Motivation Profiles at Work 1: Acknowledgements

MOTIVATION PROFILES AT WORK 1
Running Head: MOTIVATION PROFILES AT WORK
Motivation Profiles at Work: A Self-Determination Theory Approach
Joshua Howard and Marylène Gagné
University of Western Australia
Alexandre J.S. Morin
Institute for Positive Psychology and Education, Australian Catholic University
Anja Van den Broeck
KU Leuven & North-West University
Howard, J., Gagné, M., Morin, A. J. S., Van den Broeck, A. (2016). Motivation Profiles at Work: A
Self-Determination Theory Approach. Journal of Vocational Behavior. xx(x). p. xx
Acknowledgements
Preparation of this article was facilitated through funds from the Australian Postgraduate Award
scholarship awarded to the first author, as well as grants from the Fonds Québécois de la Recherche
sur la Société et la Culture and the Social Sciences and Humanities Research Council of Canada
awarded to the second author, and a grant from the Australian Research Council (LP140100100)
awarded to the third author.
Corresponding author:
Joshua Howard,
School of Psychology, University of Western Australia,
35 Stirling Hwy, Crawley 6009 WA, Australia.
Email: [email protected]
Abstract
Self-determination theory proposes that individuals experience distinct types of motivation to varying
degrees. While it is well documented that these types of motivation differentially predict outcomes,
very little attention has been paid to how they interact within individuals. The current study addresses
the simultaneous occurrence of multiple motivations types within individual workers by adopting a
person-centered approach on two samples of employees from different countries (n = 723 & 286).
Four very similar motivation profiles were found across samples, representing amotivated, externally
regulated, autonomously regulated and highly motivated employees. In Sample 1, governmental
employees presented a greater likelihood of membership into the least desirable amotivated profile,
whereas white-collar employees presented a greater likelihood of membership in the highly motivated
profile. In Sample 2, autonomously and highly motivated profiles showed superior work performance
and higher levels of wellbeing, while the amotivated profile fared the worst. The presence of external
regulation in a profile appears unimportant when combined with autonomous forms of motivation, and
detrimental to outcomes in the absence of autonomous forms of motivation. These results support the
hypothesis that autonomous forms of motivation are far more important in promoting positive
workplace outcomes than more controlling forms.

