basic, advanced, and syntax. Belmont, MA: Statistical Innovations, Inc. Verplanken, B., & Holland, R. (2002). Motivated decision making: Effects of activation and self-centrality of values on choices and behavior. Journal of Personality and Social Psychology, 82(3), 434–447. Wojciszke, B. (1989). The system of personal values and behavior. In N. Eisenberg, J. Reykowski, & E. Staub (Eds.). Social and moral values: Individual and societal perspectives (pp. 229–252). Hillsdale, NJ: Erlbaum. Woodbury, M., & Manton, K. (1982). A new procedure for analysis of medical classification. Methods Archive, 21, 210–220.
116
Contents lists available at ScienceDirect
Personality and Individual Differences journal homepage: www.elsevier.com/locate/paid
A profile-based approach for investigating the values-personality relationship
T
Georgi P. Yankov Department of Psychology, Bowling Green State University, 206 Psychology Building, 822 E. Merry St., Bowling Green, OH 43403-0232, United States
A R T I C LE I N FO
A B S T R A C T
Keywords: Values Personality Hogan Personality Inventory Profiling Clustering Latent classes Self Assessment
The relationship between values and personality was examined with latent class cluster analysis (LCCA). This innovative method combines data exploration with the confirmation of hypothesized latent structures. In this study it isolated respondents with similar value orientations and adjusted the clusters for their personality traits. This resulted in a descriptive profile of each cluster's values-personality structure. Such profiles are theoretically meaningful with regards to the individual Self, which is espoused in values, and moderated by personality. Personality traits are psychologically superordinate, but unlike values are less cognitively transparent and useful for self-attributions. The study used a test publisher's archival dataset of 1500 respondents to two established measures of values and personality. The LCCA uncovered five latent clusters which were characterized as: Traditionalists, Maximalists, Intellectuals, Climbers, and Followers. The study describes their value-personality profiles and interprets their personal strengths and weaknesses.
1. Introduction 1.1. Rationale Rokeach (1973) believed that values and personality are related, yet hierarchically different representations of the Self. To date, only few correlational studies have explored their relationship (Bilsky & Schwartz, 1994; Olver & Mooradian, 2003; Roccas, Sagiv, Schwartz, & Knafo, 2002). However, previous studies have not provided a theory of Self to integrate both constructs, nor have they used proper methodology to explore whether personality moderates the expression of values. This study fills in these gaps and suggests a non-correlational, latent profile-based investigation of the relationship between values and personality traits. In terms of the extensively used values taxonomy of Schwartz (1992), we know that agreeableness relates negatively to power and positively to benevolence; conscientiousness relates positively to achievement, security and tradition; openness to experience relates positively to stimulation, self-direction, and universalism; and extraversion relates positively to power and achievement (Parks, 2007). Also, it appears that the relationship between values and personality revolves around three clusters of values (Schwartz, 1992) which emphasize self-enhancement (power, achievement, and hedonism), selftranscendence (benevolence, universalism), and conservation (security and tradition). This study argues for reexamining the values-personality
relationship beyond the extant method of sample-based correlations. It conceives that values and personality position at different levels of a common source — one's Self (Hogan & Hogan, 1996). According to socioanalytic theory, when people respond to personality inventories they consciously think of how to represent their idealized Selves to others (Hogan, 1995). Values might provide a more direct access to the Self and its motivations, effectively recreating the personality traits in a system of conscious practical goals (Rokeach, 1973). In essence, responding to values items makes the values salient to the Self because the individual needs to order them in importance with regards to goalmotivated behavior. Conceiving values from the vantage point of the Self makes them strictly personal. Instead of suppressing the unique values-personality relationships of each person by averaging them out on a sample basis, the study examines the relationships within homogenous groups of similar individuals. As a result, more interpretative complexity results in the relationships because they can vary across these groups. For example, a group might be defined by complimentary pairs of values and personality traits such as the altruism-agreeableness one. Finding a group of altruistic, but disagreeable individuals is possible as well, and this might be a product of other values-traits pairs such as being ambitious and valuing hedonism. Isolating these groups and generating profiles of their values-personality relationships, however, is conditional on establishing an appropriate, non-correlational, methodological linkage between values and personality. Thus, based on the above arguments, the study has three objectives. First, to reexamine the
E-mail address: [email protected]. https://doi.org/10.1016/j.paid.2018.04.031 Received 17 November 2017; Received in revised form 12 April 2018; Accepted 21 April 2018 0191-8869/ © 2018 Elsevier Ltd. All rights reserved.
