creative writing

literature review of personality type

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here .

Loading metrics

Open Access

Peer-reviewed

Research Article

Personality types revisited–a literature-informed and data-driven approach to an integration of prototypical and dimensional constructs of personality description

Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing – original draft

* E-mail: [email protected]

Affiliation Department of Psychology, Freie Universität Berlin, Berlin, Germany

Roles Conceptualization, Investigation, Methodology, Supervision, Validation, Writing – review & editing

Affiliation Department of Psychology, University of Duisburg-Essen, Duisburg Germany

Affiliation Personality Psychology and Psychological Assessment Unit, Helmut Schmidt University of the Federal Armed Forces Hamburg, Hamburg, Germany

A new algorithmic approach to personality prototyping based on Big Five traits was applied to a large representative and longitudinal German dataset (N = 22,820) including behavior, personality and health correlates. We applied three different clustering techniques, latent profile analysis, the k-means method and spectral clustering algorithms. The resulting cluster centers, i.e. the personality prototypes, were evaluated using a large number of internal and external validity criteria including health, locus of control, self-esteem, impulsivity, risk-taking and wellbeing. The best-fitting prototypical personality profiles were labeled according to their Euclidean distances to averaged personality type profiles identified in a review of previous studies on personality types. This procedure yielded a five-cluster solution: resilient, overcontroller, undercontroller, reserved and vulnerable-resilient. Reliability and construct validity could be confirmed. We discuss wether personality types could comprise a bridge between personality and clinical psychology as well as between developmental psychology and resilience research.

Citation: Kerber A, Roth M, Herzberg PY (2021) Personality types revisited–a literature-informed and data-driven approach to an integration of prototypical and dimensional constructs of personality description. PLoS ONE 16(1): e0244849. https://doi.org/10.1371/journal.pone.0244849

Editor: Stephan Doering, Medical University of Vienna, AUSTRIA

Received: January 5, 2020; Accepted: December 17, 2020; Published: January 7, 2021

Copyright: © 2021 Kerber et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: The data used in this article were made available by the German Socio-Economic Panel (SOEP, Data for years 1984-2015) at the German Institute for Economic Research, Berlin, Germany. To ensure the confidentiality of respondents’ information, the SOEP adheres to strict security standards in the provision of SOEP data. The data are reserved exclusively for research use, that is, they are provided only to the scientific community. To require full access to the data used in this study, it is required to sign a data distribution contract. All contact informations and the procedure to request the data can be obtained at: https://www.diw.de/en/diw_02.c.222829.en/access_and_ordering.html .

Funding: The author(s) received no specific funding for this work.

Competing interests: The authors have declared that no competing interests exist.

Introduction

Although documented theories about personality types reach back more than 2000 years (i.e. Hippocrates’ humoral pathology), and stereotypes for describing human personality are also widely used in everyday psychology, the descriptive and variable-oriented assessment of personality, i.e. the description of personality on five or six trait domains, has nowadays consolidated its position in modern personality psychology.

In recent years, however, the person-oriented approach, i.e. the description of an individual personality by its similarity to frequently occurring prototypical expressions, has amended the variable-oriented approach with the addition of valuable insights into the description of personality and the prediction of behavior. Focusing on the trait configurations, the person-oriented approach aims to identify personality types that share the same typical personality profile [ 1 ].

Nevertheless, the direct comparison of the utility of person-oriented vs. variable-oriented approaches to personality description yielded mixed results. For example Costa, Herbst, McCrae, Samuels and Ozer [ 2 ] found a higher amount of explained variance in predicting global functioning, geriatric depression or personality disorders for the variable-centered approach using Big Five personality dimensions. But these results also reflect a methodological caveat of this approach, as the categorical simplification of dimensionally assessed variables logically explains less variance. Despite this, the person-centered approach was found to heighten the predictability of a person’s behavior [ 3 , 4 ] or the development of adolescents in terms of internalizing and externalizing symptoms or academic success [ 5 , 6 ], problem behavior, delinquency and depression [ 7 ] or anxiety symptoms [ 8 ], as well as stress responses [ 9 ] and social attitudes [ 10 ]. It has also led to new insights into the function of personality in the context of other constructs such as adjustment [ 2 ], coping behavior [ 11 ], behavioral activation and inhibition [ 12 ], subjective and objective health [ 13 ] or political orientation [ 14 ], and has greater predictive power in explaining longitudinally measured individual differences in more temperamental outcomes such as aggressiveness [ 15 ].

However, there is an ongoing debate about the appropriate number and characteristics of personality prototypes and whether they perhaps constitute an methodological artifact [ 16 ].

With the present paper, we would like to make a substantial contribution to this debate. In the following, we first provide a short review of the personality type literature to identify personality types that were frequently replicated and calculate averaged prototypical profiles based on these previous findings. We then apply multiple clustering algorithms on a large German dataset and use those prototypical profiles generated in the first step to match the results of our cluster analysis to previously found personality types by their Euclidean distance in the 5-dimensional space defined by the Big Five traits. This procedure allows us to reliably link the personality prototypes found in our study to previous empirical evidence, an important analysis step lacking in most previous studies on this topic.

The empirical ground of personality types

The early studies applying modern psychological statistics to investigate personality types worked with the Q-sort procedure [ 1 , 15 , 17 ], and differed in the number of Q-factors. With the Q-Sort method, statements about a target person must be brought in an order depending on how characteristic they are for this person. Based on this Q-Sort data, prototypes can be generated using Q-Factor Analysis, also called inverse factor analysis. As inverse factor analysis is basically interchanging variables and persons in the data matrix, the resulting factors of a Q-factor analysis are prototypical personality profiles and not hypothetical or latent variable dimensions. On this basis, personality types (groups of people with similar personalities) can be formed in a second step by assigning each person to the prototype with whose profile his or her profile correlates most closely. All of these early studies determined at least three prototypes, which were labeled resilient, overcontroler and undercontroler grounded in Block`s theory of ego-control and ego-resiliency [ 18 ]. According to Jack and Jeanne Block’s decade long research, individuals high in ego-control (i.e. the overcontroler type) tend to appear constrained and inhibited in their actions and emotional expressivity. They may have difficulty making decisions and thus be non-impulsive or unnecessarily deny themselves pleasure or gratification. Children classified with this type in the studies by Block tend towards internalizing behavior. Individuals low in ego-control (i.e. the undercontroler type), on the other hand, are characterized by higher expressivity, a limited ability to delay gratification, being relatively unattached to social standards or customs, and having a higher propensity to risky behavior. Children classified with this type in the studies by Block tend towards externalizing behavior.

Individuals high in Ego-resiliency (i.e. the resilient type) are postulated to be able to resourcefully adapt to changing situations and circumstances, to tend to show a diverse repertoire of behavioral reactions and to be able to have a good and objective representation of the “goodness of fit” of their behavior to the situations/people they encounter. This good adjustment may result in high levels of self-confidence and a higher possibility to experience positive affect.

Another widely used approach to find prototypes within a dataset is cluster analysis. In the field of personality type research, one of the first studies based on this method was conducted by Caspi and Silva [ 19 ], who applied the SPSS Quick Cluster algorithm to behavioral ratings of 3-year-olds, yielding five prototypes: undercontrolled, inhibited, confident, reserved, and well-adjusted.

While the inhibited type was quite similar to Block`s overcontrolled type [ 18 ] and the well-adjusted type was very similar to the resilient type, two further prototypes were added: confident and reserved. The confident type was described as easy and responsive in social interaction, eager to do exercises and as having no or few problems to be separated from the parents. The reserved type showed shyness and discomfort in test situations but without decreased reaction speed compared to the inhibited type. In a follow-up measurement as part of the Dunedin Study in 2003 [ 20 ], the children who were classified into one of the five types at age 3 were administered the MPQ at age 26, including the assessment of their individual Big Five profile. Well-adjusteds and confidents had almost the same profiles (below-average neuroticism and above average on all other scales except for extraversion, which was higher for the confident type); undercontrollers had low levels of openness, conscientiousness and openness to experience; reserveds and inhibiteds had below-average extraversion and openness to experience, whereas inhibiteds additionally had high levels of conscientiousness and above-average neuroticism.

Following these studies, a series of studies based on cluster analysis, using the Ward’s followed by K-means algorithm, according to Blashfield & Aldenderfer [ 21 ], on Big Five data were published. The majority of the studies examining samples with N < 1000 [ 5 , 7 , 22 – 26 ] found that three-cluster solutions, namely resilients, overcontrollers and undercontrollers, fitted the data the best. Based on internal and external fit indices, Barbaranelli [ 27 ] found that a three-cluster and a four-cluster solution were equally suitable, while Gramzow [ 28 ] found a four-cluster solution with the addition of the reserved type already published by Caspi et al. [ 19 , 20 ]. Roth and Collani [ 10 ] found that a five-cluster solution fitted the data the best. Using the method of latent profile analysis, Merz and Roesch [ 29 ] found a 3-cluster, Favini et al. [ 6 ] found a 4-cluster solution and Kinnunen et al. [ 13 ] found a 5-cluster solution to be most appropriate.

Studies examining larger samples of N > 1000 reveal a different picture. Several favor a five-cluster solution [ 30 – 34 ] while others favor three clusters [ 8 , 35 ]. Specht et al. [ 36 ] examined large German and Australian samples and found a three-cluster solution to be suitable for the German sample and a four-cluster solution to be suitable for the Australian sample. Four cluster solutions were also found to be most suitable to Australian [ 37 ] and Chinese [ 38 ] samples. In a recent publication, the authors cluster-analysed very large datasets on Big Five personality comprising more than 1,5 million online participants using Gaussian mixture models [ 39 ]. Albeit their results “provide compelling evidence, both quantitatively and qualitatively, for at least four distinct personality types”, two of the four personality types in their study had trait profiles not found previously and all four types were given labels unrelated to previous findings and theory. Another recent publication [ 40 ] cluster-analysing data of over 270,000 participants on HEXACO personality “provided evidence that a five-profile solution was optimal”. Despite limitations concerning the comparability of HEXACO trait profiles with FFM personality type profiles, the authors again decided to label their personality types unrelated to previous findings instead using agency-communion and attachment theories.

We did not include studies in this literature review, which had fewer than 199 participants or those which restricted the number of types a priori and did not use any method to compare different clustering solutions. We have made these decisions because a too low sample size increases the probability of the clustering results being artefacts. Further, a priori limitation of the clustering results to a certain number of personality types is not well reasonable on the base of previous empirical evidence and again may produce artefacts, if the a priori assumed number of clusters does not fit the data well.

To gain a better overview, we extracted all available z-scores from all samples of the above-described studies. Fig 1 shows the averaged z-scores extracted from the results of FFM clustering solutions for all personality prototypes that occurred in more than one study. The error bars represent the standard deviation of the distribution of the z-scores of the respective trait within the same personality type throughout the different studies. Taken together the resilient type was replicated in all 19 of the mentioned studies, the overcontroler type in 16, the undercontroler personality type in 17 studies, the reserved personality type was replicated in 6 different studies, the confident personality type in 4 and the non-desirable type was replicated twice.

Average Big Five z-scores of personality types based on clustering of FFM datasets with N ≥ 199 that were replicated at least once. Error bars indicate the standard deviation of the repective trait within the respective personality type found in the literature [ 5 , 6 , 10 , 22 – 25 , 27 – 31 , 33 – 36 , 38 , 39 , 41 ].

https://doi.org/10.1371/journal.pone.0244849.g001

Three implications can be drawn from this figure. First, although the results of 19 studies on 26 samples with a total N of 1,560,418 were aggregated, the Big Five profiles for all types can still be clearly distinguished. In other words, personality types seem to be a phenomenon that survives the aggregation of data from different sources. Second, there are more than three replicable personality types, as there are other replicated personality types that seem to have a distinct Big Five profile, at least regarding the reserved and confident personality types. Third and lastly, the non-desirable type seems to constitute the opposite of the resilient type. Looking at two-cluster solutions on Big Five data personality types in the above-mentioned literature yields the resilient opposed to the non-desirable type. This and the fact that it was only replicated twice in the above mentioned studies points to the notion that it seems not to be a distinct type but rather a combined cluster of the over- and undercontroller personality types. Further, both studies with this type in the results did not find either the undercontroller or the overcontroller cluster or both. Taken together, five distinct personality types were consistently replicated in the literature, namely resilient, overcontroller, undercontroller, reserved and confident. However, inferring from the partly large error margin for some traits within some prototypes, not all personality traits seem to contribute evenly to the occurrence of the different prototypes. While for the overcontroler type, above average neuroticism, below average extraversion and openness seem to be distinctive, only below average conscientiousness and agreeableness seemed to be most characteristic for the undercontroler type. The reserved prototype was mostly characterized by below average openness and neuroticism with above average conscientiousness. Above average extraversion, openness and agreeableness seemed to be most distinctive for the confident type. Only for the resilient type, distinct expressions of all Big Five traits seemed to be equally significant, more precisely below average neuroticism and above average extraversion, openness, agreeableness and conscientiousness.

Research gap and novelty of this study

The cluster methods used in most of the mentioned papers were the Ward’s followed by K-means method or latent profile analysis. With the exception of Herzberg and Roth [ 30 ], Herzberg [ 33 ], Barbaranelli [ 27 ] and Steca et. al. [ 25 ], none of the studies used internal or external validity indices other than those which their respective algorithm (in most cases the SPSS software package) had already included. Gerlach et al. [ 39 ] used Gaussian mixture models in combination with density measures and likelihood measures.

The bias towards a smaller amount of clusters resulting from the utilization of just one replication index, e.g. Cohen's Kappa calculated by split-half cross-validation, which was ascertained by Breckenridge [ 42 ] and Overall & Magee [ 43 ], is probably the reason why a three-cluster solution is preferred in most studies. Herzberg and Roth [ 30 ] pointed to the study by Milligan and Cooper [ 44 ], which proved the superiority of the Rand index over Cohen's Kappa and also suggested a variety of validity metrics for internal consistency to examine the construct validity of the cluster solutions.

Only a part of the cited studies had a large representative sample of N > 2000 and none of the studies used more than one clustering algorithm. Moreover, with the exception of Herzberg and Roth [ 30 ] and Herzberg [ 33 ], none of the studies used a large variety of metrics for assessing internal and external consistency other than those provided by the respective clustering program they used. This limitation further adds up to the above mentioned bias towards smaller amounts of clusters although the field of cluster analysis and algorithms has developed a vast amount of internal and external validity algorithms and criteria to tackle this issue. Further, most of the studies had few or no other assessments or constructs than the Big Five to assess construct validity of the resulting personality types. Herzberg and Roth [ 30 ] and Herzberg [ 33 ] as well, though using a diverse variety of validity criteria only used one clustering algorithm on a medium-sized dataset with N < 2000.

Most of these limitations also apply to the study by Specht et. al. [ 36 ], which investigated two measurement occasions of the Big Five traits in the SOEP data sample. They used only one clustering algorithm (latent profile analysis), no other algorithmic validity criteria than the Bayesian information criterion and did not utilize any of the external constructs also assessed in the SOEP sample, such as mental health, locus of control or risk propensity for construct validation.

The largest sample and most advanced clustering algorithm was used in the recent study by Gerlach et al. [ 39 ]. But they also used only one clustering algorithm, and had no other variables except Big Five trait data to assess construct validity of the resulting personality types.

The aim of the present study was therefore to combine different methodological approaches while rectifying the shortcomings in several of the studies mentioned above in order to answer the following exploratory research questions: Are there replicable personality types, and if so, how many types are appropriate and in which constellations are they more (or less) useful than simple Big Five dimensions in the prediction of related constructs?

Three conceptually different clustering algorithms were used on a large representative dataset. The different solutions of the different clustering algorithms were compared using methodologically different internal and external validity criteria, in addition to those already used by the respective clustering algorithm.

To further examine the construct validity of the resulting personality types, their predictive validity in relation to physical and mental health, wellbeing, locus of control, self-esteem, impulsivity, risk-taking and patience were assessed.

Mental health and wellbeing seem to be associated mostly with neuroticism on the variable-oriented level [ 45 ], but on a person-oriented level, there seem to be large differences between the resilient and the overcontrolled personality type concerning perceived health and well-being beyond mean differences in neuroticism [ 33 ]. This seems also to be the case for locus of control and self-esteem, which is associated with neuroticism [ 46 ] and significantly differs between resilient and overcontrolled personality type [ 33 ]. On the other hand, impulsivity and risk taking seem to be associated with all five personality traits [ 47 ] and e.g. risky driving or sexual behavior seem to occur more often in the undercontrolled personality type [ 33 , 48 ].

We chose these measures because of their empirically known differential associations to Big Five traits as well as to the above described personality types. So this both offers the opportunity to have an integrative comparison of the variable- and person-centered descriptions of personality and to assess construct validity of the personality types resulting from our analyses.

Materials and methods

The acquisition of the data this study bases on was carried out in accordance with the principles of the Basel Declaration and recommendations of the “Principles of Ethical Research and Procedures for Dealing with Scientific Misconduct at DIW Berlin”. The protocol was approved by the Deutsches Institut für Wirtschaftsforschung (DIW).

The data used in this study were provided by the German Socio-Economic Panel Study (SOEP) of the German institute for economic research [ 49 ]. Sample characteristics are shown in Table 1 . The overall sample size of the SOEP data used in this study, comprising all individuals who answered at least one of the Big-Five personality items in 2005 and 2009, was 25,821. Excluding all members with more than one missing answers on the Big Five assessment or intradimensional answer variance more than four times higher than the sample average resulted in a total Big Five sample of N = 22,820, which was used for the cluster analyses. 14,048 of these individuals completed, in addition to the Big Five, items relevant to further constructs examined in this study that were assessed in other years. The 2013 SOEP data Big Five assessment was used as a test sample to examine stability and consistency of the final cluster solution.

https://doi.org/10.1371/journal.pone.0244849.t001

The Big Five were assessed in 2005 2009 and 2013 using the short version of the Big Five inventory (BFI-S). It consists of 15 items, with internal consistencies (Cronbach’s alpha) of the scales ranging from .5 for openness to .73 for openness [ 50 ]. Further explorations showed strong robustness across different assessment methods [ 51 ].

To measure the predictive validity, several other measures assessed in the SOEP were included in the analyses. In detail, these were:

Patience was assessed in 2008 with one item: “Are you generally an impatient person, or someone who always shows great patience?”

Risk taking.

Risk-taking propensity was assessed in 2009 by six items asking about the willingness to take risks while driving, in financial matters, in leisure and sports, in one’s occupation (career), in trusting unknown people and the willingness to take health risks, using a scale from 0 (risk aversion) to 10 (fully prepared to take risks). Cronbach’s alpha was .82 for this scale in the current sample.

Impulsivity/Spontaneity.

Impulsivity/spontaneity was assessed in 2008 with one item: Do you generally think things over for a long time before acting–in other words, are you not impulsive at all? Or do you generally act without thinking things over for long time–in other words, are you very impulsive?

Affective and cognitive wellbeing.

Affect was assessed in 2008 by four items asking about the amount of anxiety, anger, happiness or sadness experienced in the last four weeks on a scale from 1 (very rare) to 5 (very often). Cronbach’s alpha for this scale was .66. The cognitive satisfaction with life was assessed by 10 items asking about satisfaction with work, health, sleep, income, leisure time, household income, household duties, family life, education and housing, with a Cronbach’s alpha of .67. The distinction between cognitive and affective wellbeing stems from sociological research based on constructs by Schimmack et al. [ 50 ].

Locus of control.

The individual attitude concerning the locus of control, the degree to which people believe in having control over the outcome of events in their lives opposed to being exposed to external forces beyond their control, was assessed in 2010 with 10 items, comprising four positively worded items such as “My life’s course depends on me” and six negatively worded items such as “Others make the crucial decisions in my life”. Items were rated on a 7-point scale ranging from “does not apply” to “does apply”. Cronbach’s alpha in the present sample for locus of control was .57.

Self-esteem.

Global self-esteem–a person’s overall evaluation or appraisal of his or her worth–was measured in 2010 with one item: “To what degree does the following statement apply to you personally?: I have a positive attitude toward myself”.

To assess subjective health, the 12-Item Short Form Health Survey (SF-12) was integrated into the SOEP questionnaire and assessed in 2002, 2004, 2006, 2008 and 2010. In the present study, we used the data from 2008 and 2010. The SF-12 is a short form of the SF-36, a self-report questionnaire to assess the non-disease-specific health status [ 52 ]. Within the SF-12, items can be grouped onto two subscales, namely the physical component summary scale, with items asking about physical health correlates such as how exhausting it is to climb stairs, and the mental component summary scale, with items asking about mental health correlates such as feeling sad and blue. The literature on health measures often distinguishes between subjective and objective health measures (e.g., BMI, blood pressure). From this perspective, the SF-12 would count as a subjective health measure. In the present sample, Cronbach’s alpha for the SF-12 items was .77.

Derivation of the prototypes

The first step was to administer three different clustering methods on the Big Five data of the SOEP sample: First, the conventional linear clustering method used by Asendorpf [ 15 , 35 , 53 ] and also Herzberg and Roth [ 30 ] combines the hierarchical clustering method of Ward [ 54 ] with the k-means algorithm [ 55 ]. This algorithm generates a first guess of personality types based on hierarchical clustering, and then uses this first guess as starting points for the k-means-method, which iteratively adjusts the personality profiles, i.e. the cluster means to minimize the error of allocation, i.e. participants with Big Five profiles that are allocated to two or more personality types. The second algorithm we used was latent profile analysis with Mclust in R [ 56 ], an algorithm based on probabilistic finite mixture modeling, which assumes that there are latent classes/profiles/mixture components underlying the manifest observed variables. This algorithm generates personality profiles and iteratively calculates the probability of every participant in the data to be allocated to one of the personality types and tries to minimize an error term using maximum likelihood method. The third algorithm was spectral clustering, an algorithm which initially computes eigenvectors of graph Laplacians of the similarity graph constructed on the input data to discover the number of connected components in the graph, and then uses the k-means algorithm on the eigenvectors transposed in a k-dimensional space to compute the desired k clusters [ 57 ]. As it is an approach similar to the kernel k-means algorithm [ 58 ], spectral clustering can discover non-linearly separable cluster formations. Thus, this algorithm is able, in contrast to the standard k-means procedure, to discover personality types having unequal or non-linear distributions within the Big-Five traits, e.g. having a small SD on neuroticism while having a larger SD on conscientiousness or a personality type having high extraversion and either high or low agreeableness.

Within the last 50 years, a large variety of clustering algorithms have been established, and several attempts have been made to group them. In their book about cluster analysis, Bacher et al. [ 59 ] group cluster algorithms into incomplete clustering algorithms, e.g. Q-Sort or multidimensional scaling, deterministic clustering, e.g. k-means or nearest-neighbor algorithms, and probabilistic clustering, e.g. latent class and latent profile analysis. According to Jain [ 60 ], cluster algorithms can be grouped by their objective function, probabilistic generative models and heuristics. In his overview of the current landscape of clustering, he begins with the group of density-based algorithms with linear similarity functions, e.g. DBSCAN, or probabilistic models of density functions, e.g. in the expectation-maximation (EM) algorithm. The EM algorithm itself also belongs to the large group of clustering algorithms with an information theoretic formulation. Another large group according to Jain is graph theoretic clustering, which includes several variants of spectral clustering. Despite the fact that it is now 50 years old, Jain states that k-means is still a good general-purpose algorithm that can provide reasonable clustering results.

The clustering algorithms chosen for the current study are therefore representatives of the deterministic vs. probabilistic grouping according to Bacher et. al. [ 59 ], as well as representatives of the density-based, information theoretic and graph theoretic grouping according to Jain [ 60 ].

Determining the number of clusters

There are two principle ways to determine cluster validity: external or relative criteria and internal validity indices.

External validity criteria.

External validity criteria measure the extent to which cluster labels match externally supplied class labels. If these external class labels originate from another clustering algorithm used on the same data sample, the resulting value of the external cluster validity index is relative. Another method, which is used in the majority of the cited papers in section 1, is to randomly split the data in two halves, apply a clustering algorithm on both halves, calculate the cluster means and allocate members of one half to the calculated clusters of the opposite half by choosing the cluster mean with the shortest Euclidean distance to the data member in charge. If the cluster algorithm allocation of one half is then compared with the shortest Euclidean distance allocation of the same half by means of an external cluster validity index, this results in a value for the reliability of the clustering method on the data sample.

As allocating data points/members by Euclidean distances always yields spherical and evenly shaped clusters, it will favor clustering methods that also yield spherical and evenly shaped clusters, as it is the case with standard k-means. The cluster solutions obtained with spectral clustering as well as latent profile analysis (LPA) are not (necessarily) spherical or evenly shaped; thus, allocating members of a dataset by their Euclidean distances to cluster means found by LPA or spectral clustering does not reliably represent the structure of the found cluster solution. This is apparent in Cohen’s kappa values <1 if one uses the Euclidean external cluster assignment method comparing a spectral cluster solution with itself. Though by definition, Cohen’s kappa should be 1 if the two ratings/assignments compared are identical, which is the case when comparing a cluster solution (assigning every data point to a cluster) with itself. This problem can be bypassed by allocating the members of the test dataset to the respective clusters by training a support vector machine classifier for each cluster. Support vector machines (SVM) are algorithms to construct non-linear “hyperplanes” to classify data given their class membership [ 61 ]. They can be used very well to categorize members of a dataset by an SVM-classifier trained on a different dataset. Following the rationale not to disadvantage LPA and spectral clustering in the calculation of the external validity, we used an SVM classifier to calculate the external validity criteria for all clustering algorithms in this study.

To account for the above mentioned bias to smaller numbers of clusters we applied three external validity criteria: Cohen’s kappa, the Rand index [ 62 ] and the Hubert-Arabie adjusted Rand index [ 63 ].

Internal validity criteria.

Again, to account for the bias to smaller numbers of clusters, we also applied multiple internal validity criteria selected in line with the the following reasoning: According to Lam and Yan [ 64 ], the internal validity criteria fall into three classes: Class one includes cost-function-based indices, e.g. AIC or BIC [ 65 ], whereas class two comprises cluster-density-based indices, e.g. the S_Dbw index [ 66 ]. Class three is grounded on geometric assumptions concerning the ratio of the distances within clusters compared to the distances between the clusters. This class has the most members, which differ in their underlying mathematics. One way of assessing geometric cluster properties is to calculate the within- and/or between-group scatter, which both rely on summing up distances of the data points to their barycenters (cluster means). As already explained in the section on external criteria, calculating distances to cluster means will always favor spherical and evenly shaped cluster solutions without noise, i.e. personality types with equal and linear distributions on the Big Five trait dimensions, which one will rarely encounter with natural data.