Motivation, generally defined as the energy, direction and persistence of behavior (Pinder, 1998),
is an inherently complex concept as evidenced by the variety of approaches to its conceptualization
and measurement. Self-determination Theory (SDT; Deci & Ryan, 1985) offers a well-supported
conceptualization which proposes that motivation is best represented by conceptually distinct, yet
complementary, types of behavioral regulations experienced by individuals to varying degrees. While
it is now well documented that these types of regulation differentially predict outcomes (e.g., Koestner
& Losier, 2002), very little attention has been paid to how they interact within individuals. The current
study addresses the simultaneous occurrence of multiple behavioral regulations within individual
workers by adopting a person-centered approach to work motivation. While variable-centered analyses,
which have dominated the field so far, have been extremely useful in their own right, the complexity
of interactions between numerous types of motivation cannot easily be examined using traditional
regression techniques, which become almost impossible to interpret when more than three interacting
variables are simultaneously considered. No such limit exists when person-centered analyses are used
to assess how configurations of motivation factors are organized within individuals.
This shift to a person-centered strategy is more than just a shift in methods. It involves a
fundamentally different way of thinking about motivation which may affect the design of interventions
(Zyphur, 2009). When conceptualizing types of motivation as variables, we are not thinking about a
whole person, but about one of the many components that make up a person’s motivational profile.
Resulting interventions are designed to increase one type of motivation (e.g., intrinsic) without taking
into consideration how the intervention will impact the other types of motivation (e.g., extrinsic). Such
an omission may well make interventions less effective. In contrast, the person-centered approach
takes into account the interplay between a person’s motives, and consequently may lead to
interventions aiming to influence the person’s whole motivational profile. This is likely to produce
better tailored and cost efficient interventions for particular subpopulations of employees (Morin &
Marsh, 2015). In practice, this approach would make SDT more compatible with how people in
positions of authority, such as managers, actually think about the motivation of their employees
(Morin, Morizot, Boudrias, & Madore, 2011; Zyphur, 2009).
As reviewed below, a few attempts have been made to conceptualize work motivation profiles.
The present study, however, does so more comprehensively by: (a) including all types of regulation
proposed by SDT (unlike Van den Broeck, Lens, De Witte, & Van Coillie, 2013), (b) using two large
heterogeneous samples of workers from two countries (unlike Graves, Cullen, Lester, Ruderman, &
Gentry, 2015) and, (c) utilizing the latest advances in latent profile analysis (unlike Moran,
Diefendorff, Kim, & Liu, 2013 and Van den Broeck et al., 2013). As such it represents an incremental
advancement in this area of research and potentially provides a more accurate representation of the
types of profiles that are likely to be found in the work domain. Furthermore, it extends previous
research by demonstrating how the relative frequency of the profiles differs across job categories
(white collar, blue collar, governmental), and the relation between the profiles and a variety of
outcomes, including in-role and extra-role performance, engagement, burnout, and job satisfaction.
Self-Determination Theory
SDT conceptualizes motivation as multiple distinguishable facets, each representing a different
form of behavioral regulation, and assumed to follow a continuum of self-determination (Deci & Ryan,
1985; Gagné & Deci, 2005). At one extreme, intrinsic motivation occurs when an individual
participates in an activity for the enjoyment inherent in the activity itself, while at the other extreme
extrinsic motivation occurs when behaviors are enacted for an instrumental reason. SDT proposes that
extrinsic motivation can be internalized to become autonomously regulated. Identified regulation, an
internalized form of extrinsic motivation, occurs when an individual elects to act because the behavior
or the outcome of the behavior is of personal significance. Identified regulation and intrinsic
motivation, are autonomous forms of motivation, while the next two regulations are controlled forms
motivation. Introjected regulation, an internalized yet controlled form of extrinsic motivation, occurs
when behaviors are undertaken in order to avoid negative self-feelings such as shame, or to attain
positive self-feelings such as pride. External regulation, a non-internalized form of extrinsic
motivation lying at the lower end of the continuum, occurs when behaviors are undertaken for
externally derived rewards or punishments. The most current conceptualization of workplace
motivation suggests that external regulation is best described through two components, external-social,
and external-material (Gagné et al., 2015). External-social regulation is characterized by the desire to
gain approval or respect from others, or to avoid criticism, whereas external-material regulation
focuses on material rewards, and the avoidance of losing one’s job.
Finally, amotivation is the absence of any desire to exert effort. Amotivation has been defined as
a state in which individuals do not associate a behavior with subsequent outcomes, and as such,
behaviors are executed for reasons unknown or not executed at all (Deci & Ryan, 1985). Accordingly,
amotivated individuals are likely to feel detached from their actions, or may feel a lack of control over
their present situation or behavior, and will therefore invest little time or energy towards such
behaviors. This state was shown to be associated with a wide range of negative workplace outcomes
including lower vitality, job satisfaction, affective commitment, adaptivity, proactivity, and job effort,
as well as greater emotional exhaustion, burnout, and turnover intention (Gagné et al., 2015; Tremblay,
Blanchard, Taylor, Pelletier, & Villeneuve, 2009). Thus, given that people are still enacting work
behaviors despite their lack of motivation, and considering the notable negative consequences
associated with amotivated behavior, it is our contention that amotivation is an important feature of the
self-determination continuum to consider.
In addition to the empirical evidence demonstrating the negative influence of amotivation on
performance and wellbeing, on a more theoretical point, a complete depiction of the continuum of
motivation should not only include a variety of motives for engaging in specific behaviors (ranging
from the intrinsic pleasure to external constraints) but also the complete lack of motive to engage in
these behaviors (which forms the opposite pole of the self-determination continuum). This
representation of the SDT continuum has been recently supported in the work area by a recent study
by Howard, Gagné, Morin and Forest (2016), in which it was found that amotivation is located along
the same continuum as the behavioral regulations, with no evidence of discontinuity.
While there is ongoing debate concerning the presence of this continuum beyond a mere heuristic
tool (Chemolli & Gagné 2014), this research will examine whether the pattern of regulations expected
from this continuum hypothesis is present in employee profiles. Specifically, support for the
continuum hypothesis would be demonstrated if profiles follow a smooth increase/decrease in the
level of the different regulations as a function of their position on the continuum. Alternatively, weak
support would be found through the presence of profiles in which people experience similar levels of
regulations assumed to be located at opposite poles of the continuum (e.g., intrinsic and external
regulations; Grolnick & Ryan, 1987).
So far, substantial research has examined how these regulations relate to various antecedents and
outcomes. Results generally demonstrate that intrinsic motivation and identified regulation yield more
positive outcomes, such as productivity and retention, than introjected and external regulations (Gagné,
2014; Gagné & Deci, 2005), though some research has found differences in the effects of intrinsic
versus identified regulation, and in the effects of introjected versus external regulation (Gagné et al.,
2015; Koestner & Losier, 2002). This approach does not take into account the multidimensional nature
of motivation, and the fact that workers may simultaneously endorse multiple reasons for doing their
job. Moreover, this research does not shed light on how distinct motivational regulations interact in
predicting outcomes. What happens when employees are motivated for both autonomous and
controlled reasons, compared to employees who are only motivated for autonomous reasons? For
instance, is it more important to have a high level of overall motivation or is the proportion of
autonomous to controlled motivation more influential? How do amotivated employees compare to
employees presenting controlled motivation? How combinations of specific regulations relate to key
outcomes also remains unknown, and essentially unexplored because of the heavy reliance on
variable-centered methods. Indeed, the complexity of interactions required to fully describe motivation
(i.e., involving six interacting types of motivation) calls for the adoption of a person-centered
approach. In response, the aims of this study are to establish which motivational profiles are most
likely to emerge in the work domain and to examine predictors and outcomes of profile membership.
Motivational Profiles
Few studies have applied a person-centered approach to motivation research across domains
(education, sport, work, etc.). Most have used cluster analysis, a method which has been criticized
(e.g., Meyer & Morin, 2016; Morin et al., 2011; Vermunt & Magidson, 2002) as being too sensitive to
the clustering algorithm and measurement scales, as lacking clear guidelines for the selection of an
optimal number of profiles, and as relying on rigid assumptions that do not always hold with real-life
data (i.e., exact assignment of employees to a single profile, conditional independence, equality of the
indicators’ variances across clusters). Furthermore, cluster analytic studies have often relied on small
samples of dubious generalizability (Boiché, Sarrazin, Grouzet, Pelletier, & Chanal, 2008; Gillet,
Berjot, & Paty, 2009; Gillet, Berjot, Vallerand, Amoura, & Rosnet, 2012; Gillet, Vallerand, & Paty,
2013; Gillet, Vallerand, & Rosnet, 2009; McNeill & Wang, 2005).
Motivational profiling has also largely been limited by the dichotomization of motivation into the
broad categories of autonomous and controlled regulations. This dichotomization is a commonly used
practice that simplifies the profiles and makes them easier to estimate, but that also reduces the
richness of potential findings and may hide potentially important configurations. Nonetheless, among
studies using this dichotomization in the educational domain, the observed profiles of academic
motivation have been relatively well replicated, and generally revealed profiles characterized by high
autonomous/low controlled motivation (HA/LC), high autonomous/high controlled motivation
(HA/HC), low autonomous/high controlled motivation (LA/HC), and low autonomous/low controlled
motivation (LA/LC; Hayenga & Corpus, 2010; Liu, Wang, Tan, Koh, & Ee, 2009; Ratelle, Guay,
Vallerand, Larose, & Senècal, 2007; Vansteenkiste, Sierens, Soenens, Luyck, & Lens, 2009). Results
from the sport domain often replicate these profiles with slight variations (e.g. HA/HC, Moderate
Autonomy/LC, HA/MC, MA/HC; Gillet, Vallerand, & Rosnet, 2009; Gillet, Vallerand, & Paty, 2013).
Given the heavy reliance on financial compensation in the work domain, motivational profiles are
likely to differ from those identified in the educational and sport domains, especially when focusing on
a more comprehensive coverage of all types of regulations. This particularity of the work domain
makes it important to look at external and introjected regulations as separate constructs. To our
knowledge, only three studies have examined motivational profiles at work (Graves, Cullen, Lester,
Ruderman, & Gentry, 2015; Moran et al., 2012; Van den Broeck et al, 2013). Van den Broeck et al.
(2013) applied cluster analysis to three samples of employees, collapsing the regulations into a
controlled-autonomous dichotomy, leading to the identification of the same set of four profiles
identified in the education and sport area. In contrast, Moran et al. (2012) applied cluster analysis to
the full range of behavioral regulations. Through this more complete representation, these authors
identified five clusters, most of which differed from those identified in the education and sport domain:
One presenting a moderate levels of motivation across regulation types, one presenting high levels of
motivation across regulation types (corresponding to the HA/HC profile), one representing low levels
of autonomy (low levels of identified and intrinsic motivation) and moderate levels on the other forms
of regulation, one presenting a more self-determined profile (high on introjected, identified and
intrinsic motivation), and one presenting moderate levels on most regulations except for a low level of
introjection. Finally, Graves et al. (2015) identified six latent profiles in a small sample of managers.
These profiles presented similar configurations of motivation (i.e., highest on intrinsic and identified
regulation, followed by introjected, and lowest on external regulation) but different overall levels, so
that one was higher on autonomous than controlled forms of motivation, while another was low on all
forms of regulations. However, this study relied on a relatively small sample of managers, and
provided insufficient information regarding model specification to allow other researchers to replicate
their results or to objectively assess the adequacy of the analyses.
This relative lack of research in the work domain, the dichotomization of regulations into
controlled or autonomous categories, and the reliance on cluster analyses performed on small samples
clearly represent significant limitations of research in this area. In contrast, the present study applied
latent profile analyses (LPA) to the full range of behavioral regulations as they occur in a work context
using large heterogeneous samples of employees from two countries (Canada and Belgium) in order to
derive a common set of work motivation profiles. Employees completed the recently validated
Multidimensional Work Motivation Scale (Gagné et al., 2015), which has been shown to have several
advantageous features (e.g., improved psychometric properties, greater content coverage in terms of
motivation types) compared to traditional measures of work motivation (e.g., Gagné et al., 2010).
In contrast to cluster analyses, LPA is a far more flexible model-based approach to classification
(Muthén, 2002). Being model-based, LPA allows for the estimation of alternative models in which the
restrictive assumptions of cluster analyses can be relaxed. Importantly, LPA aims to find the smallest
number of profiles that can describe associations among a set of continuous variables, relying on a
formal set of objective criteria to guide the identification of the optimal number of latent profiles in the
data. These profiles are called latent because they are prototypical in nature, which means that rather
than forcing each employee to correspond to a single profile, all participants are allocated a probability
of membership in all profiles based on their degree of similarity with each prototypical latent profile.
Due to the scarcity of research on motivational profiles in the work domain, especially of studies
considering the full array of motivation types, it is difficult to specify hypotheses about the nature and
number of expected profiles. Given that previous research has typically found four to six profiles, it
was expected that a relatively small number of profiles (4-6) would be identified, and would represent
not only different levels of overall motivation, but also different shapes, reflecting distinct
combinations of regulation types. Based on previous research, it was also anticipated that a profile
dominated by autonomous forms of regulation, a profile dominated by controlled forms of regulation,
and at least one profile containing both autonomous and controlled forms of regulation would be
identified. While the emergence of different profiles remains possible, in particular across the two
samples considered here, the current study aimed to introduce a broad typology of meaningful profiles
common to most workplaces. However, latent profile analyses suffer from the same limitations as
variable-centered analyses in terms of generalizability and in providing a meaningful representation of
the data (i.e., construct validity). In particular it has been previously argued that the only way to really
support a substantive interpretation of latent profiles is to embark on a process of construct validation
to demonstrate that the identified profiles either meaningfully relate to covariates (predictors, or
outcomes), or can reliably be replicated across samples (Marsh, Lüdtke, Trautwein, & Morin, 2009;
Morin, Morizot, et al., 2011; Muthén, 2003). To address this issue, we tested whether the identified set
of profiles generalized across two Western countries. Furthermore, we assessed the extent to which
these profiles were related to a series of predictors and outcomes to which we now turn our attention.
Predictors of Motivation Profiles
To date little research has examined determinants of employees’ motivation profiles. Among this
limited research, Moran et al. (2012) showed that membership into more autonomously motivated
profiles could be predicted by greater levels of satisfaction of the needs for competence, autonomy,
and relatedness, while Graves et al. (2015) showed that members of the more autonomously motivated
profiles tended to occupy hierarchically higher positions and to report receiving higher levels of
supervisor support. These results are consistent with SDT, which proposes that the satisfaction of these
needs and exposure to work-related context that support their satisfaction, are key determinants of
autonomous motivation (Deci & Ryan, 1985; Gagné & Deci, 2005). As such, it is also to be expected
that job categories allowing for greater levels of need satisfaction may result in a greater proportion of
employees corresponding to predominantly autonomously-driven profiles (i.e., higher in intrinsic
motivation and identified regulation). In particular, research shows that workplace characteristics that
influence need satisfaction, such as job design, participative leadership, and organic (vs. bureaucratic)
structures, tend to be associated with significantly higher levels of autonomous motivation (De
Cooman et al., 2013; Gillet, Gagné, Sauvagère, & Fouquereau, 2012). By this reasoning, it was
expected that manufacturing and other blue-collar industries often characterized by less skill variety,
autonomy, more directive leadership, and hourly wages, would be less likely to satisfy these needs.
For this reason, we expected motivational profiles characterized by lower levels of autonomous
motivation and higher levels of controlled motivation to be more frequent among employees working
in these sectors. In contrast, white-collar employees from the technology sector should be more likely
to experience task variety and participative leadership, which would likely facilitate need satisfaction
(Blais, Brière, Lachance, Riddle, & Vallerand, 1993; Gagné et al., 2010; Gagné, Senécal & Koestner,
1997). Thus, we expected motivational profiles characterized by higher levels of autonomous
motivation and lower levels of controlled motivation to be more frequent among these employees.
Finally, white-collar governmental employees should be more likely to experience highly bureaucratic
job structures, which may stifle motivation, making it more likely for these employees to correspond
to profiles characterized by lower levels of both autonomous and controlled motivation. This study
incorporated blue-collar manufacturing, white-collar technological, and white-collar governmental job
categories as predictors of profile membership to test these hypotheses.
Consequences of Motivation Profiles
Past research has found that profiles characterized by high levels of autonomous motivation seem
to yield better performance outcomes. However, it is less clear how controlled types of motivation
relate to performance. So far, most research conducted regarding the outcomes of motivational profiles
have been conducted in the educational area. This research has shown that the HA/LC profile tends to
be associated with higher levels of academic achievement, as well as lower levels of procrastination,
openness to cheating, and school dropout than the HA/HC profile (Hayenga & Corpus, 2010; Ratelle
et al., 2007; Vansteenkiste et al., 2009). In contrast, the LA/LC and LA/HC profiles both yielded lower
levels of academic achievement and higher levels of procrastination, but did not differ from one
another, indicating that the presence of controlled motivation had negligible effects on performance
(Vansteenkiste et al., 2009). However, additional results suggested that controlled motivation may
actually detract from optimal performance, measured by grade point average and self-perceived skill
acquisition, even when autonomous motivation is also present (Hayenga & Corpus, 2010; Liu, Wang,
Tan, Koh, & Ee, 2009). It thus appears that profile composition, or the ratio of autonomous to
controlled motivation, may represent a stronger predictor of performance outcomes than the simple
overall “quantity” of motivation that characterizes a specific profile.
However, in the work domain, researchers have theorized that some levels of introjected and
external regulation may prove beneficial in predicting positive outcomes (Boiché et al., 2008; Moran
et al., 2012; Van den Broeck et al., 2013). A meta-analysis also found that while intrinsic motivation
was more strongly related to the quality of the work completed, external regulation was more strongly
associated with the quantity of work completed (Cerasoli, Nicklin, & Ford, 2014). Likewise, work
pressure, theorized to foster external regulation, was positively related to the quantity of work effort
and engagement (De Cooman et al., 2013; Van den Broeck, De Cuyper, De Witte, & Vansteenkiste,
2010). The one profile study in the work domain that has examined performance showed that the
HA/LC and HA/HC profiles yielded comparable levels of self-reported in-role performance, and
higher levels than those observed in the LA/HC and LA/LC profiles (Moran et al., 2012).
As suggested above, the quality and quantity of performance may be promoted through different
motivational profiles (Cerasoli et al., 2014). Similarly, required (in-role) and discretionary (extra-role)
performance may also be differentially affected by motivational profiles (Gagné et al., 2015). For
instance, we might expect that profiles characterized by high levels of autonomous types of motivation
would yield greater levels of in-role and extra-role performance, while profiles presenting high levels
of controlled types of motivation would only yield greater levels of in-role performance. The question
is whether controlled types of motivation will stifle extra-role performance, as has been suggested in
some variable-centered research (Battistelli, Galletta, Portoghese, & Vandenberghe, 2013).
Past research also found that profiles characterized by high levels of autonomous motivation yield
better wellbeing outcomes (Van den Broeck et al., 2013). In this situation, unlike what is observed in
the prediction of performance, controlled motivation does not seem to have any advantage in
promoting wellbeing – it even seems to decrease it. In the educational domain, the HA/LC profile was
found to be associated with lower levels of school-related anxiety than the HA/HC profile, while the
LA/LC and LA/HC profiles were associated with the highest levels of school anxiety (Vansteenkiste et
al., 2009). In the work domain, Van den Broeck et al. (2013) and Graves et al. (2015) both found that
HA-HC and HA-LC profiles reported the greatest (and equal) levels of job satisfaction. However,
strain was lower in the HA-LC than in the HA-HC profile; followed by the LA-LC profile. Employees
from the LA-HC profile reported the highest levels of work-related strain. The present study expands
on these studies by the inclusion of work engagement (vigor, dedication, and absorption; Schaufeli &
Bakker, 2003) and burnout (emotional exhaustion, cynicism, and personal inefficacy; Maslach,
Schaufeli, & Leiter, 2001) as potential outcomes of employees’ motivational profiles.
Method
Participants and Procedure
This study incorporated two samples of data collected between 2008 and 2013. Sample 1
consisted of 723 Canadian employees recruited within three different industry sectors: 105 from the
technological sector, 319 from the government sector and 299 from the manufacturing sector (Meanage
= 44.30; Female = 15.8% [54.1% gender info missing]). The subsample of 105 white collar
technology sector employees was previously used in the MMWS validation study (Gagné et al., 2015).
These employees completed surveys containing the original English (n = 178) or French (n = 545)
versions of the MWMS. Sample 2 consisted of 286 Belgian employees (Meanage = 41.66 years;
Female = 57.7%; MeanTenure = 9.39 years) who completed Dutch versions of the outcome measures, in
addition to the Dutch MWMS. In both countries, a variety of organizations were approached with the
possibility to participate in this study of work motivation. These organizations were selected mainly
through a process of convenience based on lead investigators contacts and proximity. Employees from
the organization who agreed to participate had the possibility to complete confidential surveys on an
online platform or in paper format on their work premises. Participation was voluntary.
Measures
A variable specifying job category (e.g., blue collar manufacturing, white collar technology,
white collar governmental) was available only for Sample 1 (n = 723) and was subsequently dummy-
coded in two complementary variables to reflect white collar technology sector employees (1; n = 105)
versus others (0) and governmental employees (1; n = 319) versus others (0).
The MWMS (Gagné et al., 2015) includes 19 items assessing six distinct motivation types. Each
item is an answer to the question “Why do you or would you put effort into your current job?” along a
1 (not at all) to 7 (completely) point Likert scale. Sample items include, “I don’t know why I’m doing
this job, it’s pointless work” (Amotivation; Cronbach’s α = .74 & .87 in samples 1 and 2 respectively),
“Because others will reward me financially only if I put enough effort in my job (e.g., employer,
supervisor...)” (External regulation material; α = .60 & .70), “To get others’ approval (e.g., supervisor,
colleagues, family, clients...)” (External regulation social; α = .78 & .76), “Because otherwise I will
feel ashamed of myself” (Introjected regulation; α = .69 & .71), “Because putting efforts in this job
aligns with my personal values” (Identified regulation; α = .78 & .67), and “Because the work I do is
interesting” (Intrinsic motivation; α = .90 & .88). Validation evidence for the MWMS has
demonstrated a good fit for a six-factor structure, equivalence of the underlying measurement model
across the English, French and Dutch linguistic versions used in the present study, acceptable scale
score reliability (α from .70-.90 for all subscales), and supported the convergent and discriminant
validity of scales (Gagné et al., 2015).
The outcomes variables were available only in Sample 2. In-role performance was measured by
seven self-reported items taken from Abramis (1994). Items were rated on a 1 (really bad) to 5 (really
good) Likert scale with each item based on the question stem of, “In the last (seven days/week you
worked), how well were you…” Items included, “doing your best work,” and “showing initiative in
your work” (α = .85). Extra-role performance was measured by 9 items from Morrison (1994), with
each item rated 1 (totally disagree) to 5 (totally agree) along a Likert scale (α = .81; e.g., “I help in the
training of new colleagues” and “I take active part in meetings of the organization”). Job Satisfaction
was measured through 14 items taken from De Witte, Hooge, Vandoorne, and Glorieux (2001). Items
were rated on a 5-point scale (1, totally dissatisfied to 5, totally satisfied) in response to questions such
as, “How satisfied are you in general with your work?” (α = .89). Engagement was measured using 15
items from the Utrech Work Engagement Scale (UWES, Schaufeli & Bakker, 2003) on a 1 (very
rarely) to 6 (always) Likert scale. Subscales for vigor (5 items, e.g., “When I get up in the morning, I
feel like going to work”), dedication (5 items, e.g., “I am enthusiastic about my job”), absorption (5
items, e.g., “When I am working, I forget everything else around me”) were combined into an overall
measure of work engagement for the sake of parsimony (α = .95). Finally, burnout was measured on a
6-point scale using the Schaufeli and van Dierendonck (1993) adaptation of the Maslach Burnout
Inventory. Two subscales of emotional exhaustion (5 items; e.g., “working all day is a heavy burden
for me”) and cynicism (4 items; e.g., “I doubt the usefulness of my work”) were included and
combined in the current analyses (α = .93), and scored from 1 (very rarely) to 6 (always).
Analyses
Preliminary Measurement Models
Preliminary measurement models were estimated in both samples using the robust maximum
likelihood estimator (MLR) available in Mplus 7.3 (Muthén & Muthén, 2014), in conjunction with
Full Information Maximum Likelihood (FIML) estimation to deal with the very low level of missing
data present this data set (0% to 2.8% per item; M = 1.1%). In each sample, we contrasted a classical
confirmatory factor analytic (CFA) model, in which each of the six MWMS factors was defined on the
basis of it’s a priori items, with no cross-loading allowed between items and non-target factors, with
an exploratory structural equation modeling (ESEM; Asparouhov & Muthén, 2009; Morin, Marsh, &
Nagengast, 2013), which was defined in the same manner as the CFA model while allowing for the
free estimation of cross-loadings between items and non-target factors. These ESEM models were
specified using a confirmatory approach using target rotation (Asparouhov & Muthén, 2009), which
allows for the pre-specification of target loadings in a confirmatory manner, while cross-loadings are
targeted to be as close to zero as possible. Recent studies conducted on motivational data show the
advantages of using an ESEM measurement model (Guay, Morin, Litalien, Valois & Vallerand, 2015;
Litalien, Guay, & Morin, 2015) in terms of obtaining reduced estimates of factor correlations more in
line with theoretical expectations. This decision is also based on the results from simulation studies
(Asparouhov & Muthén, 2009; Sass & Schmitt, 2010; Schmitt & Sass, 2011) and studies of simulated
data (Marsh, Lüdtke, Nagengast, Morin, & Von Davier, 2013; Morin, Arens, & Marsh, 2015) showing
that forcing cross-loadings (even as small as .100, Marsh et al., 2013) present in the population model
to be exactly zero (as in CFA) forces these cross-loadings to be absorbed through an inflation of the
factor correlations. In contrast, these same studies show that the free estimation of cross-loadings,
even when none are present in the population model, still provides unbiased estimates of the factor
correlations (also see Asparouhov, Muthén, & Morin, 2015; Morin, Arens et al., 2015). Thus,
Asparouhov et al. (2015, p. 1564) note that:
“Overall, these studies clearly show that the inclusion of cross-loadings is neither logically
flawed nor logically incorrect but rather empirically supported by statistical research. Going back to
the flawed argument that cross-loadings “taint” the nature of the constructs, these results rather show
that it is the exclusion of these cross-loadings that modifies the meaning of the constructs.”
Given the known oversensitivity of the chi-square test of exact fit (χ²) to sample size and minor
model misspecifications (e.g., Marsh, Hau, & Grayson, 2005), we relied on goodness-of-fit indices to
describe the fit of these models (Hu & Bentler, 1999): (a) the comparative fit index (CFI), (b) the root
mean square error of approximation (RMSEA) and its 90% confidence interval (CI); (c) the
standardized root mean square residual (SRMR). Values greater than .90 and .95 for the CFI
respectively indicate adequate and excellent model fit, while values smaller than .08 or .06 for the
RMSEA and SRMR respectively support acceptable and excellent model fit. In both samples, these
results revealed the clear superiority of the ESEM measurement model [(Sample 1: χ² = 124.575, df =
72, p <.001; CFI = .986; RMSEA = .032; CI = .022 to .041; SRMR = .016); (Sample 2: χ² = 161.020,
df = 72, p <.001; CFI = .955; RMSEA= .066; CI = .052 to .079; SRMR = .020)], when compared to
the CFA model [(Sample 1: χ² = 421.443, df = 137, p <.001; CFI = .924; RMSEA = .054; CI = .048
to .059; SRMR = .058); (Sample 2: χ² = 401.719, df = 137, p <.001 CFI = .866; RMSEA = .082; CI
= .073 to .092; SRMR = .070)]. This conclusion was supported by an assessment of the parameter
estimates obtained from both models, which revealed generally well-defined factors, and reduced
factor correlations in the ESEM [(Sample 1: |r| = .015 to .761; M|r| = .281); (Sample 2: |r| = .026
to .446; M|r| = .234)], when compared to CFA model [(Sample 1: |r| = .057 to .836; M|r| = .366);
(Sample 2: |r| = .021 to .844; M|r| = .401)].
LPA were conducted using factor scores (specified to have a mean of 0 and a standard deviation
of 1) from the retained ESEM measurement models (e.g., Kam, Morin, Meyer, & Topolnytsky, 2016;
Morin & Marsh, 2015). In comparison with scale scores, factors scores have the advantage of
providing a partial control for measurement errors by giving more weight to items presenting lower
levels of measurement errors (Kam, Morin, Meyer & Topolnytsky, 2016; Morin & Marsh, 2015;
Skrondal & Laake, 2001). Correlations and estimates of scale score reliability for all variables
(including these factor scores) used in the present study are reported in Table 1.
Latent Profile Analyses
Based on our expectation that 4 to 6 latent profiles would be identified, models including 1 to 8
profiles were estimated in each sample using the robust Maximum Likelihood (MLR) estimator
available in Mplus. The means and variances of the six motivation factors were freely estimated in all
profiles (Morin, Maïano et al., 2011; Peugh & Fan, 2013), using 7,000 random sets of start values, 300
iterations for each random start, and the 200 best solutions retained for final stage optimization (Hipp
& Bauer, 2006). All models converged on well replicated solutions.
In order to determine the optimal number of profiles in each sample, it is important to consider
the substantive meaning and theoretical conformity of the profiles (Marsh et al., 2009; Muthén, 2003),
the statistical adequacy of the solution, and a variety of statistical indicators. Among these statistical
indicators, we report the Akaike Information Criterion (AIC), the Bayesian information criterion (BIC),
the Consistent AIC (CAIC), the sample-adjusted BIC (ABIC), the adjusted version of the Lo, Mendell,
and Rubin likelihood ratio test (LMR), and the Bootstrap Likelihood Ratio Test (BLRT). The entropy
was also examined, and indicates the precision with which the cases are classified into the profiles (on
a 0 to 1 scale). However, the entropy should not be used in itself to determine the optimal number of
