Content uploaded by Joeri Hofmans
Author content
All content in this area was uploaded by Joeri Hofmans on Aug 19, 2020
Content may be subject to copyright.
1
Person-centered methods in vocational research
Joeri Hofmans1, Bart Wille2, Bert Schreurs1
1 Vrije Universiteit Brussel, Belgium
2 Ghent University, Belgium
Accepted version of paper in press at Journal of Vocational Behavior. This paper is not the copy of
record and may not exactly replicate the final, authoritative version of the article. Date of acceptance:
February 6, 2020
Correspondence concerning this article should be addressed to Joeri Hofmans, Research group of
Work and Organizational Psychology, Vrije Universiteit Brussel, Belgium, Pleinlaan 2, 1050 Brussel
Email: joeri.hofmans@vub.be
2
Abstract
The vast majority of vocational research adopts a variable-centered approach. Implicit in this
approach is the assumption that the population under study is homogeneous, and that therefore we
can use a set of “averaged” parameters to describe it. Person-centered methods are a family of
methods that relax this assumption of population homogeneity, viewing the individual as holistic and
paying more attention to how specific configurations of variables, present in different subgroups of
the population, act in concert to shape behavior. Despite the potential advantages of person-centered
research, the adoption of this approach by vocational researchers has been relatively slow for both
conceptual (e.g., What exactly is person-centered research?) and methodological (e.g., Which
methods?) reasons. In response to these issues, the goal of the present article is to showcase the role
and relevance of person-centered methods for vocational research. Having discussed different
conceptualizations of the term “person-centered” we present a structured overview of the most
relevant person-centered techniques. This overview includes a description of the formal
characteristics of each technique, as well as an overview of existing applications of these techniques
in the literature. Finally, we provide a balanced discussion of both the advantages and challenges
associated with the person-centered approach.
3
In vocational research, the vast majority of studies examines relationships among variables
across individuals. Such studies have, for example, shown that charismatic personality relates
positively to career outcomes 15 years later (Vergauwe, Wille, Hofmans, & De Fruyt, 2017), or that
employment self-efficacy is positively related to job search intensity at the between-person level,
while the relation is negative at the within-person level (da Motta Veiga & Turban, 2018). Implicit in
such studies is the assumption that the population can be described by a single set of “averaged”
parameters (Morin, Bujacz, & Gagné, 2018).
Despite the prevalence of this assumption, career theories suggest that describing an entire
population using a single set of parameter estimates most likely oversimplifies reality. For example,
vocational researchers are increasingly recognizing that contemporary careers and career orientations
cannot simply be categorized as either boundaryless or protean, but that people differ in the extent to
which they hold unique combinations of different career orientations. As a result, there have been
repeated calls for the integration rather than the separation of different career orientations (Kuron,
Schweitzer, Lyons, & Ng, 2016). Also in the counseling field the idea of population homogeneity is
called into question by for example research that shows that problems associated with career
indecision are manifested in very different ways for different groups of people (e.g., Cohen,
Chartrand, & Jowdy, 1995). Such findings have led to the awareness in the counseling literature that
phenomena can only rarely be explained by a universal relationship between a small number of
variables (Frankfurt, Frazier, Syed, & Jung, 2016). Although variable-centered methods do allow
studying the interplay of variables through the inclusion of interaction terms, this quickly becomes
impractical when the number of interacting variables increases. Because of the awareness that a
single set of parameter estimates cannot be assumed to hold for a whole population and because of
the limitations of variable-centered methods to capture complex patterns of interactions,
researchers called for supplementing the body of variable-centered research with person-centered
research (e.g., Morin et al., 2018).
4
In person-centered research, the focus is no longer (exclusively) on relations between
variables, but (also) on relations among people (Zyphur, 2009). To fulfill this task, person-centered
methods do not assume population homogeneity, but model unobserved heterogeneity within the
population (Woo, Jebb, Tay, & Parrigon, 2018). Thus, person-centered methods shift the attention
away from a focus on variables to a focus on individuals by allowing the study population to be
heterogeneous (Weiss & Rupp, 2011). By doing so, person-centered methods pay more attention to
how specific configurations of variables act in concert to shape behavior (Bergman & Trost, 2006).
Despite the potential advantages of person-centered research, vocational behavior and
counseling researchers have been rather slow in adopting this approach due to conceptual (e.g., What
exactly is person-centered research?) and methodological reasons (e.g., Which methods can be used
when performing person-centered research? ). In response to these issues, the goal of the present
article is to showcase the role and relevance of person-centered research for vocational behavior and
career counseling (hereafter referred to as vocational research). More specifically, the aims of this
article are fourfold. First, we aim to improve the understanding of person-centered research and how
it is applied in vocational research. To this end, we first discuss different conceptualizations of the
term “person-centered” and differentiate it from related, yet different approaches. Based on this
outline, we then present an overview of the most relevant techniques within this approach, including
k-means and hierarchical clustering, latent profile and latent class analysis, factor mixture analysis,
mixture regression analysis, configural frequency analysis, Davison and Davenport’s (2002)
criterion-based method, latent class growth modeling (and growth mixture modeling), and latent
transition analysis. As a second contribution, we describe the state-of-the art of person-centered
approaches to vocational research by taking stock of the literature in this area. Our literature review
covers seven of the most well-established international peer-reviewed journals listed in Web of
Science that in their objectives specifically focus on careers, career counseling, and vocational
behavior: Journal of Vocational Behavior, Career Development International, Career Development
5
Quarterly, Journal of Career Development, Journal of Career Assessment, Journal of Counseling
Psychology, and The Counseling Psychologist. The results of this literature search are presented
alongside the formal description of each technique, highlighting the most important trends in
vocational research using those methods. Third, apart from describing the methods themselves and
from summarizing existing research with those methods, we highlight how those person-centered
techniques can potentially be used for advancing vocational research. Finally, a fourth objective of
this article is to provide a balanced discussion of both the advantages and challenges associated with
person-centered research.
Person-Centered Methods: Different Conceptualizations and Approaches
Although the ability of person-centered methods to account for unobserved population
heterogeneity is a key feature that distinguishes them from variable-centered methods, it is important
to note that traditional variable-centered techniques can also deal with observed heterogeneity.
Observed heterogeneity occurs when different subpopulations can be differentiated based on an
observed variable (e.g., age, occupational category). In this case, the subpopulations are referred to
as groups and traditional multi-group analytic techniques such as t-tests, ANOVA, MANOVA and
multi-group SEM can be used to test between-group differences (or heterogeneity) on the outcomes
of interest. Often, however, the variables that cause population heterogeneity are not known
beforehand and/or not observed. If this happens, heterogeneity is due to unknown reasons, which is
why this type of heterogeneity is referred to as unobserved heterogeneity and why we speak about
latent classes rather than groups. Because in this scenario it not possible to a priori divide the sample
into groups, traditional analytic techniques are of little use. It is in this particular situation that
person-centered methods, with their ability to infer subpopulation membership from the data, are
particularly useful.
Although the differentiation of variable- and person-centered methods in terms of their
treatment of observed versus unobserved heterogeneity is relatively straightforward, the term
6
“person-centered” has led to some confusion in previous writings (see Woo et al., 2018 for an
excellent treatment of these issues). According to Woo and colleagues (2018), three
conceptualizations of the term “person-centered” can be discerned in the scientific literature. First,
some researchers refer to person-centered studies as research on the characteristics of individuals (as
opposed to research on the characteristics of situations). According to this perspective, a study is
person-centered when it focuses on characteristics of people, such as personality, skills, or ability.
Second, others have used the term person-centered to refer to research that focuses on the
subjectivity of worker experiences, as opposed to research that focuses on more objective
characteristics of individuals (Weiss & Rupp, 2011). Finally, according to the third
conceptualization, the term person-centered is used to refer to a collection of methods that classify
individuals on the basis of the similarity in their scores on a set of variables (Howard & Hoffman,
2018). The approach aligns well with the idea of studying persons based on certain profiles across
multiple variables or characteristics (see further). This third conceptualization is the one that
“maximizes the level of precision in methodological discussions …” (Woo et al., 2018; p. 816).
Because of this reason, we delve a bit deeper into this conceptualization.
Using the conceptualization of person-centered research as research that clusters individuals,
such methods have been argued to be characterized by three features (Morin et al., 2018). The first
feature is that they are typological in the sense that they use a classification system that categorizes
individuals into qualitatively and quantitatively distinct subpopulations, with each of the
subpopulations being characterized by different sets of model parameters. The typological nature of
such methods is very appealing to vocational researchers since the classification system implied by
person-centered models corresponds to a way of thinking often used by managers (i.e., thinking
about employees by categorizing them in types of employees) (Morin et al., 2018) and by counseling
psychologists, who tailor their treatment based on the type of employee they have in front of them
(Cohen et al., 1995).
7
Second, person-centered research is often said to be prototypical. This means that each
individual in the sample belongs to each of the estimated profiles with a certain probability. This
probability is based on the extent to which the individual’s unique configuration of scores on the
study variables resembles the profile’s specific configuration of scores. In such probabilistic
scenario, individuals are not assigned to one of the profiles, but are assessed as being more or less
similar to each of the prototypical profiles. Accounting for the uncertainty in assignment by using
probabilistic memberships offers a way to account for the fact that the classification of individuals
into unobserved subpopulations is not without error. Although prototypicality is undisputedly a key
feature of most person-centered methods, some methods use ‘definite’ (or hard) assignment,
implying that each individual is assigned to one and only one profile. Such ‘hard clustering’ happens
in the large majority of the cluster analytic models, some of which will be discussed below.
Third, person-centered models are exploratory. Because of a lack of goodness-of-fit
information that allows for a direct assessment of the adequacy of the tested model(s), the ‘final’
model is typically obtained by comparing solutions with different numbers of profiles or clusters.
Moreover, and similar to what happens in exploratory factor analysis, in person-centered methods
the relations between the profiles and indicators are typically freely estimated (Morin, McLarnon, &
Litalien, in press). It is important to note that these methodological peculiarities do not imply that
person-centered models cannot be used for confirmatory purposes, an issue that will be elaborated on
in the Discussion section.
It is important to note that there are a number of methods that, while they are strictly
speaking not encompassed by this definition, fit the goal of person-centered approaches because they
are aimed at studying profiles or patterns of scores (e.g., Asendorpf, 2006; Davison & Davenport,
2002; von Eye, 2002). That is, whereas these methods are not classification-based, they do focus on
the pattern of scores of individuals and therefore they do consider the person as an “organized
whole” (Bergman & Magnusson, 1997, p. 291). Because of this reason, some authors consider them
8
to be person-centered (e.g., Asendorpf, 2006). Because the goal of non-typological methods closely
relates to the central goal of the “typical” person-centered approaches (i.e., moving the focus away
from studying relations between variables to studying relations between people on the study
variables), we will include two of such methods in our overview (i.e., configural frequency analysis
and Davison and Davenport’s (2002) criterion-based method).
As we mentioned above, person-centered methods relax the assumption of population
homogeneity by clustering individuals in subgroups or subpopulations. One might relax this
assumption even further, in which case inferences are made for each individual separately. Such an
approach, which in the context of Cattell’s (1952) data box has been described as the P-technique, is
referred to as an idiographic or person-specific approach. In person-specific analyses, the goal is to
build a model for each individual separately, drawing on the idea that each individual can best be
described and understood using an individualized model (Howard & Hoffman, 2018). The
philosophical differences underlying the person-specific and the person-centered approach clearly
show in the data they use as input. Unlike person-specific methods, which operate on occasions ×
variables matrices, person-centered methods operate on persons × variables matrices (except for
longitudinal person-centered methods, which use the full persons × variables × occasions data box).
In other words, whereas person-specific methods by definition analyze intra-individual variation
1
,
person-centered methods can work with both inter-individual and/or intra-individual variation (Woo
et al., 2018). In that sense, person-centered analyses offer a compromise between the parsimony of
the variable-centered approach, yielding a single set of parameters, and the person-specific approach,
yielding a set of parameters for each individual in the sample.
Following this discussion of definitional issues, in the next section, we offer an overview of
several methods that move the focus away from studying relations between variables to studying
1
Some person-specific techniques, such as dynamic factor analysis (Molenaar, 1985) and dynamic structural equation
modeling (Asparouhov, Hamaker, & Muthén, 2018) allow for the consideration of between-person variation. Those
techniques thus also operate on the full persons × variables × occasions data box.
9
relations between people. Three types of methods are discussed. First, we review methods that seek
to identify subpopulations based on their profile of scores (i.e., cluster analysis, latent class and latent
profile analysis, and factor mixture analysis). In all of those methods the profiles are based on scores
on a set of variables, without taking into account potential outcome variable(s). The second category
of models addresses the issue of a lack of criterion variables by making explicit reference to such
variable. That is, in this category of models, subpopulations are either made based on differential
relationships between a set of predictors and an outcome variable (i.e., mixture regression analysis),
or the models identify specific patterns of predictors that are associated with the criterion variable
(e.g., Davison & Davenport, 2002). Finally, and in line with recent calls for more longitudinal
studies in organizational research in general (e.g., Vantilborgh, Hofmans, & Judge, 2018), and
vocational research in particular (e.g., Zacher, Rudolph, Todorovic, & Ammann, 2019), we also pay
attention to longitudinal person-centered models, being growth mixture modeling and latent
transition analysis. This last category of models is particularly interesting as such methods allow
studying interindividual differences in intraindividual change processes (Ram & Grimm, 2007).
From Studying Variables to Studying People: An Overview of Person-Centered Methods
Modeling profiles of scores
We will review four internal person-centered methods (i.e., cluster analysis, latent class
analysis, latent profile analysis, and factor mixture analysis). The preposition “internal” refers to the
fact that in those analyses, only “internal variables” matter for profile estimation. “External
variables”—including predictors, outcomes and/or covariates of the profiles—can be included in the
analysis, but those variables do not directly contribute to the definition of the profiles. They are
instead used to provide validity evidence for the profiles obtained
2
.
2
Note that for LCA and LPA, direct inclusion of predictors, outcomes and/or covariates of profile membership can be
done within the Generalized Structural Equations Modeling framework (GSEM) (see Morin et al., 2019).