Another way not solely relying on distances to barycenters or cluster means is to calculate directly with the ratio of distances of the data points within-cluster and between-cluster. According to Desgraupes [ 67 ], this applies to the following indices: the C-index, the Baker & Hubert Gamma index, the G(+) index, Dunn and Generalized Dunn indices, the McClain-Rao index, the Point-Biserial index and the Silhouette index. As the Gamma and G(+) indices rely on the same mathematical construct, one can declare them as redundant. According to Bezdek [ 68 ], the Dunn index is very sensitive to noise, even if there are only very few outliers in the data. Instead, the authors propose several ways to compute a Generalized Dunn index, some of which also rely on the calculation of barycenters. The best-performing GDI algorithm outlined by Bezdek and Pal [ 68 ] which does not make use of cluster barycenters is a ratio of the mean distance of every point between clusters to the maximum distance between points within the cluster, henceforth called GDI31. According to Vendramin et al. [ 69 ], the Gamma, C-, and Silhouette indices are the best-performing (over 80% correct hit rate), while the worst-performing are the Point-Biserial and the McClain-Rao indices (73% and 51% correct hit rate, respectively).

Fig 2 shows a schematic overview of the procedure we used to determine the personality types Big Five profiles, i.e. the cluster centers. To determine the best fitting cluster solution, we adopted the two-step procedure proposed by Blashfield and Aldenfelder [ 21 ] and subsequently used by Asendorpf [ 15 , 35 , 53 ] Boehm [ 41 ], Schnabel [ 24 ], Gramzow [ 28 ], and Herzberg and Roth [ 30 ], with a few adjustments concerning the clustering algorithms and the validity criteria.

LPA = latent profile analysis, SVM = Support Vector Machine.

https://doi.org/10.1371/journal.pone.0244849.g002

First, we drew 20 random samples of the full sample comprising all individuals who answered the Big-Five personality items in 2005 and 2009 with N = 22,820 and split every sample randomly into two halves. Second, all three clustering algorithms described above were performed on each half, saving the 3-, 4-,…,9- and 10-cluster solution. Third, participants of each half were reclassified based on the clustering of the other half of the same sample, again for every clustering algorithm and for all cluster solutions from three to 10 clusters. In contrast to Asendorpf [ 35 ], this was implemented not by calculating Euclidean distances, but by training a support vector machine classifier for every cluster of a cluster solution of one half-sample and reclassifying the members of the other half of the same sample by the SVM classifier. The advantages of this method are explained in the section on external criteria. This resulted in 20 samples x 2 halves per sample x 8 cluster solutions x 3 clustering algorithms, equaling 960 clustering solutions to be compared.

The fourth step was to compute the external criteria comparing each Ward followed by k-means, spectral, or probabilistic clustering solution of each half-sample to the clustering by the SVM classifier trained on the opposite half of the same sample, respectively. The external calculated in this step were Cohen's kappa, Rand’s index [ 62 ] and the Hubert & Arabie adjusted Rand index [ 63 ]. The fifth step consisted of averaging: We first averaged the external criteria values per sample (one value for each half), and then averaged the 20x4 external criteria values for each of the 3-,4-…, 10-cluster solutions for each algorithm.

The sixth step was to temporarily average the external criteria values for the 3-,4-,… 10-cluster solution over the three clustering algorithms and discard the cluster solutions that had a total average kappa below 0.6.

As proposed by Herzberg and Roth [ 30 ], we then calculated several internal cluster validity indices for all remaining cluster solutions. The internal validity indices which we used were, in particular, the C-index [ 70 ], the Baker-Hubert Gamma index [ 71 ], the G + index [ 72 ], the Generalized Dunn index 31 [ 68 ], the Point-Biserial index [ 44 ], the Silhouette index [ 73 ], AIC and BIC [ 65 ] and the S_Dbw index [ 66 ]. Using all of these criteria, it is possible to determine the best clustering solution in a mathematical/algorithmic manner.

The resulting clusters where then assigned names by calculating Euclidean distances to the clusters/personality types found in the literature, taking the nearest type within the 5-dimensional space defined by the respective Big Five values.

To examine the stability and consistency of the final cluster solution, in a last step, we then used the 2013 SOEP data sample to calculate a cluster solution using the algorithm and parameters which generated the solution with the best validity criteria for the 2005 and 2009 SOEP data sample. The 2013 personality prototypes were allocated to the personality types of the solution from the previous steps by their profile similarity measure D. Stability then was assessed by calculation of Rand-index, adjusted Rand-index and Cohen’s Kappa for the complete solution and for every single personality type. To generate the cluster allocations between the different cluster solutions, again we used SVM classifier as described above.

To assess the predictive and the construct validity of the resulting personality types, the inversed Euclidean distance for every participant to every personality prototype (averaged Big Five profile in one cluster) in the 5-dimensional Big-Five space was calculated and correlated with further personality, behavior and health measures mentioned above. To ensure that longitudinal reliability was assessed in this step, Big Five data assessed in 2005 were used to predict measures which where assessed three, four or five years later. The selection of participants with available data in 2005 and 2008 or later reduced the sample size in this step to N = 14,048.

Internal and external cluster fit indices

Table 2 shows the mean Cohen’s kappa values, averaged over all clustering algorithms and all 20 bootstrapped data permutations.

https://doi.org/10.1371/journal.pone.0244849.t002

Whereas the LPA and spectral cluster solutions seem to have better kappa values for fewer clusters, the kappa values of the k-means clustering solutions have a peak at five clusters, which is even higher than the kappa values of the three-cluster solutions of the other two algorithms.

Considering that these values are averaged over 20 independent computations, there is very low possibility that this result is an artefact. As the solutions with more than five clusters had an average kappa below .60, they were discarded in the following calculations.

Table 3 shows the calculated external and internal validity indices for the three- to five-cluster solutions, ordered by the clustering algorithm. Comparing the validity criterion values within the clustering algorithms reveals a clear preference for the five-cluster solution in the spectral as well as the Ward followed by k-means algorithm.

https://doi.org/10.1371/journal.pone.0244849.t003

Looking solely at the cluster validity results of the latent profile models, they seem to favor the three-cluster model. Yet, in a global comparison, only the S_Dbw index continues to favor the three-cluster LPA model, whereas the results of all other 12 validity indices support five-cluster solutions. The best clustering solution in terms of the most cluster validity index votes is the five-cluster Ward followed by k-means solution, and second best is the five-cluster spectral solution. It is particularly noteworthy that the five-cluster K-means solution has higher values on all external validity criteria than all other solutions. As these values are averaged over 20 independent cluster computations on random data permutations, and still have better values than solutions with fewer clusters despite the fact that these indices have a bias towards solutions with fewer clusters [ 42 ], there seems to be a substantial, replicable five-component structure in the Big Five Data of the German SOEP sample.

Description of the prototypes

The mean z-scores on the Big Five factors of the five-cluster k-means as well as the spectral solution are depicted in Fig 2 . Also depicted is the five-cluster LPA solution, which is, despite having poor internal and external validity values compared to the other two solutions, more complicated to interpret. To find the appropriate label for the cluster partitions, the respective mean z-scores on the Big Five factors were compared with the mean z-scores found in the literature, both visually and by the Euclidean distance.

The spectral and the Ward followed by k-means solution overlap by 81.3%; the LPA solution only overlaps with the other two solutions by 21% and 23%, respectively. As the Ward followed by k-means solution has the best values both for external and internal validity criteria, we will focus on this solution in the following.

The first cluster has low neuroticism and high values on all other scales and includes on average 14.4% of the participants (53.2% female; mean age 53.3, SD = 17.3). Although the similarity to the often replicated resilient personality type is already very clear merely by looking at the z-scores, a very strong congruence is also revealed by computing the Euclidean distance (0.61). The second cluster is mainly characterized by high neuroticism, low extraversion and low openness and includes on average 17.3% of the participants (54.4% female; mean age 57.6, SD = 18.2). It clearly resembles the overcontroller type, to which it also has the shortest Euclidean distance (0.58). The fourth cluster shows below-average values on the factors neuroticism, extraversion and openness, as opposed to above-average values on openness and conscientiousness. It includes on average 22.5% of the participants (45% female; mean age 56.8, SD = 17.6). Its mean z-scores closely resemble the reserved personality type, to which it has the smallest Euclidean distance (0.36). The third cluster is mainly characterized by low conscientiousness and low openness, although in the spectral clustering solution, it also has above-average extraversion and openness values. Computing the Euclidean distance (0.86) yields the closest proximity to the undercontroller personality type. This cluster includes on average 24.6% of the participants (41.3% female; mean age 50.8, SD = 18.3). The fifth cluster exhibits high z-scores on every Big Five trait, including a high value for neuroticism. Computing the Euclidean distances to the previously found types summed up in Fig 1 reveals the closest resemblance with the confident type (Euclidean distance = 0.81). Considering the average scores of the Big Five traits, it resembles the confident type from Herzberg and Roth [ 30 ] and Collani and Roth [ 10 ] as well as the resilient type, with the exception of the high neuroticism score. Having above average values on the more adaptive traits while having also above average neuroticism values reminded a reviewer from a previous version of this paper of the vulnerable but invincible children of the Kauai-study [ 74 ]. Despite having been exposed to several risk factors in their childhood, they were well adapted in their adulthood except for low coping efficiency in specific stressful situations. Taken together with the lower percentage of participants in the resilient cluster in this study, compared to previous studies, we decided to name the 5 th cluster vulnerable-resilient. Consequently, only above or below average neuroticism values divided between resilient and vulnerable resilient. On average, 21.2% of the participants were allocated to this cluster (68.3% female; mean age 54.9, SD = 17.4).

Summarizing the descriptive statistics, undercontrollers were the “youngest” cluster whereas overcontrollers were the “oldest”. The mean age differed significantly between clusters ( F [4, 22820] = 116.485, p <0.001), although the effect size was small ( f = 0.14). The distribution of men and women between clusters differed significantly (c 2 [ 4 ] = 880.556, p <0.001). With regard to sex differences, it was particularly notable that the vulnerable-resilient cluster comprised only 31.7% men. This might be explained by general sex differences on the Big Five scales. According to Schmitt et al. [ 75 ], compared to men, European women show a general bias to higher neuroticism (d = 0.5), higher conscientiousness (d = 0.3) and higher extraversion and openness (d = 0.2). As the vulnerable-resilient personality type is mainly characterized by high neuroticism and above-average z-scores on the other scales, it is therefore more likely to include women. In turn, this implies that men are more likely to have a personality profile characterized mainly by low conscientiousness and low openness, which is also supported by our findings, as only 41.3% of the undercontrollers were female.

Concerning the prototypicality of the five-cluster solution compared to the mean values extracted from previous studies, it is apparent that the resilient, the reserved and the overcontroller type are merely exact replications. In contrast to previous findings, the undercontrollers differed from the previous findings cited above in terms of average neuroticism, whereas the vulnerable-resilient type differed from the previously found type (labeled confident) in terms of high neuroticism.

Stability and consistency

Inspecting the five cluster solution using the k-means algorithm on the Big Five data of the 2013 SOEP sample seemed to depict a replication of the above described personality types. This first impression was confirmed by the calculation of the profile similarity measure D between the 2005/2009 and 2013 SOEP sample cluster solutions, which yielded highest similarity for the undercontroler (D = 0.27) and reserved (D = 0.36) personality types, followed by the vulnerable-resilient (D = 0.37), overcontroler (D = 0.44) and resilient (D = 0.50) personality types. Substantial agreement was confirmed by the values of the Rand index (.84) and Cohen’ Kappa (.70) whereas the Hubert Arabie adjusted Rand Index (.58) indicated moderate agreement for the comparison between the kmeans cluster solution for the 2013 SOEP sample and the cluster allocation with an SVM classifier trained on the 2005 and 2009 kmeans cluster solution.

Predictive validity

In view of the aforementioned criticisms that (a) predicting dimensional variables will mathematically favor dimensional personality description models, and (b) using dichotomous predictors will necessarily provide less explanation of variance than a model using five continuous predictors, we used the profile similarity measure D [ 76 ] instead of dichotomous dummy variables accounting for the prototype membership. Correlations between the inversed Euclidean similarity measure D to the personality types and patience, risk-taking, spontaneity/impulsivity, locus of control, affective wellbeing, self-esteem and health are depicted in Table 4 .

https://doi.org/10.1371/journal.pone.0244849.t004

Patience had the highest association with the reserved personality type (r = .19, p < .001). The propensity to risky behavior, e.g. while driving (r = .17, p < .001), in financial matters (r = .17, p < .001) or in health decisions (r = .13, p < .001) was most highly correlated with the undercontroller personality type. This means that the more similar the Big-Five profile to the above-depicted undercontroller personality prototype, the higher the propensity for risky behavior. The average correlation across all three risk propensity scales with the undercontroller personality type is r = .21, with p < .001. This is in line with the postulations by Block and Block and subsequent replications by Caspi et al. [ 19 , 48 ], Robins et al. [ 1 ] and Herzberg [ 33 ] about the undercontroller personality type. Spontaneity/impulsivity showed the highest correlation with the overcontroller personality type (r = -.18, p<0.001). This is also in accordance with Block and Block, who described this type as being non-impulsive and appearing constrained and inhibited in actions and emotional expressivity.

Concerning locus of control, proximity to the resilient personality profile had the highest correlation with internal locus of control (r = .25, p < .001), and in contrast, the more similar the individual Big-Five profile was to the overcontroller personality type, the higher the propensity for external allocation of control (r = .22, p < .001). This is not only in line with Block and Block’s postulations that the resilient personality type has a good repertoire of coping behavior and therefore perceives most situations as “manageable” as well as with the findings by [ 33 ], but is also in accordance with findings regarding the construct and development of resilience [ 77 , 78 ].

Also in line with the predictions of Block and Block and replicating the findings of Herzberg [ 33 ], self-esteem was correlated the highest with the resilient personality profile similarity (r = .33, p < .001), second highest with the reserved personality profile proximity (r = .15, p < .001), and negatively correlated with the overcontroller personality type (r = -.27, p < .001).

This pattern also applies to affective and cognitive wellbeing as well as physical and mental health measured by the SF-12. Affective wellbeing was correlated the highest with similarity to the resilient personality type (r = .27, p < .001), and second highest with the reserved personality type (r = .23, p < .001). The overcontroller personality type, in contrast, showed a negative correlation with affective (r = -.16, p < .001) and cognitive (r = -21, p < .001) wellbeing. Concerning health, a remarkable finding is that lack of physical health impairment correlated the highest with the resilient personality profile similarity (p = -.23, p < .001) but lack of mental health impairment correlated the highest with the reserved personality type (r = -.15, p < .001). The highest correlation with mental health impairments (r = .11, p < .001), as well as physical health impairments (r = .16, p < .001) was with the overcontroller personality profile similarity. It is striking that although the undercontroller personality profile similarity was associated with risky health behavior, it had a negative association with health impairment measures, in contrast to the overcontroller personality type, which in turn had no association with risky health behavior. This result is in line with the link of internalizing and externalizing behavior with the overcontroller and undercontroller types [ 79 ], respectively. Moreover, it is also in accordance with the association of internalizing problems with somatic symptoms and/or symptoms of depressiveness and anxiety [ 80 ].

A further noteworthy finding is that these associations cannot be solely explained by the high neuroticism of the overcontroller personality type, as the vulnerable-resilient type showed a similar level of neuroticism but no correlation with self-esteem, the opposite correlation with impulsivity, and far lower correlations with health measures or locus of control. The vulnerable-resilient type showed also a remarkable distinction to the other types concerning the correlations to wellbeing. While for all other types, the direction and significance of the correlations to affective and cognitive measures of wellbeing were alike, the vulnerable-resilient type only had a significant negative correlation to affective wellbeing while having no significant correlation to measures of cognitive wellbeing.

To provide an overview of the particular associations of the Big Five values with all of the above-mentioned behavior and personality measures, Table 5 shows the bivariate correlations.

https://doi.org/10.1371/journal.pone.0244849.t005

Investigating the direction of the correlation and the relativity of each value to each other row-wise reveals, to some extent, a clear resemblance with the z-scores of the personality types shown in Fig 3 . Correlation profiles of risk taking, especially the facet risk-taking in health issues and locus of control, clearly resemble the undercontroller personality profile (negative correlations with openness and conscientiousness, positive but lower correlations with extraversion and openness). Patience had negative correlations with neuroticism and extraversion, and positive correlations with openness and conscientiousness, which in turn resembles the z-score profile of the reserved personality profile. Spontaneity/impulsivity had moderate to high positive correlations with extraversion and openness, and low negative correlations with openness and neuroticism, which resembles the inverse of the overcontroller personality profile. Self-esteem as well as affective and cognitive wellbeing correlations with the Big Five clearly resemble the resilient personality profile: negative correlations with neuroticism, and positive correlations with extraversion, openness, openness and conscientiousness. Inspecting the SF-12 health correlation, in terms of both physical and mental health, reveals a resemblance to the inversed resilient personality profile (high correlation with neuroticism, low correlation with extraversion, openness, openness and conscientiousness, as well as a resemblance with the overcontroller profile (positive correlation with neuroticism, negative correlation with extraversion).

https://doi.org/10.1371/journal.pone.0244849.g003

On the variable level, neuroticism had the highest associations with almost all of the predicted variables, with the exception of impulsivity, which was mainly correlated with extraversion and openness. It is also evident that all variables in question here are correlated with three or more Big Five traits. This can be seen as support for hypothesis that the concept of personality prototypes has greater utility than the variable-centered approach in understanding or predicting more complex psychological constructs that are linked to two or more Big Five traits.

The goal of this study was to combine different methodological approaches while overcoming the shortcomings of previous studies in order to answer the questions whether there are replicable personality types, how many of them there are, and how they relate to Big Five traits and other psychological and health-related constructs. The results revealed a robust five personality type model, which was able to significantly predict all of the psychological constructs in question longitudinally. Predictions from previous findings connecting the predicted variables to the particular Big Five dimensions underlying the personality type model were confirmed. Apparently, the person-centered approach to personality description has the most practical utility when predicting behavior or personality correlates that are connected to more than one or two of the Big Five traits such as self-esteem, locus of control and wellbeing.

This study fulfils all three criteria specified by von Eye & Bogat [ 81 ] regarding person-oriented research and considers the recommendations regarding sample size and composition by Herzberg and Roth [ 30 ]. The representative and large sample was analyzed under the assumption that it was drawn from more than one population (distinct personality types). Moreover, several external and internal cluster validity criteria were taken into account in order to validate the groupings generated by three different cluster algorithms, which were chosen to represent broad ranges of clustering techniques [ 60 , 82 ]. The Ward followed by K-means procedure covers hierarchical as well as divisive partitioning (crisp) clustering, the latent profile algorithm covers density-based clustering with probabilistic models and information theoretic validation (AIC, BIC), and spectral clustering represents graph theoretic as well as kernel-based non-linear clustering techniques. The results showed a clear superiority of the five-cluster solution. Interpreting this grouping based on theory revealed a strong concordance with personality types found in previous studies, which we could ascertain both in absolute mean values and in the Euclidean distances to mean cluster z-scores extracted from 19 previous studies. As no previous study on personality types used that many external and internal cluster validity indices and different clustering algorithms on a large data set of this size, the present study provides substantial support for the personality type theory postulating the existence of resilient, undercontroller, overcontroller, vulnerable-resilient and reserved personality types, which we will refer to with RUO-VR subsequently. Further, our findings concerning lower validity of the LPA cluster solutions compared to the k-means and spectral cluster solutions suggest that clustering techniques based on latent models are less suited for the BFI-S data of the SOEP sample than iterative and deterministic methods based on the k-means procedure or non-linear kernel or graph-based methods. Consequently, the substance of the clustering results by Specht et. al. [ 36 ], which applied latent profile analysis on the SOEP sample, may therefore be limited.

But the question, if the better validity values of the k-means and spectral clustering techniques compared to the LPA indicate a general superiority of these algorithms, a superiority in the field of personality trait clustering or only a superiority in clustering this specific personality trait assessment (BFI-S) in this specific sample (SOEP), remains subject to further studies on personality trait clustering.

When determining the longitudinal predictive validity, the objections raised by Asendorpf [ 53 ] concerning the direct comparison of person-oriented vs. variable-oriented personality descriptions were incorporated by using continuous personality type profile similarity based on Cronbach and Gleser [ 75 ] instead of dichotomous dummy variables as well as by predicting long-term instead of cross-sectionally assessed variables. Using continuous profile similarity variables also resolves the problem that potentially important information about members of the same class is lost in categorical personality descriptions [ 15 , 53 , 83 ]. Predictions regarding the association of the personality types with the assessed personality and behavior correlates, including risk propensity, impulsivity, self-esteem, locus of control, patience, cognitive and affective wellbeing as well as health measures, were confirmed.

Overcontrollers showed associations with lower spontaneity/impulsivity, with lower mental and physical health, and lower cognitive as well as affective wellbeing. Undercontrollers were mainly associated with higher risk propensity and higher impulsive behavior. These results can be explained through the connection of internalizing and externalizing behavior with the overcontroller and undercontroller types [ 5 – 7 , 78 ] and further with the connection of internalizing problems with somatic symptoms and/or symptoms of depressiveness and anxiety [ 79 ]. The dimensions or categories of internalizing and externalizing psychopathology have a long tradition in child psychopathology [ 84 , 85 ] and have been subsequently replicated in adult psychopathology [ 86 , 87 ] and are now basis of contemporary approaches to general psychopathology [ 88 ]. A central proceeding in this development is the integration of (maladaptive) personality traits into the taxonomy of general psychopathology. In the current approach, maladaptive personality traits are allocated to psychopathology spectra, such as the maladaptive trait domain negative affectivity to the spectrum of internalizing disorders. However, the findings of this study suggests that not specific personality traits are intertwined with the development or the occurrence of psychopathology but specific constellations of personality traits, in other words, personality profiles. This hypothesis is also supported by the findings of Meeus et al. [ 8 ], which investigated longitudinal transitions from one personality type to another with respect to symptoms of generalized anxiety disorder. Transitions from resilient to overcontroller personality profiles significantly predicted higher anxiety symptoms while the opposite was found for transitions from overcontroller to resilient personality profiles.

The resilient personality type had the strongest associations with external locus of control, higher patience, good health and positive wellbeing. This not only confirms the characteristics of the resilient type already described by Block & Block [ 18 ] and subsequently replicated, but also conveys the main characteristics of the construct of resilience itself. While the development of resiliency depends on the quality of attachment experiences in childhood and youth [ 89 ], resiliency in adulthood seems to be closely linked to internal locus of control, self-efficacy and self-esteem. In other words, the link between secure attachment experiences in childhood and resiliency in adulthood seems to be the development of a resilient personality trait profile. Seen the other way around, the link between traumatic attachment experiences or destructive environmental factors and low resiliency in adulthood may be, besides genetic risk factors, the development of personality disorders [ 90 ] or internalizing or externalizing psychopathology [ 91 ]. Following this thought, the p-factor [ 92 ], i.e. a general factor of psychopathology, may be an index of insufficient resilience. Although from the viewpoint of personality pathology, having a trait profile close to the resilient personality type may be an index of stable or good personality structure [ 93 ], i.e. personality functioning [ 94 ], which, though being consistently associated with general psychopathology and psychosocial functioning, should not be confused with it [ 95 ].

The reserved personality type had the strongest associations with higher patience as well as better mental health. The vulnerable-resilient personality type showed low positive correlations with spontaneity/impulsivity and low negative correlations with patience as well as health and affective wellbeing.

Analyzing the correlations of the dimensional Big Five values with the predicted variables revealed patterns similar to the mean z-scores of the personality types resilient, overcontrollers, undercontrollers and reserved. Most variables had a low to moderate correlation with just one personality profile similarity, while having at least two or three low to moderate correlations with the Big Five measures. This can be seen as support for the argument of Chapman [ 82 ] and Asendorpf [ 15 , 53 ] that personality types have more practical meaning in the prediction of more complex correlates of human behavior and personality such as mental and physical health, wellbeing, risk-taking, locus of control, self-esteem and impulsivity. Our findings further underline that the person-oritented approach may better be suited than variable-oriented personality descriptions to detect complex trait interactions [ 40 ]. E.g. the vulnerable-resilient and the overcontroller type did not differ in their high average neuroticism values, while differing in their correlations to mental and somatic health self-report measures. It seems that high neuroticism is far stronger associated to lower mental and physical health as well as wellbeing if it occurs together with low extraversion and low openness as seen in the overcontroller type. This differential association between the Big-Five traits also affects the correlation between neuroticism and self-esteem or locus of control. Not differing in their average neuroticism value, the overcontroller personality profile had moderate associations with low self-esteem and external locus of control while the vulnerable-resilient personality profile did only show very low or no association. Further remarkable is that the vulnerable-resilient profile similarity had no significant correlation with measures of cognitive wellbeing while being negatively correlated with affective wellbeing. This suggests that individuals with a Big-Five personality profile similar to the vulnerable-resilient prototype seem not to perceive impairments in their wellbeing, at least on a cognitive layer, although having high z-values in neuroticism. Another explanation for this discrepancy as well as for the lack of association of the vulnerable-resilient personality profile to low self-esteem and external locus of control though having high values in neuroticism could be found in the research on the construct of resilience. Personalities with high neuroticism values but stable self-esteem, internal locus of control and above average agreeableness and extraversion values may be the result of the interplay of multiple protective factors (e.g. close bond with primary caregiver, supportive teachers) with risk factors (e.g. parental mental illness, poverty). The development of a resilient personality profile with below average neuroticism values, on the other hand, may be facilitated if protective factors outweigh the risk factors by a higher ratio.

An interesting future research question therefore concerns to what extent personality types found in this study may be replicated using maladaptive trait assessments according to DSM-5, section III [ 96 ] or the ICD-11 personality disorder section [ 97 ] (for a comprehensive overview on that topic see e.g. [ 98 ]). As previous studies showed that both DSM-5 [ 99 ] and ICD-11 [ 100 ] maladaptive personality trait domains may be, to a large extent, conceptualized as maladaptive variants of Big Five traits, it is highly likely that also maladaptive personality trait domains align around personality prototypes and that the person-oriented approach may amend the research field of personality pathology [ 101 ].

Taken together, the findings of this study connect the variable centered approach of personality description, more precisely the Big Five traits, through the concept of personality types to constructs of developmental psychology (resiliency, internalizing and externalizing behavior and/or problems) as well as clinical psychology (mental health) and general health assessed by the SF-12. We could show that the distribution of Big Five personality profiles, at least in the large representative German sample of this study, aggregates around five prototypes, which in turn have distinct associations to other psychological constructs, most prominently resilience, internalizing and externalizing behavior, subjective health, patience and wellbeing.