profiles (Lubke & Muthén, 2007; Peugh & Fan, 2012, 2013, 2015; Tein, Coxe, & Cham, 2013).
Extensive simulation research has looked at the performance of these various indicators to help in
selecting the optimal number of latent profiles in the data in the context of latent profile analyses and
other forms of person-centered mixture models. Overall, these studies converge in supporting the
efficacy of the CAIC, the BIC, the ABIC, and the BLRT in choosing the model which best recovers
the sample’s true parameters (e.g., Henson, Reise, & Kim, 2007; McLachlan & Peel, 2000; Morgan,
2015; Nylund, Asparouhov, & Muthén, 2007; Peugh & Fan, 2012, 2013, 2015; Tein et al., 2013;
Tofighi & Enders, 2008; Tolvanen, 2007; Yang, 2006). In particular, a recent simulation study (Diallo,
Morin, & Lu, 2016) suggest that the BIC and CAIC should be privileged under conditions of high
entropy (e.g., ≥ .800), whereas the ABIC and BLRT appear to perform better in conditions of low
entropy (e.g., ≤ .500). In contrast, the bulk of current research evidence suggests that, like the entropy,
the AIC and LMR/ALMR should not be used in the class enumeration process (e.g., Diallo et al., 2016;
Henson et al., 2007; Nylund et al., 2007; Peugh & Fan, 2013; Tofighi & Enders, 2007; Yang, 2006). In
the current study, these indicators are thus simply reported to ensure a thorough disclosure of results,
but will not be used to select the optimal number of profiles. A lower value on the AIC, CAIC, BIC
and ABIC suggests a better-fitting model. Both the LMR and BLRT compare a k-profile model with a
k-1-profile model. A significant p value indicates that the k-1-profile model should be rejected in favor
of a k-profile model. However, since these tests are all variations of tests of statistical significance, the
class enumeration procedure can still be heavily influenced by sample size (Marsh et al., 2009). That
is these indicators frequently keep on improving with the addition of latent profiles to the model
without reaching a minimum. In these cases, information criteria should be graphically presented
through “elbow plots” illustrating the gains associated with additional profiles (Morin, Maïano, et al.,
2011; Morin & Marsh, 2015; Petras & Masyn, 2010). In these plots, the point after which the slope
flattens suggests the optimal number of profiles that should be examined, together with adjacent
solutions including one more and one less profile, for theoretical conformity and statistical adequacy.
Latent Profile Analyses with Predictors and Outcomes
Starting from the final LPA solution retained for Sample 1, we then proceeded to tests of the
relations between the two dummy variables created to reflect job categories and the probability of
membership into the profiles. These two variables were included to the final model through a
multinomial logistic regression. In multinomial logistic regressions, each predictor has k-1 (with k
being the number of profiles) complementary effects for each possible pairwise comparison of profiles.
The regression coefficients reflect the increase, for each unit increase in the predictor (with dummy
variables this reflects the difference between the job category coded 1 and the remaining job
categories), that can be expected in the log-odds of the outcome (i.e., the probability of membership in
one profile versus another). For simplicity, we report odds ratios (OR), reflecting the change in
likelihood of membership in a target profile versus a comparison profile associated with the target job
category. For example, an OR of 3 suggests that employees from the target job category are three-
times more likely than others to be member of the target profile (versus the comparison profile).
Then, starting from the final LPA solution retained for Sample 2, we tested the relations between
profile membership and the multiple outcome variables available in this sample (performance, extra-
role behaviors, job satisfaction, engagement, and burnout), through the direct inclusion of these
outcomes in the model as additional profile indicators (Morin & Wang, 2016). The MODEL
CONSTRAINT command of Mplus was used to systematically test mean-level differences across all
specific pairs of profiles (using the multivariate delta method: e.g., Raykov & Marcoulides, 2004).
Results
The fit indices for the alternative solutions estimated separately in both samples are reported in
Table 2. For both samples, the CAIC, BIC, ABIC, and BLRT kept on improving with the addition of
latent profiles. However, we also note that the entropy values are quite high (≥ .800) for all of the
estimated models in both samples. Following Diallo et al.’s (2016) recommendations, this suggests
that the decision of how many profiles to retain should mainly focus on the BIC and CAIC. Because
these indicators failed to reach a minimum, we relied on a graphical representation of these
information criteria (Morin, Maïano, et al., 2011; Morin & Marsh, 2015; Petras & Masyn, 2010).
These plots are reported in Figure 1, and show that the decreases in values of most information criteria
reached a plateau around 4 profiles in both samples 1 and 2. Examination of the 4-profile solutions and
of the adjacent 3- and 5- profile solutions showed that all solutions were fully proper statistically in
both samples. This examination also revealed that adding a fourth profile always resulted in the
addition of a well-defined qualitatively distinct and theoretically meaningful profile to the solution,
whereas adding a fifth profile resulted in the arbitrary division of one of the existing profile into
smaller profiles differing only quantitatively from one another. As this additional small profile did not
add anything meaningful in theoretical terms (i.e., it has the same meaning as already present profiles),
the more parsimonious 4-profile solution was thus retained for each sample, in line with the
conclusion suggested by the statistical indicators. This solution provides a reasonable level of
classification accuracy, with an entropy value of .861 in Sample 1 and .886 in Sample 2. Classification
probabilities are presented in Table 3. These results clearly demonstrate the high level of classification
accuracy of these solutions, with average posterior probabilities of class membership in the dominant
profile varying from .887 to .950 in Sample 1 and from .923 to .980 in Sample 2, with low cross-
probabilities (varying from ≤.001 to .073 in Sample 1 and from <.001 to .042 in Sample 2).
The retained 4-profile solutions are represented in Figure 2 for Sample 1, and Figure 3 for Sample
2 (with exact numerical results reported in Table 4). These figures makes it rapidly obvious that the
profile structure is remarkably similar across samples, providing clear support to the generalizability
of the profiles. For both samples, Profile 1 characterized amotivated employees (corresponding to 27.6%
of the employees in Sample 1 and 13.1% in Sample 2) presenting very high levels of amotivation and
average to low levels on all other motivation factors. For this profile, it is noteworthy that levels of
motivation decrease as a direct function of their relative degree of self-determination as proposed by
SDT. Profile 2 (11.5% in Sample 1; 27.8% in Sample 2) characterizes employees presenting very low
levels of social and material forms of external regulations, low levels of amotivation and introjection,
and average or slightly above average levels of identified regulation and intrinsic motivation. This
moderately autonomous profile thus also appears to follows the continuum structure of self-regulation
proposed by SDT in that it presents a single dominant regulation type with levels of other regulations
tapering off as they become more theoretically distant. Profile 3 characterizes highly motivated
employees (25.6% in Sample 1; 22% in Sample 2) presenting a relatively low level of amotivation and
moderate to high levels on the other types of regulations which increase as a direct function of their
relative degree of self-regulation according to SDT. This profile clearly presents the highest levels on
the more autonomous forms of motivation (identified regulation and intrinsic motivation) out of all
profiles identified in both Samples. This highly autonomous profile thus also appears to follow the
continuum structure of self-regulation proposed by SDT. Finally, Profile 4 characterizes employees
presenting average levels of all regulations although the results obtained in sample 2 suggest that this
profile may also show a tendency to have slightly above average levels of external regulation, and
slightly below average levels of autonomous forms of regulation. This profile, which also follows the
self-regulation continuum proposed by SDT, thus appears to describe employees with balanced
motivation (35.3% in Sample 1; 37.1.0% in Sample 2).
Predictors of Profile Membership
Results from the multinominal logistic regression examining relations between job category and
profile membership in Sample 1 are reported in Table 5. Given that both dummy predictors were
simultaneously considered, the blue-collar employees were used as the comparison group, with the
effects of the first dummy predictor representing differences between white-collar technology sector
employees and all other employees, and the second representing differences between white-collar
governmental employees and all other employees. These results show that white-collar technology
employees presented a lower likelihood of membership in the moderately autonomously motivated
profile (Profile 2) than in all other profiles when compared to employees from other job categories. In
contrast, white-collar governmental employees presented a greater likelihood of membership into the
least desirable amotivated profile (Profile 1) than in all other profiles when compared to all other
employees. These employees were also less likely to be in the moderately autonomously motivated
(Profile 2) or highly motivated (Profile 3) profiles than in the balanced profile (Profile 4).
Outcomes of Profile Membership
Outcomes variables were added to the final 4-profile solution retained for Sample 2. Mean levels
of each outcome across the four profiles are graphically depicted in Figure 4, while the exact mean
levels of the outcomes and the statistical significance for each pairwise comparison of outcome levels
across profiles are reported in Table 6. Most of these comparisons are statistically significant, with
only a few exceptions, supporting the predictive validity of the extracted latent profiles. Starting with
performance, the results show that levels of both in-role and extra-role performance are highest in both
the highly motivated profile (Profile 3) and the moderately autonomous profile (Profile 2), and lowest
among both the amotivated (Profile 1) and balanced (Profile 4) profiles, which could not be
distinguished from one another. Levels of job satisfaction and engagement significantly differed in a
similar manner across profiles, being highest among the highly motivated profile (Profile 3) and the
moderately autonomous profile (Profile 2), followed by the balanced profile (Profile 4), and lowest
among the amotivated profile (Profile 1). Finally, levels of burnout were highest in the balanced
profile (Profile 4), followed by the amotivated profile (Profile 1), and then by both the highly
motivated (Profile 3) and moderately autonomous (Profile 2) profiles, which could not be
distinguished from one another. 1
Discussion
This study aimed to extend motivation theory and research through the identification of profiles
of employees based on the simultaneous consideration of the six forms of behavioral regulation
assumed to form the underlying continuum of self-determination proposed by SDT (Deci & Ryan,
1985). The current study provides an incremental contribution to the literature, finding four motivation
profiles in the work domain that replicated across two reasonably large and heterogeneous samples of
employees from two different countries. Prior research has generally been plagued by the reliance on
small samples, the use of cluster analyses, and the arbitrary dichotomization of behavioral regulations
into two broad categories of autonomous and controlled regulations (Graves et al., 2015; Moran et al.,
2012; Van den Broeck et al., 2013). In contrast, this study relied on two large samples of employees
from Canada and Belgium from across multiple industries and job categories. Additionally, unlike
much of the past person-centered research, the current study used state of the art analyses to not only
identify an optimal number of profiles, but also to include antecedents and outcome variables in a
statistically more advanced and rigorous manner than previously possible. A final key contribution of
this study lies in the demonstration of the value of considering the whole range of behavioral
regulations in the estimation of motivation profiles, as opposed to dichotomizing motivation into
autonomous and controlled composite variables. In particular, the nature of the profiles observed in the
present study, which generalized across samples, supported the underlying continuum structure of
motivation proposed by SDT (Deci & Ryan, 1985). In sum, the comprehensive sampling and analyses
employed in the current research lend support to the robustness and reliability of the detected profiles.
In line with prior research conducted in the education, sport, and work domains, our results
revealed four latent profiles, which were replicated across the two samples. Particularly important is
the observation that these profiles revealed qualitative and quantitative differences in employees’
experiences of work motivation. These profiles showed that not only do employees experience varying
amounts of overall motivation or self-determination, they also tend to experience different types of
1
Upon request from a reviewer, all analyses were replicated while controlling for gender. These
additional models converged on results substantively identical to those reported here. Additional
details are available upon request from the corresponding author.
motivation. Additionally, our results revealed that the relative likelihood of membership into these
profiles differed as a function of job type, and that it was associated with a variety of work-related
performance and wellbeing outcomes. Meyer, Morin, and Vandenberghe (2015) recently noted that
the value of person-centered analyses in the work domain depends not only on their ability to identify
subgroups of employees differing from one another meaningfully on a set of variables, but also on the
ability to demonstrate that these subgroups emerge regularly across samples, can be predicted in a
meaningful manner, and are relevant to the prediction of work outcomes. As they met all of these
criteria, our results can be considered highly meaningful.
As anticipated, we found a profile containing predominantly autonomous forms of regulation, a
balanced profile containing roughly equal levels of all regulations, and at least one profile containing
both autonomous and controlled forms of regulation. External regulation seemed to stand on its own in
these profiles, whereas introjected regulation seemed to cluster more closely with autonomous forms
of regulation, showing the importance of considering regulations at this level instead of aggregating
them into global controlled and autonomous variables. For instance, the highly motivated profile was
characterized by high levels of intrinsic motivation, identified regulation, and introjected regulation,
and slightly above average levels of external regulation. Looking at the positive performance and
wellbeing outcomes associated with this profile, it appears to be one of the most desirable profiles.
Our results further revealed that white-collar technology sector employees are somewhat more
disposed to correspond to this profile compared to the moderately autonomous profile. However, these
white-collar workers were equally as likely to correspond to the amotivated and balanced profiles as to
the highly motivated profile. This suggests that job characteristics known to be more prevalent in the
white-collar technology sector, such as the more frequent use of participative management, enriched
job designs and tasks variety, and even profit-sharing schemes, may result in situations where
employee either have their basic psychological needs met and therefore experience autonomous forms
of motivation (Blais et al., 1993; Gagné et al., 1997, 2010; Gagné & Forest, 2008), or alternatively
experience amotivation or external pressure to perform – a kind of polarizing effect in which these
practices either work well or fail badly.
The moderately autonomous profile was characterized by low levels of external and introjected
regulations, and above average levels of identified regulation and intrinsic motivation. This profile is
similar to the highly motivated profile in its shape, but not in the overall level of motivation. This
profile also presented above average levels of performance and wellbeing, performing as well as the
highly motivated profile. This indicates that while the overall quantity of motivation may play some
role in influencing work outcomes, the shape of the profile appears to have more important outcome
implications. Specifically, as long as a profile is dominated by autonomous rather than controlled
forms of regulation, individuals will display above average levels of performance and wellbeing. This
finding suggests that increasing all motivation types may not improve performance or wellbeing.
Rather, it appears more important to increase identified regulation and intrinsic motivation, while
ensuring that they remain higher than external regulation.
The moderately autonomous profile becomes even more interesting when compared to the
balanced profile, given that both are characterized by similar amounts of overall motivation. However,
while the moderately autonomous profile is dominated by autonomous motivation, the balanced
profile is generally average across all regulations. Such a comparison allows for a clear examination of
the relative importance of shape effects while holding reasonably constant the overall quantity of
motivation. The results showed that the moderately autonomous profile was far more desirable than
the balanced profile, which was associated with significantly lower levels on all indicators of
performance and wellbeing. Thus, motivation profiles dominated by an emphasis on meaning and
interest appear to lead to higher performance and wellbeing, compared to the balanced or amotivated
profiles, regardless of overall amount of motivation. These results comparing the highly autonomous
and moderately autonomous profiles, as well as the moderately autonomous and balanced profiles, are
important. Indeed, these comparisons suggest that, far from being an effective motivator (Cerasoli et
al., 2014; Gerhart & Fang, 2015), an emphasis on social and material rewards may have a negative
impact on performance when it is not accompanied by a comparable emphasis on meaning, interest
and pleasure (Gagné & Deci, 2005). Worse, this negative impact may be accompanied by an equally
negative impact on wellbeing, making it doubly difficult for these employees to increase their
performance in the long term (e.g., Ryan, Deci, & Grolnick, 1995). Interestingly, the previously
discussed results regarding the fact that the moderately autonomous and highly motivated profiles are
associated with similarly desirable outcomes suggest that high levels of autonomous regulations
appear to protect employees from the effects of high levels of more controlled forms of regulations.
Finally, the amotivated profile characterizes employees for whom work is neither motivated by
meaning, guilt, enjoyment, or rewards but are rather mainly amotivated, suggesting they may possibly
feel “trapped” in their position due to high perceived sacrifices associated with leaving (i.e.,
continuance commitment; Morin, Meyer, McInerney, Marsh, & Ganotice, 2015). In line with our
expectations, white collar governmental employees, who tend to be exposed to more rigid bureaucratic
structures, presented a significantly greater likelihood of membership into this profile (De Cooman et
al., 2013; Gillet et al., 2013), followed by membership in the balanced profile, strongly suggesting that
characteristics of this job are highly detrimental to autonomous motivation. Also in line with our
expectations, employees from this amotivated profile presented the lowest levels of wellbeing out of
all profiles, and levels of performance that were undistinguishable from those observed in the
balanced profile. This profile appeared to be the least desirable.
It is interesting to note that the amotivated and highly motivated profiles both follow the expected
continuum structure so closely that it could be argued that for these profiles a single factor
representing global self-determined motivation (e.g., Howard, Gagné, Morin, & Forest, 2016) could be
sufficient to describe these employees satisfactorily. Alternately, for the moderately autonomous and
balanced profiles where the profiles do not follow the continuum structure as perfectly, it appears
necessary to take into account qualitative distinctions between the various motivation subscales in
order to obtain a complete picture of employees’ work motivation.
In regards to previous person-centered research on work motivation, the current results provide an
incremental contribution to the literature by replicating, in part, the profiles found by Graves et al.
(2015), and expanding greatly on the cluster analytic results of Van den Broeck et al. (2013) and
Moran et al. (2012). All of these studies succeeded in identifying the most extreme profiles, including
a highly motivated profile characterized by above average levels of all types of motivation, and an
amotivated profile characterized by below average levels on most types of motivation. The moderately
autonomous profile identified in the current study also largely replicates the self-determined profile
found by Graves et al. (2015) in a sample of managers. The balanced profile, which shows a slight
tendency towards an external focus, is a more novel finding of the current study. Not only has this
profile allowed for a highly insightful comparison between two profiles characterized by similar
global amounts of motivation but different shapes, but it suggests that some employees draw
motivation from multiple sources equally but do not seem to thrive in their workplace as a result of it.
Finally, the current study provides evidence of generalizability of the reported profiles. Like with
variable-centered research, the confidence with which person-centered results can be used to guide
practice depends on replicability and the convergence of results obtained from a variety of samples.
Through multiple samples and studies, it becomes possible to identify a set of core profiles which are
commonly occurring in most work contexts, and more peripheral profiles which may arise due to
specific workplace circumstances or in specific subgroups of employees (Solinger, Van Olffen, Roe &
Hofmans, 2013). The current study offers a set of four core profiles which, interestingly, replicate
some of the profiles found by Graves et al., (2015). This suggests that the subset of replicated profiles
are more likely to reflect core profiles of employee motivation, whereas the additional profiles
reported by Graves et al. may be more peripheral, arising specifically in manager sub-populations.
In sum, our results incrementally add to previous research by examining work motivation profiles
in the most rigorous manner available to date (i.e., through the incorporation of all regulation types
into state-of-the-art LPA) with reasonably large and heterogeneous samples of employees from two
countries. Additionally we provide initial evidence which demonstrates that profile membership varies
as a function of job category with white-collar technology sector employees less likely to be in the
moderately autonomously motivated profile, while government employees are more likely to be
amotivated in their work. Lastly our results show that profile membership has meaningful implications
for a wide range of work outcomes with profile characterized by predominantly autonomous forms of
motivation being associated with more positive performance and wellbeing outcomes.
Limitations and Directions for Future Research
Though the current study presents several advantages over previous research, it also presents
notable limitations. As with all cross-sectional research it is impossible to reach clear conclusions
regarding the directionality of the associations between the observed motivational profiles and the so-
called outcome variables on the basis of a single study. The possibility thus remain that the observed
associations follow reversed or even reciprocal relations as performance and wellbeing may
themselves act as predictor of employee motivation profiles. However, lending confidence to the
current interpretations, prior longitudinal research has supported the idea of directional relationships
through which motivation levels predict later levels of performance and wellbeing (e.g., Baker, 2003;
Dysvik & Kuvaas, 2013). Still, future research is needed to clarify this issue, and particularly to
investigate possible reciprocal relations among these constructs (e.g., Morin, Meyer, Bélanger,
Boudrias, Gagné, & Parker, 2016). Longitudinal studies will also be needed to examine the
development and temporal stability of motivation profiles. It would be most useful to know how, and
under which conditions, the different profiles found in the present study develop and evolve over time,
considering both organizational newcomers (Bauer & Erdogan, 2014) as well as employees at later
career stages (Gould & Hawkins, 1978). Like the present study, future person-centered research
should also strive to favor LPA over more traditional cluster analyses for reasons covered
comprehensively elsewhere (Meyer & Morin, 2016; Morin et al., 2011; Vermunt & Magidson, 2002).
In particular, LPA tends to rely on far less stringent assumptions, which can be relaxed as needed,
relative to cluster analyses, as well as a lower level of reactivity to measurement scales and clustering
algorithm. Furthermore, LPA allows for the direct incorporation of covariates into the model, without
the need to rely on suboptimal two-step strategies. Finally, research would also benefit from devoting
attention to the effects of specific modifiable organizational design factors, such as organizational
structure, job design, leadership style, and compensation systems, on membership into specific
motivational profiles. While our results suggest a clear relation between job categories and
membership into specific profiles, a finer grained analysis of the mechanisms involved in these
relations would have important practical relevance to the design of specific interventions to improve
employee motivation. In this regard, it would be particularly useful to know how organizational
changes, such as job design changes and compensation system changes, are able to predict changes in
profile membership that would affect transitions from one profile to another.
Practical Implications
In person-centered research, evidence for generalizability is built from an accumulation of studies,
from which it becomes possible to identify a core set of profiles emerging with regularity, together
with more peripheral profiles emerging irregularly under specific conditions (Solinger, Van Olffen,
Roe & Hofmans, 2013). The fact that the profiles identified in this study are in line with theoretical
expectations and emerged consistently across two independent samples of employees recruited in two
countries supports their generalizability. Though additional research is needed, we can suggest
organizations can use these four profiles to think about how employees falling into these profiles can
be best managed. For example, knowing that the balanced profile has lower than average performance,
probably because of a lack of meaning and enjoyment, organizations could try to provide meaning
(e.g., through task significance; Grant, 2008) and stimulation (e.g., through job redesign; Hackman &
Oldham, 1975) to employees. Specifically, employers may find that while a job has inherent
meaningfulness and intrinsically enjoyable factors, employee motivation, and therefore performance,
remains below expectations. Results from this study indicate that this may occur when external
motivators are equally influential as more autonomous factors (such as is the case in the balanced
profile). In these conditions, reducing the external focus and promoting more autonomously-driven
reasons could be enough to nudge employees away from the balanced profile, with its largely below
average outcomes, and into the moderately autonomous profile. Such a small adjustment could lead to
employees being driven predominately by autonomous factors and subsequently performing more
successfully and experiencing greater wellbeing. As such, knowing that autonomous motivation is
relatively more important than external regulations in promoting performance and wellbeing,
organizations may wish to focus more on meaning and enjoyment than on rewards and punishments.
The drawback of the variable-centered approach is that it often leads to thinking about an
intervention that will improve a variable (e.g., intrinsic motivation) without taking into consideration
what it may do to other forms of motivation (e.g., introjection). Conversely, the person-centered
approach allows managers to consider employees as whole entities, rather than focusing narrowly on
isolated individual characteristics. This approach recognizes the complexity of human motivation and
behavior, and as such may provide a more complete and integrated description of this reality.
Our results could also prove particularly useful in informing the long-standing debate on the
impact of incentives on work motivation. Gerhart and Fang (2015; also see Cerasoli et al., 2014)
recently argued that controlled types of motivation may yield positive outcomes and that these
motivation types could be promoted through the use of monetary incentives. Results of the current
study suggest a relatively weak association between external material regulation and performance, and
offer no support for the proposition that external rewards are successful in increasing performance
when accompanied by autonomous forms of motivation. Similar conclusions have been put forward in
previous person-centered research by Van den Broeck et al. (2013) and Moran et al., (2012), who also
found more positive outcomes associated with more autonomously driven profiles than profiles driven
by controlled regulations even when accounting for differing levels of global motivation.
In regards to the outcomes considered in this study, it is clear that organizations should attempt to
promote profiles characterized by relatively higher levels of autonomous than external forms of
regulations, through meaning making and the stimulation of people’s interests for the work they do. It
seems that as long as organizations can achieve this, they do not need to focus so much on promoting
external regulation through material and social rewards or punishments. Our results thus indicate that
it is not worth promoting controlled forms of motivation in addition to promoting autonomous forms
of motivation, as has been argued by Gerhart and Fang (2015). Furthermore, the outcomes associated
with the externally regulated profile suggest that there is an important risk associated with focusing on
the promotion of external forms of regulations. As such, it appears that organizations would benefit
more from a focus on nurturing more autonomous forms of motivation through increases in job
meaningfulness, interest, and autonomy, than from a focus on social and material rewards.
References
Abramis, D. J. (1994). Relationship of job stressors to job performance: Linear or inverted-
u? Psychological Reports, 75, 547-558.
Asparouhov, T., & Muthén, B. O. (2009). Exploratory structural equation modeling. Structural
Equation Modeling, 16, 397–438.
Asparouhov, T., Muthén, B., & Morin, A. J. S. (2015). Bayesian structural equation modeling with
cross-loadings and residual covariances: Comments on Stromeyer et al. Journal of Management,
41, 1561-1577.
Battistelli, A., Galletta, M., Portoghese, I., & Vandenberghe, C. (2013). Mindsets of commitment and
motivation. The Journal of Psychology, 147, 17-48.
Baker, S. R. (2003). A prospective longitudinal investigation of social problem-solving appraisals on
adjustment to university, stress, health, and academic motivation and performance. Personality
and Individual Differences, 35, 569-591.
Bauer, T. N., & Erdogan, B. (2014). Delineating and reviewing the role of newcomer capital in
organizational socialization. Annual Review of Organizational Psychology and Organizational
Behavior, 1, 439-457.
Blais, M. R., Brière, N. M., Lachance, L., Riddle, A. S., & Vallerand, R. J. (1993). L’inventaire des
motivations au travail de Blais. Revue québécoise de psychologie, 14(3), 185-215.
Boiché, J., Sarrazin, P., Grouzet, F., Pelletier, P., & Chanal, J. (2008). Students’ motivational profiles
and achievement outcomes in physical education: A self-determination perspective. Journal of
Educational Psychology, 100, 688–701.
Cerasoli, C., Nicklin, J., & Ford, M. (2014). Intrinsic motivation and extrinsic incentives jointly
predict performance: A 40 year meta-analysis. Psychological Bulletin, 140, 980–1008.
Chemolli, E., & Gagné, M. (2014). Evidence against the continuum structure underlying motivation
measures derived from self-determination theory. Psychological Assessment, 26, 575-585.
Deci, E. L., & Ryan, R. M. (1985). Intrinsic motivation and self-determination in human behavior.
New York: Plenum.
De Cooman, R., Stynen, D., Van den Broeck, A., Sels, L., & De Witte, H. (2013). How job
characteristics relate to need satisfaction and autonomous motivation: Implications for work
effort. Journal of Applied Psychology, 43, 1342-1352.
De Witte, H., Hooge, J., Vandoorne, J. & Glorieux, I. (2001). Prettig werken in een gezonde
samenleving. Omvang, verschillen en gevolgen van arbeidstevredenheid in Vlaanderen. In:
Administratie Planning en Statistiek (Red.), Vlaanderen gepeild! De Vlaamse overheid en
burgeronderzoek 2001 (pp. 73-98). Brussel: Ministerie van de Vlaamse Gemeenschap.
Diallo, T.M.O., Morin, A.J.S., & Lu, H. (2016, in press). The impact of total and partial inclusion or
exclusion of active and inactive time invariant covariates in growth mixture models.
Psychological Methods.
Dysvik, A., & Kuvaas, B. (2013). Intrinsic and extrinsic motivation as predictors of work effort: The
moderating role of achievement goals. British Journal Social Psychology, 52, 412–430.
Gagné, M. (2014). The Oxford handbook of work engagement, motivation, and selfdetermination
theory. New York, NY, United States: Oxford University.
Gagné, M., & Deci, E. L. (2005). Self-determination theory and work motivation. Journal of
Organizational Behavior, 26, 331-362.
Gagné, M., & Forest, J. (2008). The study of compensation systems through the lens of self-
determination theory: 35 years of debate. Canadian Psychology, 49, 225–232.
Gagné, M., Sentcal, C. B., & Koestner, R. (1997). Proximal job characteristics, feelings of
empowerment, and intrinsic motivation: A multidimensional model. Journal of Applied Social
Psychology, 27, 1222-1240.
Gagné, M., Forest, J., Gilbert, M. -H., Aubé, C., Morin, E., & Malorni, A. (2010). The Motivation at
Work Scale: Validation evidence in two languages. Educational and Psychological Measurement,
79, 628-646.
Gagné, M., Forest, J., Vansteenkiste, M., Crevier-Braud, L., Van den Broeck, A., Kristin-Aspeli, A.,
… Westbye, C. (2015). The multidimensional work motivation scale: Validation evidence in
seven languages and nine countries. Journal of Work & Organizational Psychology, 24, 178-196.
Gerhart, B., & Fang, M. (2015). Pay, intrinsic motivation, extrinsic motivation, performance, and
creativity: Revisiting long-held beliefs. Annual Review of Organizational Psychology and