10
Cluster Analysis. Cluster analysis pertains to a family of methods aimed at dividing objects
into a limited number of mutually exclusive groups (or clusters), in such way that objects belonging
to one cluster are more similar to each other than objects belonging to another cluster
3
. In general,
there are two broad cluster analytic approaches: hierarchical clustering and nonhierarchical
clustering.
In hierarchical clustering, one seeks to build a hierarchy of clusters, and this hierarchy can be
either built bottom-up (the agglomerative approach) or top-down (the divisive approach). In the
bottom-up approach, one starts from a cluster solution in which each object has its own cluster, after
which pairs of clusters are merged when moving up the hierarchy. In the top-down approach, the
initial cluster solution is one in which all objects are grouped into one cluster, after which splits are
performed while moving down the hierarchy. Thus, the idea in the bottom-up approach is to merge
pairs of clusters that are most similar to one another, while in the top-down approach one splits those
clusters that are most dissimilar to one another. An important question in hierarchical clustering
pertains to the measurement of (dis)similarity of clusters. There are several ways to measure
(dis)similarity, including the nearest neighbor (or single linkage) method, the furthest neighbor (or
complete linkage) method, and the average linkage method.
Nonhierarchical clustering, as opposed to hierarchical clustering, is not aimed at building a
hierarchy of clusters. Instead, its aim is to cluster objects into a pre-defined number of clusters by
maximizing or minimizing some criterion. Arguably the most popular type of nonhierarchical
clustering is the K-means method (Hofmans, Ceulemans, Steinley, & Van Mechelen, 2015). In K-
means clustering, the algorithm searches for a combination of a binary partitioning matrix
(containing the memberships of the I objects to the K clusters) and a centroids matrix
(containing the centroids for the K clusters) that minimizes the following least squares loss function
(with being the data matrix in which I objects are measured on J variables):
3
Note that cluster analysis can also cluster variables into groups based on their values on a set of objects.
11
(1)
Formula 1 implies that the K-means algorithm assigns objects to the cluster for which the distance to
the cluster centroid (i.e., the center of the cluster) is minimal. Readers interested in learning more
about K-means clustering can consult the overview paper of Steinley (2006), while Kaufman and
Rousseeuw (2005) provide an excellent treatment of cluster analysis in general.
Our literature review reveals that cluster analysis has a relatively long history in vocational
research, with the first studies using this technique already dating back more than half of a century
(e.g., Matthews & Tiedeman, 1964). Today, the technique is still commonly used to group
individuals and/or career events based on shared/similar features (see Table A1 for an overview) .
Latent Class Analysis (LCA) and Latent Profile Analysis (LPA). The goal of LCA and
LPA is to identify subpopulations of people, with those subpopulations being characterized by
distinct configurations of scores on a set of variables (see Figure 1). In that sense, LCA and LPA are
similar to cluster analysis. However, unlike cluster analysis, LCA and LPA (1) are model-based (i.e.,
they are based on a formal model instead of (dis)similarity measures) and (2) are prototypical, which
means that they yield probabilistic, rather than hard, assignment [note that some clustering methods,
such as fuzzy clustering (see Tan, Steinbach, Karpatne, & Kumar, 2019) also yield probabilistic
memberships].
Although the terms LCA and LPA are often used interchangeably, the difference is that LCA
uses categorical indicators, while in LPA the indicators are continuous. More formally, the
traditional latent class analysis (LCA) model can be expressed as follows:
(2)
In this formula, represents a specific response pattern (or a pattern of scores on J categorical
variables). The chance of observing this particular response profile is a function of the probability of
membership to the k latent classes (the ’s), and the probability of observing each response
12
conditional on latent class membership (the ’s). The indicator function
equals 1 when
the response to variable j equals
. If not, this indicator function is 0 (see Collins & Lanza, 2010 for
an in-depth overview of the technicalities of LCA).
As opposed to LCA, in LPA the latent variable indicators are continuous. Assuming that
these indicators are normally distributed within each latent profile, that the indicators are unrelated
within each latent profile (i.e., local independence), and that the indicator variances are equivalent
across the latent profiles (i.e., homogeneity), LPA models the distribution of observed scores on a set
of indicators ( ) as a function of the probability of membership to the K latent classes
(the ’s) and each class’ normal density
(with each class having a class-specific mean
vector and covariance matrix ):
(3)
In a more generic form, the LPA model decomposes the variance of each indicator i into two
components (see Formula 4): a between-profile component that captures how far the profile-specific
means are from the general mean (i.e.,
) and a within-profile component
containing the profile-specific variances
(i.e.,
). In both the between- and the within-
profile component, denotes the density parameter, or the probability of membership to profile k.
(4)
Although LPA can thus be used to estimate profiles differing in both means and variances,
more constrained versions in which only the means are profile-specific can also be tested (i.e.,
; Peugh & Fan, 2013). This assumption of homogeneity of variances is shared with methods such
as K-means clustering and is the default parameterization in some statistical packages, such as Mplus
(Muthén, & Muthén, 2017). Readers interested in a more in-depth discussion of LPA can consult the
book chapter by Masyn (2013) or the paper by Sterba (2013).
Our review indicates that the application of LCA and LPA in vocational research took off
around 2012-2013, with the first studies using these techniques to identify subgroups of people based
13
on their commitment profiles (e.g., Meyer, Stanley, & Parfyonova, 2012). Since then, LCA and LPA
have been adopted widely with the aim to study among other things interest profiles, motivation
profiles, and profiles of work characteristics (see tables A2 and A3 for LCA and LPA, respectively).
Factor Mixture Analysis (FMA). Whereas in LCA and LPA unobserved heterogeneity is
modeled through the inclusion of a categorical latent variable, FMA simultaneously includes a latent
categorical and one or multiple latent continuous variables within the same model (see Figure 2).
The latent categorical variable allows for the classification of individuals in groups, whereas the
latent dimensional variable(s) allow for heterogeneity within groups by modeling covariation
between observed variables within each class. Hence, FMA relaxes the conditional independence
assumption of classical LPA analyses (Lubke & Muthén, 2005). This is of particular importance to
vocational research, where the assumption of conditional independence is often unlikely due to a
global factor underlying the different indicators (e.g., in commitment research; Morin, Morizot,
Boudrias, & Madore, 2011). Moreover, because in FMA the continuous latent variable controls for
variance shared across all indicators when estimating latent profiles, FMA may result in profiles with
clearer shape differences (Morin & Marsh, 2015). Finally, by combining latent continuous and latent
categorical variables within the same model, FMA can tell us something about the underlying
continuous and categorical nature of psychological constructs (Clark et al., 2013). Formally, the
FMA model is expressed as follows (see Lubke & Muthén, 2005 for a thorough treatment of FMA):
(5)
(6)
Scores on indicator variable are expressed as a function of the regression intercept , the
regression slope or factor loading and the residual . Factor scores are denoted as . All
parameters in Formula 5 have subscript k, implying that they may vary across classes. Formula 6
shows that the factor scores are a function of the latent class variable , an intercept vector A, and
the residual factor scores .
14
Our review demonstrates that FMA has been applied scarcely in careers research. The studies
that were identified adopted FMA for the categorization of reward patterns, or for creating subgroups
based on stereotype sensitivity, or vocational interests (see Table A4).
The potential of modeling profiles of scores for vocational research. Cluster analysis,
LPA, LCA and FMA can be used to address a wide variety of questions in vocational research. First
of all, many of the constructs being studied in this domain are multidimensional. This for example
holds true for predictor variables such as personality and interests, but also for outcomes such as
performance, career success or commitment. Person-centered techniques such as cluster analysis,
LPA, LCA or FMA allow studying how those different characteristics combine into profiles. Such
insight is important because, by showing which profiles emerge and how frequent those profiles are,
these methods contribute to a better understanding of the psychological makeup of individuals. This
is crucial for vocational research, where a basic principle is that vocational behavior and attitudes
result from the unique interplay or patterning of a broad set of different characteristics. Moreover, it
aligns well with the increasing individualization of career development (Vondracek & Porfeli, 2002)
and the person-focused approach used in career counseling.
An important remark is that the techniques that allow for the modeling of scores do not
require the different scores to tap into one overall dimension (Morin et al., 2019). They can also be
used when studying profiles across a collection of variables of interest. For example, Haines, Doray-
Demers, and Martin (2018) performed LCA with the goal to develop a typology of part-time
employment on the basis of work characteristics and role occupancy. To this end, they included a
wide range of variables into their LCA, including having a partner or not, having children or not,
household income distribution, educational requirements of the part-time position and work hours.
Or in the counseling domain, Hirschi and Valero (2017) used LPA to identify five differing profiles
according to levels of perceived chance events and career decidedness.
Such endeavors have the potential to advance the career and counseling field in various ways
15
(Borgen & Barnett, 1987). First, they allow exploring the identification and structure of subgroups,
which might help in understanding the research problem better. For example, drawing on the idea
that the traditional career is declining, Gerber, Wittekind, Grote, and Staffelbach (2009) used LCA to
explore the nature and prevalence of different types of career orientation. Second, these techniques
can be used to challenge or confirm existing classifications. For example, the four-class solution by
Haines and colleagues (2018) revealed that qualifying part-time work in good and bad is too
reductionist, and that a more complex classification is warranted. Third, these techniques allow
simplifying complex datasets. For example, Ferguson and Hull (2019) identified profiles of science
career interests based on scores on science motivation, attitude, interest, and academic experiences.
Modeling predictor-outcome profiles
The methods we have reviewed up until now are all ‘internal techniques’, meaning that the
profiles are derived without taking into consideration their predictive value for outcome variable(s)
(Davison & Davenport, 2002). The second category of models addresses this issue. That is, in this
category, some models use subpopulations to capture differential relations between a set of
predictors and an outcome variable (i.e., mixture regression analysis), while others look for specific
patterns of predictors that are uniquely associated to the criterion variable [i.e., configural frequency
analysis and Davison and Davenport’s (2002) criterion-based method].
Mixture Regression Analysis (MRM). The subpopulations in mixture regression analysis
(MRM) differ from each other in the relationships between the constructs of interest. Similar to
traditional multiple regression, in MRM a criterion variable is regressed on a set of predictors. The
major difference, however, is that subpopulations are identified for whom the predictor(s)–criterion
relationship is different (Brusco, Cradit, Steinley, & Fox, 2008). In that sense, the latent categorical
variable in MRM can be thought of as an unobserved moderator of the relation between the predictor
and the criterion (see Figure 3). For example, in their study on reward satisfaction, Hofmans, De
Gieter, and Pepermans (2013) found two subpopulations with a different pattern of job reward–job
16
satisfaction relationships. For the first type, job satisfaction related to financial and psychological
reward satisfaction, whereas for the second type it related to psychological reward satisfaction only.
Formally, the MRM model can be expressed as follows:
(7)
In Formula 7, the latent subpopulations are represented by a latent categorical variable C, where C =
1,2,3,…K. Hence, represents the intercept for subpopulation (or class) k, while and
represent the regression coefficient and error term for this subpopulation. Similar to the traditional
regression model, more than one predictor variable can be included, in which case each of the
predictor variables has a class-specific regression coefficient. Moreover, as in traditional regression
models, the errors are assumed to be multivariate normal with a mean of zero and a class-specific
variance (i.e.,
). Readers interested in a more technical treatment of MRM can consult
Wedel and DeSarbo (1995).
Our literature review demonstrated that MRM has only seldom been used in vocational
behavior research (see Table A5).
Configural Frequency Analysis (CFA). The aim of Configural Frequency Analysis (CFA)
is to identify whether specific configurations or response patterns are more likely to be associated
with specific criterion groups (von Eye, 1990). This method is developed for the analysis of
categorical predictors and outcomes and draws on an analysis of frequencies in multi-way
contingency tables. In such multi-way contingency tables, individuals are categorized in disjunct
categories based on their unique profile (or configuration of scores) on the study variables. For
example, when one has two dichotomous predictor variables and one trichotomous outcome,
participants can belong to one of 2 × 2 × 3 = 12 unique profiles. After having tabulated those unique
profiles and their frequencies, the crucial test is in the comparison of the observed with the expected
frequencies of those configurations. In case n individuals are being measured on i = 1, 2, …, m
17
dichotomous variables, and assuming that all variables are independent, the expected frequencies of
a specific configuration c can be calculated as follows:
(8)
with being the probability for a member of the population to have a value of 0 on variable i,
and being the probability for a member of the population to have a value of 1 on variable i. To
test whether these expected frequencies (assuming independence of all variables) differ significantly
from the observed frequencies (denoted by ), the following -statistic can be calculated:
(9)
In case a configuration is significantly more often observed than expected, it is referred to as
a type, whereas an antitype refers to the case where a configuration is significantly less often
observed than expected. Although we have presented the default version of CFA, the -test can be
replaced by other tests (e.g., von Eye, 2002) and the expected frequencies can be calculated using
another model than the independence model (e.g., von Eye, 1990). Finally, because in CFA one
performs a (-)test to each profile or configuration, a Bonferroni-correction is often used to control
for Type I error inflation. An in-depth discussion of the technicalities of CFA can be found in the
books by von Eye (1990; 2002).
Our literature review revealed two empirical studies which used CFA in the context of
vocational research (see Table A6). First, Reitzle and Vondracek (2000) illustrated the usefulness of
this technique by identifying patterns of (categorical) career and family characteristics, including
marital status, completion of training, history of unemployment, etc. More recently, Moeller and
colleagues (2018) investigated the relationship between demands-resources profiles and engagement-
burnout profiles. For this purpose, they compared the proportions of three demands-resources
profiles in one of four engagement-burnout profiles, testing whether each profile combination was
more (or less) frequent than would be expected if the types of profiles were unrelated.
18
Davison and Davenport’s (2002) criterion-based method. Similar to mixture regression
analysis (MRM), Davison and Davenport’s (2002) criterion-based method draws on a multiple
regression-based model. Unlike MRM, however, it does not look for subpopulations with different
predictor–criterion relationships but tries to capture profile similarity as a continuous measure.
Moreover, the method has an even stronger focus on the criterion variable as its explicit goal is to
maximize the predictive value of the profile.