Personality and Individual Differences 131 (2018) 111–116
G.P. Yankov
The second requisite is to statistically operationalize the profilebased approach. When values scores are sample-centered (that is, averaged across all respondents) they are converted into ethno-cultural values and the value hierarchy of each respondent is lost. Samplecentering might be justified for cross-cultural and exploratory research, but might not be optimal for explaining individuals' values-personality relationships. Values and personality “meet” at the individual level because of their shared relationship to the Self and how it wants to represent itself to others. Therefore, unique personality traits might relate to unique value hierarchies, and the values-personality relationship should be examined through homogeneous groups of respondents with similar value orientations. An alternative to sample-centering could be a fairly new method gaining strength in the management and occupational health literatures where personalized development trajectories are of prime interest. It is called latent class cluster analysis (LCCA) and is a model-based clustering method which assumes that the latent/independent variable is nominal and the dependent/indicator data are generated by a mixture of underlying probability distributions (Vermunt & Magidson, 2002). In standard latent class analysis (LCA) class solutions are based on the locations/means of the observed dichotomous variables within each class, whereas in LCCA the cluster solutions are based on variances and covariances which allow for different homogeneity of responses within each cluster. Moreover, LCCA extends LCA for continuous data (also known as latent profile analysis) because it allows for mixed-mode indicators (nominal, ordinal, and continuous) and covariates in the model (thus making it similar to the MIMIC model in factor analysis; Vermunt, 2010). LCCA produces fewer clusters because it allows for local dependencies between pairs of indicators (e.g. power and achievement values forming a self-enhancement group of values). This dependency between indicators results in better classifications within clusters because the shared information between indicators is retained, which with regards to this study means that the values clusters will not be dominated by only one value. In essence, the proposed LCCA method might: 1) adjust for continuously measured values items, 2) derive discrete latent clusters of respondents with similar value hierarchies, 3) limit the possibility of one value dominating each cluster, and 4) allow for including personality as a regression-based covariate in the estimation of the cluster solutions. The inclusion of personality covariates recognizes the assumption that personality is the nomologically superordinate psychological factor that is reflected in one's espoused personal values. For example, a sample-centered design might show that the value of power correlates negatively with agreeableness and positively with assertiveness across all personality factors and facets (e.g., Roccas et al., 2002). This unsurprising finding, however, might be “what is left” of the individual values-personality relationships after they are sample-averaged. For example, with LCCA we might uncover a latent cluster characterized by respondents valuing high power, but also high tradition and affiliation. These individuals would disprove the correlational approach finding that people valuing power have elevated disagreeableness. Therefore, profiling with latent clusters is essentially a typological approach that aims at an alternative and more conceptually rich scheme for personality assessment.
values-personality relationship with regard to the Self. Second, to derive person-centered profiles of the relationship. Third, to examine these profiles in terms of extant values-personality research. 1.2. The values-personality relationship Values can regulate the expression of traits by being representations of abstract, context-free, and consciously selected terminal goals (Roccas et al., 2002). For example, regarding the value of honesty, we should espouse it equally in the companies of honest or dishonest people, today, and years from today. This is because during the childhood socialization individuals learn honesty as something we should do in an all-or-none manner, regardless of the circumstances (Rokeach, 1973). Values also represent group and cultural forces, and are learned in stages during early human socialization (Kohlberg, 1971). Because of this process of inculcation and the enduring societal pressure, values are more stable than attitudes or interests. In public