Organizational Behavior, 2, 489-521.
Gillet, N., Berjot, S., & Paty, B. (2009). Profil motivationnel et performance sportive [Motivational
profile and sport performance]. Psychologie Française, 54, 173-190.
Gillet, N., Berjot, S., Vallerand, R. J., Amoura, C., & Rosnet, E. (2012). Examining the motivation-
performance relationship in competitive sport: A cluster-analytic approach. International Journal
of Sport Psychology, 43, 79–102.
Gillet, N., Gagné, M., Sauvagere, S., & Fouquereau, E. (2012). The role of supervisor autonomy
support, organizational support, and autonomous and controlled motivation in predicting
employees’ satisfaction and turnover intentions. European Journal of Work and Organizational
Psychology, 22, 1–11.
Gillet, N., Vallerand, R., & Paty, P. (2013). Situational motivational profiles and performance with
elite performers, Journal of Applied Social Psychology, 43, 1200–1210
Gillet, N., Vallerand, R., & Rosnet, R. (2009). Motivational clusters and performance in a real-life
setting, Motivation and Emotion, 33, 49-62.
Gould, S., & Hawkins, B. (1978). Organizational career stage as a moderator of the satisfaction
performance relationship. Academy of Management Journal, 21, 434-450.
Grant, A. M. (2008). The significance of task significance: Job performance effects, relational
mechanisms, and boundary conditions. Journal of Applied Psychology, 93, 108-124.
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
Behaviour, 87, 32-42.
Grolnick, W. S., & Ryan, R. M. (1987). Autonomy in children's learning: an experimental and
individual difference investigation. Journal of personality and social psychology, 52(5), 890.
Guay, F., Morin, A. J. S., Litalien, D., Valois, P., & Vallerand, R.J. (2015). Application of exploratory
structural equation modeling to evaluate the academic motivation scale. The Journal of
Experimental Education, 83, 51-82.
Hackman, J. R., & Oldham, G. R. (1975). Development of the Job Diagnostic Survey. Journal of
Applied Psychology, 60, 159-170.