When performing Davison and Davenport’s (2002) criterion-based method, four steps are
taken. First, for each individual, a level score, or an average score across the predictor variables, is
calculated. Second, the criterion-related profile is identified. This is done by (1) predicting the
criterion variable from the predictor variables using multiple regression analysis, (2) calculating the
average unstandardized regression coefficient across all predictors, and (3) ipsatizing each regression
coefficient around the average unstandardized regression coefficient. This yields the criterion-related
profile, or a set of deviations around the average unstandardized regression coefficient. After having
calculated this criterion-related profile, one can calculate the profile fit score for each individual as
the average covariance between the individual’s predictor profile and the criterion-related profile.
Thus, rather than identifying different subgroups of profile scores, Davison and Davenport’s (2002)
criterion-based method treats the different profiles in a continuous manner, ranging from fit (i.e.,
high average covariance with the criterion-related profile) to misfit (i.e., low average covariance with
the criterion-related profile). Third, the level and (mis)fit scores are related to the criterion variable
using multiple regression analysis. This allows testing what percentage of the variance in the
criterion can be accounted for by level and profile effects. Finally, because the criterion-related
profile is obtained using multiple regression, and because the regression weights tend to capitalize on
the characteristics of the sample at hand, a crucial test is evaluating whether the level and profile
effects cross-validate. This can be done by splitting the data in half, after which one can estimate the
criterion-related profile using the first half of the data after which the level and (mis)fit scores can be
19
related to the criterion variable in the second half. Note that, although Davison and Davenport’s
(2002) criterion-based method is explicitly criterion-focused, it does not explain additional variance
in the criterion above and beyond a traditional multiple regression analysis. Instead, it separates the
predictor variance into level and profile effects, thereby providing insight into the usefulness of using
profiles or patterns in applied prediction.
Our literature review identified two articles using criterion profile/pattern analysis in the
context of vocational research (see Table A7). In a first application of this technique, Perry (2008)
investigated the effects of vocational exploration and racial identity on behavioral (attendance,
attention, time spent on class work) and psychological (identification with school) factors of school
engagement among urban youth of color. The criterion-based method revealed a predictive profile
marked by high levels of positive racial internalization and career planning combined with low levels
of racial dissonance. As a second application, Wiernik (2016) identified patterns in the predictive
relationships between personality traits and Realistic vocational interests. In two studies, he
demonstrated that one’s personality profile pattern, rather than the absolute levels of those traits,
drove the validity of personality traits in explaining Realistic vocational interest.
The potential of modeling predictor-outcome profiles for vocational research. The
usefulness of MRM for vocational research lies in the fact that in this domain people’s behaviors and
attitudes are thought to result from the unique interplay of a broad set of characteristics. For
example, in their overview of 100 years of research on career management and retirement, Wang and
Wanberg (2017) noted that career choices are impacted by among other things ability, personality
characteristics and biographical data such as socioeconomic status and parental involvement.
Similarly, De Vos, Van der Heijden, and Akkermans (in press) note that “careers form a complex
mosaic of objective experiences and subjective evaluations, resulting in an enormous diversity in
terms of how careers can take shape” and that “different levels of influential factors have to be taken
into account” to understand the nature of contemporary careers. Importantly, those factors are not
20
only assumed to have unique effects, but they interact in complicated manners. For example,
Chlosta, Patzelt, Klein, and Dormann (2012) demonstrated that the likelihood to become self-
employed depends on the unique interplay of the presence of parental role models and the person’s
score on trait openness. In such situation, where behaviors and attitudes are believed to result from
the unique interplay of multiple determinants, assuming that predictors relate to outcomes in the
same way for everyone is counterintuitive at best. Considering this complexity, methods for
modeling predictor-outcome profiles are very useful, particularly because those predictor-outcome
relations are likely to be affected by not one variable, but by a wide range of variables (and their
unique interplay), some of which are not known a priori.
The goal of configural frequency analysis and Davison and Davenport’s (2002) criterion-
based method are somewhat different from that of MRM in the sense that the former focuses on
capturing unobserved heterogeneity in predictor-outcome relations, whereas the latter techniques are
explicitly designed to test the predictive validity of patterns or profiles of predictor variables. Hence,
CFA and Davison and Davenport’s (2002) criterion-based method are well suited to test the idea that
people develop interests for jobs that align with their relative strengths (i.e., the peaks in their profile
of trait scores), rather than their absolute trait levels (Weirnik, 2016), or the hypothesis that it is the
particular patterning of specific job demands and resources that is predictive of burnout and
engagement, rather than the demands and resources as such (Moeller et al., 2018). In sum, those
techniques can be particularly helpful in expanding our knowledge on how specific patterns of
variables matter for vocational outcomes.
Modeling profiles of intraindividual change processes
In this last category of models, we review two longitudinal person-centered models, being
growth mixture modeling and latent transition analysis. Those models are well suited to model
stability and change over time, allowing for example for an assessment of the impact of important
transitions in employees’ lives (e.g., starting a job, promotion, retirement; see Solinger, van Olffen,
21
Roe, & Hofmans, 2013), or for analyzing the impact of the occurrence of critical events in
organizations (e.g., organizational change). Because of their ability to examine inter-individual
differences in intra-individual processes, these models are ideally placed as the analytical solution to
calls for more longitudinal, within-person research in vocational research (e.g., Zacher et al., 2019).
Growth Mixture Modeling (GMM). Growth Mixture Modeling (GMM) aims at identifying
subpopulations that follow different longitudinal growth trajectories over time, thereby being a
mixture extension of latent growth or latent curve models (see Bollen & Curran, 2006)
4
. In such
latent growth models, one or more variables is measured repeatedly and growth in the level of these
variables across time is estimated via random intercept and slope(s) factors. The random intercept
factor(s) capture each individual’s initial level on the repeated measures, while the random slope
factor(s) capture each individual’s change in those repeated measures as a function of time. At its
simplest, growth is characterized by a random intercept and a random linear slope factor, although
more complicated growth trajectories can be modeled by adding additional, higher-order slope
factors (e.g., quadratic, cubic, ….).
GMM, being a mixture extension of the latent growth model, aims to identify subpopulations
following different growth trajectories over time (see Figure 4). In this sense, GMM is similar to
multi-group growth curve modeling, where different growth models are tested for each group.
However, unlike in multi-group growth curve modeling, where the groups are observed, in GMM the
grouping variable is latent or unobserved (Ram & Grimm, 2009). In its simplest form, the latent
subpopulations are only allowed to differ regarding their average level on the growth factors.
However, more complex GMMs can also be estimated, with the subpopulations being allowed to
vary not only on intercept and slope(s) averages, but also intercept and slope(s) variances and
4
Latent Class Growth Analysis (LCGA) is a special case of GMM in which the variances and covariances of the growth
factors in each latent class are fixed to zero (see e.g., Jung & Wickrama, 2008).
22
covariances, and even time-specific residuals. Moreover, the subpopulations can also be allowed to
follow a different functional form. Formally, a general linear GMM can be expressed as:
(10)
(11)
(12)
In Formula 10, —or the level of variable y for person i at time t— is a function of (1) the
profile-specific random intercepts , linear slopes , and error terms (with k = 1, 2, …, K
being the latent profiles), and (2) the probability of belonging to each of the latent subpopulation or
profiles, (with all and
=1). In other words, the raw repeated measures data for
each individual are conceived of as a mixture (i.e., a weighted sum) of the K different latent growth
profiles. Time in formula 10 is represented by , being the factor loading matrix relating the
repeated measures of y to the slope factor. In GMM, should be coded in such way that it reflects
the interval between measurement occasions (for example λ1 = 0, λ2 = 1, λ3 = 2, λ4 = 3 with four
equally spaced measures or λ1 = 0, λ2 = 1, λ3 = 1.5, λ4 = 2 in case the measurement one and two are
separated by a period double the period separating measurement two and three and three and four).
More information on the technicalities involved in defining the time codes can be found in Biesanz,
Deeb-Sossa, Papadakis, Bollen, and Curran (2004). The random intercepts in Formula 10 can
further be decomposed into , or the average intercept for each profile, and , or the deviation
of each person’s profile intercept from this average intercept (see Formula 11). Similarly, the random
linear slope is decomposed into , or the average slope for each profile and , or the
deviation from this average slope for each person I (see Formula 12). Interestingly, because and
capture deviations from the average intercept and slope, respectively, they represent the
variability of the intercepts and slopes across cases within profiles. Of particular importance is that,
because all terms in formulas 11 and 12 have a subscript k, each of the profiles can have a unique
growth function. Although the GMM in formulas 10-12 is a linear GMM, other functional forms can
23
be tested as well, such as a quadratic or cubic GMM. Readers interested in a more thorough
treatment of GMM can consult the papers by Jung and Wickrama (2009) or Ram and Grimm (2009).
Our literature review revealed a modest number of studies applying this longitudinal person-
centered technique in vocational research, with the first studies using this technique being published
around the year 2010 (see Table A8). For instance, Hirschi (2011c) used LCGA to identify different
developmental trajectories of career-choice readiness: (1) “high increasing” describes a class of
people with high initial readiness and a linear increase of readiness over time; (2) “high decreasing”
is characterized by a very high initial level of readiness followed by a decline in readiness over time;
(3) “moderate increasing” showed a moderate initial level of readiness and a linear subsequent
increase in readiness; and finally (4) “low stable” showed a low initial level of readiness and almost
no increase in readiness over time.
Latent Transition Analysis (LTA). Latent Transition Analysis (LTA) is a longitudinal
extension of LCA/LPA (Collins & Lanza, 2010). That is, in LTA people can transition from one
latent class to another over time (see Figure 5). Because the latent classes in LTA refer to subgroup
memberships at that particular point in time, they are referred to as latent statuses, rather than latent
classes. A good illustration of this technique comes from research conducted by Mäkikangas (2018),
who studied latent profiles of job crafting strategies across time. Using latent profile analysis, she
first demonstrated that in a sample of Finnish rehabilitation center employees a distinction can be
made between ‘active’ and ‘passive’ job crafters, with the latter only trying to decrease their
hindering job demands to some extent, without trying to increase their job resources or challenging
job demands. In a next step, she used LTA to investigate the stayer-mover patterns across job
crafting profiles over time. In this specific example, the latent transition probabilities were zero,
indicating that no transitions occurred across a one-week interval.
In a LTA for categorical indicators, three sets of parameters are estimated. First, at each time
point the proportion of individuals that is expected to belong to each latent status is estimated. This is
24
referred to as the latent status membership probabilities ( in Formula 13). Second, the transition
probabilities capture the probability of transitioning from a specific latent status at time t to another
latent status at time t+1 ( in Formula 13). Third, item-response probabilities tap into the connection
between latent status membership and the observed categorical indicators at each time point (the ’s
in Formula 13). By doing so, item-response probabilities provide information on the differentiation
of the latent statuses. Formally, a LTA model for two measurement occasions (i.e., t and t+1),
latent statuses (with a denoting a latent status at measurement occasion t, and b denoting a
latent status at measurement occasion t+1), and four indicators at each measurement occasion (i.e.,
, , , ) can be expressed as follows:
(13)
In Formula 13, y represents a specific response pattern on the categorical indicators across both
measurement occasions (i.e., ), represents the proportion of
individuals in latent status a at time t, and is the probability of membership in latent status b at
measurement occasion t+1, conditional on membership in latent status a at measurement occasion t.
Finally, is the probability of response i to the first item at measurement occasion t, conditional
on membership in latent status a at measurement occasion t. Readers interested in a more thorough
treatment of LTA can consult the book by Collins and Lanza (2010).
Our literature review identified only three studies that used LTA in vocational research so far
(see Table A9). In addition to Mäkikangas (2018; see above), Kunst, van Woerkom, van Kollenburg
and Poell (2018) used LTA to identify trajectories of goal orientation profiles in a teacher sample.
Although the majority of teachers remained in the same goal orientation profile over the one-year
interval (i.e., success-oriented, diffuse, low-performance, or high-avoidance), a small percentage of
teachers shifted towards a different profile, and this shift was supported by a specific type of
managerial coaching. Rice, Ray, Davis, DeBlaere and Ashby (2015) used LTA to study the stress
25
trajectories of different types of perfectionists, showing that maladaptive perfectionists never
transitioned to low stress whereas only 4% of the adaptive perfectionists transitioned to high stress.
The potential of modeling profiles of intraindividual change processes for vocational
research. Methods for detecting profiles of intraindividual change have direct relevance to
vocational research because careers by definition develop and evolve over time (De Vos et al., in
press; Hall, 2002). Traditional longitudinal models, such as the latent growth model, however, make
the strong assumption that change can be described using the same functional form (e.g., linear or
quadratic) for everyone. Whereas this might be true in very specific circumstances, the awareness
that careers and career choices are driven by the complex interplay of a wide set of person
characteristics, as well as the situations one encounters, suggests that heterogeneity in change might
be the rule rather than the exception. In response to this awareness, GMM is particularly interesting
because it allows the change over time to be qualitatively different for different groups of
individuals. Using GMM, Hirschi (2011) for example identified distinct developmental trajectories
of career-choice readiness in adolescents and demonstrated that students in those trajectories differed
on core-self evaluations, occupational knowledge and barriers. Also for the counseling field GMM
shows a lot of promise because it for example allows studying patterns of responses to treatment,
showing “what works or does not work for whom?” (Frankfurt et al., 2016; p. 624). Such insights
gained through GMM might help counseling psychologists tailoring their treatments and intervene
more effectively.
LTA, being a longitudinal extension of LCA/LPA, is a method holding a lot of promise for
the careers field because it aligns well with the definition of careers as “the individually perceived
sequence of work-related experiences and activities over the span of a person’s life” (Hall, 2002, p.
12). In LTA—and LPA and LCA more broadly—those work-related experiences are not studied in
isolation, rather the combined profile states of those work-related experiences are the unit of
analysis. Moreover, because of its ability to model transitions between those profile states, the
26
treatment of careers in LTA closely resembles our theoretical conceptualization of it. This is not only
important from a substantive-methodological fit perspective, but studying careers in this way might
also provide novel and unique information. For example, Xu and Payne (2018) used LTA for
studying changes (or transitions) in organizational commitment profiles over time, and they
demonstrated that those transitions themselves (e.g., from a value-based commitment profile to a
weak commitment profile) were predictive of turnover hazards. Future studies could for example use
LTA to investigate transitions in career profiles (e.g., from “protean career architects” to “solid
citizen”, Briscoe & Hall, 2006) or in work versus family commitment profiles (e.g., from “work
profile” to “family profile”, Cinamon & Rich, 2002).