Limitations

Several limitations of the present study need to be considered: One problem concerns the assessment of patience, self-esteem and impulsivity. From a methodological perspective, these are not suitable for the assessment of construct validity as they were assessed with only one item. A further weakness is the short Big Five inventory with just 15 items. Though showing acceptable reliability, 15 items are more prone to measurement errors than measures with more items and only allow a very broad assessment of the 5 trait domains, without information on individual facet expressions. A more big picture question is if the Big Five model is the best way to assess personality in the first place. A further limitation concerns the interpretation of the subjective health measures, as high neuroticism is known to bias subjective health ratings. But the fact that the vulnerable-resilient and the overcontroler type had similar average neuroticism values but different associations with the subjective health measures speaks against a solely neuroticism-based bias driven interpretation of the associations of the self-reported health measures with the found personality clusters. Another limitation is the correlation between the personality type similarities: As they are based on Euclidean distances and the cluster algorithms try to maximize the distances between the cluster centers, proximity to one personality type (that is the cluster mean) logically implies distance from the others. In the case of the vulnerable-resilient and the resilient type, the correlation of the profile similarities is positive, as they mainly differ on only one dimension (neuroticism). These high correlations between the profile similarities prevents or diminishes, due to the emerging high collinearity, the applicability of general linear models, i.e. regression to calculate the exact amount of variance explained by the profile similarities.

The latter issue could be bypassed by assessing types and dimensions with different questionnaires, i.e. as in Asendorpf [ 15 ] with the California Child Q-set to determine the personality type and the NEO-FFI for the Big Five dimensions. Another possibility is to design a new questionnaire based on the various psychological constructs that are distinctly associated with each personality type, which is probably a subject for future person-centered research.

Acknowledgments

The data used in this article were made available by the German Socio-Economic Panel (SOEP, Data for years 1984–2015) at the German Institute for Economic Research, Berlin, Germany. However, the findings and views reported in this article are those of the authors. To ensure the confidentiality of respondents’ information, the SOEP adheres to strict security standards in the provision of SOEP data. The data are reserved exclusively for research use, that is, they are provided only to the scientific community. All users, both within the EEA (and Switzerland) and outside these countries, are required to sign a data distribution contract.

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Personality types revisited–a literature-informed and data-driven approach to an integration of prototypical and dimensional constructs of personality description

1 Department of Psychology, Freie Universität Berlin, Berlin, Germany

Marcus Roth

2 Department of Psychology, University of Duisburg-Essen, Duisburg Germany

Philipp Yorck Herzberg

3 Personality Psychology and Psychological Assessment Unit, Helmut Schmidt University of the Federal Armed Forces Hamburg, Hamburg, Germany

Associated Data

The data used in this article were made available by the German Socio-Economic Panel (SOEP, Data for years 1984-2015) at the German Institute for Economic Research, Berlin, Germany. To ensure the confidentiality of respondents’ information, the SOEP adheres to strict security standards in the provision of SOEP data. The data are reserved exclusively for research use, that is, they are provided only to the scientific community. To require full access to the data used in this study, it is required to sign a data distribution contract. All contact informations and the procedure to request the data can be obtained at: https://www.diw.de/en/diw_02.c.222829.en/access_and_ordering.html .

A new algorithmic approach to personality prototyping based on Big Five traits was applied to a large representative and longitudinal German dataset (N = 22,820) including behavior, personality and health correlates. We applied three different clustering techniques, latent profile analysis, the k-means method and spectral clustering algorithms. The resulting cluster centers, i.e. the personality prototypes, were evaluated using a large number of internal and external validity criteria including health, locus of control, self-esteem, impulsivity, risk-taking and wellbeing. The best-fitting prototypical personality profiles were labeled according to their Euclidean distances to averaged personality type profiles identified in a review of previous studies on personality types. This procedure yielded a five-cluster solution: resilient, overcontroller, undercontroller, reserved and vulnerable-resilient. Reliability and construct validity could be confirmed. We discuss wether personality types could comprise a bridge between personality and clinical psychology as well as between developmental psychology and resilience research.

Introduction

Although documented theories about personality types reach back more than 2000 years (i.e. Hippocrates’ humoral pathology), and stereotypes for describing human personality are also widely used in everyday psychology, the descriptive and variable-oriented assessment of personality, i.e. the description of personality on five or six trait domains, has nowadays consolidated its position in modern personality psychology.

In recent years, however, the person-oriented approach, i.e. the description of an individual personality by its similarity to frequently occurring prototypical expressions, has amended the variable-oriented approach with the addition of valuable insights into the description of personality and the prediction of behavior. Focusing on the trait configurations, the person-oriented approach aims to identify personality types that share the same typical personality profile [ 1 ].

Nevertheless, the direct comparison of the utility of person-oriented vs. variable-oriented approaches to personality description yielded mixed results. For example Costa, Herbst, McCrae, Samuels and Ozer [ 2 ] found a higher amount of explained variance in predicting global functioning, geriatric depression or personality disorders for the variable-centered approach using Big Five personality dimensions. But these results also reflect a methodological caveat of this approach, as the categorical simplification of dimensionally assessed variables logically explains less variance. Despite this, the person-centered approach was found to heighten the predictability of a person’s behavior [ 3 , 4 ] or the development of adolescents in terms of internalizing and externalizing symptoms or academic success [ 5 , 6 ], problem behavior, delinquency and depression [ 7 ] or anxiety symptoms [ 8 ], as well as stress responses [ 9 ] and social attitudes [ 10 ]. It has also led to new insights into the function of personality in the context of other constructs such as adjustment [ 2 ], coping behavior [ 11 ], behavioral activation and inhibition [ 12 ], subjective and objective health [ 13 ] or political orientation [ 14 ], and has greater predictive power in explaining longitudinally measured individual differences in more temperamental outcomes such as aggressiveness [ 15 ].

However, there is an ongoing debate about the appropriate number and characteristics of personality prototypes and whether they perhaps constitute an methodological artifact [ 16 ].

With the present paper, we would like to make a substantial contribution to this debate. In the following, we first provide a short review of the personality type literature to identify personality types that were frequently replicated and calculate averaged prototypical profiles based on these previous findings. We then apply multiple clustering algorithms on a large German dataset and use those prototypical profiles generated in the first step to match the results of our cluster analysis to previously found personality types by their Euclidean distance in the 5-dimensional space defined by the Big Five traits. This procedure allows us to reliably link the personality prototypes found in our study to previous empirical evidence, an important analysis step lacking in most previous studies on this topic.

The empirical ground of personality types

The early studies applying modern psychological statistics to investigate personality types worked with the Q-sort procedure [ 1 , 15 , 17 ], and differed in the number of Q-factors. With the Q-Sort method, statements about a target person must be brought in an order depending on how characteristic they are for this person. Based on this Q-Sort data, prototypes can be generated using Q-Factor Analysis, also called inverse factor analysis. As inverse factor analysis is basically interchanging variables and persons in the data matrix, the resulting factors of a Q-factor analysis are prototypical personality profiles and not hypothetical or latent variable dimensions. On this basis, personality types (groups of people with similar personalities) can be formed in a second step by assigning each person to the prototype with whose profile his or her profile correlates most closely. All of these early studies determined at least three prototypes, which were labeled resilient, overcontroler and undercontroler grounded in Block`s theory of ego-control and ego-resiliency [ 18 ]. According to Jack and Jeanne Block’s decade long research, individuals high in ego-control (i.e. the overcontroler type) tend to appear constrained and inhibited in their actions and emotional expressivity. They may have difficulty making decisions and thus be non-impulsive or unnecessarily deny themselves pleasure or gratification. Children classified with this type in the studies by Block tend towards internalizing behavior. Individuals low in ego-control (i.e. the undercontroler type), on the other hand, are characterized by higher expressivity, a limited ability to delay gratification, being relatively unattached to social standards or customs, and having a higher propensity to risky behavior. Children classified with this type in the studies by Block tend towards externalizing behavior.

Individuals high in Ego-resiliency (i.e. the resilient type) are postulated to be able to resourcefully adapt to changing situations and circumstances, to tend to show a diverse repertoire of behavioral reactions and to be able to have a good and objective representation of the “goodness of fit” of their behavior to the situations/people they encounter. This good adjustment may result in high levels of self-confidence and a higher possibility to experience positive affect.

Another widely used approach to find prototypes within a dataset is cluster analysis. In the field of personality type research, one of the first studies based on this method was conducted by Caspi and Silva [ 19 ], who applied the SPSS Quick Cluster algorithm to behavioral ratings of 3-year-olds, yielding five prototypes: undercontrolled, inhibited, confident, reserved, and well-adjusted.

While the inhibited type was quite similar to Block`s overcontrolled type [ 18 ] and the well-adjusted type was very similar to the resilient type, two further prototypes were added: confident and reserved. The confident type was described as easy and responsive in social interaction, eager to do exercises and as having no or few problems to be separated from the parents. The reserved type showed shyness and discomfort in test situations but without decreased reaction speed compared to the inhibited type. In a follow-up measurement as part of the Dunedin Study in 2003 [ 20 ], the children who were classified into one of the five types at age 3 were administered the MPQ at age 26, including the assessment of their individual Big Five profile. Well-adjusteds and confidents had almost the same profiles (below-average neuroticism and above average on all other scales except for extraversion, which was higher for the confident type); undercontrollers had low levels of openness, conscientiousness and openness to experience; reserveds and inhibiteds had below-average extraversion and openness to experience, whereas inhibiteds additionally had high levels of conscientiousness and above-average neuroticism.

Following these studies, a series of studies based on cluster analysis, using the Ward’s followed by K-means algorithm, according to Blashfield & Aldenderfer [ 21 ], on Big Five data were published. The majority of the studies examining samples with N < 1000 [ 5 , 7 , 22 – 26 ] found that three-cluster solutions, namely resilients, overcontrollers and undercontrollers, fitted the data the best. Based on internal and external fit indices, Barbaranelli [ 27 ] found that a three-cluster and a four-cluster solution were equally suitable, while Gramzow [ 28 ] found a four-cluster solution with the addition of the reserved type already published by Caspi et al. [ 19 , 20 ]. Roth and Collani [ 10 ] found that a five-cluster solution fitted the data the best. Using the method of latent profile analysis, Merz and Roesch [ 29 ] found a 3-cluster, Favini et al. [ 6 ] found a 4-cluster solution and Kinnunen et al. [ 13 ] found a 5-cluster solution to be most appropriate.

Studies examining larger samples of N > 1000 reveal a different picture. Several favor a five-cluster solution [ 30 – 34 ] while others favor three clusters [ 8 , 35 ]. Specht et al. [ 36 ] examined large German and Australian samples and found a three-cluster solution to be suitable for the German sample and a four-cluster solution to be suitable for the Australian sample. Four cluster solutions were also found to be most suitable to Australian [ 37 ] and Chinese [ 38 ] samples. In a recent publication, the authors cluster-analysed very large datasets on Big Five personality comprising more than 1,5 million online participants using Gaussian mixture models [ 39 ]. Albeit their results “provide compelling evidence, both quantitatively and qualitatively, for at least four distinct personality types”, two of the four personality types in their study had trait profiles not found previously and all four types were given labels unrelated to previous findings and theory. Another recent publication [ 40 ] cluster-analysing data of over 270,000 participants on HEXACO personality “provided evidence that a five-profile solution was optimal”. Despite limitations concerning the comparability of HEXACO trait profiles with FFM personality type profiles, the authors again decided to label their personality types unrelated to previous findings instead using agency-communion and attachment theories.

We did not include studies in this literature review, which had fewer than 199 participants or those which restricted the number of types a priori and did not use any method to compare different clustering solutions. We have made these decisions because a too low sample size increases the probability of the clustering results being artefacts. Further, a priori limitation of the clustering results to a certain number of personality types is not well reasonable on the base of previous empirical evidence and again may produce artefacts, if the a priori assumed number of clusters does not fit the data well.

To gain a better overview, we extracted all available z-scores from all samples of the above-described studies. Fig 1 shows the averaged z-scores extracted from the results of FFM clustering solutions for all personality prototypes that occurred in more than one study. The error bars represent the standard deviation of the distribution of the z-scores of the respective trait within the same personality type throughout the different studies. Taken together the resilient type was replicated in all 19 of the mentioned studies, the overcontroler type in 16, the undercontroler personality type in 17 studies, the reserved personality type was replicated in 6 different studies, the confident personality type in 4 and the non-desirable type was replicated twice.

An external file that holds a picture, illustration, etc.
Object name is pone.0244849.g001.jpg

Average Big Five z-scores of personality types based on clustering of FFM datasets with N ≥ 199 that were replicated at least once. Error bars indicate the standard deviation of the repective trait within the respective personality type found in the literature [ 5 , 6 , 10 , 22 – 25 , 27 – 31 , 33 – 36 , 38 , 39 , 41 ].

Three implications can be drawn from this figure. First, although the results of 19 studies on 26 samples with a total N of 1,560,418 were aggregated, the Big Five profiles for all types can still be clearly distinguished. In other words, personality types seem to be a phenomenon that survives the aggregation of data from different sources. Second, there are more than three replicable personality types, as there are other replicated personality types that seem to have a distinct Big Five profile, at least regarding the reserved and confident personality types. Third and lastly, the non-desirable type seems to constitute the opposite of the resilient type. Looking at two-cluster solutions on Big Five data personality types in the above-mentioned literature yields the resilient opposed to the non-desirable type. This and the fact that it was only replicated twice in the above mentioned studies points to the notion that it seems not to be a distinct type but rather a combined cluster of the over- and undercontroller personality types. Further, both studies with this type in the results did not find either the undercontroller or the overcontroller cluster or both. Taken together, five distinct personality types were consistently replicated in the literature, namely resilient, overcontroller, undercontroller, reserved and confident. However, inferring from the partly large error margin for some traits within some prototypes, not all personality traits seem to contribute evenly to the occurrence of the different prototypes. While for the overcontroler type, above average neuroticism, below average extraversion and openness seem to be distinctive, only below average conscientiousness and agreeableness seemed to be most characteristic for the undercontroler type. The reserved prototype was mostly characterized by below average openness and neuroticism with above average conscientiousness. Above average extraversion, openness and agreeableness seemed to be most distinctive for the confident type. Only for the resilient type, distinct expressions of all Big Five traits seemed to be equally significant, more precisely below average neuroticism and above average extraversion, openness, agreeableness and conscientiousness.

Research gap and novelty of this study

The cluster methods used in most of the mentioned papers were the Ward’s followed by K-means method or latent profile analysis. With the exception of Herzberg and Roth [ 30 ], Herzberg [ 33 ], Barbaranelli [ 27 ] and Steca et. al. [ 25 ], none of the studies used internal or external validity indices other than those which their respective algorithm (in most cases the SPSS software package) had already included. Gerlach et al. [ 39 ] used Gaussian mixture models in combination with density measures and likelihood measures.

The bias towards a smaller amount of clusters resulting from the utilization of just one replication index, e.g. Cohen's Kappa calculated by split-half cross-validation, which was ascertained by Breckenridge [ 42 ] and Overall & Magee [ 43 ], is probably the reason why a three-cluster solution is preferred in most studies. Herzberg and Roth [ 30 ] pointed to the study by Milligan and Cooper [ 44 ], which proved the superiority of the Rand index over Cohen's Kappa and also suggested a variety of validity metrics for internal consistency to examine the construct validity of the cluster solutions.

Only a part of the cited studies had a large representative sample of N > 2000 and none of the studies used more than one clustering algorithm. Moreover, with the exception of Herzberg and Roth [ 30 ] and Herzberg [ 33 ], none of the studies used a large variety of metrics for assessing internal and external consistency other than those provided by the respective clustering program they used. This limitation further adds up to the above mentioned bias towards smaller amounts of clusters although the field of cluster analysis and algorithms has developed a vast amount of internal and external validity algorithms and criteria to tackle this issue. Further, most of the studies had few or no other assessments or constructs than the Big Five to assess construct validity of the resulting personality types. Herzberg and Roth [ 30 ] and Herzberg [ 33 ] as well, though using a diverse variety of validity criteria only used one clustering algorithm on a medium-sized dataset with N < 2000.

Most of these limitations also apply to the study by Specht et. al. [ 36 ], which investigated two measurement occasions of the Big Five traits in the SOEP data sample. They used only one clustering algorithm (latent profile analysis), no other algorithmic validity criteria than the Bayesian information criterion and did not utilize any of the external constructs also assessed in the SOEP sample, such as mental health, locus of control or risk propensity for construct validation.

The largest sample and most advanced clustering algorithm was used in the recent study by Gerlach et al. [ 39 ]. But they also used only one clustering algorithm, and had no other variables except Big Five trait data to assess construct validity of the resulting personality types.

The aim of the present study was therefore to combine different methodological approaches while rectifying the shortcomings in several of the studies mentioned above in order to answer the following exploratory research questions: Are there replicable personality types, and if so, how many types are appropriate and in which constellations are they more (or less) useful than simple Big Five dimensions in the prediction of related constructs?

Three conceptually different clustering algorithms were used on a large representative dataset. The different solutions of the different clustering algorithms were compared using methodologically different internal and external validity criteria, in addition to those already used by the respective clustering algorithm.

To further examine the construct validity of the resulting personality types, their predictive validity in relation to physical and mental health, wellbeing, locus of control, self-esteem, impulsivity, risk-taking and patience were assessed.

Mental health and wellbeing seem to be associated mostly with neuroticism on the variable-oriented level [ 45 ], but on a person-oriented level, there seem to be large differences between the resilient and the overcontrolled personality type concerning perceived health and well-being beyond mean differences in neuroticism [ 33 ]. This seems also to be the case for locus of control and self-esteem, which is associated with neuroticism [ 46 ] and significantly differs between resilient and overcontrolled personality type [ 33 ]. On the other hand, impulsivity and risk taking seem to be associated with all five personality traits [ 47 ] and e.g. risky driving or sexual behavior seem to occur more often in the undercontrolled personality type [ 33 , 48 ].

We chose these measures because of their empirically known differential associations to Big Five traits as well as to the above described personality types. So this both offers the opportunity to have an integrative comparison of the variable- and person-centered descriptions of personality and to assess construct validity of the personality types resulting from our analyses.

Materials and methods

The acquisition of the data this study bases on was carried out in accordance with the principles of the Basel Declaration and recommendations of the “Principles of Ethical Research and Procedures for Dealing with Scientific Misconduct at DIW Berlin”. The protocol was approved by the Deutsches Institut für Wirtschaftsforschung (DIW).

The data used in this study were provided by the German Socio-Economic Panel Study (SOEP) of the German institute for economic research [ 49 ]. Sample characteristics are shown in Table 1 . The overall sample size of the SOEP data used in this study, comprising all individuals who answered at least one of the Big-Five personality items in 2005 and 2009, was 25,821. Excluding all members with more than one missing answers on the Big Five assessment or intradimensional answer variance more than four times higher than the sample average resulted in a total Big Five sample of N = 22,820, which was used for the cluster analyses. 14,048 of these individuals completed, in addition to the Big Five, items relevant to further constructs examined in this study that were assessed in other years. The 2013 SOEP data Big Five assessment was used as a test sample to examine stability and consistency of the final cluster solution.

Exclusion of participants in the derivation and test samples based on missing answers or intradimensional answer variance more than four times higher than the sample average on the Big Five assessment. Longitudinal construct validity sample consistent of participants with available data on assessments of patience, risk taking, impulsivity, affective and cognitive wellbeing, locus of control, self-esteem and health. SOEP = German Socio-Economic Panel, M = mean, SD = standard deviation, Rg = Range, F = female.

The Big Five were assessed in 2005 2009 and 2013 using the short version of the Big Five inventory (BFI-S). It consists of 15 items, with internal consistencies (Cronbach’s alpha) of the scales ranging from .5 for openness to .73 for openness [ 50 ]. Further explorations showed strong robustness across different assessment methods [ 51 ].

To measure the predictive validity, several other measures assessed in the SOEP were included in the analyses. In detail, these were:

Patience was assessed in 2008 with one item: “Are you generally an impatient person, or someone who always shows great patience?”

Risk taking

Risk-taking propensity was assessed in 2009 by six items asking about the willingness to take risks while driving, in financial matters, in leisure and sports, in one’s occupation (career), in trusting unknown people and the willingness to take health risks, using a scale from 0 (risk aversion) to 10 (fully prepared to take risks). Cronbach’s alpha was .82 for this scale in the current sample.

Impulsivity/Spontaneity

Impulsivity/spontaneity was assessed in 2008 with one item: Do you generally think things over for a long time before acting–in other words, are you not impulsive at all? Or do you generally act without thinking things over for long time–in other words, are you very impulsive?

Affective and cognitive wellbeing

Affect was assessed in 2008 by four items asking about the amount of anxiety, anger, happiness or sadness experienced in the last four weeks on a scale from 1 (very rare) to 5 (very often). Cronbach’s alpha for this scale was .66. The cognitive satisfaction with life was assessed by 10 items asking about satisfaction with work, health, sleep, income, leisure time, household income, household duties, family life, education and housing, with a Cronbach’s alpha of .67. The distinction between cognitive and affective wellbeing stems from sociological research based on constructs by Schimmack et al. [ 50 ].

Locus of control

The individual attitude concerning the locus of control, the degree to which people believe in having control over the outcome of events in their lives opposed to being exposed to external forces beyond their control, was assessed in 2010 with 10 items, comprising four positively worded items such as “My life’s course depends on me” and six negatively worded items such as “Others make the crucial decisions in my life”. Items were rated on a 7-point scale ranging from “does not apply” to “does apply”. Cronbach’s alpha in the present sample for locus of control was .57.

Self-esteem

Global self-esteem–a person’s overall evaluation or appraisal of his or her worth–was measured in 2010 with one item: “To what degree does the following statement apply to you personally?: I have a positive attitude toward myself”.

To assess subjective health, the 12-Item Short Form Health Survey (SF-12) was integrated into the SOEP questionnaire and assessed in 2002, 2004, 2006, 2008 and 2010. In the present study, we used the data from 2008 and 2010. The SF-12 is a short form of the SF-36, a self-report questionnaire to assess the non-disease-specific health status [ 52 ]. Within the SF-12, items can be grouped onto two subscales, namely the physical component summary scale, with items asking about physical health correlates such as how exhausting it is to climb stairs, and the mental component summary scale, with items asking about mental health correlates such as feeling sad and blue. The literature on health measures often distinguishes between subjective and objective health measures (e.g., BMI, blood pressure). From this perspective, the SF-12 would count as a subjective health measure. In the present sample, Cronbach’s alpha for the SF-12 items was .77.

Derivation of the prototypes

The first step was to administer three different clustering methods on the Big Five data of the SOEP sample: First, the conventional linear clustering method used by Asendorpf [ 15 , 35 , 53 ] and also Herzberg and Roth [ 30 ] combines the hierarchical clustering method of Ward [ 54 ] with the k-means algorithm [ 55 ]. This algorithm generates a first guess of personality types based on hierarchical clustering, and then uses this first guess as starting points for the k-means-method, which iteratively adjusts the personality profiles, i.e. the cluster means to minimize the error of allocation, i.e. participants with Big Five profiles that are allocated to two or more personality types. The second algorithm we used was latent profile analysis with Mclust in R [ 56 ], an algorithm based on probabilistic finite mixture modeling, which assumes that there are latent classes/profiles/mixture components underlying the manifest observed variables. This algorithm generates personality profiles and iteratively calculates the probability of every participant in the data to be allocated to one of the personality types and tries to minimize an error term using maximum likelihood method. The third algorithm was spectral clustering, an algorithm which initially computes eigenvectors of graph Laplacians of the similarity graph constructed on the input data to discover the number of connected components in the graph, and then uses the k-means algorithm on the eigenvectors transposed in a k-dimensional space to compute the desired k clusters [ 57 ]. As it is an approach similar to the kernel k-means algorithm [ 58 ], spectral clustering can discover non-linearly separable cluster formations. Thus, this algorithm is able, in contrast to the standard k-means procedure, to discover personality types having unequal or non-linear distributions within the Big-Five traits, e.g. having a small SD on neuroticism while having a larger SD on conscientiousness or a personality type having high extraversion and either high or low agreeableness.

Within the last 50 years, a large variety of clustering algorithms have been established, and several attempts have been made to group them. In their book about cluster analysis, Bacher et al. [ 59 ] group cluster algorithms into incomplete clustering algorithms, e.g. Q-Sort or multidimensional scaling, deterministic clustering, e.g. k-means or nearest-neighbor algorithms, and probabilistic clustering, e.g. latent class and latent profile analysis. According to Jain [ 60 ], cluster algorithms can be grouped by their objective function, probabilistic generative models and heuristics. In his overview of the current landscape of clustering, he begins with the group of density-based algorithms with linear similarity functions, e.g. DBSCAN, or probabilistic models of density functions, e.g. in the expectation-maximation (EM) algorithm. The EM algorithm itself also belongs to the large group of clustering algorithms with an information theoretic formulation. Another large group according to Jain is graph theoretic clustering, which includes several variants of spectral clustering. Despite the fact that it is now 50 years old, Jain states that k-means is still a good general-purpose algorithm that can provide reasonable clustering results.

The clustering algorithms chosen for the current study are therefore representatives of the deterministic vs. probabilistic grouping according to Bacher et. al. [ 59 ], as well as representatives of the density-based, information theoretic and graph theoretic grouping according to Jain [ 60 ].

Determining the number of clusters

There are two principle ways to determine cluster validity: external or relative criteria and internal validity indices.

External validity criteria

External validity criteria measure the extent to which cluster labels match externally supplied class labels. If these external class labels originate from another clustering algorithm used on the same data sample, the resulting value of the external cluster validity index is relative. Another method, which is used in the majority of the cited papers in section 1, is to randomly split the data in two halves, apply a clustering algorithm on both halves, calculate the cluster means and allocate members of one half to the calculated clusters of the opposite half by choosing the cluster mean with the shortest Euclidean distance to the data member in charge. If the cluster algorithm allocation of one half is then compared with the shortest Euclidean distance allocation of the same half by means of an external cluster validity index, this results in a value for the reliability of the clustering method on the data sample.

As allocating data points/members by Euclidean distances always yields spherical and evenly shaped clusters, it will favor clustering methods that also yield spherical and evenly shaped clusters, as it is the case with standard k-means. The cluster solutions obtained with spectral clustering as well as latent profile analysis (LPA) are not (necessarily) spherical or evenly shaped; thus, allocating members of a dataset by their Euclidean distances to cluster means found by LPA or spectral clustering does not reliably represent the structure of the found cluster solution. This is apparent in Cohen’s kappa values <1 if one uses the Euclidean external cluster assignment method comparing a spectral cluster solution with itself. Though by definition, Cohen’s kappa should be 1 if the two ratings/assignments compared are identical, which is the case when comparing a cluster solution (assigning every data point to a cluster) with itself. This problem can be bypassed by allocating the members of the test dataset to the respective clusters by training a support vector machine classifier for each cluster. Support vector machines (SVM) are algorithms to construct non-linear “hyperplanes” to classify data given their class membership [ 61 ]. They can be used very well to categorize members of a dataset by an SVM-classifier trained on a different dataset. Following the rationale not to disadvantage LPA and spectral clustering in the calculation of the external validity, we used an SVM classifier to calculate the external validity criteria for all clustering algorithms in this study.