Hayenga, A., & Corpus, J. (2010). Profiles of intrinsic and extrinsic motivations: A person-centered
approach to motivation and achievement in school. Motivation and Emotion, 34, 371–383.
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.
Hipp, J. R., & Bauer, D. J. (2006). Local solutions in the estimation of growth mixture models.
Psychological Methods, 11, 36-53.
Howard, J. L., Gagné, M., Morin, A. J. S., & Forest, J. (2016, in press). Using bifactor exploratory
structural equation modeling to test for a continuum structure of motivation. Journal of
Management. Early View: doi 10.1177/0149206316645653
Hu, L.-T., & Bentler, P.M. (1999). Cutoff criteria for fit indexes in covariance structure analysis:
Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55.
Kam, C., Morin, A. J. S., Meyer, J. P., & Topolnytsky, L. (2016, in press). Are commitment profiles
stable and predictable? A latent transition analysis. Journal of Management. Early View, doi:
10.1177/0149206313503010.
Koestner, R., & Losier, G. F. (2002). Distinguishing three ways of being internally motivated. In E. L.
Deci & R. M. Ryan (Eds.), Handbook of self-determination research (pp. 101–121). Rochester,
NY: University of Rochester.
Litalien, D., Guay, F., & Morin, A. J. S. (2015). Motivation for PhD studies: Scale development and
validation. Learning and Individual Differences, 41, 1-13.
Liu, W., Wang, J., Tan, O., Koh, C., & Ee, J. (2009). A self-determination approach to understanding
students' motivation in project work. Learning and Individual Differences, 19, 139–145.
Lubke, G., & Muthén, B.O. (2007). Performance of factor mixture models as a function of model size,
covariate effects, and class-specific parameters. Structural Equation Modeling, 14, 26-47.
Marsh, H. W., Hau, K., & Grayson, D. (2005). Goodness of Fit in Structural Equation Models. In
Maydeu-Olivares, Albert (Ed); McArdle, John J. (Ed), Contemporary psychometrics: A festschrift
for Roderick P. McDonald. (pp. 275-340). Mahwah, NJ: Erlbaum.
Marsh, H.W., Lüdtke, O., Nagengast, B., Morin, A.J.S., & Von Davier, M. (2013). Why item parcels
are (almost) never appropriate: Two wrongs do not make a right-Camouflaging misspecification
with item parcels in CFA models. Psychological Methods, 18, 257-284.
Marsh, H. W., Lüdtke, O., Trautwein, U., & Morin, A. J. S. (2009). Classical latent profile analysis of
academic self-concept dimensions: Synergy of person- and variable-centered approaches to
theoretical models of self-concept. Structural Equation Modeling, 16, 191-225.
Maslach, C., Schaufeli, W. B., & Leiter, M. P. (2001). Job burnout. Annual Review of Psychology, 52,
397–422.
McLachlan, G., & Peel, D. (2000). Finite Mixture Models. New York: Wiley.
McNeill, M., & Wang, C. K. J. (2005). Psychological profiles of elite sports players in Singapore.
Psychology of Exercise and Sport, 6, 117-128.
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., Morin, A. J. S., & Vandenberghe, C. (2015). Dual Commitment to Organization and
Supervisor: A Person-centered Approach. Journal of Vocational Behavior, 88, 56-72.
Moran, M., Diefendorff, J., Kim, T., & Liu, Z. (2012). A profile approach to self-determination theory
motivations at work. Journal of Vocational Behavior, 81, 354–363.
Morgan, G.B. (2015). Mixed mode latent class analysis: An examination of fit index performance for
classification. Structural Equation Modeling, 22, 76-86.
Morin, A. J. S., Arens, A. K., & Marsh, H. W. (2015). A bifactor exploratory structural equation
modeling framework for the identification of distinct sources of construct-relevant psychometric
multidimensionality. Structural Equation Modeling, 1-24.
Morin, A. J. S., Maïano, C., Nagengast, B., Marsh, H. W., Morizot, J., & Janosz, M. (2011). Growth
mixture modeling of adolescents trajectories of anxiety across adolescence: The impact of
untested invariance assumptions. Structural Equation Modeling, 18, 613–648.
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., Marsh, H.W., & Nagengast, B. (2013). Chapter 10. Exploratory structural equation
modeling. In Hancock, G. R., & Mueller, R. O. (Eds.). (2013). Structural equation modeling: A
second course (2nd ed.). Charlotte, NC: Information Age Publishing, Inc.
Morin, A. J. S., Meyer, J. P., Bélanger, É., Boudrias, J. -S., Gagné, M., & Parker, P. D. (2016, In
Press). Longitudinal associations between employees’ perceptions of the quality of the change
management process, affective commitment to change and empowerment. Human Relations.
Morin, A. J. S., Meyer, J. P., McInerney, D. M., Marsh, H. W., & Ganotice, F. A. Jr. (2015). Profiles
of dual commitment to the occupation and organization: Relations to wellbeing and turnover
intentions. Asia-Pacific Journal of Management, 32, 717-744.
Morin, A. J. S., Morizot, J., Boudrias, J., & 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, UK: Wiley.
Morrison, E. W. (1994). Role definitions and organisational citizenship behaviour: The importance of
the employee’s perspective. Academy of Management Journal, 37, 1543-1567.
Muthén, B. O. (2002). Beyond SEM: General latent variable modeling. Behaviormetrika, 29, 81-117.
Muthén, B. O. (2003). Statistical and substantive checking in growth mixture modeling: Comment on
Bauer and Curran (2003). Psychological Methods, 8, 369-377.
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.
Petras, H., & Masyn, K. (2010). General growth mixture analysis with antecedents and consequences
of change. In A. R. Piquero & D.Weisburd (Eds.), Handbook of quantitative criminology (pp. 69-
100). New York, NY: Springer.
Peugh, J. & Fan, X. (2012). How well does growth mixture modeling identify heterogenous growth
trajectories? A simulation study examining GMM’s performance characteristics. Structural
Equation Modeling, 19, 204-226.