Important issues in person-centered research
There are a number of issues that are fundamental and practical to many contemporary
person-centered analyses, including class enumeration, profile labeling, inclusion of covariates, and
multi-group invariance testing. Because these issues apply to most methods discussed above, we
review them in a separate section.
Class enumeration
Selecting the optimal number of latent profiles is a thorny issue. Typically, models with an
increasing number of latent profiles are tested after which the most optimal one is selected based on
interpretability and theoretical conformity of the solution, statistical adequacy (e.g., no negative
residual variances), and statistical indicators. Regarding the latter, several indicators are available,
with simulation research showing that the Bayesian Information Criterion (BIC), the sample-adjusted
BIC (SABIC), the Consistent Akaike Information Criterion (CAIC), and the Bootstrap Likelihood
Ratio Test (BLRT) are among the most effective ones (e.g., Henson, Reise, & Kim, 2007; Nylund,
Asparouhov, & Muthén, 2007). However, because of the sample size dependency of those indicators,
they might suggest keeping on adding profiles in case one’s sample size is large. If this happens,
27
Morin and colleagues (2011) suggest looking at additional gains in fit when adding more latent
profiles using so-called “elbow plots”.
Labeling of profiles
The latent profiles in a profile solution can differ in many ways, including differences in the
unique pattern of high and low mean scores on the indicators (i.e., shape differences), differences in
the mean score across all indicators (i.e., level differences), and differences in the degree of
differentiation among indicators within a profile (i.e., scatter differences) (Meyer & Morin, 2016).
When it comes to labeling of the profiles, any of these differences can be referred to, with different
labeling schemes being used in different research fields. For example, in the commitment literature,
researchers have predominantly focused on shape differences, with the most common labeling
scheme being one in which the commitment component with the highest score is referred to as
“dominant” (e.g., affective commitment dominant or continuance commitment dominant). The
advantage of focusing on only one of the differences is simplicity. The downside, however, is that it
comes with decreased accuracy because other between-profile differences are not taken into account.
One solution adopted by Meyer and Morin (2016) is to add level and scatter information whenever
relevant (i.e., whenever level or scatter are either high or low).
Incorporating covariates
When engaging in person-centered research, one is often interested in learning how profile
membership relates to covariates. Research generally shows that covariates should only be included
once the optimal unconditional profile solution (i.e., the profile solution based on only those
variables making up the profile) is selected (Morin et al., 2019; Nylund-Gibson & Masyn, 2016).
Moreover, inclusion of covariates in the model should not change the nature of the profiles as this
causes the latent categorical variable to “lose its meaning” (Asparouhov & Muthén, 2014; p. 329).
Looking at different ways in which covariates can be included in the analysis, a first way to
test predictors and/or outcomes is to directly include them in the final solution. For example, one
28
might include profile outcomes by specifying them as additional profile indicators. Whereas direct
inclusion of covariates might help to reduce biases in the estimation of the profile-covariate relations
and although this helps limiting Type 1 errors (Diallo & Lu, 2017), one needs to make sure that
including the covariates does not change the optimal unconditional profile solution (see above). In
case the profile solution is modified by the inclusion of the covariates, a different approach can be
taken. This approach, referred to as the automated auxiliary approach, is specifically designed to
prevent this from happening. In fact, there is not one but different automated auxiliary approaches,
with Morin and colleagues (2019) suggesting that the preferred automated auxiliary approach
depends on whether you look at predictors, outcomes or correlates of profile membership. In case
one is interested in predictors, the “three-step” approach seems to perform well (see Asparouhov &
Muthén, 2014 for more information). For outcomes, either the three-step approach, the approach by
Lanza, Tan, and Bray (2013), or the BCH approach (see Asparouhov & Muthén, 2014 for more
information) can be used. According to Meyer and Morin (2016), correlates can best be tested using
the E function in Mplus because this approach does not assume directionality of the associations.
Finally, McLarnon and O’Neill (2018) discuss how one can manually implement the BCH and three-
step approach when one wants to test more complex mediation and moderation models or models
that look at the effect on an outcome after accounting for control variables.
Multi-group invariance testing
An important issue in person-centered research is whether profiles found in one sample
generalize across known subpopulations (Morin, Meyer, Creusier, & Biétry, 2016). For LCA, multi-
group invariance has typically been tested using a three-step approach in which one tests whether (1)
the same numbers of latent classes are extracted within each group, (2) the response probabilities are
the same across groups, and (3) the relative size of the profiles is the same across groups (see Eid,
Langeheine, & Diener, 2003).
29
Recently, Morin and colleagues (2016) extended this approach by revising the second step for
LPA rather than LCA and by including tests of similarity between the profiles, antecedents and
outcomes across subpopulations. This approach consists of six steps that test (1) whether the same
number of latent profiles is found in each group (i.e., configural similarity), (2) whether the
indicator’s levels are equal across groups (i.e., structural similarity), (3) whether the indicator’s
variability is equal across groups (i.e., dispersion similarity), (4) whether the relative size of the
profiles is the same across groups (i.e., distributional similarity), (5) whether the predictor-profile
relations are the same across groups (i.e., predictive similarity), and (6) whether the profile-outcome
relations are the same across groups (i.e., explanatory similarity). Morin and Wang (2016) extended
this approach to MRM, which essentially requires one additional step between the first and second
one in which the invariance of regression coefficients is tested across groups. Readers interested in
learning more about multi-group invariance testing in the context of LPA can consult the paper and
accompanying Mplus code by Morin and colleagues (2016), while the chapter by Morin and Wang
(2016) shows how to perform multi-group invariance testing for MRM.
Finally, as argued by Morin and colleagues (2019), the six-step multi-group profile similarity
framework can also be used to test for longitudinal invariance in LTA, although in the presence of
distributional similarity (i.e., the profiles account for equal proportions of the sample over time) one
cannot directly impose equality constraints on the relative size of the profiles over time. In this case,
the approach described by Morin and Litalien (2017) is needed.
Discussion: Critical Reflections on the Use of Person-Centered Methods
Despite their promise to vocational research, as evidenced by our literature review, some
researchers remain reluctant to adopt person-centered methods because of their exploratory nature
and their choice for a categorical rather than a continuous latent variable. In what follows, we aim to
offer a balanced discussion of these issues, hoping that this helps researchers to take a stance and
make informed decisions when designing their studies and plans of analysis.
30
The exploratory nature of person-centered methods
As mentioned earlier, person-centered models are exploratory in the sense that a ‘final’
model is typically obtained by comparing solutions with different numbers of profiles (or
subpopulations) after which the ‘optimal’ one is selected. One concern is that such an exploratory
procedure is highly sample-dependent, thus limiting the generalizability of one’s findings.
First, it is important to note that balancing model fit and model parsimony does not preclude
the generation of expectations regarding the number and/or the structure of the profiles (Morin et al.,
2018). For example, in case one would study how people’s job satisfaction develops after starting a
new job, it would be good practice to build on previous research that has demonstrated that job
satisfaction generally shows a trend of steady decline after entering a new job (e.g., Boswell,
Boudreau, & Tichy, 2005). Hence, in that particular case one expects a hypothesis that at least one of
the subpopulations follows such a hangover-pattern (see Solinger et al., 2013). Thus, although the
exact number of profiles can often not be predicted when performing person-centered analyses, one
might still have expectations concerning the nature of some of the profiles. Morin and colleagues
(2018) make exactly the same point, using the analogy of fishing. Whereas a fully a-theoretical
undertaking (which they refer to as dustbowl empiricism) corresponds to dynamite fishing, in which
one throws sticks of dynamite into the water and catches whatever floats to the surface, valuable
exploratory research is like fly fishing. In fly fishing, one starts by carefully selecting the appropriate
bait and fishing location, anticipating catching a particular type of fish. While in the fly fishing
scenario the number of fish, their size and even the type of fish is not known in advance, the
difference with dynamite fishing is that one goes well prepared to the expedition, knowing that
something valuable will come out of it (Morin et al., 2018).
Even though exploratory research, when well-planned, often leads to interesting findings, the
lack of a comprehensive theory that serves as a basis for clear hypotheses makes replication of one’s
findings increasingly important. This is particularly true provided that in some cases—for example
31
when the model’s distributional assumptions are violated—spurious profiles can emerge (Bauer &
Curran, 2003). Therefore, construct validation of one’s solution is essential (Morin et al., 2018).
According to Morin and colleagues (2018), such construct validation involves the following steps:
(1) demonstrating that the profiles have theoretical value, (2) demonstrating that the profiles relate in
a meaningful way to key covariates, and (3) demonstrating that the profiles generalize to new
samples or are (at least somewhat) stable across time.
Finally, we believe that person-centered methods are useful for inductive theorizing
(Hofmans, Vantilborgh, & Solinger, 2018). In case suitable theory is scarce or even non-existing,
restricting oneself to deductive logic in which one draws on a theory to build a general rule, after
which one tests whether the rule also applies to one’s data, might be too limiting (Ketokivi &
Mantere, 2010). In the scenario where there is little theory, person-centered methods are particularly
interesting because they offer a way to discover new aspects of phenomena through inductive
thinking. This inductive thinking might take the form of contextual or theoretical induction (Ketokivi
& Mantere, 2010). With contextual induction, the reason for existence of (some of) the
subpopulations is looked for in the research context. In theory-based induction, the subpopulations
are not assumed to be the result of a particular sample setting; instead one tries to achieve a
theoretical understanding of the subject matter. Apart from contextual and theoretical induction,
researchers can also look for counter-factuals in their findings. Such counter-factuals are findings
that are counter to one’s set of theoretical assumptions, and typically give rise to imaginative and
innovative research because they are followed by a problematization of assumptions from the part of
the researcher and the presentation of an alternative (Cornelissen & Durand, 2014).
A categorical versus dimensional approach to latent variables
As we have argued above, person-centered methods are typically classification-based. This
means that, in case the method is performed in a latent variable framework, it posits a categorical
latent variable. A logical question then is whether the choice for a categorical latent variable, rather
32
than a continuous one, makes sense. This issue is particularly important because, as Molenaar and
von Eye (1994) demonstrated, under certain conditions, an m-factor common factor model can be
perfectly reproduced with a K = m +1 class latent profile model (see also Bauer & Curran, 2003).
This equivalence is important because it places great onus on researchers to argue for the tenability
of one representation versus the other. We feel that there are two possible ways to go about this (see
also Collins & Lanza, 2010).
First, one may have a strong belief that the latent variable is categorical, and that it therefore
should be modeled in a person-centered, categorical way. Although such discussions are not very
prominent in vocational research, there are a number of research domains in which the issue of
dimensionality versus categoricity is a key question. Psychopathology research is such a domain,
with the crucial question being whether a dimensional or categorical classification of personality
disorders should be used (Ruscio, Ruscio, & Carney, 2011). An important criterion that is used to
argue for the meaningfulness of a categorical (or person-centered) rather than a dimensional (or
variable-centered) solution is the presence of qualitative, rather than quantitative differences between
profiles (Chen, Morin, Parker, & Marsh, 2015; De Boeck, Wilson, & Acton, 2005). The rationale
behind this idea is that quantitative differences, or ordered profiles that differ only in level, can be
well accommodated by a model with a continuous latent variable. Qualitative differences, or profiles
that differ in shape, instead, support the meaningfulness of a categorical, person-centered approach
because such differences cannot be captured well using the traditional, variable centered approach.
Although the issue of testing for dimensionality versus categoricity goes beyond the scope of the
present article, it is important to know that empirical tests have been developed to evaluate whether a
construct is categorical versus dimensional. Readers interested in this issue can consult the
taxometric method developed by Meehl (1992) or the article of Ruscio and colleagues (2011).
Second, rather than debating whether a construct is continuous or categorical, one might
consider that both the continuous as well as the categorical approach provide separate, but equally
33
useful information (Collins & Lanza, 2010). When such an agnostic perspective on the nature of
constructs is adopted, the choice for a continuous or a categorical latent variable is dictated by the
research question at hand. For example, although few people would dispute a continuous treatment
of personality traits, according to which individual differences in personality traits are expressed as
the degree to which the trait is characteristic of those individuals, most people also see value in a
categorical, profile-perspective on traits. The reason is that those two approaches to personality traits
address different questions. While the dimensional perspective is well suited to study the effect of
individual differences in one or more traits, thereby taking those traits as the focal point of analysis,
the categorical perspective looks at the effects of specific combinations of trait scores, which implies
shifting the focus from the traits to the individual. This example also illustrates a broader point.
Despite our plea for more person-centered vocational research, it is important to realize that person-
and variable-centered approaches are not conflicting but rather complementary. Ultimately, person-
and variable-centered approaches can even be used in tandem to provide a more comprehensive view
of the same phenomena (e.g., Morin, Boudrias, Marsh, Madore, & Desrumaux, 2016).
References
Asendorpf, J. B. (2006). Typeness of personality profiles: A continuous person-centred approach to
personality data. European Journal of Personality, 20, 83-106.
Asparouhov, T., Hamaker, E. L., & Muthén, B. (2018). Dynamic structural equation models.
Structural Equation Modeling, 25, 359-388.
Asparouhov, T., & Muthén, B.O. (2014). Auxiliary variables in mixture modeling: Three-step
approaches using Mplus. Structural Equation Modeling, 21, 329-341.
Bauer, D.J., & Curran, P.J. (2003). Distributional assumptions of growth mixture models over-
extraction of latent trajectory classes. Psychological Methods, 8, 338-363.
Bergman, L. R., & Magnusson, D. (1997). A person-oriented approach in research on developmental
psychopathology. Development and Psychopathology, 9, 291-319.
34
Bergman, L. R., & Trost, K. (2006). The person-oriented versus the variable-oriented approach: Are
they complementary, opposites, or exploring different worlds? Merrill-Palmer Quarterly, 52,
601-632.
Biesanz, J. C., Deeb-Sossa, N., Aubrecht, A. M., Bollen, K. A., & Curran, P. J. (2004). The role of
coding time in estimating and interpreting growth curve models. Psychological Methods, 9, 30-
52.
Bollen, K.A., & Curran, P.J. (2006). Latent curve models: A structural equation perspective.