settings, what anthropologists have called “the social requiredness” and “the desirable” (Kluckhohn, 1951), makes people conform to dominant cultural values. In private life, values form and elaborate our internalized ideal Self. It sanctions our behavior by provoking guilt and shame, and thus ensures compliance with values. Being the most abstract layer of the conscious ideal Self (Wojciszke, 1989), values can activate behavior when the respective value is central to one's Self-concept (Verplanken & Holland, 2002). This process becomes apparent and verbally problematized in moral dilemmas (Lefkowitz, 2006). In comparison to values, traits do not have prescriptive nature; they are dispositions to show endogenous patterns/temperaments of thoughts, feelings, and actions (McCrae & Costa, 1990). Whereas values are measured in relation to their importance to the Self, traits can indicate positive or negative adjustment. It is probable that individuals attribute valence to traits through their value orientations, which might temper the behavioral expression of traits (Parks & Guay, 2009). This is because of two reasons. First, traits are frequently conceptualized and measured in terms of specific behaviors, and, therefore, they are not used as elaborated justifications for goal-directed behavior. Second, some personality traits and their relevant behaviors are undesirable and kept in check (e.g. hostility) because our cultural and personal values (e.g. benevolence) require us so. A hostile person can choose to suppress their hostility and be benevolent and respectful to the well-being of others. Alternatively, this person can justify hostile behavior because he or she highly endorses other two values — power and achievement. Thus, when salient to the Self, values provide for the conscious attribution and explanation of behavior, but in actuality the less salient to the Self personality traits might be driving automatic everyday behavior. 1.3. A profile-based approach for the values-personality relationship The desired profile-based approach depends on establishing reliable psychometric procedures that guarantee that responses reflect the values hierarchy of the Self and its relationship to personality. Specifically, the procedures have to guarantee not only that 1) reported values relate in various strength to the Self, but also that 2) a method exists for the statistical profiling of values and personality. Regarding the first requisite, Rokeach (1973) assumed that individuals respond to values items by comparing them for concordance with their Self-image. As a result, two measurement issues have risen regarding the responding to values items. First, some social desirability is expected but latest research indicates that it is not equivalent with error variance and can have substantive meaning depending on the examined value (Fisher & Katz, 2008). Second, ranking values on their importance in one's life, might provide better differentiation and validity over rating each value for its own importance in one's life (Krosnick & Alwin, 1988). However, current research is much less conclusive on the preferable approach (Maio, Roese, Seligman, & Katz, 1996).
2. Method 2.1. Sample Hogan Assessments allowed the author to use for research-purposes a dataset of 1500 US individuals who responded to two instruments: the Motives, Values, Preferences Inventory (MVPI; Hogan & Hogan, 1996) and the Hogan Personality Inventory (HPI; Hogan & Hogan, 1995). The sample was 30% female and 39.6 years (SD = 10.98) old on average. Schwartz and Rubel (2005) uncovered gender differences among values but they were of low effect size and varied across cultures. Given that 112
Personality and Individual Differences 131 (2018) 111–116
G.P. Yankov
Table 1 Descriptive statistics and correlations. Adjustment Aesthetics Affiliation Altruism Commercial Hedonism Power Recognition Scientific Security Tradition Mean SD ⁎ ⁎⁎
−0.01 0.21⁎⁎ 0.07⁎ 0.12⁎⁎ −0.14⁎⁎ 0.07⁎ −0.13⁎⁎ −0.07⁎ 0.00 0.13⁎⁎ 2.54 0.81
Ambition ⁎⁎
0.09 0.30⁎⁎ 0.16⁎⁎ 0.21⁎⁎ 0.01 0.23⁎⁎ 0.18⁎⁎ 0.11⁎⁎ −0.09⁎⁎ 0.06⁎ 2.66 0.63
Sociability ⁎⁎
0.12 0.56⁎⁎ 0.16⁎⁎ 0.18⁎⁎ 0.21⁎⁎ 0.20⁎⁎ 0.29⁎⁎ 0.07⁎ −0.04 0.06⁎ 1.64 1.02
Interpersonal 0.04 0.14⁎⁎ 0.13⁎⁎ −0.03 −0.16⁎⁎ −0.15⁎⁎ −0.19⁎⁎ −0.09⁎⁎ 0.02 0.03 2.08 0.80
Prudence
Inquisitive
−0.02 0.00 0.09⁎⁎ −0.01 −0.23⁎⁎ −0.13⁎⁎ −0.19⁎⁎ −0.09⁎⁎ 0.21⁎⁎ 0.11⁎⁎ 1.95 0.89
⁎⁎
0.17 0.07⁎ 0.16⁎⁎ 0.13⁎⁎ −0.03 0.11⁎⁎ 0.07⁎⁎ 0.40⁎⁎ −0.09⁎⁎ 0.07⁎ 2.36 0.75
Learning ⁎⁎
0.17 0.06⁎ 0.18⁎⁎ 0.05 −0.04 0.07⁎⁎ 0.00 0.21⁎⁎ 0.01 0.07⁎⁎ 2.00 0.89
Mean
SD
5.13 6.91 6.86 7.21 6.14 6.81 5.68 6.19 6.11 6.57
1.87 1.55 1.67 1.76 1.92 1.69 1.79 2.00 1.88 1.75
p < 0.05. p < 0.01.