Peugh, J. & Fan, X. (2013). Modeling unobserved heterogeneity using latent profile analysis: A Monte
Carlo simulation. Structural Equation Modeling, 20, 616-639.
Peugh, J. & Fan, X. (2015). Enumeration index performance in generalized growth mixture models: A
Monte Carlo test of Muthén’s (2003) hypothesis. Structural Equation Modeling, 22, 115-131.
Pinder, C. C. (1998). Motivation in work organizations. Upper Saddle River, NJ.
Ratelle, C., Guay, F., Vallerand, R., Larose, S., & Senécal, C. (2007). Autonomous, controlled, and
amotivated types of academic motivation: A person-oriented analysis. Journal of Educational
Raykov, T., & Marcoulides, G. A. (2004). Using the delta method for approximate interval estimation
of parameter functions in SEM. Structural Equation Modeling, 11, 621-637.
Ryan, R. M., Deci, E. L., & Grolnick, W. S. (1995). Autonomy, relatedness, and the self: Their
relation to development and psychopathology. In D. Cicchetti & D. J. Cohen (Eds.),
Developmental psychopathology: Theory and methods (Vol 1, pp. 618-655). New York: Wiley.
Sass, D.A., & Schmitt, T.A. (2010). A comparative investigation of rotation criteria within exploratory
factor analysis. Multivariate Behavioral Research, 45, 73-103.
Schaufeli, W. B., & Bakker, A. B. (2003). UWES-Utrecht Work Engagement Scale: test manual.
Unpublished manuscript, Department of Psychology, Utrecht University.
Schaufeli, W. B., & van Dierendonck, D. (1993). The construct validity of two burnout measures.
Journal of Organizational Behavior, 14, 631-647.
Schmitt, T.A., & Sass, D.A. (2011). Rotation criteria and hypothesis testing for exploratory factor
analysis: implications for factor pattern loadings and interfactor correlations. Educational &
Psychological Measurement, 71, 95-113.
Skrondal, A., & Laake, P. (2001). Regression among factor scores. Psychometrika, 66, 563-576.
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.
Tein, J.-Y., Coxe, S., & Cham, H. (2013). Statistical power to detect the correct number of classes in
latent profile analysis. Structural Equation Modeling, 20, 640-657.
Tofighi, D., & Enders, C. K. (2008). Identifying the correct number of classes in growth mixture
models. In G. R. H. & & K. M. Samuelsen (Eds.), Advances in latent variable mixture models
(pp. 317-341). Charlotte, NC: Information Age.
Tolvanen, A. (2007). Latent growth mixture modeling: A simulation study. PhD dissertation,
Department of Mathematics, University of Jyväskylä, Jyväskylä, Finland.
Tremblay, M. A., Blanchard, C. M., Taylor, S., Pelletier, L. G., & Villeneuve, M. (2009). Work
Extrinsic and Intrinsic Motivation Scale: Its value for organizational psychology
research. Canadian Journal of Behavioural Science/Revue canadienne des sciences du
comportement, 41(4), 213.
Van den Broeck, A., De Cuyper, N., De Witte, H., & Vansteenkiste, M. (2010). Not all job demands
are equal: Differentiating job hindrances and job challenges in the Job Demands–Resources
model. European Journal of Work and Organizational Psychology, 19(6), 735–759.
Van den Broeck, A., Lens, W., De Witte, H., & Van Coillie, H. (2013). Unraveling the importance of
the quantity and the quality of workers’ motivation for wellbeing: A person-centered perspective.
Journal of Vocational Behavior, 82, 69-78.
Vansteenkiste, M., Sierens, E., Soenens, B., Luyck, K., & Lens, W. (2009). Motivational profiles from
a self-determination perspective: The quality of motivation matters. Journal of Educational
Vermunt, J.K., & Magidson, J. (2002). Latent class cluster analysis. In J. Hagenaars & A.
McCutcheon (Eds.), Applied latent class models (pp. 89-106). New York: Cambridge.
Yang, C. (2006). Evaluating latent class analyses in qualitative phenotype identification.
Computational Statistics & Data Analysis, 50, 1090–1104.
Zyphur, M. J. (2009). When mindsets collide: Switching analytical mindsets to advance organizational
science. Academy of Management Review, 34, 677–688.