Hoboken, NJ: Wiley.
Borgen, F. H., & Barnett, D. C. (1987). Applying cluster analysis in counseling psychology research.
Journal of Counseling Psychology, 34, 456-468.
Boswell, W.R., Boudreau, J.W., & Tichy, J. (2005). The relationship between employee job change
and job satisfaction: The honeymoon-hangover effect. Journal of Applied Psychology, 90, 882–
892.
Briscoe, J. P., & Hall, D. T. (2006). The interplay of boundaryless and protean careers:
Combinations and implications. Journal of Vocational Behavior, 69, 4-18.
Brusco, M. J., Cradit, J. D., Steinley, D., & Fox, G. L. (2008). Cautionary remarks on the use of
clusterwise regression. Multivariate Behavioral Research, 43, 29–49.
Cattell, R.B. (1952). The three basic factor-analytic designs: Their interrelations and derivatives.
Psychological Bulletin, 49, 499–520.
Chen, X., Morin, A. J., Parker, P. D., & Marsh, H. W. (2015). Developmental investigation of the
domain-specific nature of the life satisfaction construct across the post-school transition.
Developmental Psychology, 51, 1074-1085.
Chlosta, S., Patzelt, H., Klein, S. B., & Dormann, C. (2012). Parental role models and the decision to
become self-employed: the moderating effect of personality. Small Business Economics, 38, 121-
138.
35
Cinamon, R. G., & Rich, Y. (2002). Profiles of attribution of importance to life roles and their
implications for the work–family conflict. Journal of Counseling Psychology, 49, 212-220.
Clark, S. L., Muthén, B., Kaprio, J., D'Onofrio, B. M., Viken, R., & Rose, R. J. (2013). Models and
strategies for factor mixture analysis: An example concerning the structure underlying
psychological disorders. Structural Equation Modeling, 20, 681-703.
Cohen, C. R., Chartrand, J. M., & Jowdy, D. P. (1995). Relationships between career indecision
subtypes and ego identity development. Journal of Counseling Psychology, 42, 440-447.
Collins, L.M., & Lanza, S.T. (2010). Latent class and latent transition analysis with applications in
the social, behavioral, and health sciences. Hoboken, NJ: Wiley.
Cornelissen, J. P., & Durand, R. (2014). Moving forward: Developing theoretical contributions in
management studies. Journal of Management Studies, 51, 995-1022.
da Motta Veiga, S.P. & Turban, D.B. (2018). Insight into job search self-regulation: Effects of
employment self-efficacy and perceived progress on job search intensity. Journal of Vocational
Behavior, 108, 57-66.
Davison, M. L., & Davenport, E. C., Jr. (2002). Identifying criterion-related patterns of predictor
scores using multiple regression. Psychological Methods, 7, 468–484.
De Boeck, P., Wilson, M., & Acton, G. S. (2005). A conceptual and psychometric framework for
distinguishing categories and dimensions. Psychological Review, 112, 129-158.
De Vos, A., Van der Heijden, B. I., & Akkermans, J. (in press). Sustainable careers: towards a
conceptual model. Journal of Vocational Behavior.
Diallo, T. M., & Lu, H. (2017). Consequences of misspecifying across-cluster time-specific residuals
in multilevel latent growth curve models. Structural Equation Modeling, 24, 359-382.
Eid, M., Langeheine, R., & Diener, E. 2003. Comparing typological structures across cultures by
multigroup latent class analysis. Journal of Cross-Cultural Psychology, 34, 195-210.
Frankfurt, S., Frazier, P., Syed, M., & Jung, K. R. (2016). Using group-based trajectory and growth
36
mixture modeling to identify classes of change trajectories. The Counseling Psychologist, 44,
622-660.
Ferguson, S. L., & Hull, D. M. (2019). Exploring science career interest: Latent profile analysis of
high school occupational preferences for science. Journal of Career Development, 46, 583-598.
Gerber, M., Wittekind, A., Grote, G., & Staffelbach, B. (2009). Exploring types of career orientation:
A latent class analysis approach. Journal of Vocational Behavior, 75, 303-318.
Haines III, V. Y., Doray-Demers, P., & Martin, V. (2018). Good, bad, and not so sad part-time
employment. Journal of Vocational Behavior, 104, 128-140.
Hall, D.T. (2002). Careers in and out of organisations. Thousand Oaks, CA: Sage.
Henson, J.M., Reise, S.P., & Kim, K.H. (2007). Detecting mixtures from structural model
differences using latent variable mixture modeling: A comparison of relative model fit statistics.
Structural Equation Modeling, 14, 202-226.
Hirschi, A., & Valero, D. (2017). Chance events and career decidedness: Latent profiles in relation to
work motivation. Career Development Quarterly, 65, 2-15.
Hofmans, J., Ceulemans, E., Steinley, D., & Van Mechelen, I. (2015). On the added value of
bootstrap analysis for K-means clustering. Journal of Classification, 32, 268-284.
Hofmans, J., De Gieter, S., & Pepermans, R. (2013). Individual differences in the relationship
between satisfaction with job rewards and job satisfaction. Journal of Vocational Behavior, 82, 1-
9.
Hofmans, J., Vantilborgh, T., & Solinger, O.N. (2018). k-centres Functional Clustering: A person-
centered approach to modeling complex nonlinear growth trajectories. Organizational Research
Methods, 21, 905-930.
Howard, M. C., & Hoffman, M. E. (2018). Variable-centered, person-centered, and person-specific
approaches: Where theory meets the method. Organizational Research Methods, 21, 846-876.
Jung, T., & Wickrama, K. A. S. (2008). An introduction to latent class growth analysis and growth
37
mixture modeling. Social and Personality Psychology Compass, 2, 302-317.
Kaufman, L., & Rousseeuw, P. J. (2005). Finding groups in data: An introduction to cluster
analysis. Wiley Hoboken, N.J.
Ketokivi, M., & Mantere, S. (2010). Two strategies for inductive reasoning in organizational
research. Academy of Management Review, 35, 315-333.
Kunst, E. M., van Woerkom, M., van Kollenburg, G. H., & Poell, R. F. (2018). Stability and change
in teachers' goal orientation profiles over time: Managerial coaching behavior as a predictor of
profile change. Journal of Vocational Behavior, 104, 115-127.
Kuron, L. K. J., Schweitzer, L., Lyons, S., & Ng, E. S. W. (2016). Career profiles in the "new
career": evidence of their prevalence and correlates. Career Development International, 21, 355-
377.
Lanza, S.T., Tan, X., & Bray, B.C. (2013). Latent class analysis with distal outcomes: A flexible
model-based approach. Structural Equation Modeling, 20, 1-26.
Lubke, G. H., & Muthén, B. (2005). Investigating population heterogeneity with factor mixture
models. Psychological Methods, 10, 21-39.
Mäkikangas, A. (2018). Job crafting profiles and work engagement: A person-centered approach.
Journal of Vocational Behavior, 106, 101-111.
Masyn, K.E. (2013). Latent class analysis and finite mixture modeling. In T.D. Little (Ed.), The
Oxford handbook of quantitative methods (pp. 551-611). NY: Oxford University Press.
Matthews, E., & Tiedeman, D. V. (1964). Attitudes toward career and marriage and the development
of life style in young women. Journal of Counseling Psychology, 11, 375-384.
McLarnon, M.J.W. & O’Neill, T.A. (2018). Extensions of auxiliary variable approaches for the
investigation of mediation, moderation, and conditional effects in mixture models. Organizational
Research Methods, 21, 955-982.
Meehl, P. E. (1992). Factors and taxa, traits and types, differences of degree and differences in kind.
38
Journal of Personality, 60, 117–174.
Meyer, J.P., & Morin, A.J.S. (2016). A person-centered approach to commitment research: Theory,
research, and methodology. Journal of Organizational Behavior, 37, 584-612.
Meyer, J. P., Stanley, L. J., & Parfyonova, N. M. (2012). Employee commitment in context: The
nature and implications of commitment profiles. Journal of Vocational Behavior, 80, 1-16.
Moeller, J., Ivcevic, Z., White, A. E., Menges, J. I., & Brackett, M. A. (2018). Highly engaged but
burned out: intra-individual profiles in the US workforce. Career Development International, 23,
86-105.
Molenaar, P. C. M. (1985). A dynamic factor model for the analysis of multivariate time series.
Psychometrika, 50, 181–202.
Molenaar, P. C. M., & von Eye, A. (1994). On the arbitrary nature of latent variables. In A. von Eye
& C. C. Clogg (Eds.), Latent variables analysis: Applications for developmental research (pp.
226-242). Thousand Oaks, CA, US: Sage Publications, Inc.
Morin, A.J.S, Boudrias, J.S., Marsh, H.W., Madore, I., & Desrumaux, P. (2016). Further reflections
on disentangling shape and level effects in person-centered analyses: An illustration exploring the
dimensionality of psychological health. Structural Equation Modeling, 23, 438-454.
Morin, A.J.S., Bujacz, A., & Gagné, M. (2018). Person-centered methodologies in the organizational
sciences: introduction to the feature topic. Organizational Research Methods, 21, 803-813.
Morin, A.J.S., & Litalien, D. (2017). Webnote: Longitudinal tests of profile similarity and latent
transition analyses. Retrieved October, 14 2019 from
https://smslabstats.weebly.com/uploads/1/0/0/6/100647486/lta_distributional_similarity_v02.pdf.
Montreal, QC: Substantive Methodological Synergy Research Laboratory.
Morin, A.J.S., Maano, C., Nagengast, B., Marsh, H.W., Morizot, J., & Janosz, M. (2011). Growth
mixture modeling of adolescents trajectories of anxiety: The impact of untested invariance
assumptions on substantive interpretations. Structural Equation Modeling, 18, 613-648.
39
Morin, A. J. S., & Marsh, H. W. (2015). Disentangling shape from level effects in person-centered
analyses: An illustration based on university teachers’ multidimensional profiles of effectiveness.
Structural Equation Modeling, 22, 39-59.
Morin, A.J.S., McLarnon, M.J.W., & Litalien, D. (in press). Mixture modeling for organizational
behavior research. In Y. Griep, & S.D. Hansen (Eds.), Handbook of dynamic organizational
behavior. Cheltenham, UK: Edward Elgar.
Morin, A.J.S., Meyer, J.P., Creusier, J., & Biétry, F. (2016). Multiple-group analysis of similarity in
latent profile solutions. Organizational Research Methods, 19, 231-254.
Morin, A. J. S., Morizot, J., Boudrias, J. -S., & Madore, I. (2011). A multifoci person-centered
perspective on workplace affective commitment: A latent profile/factor mixture analysis.
Organizational Research Methods, 14, 58–90.
Morin, A. J. S., & Wang, J. C. K. (2016). A gentle introduction to mixture modeling using physical
fitness data. In N. Ntoumanis & N. Myers (Eds.), An introduction to intermediate and advanced
statistical analyses for sport and exercise scientists (pp. 183–210). London: Wiley.
Muthén, L.K., & Muthén, B.O. (2017). Mplus User’s Guide (8th ed.). Los Angeles, CA: Muthén &
Muthén.
Nylund, K.L., Asparouhov, T., & Muthén, B. (2007). Deciding on the number of classes in latent
class analysis and growth mixture modeling: A Monte Carlo simulation study. Structural
Equation Modeling, 14, 535–569.
Nylund-Gibson, K., & Masyn, K.E. (2016). Covariates and mixture modeling: Results of a
simulation study exploring the impact of misspecified effects on class enumeration. Structural
Equation Modeling, 23, 782-797.
Perry, J. C. (2008). School engagement among urban youth of color - Criterion pattern effects of
vocational exploration and racial identity. Journal of Career Development, 34, 397-422.
Peugh, J. & Fan, X. (2013). Modeling unobserved heterogeneity using latent profile analysis: A
40
Monte Carlo simulation. Structural Equation Modeling, 20, 616-639.
Ram, N., & Grimm, K. (2007). Using simple and complex growth models to articulate
developmental change: Matching theory to method. International Journal of Behavioral
Development, 31, 303-316.
Ram, N., & Grimm, K. J. (2009). Growth mixture modeling: a method for identifying differences in
longitudinal change among unobserved groups. International Journal of Behavioral Development,
33, 565–576.
Reitzle, M., & Vondracek, F. W. (2000). Methodological avenues for the study of career pathways.
Journal of Vocational Behavior, 57, 445-467.
Rice, K. G., Ray, M. E., Davis, D. E., DeBlaere, C., & Ashby, J. S. (2015). Perfectionism and
longitudinal patterns of stress for STEM majors: Implications for academic performance. Journal
of Counseling Psychology, 62, 718-731.
Ruscio, J., Ruscio, A. M., & Carney, L. M. (2011). Performing taxometric analysis to distinguish
categorical and dimensional variables. Journal of Experimental Psychopathology, 2, 170-196.
Solinger, O. N., Van Olffen, W., Roe, R. A., & Hofmans, J. (2013). On becoming (un) committed: A
taxonomy and test of newcomer onboarding scenarios. Organization Science, 24, 1640-1661.
Steinley, D. (2006). k-means clustering: A half-century synthesis. British Journal of Mathematical
and Statistical Psychology, 59, 1-34.
Sterba, S. K. (2013). Understanding linkages among mixture models. Multivariate Behavioral
Research, 48, 775-815.
Tan, P.-N., Steinbach, M., Karpatne, A., & Kumar, V. (2019). Introduction to data mining. New
York, NY: Pearson.
Vantilborgh, T., Hofmans, J., & Judge, T. A. (2018). The time has come to study dynamics at work.
Journal of Organizational Behavior, 39, 1045–1049.
Vergauwe, J., Wille, B., Hofmans, J., & De Fruyt, F. (2017). Development of a Five-Factor Model
41
charisma compound and its relations to career outcomes. Journal of Vocational Behavior, 99, 24-
39.
Vondracek, F.W., & Porfeli, E. (2002). Integrating person- and function-centered approaches in
career development theory and research. Journal of Vocational Behavior, 61, 386-397.
von Eye, A. (1990). Introduction to configural frequency analysis. Cambridge: Cambridge
University Press.
von Eye, A. (2002). Configural frequency analysis: Methods, models, and applications. Mahwah,
NJ: Erlbaum.