2.3. Procedure
the sample was US-based and fairly large, such differences might be minimal.
The program Latent GOLD 5.1 (Vermunt & Magidson, 2016) generated the cluster solutions. The HPI scale scores served as covariates in the LCCA model. The study followed a 3-step analysis because it is more intuitive to first build a latent class model, and then relate it to predictors and covariates (Vermunt & Magidson, 2016). The step 3 submodule in Latent GOLD implements a bias adjustment which eliminates the threat of underestimating parameter estimates in this step (Bakk, Tekle, & Vermunt, 2013). In essence, the 3-step analysis starts by estimating a latent class model. Then cases are assigned to latent classes, and this classification information is saved. In the third step the latent classification scores saved in step 2 are regressed on the covariate variables correcting for the misclassification in the second step error to prevent bias (Asparouhov & Muthén, 2014).
2.2. Measures 2.2.1. Values The Schwartz Value Survey (Schwartz, 1992) is a widely used instrument for measuring personal values in cross-cultural psychology. It was developed with multidimensional scaling of international samples of students and teachers. Its exploratory nature and sample specificity might not guarantee that the measured values have sufficient construct validity. In contrast, the MVPI is a theoretically-grounded and empirically validated instrument which is also extensively used for values assessment. The MVPI is more amenable to the proposed LCCA methodology because it has reliable values scales for the maximum-likelihood clustering algorithms. The study used MVPI's ten beliefs scales measuring “shoulds”, ideals, and life goals. Their names follow the ten broad MVPI factors: Aesthetics, Affiliation, Altruism, Commerce, Hedonism, Power, Recognition, Science, Security, and Tradition.
3. Results 3.1. Descriptives Table 1 presents the descriptive statistics of the study's variables. Notable is the stable pattern of correlations between Ambition and Sociability and all values with the exclusion of the insignificant Hedonism-Ambition and Security-Sociability pairs. As expected, the selfenhancement values of Hedonism, Power, and Recognition relate negatively to Interpersonal Sensitivity, but even more so to Prudence. The values of Aesthetics and Science have stable relationships to the Inquisitive and Learning Approach. In general, two clusters of values have the highest and most consistent patterns of relationships to personality — the self-transcendence (Affiliation and Altruism) and the self-enhancement values. The conservation values (Security and Tradition) appear to relate most substantially only with Prudence.
2.2.2. Personality The HPI is extensively used by the HR industry because of its wellresearched validity for various jobs, and because it generally conforms to the authoritative Five-Factor Model of personality (FFM; John & Srivastava, 1999). HPI has seven personality factors: Adjustment, Ambition, Sociability, Interpersonal Sensitivity, Prudence, Inquisitive, and Learning Approach. In broad FFM language, Ambition and Sociability form Extraversion, Adjustment is Neuroticism, Prudence is Conscientiousness, Interpersonal Sensitivity is Agreeableness, and Learning Approach is an additional factor not captured by Openness to Experience and its HPI equivalent of Inquisitive. However, Smith, Hanges and Dickson (2001) found that HPI maps better for some FFM factors, and the desirable equivalence should reflect the HPI facets, called homogenous item clusters (HICs). For example, Adjustment's Trusting and Empathy HICs might relate better to Agreeableness, whereas Interpersonal Sensitivity's No Hostility HIC to Neuroticism. The archival dataset contained three items from each scale, with sufficient reliability and validity to include in research as markers for full-scale constructs. The corrected correlations (Levy, 1967) between the original long scales and the short, research-use scales used in the study are: Aesthetics (0.83), Affiliation (0.79), Altruism (0.82), Commerce (0.75), Hedonism (0.80), Power (0.75), Recognition (0.81), Science (0.83), Security (0.80), Tradition (0.75), Adjustment (0.45), Ambition (0.56), Sociability (0.78), Interpersonal Sensitivity (0.59), Prudence (0.66), Inquisitive (0.67), and Learning Approach (0.78). Also, because HPI uses a dichotomous (True/False) response scale, and the MVPI a trichotomous (Agree/Disagree/Neutral) scale, HPI scale scores varied between 0 and 3, and MVPI scale scores between 3 and 9.