Table 1.
Correlations and Scale Score Reliability (α) Estimates for the Variables Used in the Present Study
1 2 3 4 5 6 7 8 9 10 11
1. Amotivation - 0.137* 0.190* -0.200* -0.396* -0.401*
2. Ext-Material 0.107 - 0.465* 0.324* 0.191* 0.137*
3. Ext-Social 0.79* 0.304* - 0.297* 0.015 -0.095*
4. Introjected -0.039 0.200* 0.218* - 0.357* 0.246*
5. Identified -0.133 0.108 0.345* 0.441* - 0.761*
6. Intrinsic -0.361* 0.209* -0.026 0.399* 0.446* -
7. In-role Performance -0.161* -0.084 -0.029 0.222* 0.247* 0.252* -
8. Extra-role Performance -0.054 -0.133* -0.059 0.207* 0.247* 0.264* 0.329* -
9. Engagement -0.453* 0.035 -0.142* 0.265* 0.374* 0.660* 0.345* 0.359* -
10. Burnout 0.426* 0.029 -0.175* -0.185* -0.287* -0.456* -0.292* -0.165* -0.438* -
11. Job Satisfaction -0.506* 0.021 -0.175* 0.240* 0.331* 0.612* 0.234* 0.220* 0.646* -0.500*
α (Sample 1) 0.741 0.781 0.600 0.692 0.775 0.898
α (Sample 2) 0.886 0.695 0.761 0.711 0.671 0.882 0.846 0.810 0.946 0.927 0.885
Note: * p < .05. Sample 1 is above diagonal. Sample 2 is below diagonal. External-M = External-Material Regulation; External-S = External-Social
Regulation. Scores are all factor scores from preliminary models with a mean of 0 and standard deviation of 1.
Table 2.
Class Enumeration
Log Likelihood #fp scaling AIC CAIC BIC ABIC Entropy LMR BLRT
Sample 1 (n = 723)
1 Profile -5746.162 12 1.163 11516.324 11583.325 11571.325 11533.222 Na Na Na
2 Profiles -5054.193 25 1.020 10158.385 10297.971 10272.971 10193.588 0.816 <.001 <.001
3 Profiles -4808.461 38 1.135 9692.922 9905.092 9867.092 9746.431 0.840 0.002 <.001
4 Profiles -4611.800 51 1.196 9325.600 9610.354 9559.354 9397.414 0.861 0.086 <.001
5 Profiles -4491.730 64 1.118 9111.461 9468.799 9404.799 9201.581 0.851 0.018 <.001
6 Profiles -4384.863 77 1.093 8923.726 9353.648 9276.648 9032.151 0.867 <.001 <.001
7 Profiles -4291.002 90 1.044 8762.005 9264.512 9174.512 8888.735 0.861 0.002 <.001
8 Profiles -4226.600 103 1.099 8659.200 9234.291 9131.291 8804.236 0.853 0.162 <.001
Sample 2 (n = 286)
1 Profile -2281.653 12 2.0090 4587.305 4643.177 4631.177 4593.124 Na Na Na
2 Profiles -1714.199 25 0.9661 3478.397 3594.797 3569.797 3490.520 0.930 <.001 <.001
3 Profiles -1589.459 38 0.9764 3254.917 3431.845 3393.845 3273.344 0.897 <.001 <.001
4 Profiles -1473.405 51 1.1226 3048.810 3286.266 3235.266 3073.540 0.886 0.023 <.001
5 Profiles -1416.272 64 1.0316 2960.545 3258.528 3194.528 2991.579 0.890 0.012 <.001
6 Profiles -1380.270 77 1.0258 2914.539 3273.051 3196.051 2951.877 0.906 0.033 <.001
7 Profiles -1347.972 90 1.0299 2875.944 3294.983 3204.983 2919.585 0.917 0.232 <.001
8 Profiles -1315.187 103 1.0627 2836.373 3315.940 3212.940 2886.319 0.911 0.227 <.001
Note: #fp = Number of free parameters; AIC = Akaike information criterion; CAIC = Constant AIC; BIC = Bayesian information criterion; ABIC = Sample
size adjusted BIC; LMR = p value associated with the adjusted Lo-Mendell-Rubin likelihood ratio test; BLRT= p value associated with the bootstrap
likelihood ratio test.
Table 3.
Posterior Classification Probabilities for the Most Likely Latent Profile Membership (Row) by Latent
Profile (Column).
Amotivated Moderately Highly Motivated Balanced
(P.1) Autonomous (P.2) (P.3) (P.4)
Sample 1
Amotivated (P.1) 0.902 0.001 0.001 0.096
Moderately Autonomous (P.2) 0.003 0.976 0.014 0.007
Highly Motivated (P.3) 0.000 0.008 0.938 0.054
Balanced (P.4) 0.041 0.010 0.032 0.917
Sample 2
Amotivated (P.1) 0.942 0.004 0.000 0.054
Moderately Autonomous (P.2) 0.000 0.925 0.034 0.041
Highly Motivated (P.3) 0.000 0.031 0.941 0.028
Balanced (P.4) 0.007 0.023 0.021 0.949
Note. P: Profile.
Table 4
Mean Levels of Motivation in the Retained Latent Profile Models.
Moderately
Amotivated (P.1) Highly Motivated (P.3) Balanced (P.4)
Autonomous (P.2)
Mean Variance Mean Variance Mean Variance Mean Variance
Sample 1
Amotivation 1.025 1.169 -0.554 0.019 -0.515 0.019 -0.249 0.080
External-M 0.053 0.841 -1.075 0.112 0.403 0.632 0.015 0.581
External-S 0.242 0.786 -1.308 0.006 0.192 0.889 0.095 0.63
Introjected -0.331 0.764 -0.467 0.761 0.605 0.489 -0.027 0.532
Identified -0.840 1.005 0.143 0.498 0.901 0.119 -0.041 0.222
Intrinsic -0.867 1.009 0.288 0.54 0.961 0.126 -0.11 0.250
Sample 2
Amotivation 1.679 4.131 -0.338 0.002 -0.264 0.002 -0.183 0.004
External-M -0.050 1.152 -0.675 0.193 0.514 0.461 0.218 0.608
External-S 0.292 0.624 -0.805 0.260 0.548 0.824 0.175 0.608
Introjected -0.335 1.301 -0.236 0.510 0.909 0.183 -0.243 0.504
Identified -0.597 1.352 -0.034 0.611 0.823 0.276 -0.251 0.399
Intrinsic -1.450 0.974 0.437 0.267 0.961 0.091 -0.384 0.424
Note. P: Profile; External-M = External-Material Regulation; External-S = External-Social Regulation;
Indicators are estimated from factor scores with mean of 0 and a standard deviation of 1.
Table 5.
Results from Multinominal Logistic Regression Evaluating Relations between Job Type and Latent
Profile Membership (Sample 1)
Profile 1 vs. 2 Profile 1 vs. 3 Profile 1 vs. 4
Job Category Coefficient (SE) OR Coefficient (SE) OR Coefficient (SE) OR
White Collar 1.118 (0.413)** 3.059** -0.320 (0.429) 0.726 0.235 (0.397) 1.265
Government 1.544 (0.397)** 4.683** 1.557 (0.303)** 4.745** 0.803 (0.363)* 2.232*
Profile 2 vs. 3 Profile 2 vs. 4 Profile 3 vs. 4
Coefficient (SE) OR Coefficient (SE) OR Coefficient (SE) OR
White Collar -1.438 (0.485)** 0.237** -0.883 (0.407)* 0.413* 0.555 (0.444) 1.742
Government 0.014 (0.351) 1.014 -0.740 (0.351)* 0.477* -0.754 (0.267)** 0.470**
Note. *p <.05. **p <.01; OR = Odds Ratio; SE = Standard error of the coefficient.
Table 6.
Outcome Means and Pairwise Comparisons between Profiles (Sample 2)
Standardized Profile Means Profile Comparisons Summary of
Amotivated Moderately Highly Balanced comparisons
1 vs 2 1 vs 3 1 vs 4 2 vs 3 2 vs 4 3 vs 4
(P.1) Autonomous (P.2) Motivated (P.3) (P.4)
In-Role Performance -0.408 0.267 0.400 -0.249 -0.675** -0.808** -0.159 -0.133 0.516** 0.650** 1 = 4 < 2 = 3
Extra-Role 1=4<2=3
-0.202 0.319 0.496 -0.408 -0.521** -0.697** 0.206 -0.177 0.727** 0.904**
Performance
Job Satisfaction -1.544 0.505 0.646 -0.138 -2.049** -2.190** -1.406** -0.140 0.643** 0.784** 1< 4 < 2 = 3
Engagement -1.283 0.538 0.684 -0.271 -1.821** -1.967** -1.012** -0.146 0.809** 0.955** 1 < 4 < 2 = 3
Burnout 1.257 -0.372 -0.423 0.030 1.629** 1.681** -1.228 ** 0.051 -0.401** -0.453** 2 = 3 < 4 < 1
Note. *p <.05; **p <.01; Indicators are estimated from factor scores with mean of 0 and a standard deviation of 1.
Figure 1. Elbow Plot for the Information Criteria in Sample 1 (left) and 2 (right).
Amotivated Moderately Autonomous Highly Motivated Balanced