Wang, M., & Wanberg, C. R. (2017). 100 years of applied psychology research on individual
careers: From career management to retirement. Journal of Applied Psychology, 102, 546–563.
Wedel, M., & DeSarbo, W. S. (1995). A Mixture Likelihood Approach for Generalized Linear
Models. Journal of Classification, 12, 21–55.
Weiss, H. M., & Rupp, D. E. (2011). Experiencing work: An essay on a person-centric work
psychology. Industrial and Organizational Psychology, 4, 83-97.
Wiernik, B. M. (2016). Intraindividual personality profiles associated with realistic interests. Journal
of Career Assessment, 24, 460-480.
Woo, S. E., Jebb, A., Tay, L., & Parrigon, S. (2018). Putting the “person” in the center: Review and
synthesis of person-centered approaches and methods in organizational science. Organizational
Research Methods, 21, 814-845.
Xu, X., & Payne, S. C. (2018). Predicting retention duration from organizational commitment profile
transitions. Journal of Management, 44, 2142-2168.
Zacher, H., Rudolph, C. W., Todorovic, T., & Ammann, D. (2019). Academic career development:
A review and research agenda. Journal of Vocational Behavior, 110, 357-373.
Zyphur, M.J. (2009). When mindsets collide: Switching analytical mindsets to advance organization
research. Academy of Management Review, 34, 677-688.
42
Figure 1: Latent Profile Analysis (LPA). Scores on the continuous (LPA) indicators (y’s) are caused
by the categorical latent variable C (with k latent classes). In the most constrained model, the latent
classes differ in mean scores on the indicators only, but in alternative formulations, indicator
variances can be class-specific and correlated residuals can be added. In case the indicators are not
continuous but categorical, LPA becomes Latent Class Analysis (LCA).
43
Figure 2: Factor Mixture Analysis (FMA). The indicators (y’s) are caused by both a continuous
latent variable f and a categorical latent variable C (with k latent classes). The dashed lines indicate
that the factor structure can be different in each latent class. In FMA, factor loadings, factor means,
the factor covariance matrix and item intercepts/thresholds can be class-specific.
44
Figure 3: Mixture Regression Analysis (MRM). The latent variable C (with k latent classes)
moderates the relation between X and Y. In the basic MRM, the means and variances of the
outcome(s) are class-specific, while in a more flexible representation the means and variances of the
predictor(s) can also be class-specific.
45
Figure 4: Growth Mixture Modeling (GMM). k latent classes are estimated, each having class-
specific growth parameters. In GMM, any part of the model can be class-specific (including the
means and variances of the latent growth parameters, the indicator variances, etc.).
46
Figure 5: Latent Transition Analysis (LTA). The LTA model estimates on both measurement
occasions k latent classes (from repeated measures of the same four items at t1 and t2), as well as the
probabilities to transition from classes in Ck1 to classes in Ck2 over time. The number and structure
of profiles can be different on both measurement occasions, and indicators can be categorical
(implying that the latent transition model is an extension of Latent Class Analysis) or continuous
(implying that the latent transition model is an extension of Latent Profile Analysis).
47
Table A1: Overview of studies in research on careers, career counseling and/or vocational
behavior that used cluster analysis, along with the substantive question they wanted to address
Clustering dimensions of vocational identity and career beliefs
Hechtlinger, Levin, & Gati
(2019)
Investigating the factor structure and psychometric properties of the
Dysfunctional Career Decision-Making Beliefs questionnaire.
Shimizu, Dik, & Conner
(2019)
Investigating characteristics of subgroups of individuals who
identified as having a calling.
Capitano, DiRenzo, Aten, &
Greenhaus (2017)
Identifying role identity by means of profiles that reflect the
salience of three roles: work, home, and military service.
Rhee, Lee, Kim, Ha, & Lee
(2016)
Identifying statuses of vocational identity and how these are related
to planned happenstance skills.
Sestito, Sica, Ragozini, Porfeli,
Weisblat, & Di Palma (2015)
Exploring the configuration of vocational and overall identity
domains in young adults.
Vilhjalmsdottir & Arnkelsson
(2013)
Examining the relationship between habitus (i.e., a cognitive
structure based on configurations of cultural and leisure activities)
and career choice.
Kossek, Ruderman, Braddy, &
Hannum, (2012)
Exploring how boundary management profiles, reflecting
interruption behaviors, identity centralities, and boundary control,
relate to key work-family outcomes.
Zhou, Leung, & Li (2012)
Examining the meaning of work among Chinese university
students.
Santos & Ferreira (2012)
Identifying groupings that underly the concept of career indecision
based on a battery of instruments designed to assess career and
personality dimensions.
48
Hirschi (2011a)
Identifying essential and optional components of a presence of
calling.
Hirschi (2011b)
Identifying groups of students based on the dimensions of career
exploration and career commitment; investigating whether these
different identity statuses relate to differences in interest structure
in terms of differentiation, coherence, elevation, and interest-
aspiration congruence.
Luyckx, Duriez, Klimstra, &
De Witte (2010)
Identifying identity clusters or statuses and investigating concurrent
and prospective relations with work engagement and burnout.
Moen, Kelly, & Huang (2008)
Identifying job and home ‘ecologies’ based on work-family and
demands-control variables.
Argyropoulou, Sidiropoulou-
Makakou, & Besevegis (2007)
Classifying students based on their career decision status;
investigating the relationship between career decision status groups
and generalized self-efficacy, coping strategies, and vocational
interests.
Akos, Konold, & Niles (2004)
Exploring a career readiness typology of 8th-graders using the
Career Factors Inventory.
Kelly & Lee (2002)
Exploring the structure of career indecision based on six factors:
lack of information, need for information, trait indecision,
disagreement with others, identity diffusion, and choice anxiety.
Tracey & Darcy (2002)
Examining the relationships between career decidedness and
vocational interest.
Cinamon & Rich (2002)
Identifying profiles of attribution of importance to life roles and
examining their implications for the work-family conflict.
49
Larson & Majors (1998)
Identifying subtypes of undecided students based on the clustering
of a variety of career planning measures.
Meldahl & Muchinsky (1997)
Identifying different clusters of career indecision based on
measures of career indecision and neuroticism.
Gati, Krausz, & Osipow (1996)
Investigating the internal structure of a career decision-making
difficulties model and questionnaire.
Cohen & Chartrand (1995)
Investigating the relationships between career indecision subtypes
and ego identity development.
Rojewski (1994)
Identifying different career indecision types in a group of
adolescents from rural areas.
Wanberg & Muchinsky (1992)
Identifying groups of college students on the basis of their scores
on a range of personality and vocational indecision constructs.
Savickas & Jarjoura (1991)
Identifying groups of college students on the basis of their
responses to career decision scale items.
Lucas & Epperson (1990)
Identifying types of vocational undecidedness based on a battery of
personality and work orientation questionnaires.
Larson, Heppner, Ham, &
Dugan (1988)
Investigating multiple subtypes of career indecision.
Hamilton (1977)
Identifying different types of professional identity in doctorate
level programs in clinical and counseling psychology.
Matthews & Tiedeman (1964)
Investigating attitudes toward career and marriage and the
development of life style in young women.
Clustering career events and occupational characteristics
Smith & Campbell (2006)
Investigating the structure of ONET occupational values.
50
Armstrong, Smith, Donnay, &
Rounds (2004)
Creating a classification system and spatial map of occupations
using the Basic Interest Scale profiles of occupational incumbent
samples.
Bruce & Scott (1994)
Identifying types of inter-role transitions based on desirability and
magnitude ratings of 15 career events. Investigating whether
transition outcomes (strain, role ambiguity, adjustment difficulty,
transition eagerness, perceived gains and losses) differ across
transition types.
Claes & Quintanilla (1994)
Constructing career patterns based on self-reported activities taking
place (i.e., employment, educational preparations, unemployment,
military or civil service), personal and work-related variables, as
well as by means of work indices (work centrality, intrinsic versus
extrinsic work orientations, and societal norms about working).
Pickering & Galvin-Schaefers
(1988)
Comparing reentry women with career women on a set of
demographic and personality variables.
Clustering attitude dimensions (commitment – motivation – balance)
Vieira, Matias, Lopez, &
Matos (2018)
Identifying couple-level profiles of conflictual and enriching
dimension of work-family balance and investigating their
associations with individuals’ work- and family-related satisfaction.
Paixão & Gamboa (2017)
Identifying distinct motivational profiles in a sample of high school
students and investigating differences between and among these
profiles across career exploration and career indecision levels.
Kuron, Schweitzer, Lyons, &
Ng (2016)
Identifying career profiles based on protean and boundaryless
career attitudes and to examine differences between profiles in
51
terms of agency (i.e. career commitment, self-efficacy, and work
locus of control) and career attitudes (i.e. salience and satisfaction).
Moran, Diefendorff, Kim, &
Liu (2012)
Exploring how different combinations or patterns of motivations
(based on self-determination theory) relate to organizational
factors.
Tsoumbris & Xenikou (2010)
Creating profiles of commitment, based on the three components
(i.e., affective, continuance, normative) of organizational and
occupational commitment.
Segers, Inceoglu, Vloeberghs,
Bartram, & Henderickx (2008)
Identifying motivational groups based on a combination of work
motives and protean and boundaryless career attitudes.
Wasti (2005)
Exploring how affective, continuance, and normative commitment
combine to create distinct profiles of commitment; investigating
how commitment profiles relate to desirable job behaviors.
Clustering behaviors and problems
Maher, Gallagher, Rossi,
Ferris, & Perrewe (2018)
Investigating configurations of impression management tactics; and
to test political skill and political will as predictors of impression
management configurations.
Poynton, Lapan, & Marcotte
(2015)
Identifying distinct financial planning groups in a sample of 12th
graders and investigating how these groups differ on various
college and career readiness characteristics.
Solberg, Carlstrom, Howard, &
Jones (2007)
Classifying high school youth into varying academic at-risk
profiles using self-reported levels of academic confidence,
motivation to attend school, perceived family support, connections
with teachers and peers, and exposure to violence.
52
Multon, Wood, & Gysbers
(2007)
Identifying different types of career counseling clients based on a
variety of career-related variables (e.g., vocational identity) and
psychological issues that may affect career concerns (e.g., level of
psychological distress).
Gore, Bobek, Robbins, &
Shayne (2006)
Identifying a typology of computerized career guidance users,
based on a clustering of career exploratory behavior.
Rochlen, Milburn, & Hill
(2004)
Identifying different types of career counseling clients based on
(personal and career-related) distress, discomfort, uncertainty,
concerns, and stigma about career counseling.
Niles, Anderson, &
Goodnough (1998)
Identifying different ways in which adults use exploratory behavior
to cope with career development tasks.
Clustering interests and preferences
Einarsdottir, Eyjolfsdottir, &
Rounds (2013)
Clustering vocational interest items into basic interest scales to
describe the vocational interest landscape in Iceland.
Tay, Su, & Rounds (2011)
Investigating the structure and meaning of the people—things and
data—ideas interest dimensions.
Armstrong & Vogel (2009)
Investigating interest-efficacy associations from a RIASEC
perspective.
Armstrong, Rounds, & Hubert
(2008)
Exploring whether specific interest measures are best clustered
according to Holland’s higher order RIASEC types.
Stratton, Witzke, Elam, &
Cheever (2005)
Generating instructional profiles, reflecting participants’
comparative preferences for self-study/lecture versus group
discussion/computers.
Hansen & Scullard (2002)
Investigating the structure of leisure interests.
53
Shivy, Rounds, & Jones (1999)
Examining the structure of naturally occurring occupational
perceptions.
Clustering personality variables
Viola, Musso, Inguglia, & Lo
Coco (2016)
Identifying profiles of hardiness and to explore the moderating role
of hardiness in the association between psychological well-being
and career indecision.
Oztemel (2013)
Examining the validity of the classification system of the emotional
and personality-related career decision-making difficulties model
and questionnaire.
De Fruyt (2002)
Demonstrating that clusters of individuals, based on their FFM
personality scores (i.e., internalizers/externalizers and resilients),
show different positions on the labor market and demonstrate
differential initial career outcomes.
Gustafson & Mumford (1995)
Identifying clusters of ‘personal style’ based on seven job-relevant
personality variables. Identifying environmental constraints and
opportunities based on nine workgroup characteristics.
Note: Search terms were “cluster”, “hierarchical cluster”, and “k-means”
54
Table A2: Overview of studies in research on careers, career counseling and/or vocational
behavior that used latent class analysis, along with the substantive question they wanted to address
Classifying career preferences
Johnson & Bouchard (2009)
Investigating links between general intelligence and eight occupational
interest dimensions.
Gerber, Wittekind, Grote, &
Staffelbach (2009)
Identifying types of career orientation; to explore their prevalence; and
investigating differences in work attitudes and sociodemographic
variables between types.
Classifying employment types and outcomes
Haines, Doray-Demers, &
Martin (2018)
Developing a typology of part-time employment on the basis of role
occupancy and work characteristics; investigating attitudinal and
health-related outcomes associated with different employment
forms.
Hyvönen, Räikkönen, Feldt,
Mauno, Dragano, & Matthewman
(2017)
Forming reward patterns on the basis of perceived and objective
career rewards (i.e., career stability and promotions) and
investigating the impact of these patterns on personal work goals.
Van Aerden, Moors, Levecque,
& Vanroelen (2015)
Creating a typology of employment arrangements; investigating
differences between employment arrangements in terms of work-
related well-being indicators.
Classifying mobility patterns
Majeed, Forder, Mishra,
Kendig, & Byles (2015)
Identifying workforce participation patterns across the adult life
course, and exploring the influences of various early and adult life
socio-demographic circumstances.
55
Woo (2011)
To validate Ghiselli’s “hobo syndrome” as a career pattern
characterized by frequent job movement behavior and positive
attitudes about such behavior. To explore the dispositional roots of
hobo syndrome and its work-related outcomes.
Note: The search term was “latent class”
56
Table A3: Overview of studies in research on careers, career counseling and/or vocational
behavior that used latent profile analysis, along with the substantive question they wanted to address
Profiling interests
Ferguson & Hull (2019)
Identifying distinct profiles of science interest; investigating gender,
vocabulary ability, and personality as predictors of profile
membership.