To account for the above mentioned bias to smaller numbers of clusters we applied three external validity criteria: Cohen’s kappa, the Rand index [ 62 ] and the Hubert-Arabie adjusted Rand index [ 63 ].

Internal validity criteria

Again, to account for the bias to smaller numbers of clusters, we also applied multiple internal validity criteria selected in line with the the following reasoning: According to Lam and Yan [ 64 ], the internal validity criteria fall into three classes: Class one includes cost-function-based indices, e.g. AIC or BIC [ 65 ], whereas class two comprises cluster-density-based indices, e.g. the S_Dbw index [ 66 ]. Class three is grounded on geometric assumptions concerning the ratio of the distances within clusters compared to the distances between the clusters. This class has the most members, which differ in their underlying mathematics. One way of assessing geometric cluster properties is to calculate the within- and/or between-group scatter, which both rely on summing up distances of the data points to their barycenters (cluster means). As already explained in the section on external criteria, calculating distances to cluster means will always favor spherical and evenly shaped cluster solutions without noise, i.e. personality types with equal and linear distributions on the Big Five trait dimensions, which one will rarely encounter with natural data.

Another way not solely relying on distances to barycenters or cluster means is to calculate directly with the ratio of distances of the data points within-cluster and between-cluster. According to Desgraupes [ 67 ], this applies to the following indices: the C-index, the Baker & Hubert Gamma index, the G(+) index, Dunn and Generalized Dunn indices, the McClain-Rao index, the Point-Biserial index and the Silhouette index. As the Gamma and G(+) indices rely on the same mathematical construct, one can declare them as redundant. According to Bezdek [ 68 ], the Dunn index is very sensitive to noise, even if there are only very few outliers in the data. Instead, the authors propose several ways to compute a Generalized Dunn index, some of which also rely on the calculation of barycenters. The best-performing GDI algorithm outlined by Bezdek and Pal [ 68 ] which does not make use of cluster barycenters is a ratio of the mean distance of every point between clusters to the maximum distance between points within the cluster, henceforth called GDI31. According to Vendramin et al. [ 69 ], the Gamma, C-, and Silhouette indices are the best-performing (over 80% correct hit rate), while the worst-performing are the Point-Biserial and the McClain-Rao indices (73% and 51% correct hit rate, respectively).

Fig 2 shows a schematic overview of the procedure we used to determine the personality types Big Five profiles, i.e. the cluster centers. To determine the best fitting cluster solution, we adopted the two-step procedure proposed by Blashfield and Aldenfelder [ 21 ] and subsequently used by Asendorpf [ 15 , 35 , 53 ] Boehm [ 41 ], Schnabel [ 24 ], Gramzow [ 28 ], and Herzberg and Roth [ 30 ], with a few adjustments concerning the clustering algorithms and the validity criteria.

An external file that holds a picture, illustration, etc.
Object name is pone.0244849.g002.jpg

LPA = latent profile analysis, SVM = Support Vector Machine.

First, we drew 20 random samples of the full sample comprising all individuals who answered the Big-Five personality items in 2005 and 2009 with N = 22,820 and split every sample randomly into two halves. Second, all three clustering algorithms described above were performed on each half, saving the 3-, 4-,…,9- and 10-cluster solution. Third, participants of each half were reclassified based on the clustering of the other half of the same sample, again for every clustering algorithm and for all cluster solutions from three to 10 clusters. In contrast to Asendorpf [ 35 ], this was implemented not by calculating Euclidean distances, but by training a support vector machine classifier for every cluster of a cluster solution of one half-sample and reclassifying the members of the other half of the same sample by the SVM classifier. The advantages of this method are explained in the section on external criteria. This resulted in 20 samples x 2 halves per sample x 8 cluster solutions x 3 clustering algorithms, equaling 960 clustering solutions to be compared.

The fourth step was to compute the external criteria comparing each Ward followed by k-means, spectral, or probabilistic clustering solution of each half-sample to the clustering by the SVM classifier trained on the opposite half of the same sample, respectively. The external calculated in this step were Cohen's kappa, Rand’s index [ 62 ] and the Hubert & Arabie adjusted Rand index [ 63 ]. The fifth step consisted of averaging: We first averaged the external criteria values per sample (one value for each half), and then averaged the 20x4 external criteria values for each of the 3-,4-…, 10-cluster solutions for each algorithm.

The sixth step was to temporarily average the external criteria values for the 3-,4-,… 10-cluster solution over the three clustering algorithms and discard the cluster solutions that had a total average kappa below 0.6.

As proposed by Herzberg and Roth [ 30 ], we then calculated several internal cluster validity indices for all remaining cluster solutions. The internal validity indices which we used were, in particular, the C-index [ 70 ], the Baker-Hubert Gamma index [ 71 ], the G + index [ 72 ], the Generalized Dunn index 31 [ 68 ], the Point-Biserial index [ 44 ], the Silhouette index [ 73 ], AIC and BIC [ 65 ] and the S_Dbw index [ 66 ]. Using all of these criteria, it is possible to determine the best clustering solution in a mathematical/algorithmic manner.

The resulting clusters where then assigned names by calculating Euclidean distances to the clusters/personality types found in the literature, taking the nearest type within the 5-dimensional space defined by the respective Big Five values.

To examine the stability and consistency of the final cluster solution, in a last step, we then used the 2013 SOEP data sample to calculate a cluster solution using the algorithm and parameters which generated the solution with the best validity criteria for the 2005 and 2009 SOEP data sample. The 2013 personality prototypes were allocated to the personality types of the solution from the previous steps by their profile similarity measure D. Stability then was assessed by calculation of Rand-index, adjusted Rand-index and Cohen’s Kappa for the complete solution and for every single personality type. To generate the cluster allocations between the different cluster solutions, again we used SVM classifier as described above.

To assess the predictive and the construct validity of the resulting personality types, the inversed Euclidean distance for every participant to every personality prototype (averaged Big Five profile in one cluster) in the 5-dimensional Big-Five space was calculated and correlated with further personality, behavior and health measures mentioned above. To ensure that longitudinal reliability was assessed in this step, Big Five data assessed in 2005 were used to predict measures which where assessed three, four or five years later. The selection of participants with available data in 2005 and 2008 or later reduced the sample size in this step to N = 14,048.

Internal and external cluster fit indices

Table 2 shows the mean Cohen’s kappa values, averaged over all clustering algorithms and all 20 bootstrapped data permutations.

Each cell is an average value over 20 independent cluster computations on random data permutations; the mean value in the last row is the average over all cluster algorithms. LPA = latent profile analysis, k-Means = k-Means Clustering algorithm, Spectral = Spectral clustering algorithm.

Whereas the LPA and spectral cluster solutions seem to have better kappa values for fewer clusters, the kappa values of the k-means clustering solutions have a peak at five clusters, which is even higher than the kappa values of the three-cluster solutions of the other two algorithms.

Considering that these values are averaged over 20 independent computations, there is very low possibility that this result is an artefact. As the solutions with more than five clusters had an average kappa below .60, they were discarded in the following calculations.

Table 3 shows the calculated external and internal validity indices for the three- to five-cluster solutions, ordered by the clustering algorithm. Comparing the validity criterion values within the clustering algorithms reveals a clear preference for the five-cluster solution in the spectral as well as the Ward followed by k-means algorithm.

Best value across all solutions for each validity criterion is highlighted in yellow, best value within the respective algorithm in blue. GDI31 = Generalized Dunn Index 31, AIC = Akaike’s information criterion, BIC = Bayesian information criterion, LPA = latent profile analysis, k-Means = k-Means Clustering algorithm, Spectral = Spectral clustering algorithm.

Looking solely at the cluster validity results of the latent profile models, they seem to favor the three-cluster model. Yet, in a global comparison, only the S_Dbw index continues to favor the three-cluster LPA model, whereas the results of all other 12 validity indices support five-cluster solutions. The best clustering solution in terms of the most cluster validity index votes is the five-cluster Ward followed by k-means solution, and second best is the five-cluster spectral solution. It is particularly noteworthy that the five-cluster K-means solution has higher values on all external validity criteria than all other solutions. As these values are averaged over 20 independent cluster computations on random data permutations, and still have better values than solutions with fewer clusters despite the fact that these indices have a bias towards solutions with fewer clusters [ 42 ], there seems to be a substantial, replicable five-component structure in the Big Five Data of the German SOEP sample.

Description of the prototypes

The mean z-scores on the Big Five factors of the five-cluster k-means as well as the spectral solution are depicted in Fig 2 . Also depicted is the five-cluster LPA solution, which is, despite having poor internal and external validity values compared to the other two solutions, more complicated to interpret. To find the appropriate label for the cluster partitions, the respective mean z-scores on the Big Five factors were compared with the mean z-scores found in the literature, both visually and by the Euclidean distance.

The spectral and the Ward followed by k-means solution overlap by 81.3%; the LPA solution only overlaps with the other two solutions by 21% and 23%, respectively. As the Ward followed by k-means solution has the best values both for external and internal validity criteria, we will focus on this solution in the following.

The first cluster has low neuroticism and high values on all other scales and includes on average 14.4% of the participants (53.2% female; mean age 53.3, SD = 17.3). Although the similarity to the often replicated resilient personality type is already very clear merely by looking at the z-scores, a very strong congruence is also revealed by computing the Euclidean distance (0.61). The second cluster is mainly characterized by high neuroticism, low extraversion and low openness and includes on average 17.3% of the participants (54.4% female; mean age 57.6, SD = 18.2). It clearly resembles the overcontroller type, to which it also has the shortest Euclidean distance (0.58). The fourth cluster shows below-average values on the factors neuroticism, extraversion and openness, as opposed to above-average values on openness and conscientiousness. It includes on average 22.5% of the participants (45% female; mean age 56.8, SD = 17.6). Its mean z-scores closely resemble the reserved personality type, to which it has the smallest Euclidean distance (0.36). The third cluster is mainly characterized by low conscientiousness and low openness, although in the spectral clustering solution, it also has above-average extraversion and openness values. Computing the Euclidean distance (0.86) yields the closest proximity to the undercontroller personality type. This cluster includes on average 24.6% of the participants (41.3% female; mean age 50.8, SD = 18.3). The fifth cluster exhibits high z-scores on every Big Five trait, including a high value for neuroticism. Computing the Euclidean distances to the previously found types summed up in Fig 1 reveals the closest resemblance with the confident type (Euclidean distance = 0.81). Considering the average scores of the Big Five traits, it resembles the confident type from Herzberg and Roth [ 30 ] and Collani and Roth [ 10 ] as well as the resilient type, with the exception of the high neuroticism score. Having above average values on the more adaptive traits while having also above average neuroticism values reminded a reviewer from a previous version of this paper of the vulnerable but invincible children of the Kauai-study [ 74 ]. Despite having been exposed to several risk factors in their childhood, they were well adapted in their adulthood except for low coping efficiency in specific stressful situations. Taken together with the lower percentage of participants in the resilient cluster in this study, compared to previous studies, we decided to name the 5 th cluster vulnerable-resilient. Consequently, only above or below average neuroticism values divided between resilient and vulnerable resilient. On average, 21.2% of the participants were allocated to this cluster (68.3% female; mean age 54.9, SD = 17.4).

Summarizing the descriptive statistics, undercontrollers were the “youngest” cluster whereas overcontrollers were the “oldest”. The mean age differed significantly between clusters ( F [4, 22820] = 116.485, p <0.001), although the effect size was small ( f = 0.14). The distribution of men and women between clusters differed significantly (c 2 [ 4 ] = 880.556, p <0.001). With regard to sex differences, it was particularly notable that the vulnerable-resilient cluster comprised only 31.7% men. This might be explained by general sex differences on the Big Five scales. According to Schmitt et al. [ 75 ], compared to men, European women show a general bias to higher neuroticism (d = 0.5), higher conscientiousness (d = 0.3) and higher extraversion and openness (d = 0.2). As the vulnerable-resilient personality type is mainly characterized by high neuroticism and above-average z-scores on the other scales, it is therefore more likely to include women. In turn, this implies that men are more likely to have a personality profile characterized mainly by low conscientiousness and low openness, which is also supported by our findings, as only 41.3% of the undercontrollers were female.

Concerning the prototypicality of the five-cluster solution compared to the mean values extracted from previous studies, it is apparent that the resilient, the reserved and the overcontroller type are merely exact replications. In contrast to previous findings, the undercontrollers differed from the previous findings cited above in terms of average neuroticism, whereas the vulnerable-resilient type differed from the previously found type (labeled confident) in terms of high neuroticism.

Stability and consistency

Inspecting the five cluster solution using the k-means algorithm on the Big Five data of the 2013 SOEP sample seemed to depict a replication of the above described personality types. This first impression was confirmed by the calculation of the profile similarity measure D between the 2005/2009 and 2013 SOEP sample cluster solutions, which yielded highest similarity for the undercontroler (D = 0.27) and reserved (D = 0.36) personality types, followed by the vulnerable-resilient (D = 0.37), overcontroler (D = 0.44) and resilient (D = 0.50) personality types. Substantial agreement was confirmed by the values of the Rand index (.84) and Cohen’ Kappa (.70) whereas the Hubert Arabie adjusted Rand Index (.58) indicated moderate agreement for the comparison between the kmeans cluster solution for the 2013 SOEP sample and the cluster allocation with an SVM classifier trained on the 2005 and 2009 kmeans cluster solution.

Predictive validity

In view of the aforementioned criticisms that (a) predicting dimensional variables will mathematically favor dimensional personality description models, and (b) using dichotomous predictors will necessarily provide less explanation of variance than a model using five continuous predictors, we used the profile similarity measure D [ 76 ] instead of dichotomous dummy variables accounting for the prototype membership. Correlations between the inversed Euclidean similarity measure D to the personality types and patience, risk-taking, spontaneity/impulsivity, locus of control, affective wellbeing, self-esteem and health are depicted in Table 4 .

N = 14048. Except the ones in brackets, only correlations with a significance level ≤ 0.001 are depicted. The highest and lowest correlation in each row are marked in bold. SF-12 = 12-Item Short Form Health Survey.

Patience had the highest association with the reserved personality type (r = .19, p < .001). The propensity to risky behavior, e.g. while driving (r = .17, p < .001), in financial matters (r = .17, p < .001) or in health decisions (r = .13, p < .001) was most highly correlated with the undercontroller personality type. This means that the more similar the Big-Five profile to the above-depicted undercontroller personality prototype, the higher the propensity for risky behavior. The average correlation across all three risk propensity scales with the undercontroller personality type is r = .21, with p < .001. This is in line with the postulations by Block and Block and subsequent replications by Caspi et al. [ 19 , 48 ], Robins et al. [ 1 ] and Herzberg [ 33 ] about the undercontroller personality type. Spontaneity/impulsivity showed the highest correlation with the overcontroller personality type (r = -.18, p<0.001). This is also in accordance with Block and Block, who described this type as being non-impulsive and appearing constrained and inhibited in actions and emotional expressivity.

Concerning locus of control, proximity to the resilient personality profile had the highest correlation with internal locus of control (r = .25, p < .001), and in contrast, the more similar the individual Big-Five profile was to the overcontroller personality type, the higher the propensity for external allocation of control (r = .22, p < .001). This is not only in line with Block and Block’s postulations that the resilient personality type has a good repertoire of coping behavior and therefore perceives most situations as “manageable” as well as with the findings by [ 33 ], but is also in accordance with findings regarding the construct and development of resilience [ 77 , 78 ].

Also in line with the predictions of Block and Block and replicating the findings of Herzberg [ 33 ], self-esteem was correlated the highest with the resilient personality profile similarity (r = .33, p < .001), second highest with the reserved personality profile proximity (r = .15, p < .001), and negatively correlated with the overcontroller personality type (r = -.27, p < .001).

This pattern also applies to affective and cognitive wellbeing as well as physical and mental health measured by the SF-12. Affective wellbeing was correlated the highest with similarity to the resilient personality type (r = .27, p < .001), and second highest with the reserved personality type (r = .23, p < .001). The overcontroller personality type, in contrast, showed a negative correlation with affective (r = -.16, p < .001) and cognitive (r = -21, p < .001) wellbeing. Concerning health, a remarkable finding is that lack of physical health impairment correlated the highest with the resilient personality profile similarity (p = -.23, p < .001) but lack of mental health impairment correlated the highest with the reserved personality type (r = -.15, p < .001). The highest correlation with mental health impairments (r = .11, p < .001), as well as physical health impairments (r = .16, p < .001) was with the overcontroller personality profile similarity. It is striking that although the undercontroller personality profile similarity was associated with risky health behavior, it had a negative association with health impairment measures, in contrast to the overcontroller personality type, which in turn had no association with risky health behavior. This result is in line with the link of internalizing and externalizing behavior with the overcontroller and undercontroller types [ 79 ], respectively. Moreover, it is also in accordance with the association of internalizing problems with somatic symptoms and/or symptoms of depressiveness and anxiety [ 80 ].

A further noteworthy finding is that these associations cannot be solely explained by the high neuroticism of the overcontroller personality type, as the vulnerable-resilient type showed a similar level of neuroticism but no correlation with self-esteem, the opposite correlation with impulsivity, and far lower correlations with health measures or locus of control. The vulnerable-resilient type showed also a remarkable distinction to the other types concerning the correlations to wellbeing. While for all other types, the direction and significance of the correlations to affective and cognitive measures of wellbeing were alike, the vulnerable-resilient type only had a significant negative correlation to affective wellbeing while having no significant correlation to measures of cognitive wellbeing.

To provide an overview of the particular associations of the Big Five values with all of the above-mentioned behavior and personality measures, Table 5 shows the bivariate correlations.

N = 14,048. Except the one in brackets, only correlations with a significance level ≤ 0.001 are depicted. SF-12 = 12-Item Short Form Health Survey.

Investigating the direction of the correlation and the relativity of each value to each other row-wise reveals, to some extent, a clear resemblance with the z-scores of the personality types shown in Fig 3 . Correlation profiles of risk taking, especially the facet risk-taking in health issues and locus of control, clearly resemble the undercontroller personality profile (negative correlations with openness and conscientiousness, positive but lower correlations with extraversion and openness). Patience had negative correlations with neuroticism and extraversion, and positive correlations with openness and conscientiousness, which in turn resembles the z-score profile of the reserved personality profile. Spontaneity/impulsivity had moderate to high positive correlations with extraversion and openness, and low negative correlations with openness and neuroticism, which resembles the inverse of the overcontroller personality profile. Self-esteem as well as affective and cognitive wellbeing correlations with the Big Five clearly resemble the resilient personality profile: negative correlations with neuroticism, and positive correlations with extraversion, openness, openness and conscientiousness. Inspecting the SF-12 health correlation, in terms of both physical and mental health, reveals a resemblance to the inversed resilient personality profile (high correlation with neuroticism, low correlation with extraversion, openness, openness and conscientiousness, as well as a resemblance with the overcontroller profile (positive correlation with neuroticism, negative correlation with extraversion).

An external file that holds a picture, illustration, etc.
Object name is pone.0244849.g003.jpg

On the variable level, neuroticism had the highest associations with almost all of the predicted variables, with the exception of impulsivity, which was mainly correlated with extraversion and openness. It is also evident that all variables in question here are correlated with three or more Big Five traits. This can be seen as support for hypothesis that the concept of personality prototypes has greater utility than the variable-centered approach in understanding or predicting more complex psychological constructs that are linked to two or more Big Five traits.

The goal of this study was to combine different methodological approaches while overcoming the shortcomings of previous studies in order to answer the questions whether there are replicable personality types, how many of them there are, and how they relate to Big Five traits and other psychological and health-related constructs. The results revealed a robust five personality type model, which was able to significantly predict all of the psychological constructs in question longitudinally. Predictions from previous findings connecting the predicted variables to the particular Big Five dimensions underlying the personality type model were confirmed. Apparently, the person-centered approach to personality description has the most practical utility when predicting behavior or personality correlates that are connected to more than one or two of the Big Five traits such as self-esteem, locus of control and wellbeing.

This study fulfils all three criteria specified by von Eye & Bogat [ 81 ] regarding person-oriented research and considers the recommendations regarding sample size and composition by Herzberg and Roth [ 30 ]. The representative and large sample was analyzed under the assumption that it was drawn from more than one population (distinct personality types). Moreover, several external and internal cluster validity criteria were taken into account in order to validate the groupings generated by three different cluster algorithms, which were chosen to represent broad ranges of clustering techniques [ 60 , 82 ]. The Ward followed by K-means procedure covers hierarchical as well as divisive partitioning (crisp) clustering, the latent profile algorithm covers density-based clustering with probabilistic models and information theoretic validation (AIC, BIC), and spectral clustering represents graph theoretic as well as kernel-based non-linear clustering techniques. The results showed a clear superiority of the five-cluster solution. Interpreting this grouping based on theory revealed a strong concordance with personality types found in previous studies, which we could ascertain both in absolute mean values and in the Euclidean distances to mean cluster z-scores extracted from 19 previous studies. As no previous study on personality types used that many external and internal cluster validity indices and different clustering algorithms on a large data set of this size, the present study provides substantial support for the personality type theory postulating the existence of resilient, undercontroller, overcontroller, vulnerable-resilient and reserved personality types, which we will refer to with RUO-VR subsequently. Further, our findings concerning lower validity of the LPA cluster solutions compared to the k-means and spectral cluster solutions suggest that clustering techniques based on latent models are less suited for the BFI-S data of the SOEP sample than iterative and deterministic methods based on the k-means procedure or non-linear kernel or graph-based methods. Consequently, the substance of the clustering results by Specht et. al. [ 36 ], which applied latent profile analysis on the SOEP sample, may therefore be limited.

But the question, if the better validity values of the k-means and spectral clustering techniques compared to the LPA indicate a general superiority of these algorithms, a superiority in the field of personality trait clustering or only a superiority in clustering this specific personality trait assessment (BFI-S) in this specific sample (SOEP), remains subject to further studies on personality trait clustering.

When determining the longitudinal predictive validity, the objections raised by Asendorpf [ 53 ] concerning the direct comparison of person-oriented vs. variable-oriented personality descriptions were incorporated by using continuous personality type profile similarity based on Cronbach and Gleser [ 75 ] instead of dichotomous dummy variables as well as by predicting long-term instead of cross-sectionally assessed variables. Using continuous profile similarity variables also resolves the problem that potentially important information about members of the same class is lost in categorical personality descriptions [ 15 , 53 , 83 ]. Predictions regarding the association of the personality types with the assessed personality and behavior correlates, including risk propensity, impulsivity, self-esteem, locus of control, patience, cognitive and affective wellbeing as well as health measures, were confirmed.

Overcontrollers showed associations with lower spontaneity/impulsivity, with lower mental and physical health, and lower cognitive as well as affective wellbeing. Undercontrollers were mainly associated with higher risk propensity and higher impulsive behavior. These results can be explained through the connection of internalizing and externalizing behavior with the overcontroller and undercontroller types [ 5 – 7 , 78 ] and further with the connection of internalizing problems with somatic symptoms and/or symptoms of depressiveness and anxiety [ 79 ]. The dimensions or categories of internalizing and externalizing psychopathology have a long tradition in child psychopathology [ 84 , 85 ] and have been subsequently replicated in adult psychopathology [ 86 , 87 ] and are now basis of contemporary approaches to general psychopathology [ 88 ]. A central proceeding in this development is the integration of (maladaptive) personality traits into the taxonomy of general psychopathology. In the current approach, maladaptive personality traits are allocated to psychopathology spectra, such as the maladaptive trait domain negative affectivity to the spectrum of internalizing disorders. However, the findings of this study suggests that not specific personality traits are intertwined with the development or the occurrence of psychopathology but specific constellations of personality traits, in other words, personality profiles. This hypothesis is also supported by the findings of Meeus et al. [ 8 ], which investigated longitudinal transitions from one personality type to another with respect to symptoms of generalized anxiety disorder. Transitions from resilient to overcontroller personality profiles significantly predicted higher anxiety symptoms while the opposite was found for transitions from overcontroller to resilient personality profiles.

The resilient personality type had the strongest associations with external locus of control, higher patience, good health and positive wellbeing. This not only confirms the characteristics of the resilient type already described by Block & Block [ 18 ] and subsequently replicated, but also conveys the main characteristics of the construct of resilience itself. While the development of resiliency depends on the quality of attachment experiences in childhood and youth [ 89 ], resiliency in adulthood seems to be closely linked to internal locus of control, self-efficacy and self-esteem. In other words, the link between secure attachment experiences in childhood and resiliency in adulthood seems to be the development of a resilient personality trait profile. Seen the other way around, the link between traumatic attachment experiences or destructive environmental factors and low resiliency in adulthood may be, besides genetic risk factors, the development of personality disorders [ 90 ] or internalizing or externalizing psychopathology [ 91 ]. Following this thought, the p-factor [ 92 ], i.e. a general factor of psychopathology, may be an index of insufficient resilience. Although from the viewpoint of personality pathology, having a trait profile close to the resilient personality type may be an index of stable or good personality structure [ 93 ], i.e. personality functioning [ 94 ], which, though being consistently associated with general psychopathology and psychosocial functioning, should not be confused with it [ 95 ].

The reserved personality type had the strongest associations with higher patience as well as better mental health. The vulnerable-resilient personality type showed low positive correlations with spontaneity/impulsivity and low negative correlations with patience as well as health and affective wellbeing.

Analyzing the correlations of the dimensional Big Five values with the predicted variables revealed patterns similar to the mean z-scores of the personality types resilient, overcontrollers, undercontrollers and reserved. Most variables had a low to moderate correlation with just one personality profile similarity, while having at least two or three low to moderate correlations with the Big Five measures. This can be seen as support for the argument of Chapman [ 82 ] and Asendorpf [ 15 , 53 ] that personality types have more practical meaning in the prediction of more complex correlates of human behavior and personality such as mental and physical health, wellbeing, risk-taking, locus of control, self-esteem and impulsivity. Our findings further underline that the person-oritented approach may better be suited than variable-oriented personality descriptions to detect complex trait interactions [ 40 ]. E.g. the vulnerable-resilient and the overcontroller type did not differ in their high average neuroticism values, while differing in their correlations to mental and somatic health self-report measures. It seems that high neuroticism is far stronger associated to lower mental and physical health as well as wellbeing if it occurs together with low extraversion and low openness as seen in the overcontroller type. This differential association between the Big-Five traits also affects the correlation between neuroticism and self-esteem or locus of control. Not differing in their average neuroticism value, the overcontroller personality profile had moderate associations with low self-esteem and external locus of control while the vulnerable-resilient personality profile did only show very low or no association. Further remarkable is that the vulnerable-resilient profile similarity had no significant correlation with measures of cognitive wellbeing while being negatively correlated with affective wellbeing. This suggests that individuals with a Big-Five personality profile similar to the vulnerable-resilient prototype seem not to perceive impairments in their wellbeing, at least on a cognitive layer, although having high z-values in neuroticism. Another explanation for this discrepancy as well as for the lack of association of the vulnerable-resilient personality profile to low self-esteem and external locus of control though having high values in neuroticism could be found in the research on the construct of resilience. Personalities with high neuroticism values but stable self-esteem, internal locus of control and above average agreeableness and extraversion values may be the result of the interplay of multiple protective factors (e.g. close bond with primary caregiver, supportive teachers) with risk factors (e.g. parental mental illness, poverty). The development of a resilient personality profile with below average neuroticism values, on the other hand, may be facilitated if protective factors outweigh the risk factors by a higher ratio.