3.2. Cluster solution The number of clusters was determined through the complex models-penalizing Bayesian Information Criterion (BIC), the sample size-penalizing Consistent Akaike Information Criterion (CAIC), the parsimony of the model, its error terms, and its substantive explanatory value (Celeux, Biernacki, & Govaert, 1997). Table 2 presents the fit statistics of six initial exploratory models. The 5-cluster model shows the greatest decrement in BIC and CAIC. However, the chi-square distributed L2 statistic, which indicates the unexplained amount of association among the variables after the estimation, needs to be as low as possible, and it started tapering off from the 4-cluster model. Also, parameters-wise, the 4-cluster model was more parsimonious. Both models' p-values were significantly lower than 0.001 whereas the best model should have the largest and insignificant 113
Personality and Individual Differences 131 (2018) 111–116
G.P. Yankov
Table 2 Fits of the initial cluster solutions.
Model Model Model Model Model Model
1 2 3 4 5 6
No. clusters
BIC(LL)
CAIC(LL)
No. par
L2
df
p-Value
Class. err.
1-Cluster 2-Cluster 3-Cluster 4-Cluster 5-Cluster 6-Cluster
51,328.68 50,831.96 50,748.99 50,686.06 50,651.41 50,660.03
51,388.68 50,902.96 50,830.99 50,779.06 50,755.41 50,775.03
60.00 71.00 82.00 93.00 104.00 115.00
29,961.20 29,384.48 29,221.50 29,078.57 28,963.91 28,892.53
1381.00 1370.00 1359.00 1348.00 1337.00 1326.00
3.3e−5287 4.1e−5175 1.1e−5148 2.0e−5126 5.0e−5110 1.4e−5102
0.00 0.14 0.22 0.24 0.26 0.26
Fig. 1. Profile plots of the five clusters based on values.
p-value. Bootstrapping the L2's p-value provides better estimation because the assumption of its chi-squared distribution is relaxed. Bootstrapping with 500 iterations resulted in p's of 0.34 (5-cluster model) and 0.37 (4-cluster model). It seemed that the 4-cluster model was superior, but its residuals had more values substantially larger than 1 suggesting that the model fell short of reproducing the association between the variables. The 5-cluster model's residuals were also inflated but presented a more balanced pattern especially for the value of Aesthetics whose residuals averaged 1.84 vs. 5.11 for the 4-cluster model. Thus, apart from being slightly less parsimonious (i.e. having more parameters to estimate) and misclassifying 2% more cases than the 4-cluster model (the last column in Table 2), the 5-cluster model was the better model in terms of general fit.1 After randomly splitting the sample, the 5-cluster solution was the most robust in the two subsamples.
tests, it seems that cluster 5 is very similar to cluster 3 in terms of the self-enhancement values, to cluster 4 in terms of conservation values, and to cluster 1 in terms of Aesthetics and Science. The lower half of Table 3 (re-expressed in Fig. 2) displays the covariates means after adjusting the clusters for the personality covariation. The LCCA's third step models the probabilities for each personality trait to characterize the clusters, not the personality traits as dependent variables. Therefore, no SEs are available for the covariates means. The covariates Wald tests were all significant but for ambition (p = 0.07). Thus, all other covariates significantly contributed to the model's ability to discriminate between the clusters. In terms of substantive interpretation, Adjustment did not discriminate well between clusters but for the low value for cluster 4. Cluster 1 is dominated by high Prudence, but low Sociability and Inquisitive; cluster 2 by very high Sociability, Inquisitive and Learning Approach; cluster 3 by very high Interpersonal Sensitivity and Inquisitive; cluster 4 by lower Adjustment, Interpersonal Sensitivity, and rather low Prudence, but high Sociability; cluster 5 by very low Sociability, Inquisitive and Learning Approach, but high Interpersonal Sensitivity. Notably, cluster 5 is alike cluster 1 but for its low Prudence. Also, cluster 3 is alike cluster 2, but has lower Sociability and higher Interpersonal Sensitivity. Below are the suggested profiles of each cluster based on conjoint analyses of their values and personality means. They are interpreted in the language of the FFM, but taking into account the specificity of HPI.