(27.6%) (11.5%) (25.6%) (35.3%)
1.5
0.5
-0.5
-1
-1.5
Amotivation External-Material External-Social Introjected Identified Intrinsic
Figure 2. Sample 1 Profiles (n = 723)
Note. Indicators are estimated from factor scores with mean of 0 and a standard deviation of 1.

(13.1%) (27.8%) (22%) (37.1%)
2
1.5
0.5
-0.5
-1
-1.5
-2
Amotivation External-Material External-Social Introjected Identified Intrinsic
Figure 3. Sample 2 Profiles (n = 286)

(13.1%) (27.8%) (22%) (37.1%)
2.5
2
1.5
1
0.5
0
-0.5
-1
-1.5
-2
-2.5
Extra-Role Performance In-Role Performance Job Satisfaction Engagement Burnout
Figure 4. Outcomes Associated with Profile Membership

Motivation Profiles at Work 1: Acknowledgements

Uploaded by

Copyright:

Available Formats

Motivation Profiles at Work 1: Acknowledgements

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Motivation Profiles at Work 1: Acknowledgements

Uploaded by

Copyright:

Available Formats

MOTIVATION PROFILES AT WORK 1

Running Head: MOTIVATION PROFILES AT WORK

Motivation Profiles at Work: A Self-Determination Theory Approach

Joshua Howard and Marylène Gagné

University of Western Australia

Alexandre J.S. Morin

Institute for Positive Psychology and Education, Australian Catholic University

Anja Van den Broeck

KU Leuven & North-West University

awarded to the third author.

School of Psychology, University of Western Australia,

35 Stirling Hwy, Crawley 6009 WA, Australia.

regulated, autonomously regulated and highly motivated employees. In Sample 1, governmental

workplace outcomes than more controlling forms.

is an inherently complex concept as evidenced by the variety of approaches to its conceptualization

complementary, types of behavioral regulations experienced by individuals to varying degrees. While

workers by adopting a person-centered approach to work motivation. While variable-centered analyses,

to assess how configurations of motivation factors are organized within individuals.

(Morin, Morizot, Boudrias, & Madore, 2011; Zyphur, 2009).

SDT conceptualizes motivation as multiple distinguishable facets, each representing a different

extrinsic motivation can be internalized to become autonomously regulated. Identified regulation, an

positive self-feelings such as pride. External regulation, a non-internalized form of extrinsic

externally derived rewards or punishments. The most current conceptualization of workplace

focuses on material rewards, and the avoidance of losing one’s job.

self-determination continuum to consider.

In addition to the empirical evidence demonstrating the negative influence of amotivation on

the same continuum as the behavioral regulations, with no evidence of discontinuity.

continuum hypothesis would be demonstrated if profiles follow a smooth increase/decrease in the

regulations; Grolnick & Ryan, 1987).

autonomous to controlled motivation more influential? How do amotivated employees compare to

autonomous/low controlled motivation (HA/LC), high autonomous/high controlled motivation

their results or to objectively assess the adequacy of the analyses.

dominated by autonomous forms of regulation, a profile dominated by controlled forms of regulation,

variable-centered analyses in terms of generalizability and in providing a meaningful representation of

support a substantive interpretation of latent profiles is to embark on a process of construct validation

Predictors of Motivation Profiles

employees corresponding to predominantly autonomously-driven profiles (i.e., higher in intrinsic

1997). Thus, we expected motivational profiles characterized by higher levels of autonomous

incorporated blue-collar manufacturing, white-collar technological, and white-collar governmental job

categories as predictors of profile membership to test these hypotheses.

Consequences of Motivation Profiles

overall “quantity” of motivation that characterizes a specific profile.

some variable-centered research (Battistelli, Galletta, Portoghese, & Vandenberghe, 2013).

Schaufeli, & Leiter, 2001) as potential outcomes of employees’ motivational profiles.

Participants and Procedure

validity of scales (Gagné et al., 2015).

Preliminary Measurement Models

Asparouhov et al. (2015, p. 1564) note that:

(Sample 2: |r| = .021 to .844; M|r| = .401)].

Latent Profile Analyses

& Bauer, 2006). All models converged on well replicated solutions.

Latent Profile Analyses with Predictors and Outcomes

these indicators failed to reach a minimum, we relied on a graphical representation of these

motivation decrease as a direct function of their relative degree of self-determination as proposed by

continuum structure of self-regulation proposed by SDT. Finally, Profile 4 characterizes employees

motivation (35.3% in Sample 1; 37.1.0% in Sample 2).

Predictors of Profile Membership

employees presented a lower likelihood of membership in the moderately autonomously motivated

Outcomes of Profile Membership