Perera & McIlveen (2018)
Identifying distinct profiles of interests; examining the likelihood of
STEM degree choice as a function of profile membership; and
investigating personality predictors of interest profile membership.
Profiling attitudes (motivation – engagement – burnout – commitment)
Moeller, Ivcevic, White,
Menges, & Brackett (2018)
Investigating intra-individual engagement-burnout profiles, and
demands-resources profiles.
Gillet, Morin, Sandrin, &
Houle (2018)
Exploring combinations of work engagement and workaholism
levels; and investigate their relations with negative outcomes.
Valero & Hirschi (2016)
Identifying profiles of work-related motivation among adolescents
(i.e., autonomous goals, positive affect, and occupational self-
efficacy). Investigating whether motivational profiles predict
changes in desirable work outcomes.
Howard, Gagne, Morin, & Van
den Broeck (2016)
Identifying the simultaneous occurrence of multiple motivation
types within individual workers. Investigating the relationship
between motivation profile and work performance.
Graves, Cullen, Lester,
Ruderman, & Gentry (2015)
Identifying managers' motivational profiles based on four
motivational types delineated by self-determination theory (i.e.,
external, introjected, identified, intrinsic). To test a model of the
57
antecedents (i.e., perceived supervisor support and organizational
politics) and consequences (i.e., work attitudes and promotability)
of these profiles.
Meyer, Morin, &
Vandenberghe (2015)
Investigating profiles of (affective, normative, continuance)
commitment to two interrelated targets, the organization and
supervisor.
Lopez, McDermott, & Fons-
Scheyd (2014)
Identifying clusters of participants with distinct profiles of multiple
role planning attitudes; and investigating the well-being outcomes
of profile-membership.
Stanley, Vandenberghe,
Vandenberg, & Bentein (2013)
Identifying profiles of commitment based on combinations of
affective, normative, perceived sacrifice, and few alternative
commitments. Investigating how these profiles determine turnover
intention and turnover.
Meyer, Stanley, & Parfyonova
(2012)
Identifying distinct profiles of (affective, normative, continuance)
commitment and how these relate to a range of outcomes (i.e., need
satisfaction, regulation, affect, engagement, organizational
citizenship behavior, and well-being).
Profiling personality (traits – adaptability)
Barbaranelli, Fida, Paciello, &
Tramontano (2018)
Combining work self-efficacy dimensions into different patterns;
investigating whether self-efficacy profiles associate with different
levels of adjustment.
Perera & McIlveen (2017)
Identifying distinct profiles of adaptivity based on combinations of
the Big-Five personality dimensions. Linking these adaptivity
profiles with adapting and adaptation outcomes.
58
Hirschi & Valero (2015)
Identifying subgroups with distinct adaptability profiles in terms of
concern, control, curiosity and confidence. Exploring the
relationship between the various adaptability profiles and adapting
(career planning, career decision-making difficulties, career
exploration, and occupational self-efficacy beliefs) and adaptivity
(core self-evaluations and proactivity).
Rice, Ray, Davis, DeBlaere, &
Ashby (2015)
Identifying different types of perfectionism (i.e., adaptive,
maladaptive, and nonperfectionist).
Rice, Lopez, & Richardson
(2013)
Creating profiles based on measures of perfectionism and
personality (i.e., conscientiousness, neuroticism) and investigate
how this relates to STEM performance.
Profiling behaviors
Mäkikangas (2018)
Examining whether discernable profiles can be identified based on
scores on four job crafting behaviors, and if so, whether such
profiles differ in relation to work engagement.
Profiling the (perceived) environment
Hirschi & Valero (2017)
Identifying qualitatively differing profiles according to levels of
perceived chance events and career decidedness. Investigating
whether these groups differ in work motivation (i.e., occupational
self-efficacy beliefs, perceived person-job fit, and work
engagement).
Dahling, Gabriel, &
MacGowan (2017)
Identifying profiles of feedback environment perceptions; link these
profiles to antecedents grounded in social exchange theory; and test
59
the relations of these profiles with important feedback environment
criteria.
Note: The search term was “latent profile”
60
Table A4: Overview of studies in research on careers, career counseling and/or vocational
behavior that used factor mixture analysis, along with the substantive question they wanted to
address
Hyvönen, Räikkönen, Feldt,
Mauno, Dragano, & Matthewman
(2017)
Identifying different reward patterns on the basis of perceived and
objective career rewards (i.e., career stability and promotions)
across four measurements and investigating the impact of long-term
rewards patterns on contents of personal work goals.
Leuty, Hansen, & Speaks (2016)
Identifying groups of college students with similar profiles of vocational
and leisure interests.
McLarnon, Carswell, &
Schneider (2015)
Identifying qualitatively and quantitatively distinct subgroups or
types of individuals differentiated on the basis of interests in the
RIASEC variables.
Deemer, Lin, Graham, & Soto
(2009)
Identifying subgroups of STEM-students on the base of their
vulnerabilities and affective responses to threatening stereotypes.
Note: The search term was “factor mixture”
61
Table A5: Overview of studies in research on careers, career counseling and/or vocational
behavior that used mixture regression analysis, along with the substantive question they wanted to
address
Gillet, Morin, Sandrin, &
Houle (2018)
Exploring combinations of work engagement and workaholism
levels; and investigating their relations with negative outcomes.
Chénard-Poirier, Morin, &
Boudrias (2017)
Exploring patterns of relations among three leadership
empowerment practices (i.e., delegation, coaching, and recognition)
and five indicators of behavioral empowerment
Hofmans, De Gieter &
Pepermans (2013)
Identifying types of individuals based on different job reward-job
satisfaction relationships. Investigating whether these different
person types have differential associations with turnover intention
and organizational commitment.
Note: Search terms were “mixture regression”, “clusterwise regression”, and “latent class regression”
62
Table A6: Overview of studies in research on careers, career counseling and/or vocational
behavior that used configural frequency analysis, along with the substantive question they wanted to
address
Moeller, Ivcevic, White,
Menges, & Brackett (2018)
Examining associations between demands-resources profiles and
engagement-burnout profiles.
Reitzle & Vondracek (2000)
Identifying patterns of (categorical) career and family
characteristics (e.g., marital status, completion of training, history of
unemployment, etc.).
Note: The search term was “configural frequency”
63
Table A7: Overview of studies in research on careers, career counseling and/or vocational
behavior that used Davison and Davenport’s (2002) criterion-based method, along with the
substantive question they wanted to address
Wiernik (2016)
Identifying patterns in the predictive relationships between
personality traits and Realistic vocational interests
Perry (2008)
Identifying a predictive profile of vocational exploration and racial
identity for behavioral and psychological factors of school
engagement.
Note: We searched for articles that referred to the Davison and Davenport (2002) article and also looked for the terms
“criterion profile” and “criterion pattern”
64
Table A8: Overview of studies in research on careers, career counseling and/or vocational
behavior that used growth mixture modeling, along with the substantive question they wanted to
address
Modeling career trajectories and success
Huang, Evans, Hara, Weiss, &
Hser (2011)
Identifying employment trajectory groups and the impact of gender
and drug use.
Zwaan, ter Bogt, &
Raaijmakers (2010)
Identifying groups of musicians with different career patterns; and
investigating how career success was influenced by social support,
professional attitude and professional network.
Modeling vocational identity trajectories
Hirschi (2011c)
Identifying distinct developmental trajectories of career-choice
readiness in adolescents; and investigating the impact of
environmental demands and individual differences on these
developmental trends.
Modeling performance trajectories
Miraglia, Alessandri, &
Borgogni (2015)
Identifying trajectory classes of job performance based on repeated
supervisory ratings and data on employees’ organizational tenure
and self-efficacy.
Modeling attitude trajectories (motivation – engagement)
Gillet, Morin, Huart, Odry,
Chevalier, Coillot, &
Fouquereau (2018)
Identifying distinct trajectories of self-determined motivation for a
vocational training, and investigating the implications of these
65
trajectories for a variety of outcomes (i.e., positive and negative
affect, and performance).
Upadyaya & Salmela-Aro
(2015)
Identifying latent trajectory groups based on repeated assessments
of career engagement and satisfaction among young adults.
Rice, Ray, Davis, DeBlaere, &
Ashby (2015)
Identifying distinctly low, moderate, and high patterns of academic
stress over the year.
Note: Search terms were “growth mixture”, and “latent class growth”
66
Table A9: Overview of studies in research on careers, career counseling and/or vocational
behavior that used latent transition analysis, along with the substantive question they wanted to
address
Mäkikangas (2018)
Identifying profiles of job crafters based on self-report crafting
behaviors; and investigating whether employees maintain their
profile membership over time.
Kunst, van Woerkom, van
Kollenburg & Poell (2018)
Identifying goal orientation profiles; evaluate their stability over
time; and assessing the impact of managerial coaching behavior
change in employees’ goal orientation profiles.
Rice, Ray, Davis, DeBlaere, &
Ashby (2015)
Examining the relationships between different types of
perfectionism and the experience of stress levels across time.
Note: The search term was “latent transition”
67
References in Appendices
Akos, P., Konold, T., & Niles, S. G. (2004). A career readiness typology and typal membership in
middle school. Career Development Quarterly, 53(1), 53-66.
Argyropoulou, E. P., Sidiropoulou-Makakou, D., & Besevegis, E. G. (2007). Generalized self-
efficacy, coping, career indecision, and vocational choices of senior high school students in
Greece - Implications for career guidance practitioners. Journal of Career Development, 33(4),
316-337.
Armstrong, P. I., Rounds, J., & Hubert, L. (2008). Re-conceptualizing the past: Historical data in
vocational interest research. Journal of Vocational Behavior, 72(3), 284-297.
Armstrong, P. I., Smith, T. J., Donnay, D. A., & Rounds, J. (2004). The strong ring: A basic interest
model of occupational structure. Journal of Counseling Psychology, 51(3), 299-313.
Armstrong, P. I., & Vogel, D. L. (2009). Interpreting the interest–efficacy association from a
RIASEC perspective. Journal of Counseling Psychology, 56(3), 392-407.
Barbaranelli, C., Fida, R., Paciello, M., & Tramontano, C. (2018). 'Possunt, quia posse videntur':
They can because they think they can. Development and validation of the Work Self-Efficacy
scale: Evidence from two studies. Journal of Vocational Behavior, 106, 249-269.
Bruce, R. A., & Scott, S. G. (1994). Varieties and commonalities of career transitions: Louis
typology revisited. Journal of Vocational Behavior, 45(1), 17-40.
Capitano, J., DiRenzo, M. S., Aten, K. J., & Greenhaus, J. H. (2017). Role identity salience and
boundary permeability preferences: An examination of enactment and protection effects.
Journal of Vocational Behavior, 102, 99-111.
Chénard-Poirier, L. A., Morin, A. J. S., & Boudrias, J. S. (2017). On the merits of coherent
leadership empowerment behaviors: A mixture regression approach. Journal of Vocational
Behavior, 103, 66-75.
Cinamon, R. G., & Rich, Y. (2002). Profiles of attribution of importance to life roles and their
68
implications for the work–family conflict. Journal of Counseling Psychology, 49(2), 212-220.
Claes, R., & Quintanilla, S. A. R. (1994). Initial career and work meanings in 7 European countries.
Career Development Quarterly, 42(4), 337-352.
Cohen, C. R., Chartrand, J. M., & Jowdy, D. P. (1995). Relationships between career indecision
subtypes and ego identity development. Journal of Counseling Psychology, 42, 440-447.
Dahling, J. J., Gabriel, A. S., & MacGowan, R. (2017). Understanding typologies of feedback
environment perceptions: A latent profile investigation. Journal of Vocational Behavior, 101,
133-148.
Deemer, E. D., Lin, C., Graham, R., & Soto, C. (2016). Development and validation of a measure of
threatening gender stereotypes in science: A factor mixture analysis. Journal of Career
Assessment, 24(1), 145-161.
De Fruyt, F. (2002). A person-centered approach to P-E fit questions using a multiple-trait model.
Journal of Vocational Behavior, 60(1), 73-90.
Einarsdottir, S., Eyjolfsdottir, K. O., & Rounds, J. (2013). Development of indigenous basic interest
scales: Re-structuring the Icelandic interest space. Journal of Vocational Behavior, 82(2), 105-
115.
Ferguson, S. L., & Hull, D. M. (2019). Exploring science career interest: Latent profile analysis of
high school occupational preferences for science. Journal of Career Development, 46(5), 583-
598.
Gati, I., Krausz, M., & Osipow, S. H. (1996). A taxonomy of difficulties in career decision making.
Journal of Counseling Psychology, 43(4), 510-526.
Gerber, M., Wittekind, A., Grote, G., & Staffelbach, B. (2009). Exploring types of career orientation:
A latent class analysis approach. Journal of Vocational Behavior, 75(3), 303-318.
Gillet, N., Morin, A. J. S., Huart, I., Odry, D., Chevalier, S., Coillot, H., & Fouquereau, E. (2018).
Self-determination trajectories during police officers' vocational training program: A growth
69
mixture analysis. Journal of Vocational Behavior, 109, 27-43.
Gillet, N., Morin, A. J. S., Sandrin, E., & Houle, S. A. (2018). Investigating the combined effects of
workaholism and work engagement: A substantive-methodological synergy of variable-centered
and person-centered methodologies. Journal of Vocational Behavior, 109, 54-77.
Gore, P. A., Bobek, B. L., Robbins, S. B., & Shayne, L. (2006). Computer-based career exploration:
Usage patterns and a typology of users. Journal of Career Assessment, 14(4), 421-436.
Graves, L. M., Cullen, K. L., Lester, H. F., Ruderman, M. N., & Gentry, W. A. (2015). Managerial
motivational profiles: Composition, antecedents, and consequences. Journal of Vocational
Behavior, 87, 32-42.
Gustafson, S. B., & Mumford, M. D. (1995). Personal style and person-environment fit: A pattern
approach. Journal of Vocational Behavior, 46(2), 163-188.
Haines III, V. Y., Doray-Demers, P., & Martin, V. (2018). Good, bad, and not so sad part-time
employment. Journal of Vocational Behavior, 104, 128-140.
Hamilton, M. K. (1977). Graduate training and professional identity. The Counseling Psychologist,
7(2), 26-29.