3.3. Personality covariation The upper half of Table 3 (re-expressed in Fig. 1) provides the clusters' values scores from the first estimation step. The Wald tests for each value across the five clusters indicate which clusters means significantly differ. Cluster 1 is dominated by the values of Commerce, Power, Security, and Tradition; cluster 2 by all values but for Aesthetics and to some extent Hedonism; cluster 3 by Altruism, Commerce, and Science; cluster 4 by Affiliation, Commerce, Hedonism, and Power; cluster 5 by Affiliation and Commerce. Given the Wald significance
• Cluster 1/Traditionalists value the conservation and self-enhance-
ment values of commerce, power, security, and tradition. They do not value aesthetics, recognition, nor science. They are the least extroverted and most self-controlled and moralistic cluster; they are
1
Personal communication with LATENTGOLD's creators Jeroen Vermunt and Jay Magidson (August 2017).
114
Personality and Individual Differences 131 (2018) 111–116
G.P. Yankov
Table 3 Parameters of the 5-cluster model. Cluster 1
Cluster 2
Cluster 3
Cluster 4
0.26
0.26
0.20
0.17
Cluster size
Cluster 5 0.11
First step
Mean
SE
Mean
SE
Mean
SE
Mean
SE
Mean
SE
Aesthetics Affiliation Altruism Commercial Hedonism Power Recognition Scientific Security Tradition
4.07 6.46 6.48 7.28 6.19 6.92 5.45 5.28 7.20 7.33
0.12 0.13 0.14 0.13 0.15 0.16 0.14 0.16 0.16 0.13
5.81 7.73 8.00 8.07 6.76 7.87 6.69 7.05 6.50 7.54
0.14 0.09 0.09 0.09 0.12 0.09 0.12 0.13 0.13 0.11
6.34 6.51 7.44 6.60 5.00 5.75 4.85 6.83 5.55 6.33
0.18 0.13 0.15 0.17 0.17 0.17 0.15 0.17 0.16 0.15
4.83 7.27 5.98 7.25 7.12 7.34 6.29 6.33 5.02 5.04
0.18 0.14 0.19 0.17 0.21 0.17 0.22 0.22 0.19 0.23
4.20 6.24 5.23 6.00 5.16 5.18 4.43 4.83 5.22 5.18
0.21 0.20 0.22 0.27 0.23 0.23 0.20 0.26 0.26 0.26
Third step
Mean
Mean
Mean
Mean
Mean
Adjustment Ambition Sociability Interpersonal sensitivity Prudence Inquisitive Learning approach
2.61 2.51 1.08 2.12 2.32 2.01 1.79
2.70 2.93 2.40 2.07 1.94 2.63 2.30
2.65 2.70 1.53 2.33 2.05 2.67 2.23
2.32 2.83 2.20 1.68 1.27 2.41 1.93
2.58 2.40 1.21 2.28 2.01 1.97 1.57
•
1–5 1–3/5; 3–5 1–3 1–4; 3–5 2–4; 3–5 1–4; 3–5 2–4; 3–5 1–5; 2–3; 3–4 3–5; 4–5 1–2; 3–4; 4–5
1–2/3/5; 2–3/5; 3–5 1–3/5; 2–4; 3–4/5 1–3/5; 2–4 1–2/3/5; 3–5 2–3; 3–5 1–5; 2–3 1–4/5; 2–3; 3–4
and temperamental too.
low on Openness to Experience.
• Cluster 2/Maximalists has the highest values loadings from all five, •
Cluster pairs with non-sig. (p > 0.05) Wald tests for means differences
• Cluster 5/Followers are a mix of the strengths and weaknesses of the
and its respondents are “high-valuers.” Maximalists are the most extraverted cluster, and are very open to experience and academically capable. Cluster 3/Intellectuals value altruism, and science, but not the values of hedonism, power, or recognition. Being so humble, they are the easiest to live with, sensitive, non-aggressive, and intellectual cluster. They might be the less extraverted and self-centered type of Maximalists. Cluster 4/Climbers are alike Maximalists in their quest for self-enhancement, and very different than the Traditionalists' emphasis on conservation. Climbers are the least conscientious cluster, they cooperate and understand others with difficulty, and might be anxious
other clusters. They are as humble as the Intellectuals, but do not share their openness; are emotionally stable and introverted as the Traditionalists, but not nearly as self-controlled and moralistic as them; like the Climbers disregard security and tradition, but unlike them cooperate easily.