An interesting future research question therefore concerns to what extent personality types found in this study may be replicated using maladaptive trait assessments according to DSM-5, section III [ 96 ] or the ICD-11 personality disorder section [ 97 ] (for a comprehensive overview on that topic see e.g. [ 98 ]). As previous studies showed that both DSM-5 [ 99 ] and ICD-11 [ 100 ] maladaptive personality trait domains may be, to a large extent, conceptualized as maladaptive variants of Big Five traits, it is highly likely that also maladaptive personality trait domains align around personality prototypes and that the person-oriented approach may amend the research field of personality pathology [ 101 ].

Taken together, the findings of this study connect the variable centered approach of personality description, more precisely the Big Five traits, through the concept of personality types to constructs of developmental psychology (resiliency, internalizing and externalizing behavior and/or problems) as well as clinical psychology (mental health) and general health assessed by the SF-12. We could show that the distribution of Big Five personality profiles, at least in the large representative German sample of this study, aggregates around five prototypes, which in turn have distinct associations to other psychological constructs, most prominently resilience, internalizing and externalizing behavior, subjective health, patience and wellbeing.

Limitations

Several limitations of the present study need to be considered: One problem concerns the assessment of patience, self-esteem and impulsivity. From a methodological perspective, these are not suitable for the assessment of construct validity as they were assessed with only one item. A further weakness is the short Big Five inventory with just 15 items. Though showing acceptable reliability, 15 items are more prone to measurement errors than measures with more items and only allow a very broad assessment of the 5 trait domains, without information on individual facet expressions. A more big picture question is if the Big Five model is the best way to assess personality in the first place. A further limitation concerns the interpretation of the subjective health measures, as high neuroticism is known to bias subjective health ratings. But the fact that the vulnerable-resilient and the overcontroler type had similar average neuroticism values but different associations with the subjective health measures speaks against a solely neuroticism-based bias driven interpretation of the associations of the self-reported health measures with the found personality clusters. Another limitation is the correlation between the personality type similarities: As they are based on Euclidean distances and the cluster algorithms try to maximize the distances between the cluster centers, proximity to one personality type (that is the cluster mean) logically implies distance from the others. In the case of the vulnerable-resilient and the resilient type, the correlation of the profile similarities is positive, as they mainly differ on only one dimension (neuroticism). These high correlations between the profile similarities prevents or diminishes, due to the emerging high collinearity, the applicability of general linear models, i.e. regression to calculate the exact amount of variance explained by the profile similarities.

The latter issue could be bypassed by assessing types and dimensions with different questionnaires, i.e. as in Asendorpf [ 15 ] with the California Child Q-set to determine the personality type and the NEO-FFI for the Big Five dimensions. Another possibility is to design a new questionnaire based on the various psychological constructs that are distinctly associated with each personality type, which is probably a subject for future person-centered research.

Acknowledgments

The data used in this article were made available by the German Socio-Economic Panel (SOEP, Data for years 1984–2015) at the German Institute for Economic Research, Berlin, Germany. However, the findings and views reported in this article are those of the authors. To ensure the confidentiality of respondents’ information, the SOEP adheres to strict security standards in the provision of SOEP data. The data are reserved exclusively for research use, that is, they are provided only to the scientific community. All users, both within the EEA (and Switzerland) and outside these countries, are required to sign a data distribution contract.

Funding Statement

The author(s) received no specific funding for this work.

Data Availability

PLoS One. 2021; 16(1): e0244849.

Decision Letter 0

PONE-D-20-00337

Personality Types Revisited – a Comprehensive Algorithmic Approach to an Integration of Prototypical and Dimensional Constructs of Personality Description

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

We would appreciate receiving your revised manuscript by April 23, 2020. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Stephan Doering, M.D.

Academic Editor

Journal requirements:

1. When submitting your revision, we need you to address these additional requirements.

Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

http://www.journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and http://www.journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions .

* In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially identifying or sensitive patient information) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. Please see http://www.bmj.com/content/340/bmj.c181.long for guidelines on how to de-identify and prepare clinical data for publication. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories .

We will update your Data Availability statement on your behalf to reflect the information you provide.

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

2. Has the statistical analysis been performed appropriately and rigorously?

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: This is a sophisticated and methodologically exhaustive study.

I must confess that much of this material is beyond my scope of knowledge and my capacity to comprehend.

After reading the abstract I am still not very clear about what this study is all about. It all seems like at novel approach - or maybe a pioneering approach is a more appropriate term. Therefore, I suggest that nothing should be too obvious in the communication of this study.

1] For example, the abstract only refers to “a large representative German dataset” without providing the N? What is the abbreviation Ward/k standing for?

In general, I encourage the authors to rephrase the abstract and parts of the introduction as a service for the reader.

2] General comment: During the introduction on the first 3-4 pages, I feel somewhat lost as reader. I suppose the authors could sharpen up this part. It may also be helpful to link the rationale to some more familiar/contemporary theory and research within the field.

The 10 first pages seem to work as a review of the literature.

The aim is not presented until page 9 line 207.

3] It could be helpful with a more clear distinguishing between types and traits?

4] Page 6, line 127: “In a recent nature human behavior publication” – are the authors referring to a journal here or a particular issue or paper? It is not evident.

5] Page 7: ”total N of 1560418” – please use comma separators.

6] The authors cite the HiTOP and related scientific papers (e.g., Forbush et al, Kotov et al., Krueger et al.). However, the authors did not relate their findings or discussions to the more authoritative diagnostic frameworks such as the approved ICD-11 dimensional classifications of PDs as well as the DSM-5 alternative model – with particular emphasis on their trait systems.

7] On page 4 the authors write: “it can be said that the human goal is to be as undercontrolled as possible and as overcontrolled as necessary. When one is more undercontrolled than is adaptively effective or more overcontrolled than is adaptively required, one is not resilient”

In relation to “resilience”, it is remarkable that the authors have not related their findings or discussion to Fonagy and Sharp as well as Caspi’s P-factor (see references below). I particularly refer to the P-factor as an index of insufficient resilience, which may be something that could be more clearly incorporated into the manuscript?

Caspi, A., Houts, R. M., Belsky, D. W., & Goldman-mellor, S. J. (2015). The p factor: One general psychopathology factor in the structure of psychiatric disorders? Clinical Psychological Science, 2(2), 119–137. https://doi.org/10.1177/2167702613497473.The

Sharp, C., Wright, A. G. C., Fowler, J. C., Frueh, B. C., Allen, J. G., Oldham, J., & Clark, L. A. (2015). The structure of personality pathology: Both general (‘g’) and specific (‘s’) factors? Journal of Abnormal Psychology, 124(2), 387–398. https://doi.org/10.1037/abn0000033

Fonagy, P., Luyten, P., Allison, E., & Campbell, C. (2017). What we have changed our minds about: Part 1. Borderline personality disorder as a limitation of resilience. Borderline Personality Disorder and Emotion Dysregulation, 4(1), 11. https://doi.org/10.1186/s40479-017-0061-9

Reviewer #2: congratulations to the authors, this is an excellent work which, however, has two fundamental limitations: 1. it includes a long part, not consistent with the title and the abstract, which can be eliminated; 2 the description of the statistical methodology is poorly understood by colleagues who are not experts in data analysis. The text is weighted and complex to read.

I will point out my thoughts step by step. following them the writing becoming more agile and accessible will bring out the fantastic work behind it.

from row 48 to row 51

The difference between the two approaches should be clearly explained

from row 69 to row 70

The Q procedure should be clearly explained

from row 88 to row 91

I would delete this sentence

from row 127 to row 130

I would explain this study in more detail

rom row 131 to row 133

the reasons for this choice should be explained

from row 143 to row 206

I would eliminate this part

(it seems to me, to all intents and purposes, something that may belong to an interesting review of the literature. this part proposed in this stringed way is obviously inadequate, inconsistent with the title and unnecessarily burdens the text)

Clearly this implies the elimination also of figure 1 and of the results and discussion that refer to the comparison between figure 1 and figure 2

from row 227 to row 229

I would extend this concept

from row 233 to row 244

I would insert a table representing the elements described

from row 291 to row 321

from row 325 to row 353

The meaning of these methods should be clarified in relation to the type of data examined. this will allow a perfect understanding of the results even for non-expert colleagues in data analysiss

from row 354 to row 357

I would delete this sentence, there are recent studies that question this claim

(see Matthijs J Warrens On the Equivalence of Cohen’s Kappa and the Hubert-Arabie Adjusted Rand Index

February 2008 Journal of Classification 25 (2): 177-183)

from row 358 to row 383

These are basic concepts for which the paragraph can be reduced in size

from row 384 to row 435

Insert a figure that graphically describes the procedure

rom row 677 to row 680

delete this sentence

6. PLOS authors have the option to publish the peer review history of their article ( what does this mean? ). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy .

Reviewer #2: Yes: Raffaele Sperandeo

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files to be viewed.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/ . PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at [email protected] . Please note that Supporting Information files do not need this step.

Author response to Decision Letter 0

Please see the attached document "Response to the reviewers".

Submitted filename: Response to the Reviewers.docx

Decision Letter 1

PONE-D-20-00337R1

Personality Types Revisited – a Literature-Informed and Data-Driven Approach to an Integration of Prototypical and Dimensional Constructs of Personality Description

Dear Dr. Kerber,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process. There are only very minor points raised by reviewer 1 that need to be addressed.

Please submit your revised manuscript by December 7, 2020. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at [email protected] . When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #2: All comments have been addressed

2. Is the manuscript technically sound, and do the data support the conclusions?

Reviewer #1: Yes

Reviewer #2: Yes

3. Has the statistical analysis been performed appropriately and rigorously?

4. Have the authors made all data underlying the findings in their manuscript fully available?

Reviewer #1: No

5. Is the manuscript presented in an intelligible fashion and written in standard English?

6. Review Comments to the Author

Reviewer #1: I feel the authors overall adressed the issues I raised.

I only have the following minor comments:

1) The tables have no definitions in the legend for the different terms and abbreivations - I am not entirely aware of the author guidelines for this journal, but I think it is much needed.

2) The authors rigthly included a reference to the now approved ICD-11 PD classification (line 825). However, the authors should provide the correct reference:

WHO. (2019). ICD-11 Clinical Descriptions and Diagnostic Guidelines for Mental and Behavioural Disorders. World Health Organisation. gcp.network/en/private/icd-11-guidelines/disorders

3) Moreover, they only refer to studies on big five convergence with DSM-5 Section III traits - but not with the ICD-11 traits. See for example the following papers:

Somma, A., Gialdi, G., & Fossati, A. (2020). Reliability and construct validity of the Personality Inventory for ICD-11 (PiCD) in Italian adult participants. Psychological Assessment, 32(1), 29–39. https://doi.org/10.1037/pas0000766

Oltmanns, J. R., & Widiger, T. A. (2018). A self-report measure for the ICD-11 dimensional trait model proposal: The Personality Inventory for ICD-11. Psychological Assessment, 30(2), 154–169. https://doi.org/10.1037/pas0000459

Oltmanns, J. R., & Widiger, T. A. (2019). The Five-Factor Personality Inventory for ICD-11: A facet-level assessment of the ICD-11 trait model. Psychological Assessment. https://doi.org/10.1037/pas0000763

Reviewer #2: I read this study and reviewed it with great pleasure. I congratulate you on this innovative work which appears to be a milestone in the study of personality

7. PLOS authors have the option to publish the peer review history of their article ( what does this mean? ). If published, this will include your full peer review and any attached files.

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/ . PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at [email protected] . Please note that Supporting Information files do not need this step.

Author response to Decision Letter 1

Response: Thanks to this suggestion we have reviewed all our tables for abbreviations that are not explained and included them in the respective notes.

Response: We have rephrased ll. 838-841 to also include a reference to the ICD-11 PD model.

Decision Letter 2

PONE-D-20-00337R2

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/ , click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at [email protected] .

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact [email protected] .

Acceptance letter

Personality Types Revisited – a Literature-Informed and Data-Driven Approach to an Integration of Prototypical and Dimensional Constructs of Personality Description

Dear Dr. Kerber:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact [email protected] .

If we can help with anything else, please email us at [email protected] .

Thank you for submitting your work to PLOS ONE and supporting open access.

PLOS ONE Editorial Office Staff

Professor Stephan Doering

Similar articles being viewed by others

Slider with three articles shown per slide. Use the Previous and Next buttons to navigate the slides or the slide controller buttons at the end to navigate through each slide.

literature review of personality type

Does personality still matter in e-commerce? How perceived hubris influences the assessment of founders’ trustworthiness using the example of reward-based crowdfunding

14 September 2022

Janina Sundermeier & Tyge-F. Kummer

The narcissism of crowdfunding entrepreneurs

22 February 2019

Helen Bollaert, Gaël Leboeuf & Armin Schwienbacher

literature review of personality type

Hardwired to attack. Candidates’ personality traits and negative campaigning in three European countries

01 January 2022

Alessandro Nai, Anke Tresch & Jürgen Maier

literature review of personality type

The Dark Side of Machiavellian Rhetoric: Signaling in Reward-Based Crowdfunding Performance

02 November 2021

Goran Calic, Rene Arseneault & Maryam Ghasemaghaei

It’s Who’s on the Inside that Counts: Campaign Practitioner Personality and Campaign Electoral Integrity

20 February 2019

Hans J. G. Hassell

literature review of personality type

We vote for the person, not the policies: a systematic review on how personality traits influence voting behaviour

10 January 2023

Filipe Falcão, Bárbara Sousa, … Patrício Costa

literature review of personality type

Success of Crowdfunding: A Trustworthiness Perspective

11 January 2022

Steven S. Lui, Zhijing Zhu & Jinjing Liu

literature review of personality type

Personality trait differences across types of entrepreneurs: a systematic literature review

Florentine U. Salmony & Dominik K. Kanbach

literature review of personality type

Disagreeable narcissists, extroverted psychopaths, and elections: a new dataset to measure the personality of candidates worldwide

07 September 2018

Million dollar personality: a systematic literature review on personality in crowdfunding

Management Review Quarterly volume 72 , pages 309–345 ( 2022 ) Cite this article

Metrics details

Expressed personality traits can play a pivotal role in convincing investors in crowdfunding. Our study answers the research question: What is the current body of knowledge regarding the relationship between personality factors and crowdfunding success and where are knowledge gaps where the literature is silent? In our literature review, we therefore analyze and categorize (1) the results provided by quantitative studies on the relationship between the personality of entrepreneurs and crowdfunding success and (2) the research gaps identified by the authors investigating personality in crowdfunding. We find that studies investigating the entrepreneur's personality, i.e. the Big Five, other baseline personality traits (self-efficacy, innovativeness, locus of control, and need for achievement) and the Dark Triad, find positive relationships between openness and crowdfunding success, while narcissism shows an inverted u-shaped relationship with crowdfunding success across articles. However, the effects of other personality traits on crowdfunding success are largely inconclusive. Further, we identify four main gaps in the literature. First, future studies should examine non-linear relationships between expressed personality traits and crowdfunding success. Second, there is a need for more studies that employ different methods like qualitative or mixed-method approaches. Third, replication studies in similar and different contexts are urgently needed. Fourth, a plurality of personality perspectives would strengthen future research (e.g., investor perspective, third party perspective). To our knowledge this is the first literature review of personality traits in crowdfunding. Our work aims to enrich our understanding of individual-level components in the underexplored alternative finance market.

Working on a manuscript?

Avoid the most common mistakes and prepare your manuscript for journal editors.

1 Introduction

Young firms face the challenge of acquiring early stage venture capital (Drover et al. 2017 ) which more than doubles their chances of survival (Puri and Zarutskie 2012 ). To finance their venture, entrepreneurs increasingly face a number of options outside of traditional venture capital funding or business angel investments. One example of such alternative financing methods is crowdfunding, which opens new pathways for young firms to raise capital in a less regulated way than via classical funding instruments (Cumming et al. 2019b ). Crowdfunding presents a financing method in which firms acquire capital from a crowd of individuals via an open call (Belleflamme et al. 2010 ). Entrepreneurs turn to crowdfunding when they need financial assistance to realize a project. Via crowdfunding, entrepreneurs can also acquire customers and validate their business models or ideas at an early stage while simultaneously retaining a high degree of independence from individual investors. Types of crowdfunding include borrowing money online for investments (lending-based), offering products or rewards for pre-sale (reward-based), collecting donations to realize charitable projects (donation-based), or selling equity shares of a company to a crowd of investors (equity-based). The types of crowdfunding significantly differ from each other. For example, equity crowdfunding gears towards long-term investments, whereas other types of crowdfunding typically involve pre-selling, short-term loans, or donations regarding future projects. Similarly, entrepreneurs seeking equity crowdfunding are in a somewhat similar stage to those that receive classical venture capital or angel financing, as these settings both involve a (long-term) stake in the venture. This similarity does not hold for most other crowdfunding forms.

A growing stream of literature investigates factors that lead to successful crowdfunding (Wiklund et al. 2011 ). Authors find that several “hard facts” such as the target investment amount, the number of investors/backers to date, provided roadmaps, Facebook shares, or the location of a company impact the outcome of a crowdfunding campaign (Ahlers et al. 2015 ; Bertrand and Schoar 2003 ; Bi et al. 2017 ; Block et al. 2018 ; Chan and Parhankangas 2017 ; Courtney et al. 2017 ; Davis et al. 2017 ; Janku and Kucerova 2018 ; Prodromos et al. 2014 ). “Softer factors” that include media richness (e.g., use of photos and videos), third-party endorsement, and campaign updates can also drive the funding process (Courtney et al. 2017 ; Wang et al. 2020 ). In addition, individual-level factors are critical for crowdfunding success. For example, entrepreneurs’ education and professional background, previous funding experience, and gender or ethnic background can influence the crowd’s contributions to a given campaign (Allison et al. 2017 ; Barbi and Mattioli 2019 ; Courtney et al. 2017 ; Moleskis et al. 2019 ; Younkin and Kuppuswamy 2018 ).

Within this stream, a unique discourse relates to the entrepreneur's personality. Personalities describe the unique combinations of traits that form people's individual character. In line with the entrepreneurship field in general, research in crowdfunding has also begun to study the impact of personality on funding success. Two studies examine the influence of the Big Five personality traits on reward-based crowdfunding success on the website Kickstarter (Gera and Kaur 2018 ; Thies et al. 2016 ). Further, Bollaert et al. ( 2019 ) research indicates a negative impact of narcissistic personality traits on funding success, while other authors find inconclusive relations of narcissistic rhetoric to crowdfunding success depending on the compliance with other characteristics of the entrepreneur (Anglin et al. 2018b ). Regarding hubris and charisma, researchers have found that entrepreneurs perceived as scoring high on these traits are more successful in raising funds (Sundermeier and Kummer 2019 ). Moritz et al. ( 2015 ) argue that perceived sympathy, openness, and trustworthiness are essential in reducing information asymmetries (e.g., where one party knows more than the other and could exploit this information supremacy) between entrepreneurs and investors in the crowdfunding context.

As an alternative method of financial resource acquisition, crowdfunding is of special interest for entrepreneurship research (Landström and Harirchi 2019 ), especially when combined with the “most promising topical areas in entrepreneurship research” (Kuckertz and Prochotta 2018 , p. 3), e.g., entrepreneurial behavior and psychology. Although promising, crowdfunding does not come without challenges for entrepreneurs seeking capital and particularly for investors when trying to discern entrepreneurs’ chances of success. On the one hand, investors face increased information asymmetries than they would in other funding types (Cumming et al. 2019a ). These arise from reduced disclosure requirements for fund-seeking entrepreneurs (Cumming et al. 2019a ), the use of new media tools, and the lack of opportunity to directly question campaign initiators. Such circumstances increase the need for cognitive shortcuts to make investment decisions. These are based (among others) on impressions of entrepreneurs’ personality and used, for example, to access the entrepreneur's capability to lead a successful venture. For entrepreneurs, on the other hand, funds are not acquired via direct interaction but through means of computer-mediated communication (Pollack et al. 2021 ). Investments are mediated by online fundraising platforms where personality is displayed and perceived in a very different way than in traditional and interpersonal settings (e.g., with an angel investor or loan agent). For entrepreneurs in the context of crowdfunding, knowing which personality displays convince the crowd to invest in their campaigns is of particular practical relevance, as it can shape investor perception and therefore campaign success. In crowdfunding, the personality impression perceived by investors is literally worth up to a million dollars (JOBS Act; (Ahlers et al. 2015 )), inspiring the title of this paper.

Although a growing body of literature summarizes and evaluates crowdfunding success factors, personality plays no role in these reviews. To our knowledge, no previously published literature review focuses on personality factors in crowdfunding, although the implications both for practice (as explained above) and for the scientific community are essential. Combining the representative findings on crowdfunding and personality from disparate studies into one literature review would focus future research on relevant gaps and broaden the impact of this field. Additionally, identifying areas where the results from crowdfunding are generalizable to other forms of entrepreneurial financing would create the opportunity to transfer implications from crowdfunding, with its easy accessibility and high sample sizes, to other areas where research is scarce due to difficulties to access data (e.g., business angel financing) (Cumming et al. 2019a , b ). We address this gap by examining the following research question: What is the current body of knowledge regarding the relationship between personality factors and crowdfunding success, and where are knowledge gaps where the literature is silent?

Our study finds a trend towards more research on entrepreneurial personality in crowdfunding and a tendency to employ software-based narrative methods and questionnaires. We identified four main gaps that should be addressed by future research studies. First, future quantitative studies should examine nonlinear (e.g., quadratic) relations between expressed personality traits and crowdfunding outcomes. Second, future studies should employ different methods e.g., mixed-methods approaches in order to validate existing narrative methods, such as by combining them with questionnaires. Third, authors should conduct replications in highly similar settings to strengthen results as well as in different contexts, e.g., crowdfunding types, to explore different effects of personality. Fourth, studies are required that investigate not only the personality of entrepreneurs but change/flip the perspective and also investigate the personality of investors and how they interact during the crowdfunding process.

In the following, we first describe the conceptual background of personality constructs and the chosen methodology, as well as our analysis of the selected literature. Finally, we highlight commonalities, differences and gaps, in addition to implications and suggestions for future research.

2 Conceptual Background on Personality in the Entrepreneurial Context

The personality of an individual is the basis that effects a person’s decisions and behavior in everyday life situations as well as in the economic aspects of life (McAdams and Pals 2006 ; Rauch and Frese 2014 ). The broad concept of personality includes a range of aspects from abilities such as different forms of intelligence, motives, attitudes up to a person’s characteristics and temper (Brandstätter 2011 ). Taken together, personality can be seen as the foundation for individual differences between humans (Mairnesse et al. 2007 ). Studies suggest that personality is an underlying system that develops until the age of 30 and then stays stable over adolescent life (Costa and McCrae 1988 ). In the entrepreneurship literature, authors investigate a wide variety of personality aspects.

The personality theory most frequently investigated in entrepreneurship is the Big Five Personality Theory from psychology (e.g., Brandstätter 2011 ; Kerr et al. 2017 ; Mueller and Thomas 2001 ; Rauch and Frese 2014 ). Research in entrepreneurial finance finds effects of the Big Five on business angel syndication, investment management, and loss aversion in the financial domain (Block et al. 2019 ; Boyce et al. 2016 ; Mayfield et al. 2008 ). The concept focuses on five key traits: First, openness, when strongly expressed, is a driver for the need for variety and intellectual curiosity (Costa Jr and McCrae 1995 ). People that rate high on openness seek new experiences. In a business-related context, people with high openness ratings are socially skilled. Scientists suggest that they are good salespeople and have managerial skills (Almlund et al. 2011 ). People who rate low on openness are risk-averse (Almlund et al. 2011 ). Researchers associate openness with intelligence and creativity, but also with negative aspects such as sensation-seeking and a tendency to question authority (Costa Jr and McCrae 1995 ). Second, conscientiousness relates to striving for achievement, hard work, dutifulness, and self-discipline (Almlund et al. 2011 ; Bozionelos 2004 ). In the business context, conscientiousness is a predictor for career success, job performance, and wages (Almlund et al. 2011 ; Hogan and Ones 1997 ; Judge et al. 1999 ). Third, extraversion is associated with sociability, optimism, ambition, positive emotionality, cheerfulness, dominance, and excitement seeking (Barrick et al. 2001 ; Bozionelos 2004 ; Watson and Clark 1997 ). High scores in extraversion predict effective job performance, the likelihood to reach a leadership role, and wages (Almlund et al. 2011 ; Barrick and Mount 1991 ; Bozionelos 2004 ; Ciavarella et al. 2004 ; Judge et al. 1999 ). Fourth, agreeableness is a trait often summarized as warmness. People with high scores on agreeableness tend to be altruistic, friendly, flexible, courteous, forgiving, modest, and trustworthy (Almlund et al. 2011 ; Barrick et al. 2001 ; Bozionelos 2004 ). Studies demonstrate a negative relationship between agreeableness and career success or work involvement (Bozionelos 2004 ). Fifth, neuroticism (also referred to as emotional instability) is related to the experience of negative emotions, insecurity, low goal-orientation, and low self-esteem (Almlund et al. 2011 ; Bozionelos 2004 ). Research also finds negative associations between neuroticism and job search efforts, work performance, performance motivation, and extrinsic success (Almlund et al. 2011 ; Judge and Ilies 2002 ).

Other baseline key personality traits frequently studied in entrepreneurship (aside from the Big Five) are self-efficacy, innovativeness, locus of control, and need for achievement (Kerr et al. 2017 ; Rauch and Frese 2014 ), explained hereafter. First, self-efficacy as part of the personality is of particular interest regarding entrepreneurs as it describes a person's inclination to see themselves as capable of performing actions and aligning themselves with self-set goals (Chen et al. 1998 ; Rauch and Frese 2014 ). Overcoming failure can also be counted as self-efficacy (Harburg et al. 2015 ). Second, innovativeness is strongly linked to a person's ability to engage in new things. Innovative people are those in a society who adapt to change faster than the average (Manning et al. 1995 ; Rogers and Shoemaker 1971 ). Since innovativeness is a prerequisite for innovation, it is a crucial personality component in entrepreneurship. Third, locus of control is closely linked to a person's belief in their ability to determine their destiny (Hoffman et al. 2003 ). Researchers differentiate between external and internal locus of control. An external locus of control refers to when people perceive their future to be shaped by their environment and not by their own actions. In general, founders tend to have an internal locus of control, which refers to situations where people are convinced that they can shape their future by their actions and decisions (Rotter 1966 ). Fourth, the need for achievement is a personality factor that goes back to David McClelland's Motivation Theory (Johnson and McClelland 1984 ). A high need for achievement describes people who are not satisfied with routine tasks but strive for challenges and continuous improvement (Rauch and Freser 2014 ). They take responsibility for the results they achieve and demand feedback for their actions. Many studies highlight the relevance of this trait for founders (Rauch and Frese 2007 ), as it can influence venture size and growth (Lee and Tsang 2001 ).