Hansen, J. I. C., & Scullard, M. G. (2002). Psychometric evidence for the Leisure Interest
Questionnaire and analyses of the structure of leisure interests. Journal of Counseling
Psychology, 49(3), 331-341.
Hechtlinger, S., Levin, N., & Gati, I. (2019). Dysfunctional career decision-making beliefs: A
multidimensional model and measure. Journal of Career Assessment, 27(2), 209-229.
Hirschi, A. (2011a). Callings in career: A typological approach to essential and optional components.
Journal of Vocational Behavior, 79(1), 60-73.
Hirschi, A. (2011b). Relation of vocational identity statuses to interest structure among Swiss
adolescents. Journal of Career Development, 38(5), 390-407.
Hirschi, A. (2011c). Career-choice readiness in adolescence: Developmental trajectories and
70
individual differences. Journal of Vocational Behavior, 79(2), 340-348.
Hirschi, A., & Valero, D. (2015). Career adaptability profiles and their relationship to adaptivity and
adapting. Journal of Vocational Behavior, 88, 220-229.
Hirschi, A., & Valero, D. (2017). Chance events and career decidedness: Latent profiles in relation to
work motivation. Career Development Quarterly, 65(1), 2-15.
Hofmans, J., De Gieter, S., & Pepermans, R. (2013). Individual differences in the relationship
between satisfaction with job rewards and job satisfaction. Journal of Vocational Behavior,
82(1), 1-9.
Howard, J., Gagne, M., Morin, A. J. S., & Van den Broeck, A. (2016). Motivation profiles at work:
A self-determination theory approach. Journal of Vocational Behavior, 95-96, 74-89.
Huang, D. Y. C., Evans, E., Hara, M., Weiss, R. E., & Hser, Y. I. (2011). Employment trajectories:
Exploring gender differences and impacts of drug use. Journal of Vocational Behavior, 79(1),
277-289.
Hyvönen, K., Räikkönen, E., Feldt, T., Mauno, S., Dragano, N., & Matthewman, L. (2017). Long-
term reward patterns contribute to personal goals at work among Finnish managers. Journal of
Career Development, 44(5), 394-408.
Johnson, W., & Bouchard, T. J. (2009). Linking abilities, interests, and sex via latent class analysis.
Journal of Career Assessment, 17(1), 3-38.
Kelly, K. R., & Lee, W. C. (2002). Mapping the domain of career decision problems. Journal of
Vocational Behavior, 61(2), 302-326.
Kossek, E. E., Ruderman, M. N., Braddy, P. W., & Hannum, K. M. (2012). Work-nonwork boundary
management profiles: A person-centered approach. Journal of Vocational Behavior, 81(1), 112-
128.
Kunst, E. M., van Woerkom, M., van Kollenburg, G. H., & Poell, R. F. (2018). Stability and change
in teachers' goal orientation profiles over time: Managerial coaching behavior as a predictor of
71
profile change. Journal of Vocational Behavior, 104, 115-127.
Kuron, L. K. J., Schweitzer, L., Lyons, S., & Ng, E. S. W. (2016). Career profiles in the "new
career": evidence of their prevalence and correlates. Career Development International, 21(4),
355-377.
Larson, L. M., Heppner, P. P., Ham, T., & Dugan, K. (1988). Investigating multiple subtypes of
career indecision through cluster analysis. Journal of Counseling Psychology, 35(4), 439-446.
Larson, L. M., & Majors, M. S. (1998). Applications of the coping with career indecision instrument
with adolescents. Journal of Career Assessment, 6(2), 163-179.
Leuty, M. E., Hansen, J. I. C., & Speaks, S. Z. (2016). Vocational and leisure interests: A profile-
level approach to examining interests. Journal of Career Assessment, 24(2), 215-239.
Lopez, F. G., McDermott, R. C., & Fons-Scheyd, A. L. (2014). Profiling the multiple role planning
attitudes of college women. Journal of Career Assessment, 22(4), 700-714.
Lucas, M. S., & Epperson, D. L. (1990). Types of vocational undecidedness: A replication and
refinement. Journal of Counseling Psychology, 37(4), 382-388.
Luyckx, K., Duriez, B., Klimstra, T. A., & De Witte, H. (2010). Identity statuses in young adult
employees: Prospective relations with work engagement and burnout. Journal of Vocational
Behavior, 77(3), 339-349.
Maher, L. P., Gallagher, V. C., Rossi, A. M., Ferris, G. R., & Perrewe, P. L. (2018). Political skill
and will as predictors of impression management frequency and style: A three-study
investigation. Journal of Vocational Behavior, 107, 276-294.
Majeed, T., Forder, P., Mishra, G., Kendig, H., & Byles, J. (2015). A gendered approach to
workforce participation patterns over the life course for an Australian baby boom cohort.
Journal of Vocational Behavior, 87, 108-122.
Mäkikangas, A. (2018). Job crafting profiles and work engagement: A person-centered approach.
Journal of Vocational Behavior, 106, 101-111.
72
Matthews, E., & Tiedeman, D. V. (1964). Attitudes toward career and marriage and the development
of life style in young women. Journal of Counseling Psychology, 11(4), 375-384.
McLarnon, M. J. W., Carswell, J. J., & Schneider, T. J. (2015). A case of mistaken identity? Latent
profiles in vocational interests. Journal of Career Assessment, 23(1), 166-185.
Meldahl, J. M., & Muchinsky, P. M. (1997). The neurotic dimension of vocational indecision:
Gender comparability? Journal of Career Assessment, 5(3), 317-331.
Meyer, J. P., Morin, A. J. S., & Vandenberghe, C. (2015). Dual commitment to organization and
supervisor: A person-centered approach. Journal of Vocational Behavior, 88, 56-72.
Meyer, J. P., Stanley, L. J., & Parfyonova, N. M. (2012). Employee commitment in context: The
nature and implications of commitment profiles. Journal of Vocational Behavior, 80, 1-16.
Miraglia, M., Alessandri, G., & Borgogni, L. (2015). Trajectory classes of job performance: The role
of self-efficacy and organizational tenure. Career Development International, 20(4), 424-442.
Moeller, J., Ivcevic, Z., White, A. E., Menges, J. I., & Brackett, M. A. (2018). Highly engaged but
burned out: intra-individual profiles in the US workforce. Career Development International,
23(1), 86-105.
Moen, P., Kelly, E., & Huang, Q. L. (2008). Work, family and life-course fit: Does control over
work time matter? Journal of Vocational Behavior, 73(3), 414-425.
Moran, C. M., Diefendorff, J. M., Kim, T. Y., & Liu, Z. Q. (2012). A profile approach to self-
determination theory motivations at work. Journal of Vocational Behavior, 81(3), 354-363.
Multon, K. D., Wood, R., Heppner, M. J., & Gysbers, N. C. (2007). A cluster-analytic investigation
of subtypes of adult career counseling clients: Toward a taxonomy of career problems. Journal
of Career Assessment, 15(1), 66-86.
Niles, S. G., Anderson, W. P., & Goodnough, G. (1998). Exploration to foster career development.
Career Development Quarterly, 46(3), 262-275.
Oztemel, K. (2013). Testing the validity of the emotional and personality-related career decision-
73
making difficulties questionnaire in Turkish culture. Journal of Career Development, 40(5),
390-407.
Paixão, O., & Gamboa, V. (2017). Motivational profiles and career decision making of high school
students. Career Development Quarterly, 65(3), 207-221.
Perera, H. N., & McIlveen, P. (2018). Vocational interest profiles: Profile replicability and relations
with the STEM major choice and the Big-Five. Journal of Vocational Behavior, 106, 84-100.
Perera, H. N., & McIlveen, P. (2017). Profiles of career adaptivity and their relations with
adaptability, adapting, and adaptation. Journal of Vocational Behavior, 98, 70-84.
Perry, J. C. (2008). School engagement among urban youth of color - Criterion pattern effects of
vocational exploration and racial identity. Journal of Career Development, 34(4), 397-422.
Pickering, G. S., & Galvin-Schaefers, K. (1988). An empirical study of reentry women. Journal of
Counseling Psychology, 35(3), 298-303.
Poynton, T. A., Lapan, R. T., & Marcotte, A. M. (2015). Financial planning strategies of high school
seniors: Removing barriers to career success. Career Development Quarterly, 63(1), 57-73.
Reitzle, M., & Vondracek, F. W. (2000). Methodological avenues for the study of career pathways.
Journal of Vocational Behavior, 57(3), 445-467.
Rhee, E., Lee, B. H., Kim, B., Ha, G., & Lee, S. M. (2016). The relationship among the six
vocational identity statuses and five dimensions of planned happenstance career skills. Journal
of Career Development, 43(4), 368-378.
Rice, K. G., Lopez, F. G., & Richardson, C. M. E. (2013). Perfectionism and performance among
STEM students. Journal of Vocational Behavior, 82(2), 124-134.
Rice, K. G., Ray, M. E., Davis, D. E., DeBlaere, C., & Ashby, J. S. (2015). Perfectionism and
longitudinal patterns of stress for STEM majors: Implications for academic performance.
Journal of Counseling Psychology, 62(4), 718-731.
Rochlen, A. B., Milburn, L., & Hill, C. E. (2004). Examining the process and outcome of career
74
counseling for different types of career counseling clients. Journal of Career Development,
30(4), 263-275.
Rojewski, J. W. (1994). Career indecision types for rural adolescents from disadvantaged and
nondisadvantaged backgrounds. Journal of Counseling Psychology, 41(3), 356-363.
Santos, P. J., & Ferreira, J. A. (2012). Career decision statuses among Portuguese secondary school
students: A cluster analytical approach. Journal of Career Assessment, 20(2), 166-181.
Savickas, M. L., & Jarjoura, D. (1991). The Career Decision Scale as a type indicator. Journal of
Counseling Psychology, 38(1), 85-90.
Segers, J., Inceoglu, I., Vloeberghs, D., Bartram, D., & Henderickx, E. (2008). Protean and
boundaryless careers: A study on potential motivators. Journal of Vocational Behavior, 73(2),
212-230.
Sestito, L. A., Sica, L. S., Ragozini, G., Porfeli, E., Weisblat, G., & Di Palma, T. (2015). Vocational
and overall identity: A person-centered approach in Italian university students. Journal of
Vocational Behavior, 91, 157-169.
Shimizu, A. B., Dik, B. J., & Conner, B. T. (2019). Conceptualizing calling: Cluster and taxometric
analyses. Journal of Vocational Behavior, 114, 7-18.
Shivy, V. A., Rounds, J., & Jones, L. E. (1999). Applying vocational interest models to naturally
occurring occupational perceptions. Journal of Counseling Psychology, 46(2), 207-217.
Smith, T. J., & Campbell, C. (2006). The structure of ONET occupational values. Journal of Career
Assessment, 14(4), 437-448.
Solberg, V. S. H., Carlstrom, A. H., Howard, K. A. S., & Jones, J. E. (2007). Classifying at-risk high
school youth: The influence of exposure to community violence and protective factors on
academic and health outcomes. Career Development Quarterly, 55(4), 313-327.
Stanley, L., Vandenberghe, C., Vandenberg, R., & Bentein, K. (2013). Commitment profiles and
employee turnover. Journal of Vocational Behavior, 82(3), 176-187.
75
Stratton, T. D., Witzke, D. B., Elam, C. L., & Cheever, T. R. (2005). Learning and career specialty
preferences of medical school applicants. Journal of Vocational Behavior, 67(1), 35-50.
Tay, L., Su, R., & Rounds, J. (2011). People–things and data–ideas: Bipolar dimensions? Journal of
Counseling Psychology, 58(3), 424-440.
Tracey, T. J., & Darcy, M. (2002). An idiothetic examination of vocational interests and their
relation to career decidedness. Journal of Counseling Psychology, 49(4), 420-427.
Tsoumbris, P., & Xenikou, A. (2010). Commitment profiles: The configural effect of the forms and
foci of commitment on work outcomes. Journal of Vocational Behavior, 77(3), 401-411.
Upadyaya, K., & Salmela-Aro, K. (2015). Development of early vocational behavior: Parallel
associations between career engagement and satisfaction. Journal of Vocational Behavior, 90,
66-74.
Valero, D., & Hirschi, A. (2016). Latent profiles of work motivation in adolescents in relation to
work expectations, goal engagement, and changes in work experiences. Journal of Vocational
Behavior, 93, 67-80.
Van Aerden, K., Moors, G., Levecque, K., & Vanroelen, C. (2015). The relationship between
employment quality and work-related well-being in the European Labor Force. Journal of
Vocational Behavior, 86, 66-76.
Vieira, J. M., Matias, M., Lopez, F. G., & Matos, P. M. (2018). Work-family conflict and
enrichment: An exploration of dyadic typologies of work-family balance. Journal of Vocational
Behavior, 109, 152-165.
Vilhjalmsdottir, G., & Arnkelsson, G. B. (2013). Social aspects of career choice from the perspective
of habitus theory. Journal of Vocational Behavior, 83(3), 581-590.
Viola, M. M., Musso, P., Inguglia, C., & Lo Coco, A. (2016). Psychological well-being and career
indecision in emerging adulthood: The moderating role of hardiness. Career Development
Quarterly, 64(4), 387-396.
76
Wanberg, C. R., & Muchinsky, P. M. (1992). A typology of career decision status: Validity
extension of the vocational decision status model. Journal of Counseling Psychology, 39(1), 71-
80.
Wasti, S. A. (2005). Commitment profiles: Combinations of organizational commitment forms and
job outcomes. Journal of Vocational Behavior, 67(2), 290-308.
Wiernik, B. M. (2016). Intraindividual personality profiles associated with realistic interests. Journal
of Career Assessment, 24(3), 460-480.
Woo, S. E. (2011). A study of Ghiselli's hobo syndrome. Journal of Vocational Behavior, 79(2),
461-469.
Zhou, S., Leung, S. A., & Li, X. (2012). The meaning of work among Chinese university students:
Findings from prototype research methodology. Journal of Counseling Psychology, 59(3), 408-
423.
Zwaan, K., ter Bogt, T. F. M., & Raaijmakers, Q. (2010). Career trajectories of Dutch pop musicians:
A longitudinal study. Journal of Vocational Behavior, 77(1), 10-20.