4. Discussion Only recently have psychologists started to profile with LCCA (e.g. Lawrence and Zyphur, 2011). In this study, it produced a result that is descriptively and interpretatively richer than the sample-based correlations between values and personality traits. With LCCA values and
Fig. 2. Profile plots of the five clusters with the personality covariates. Note: no scale scores were below 1, hence the scaling axis starts from 1. 115
Personality and Individual Differences 131 (2018) 111–116
G.P. Yankov
Acknowledgements
personality are dynamically integrated within the ideal Self. The latter safeguards the suggested profiles from dustbowl empiricism. This study claimed that values organize and “express” the Self, but personality, being psychologically superordinate and thus less transparent to the Self, moderates the expression of values. Thus, we can interpret the five profiles from a value and from a personality-based perspective, but insight is conditional on a conjoint analysis. Values, being more face-valid and cognitively assessable, might be central to the Self, but personality characteristics would probably sway behavior in social contexts where values might conflict (i.e. dilemmas) and individuals need to act. Thus, personality becomes the superordinate factor which often decides the course of our “valued” actions. For example, the Traditionalists endorse tangible rewards, rules, and status. Knowing that they are not very sociable, but are very controlled and rule-bound, they might not thrive in very innovative and peoplerelated jobs. Maximalists are high valuers, but might not thrive in workplaces that do not match their sociability and intellect. The Intellectuals value science and harmony. However, these quiet individuals might become unproductive among very extroverted and competitive coworkers. The Climbers want to be powerful, rich, and liked. Yet, to be productive at work and private life, they should be more cooperative and sensitive, less impulsive, obey the rules and follow society's norms. The Followers, although limited in size, seem to mostly want to be in people's good books. Personality-wise, this “silent class” should focus on building their assertiveness, and start learning, organizing, and working on goals that would grow their self-efficacy and expertise. Given the cluster sizes, almost three fourths of respondents are either Traditionalists, Maximalists, or Intellectuals. Climbers, and especially Followers, are limited in size and show more unfavorable valuespersonality profiles, which is in tune with the assumption that variation in the value-personality profiles can become more and more idiosyncratic when the number of isolated clusters increases. This has implications for extant values-personality research. Instead of replicating Schwartz's (1992) values hierarchy this study indicated that individuals not only have clear-cut value preferences like the Traditionalists (conservation values), but also can espouse values from conflicting groups like the Maximalists (self-enhancement and self-transcendence values). This is a further proof that the values hierarchy is personal, and appraising it with zero-order correlations on a sample basis might conceal substantial inter-individual idiosyncrasies. This study poses several caveats. Statistically, cross-validation with larger population-representative samples is required to verify the five profiles. This is because clustering algorithms overfit the sample and do not hypothesize latent structures that generalize to the population. Also, other profiling methods such as grade-of-membership analysis (Woodbury & Manton, 1982) can be used for triangulation. Exploring the profiles with facet-level personality covariates would also convey a richer, more discriminative interpretation. Knowing this information would be helpful for understanding the consequences of hiring individuals in the suggested profiles. Practically, by profiling individuals conjointly on values and personality, selection practitioners might assess more fully the person-organization fit of applicants. That is, both candidate personality and value fit would be considered. Also, talent managers might use the profiles to design specific developmental trajectories and trainings. The last two could target the potential personality derailers of each profile. By exploring the Self from a values-personality perspective we might help individuals to better align what they value and want to be with who they are.