A personality aspect of increasing interest to researchers is narcissism (Bollaert et al. 2019 ; Butticé and Rovelli 2020 ). Narcissistic individuals are generally perceived as arrogant and self-centered. They usually have an elevated image of their achievements and react with offense or even aggression when questioned (Miller et al. 2010 ). On the other hand, narcissism can also have positive effects, e.g., on self-confidence and self-respect, if not overly expressed (Paulhus and Williams 2002 ). Therefore, these characteristics are clearly relevant for entrepreneurs. Narcissism is one of three characteristics summarized as the “Dark Triad” (Paulhus and Williams 2002 ) which refers to the three socially aversive traits narcissism, Machiavellianism, and psychopathy. These traits reflect self-promotion, emotional coldness, and aggressive behavior in a person's character (Paulhus and Williams 2002 ). Focusing on manager characteristics, the dark triad and, in particular, narcissism diminish the positive effect of entrepreneurial orientation and thereby negatively influence firm performance (Bouncken et al. 2020 ; Engelen et al. 2016 ). Narcissism and psychopathy are officially classified as psychological disorders in the U.S. and Europe (e.g., in DSM 4 and 5) (Furnham et al. 2013 ). However, the entrepreneurial literature uses them to describe personality aspects that tend towards the clinical definition but do not necessarily fit this pathological description of narcissism.

In the following section, we focus on those traits most frequently addressed in entrepreneurship and introduced above (Kerr et al. 2017 ; Mueller and Thomas 2001 ; Rauch and Frese 2014 ). These are the Big Five personality model, the additional baseline traits innovativeness, self-efficacy, locus of control, need for achievement, and the Dark Triad.

3 Methodology

3.1 data collection.

To answer our research question, we followed the guidelines set forth by Fisch and Block ( 2018 ). Therefore, we began by screening the existing literature. We collected the articles for this review in May of 2021, allowing us to take a snapshot of the literature on personality in crowdfunding. To obtain a comprehensive overview of literature on the topic, we did not limit our search to specific journals (Webster and Watson 2002 ). Instead, we rely on the leading databases of the field, such as EBSCO Host, Scopus, and Web of Science. Our literature search involved four steps:

First, we searched the databases. For each of these we used the closest corresponding filter criteria available (abstract search in EBSCO Host, abstract and title search in Scopus, and topic search in Web of Science). For the search we combined the term “crowdfunding”, “P2P lending”, or “peer-to-peer lending” and one of the following terms on personality: “personality”, “big five”, “openness”, “conscientiousness”, “extraversion”, “agreeableness”, “neuroticism”, “dark triad”, “narcissism”, “self-efficacy”, “innovativeness”, “locus of control”, and “need for achievement”. Table 1 provides further information on the search strings employed and the respectively resulting number of articles. The initial search generated 20 unique EBSCO host articles, 65 unique Scopus articles, and 45 unique Web of Science articles resulting in 81 unique articles over all three platforms (removing duplicates).

In a second step, we screened all retrieved articles and included them in our review based on subject matter fit. We therefore excluded all articles with no clear focus on crowdfunding or on personality. We also exclude those studies that solely mention personality, but do not actually include one or more personality constructs or crowdfunding in their research. In case of personality this exclusion criterion is complicated to assess because researchers often use the term personality to describe personal characteristics (e.g., optimism) rather than concrete personality constructs (e.g., agreeableness). To differentiate the papers that actually explore personality constructs in the context of crowdfunding from those that do not, we asked ourselves the following three questions while examining the papers:

(1) Do the search terms appear within the title, abstract, or keywords of the paper, or is it a mismatched result (i.e., where the terms do not really appear as expected)? For example, we excluded Borst et al. ( 2018 ) as none of our personality-related terms were mentioned within the title, abstract, or keywords (“From friendfunding to crowdfunding: Relevance of relationships, social media, and platform activities to crowdfunding performance”).

(2) Is personality/crowdfunding a core concept of the paper or just used as an example to research a related topic? For example, we excluded Gruda et al. ( 2021 ) as crowdfunding is just a concept to which the paper's results are compared (i.e. “We discuss and compare our findings to previous work on narcissism and crowdfunding.” (Gruda et al. 2021 , p. 1)); another example is the exclusion of Wang et al. ( 2017 ) who investigate sentiments rather than personality (“The study proves that positive sentiment in the blurb and detailed description promotes the successful campaigns” (Wang et al., 2017 , p. 2)).

(3) Is the construct related to a person/group? For example, we excluded Ceballos et al. ( 2017 ) as product innovativeness is not a characteristic of the entrepreneur (“the innovativeness of a project, […] can positively affect crowdfunding achievement.” (Ceballos et al. 2017 , p. 79)).

For the 81 articles, two researchers assessed the relevance of each article by screening the title, abstract, and keywords and by employing the three questions as additional fit criteria to decide on the relevance for the literature review. If the title, abstract, and keywords were insufficient to assess whether or not the article should be included in the review, the whole paper was read to reach a clear conclusion (8 articles, e.g., Shin and Lee 2020 ). This rating method was conducted by two authors independently. In cases of disagreement (12 articles, e.g., Tseng 2020 ), the articles were discussed until a consensus was reached. This procedure led to the inclusion of 25 (out of 81) articles.

In the third step, we performed subsequent forward and backward searches, using both the reference lists of the articles and Google Scholar. We used the aforementioned criteria to assess the relevance of the retrieved articles, yielding three additional articles for our data set, for a total of 28.

As the last step, we also examined other literature reviews on crowdfunding. In these, however, the focus was mostly on general success factors (Alegre and Moleskis 2019 ; Bouncken et al. 2015 ; Butticé et al. 2018 ; Cai et al. 2021 ; Dalla Chiesa and Handke 2020 ; Iurchenko 2019 ; Jovanović 2019 ; Kaartemo 2017 ; Mochkabadi and Volkmann 2020 ; Moleskis and Alegre 2018 ; Moritz and Block 2016 ; Salido-Andres et al. 2020 ; Shneor and Vik 2020 ; Zhao and Ryu 2020 ). Overall, personality was only mentioned as a success factor in one of the reviews (Butticé et al. 2018 ), which further illustrates the necessity of our work.

For our review, we only included articles written in English and published in peer-reviewed academic journals, research compilations or conference proceedings. The only exception to this was a dissertation on the Dark Triad by an expert in the field (Creek 2018 ). Overall, our literature screening resulted in a collection of articles that very clearly examine crowdfunding and personality with a particular emphasis on the personality aspects we included in our search terms. The steps of the literature search and selection are summarized in Fig. 1 below.

Systematic Data Collection Process

3.2 Data analysis

After carefully screening the articles, we decided on a topic-centered analysis. Therefore, we first collected classical descriptive data on the articles in our dataset (e.g., publication date, outlet, research method). We also identified and recorded topic-specific descriptive data; for example, we determined the crowdfunding type described in the articles (reward-based, equity-based, lending-based, or donation-based), categorized the theoretical approach (e.g., signaling theory, social identity theory), the methodologies utilized (e.g., questionnaire, narrative analysis, etc.), and the variables employed (e.g., Big Five personality, innovativeness) in more detail. We also identified the authors' perspectives on their investigation and categorized these as campaign owner-centered, investor-centered, or as a hybrid approach (Table 2 ). After the articles were categorized by one author using the citation management software Citavi, they were reviewed by another researcher without significant discrepancies after discussion.

For the content analysis, we followed the direction of our research question and best practices (e.g., Colombo 2020 ; Jones et al. 2011 ; Mochkabadi and Volkmann 2020 ). We analyzed (1) the contents of the qualitative articles, (2) the results of the quantitative papers focusing on crowdfunding outcomes, and (3) the limitations and future research opportunities suggested by the authors of the reviewed papers.

(1) We summarized the results of the three articles in our sample that utilize a quantitative approach and provide an overview of these within Table 3 .

(2) We examined the subset of twelve quantitative papers focusing on crowdfunding success from our literature selection in more detail. First, for each quantitative study reviewed, we extracted the personality variables examined by the authors. We then supplement these variables with the personality constructs identified within the conceptual background and use them as the basis for our subsequent analysis in Table 4 . We examined the findings of the quantitative analysis conducted in detail and extracted all significant and non-significant findings regarding personality variables. Next to these variables the findings were assigned to the crowdfunding type and success variable (e.g., funding success, amount raised, total backers) researched by the authors of the representing article (Table 4 ). As some authors examine multiple personality variables or different crowdfunding types simultaneously, one article can account for more than one effect displayed in Table 4 . As before, one researcher conducted the assignment of the quantitative findings, followed by a review by another researcher and subsequent discussions to eliminate differing assessments.

(3) Next, we closely examined all studies' limitations and the suggested future research identified by the authors of all 28 articles. Hereby, we employed three steps, following a similar approach to that of Jones et al. ( 2011 ) for identifying and subsequently coding topic themes. First, we extracted the mentioned limitations and future research sections for each paper. Second, we summarized these sections to reflect their key points (Table 5 ). One author conducted this step, followed by the mentioned review and discussion process with another researcher. Third, as future research opportunities are of particular interest to the scientific community, we then continued to cluster the mentioned research opportunities into categories. Therefore, two authors independently categorized the future research opportunities mentioned by the respective authors of the reviewed papers, clustering them by similarity (e.g., “We thus advise scholars to extend our work to alternative types of crowdfunding campaigns and platforms.” (Butticé and Rovelli 2020 , p. 5) and “future research can be extended to other forms of crowdfunding, such as peer-to-peer lending” (Leonelli et al. 2020 , p. 55)). Next we compared the clusters and resolved the remaining differences by reaching consensus between the authors (e.g., splitting the topic “perspectives” into the topics “perspective” and “context”). We next discussed and subsequently assigned topic and subtopic names to the five resulting clusters and twelve subclusters. In many cases, articles reviewed pointed out multiple future research opportunities (e.g., the use of alternate methods and variables, larger samples, etc.). Therefore, we counted some articles into more than one topic cluster (e.g., Butticé and Rovelli ( 2020 ) state: “We thus advise scholars to extend our work to alternative types of crowdfunding campaigns and platforms” categorized in our topic “Context” and subtopic “Crowdfunding Type”, but the authors also advise: “replicate our study on a subsample of entrepreneurs administering them the Narcissistic Personality Inventory” categorized in our topic “Methods” and subtopic “Approach” (Butticé and Rovelli 2020 , p. 5)). Figure 5 provides an overview of how many of the reviewed articles mentioned one or more of the five future research topics.

4.1 Descriptive results

Our analysis spans 28 articles. These were published between 2015 and March 2021 with a low point of no published papers in 2017 and an increasing trend in more recent years (Fig. 2 ).

Number of Publications per Year

Our search returned papers focusing on the following personality constructs in line with our search terms: the Big Five in general (Bernardino and Santos 2016 ; Davidson and Poor 2015 ; Gera and Kaur 2018 ; Kim and Hall 2021 ; Kim et al. 2021 ; Rottler et al. 2020 ; Ryu and Kim 2016 ; Thies et al. 2016 ), only openness (Moritz et al. 2015 ), only conscientiousness (Moss et al. 2015 ; Short and Anglin 2019 ), only extraversion (Netzer et al. 2019 ) the Dark Triad (Creek 2018 ; Leonelli et al. 2020 ), only narcissism (Anglin et al. 2018b ; Bollaert et al. 2019 ; Butticé and Rovelli 2020 ), self-efficacy (Harburg et al. 2015 ; Macht and Chapman 2019 ; Shneor and Munim 2019 ; Stevenson et al. 2019 ; Troise and Tani 2020 ), innovativeness (Calic and Shevchenko 2020 ; Moss et al. 2015 ; Rodriguez-Ricardo et al. 2018 ; Shin and Lee 2020 ; Short and Anglin 2019 ; Tseng 2020 ), and locus of control (Rodriguez-Ricardo et al. 2018 ). Also, the broad search for the term “personality” in general also revealed additional traits investigated by researchers in the context of crowdfunding: risk-taking (Calic and Shevchenko 2020 ; Moss et al. 2015 ; Short and Anglin 2019 ), autonomy (Moss et al. 2015 ; Short and Anglin 2019 ), as well as charisma and hubris (Sundermeier and Kummer 2019 ). The crowdfunding literature does not yet reflect the term “need for achievement” as a personality construct.

In 17 of these articles, authors primarily investigate personality aspects in reward-based crowdfunding rather than in other crowdfunding types (Table 2 ). This trend might be due to the easy accessibility of Kickstarter data via openly available tracking platforms such as Kickspy. It is also noteworthy that both reward- and lending-based crowdfunding permit the authors to use larger samples of campaign data (on average) compared to donation-based and particularly equity-based forms of crowdfunding (Fig. 3 ).

Average Examined Campaigns per Crowdfunding Type

The methods used within the selected papers are based on questionnaires, narrative analysis, experiments, and interviews (Fig. 4 a). Most of the articles are based on methods that focus on questionnaires or the text of a given campaign. The software tools most frequently employed for narrative analysis conducted in 11 articles are Linguistic Analysis and Word Count (LIWC) and CAT Scanner. Further, two authors used the artificial intelligence based tool IBM Personality Insights (Fig. 4 b).

Approach: Percentage Distribution and Investigation Method

Of the 28 articles, only three base their research on qualitative approaches. These conducted semi-structured interviews in two cases (Harburg et al. 2015 ; Moritz et al. 2015 ) and in the third case coded comments on crowdfunding pages regarding e.g., moral support provided by the investors (Macht and Chapman 2019 ). The remaining articles follow a quantitative approach largely based on regression models (Table 2 ).

The authors of the articles selected for our review employ a number of theories. Three articles base their research on Signaling Theory (Spence 1978 ). Social Role Theory (Eagly and Wood 2012 ) and Self-Determination Theory (Deci and Ryan 2008 ) were also used by more than one author team. Additional theories utilized in the articles can be derived from Table 2 .

Regarding the perspective taken in the articles, across all 28 studies, 18 focus on the entrepreneur’s or campaign creator’s view. Nine articles take the investor perspective. Strikingly, only one author team took a more comprehensive approach (Moritz et al. 2015 ) by investigating all parties involved: the entrepreneurs, investors, and any third parties involved, e.g., platform representatives (Table 2 ).

4.2 Results of the thematic analysis

For a more in-depth thematic analysis, we set three priorities. First, we summarized the results of the three qualitative studies. Second, we categorized previous quantitative studies in a way that can be easily utilized by future authors. Third, we summarize and categorize what other authors consider to be the essential future research steps in personality research on crowdfunding.

4.2.1 Summary of the qualitative articles reviewed

Three out of the 28 research papers within this literature review are qualitative in nature (Table 3 ). First, the qualitative-empirical study of Moritz et al. ( 2015 ) inductively investigates the role of investor communication as a medium for overcoming information asymmetries. Therefore, the authors conducted 23 interviews with investors, representatives of new ventures, and third party stakeholders such as platform operators. The study finds that within the crowdfunding process, personal communication is replaced by pseudo-personal communication via the Internet and that communicating soft personality factors, e.g., openness is vital to reduce perceived information asymmetry , i.e., when one party has more (private) information than the other. In so doing, the authors took the perspective of different participants in the crowdfunding process and thereby provided the only paper that simultaneously investigates multiple perspectives and goes on to build theory from cases.

Second, Harburg et al. ( 2015 ) investigate the influence of crowdfunding ecosystems on the entrepreneurs' self-efficacy. The authors thereby conducted 53 semi-structured interviews and rely on Bandura’s social cognitive theory (Bandura et al. 1999 )– which maintains that people’s knowledge acquisition is based on observing others in social context and the media. Therefore, the study is clearly deductive in nature. The authors report that entrepreneurs gain self-efficacy via the received feedback and number of backers supporting them, metrics showing their progress on the funding page, and examples of succeeding entrepreneurs. Nevertheless, the entrepreneurs' self-efficacy can also decrease when facing a lack of public validation or their project fails in front of the crowd (e.g., experiencing shame).

Third, Macht and Chapman ( 2019 ) also examine self-efficacy supplemented by other psychological capital aspects like optimism and resilience in the context of crowdfunding. Their qualitative interpretative work investigates the associations between the crowds' comments within a given campaign and fund seekers' human, social, and psychological capital. By coding and thematically analyzing 475 comments from ten crowdfunding campaigns (examining only those with a minimum of 30 comments in a selection process that can at best be described as semi-random), the authors core finding is that the crowd can increase the entrepreneurs' self-efficacy, hope, optimism, and resilience by providing support and by showing support and criticism within their comments. The generalizability of this finding is limited, given the moderate sample size. Also, the methodology used is not clearly specified and it is unclear if this work is inductive or rather a more deductive approach that begins with psychological capital and goes on to “test” this qualitatively.

With the exception of the study of Moritz et al. ( 2015 ), the qualitative studies focus not on the personality displayed within the crowdfunding process but on gaining self-efficacy via the crowdfunding process itself. While the degree to which an individual’s personality can change through a single crowdfunding campaign may be questionable, these studies focus on an angle of personality in crowdfunding that has clearly been neglected by the other studies within this literature review. Thereby, such qualitative studies can help explore future research avenues not yet represented in the body of literature.

4.2.2 Categorization of results of the quantitative articles reviewed

Only twelve articles quantitatively analyze the effects of personality on campaign outcomes. We focus on the independent personality variables reflected by the papers retrieved in our literature search. The outcome of a campaign is measured either by a dummy variable for success (goal reached yes/no), the actual amount raised (a continuous variable), the number of contributors to a campaign (as a count variable), or a combination of these three.

Three articles study the Big Five traits (Gera and Kaur 2018 ; Rottler et al. 2020 ; Thies et al. 2016 ) and two additional studies examine the single Big Five trait conscientiousness (Moss et al. 2015 ; Short and Anglin 2019 ). The authors find strong evidence for a positive impact of openness on crowdfunding success and suggest a positive influence of agreeableness and extraversion and a negative influence of neuroticism (Gera and Kaur 2018 ; Rottler et al. 2020 ; Thies et al. 2016 ). It is noteworthy that for most Big Five factors, the authors do not report similar findings, but find both significant and non-significant effects. Only openness and its positive influence on campaign success in reward-based crowdfunding seems to be a robust relationship across the quantitative studies reviewed (Table 4 ).

Focusing on the Dark Triad, we see that while existing results for other crowdfunding types often contradict each other, in some cases there are clear tendencies, such as for the negative but inverse u-shaped effect of narcissism on crowdfunding success (even across different measures of success). Although the articles report no significant results for Machiavellianism, they report some evidence for the effects of psychopathy. For example, Creek ( 2018 ) finds a positive relationship between the amount raised and psychopathy in equity-based crowdfunding, contrary to the opposite finding of Leonelli et al. ( 2020 ) regarding campaign success.

Finally, we report our findings on the study of the additional (frequently used) personality traits within the identified crowdfunding literature. First, Shneor and Munim ( 2019 ) find an indirect effect of self-efficacy in reward-based crowdfunding, in particular a significant influence on their mediator variable “financial contribution intention”. Second, Short and Anglin ( 2019 ) find a significant negative effect of innovativeness on the amount raised, and Calic and Shevchenko ( 2020 ) find positive but also significant inverted u-shaped relations for innovativeness in all three crowdfunding performance measurements (success, amount raised, and number of backers). Both studies were conducted in a reward-based crowdfunding setting. Third, some authors find that risk-taking entrepreneurs succeed more often in lending-based crowdfunding campaigns (Moss et al. 2015 ), while Calic and Shevchenko ( 2020 ) report inverted u-shaped relationships between risk-taking and campaign success in reward-based crowdfunding. Further, it is noteworthy that, while risk-takers are more likely to receive crowdfunded loans, they are less likely to succeed with other types of crowdfunding. Fourth, autonomy negatively affects the amount raised in reward-based crowdfunding (Short and Anglin 2019 ) and shows an inverted u-shaped relation across all performance measurements (Calic and Shevchenko 2020 ). In lending-based crowdfunding, however, Moss et al. ( 2015 ) report a positive effect of autonomy.

4.2.3 Analysis of the Future Research Sections

The analysis of the critical gaps for future research in personality and crowdfunding is based on all 28 articles included in the literature review. Table 5 provides detailed insights into what the representative authors identified as limitations in their articles and how they would like to see future research evolve to address these concerns. We summarize, categorize and quantify the individual statements in Fig. 5 .

Future Research Suggestions from the Articles Reviewed categorized in Topics and Subtopics. *Number of articles in a subtopic may add up to more than the number of articles within a topic as some articles point out multiple future research opportunities (e.g., the use of alternate methods and variables, larger samples, etc.)

Overall, we found that first, the authors call for future studies that employ more comprehensive methods (e.g., other approaches or larger sample sizes). Second, the inclusion of more variables is important for the authors to reduce omitted variable bias and endogeneity concerns. Many of them suggest including not only additional controls, but further constructs such as trust, credibility, commitment, and intention (Gera and Kaur 2018 ). Third, nearly equally frequently, authors request future authors to the transfer their analysis to other contexts, such as to other types of crowdfunding. Sixteen articles mentioned this aspect, whereby eight specifically refer to shifting the focus from one crowdfunding type to another. Finally, other ideas for future research identified across the articles are: a change of perspective, for example by investigating other stakeholders, and the inclusion of other theories, e.g., Social Capital Theory or Social Cognitive Theory (Bandura et al. 1999 ; Shneor and Munim 2019 ).

5 Discussion

Personality is an important and under-researched topic in entrepreneurial finance, especially in the crowdfunding context, expressed in a growing body of research that has peaked in 2020. In this literature review, we retrieved articles focusing on nearly every personality construct included in the search terms (except for the “need for achievement”). Further, the more generalized search term “personality” uncovered additional personality constructs, which are risk-taking (Calic and Shevchenko 2020 ; Moss et al. 2015 ), autonomy (Calic and Shevchenko 2020 ; Moss et al. 2015 ), and traits associated with charisma and hubris (Sundermeier and Kummer 2019 ). Risk-taking describes the tendency to make risky decisions in the presence of uncertainty (Knight 1921 ); autonomy stands for the need for independence. Charisma and hubris combine personality traits attributed to entrepreneurs, such as excessive pride and self-confidence (hubris) or charm and persuasion (charisma) (Sundermeier and Kummer 2019 ).

We further find that within studies that focus on the Dark Triad, more studies cover narcissism than psychopathy or Machiavellianism. This difference could be rooted in the relatively high salience of the narcissism construct in narratives relative to the other traits. However, the popular and well-known measurement of narcissistic rhetoric introduced by Chatterjee and Hambrick (2007), while measuring CEO narcissism, might also be why many researchers focus on this trait.

5.1 Gaps and future research

In the following, we discuss key findings from our results in order of importance. We thereby not only examine the results of the quantitative articles included in the analysis of personality effects on crowdfunding performance, but combine these with the literature gaps identified by all articles included in the review. Therefore, we take a closer look at personality traits as non-linear, the use of narrative analysis methods, the context dependency of personality research in crowdfunding, and the specific personality perspective taken by the authors.

5.1.1 Personality as non-linear

Apart from the rather consistent results for openness and narcissism, the results differ from article to article and show no consistent pattern (Table 4 ). However, it is important to mention the inverted u-shape that authors often find for several personality traits. Miller ( 2015 ) argues convincingly that personality attributes are Janus-faced and that the negative aspects of the entrepreneurial personality have been largely ignored so far. Similarly, Calic and Shevchenko ( 2020 ) conclude that personality components such as innovativeness or risk propensity can be perceived as desirable by investors to a certain degree, but lose their positive appeal when over-expressed and hence are subject to a threshold effect. Although such nonlinear relationships appear to make sense when investigating personality in a complex context like crowdfunding, only a few authors analyze nonlinear relationships (e.g., quadratic relations) and surprisingly none mention this approach as potential for future research. We nevertheless argue that future research must pay special attention to these findings by testing for or including quadratic terms when examining personality effects in crowdfunding. A research question focusing on this non-linear relationship could entail: Do personality traits displayed in crowdfunding campaigns reach a saturation point at which they are overexpressed and consequently diminish the engagement/contribution level of the crowd? Answering this question would resolve inconsistencies in the current literature and fill a research gap regarding potentially underexplored quadratic effects of expressed personality in crowdfunding. Further, it would contribute to research on the effects of perceived personality expressions on impression formation (Hamilton et al. 1980 ). In practice, answering this question would also help crowdfunding entrepreneurs evaluate campaign material (e.g., videos) in a more nuanced way.

5.1.2 Use of different methods

Eleven of the studies examined base their research on software-based text analysis methods which are increasingly popular in entrepreneurship research, particularly so in studies related to personality. The perks are undeniable: employing this method facilitates access to larger samples that were not previously accessible. Using these methods, researchers rely on publicly available online text snippets such as letters to shareholders, IPO prospectuses, tweets, campaign page text, and even transcribed voice and video recordings, e.g., manager earning calls (Aerts and Yan 2017 ; Golbeck et al. 2011 ; Harrison et al. 2019 ; Loughran and McDonald 2013 ). However, the disadvantages of such methods should not be underestimated. On the one hand, there is the problem of validity. The methods employed are often validated only based on self-written imaginary text, generated in experimental settings and not on topic-specific text with an economic focus (Mairnesse et al. 2007 ; Pennebaker and King 1999 ). On the other hand, campaign pages’ texts are not necessarily authored by the entrepreneurs themselves, although assumed by this method of text-based personality assessment. It is also possible, that third parties such as public relations firms are hired to craft the campaign text on behalf of the entrepreneur or startup team. Analyzing these campaign texts, we must question whether the traits measured actually capture the campaign creator's personality.

So what do these studies actually measure? Some authors argue that they might have measured perceived personality rather than the entrepreneurs' true personality (Moss et al. 2015 ). Often, researchers are simply interested in the impact of personality traits as perceived by investors on crowdfunding success and do not require knowledge about the true underlying personality of the entrepreneur. As long as the studies find a correlation between the measured construct and crowdfunding success, the results suggest that the method is functioning as intended. Also, perceived personality could be a valid measure for a number of research questions, because investors are limited to the information presented on the campaign page. For instance, this could be the case for big data researchers or in entrepreneurial finance (Harrison et al. 2019 ), but may not be the case for psychologists that study personality in more personal context (Bozionelos 2017 ). In cases where the true personality of an entrepreneur is needed to answer a particular research question, text-based methods along with the stated limitations regarding perceived personality could present a real challenge. Future research could tackle this issue by combining, different methods such as combining text-based methods with psychological questionnaires as argued by Butticé and Rovelli ( 2020 ). Also, other studies analyzed within this paper highlight the need for the use of different methods while investigating personality in the crowdfunding context (see Table 5 ). Letting some of these authors speak for themselves they “encourage future researchers in crowdfunding to analyze empirical measures from crowdfunding platforms” (Rodriguez-Ricardo et al. 2019 , p. 12), argue that “qualitative and quantitative tools” (Davidson and Poor 2015 , p. 303) are needed in this research area, and emphasize that including e.g., questionnaires in their research model “would contribute to add reliability to our study and to rule out possible alternative explanations” (Butticé and Rovelli 2020 , p. 5). An unanswered research question focusing on the combination of different personality measurements, therefore, is: Does a narrative analysis of crowdfunding campaign texts reveal similar personality trait expressions as validated personality questionnaires conducted by the campaign owners? Research focusing on this question could contribute to the ongoing debate on the effect of individual-level attributes of the entrepreneur on campaign success. Revealing if the effect of perceived personality outweighs the effect of inner personality (or vice versa) in terms of venture financing success in crowdfunding could monumentally influence crowdfunding practice as entrepreneurs can shape their narratives, and by extension, their impressions on people, but their internal personality is more or less fixed (Costa and McCrae 1988 ).

5.1.3 Context dependence

Due to the newness of the crowdfunding research field and the use of highly recent methodologies still under development, there are few studies in general and even fewer replication studies in this area. Only one article intentionally replicates the work of another author team (Short and Anglin 2019 ). In their article, the authors conclude that “individuals should exercise extreme caution in regard to assuming that findings in one context can be generalized to others” as they “failed to replicate any of the hypotheses where the authors originally found support” in one of the included replication studies (Short and Anglin 2019 , p. 12). This comment by Short and Anglin ( 2019 ) is strikingly similar to what we actually observe in our review of studies in this field. Trying to summarize the relationships tested by the quantitative studies on personality and crowdfunding campaign success does not result in a clear picture (see Table 4 ). Instead, many studies find no effects, where others find effects or even contradictory results (e.g., Creek 2018 ; Leonelli et al. 2020 ).

One reason for this could lie in the different settings of the studies. Short and Anglin ( 2019 ) replicated the study by Moss et al. ( 2015 ) in a reward-based crowdfunding context whereas it was initially conducted with lending-based crowdfunding data. With this change in settings there is are also implicit changes in the basic features of the investigated construct, such as investor motivation. For example, while investors in reward-based crowdfunding are often assumed to be intrinsically motivated, investors in other crowdfunding types might behave differently (Cholakova and Clarysse 2015 ).

Further, it is somewhat puzzling why studies that measure the same constructs in similar settings obtain different results. For example, even in studies conducted in the same setting, e.g., reward-based crowdfunding and studying the same relationship, e.g., between perceived Big Five personality traits of entrepreneurs and campaign success and on the same platform (often Kickstarter), the results can differ (Gera and Kaur 2018 ; Thies et al. 2016 ). Although addressing a similar research question, there are striking differences in the methodologies of the full paper by Thies et al. ( 2016 ) and the short paper by Gera and Kaur ( 2018 ). First, the text used for the calculations in Thies et al. ( 2016 ) included the campaign text and the campaign description separately with similar results. On the other hand, Gera and Kaur ( 2018 ) use campaign descriptions and profile descriptions from the campaign owners. Second, whereas Thies et al. ( 2016 ) base their analysis on a regression model, Gera and Kaur ( 2018 ) (although mentioning logistic regressions) report only correlations as results. Third, Thies et al. ( 2016 ) analyze 33,420 campaign texts and 12,859 video transcripts, while Gera and Kaur ( 2018 ) do not include videos but instead opted to analyze a smaller number of 4059 campaign descriptions and 1721 creator profiles. Fourth, both author teams include different control variables in their analysis. Fifth, using a different time period to obtain the data and regulatory changes could cause systematically different results (Pollack et al. 2021 ). The example of these two papers (Gera and Kaur 2018 ; Thies et al. 2016 ), which appear similar at first, illustrates the problems that future researchers could solve by conducting replication studies. It is undeniable that personality constructs affect crowdfunding outcomes, but since the strength of the influence depends on the circumstances, researchers must pay particular attention to such details.

Therefore, we think that replication studies are particularly important for future research to determine differences in the effects of personality. First, replications are needed across types of crowdfunding and different platforms to observe the effect of this contextualization. This point was made by eight articles included in this research (Fig. 5 ; e.g., Bollaert et al. 2019 ; Leonelli et al. 2020 ) Second, even when the type of crowdfunding and platform are held constant, such replication studies are crucial to generate a reliable knowledge base about the relationships between personality constructs and crowdfunding outcomes. Third, as cultural and geographic factors could also influence crowdfunding outcomes, authors should consider including different regions in their studies as suggested by Bernardino and Santos ( 2016 ) and others (Table 5 ). A specific research question is: Which context-dependent variables moderate the effects of personality on crowdfunding? Answering this question could change how entrepreneurial science sees crowdfunding in that the role of personality could illustrate how the different types of crowdfunding might differ from each other more than they do from other forms of venture finance. Entrepreneurial displays of agreeableness to an audience of equity crowdfunding investors could have more implications for angel investments or IPOs than for reward-based crowdfunding and thereby open the opportunity for researchers to transfer findings from the accessible crowdfunding context to more traditional investment settings. Also, the scientific community could learn more about the role of individual crowdfunding platforms within a given type of crowdfunding (e.g., StartEngine and Wefunder for equity crowdfunding) in shaping the effect of individual characteristics like personality on campaign outcomes. Finally, we could also learn more about the role of national culture or geographic context in shaping how personality factors leading to crowdfunding success. This knowledge could help entrepreneurs who are thinking about entering new markets or expanding across borders.

5.1.4 Change of personality perspective

In the literature reviewed, we see a focus on studying the personality of the entrepreneur who is assumed to be the campaign creator. Studies on investors' personality, on the other hand, are less frequently conducted, even though there are relatively easy to investigate by survey studies while entrepreneurs are more difficult to access directly regarding their personality (Hambrick and Mason 1984 ). Studies on investors' personality typically use inventory-based questionnaires (Rodriguez-Ricardo et al. 2019 ; Shneor and Munim 2019 ), but have so far neglected studying investor comments for example. There have, however, been studies that investigate investor comment sentiment (Wang et al. 2018 ) which seems to be leading in a fruitful direction.

Only a few of the articles reviewed focus on the investor personality perspective. They find that social identification with the crowdfunding community and the individual level of innovativeness, unlike internal locus of control, positively affect the intention to participate in crowdfunding (Rodriguez-Ricardo et al. 2018 , 2019 ). Further, Ryu and Kim ( 2016 ) categorize crowdfunders into four groups (angelic backers, reward hunters, avid fans, tasteful hermits) employing various factors including the Big Five personality traits, whereas Shneor and Munim ( 2019 ) find differences in self-efficacy between investors that contribute higher vs. lower amounts to campaigns.

Only one article by Moritz et al. ( 2015 ) includes more than one personality perspectives (e.g., investor, entrepreneur, involved third parties such as platforms). In their qualitative study, they investigate how information asymmetries within the crowdfunding process can be reduced by communication (e.g., of soft factors) between the parties involved via the internet (Moritz et al. 2015 ). Nevertheless, the authors of the analyzed articles also recognize the potential that arises from investigating other perspectives (Table 5 ). They argue that future research “should consider the role that [all actors (crowdfunders, fund seeker and platforms)] play in this new phenomenon” (Rodriguez-Ricardo et al. 2018 , p. 178) and that it is important to “further analyze the relationship between lender characteristics and those of borrowers” (Moss et al. 2015 , p. 47).

Including several perspectives is a promising task for future research. As the saying “Birds of a feather flock together” implies, people that share specific characteristics get along better. In his paper on homogeneity, Marsden ( 1988 ) discovers that people that have strong social relations are more likely to share similar attributes. Transferring this idea to the crowdfunding context, Venturelli et al. ( 2020 ) investigated the effects of ethnic and gender similarities between investors and entrepreneurs and the positive impact on funding in equity-based crowdfunding. Oo et al. ( 2019 ) focus on the mediating effect of similarity (in-group favoritism) between entrepreneurs and investors in reward-based crowdfunding. Additionally, Burtch et al. ( 2014 ) found that crowdfunders prefer culturally similar and geographically proximate fund-seekers. Lin and Viswanathan (2016) refer to this phenomenon as “home bias”. Similarly, Mollick ( 2014 ) suggests that geography may play an important role. These studies demonstrate the importance of investigating the relationship between funding seekers and investors in the crowd. Therefore, we strongly encourage research on the personality of all parties involved in the crowdfunding process and especially the interaction between investors’ and entrepreneurs’ personality. A concrete research question dealing with this change of perspective is: Are there interactions between the personalities displayed by entrepreneurs and those of the contributing investors in the crowd? Answering this question could impact how entrepreneurs approach investors in the crowd. It would also shed light on investors' selection processes when finding crowdfunding campaigns to invest in.

5.2 Implications

Our results have a number of implications for research and practice. First, our study implies that quantitative crowdfunding researchers should pay particular attention to the type of crowdfunding, the measure of success utilized and the selected personality traits when designing their studies. Second, the mixed results for many traits imply a strong need for replication studies to validate the results and methods used. Third, authors should consider qualitative and mixed-methods approaches in future studies to advance and deepen our theoretical knowledge and not just test existing knowledge or theory. Fourth, personality researchers, our results imply that many of these constructs may not be fully distinctive from one another or optimally measured in crowdfunding by using narrative approaches alone. Therefore, it could be helpful to combine different types of analysis to better capture personality traits (e.g., the analysis of campaign text narratives with the analysis of pitch videos, observer ratings or questionnaires). Finally, our results can feed into big data approaches and into studies on deception in crowdfunding and other forms of entrepreneurial finance (e.g., Siering et al. 2016 ; von Selasinsky and Isaak 2020 ).

Our study also has several practical implications. First, for entrepreneurs seeking capital from the crowd, our results imply that displaying certain types of personality when crafting their campaign narratives (e.g., openness) in certain types of crowdfunding (e.g., reward-based) can indeed impact the success of their campaign (see Table 4 ). Entrepreneurs that display openness are presumably more likely to be perceived as having the necessary networking capabilities to succeed with a startup venture.

Second, by examining the results in comparison, investors in the crowd could screen campaigns for traits in which entrepreneurs display personality that improves (or reduces) the probability of a successful outcome, guiding their investment decision beyond just utilization of hard facts (e.g., the number of backers so far and the amount collected so far). Third, crowdfunding platforms could add personality screening inventories when conducting their project due diligence when evaluating project risks (together with other existing factors such as screening for typos and completeness of the campaign text and multimedia) to better pick the winners and improve their preselection of which projects are allowed to enter the crowdfunding process.

5.3 Limitations

Our study also has a number of limitations. First, due to the specialized nature of the subject which requires interdisciplinary approaches, our review covers only a limited number of articles. Second, which factors should be considered as personality traits in a narrower sense is not always clear. We included those which are mostly unquestioned in psychology (particularly the Big Five and the Dark Triad traits) and a number of additional traits that are frequently used in studies that appear in top entrepreneurship journals (e.g., ETP, JBV, etc.) in our literature review (Costa Jr and McCrae 1995 ; Paulhus and Williams 2002 ; Rauch and Frese 2007 ). Nonetheless, this could be further extended by incorporating studies on what some psychologists now refer to as the sixth basic component of personality (the Honesty-Humility trait, yielding the Big Six, also known under the acronym HEXACO) (Ashton and Lee 2007 ; Saucier 2009 ). Third, researchers often refer to other psychological constructs while investigating entrepreneurial behavior. These include passion, which describes a strong inclination towards a specific activity (Murnieks et al. 2014 ) and altruism, i.e., prosocial behavior (Batson and Powell, 2003 ). Although passion is more of an emotional (Anglin et al. 2018a ; Avey et al. 2008 ) and altruism is more a motivational construct (Rushton et al. 1981 ) than a personality trait, further research could investigate both in the context of crowdfunding. While including these would have been out of the scope of this study, in an additional informal screening of such literature, we found very few such studies, highlighting a significant research gap regarding plurality of actor perspectives when examining crowdfunding and personality.

5.4 Conclusion

We conclude our literature review on personality research in crowdfunding by noting that this is a very young and budding research field, which still offers considerable room for further research. Our results question a finding of the article “How Should Crowdfunding Research Evolve” that reports no interest by leading editors surveyed in the research field of ‘personality theories’ in crowdfunding (McKenny et al. 2017 ). Recently, however, we observe an increase in published studies in this research field which indicates growing interest by the scientific community. Newly available analysis methods might be driving this trend. For example, scraping techniques have evolved to more easily gather online data; also, new software tools such as those based on artificial intelligence capitalize on big data approaches and permit the investigation of personality in novel ways.

By identifying crucial gaps in the literature for future research and by highlighting which approaches are needed for this research stream to evolve our review contributes to research on crowdfunding and personality (e.g., Anglin et al. 2018a ; Moss et al. 2015 ) and to research on the entrepreneurial personality more generally (e.g., Kets de Vries 1977 ; Rauch and Frese 2014 ). First, future studies should examine non-linear relations between expressed personality traits and crowdfunding success, as personality traits are not dichotomous and can cause different behavior depending on the intensity of expression. Second, there is a need for studies that employ different methods such as mixed-methods approaches to validate narrative analysis techniques with, for example questionnaires or experiments. Third, to obtain a clear picture of personality effects in crowdfunding, replication studies in similar and different contexts are of crucial importance to this scientific field. Fourth, our review revealed that a plurality of personality perspectives would strengthen future research. We hope that our review article will help to encourage research in this area and provide researchers with a first systematic overview of the field.

Aerts W, Yan B (2017) Rhetorical impression management in the letter to shareholders and institutional setting. Account Audit Account J 30(2):404–432

Article Google Scholar

Ahlers GK, Cumming D, Günther C, Schweizer D (2015) Signaling in Equity Crowdfunding. Entrep Theory Pract 39(4):955–980

Alegre I, Moleskis M (2019) Beyond financial motivations in crowdfunding: a systematic literature review of donations and rewards. Voluntas Int J Voluntary Nonprofit Org 32(2):276–287

Allison TH, Davis BC, Webb JW, Short JC (2017) Persuasion in crowdfunding: An elaboration likelihood model of crowdfunding performance. J Bus Ventur 32(6):707–725

Almlund M, Duckworth AL, Heckman J, Kautz T (2011) Personality Psychology and Economics. In: Ha-nu-s-hek EA, Ma-chin S, Woess-mann L (eds) Handbook of the Economics of Education. Handbook of The Economics of Education, Elsevier, Amsterdam p, pp 1–181

Anglin AH, Short JC, Drover W, Stevenson RM, McKenny AF, Allison TH (2018a) The power of positivity? The influence of positive psychological capital language on crowdfunding performance. J Bus Ventur 33(4):470–492

Anglin AH, Wolfe MT, Short JC, McKenny AF, Pidduck RJ (2018b) Narcissistic rhetoric and crowdfunding performance: A social role theory perspective. J Bus Ventur 33(6):780–812

Ashton MC, Lee K (2007) Empirical, theoretical, and practical advantages of the HEXACO model of personality structure. Pers Soc Psychol Rev 11(2):150–166

Avey JB, Wernsing TS, Luthans F (2008) Can positive employees help positive organizational change? Impact of psychological capital and emotions on relevant attitudes and behaviors. J Appl Behav Sci 44:48–70

Bandura A, Freeman WH, Lightsey R (1999) Self-Efficacy: The Exercise of Control. J Cogn Psychother 13(2):158–166

Barbi M, Mattioli S (2019) Human Capital, Investor Trust, and Equity Crowdfunding. Res Int Bus Financ 49:1–12

Barrick MR, Mount MK (1991) The big five personality dimensions and job performance: a meta-analysis. Pers Psychol 44(1):1–26

Barrick MR, Mount MK, Judge TA (2001) Personality and Performance at the Beginning of the New Millennium: What Do We Know and Where Do We Go Next? Int J Sel Assess 9(1 and 2):9–30

Batson CD, Powell AA (2003) Altruism and Prosocial Behavior. In: Weiner IB (ed) Handbook of Psychology. Wiley, Inc, Hoboken, NJ, USA

Belleflamme P, Lambert T, Schwienbacher A (2010) Crowdfunding: An industrial organization perspective. Prepared for the workshop Digital Business Models: Understanding Strategies’, held in Paris on June, 25–26.

Bernardino S, Santos JF (2016) Financing social ventures by crowdfunding. Int J Entrep Innov 17(3):173–183

Bertrand M, Schoar A (2003) Managing with Style: The Effect of Managers on Firm Policies. Q J Econ 118(4):1169–1208

Bi S, Liu Z, Usman K (2017) The influence of online information on investing decisions of reward-based crowdfunding. J Bus Res 71:10–18

Block J, Hornuf L, Moritz A (2018) Which Updates during an Equity Crowdfunding Campaign Increase Crowd Participation? Small Bus Econ 50(1):3–27

Block JH, Fisch CO, Obschonka M, Sandner PG (2019) A personality perspective on business angel syndication. J Bank Finance 100:306–327

Bollaert H, Leboeuf G, Schwienbacher A (2019) The narcissism of crowdfunding entrepreneurs. Small Bus Econ 47(4):819

Borst I, Moser C, Ferguson J (2018) From friendfunding to crowdfunding: Relevance of relationships, social media, and platform activities to crowdfunding performance. New Media Soc 20(4):1396–1414

Bouncken RB, Komorek M, Kraus S (2015) Crowdfunding: The Current State of Research. Int Bus Econom Res J (IBER) 14(3):407

Bouncken R, Cesinger B, Tiberius V (2020) Narcissism, Machiavellianism, and psychopathy of top managers: Can Entrepreneurial Orientation secure performance. Int J Entrep Ventur 12(3):273

Boyce CJ, Wood AM, Ferguson E (2016) Individual differences in loss aversion: conscientiousness predicts how life satisfaction responds to losses versus gains in income. Pers Soc Psychol Bull 42(4):471–484

Bozionelos N (2004) The big five of personality and work involvement. J Manag Psychol 19(1):69–81

Bozionelos G (2017) The relationship of the big-five with workplace network resources: More quadratic than linear. Personality Individ Differ 104:374–378

Brandstätter H (2011) Personality aspects of entrepreneurship: A look at five meta-analyses. Personality Individ Differ 51(3):222–230

Burtch G, Ghose A, Wattal S (2014) Cultural differences and geography as determinants of online prosocial lending. MIS Q 38(3):773–794

Butticé V, Rovelli P (2020) “Fund me, I am fabulous!” Do narcissistic entrepreneurs succeed or fail in crowdfunding? Personal Individ Differ 162:110037

Butticé V, Franzoni C, Rossi-Lamastra C, Rovelli P (2018) The Road to Crowdfunding Success: A Review of Extant Literature. In: Tucci CL (ed) Creating and capturing value through crowdsourcing, 1st edn. Oxford University Press, Oxford, pp 97–126

Cai W, Polzin F, Stam E (2021) Crowdfunding and social capital: A systematic review using a dynamic perspective. Technol Forecast Soc Change 162:120412

Calic G, Shevchenko A (2020) How signal intensity of behavioral orientations affects crowdfunding performance: The role of entrepreneurial orientation in crowdfunding business ventures. J Bus Res 115:204–220

Ceballos R, Hou W, Levitas E, Price SM (2017) The product, the mind and the heart of crowdfunding: The effect of signals on technology projects. Int J Serv Stand 12(1):79–99

Chan CSR, Parhankangas A (2017) Crowdfunding Innovative Ideas: How Incremental and Radical Innovativeness Influence Funding Outcomes. Entrep Theory Pract 41(2):237–263

Chen CC, Greene PG, Crick A (1998) Does entrepreneurial self-efficacy distinguish entrepreneurs from managers? J Bus Ventur 13(4):295–316

Cholakova M, Clarysse B (2015) Does the Possibility to Make Equity Investments in Crowdfunding Projects Crowd Out Reward-Based Investments? Entrep Theory Pract 39(1):145–172

Ciavarella MA, Buchholtz AK, Riordan CM, Gatewood RD, Stokes GS (2004) The Big Five and venture survival: Is there a linkage? J Bus Ventur 19(4):465–483

Colombo O (2020) The Use of Signals in New-Venture Financing: A Review and Research Agenda. J Manag 38:014920632091109

Costa PT, McCrae RR (1988) Personality in adulthood: A six-year longitudinal study of self-reports and spouse ratings on the NEO Personality Inventory. J Pers Soc Psychol 54(5):853–863

Costa PT Jr, McCrae RR (1995) Domains and facets: hierarchical personality assessment using the revised NEO personality inventory. J Pers Assess 64(1):21–50

Courtney C, Dutta S, Li Y (2017) Resolving Information Asymmetry: Signaling, Endorsement, and Crowdfunding Success. Entrep Theory Pract 41(2):265–290

Creek SA (2018) The dark triad and entrepreneurial fundraising: a comparison of rewards-based and equity crowdfunding campaigns. Acad Manag Proceed 2019(1):15932

Cumming D, Deloof M, Manigart S, Wright M (2019a) New directions in entrepreneurial finance. J Bank Finance 100:252–260

Cumming D, Meoli M, Vismara S (2019b) Investors’ choices between cash and voting rights: Evidence from dual-class equity crowdfunding. Res Pol 48(8):103740

Dalla Chiesa C, Handke C (2020) Crowdfunding. In: Towse R (ed) Handbook of Cultural Economics. Edward Elgar Publishing, Third Edition

Davidson R, Poor N (2015) The barriers facing artists’ use of crowdfunding platforms: Personality, emotional labor, and going to the well one too many times. New Media Soc 17(2):289–307

Davis BC, Hmieleski KM, Webb JW, Coombs JE (2017) Funders’ positive affective reactions to entrepreneurs’ crowdfunding pitches: The influence of perceived product creativity and entrepreneurial passion. J Bus Ventur 32(1):90–106

Deci EL, Ryan RM (2008) Self-determination theory: A macrotheory of human motivation, development, and health. Can Psychol 49(3):182–185

Drover W, Busenitz L, Matusik S, Townsend D, Anglin A, Dushnitsky G (2017) A Review and Road Map of Entrepreneurial Equity Financing Research: Venture Capital, Corporate Venture Capital, Angel Investment, Crowdfunding, and Accelerators. J Manag 43(6):1820–1853

Eagly AH, Wood W (2012) Social Role Theory. In: van Lange P, Kruglanski A, Higgins E (eds) Handbook of Theories of Social Psychology. SAGE Publications Ltd, London, United Kingdom, pp 458–476

Chapter Google Scholar

Engelen A, Neumann C, Schmidt S (2016) Should Entrepreneurially Oriented Firms Have Narcissistic CEOs? J Manag 42(3):698–721

Fisch C, Block J (2018) Six tips for your (systematic) literature review in business and management research. Management Review Quarterly 68(2):103–106

Furnham A, Richards SC, Paulhus DL (2013) The Dark Triad of Personality: A 10 Year Review. Soc Pers Psychol Compass 7(3):199–216

Gera J, Kaur H (2018) Influence of Personality Traits on Campaign Success. In 8 th International Conference on Cloud Computing, Data Science and Engineering (Confluence) (pp. 14–15). IEEE.

Golbeck J, Robles C, Edmondson M, Turner K (2011) Predicting Personality from Twitter. In 2011 IEEE Third Int l Conference on Privacy, Security, Risk and Trust and 2011 IEEE Third Int l Conference on Social Computing (pp. 149–156). IEEE.

Gruda D, McCleskey J, Karanatsiou D, Vakali A (2021) I’m simply the best, better than all the rest: Narcissistic leaders and corporate fundraising success. Personal Indiv Differ 168:110317

Hambrick DC, Mason PA (1984) Upper Echelons: The Organization as a Reflection of Its Top Managers. Acad Manag Rev 9(2):193–206

Hamilton DL, Katz LB, Leirer VO (1980) Cognitive representation of personality impressions: Organizational processes in first impression formation. J Pers Soc Psychol 39(6):1050–1063

Harburg E, Hui J, Greenberg M, Gerber EM (2015) Understanding the effects of crowdfunding on entrepreneurial self-efficacy. CSCW 2015 Proceedings of the 2015 ACM International conference on computer-supported cooperative work and social computing.

Harrison JS, Thurgood GR, Boivie S, Pfarrer MD (2019) Measuring CEO personality: Developing, validating, and testing a linguistic tool. Strateg Manag J 9(1):9

Hoffman DL, Novak TP, Schlosser AE (2003) Locus of Control, Web use, and Consumer Attitudes toward Internet Regulation. J Public Policy Mark 22(1):41–57

Hogan J, Ones DS (1997) Conscientiousness and Integrity at Work. In: Hogan R, Briggs S, Johnson J (eds) Handbook of Personality Psychology. Elsevier, pp 849–870

Iurchenko D (2019) Three Essays on Equity Crowdfunding and the Digitalization of Entrepreneurial Finance. Dissertation: Université de Lausanne, Faculté des hautes études commerciales.

Janku J, Kucerova Z (2018) Successful crowdfunding campaigns: the role of project specifics, competition and founders’ experience. Czech J Econom Finance 68(4):351–373

Johnson EW, McClelland DC (1984) Learning to achieve. Scott, Foresman, Glenview, IL

Jones MV, Coviello N, Tang YK (2011) International Entrepreneurship research (1989–2009): A domain ontology and thematic analysis. J Bus Ventur 26(6):632–659

Jovanović T (2019) Crowdfunding: what do we know so far? Int J Innov Technol Manag 16(01):1950009

Judge TA, Ilies R (2002) Relationship of personality to performance motivation: a meta-analytic review. J Appl Psychol 87(4):797–807

Judge TA, Higgins CA, Thoresen CJ, Barrick MR (1999) The big five personality traits, general mental ability and career success. Person Psychol 52(3):621–652

Kaartemo V (2017) The elements of a successful crowdfunding campaign: a systematic literature review of crowdfunding performance. Int Rev Entrepreneurship 15(3):291–318

Kerr SP, Kerr W, Xu T (2017) Personality Traits of Entrepreneurs: A Review of Recent Literature. National Bureau of Economic Research Working Paper Series.