I would like to thank Mike Zickar and Jeff Foster for their help and advice. References Asparouhov, T., & Muthén, B. (2014). Auxiliary variables in mixture modeling: Three-step approaches using M plus. Structural Equation Modeling, 21(3), 329–341. Bakk, Z., Tekle, F., & Vermunt, J. (2013). Estimating the association between latent class membership and external variables using bias-adjusted three-step approaches. Sociological Methodology, 43(1), 272–311. Bilsky, W., & Schwartz, S. (1994). Values and personality. European Journal of Personality, 8(3), 163–181. Celeux, G., Biernacki, C., & Govaert, G. (1997). Choosing models in model-based clustering and discriminant analysis. Technical report. Rhone-Alpes: INRIA. Fisher, R., & Katz, J. (2008). Social-desirability bias and the validity of self-reported values. Psychology & Marketing, 17(2), 105–120. Hogan, R. (1995). A socioanalytic perspective on the Five-Factor Model. In J. S. Wiggins (Ed.). Theories of the Five-Factor Model. New York: Guilford. Hogan, J., & Hogan, R. (1995). Hogan personality inventory manual (2nd ed.). Tulsa, OK: HAS. Hogan, R., & Hogan, J. (1996). Motives, values, and preferences inventory manual. Tulsa, OK: HAS. John, O., & Srivastava, S. (1999). The big five trait taxonomy: History, measurement, and theoretical perspectives. In L. Pervin, & O. John (Eds.). Handbook of personality: Theory and research (pp. 102–138). (2nd ed.). New York: Guilford. Kluckhohn, C. (1951). Values and value-orientations in the theory of action. In T. Parsons, & E. Shils (Eds.). Toward a general theory of action. Cambridge, MA: Harvard University Press. Kohlberg, L. (1971). Stages of moral development. Moral Education, 1, 23–92. Krosnick, J., & Alwin, D. (1988). A test of the form-resistant correlation hypothesis: Ratings, rankings, and the measurement of values. Public Opinion Quarterly, 52(4), 526–538. Lawrence, B., & Zyphur, M. (2011). Identifying organizational faultlines with latent class cluster analysis. Organizational Research Methods, 14(1), 32–57. Lefkowitz, J. (2006). The constancy of ethics amidst the changing world of work. Human Resource Management Review, 16, 245–268. Levy, P. L. (1967). The correction for spurious correlation in the evaluation of short-form tests. Journal of Clinical Psychology, 23(1), 84–86. Maio, G., Roese, N., Seligman, C., & Katz, A. (1996). Rankings, ratings, and the measurement of values: Evidence for the superior validity of ratings. Basic and Applied Social Psychology, 18(2), 171–181. McCrae, R., & Costa, P. (1990). Personality in adulthood. New York: Guilford Press. Olver, J., & Mooradian, T. (2003). Personality traits and personal values: A conceptual and empirical integration. Personality and Individual Differences, 35(1), 109–125. Parks, L. (2007). Personality and values: A meta-analysis. Paper presented at the annual conference for the Society of Industrial and Organizational Psychology. New York: New York. Parks, L., & Guay, R. (2009). Personality, values, and motivation. Personality and Individual Differences, 47(7), 675–684. Roccas, S., Sagiv, L., Schwartz, S., & Knafo, A. (2002). The big five personality factors and personal values. Personality and Social Psychology Bulletin, 28(6), 789–801. Rokeach, M. (1973). The nature of human values. New York: Free Press. Schwartz, S. (1992). Universals in the content and structure of values: Theoretical advances and empirical tests in 20 countries. Advances in Experimental Social Psychology, 25, 1–65. Schwartz, S., & Rubel, T. (2005). Sex differences in value priorities: Cross-cultural and multimethod studies. Journal of Personality and Social Psychology, 89(6), 1010–1028. Smith, B., Hanges, J., & Dickson, W. (2001). Personnel selection and the five-factor model: Reexamining the effects of applicant's frame of reference. Journal of Applied Psychology, 86(2), 304–315. Vermunt, J. (2010). Latent class modeling with covariates: Two improved three-step approaches. Political Analysis, 18(4), 450–469. Vermunt, J., & Magidson, J. (2002). Latent class cluster analysis. Applied Latent Class Analysis, 11, 89–106. Vermunt, J., & Magidson, J. (2016). Technical guide for latent Gold 5.1: Basic, advanced, and syntax. Belmont, MA: Statistical Innovations, Inc. Verplanken, B., & Holland, R. (2002). Motivated decision making: Effects of activation and self-centrality of values on choices and behavior. Journal of Personality and Social Psychology, 82(3), 434–447. Wojciszke, B. (1989). The system of personal values and behavior. In N. Eisenberg, J. Reykowski, & E. Staub (Eds.). Social and moral values: Individual and societal perspectives (pp. 229–252). Hillsdale, NJ: Erlbaum. Woodbury, M., & Manton, K. (1982). A new procedure for analysis of medical classification. Methods Archive, 21, 210–220.
116