Kets de Vries MF (1977) The Entrepreneurial Personality: A Person at the Crossroads. J Manage Stud 14(1):34–57

Kim MJ, Hall CM (2021) Do value-attitude-behavior and personality affect sustainability crowdfunding initiatives? J Environ Manag 280:111827

Kim MJ, Hall CM, Han H (2021) Behavioral influences on crowdfunding SDG initiatives: the importance of personality and subjective well-being. Sustainability 13(7):3796

Knight FH (1921) Risk, Uncertainty and Profit, vol 31. Houghton Mifflin, Boston, MA

Kuckertz A, Prochotta A (2018) What’s Hot in Entrepreneurship Research 2018? Hohenheim Entrepreneurship Research Brief . University of Hohenheim, Stuttgart

Landström H, Harirchi G (2019) “That’s Interesting!” in Entrepreneurship Research. J Small Bus Manage 57(sup2):507–529

Lee DY, Tsang EWK (2001) The effects of entrepreneurial personality, background and network activities on venture growth*. J Manage Stud 38(4):583–602

Leonelli S, Di Pietro F, Masciarelli F (2020) Narcissism, Machiavellianism and Psychopathy: How Do Displayed Entrepreneurs’ Personality Dark Traits Influence Crowdfunding Success? Unveiling the cognitive and emotional aspect of entrepreneurship In The Entrepreneurial Behaviour. Emerald Publishing Limited, England

Loughran T, McDonald B (2013) IPO first-day returns, offer price revisions, volatility, and form S-1 language. J Financ Econ 109(2):307–326

Macht S, Chapman G (2019) Getting more than money through online crowdfunding. Asia-Pacific J Bus Admin 11(2):171–186

Mairnesse F, Walker MA, Mehl MR, Moore RK (2007) Using linguistic cues for the automatic recognition of personality in conversation and text. J Artif Intell Res 30:457–500

Manning KC, Bearden WO, Madden TJ (1995) Consumer Innovativeness and the Adoption Process. J Consum Psychol 4(4):329–345

Marsden PV (1988) Homogeneity in confiding relations. Social Networks 10(1):57–76

Mayfield, C., Perdue, G., and Wooten, K. (2008). Investment management and personality type. Financial Services Review, 17 (3), 219–236, from http://search.ebscohost.com/login.aspx?direct=trueanddb=buhandAN=35767948andsite=ehost-live .

McAdams DP, Pals JL (2006) A new Big Five: fundamental principles for an integrative science of personality. Am Psychol 61(3):204–217

McKenny AF, Allison TH, Ketchen DJ, Short JC, Ireland RD (2017) How Should Crowdfunding Research Evolve? A Survey of the Entrepreneurship Theory and Practice Editorial Board. Entrep Theory Pract 41(2):291–304

Miller D (2015) A downside to the entrepreneurial personality? Ent Theory & Pract 39:1–8

Miller JD, Widiger TA, Campbell WK (2010) Narcissistic personality disorder and the DSM-V. J Abnorm Psychol 119(4):640–649

Mochkabadi K, Volkmann CK (2020) Equity crowdfunding: a systematic review of the literature. Small Bus Econ 54(1):75–118

Moleskis M, Alegre I, Canela MA (2019) Crowdfunding Entrepreneurial or Humanitarian Needs? The Influence of Signals and Biases on Decisions. Nonprofit Volunt Sect Q 48(3):552–571

Moleskis M, Alegre I (2018) Crowdfunding: A Short Past and Long Future. Available at SSRN 3163006 .

Mollick E (2014) The dynamics of crowdfunding: An exploratory study. J Bus Ventur 29(1):1–16

Moritz A, Block JH (2016) Crowdfunding: A Literature Review and Research Directions. In: Brüntje D, Gajda O (eds) FGF Studies in Small Business and Entrepreneurship Crowdfunding in Europe. Springer International Publishing, Cham, pp 25–53

Moritz A, Block J, Lutz E (2015) Investor communication in equity-based crowdfunding: a qualitative-empirical study. Qual Res Financ Markets 7(3):309–342

Moss TW, Neubaum DO, Meyskens M (2015) The effect of virtuous and entrepreneurial orientations on microfinance lending and repayment: a signaling theory perspective. Entrep Theory Pract 39(1):27–52

Mueller SL, Thomas AS (2001) Culture and entrepreneurial potential. J Bus Ventur 16(1):51–75

Murnieks CY, Mosakowski E, Cardon MS (2014) Pathways of passion. J Manag 40:1583–1606

Netzer O, Lemaire A, Herzenstein M (2019) When Words Sweat: Identifying Signals for Loan Default in the Text of Loan Applications. J Mark Res 56(6):960–980

Oo PP, Allison TH, Sahaym A, Juasrikul S (2019) User entrepreneurs’ multiple identities and crowdfunding performance: Effects through product innovativeness, perceived passion, and need similarity. J Bus Ventur 34(5):105895

Paulhus DL, Williams KM (2002) The Dark Triad of personality: Narcissism, Machiavellianism, and psychopathy. J Res Pers 36(6):556–563

Pennebaker JW, King LA (1999) Linguistic styles: Language use as an individual difference. J Pers Soc Psychol 77(6):1296–1312

Pollack JM, Maula M, Allison TH, Renko M, Günther CC (2021) Making a Contribution to Entrepreneurship Research by Studying Crowd-Funded Entrepreneurial Opportunities. Entrep Theory Pract 6(3):104225871988864

Prodromos K, Theriou G, Sarigiannidis L (2014) Crowdfunding: the factors that affect the intention of users/ consumers to participate financially at crowdfunding projects. Practical Issues Manag Econom 7(1):49–70

Puri M, Zarutskie R (2012) On the life cycle dynamics of venture-capital- and non-venture-capital-financed firms. J Financ 67(6):2247–2293

Rauch A, Frese M (2014) Born to Be an Entrepreneur? Revisiting the Personality Approach to Entrepreneurship. In J. R. Baum, M. Frese, and R. A. Baron (Eds.), SIOP organizational frontier series The psychology of entrepreneurship. East Sussex: Psychology Press, England. p. 73–98

Rauch A, Frese M (2007) Let’s put the person back into entrepreneurship research: A meta-analysis on the relationship between business owners’ personality traits, business creation, and success. Eur J Work Organ Psy 16(4):353–385

Rodriguez-Ricardo Y, Sicilia M, López M (2018) What drives crowdfunding participation? The influence of personal and social traits [¿Qué determina la participacíon en proyectos de crowdfunding? La influencia de las características sociales y personales]. Span J Market ESIC 22(2):163–182

Rodriguez-Ricardo Y, Sicilia M, Lopez M (2019) Altruism and Internal Locus of Control as Determinants of the Intention to Participate in Crowdfunding: The Mediating Role of Trust. J Theor Appl Electron Commer Res 14(3):1–16

Rogers EM, Shoemaker FF (1971) Communication of Innovations; a cross-cultural approach. The Free Press, New York, NY

Rotter JB (1966) Generalized expectancies for internal versus external control of reinforcement. Psychol Monogr Gen Appl 80(1):1–28

Rottler M, Helmig B, Ahrens J-P (2020) The Biased Crowd? Personality Perceptions in Crowdfunding. In Proceedings of the International Conference on Information Systems - ICIS 2020, Munich, Germany, December 15–18, 2019. Atlanta, GA: AISeL.

Rushton PJ, Chrisjohn RD, Fekken CG (1981) The altruistic personality and the self-report altruism scale. Pers Individ Differ 2:293–302

Ryu S, Kim Y-G (2016) A typology of crowdfunding sponsors: Birds of a feather flock together? Electron Commer Res Appl 16:43–54

Salido-Andres N, Rey-Garcia M, Alvarez-Gonzalez LI, Vazquez-Casielles R (2020) Mapping the field of donation-based crowdfunding for charitable causes: systematic review and conceptual framework. Voluntas: International Journal of Voluntary and Nonprofit Organizations, 1–15, 288-302

Saucier G (2009) Recurrent personality dimensions in inclusive lexical studies: indications for a big six structure. J Pers 77(5):1577–1614

von Selasinsky C, Isaak AJ (2020) It’s all in the (Sub-)title? Expanding Signal Evaluation in Crowdfunding Research. ECIS 2020 Research-in-Progress Papers. 56. https://aisel.aisnet.org/ecis2020_rip/56 .

Shin M-H, Lee Y-M (2020) The Effects of Consumption Value and Consumer Trust on Crowdfunding Participation Intention. The J Distrib Sci 18(6):93–101

Shneor R, Munim ZH (2019) Reward crowdfunding contribution as planned behaviour: An extended framework. J Bus Res 103:56–70

Shneor R, Vik AA (2020) Crowdfunding success: a systematic literature review 2010–2017. Balt J Manag 15(2):149–182

Short JC, Anglin AH (2019) Is leadership language ‘rewarded’in crowdfunding? Replicating social entrepreneurship research in a rewards-based context. J Bus Ventur Insights 11:e00121

Siering M, Koch J-A, Deokar AV (2016) Detecting fraudulent behavior on crowdfunding platforms: the role of linguistic and content-based cues in static and dynamic contexts. J Manag Inf Syst 33(2):421–455

Spence M (1978) Job Marketing Signaling. Uncertainty in Economics. Elsevier, Cambridge, MA, pp 281–306

Stevenson RM, Ciuchta MP, Letwin C, Dinger JM, Vancouver JB (2019) Out of control or right on the money? Funder self-efficacy and crowd bias in equity crowdfunding. J Bus Ventur 34(2):348–367

Sundermeier J, Kummer T (2019) Startup Founders’ Personality Attributes in Crowdfunding Campaigns: The Relevance of Hubris and Charisma in Raising Seed Funding Online. In T. Bui (Ed.), Proceedings of the 52 nd Hawaii International Conference on System Sciences .

Thies F, Wessel M, Rudolph J, Benlian A (2016) Personality Matters: How Signaling Personality Traits Can Influence the Adoption and Diffusion of Crowdfunding Campaigns. Research Papers. 36, from https://aisel.aisnet.org/ecis2016_rp/36 .

Troise C, Tani M (2020) Exploring entrepreneurial characteristics, motivations and behaviours in equity crowdfunding: some evidence from Italy. Manag Decision. 59(5):995–1024

Tseng F-C (2020) Joint contribution and consumption through online crowdfunding campaigns. Cogent Bus Manage 7(1):1843308

Venturelli V, Pedrazzoli A, Gallo G (2020) Birds of a feather flock together: the inclusive effect of similarity patterns in equity crowdfunding. Sustainability 12(9):3539

Wang W, Zhu K, Wang H, Wu Y-CJ (2017) The impact of sentiment orientations on successful crowdfunding campaigns through text analytics. IET Softw 11:229–238

Wang N, Li Q, Liang H, Ye T, Ge S (2018) Understanding the importance of interaction between creators and backers in crowdfunding success. Electron Commer Res Appl 27:106–117

Wang W, Chen W, Zhu K, Wang H (2020) Emphasizing the entrepreneur or the idea? The impact of text content emphasis on investment decisions in crowdfunding. Decision Support Sys 136:113341

Watson D, Clark LA (1997) Extraversion and Its Positive Emotional Core. In: Hogan R, Briggs S, Johnson J (eds) Handbook of Personality Psychology. Elsevier, pp 767–793

Webster J, Watson RT (2002) Analyzing the past to prepare for the future: Writing a literature review. Mis Quarterly, xiii–xxiii.

Wiklund J, Davidsson P, Audretsch DB, Karlsson C (2011) The Future of Entrepreneurship Research. Entrep Theory Pract 35(1):1–9

Younkin P, Kuppuswamy V (2018) The Colorblind Crowd? Founder Race and Performance in Crowdfunding. Manage Sci 64(7):3269–3287

Zhao L, Ryu S (2020) Reward-based crowdfunding research and practice. In: Shneor R, Zhao L, Flåten B-T (eds) Advances in crowdfunding. Springer International Publishing, Cham, pp 119–143

Download references

Open Access funding enabled and organized by Projekt DEAL. Research generously supported by the Manchot Graduate School “Competitiveness of Young Enterprises” of the Heinrich-Heine-University of Düsseldorf. Funding was provided by the Jürgen Manchot Stiftung.

Author information

Authors and affiliations.

Manchot Graduate School, Competitiveness of Young Enterprises (WEJU), Heinrich-Heine-University of Düsseldorf, Universitätsstrasse 1, 40225, Düsseldorf, Germany

Julia Neuhaus, Andrew Isaak & Denefa Bostandzic

Corporate Finance, Witten/Herdecke University, Alfred-Herrhausen-Straße 50, 58448, Witten, Germany

Denefa Bostandzic

You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Julia Neuhaus .

Ethics declarations

Conflict of interest.

The authors declare that they have no conflict of interest.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and Permissions

About this article

Cite this article.

Neuhaus, J., Isaak, A. & Bostandzic, D. Million dollar personality: a systematic literature review on personality in crowdfunding. Manag Rev Q 72 , 309–345 (2022). https://doi.org/10.1007/s11301-021-00242-9

Download citation

Received : 07 December 2020

Accepted : 08 October 2021

Published : 10 November 2021

Issue Date : June 2022

DOI : https://doi.org/10.1007/s11301-021-00242-9

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

JEL Classification

Essay Service Examples Psychology Big Five Personality Traits

The Big Five Personality Traits: Review Of Literature And Reflection

Introduction

Personality plays a big part of who you are as an individual. There are many types of personality traits. In this research paper it gives you a definition of what personality is, how does someone personality affect their development. It tells you about the “Big Five” personality traits, it then goes into details about every one of them to let you know how it affect you as a person whether it’s negative or positive. It explains how personality are assessed and examples of projective and objectives of personality assessments. It has an interesting part which is a brief autobiography on my life in relation to how it contributed to my personality development. It ends with a discussion concerning strategies I will put in place to engage in reflective exercises to remain self-aware as an early childhood educator.

A Review of Literature

Personality traits are a reflection of who you are as an individual, it give others an idea of your characteristics. The big five personality traits are extroversion, conscientiousness, agreeableness, openness and neuroticism. (Pappas). Extroversion is when someone takes the pleasure in being around other people rather than being alone. (Cherry) .They are usually the center of attention, they are usually a people’s person, easy to get along with and they can be positive, cheerful and outgoing. Extroversion is a combination of both nature and nurture. People who were overly extroverted as children may find that they are more of an introvert later on in their lives as they go into adulthood.

Conscientiousness is a desire to do one’s duty and to take obligations to others seriously. (Thiel). People who conscientious tend to be well organized and have a strong sense of duty. They like to plan ahead of time which makes them not spontaneous, they are self-discipline, and they are also responsible and reliable. People who have a high level of conscientiousness can be self-efficacy, achievement-striving, perfectionist as they set excessively high performance standards and they can come across as workaholics, they tend to enjoy their work which may cause them to work a lot.

Agreeableness is when a person goes along with whatever is being said even though they don’t think it’s right, it is the quality of being pleasant and friendly(Thiel) . Agreeableness is important because it can create better relationships, it prevents arguments and it helps prevent aggression. Someone who turns out to be too agreeable can be taken advantage of because it may seem like they don’t have a mind of their own. Being too agreeable makes a person more prone to disappointment and they can worry too much about not being liked.

Openness is the quality of letting anyone in your life, it’s a lack of restriction, letting people have accessibility to your life. (Cherry). Being open allows you to learn new things, it allows people to get to know the real you, it also allows you to meet new people. Most people are afraid to be open with other people because they think others may judge them and try to use the information gain against them which may not be the case. Persons who are open minded tend to be curious, creative, adventurous and flexible.

Neuroticism is a form of emotional reaction which can be disproportionate worrying and anxiety, whereas neurotic people tend to worry about everything which may cause them to easily slip into a state of depression. (Thiel) If their lives are going all too well, they then find something to worry about. Neurotic people dwell more on the negative aspects of situation rather than the positive side of things. When they struggle with life and all the stress that comes with life, they become very frustrated and angry. They basically can’t or don’t know how to deal with obstacles that come their way. They don’t know how to deal with their emotions. Some traits of people who are neurotic are immoderate, anger, vulnerable, depression, anxiety and self-conscious.

Personality is assessed by the measurement of someone’s characteristic. Projective assessments are tests where you are shown an ambiguous image and you say the first thing that comes to your mind. It is a test that expose a person’s unconscious perceptions. (Cherry). Two examples are the Thematic Apperception Test and the Rorschach test. The Thematic Apperception Test is a test that consist of twenty cards with one being blank, each card is with an ambiguous drawing and you are then asked to tell a story about each one. This test tells the person that’s giving the test about your characteristics, personality and emotional functioning. The Rorschach test consists of ten cards of ambiguous images and your imagination. It begins with ten cards where the examiner shows you one image at a time. The examiner then ask you to describe the image you see. The examiner then record your tone of voice, reactions and your responses.

Objectives of personality assessment are tests used to restrict response format. Two examples are the Minnesota multiphasic personality inventory (MMPI) and the Rotten Incomplete Sentence Blank. The Minnesota multiphasic personality inventory which are true or false questions where you have forty incomplete sentences or question and you can either tick true or false. The other example is the Rotten Incomplete Sentence Blank where you are given a sentence with at least two-three words and the person have to finish the sentence.

I think a teacher’s personality can indeed affect student’s development and performance. I think so because education is the key to a successful life. You get an education by attending school and applying what you learn. In schools you meet teachers and a teacher knowledge and personality will in fact determine how a child cope with their studies. A teacher that is passionate about her job will make learning so much easier and fun for student. It is important that a teacher motivates and help her students, after all, school is where they spend most of their time. I think teachers should be role models to their students, they have a big impact on their students. If a teacher goes to class unprepared and with a negative attitude, her students will return the favor, they would not have the right attitude to learn which will definitely affect their performance, the teacher can either hurt or help her student’s well-being, motivation and achievement, her attitude can cause her students to be stressed. I think when a teacher expects good results from her students, she gets them and this means teachers are supposed to encourage their students rather than insult them. Students who are internally motivated to learn generally do better, you get better results.

literature review of personality type

Autobiography

This is me. Who am I? My name is Kerisha Shamika Harridass. I am the third of four children. I have a little girl of my own and she is a complete joy to be around and so my story begins. I never attended preschool, I went straight to primary school at the age of five. I was always a shy individual and ever since I know myself I have been an introvert type of person. I made a couple friends in primary school and ended up with one best friend where we went onto high school together.

I was the same person there, shy and would always stick to myself, didn’t have much friends. I ended up with a bad allergic reaction in form three where the majority of the students made fun of me, including the ones I thought were my friends. They would laugh, point at me, and call me names which damage me for a while because I didn’t want to go back to school. I stick to myself even more. Things changed as time went by, I had friends, still wasn’t the popular girl and I never wanted to be anyways. I graduated and moved on with my life.

At the age of nineteen, something happened that changed my life completely. I got pregnant, it was an experience that I wouldn’t wish on my worst enemy. During my entire pregnancy I would have a lot and I mean a lot of vivid dreams. I would wake up in pain, crying. I went into labor at seven months, my baby was alive, and he looked perfect and healthy to me. He was supposed to be in the hospital for a week, which lead to four entire months. It was the worst four months of my life, he did surgeries back to back, wires all over him, tubes and he was always on IV fluids, never drank actual milk like other babies. The doctors said they did everything for him, yet still they couldn’t find the problem. He eventually died.

I went into a state of depression, it was so horrible that I gave up on prays, I literally hated him, that phrase of my life affected my personality so much. I lost so much of who I was, I was so isolated and in a dark place, this was happening for some time. I then decided that I would like to try again with life with pregnancy and so I did. I never really fully recovered from that experience but I was going to have a baby girl, so I had to be better for her. I am still that shy, introverted young lady I was at primary school. Knowing what I know now about people and about my life, I’m more protective of myself and the people I love more than ever. I love who I am, I love my personality and I love everything about me, so I would say that my very first pregnancy has a lot to do with my personality development, it most definitely changed me.

Five strategies I will use to remain grounded or self-aware as I teach are by keeping a personal journal. It would allow me to see how I have matured and grown throughout the year. It would help me to keep track of my feelings that I was feeling at that particular day, it would also allow me to go back on certain events that happened in my life whether good or bad. It allows me to reflect on past experiences and how I dealt with it. Keeping a journal is a great way to remain self-aware and a great way to keep track of my memories.

Secondly, writing down five things I’m grateful for each day or a list of my most important tasks regularly. By writing down five things that I’m grateful for each day can make me feel so much better whether I’m in a good or bad mood. Those five things can change every day or it can remain the same but the goal is to come up with five different things each day. At the end of the week it will make me realize how many things I’m grateful for in life.

Thirdly, writing down my most important task regularly at the beginning of each day. It would help me work towards something when the day comes, so at the ending of the day I would know for sure I have made progress and that I have accomplished my daily task.

Taking a morning walk, it can help me relax, it gives me time to think, gather my thoughts and feelings. It can be a good morning exercise before starting my day. It gives me time to myself, I’ll be full of energy and in a great mood, that way I’ll want to teacher and it’s a great way to begin the long day ahead.

Lastly, I can read books, I love to read. I find it very relaxing. Reading helps improve our vocabulary, makes us smarter, keep our minds active and we can gain more knowledge. I find it very entertaining but it all depends on what I’m reading. It helps you keep up with what’s going on around you. I strongly think those five strategies would help me remain grounded and self-aware as an early childhood educator.

Our writers will provide you with an essay sample written from scratch: any topic, any deadline, any instructions.

Cite this Page

Get your paper done in as fast as 3 hours, 24/7.

Related essay Topics

Popular Categories

Most Popular Essays

Big Five Personality Traits

Introduction Personality development is the development of the organized pattern of behavior and attitudes that makes a person distinctive. Personality development occurs by the ongoing interaction of temperament, character and environment. One of the most important measures of personality created is the Big Five Inventory. The big five was originally derived in the 1970s by two independent research teams – Paul Costa and Robert McCrae. These five dimensions were derived by asking thousands of people questions and then analyzing the...

Academic Performance

In the society we live in today getting a degree or being educated is essential to go up in the socio-economic hierarchy.This research paper is a study on the effect of people’s personalities on their academic performance. Personality is a set of personal traits that make up a person’s behavioral pattern. Academic performance is how we measure our individual achievements in various academic subjects. There are five types of personality traits; Openness,Conscientious,Extraversion,Agreeableness,Neuroticism (Pappas,S,201). In the pursuit of academic excellence different...

Personality

Abstract In this research it includes that how personality varies at each person and with the same person how it differs as per the different situation. It is also mentioned that how people sometime behave in such manner they do, the way they do and for what reason. Also on what basis we judge people and their personality at what extent of level. It is likely suggest that how we can judge each person’s personality and at the same time...

Personality factors contribute to many aspects of human life, including the determinants of those who are likely to be ambitious in their careers. These personality factors which play a role in this topic can be observed in the context of two theories of personality; the social cognitive theory as well as the Big Five trait theory. The idea of ambitious careers has changed in the last few years. Ambitious careers tended to be thought of as careers such as doctors,...

Introduction These assessments helped me understand what kind of leader I want to be and have given me a plan to improve my leadership skills. I have developed an idea of the potential my leadership has in my professional and personal life. With the help of leadership, we can uplift, inspire, and, most of all, help others realize their potential. With a lot of hard-working and patience, it is a skill that will improve over time by learning from more...

Organisational behavior is a huge concept which highly focuses on human behavior and performance at work place. Personality is an object which plays the most important role when working with others. Its development helps an individual to develop a positive attitude towards work and improve not only outer, but also the inner self. Looking further at the broad theory of personality, lets look over the Big Five dimensions which differentiate different people in terms of organisational behavior. Ability to solve...

Introduction Entrepreneurship brings forth economic freedom in its ability to allow individuals to choose how they would like to make the money needed to support themselves and their family members. Benefits of entrepreneurship include but are not limited to, providing a means for boosting the economy and reducing unemployment rates by creating employment opportunities and utilizing innovative technological processes. (Yang & Ai, 2019, p. 2) However, entrepreneurship is not for everyone and it takes a certain kind of individual to...

Abstract Self-Determination Theory (SDT) posits that as humans we constantly seek mastery and growth within our lives, and that this can only be achieved by the fulfilment of certain psychological needs which entail relatedness, autonomy and competence (Deci & Ryan, 2000). Our personality has oft been suspected to impact these results, with particular focus as to whether certain ‘personality types’ have more or less influence in their psychological wellbeing than others. The generally accepted system of classifying personality types is...

INTRODUCTION The topic of research moves around the detail analysis of correlation between Irrational belief, Aggression and Big five personality traits. Whenever irrational believes comes in spotlight of discussion, attention draws toward the consequences after thinking irrationally thus aggression comes as one of its consequences whereas irrational thinking is found to be a dominant trait of personality thus it is also an interesting thing to find out if any personality trait is correlated with either irrational belief or aggression. This...

Fair Use Policy

EduBirdie considers academic integrity to be the essential part of the learning process and does not support any violation of the academic standards. Should you have any questions regarding our Fair Use Policy or become aware of any violations, please do not hesitate to contact us via [email protected]

We are here 24/7 to write your paper in as fast as 3 hours.

IMAGES

Personality types by music and literature choices?
√ What's My Personality
Personality type
(PDF) A Systematic Literature Review of Personality Trait Classification from Textual Content
Literature and the Cult of Personality by Gregory Maertz · OverDrive: ebooks, audiobooks, and
Writing a Literature Review

VIDEO

My two personality traits #books #kpop
அண்ணன் start சொல்றதுக்கு முன்னாடி என்ன பண்றிங்க அக்கா @Barbie Remo #shorts
Chapter 1: PERSONALITY (Part 1)
TYPE A PERSONALITY
Why Personality Type Matters at Work
The Study of Personality

COMMENTS

Personality types revisited–a literature-informed and data-driven
A new algorithmic approach to personality prototyping based on Big Five traits was applied to a large representative and longitudinal German
A Systematic Literature Review of Personality Trait Classification
(2) After investigating the gradual research regardingthe field of personality classification, the need of asystematic literature survey is
A Literature Review on Personality, Creativity and Innovative Behavior
What aspects of innovative behavior and personality traits need to be discussed further? So this study will review past studies on concept of personality;.
Personality: bridging the literatures from human psychology ...
Personality in the behavioural ecology (henceforth, BE) literature refers to consistent differences in behaviour among individuals from the same species or
Personality types revisited–a literature-informed and data ...
Reliability and construct validity could be confirmed. We discuss wether personality types could comprise a bridge between personality and clinical psychology
Literature review on relationship between Personality traits and
There are many organizational parameters like job performance, job satisfaction, leadership qualities etc. where personality traits play a
(PDF) Personality : A Review of Related Literature along with
According to Eyesenck (1971), personality is more or less stable and enduring organization of a person's character, temperament,. intellect
11 CHAPTER II REVIEW OF RELATED LITERATURE A. Theoretical
Big five personality traits also known as five factor model. (FFM) is a set of five broad trait dimensions or domains to describe the one's traits of
Million dollar personality: a systematic literature review on
Expressed personality traits can play a pivotal role in convincing investors in crowdfunding. Our study answers the research question: What
The Big Five Personality Traits: Review Of Literature And Reflection
A Review of Literature ... Personality traits are a reflection of who you are as an individual, it give others an idea of your characteristics. The big five