Financially focused self-concept:
Measurement invariance, stability, and concurrent validity

Nassim Tabri1
Kahlil S. Philander2,3
Richard T. Wood4
and
Michael J. A. Wohl1

Department of Psychology, Carleton University, Ottawa, ON, Canada

School of Hospitality Business Management, Carson College of Business, Washington State
University, Everett, WA, USA

School of Psychology, University of Sydney, 94 Mallet Street, Camperdown, NSW, Australia

GamRes Limited, Hawkesbury, ON, Canada.

Correspondence concerning this article should be addressed to Nassim Tabri, Department of
Psychology, Carleton University, 1125 Colonel By Drive, Ottawa, Ontario, Canada, K1S 5B6.
E-mail: nassim.tabri@carleton.ca

Abstract
People with a financially focused self-concept overvalue the importance of financial success for
self-definition and self-worth. Herein, we examined whether the measurement properties of the
financially focused self-concept scale (FFS) are invariant over time and across different
demographic groups of people who gamble. We also assessed the temporal stability and
concurrent validity of the FFS. In Study 1 (N=147), community players completed the 4-item
FFS and problem gambling severity index (PGSI) on two assessments about four weeks apart. In
Study 2 (N=135), students who gamble completed the FFS and PGSI about 6-7 months apart. In
Studies 1 and 2, FFS had strong temporal measurement invariance and moderate-to-high
temporal stability. FFS and PGSI were also positively associated within and across assessments
in Studies 1 and 2. Additionally, in Study 2, an increase in FFS over time was accompanied by
an increase in PGSI. FFS also had partial measurement invariance across community and student
players. Findings provide support for assumptions of longitudinal and comparative research
designs. Findings are anchored in prior research on the effects of a focused self-concept in other
domains. We also call for longitudinal and comparative gambling research on the antecedents
and consequences of a financially focused self-concept.

Keywords: self-concept; disordered gambling; financial success; measurement invariance.

Financially focused self-concept:
Measurement invariance, stability, and concurrent validity
Some individuals overvalue the importance of their financial success for self-definition
and self-worth (Tabri, Wohl, Eddy, & Thomas, 2017). A growing body of research indicates
having a financially focused self-concept has predictive utility to identify players at risk of
developing gambling problems (Tabri, Hollingshead, & Wohl, 2019; Tabri, Salmon, & Wohl, in
press; Tabri, Shead, & Wohl, 2017; Tabri, Werner, Milyavkaya, & Wohl, 2018; Tabri, Wohl, et
al., 2017; Tabri, Wohl, Wood, & Philander, 2018; Tabri, Wood, Philander, & Wohl, in press;
Wood, Wohl, Tabri, & Philander, 2017). For example, among community players, Tabri, Wohl,
et al. (2017) showed that greater financial focus was uniquely and moderately associated with
disordered gambling severity above and beyond known risk factors for gambling, including
personal income, Big-Five personality domains, global self-esteem, personal relative deprivation
and materialism. Critically, however, this body of work used a cross-sectional research design,
which limits the ability to draw causal inferences about the nature of the relation between
financially focused self-concept and gambling outcomes.
In the current research, we tested whether the same financially focused self-concept
construct can be measured over time and across different types of players. To this end, we
conducted a longitudinal study of community members (Study 1) and university students (Study
2) who gamble that assessed financial focus at two time points and examined the measurement
invariance of the financially focused self-concept construct over time and between different
types of players.
Financial focus and disordered gambling

A defining aspect of some individuals’ self-concept is their perceived financial success
(Park, Ward, & Naragon-Gainey, 2017; Tabri, Wohl, et al., 2017). Such individuals are
motivated to act in ways they believe will increase their wealth. This is because individuals
generally seek means to bolster their self-worth in domains on which their self-worth is staked
(Crocker & Knight, 2005; Crocker & Park, 2004). Accordingly, financially focused people often
try to increase the amount of money they have through conventional avenues (e.g., hard work
and saving; see Tabri, Wohl et al., 2017). However, people with a financial focus may also try to
boost their wealth (and thus self-worth) through unconventional avenues such as gambling,
especially if they believe conventional avenues to financial success are inaccessible (Tabri,
Dupuis, Kim, & Wohl, 2015).
Unfortunately, gambling for financial gain is a well-established predictor of disordered
gambling (e.g., Clarke et al., 2006; Hahn, Wilson, McRae, & Gilbert, 2013; Lister, Nower, &
Wohl, 2016; MacLaren, Ellery, & Knoll, 2015; Tabri et al., 2015). Players who are high in
financial focus are therefore at heightened risk for disordered gambling compared to players low
in financial focus (Tabri et al., 2019; Tabri, Salmon, et al., in press; Tabri, Shead et al., 2017;
Tabri, Werner et al., 2018; Tabri, Wohl, et al., 2017).
One reason for the link between a financially focused self-concept and disordered
gambling symptomatology is that players high in financial focus are more likely to gamble
irresponsibly (Tabri et al., 2019; Tabri, Wood, et al., in press; Wood et al., 2017). For instance,
players high in financial focus score low on measures of responsible gambling beliefs (e.g.
believing that gambling is not a good way to make money) and behaviors (e.g. setting a money
limit on gambling expenditures), compared to players low in financial focus (Tabri, Wood, et al.,
in press; Wood et al., 2017). Electronic gaming machine players high in financial focus were

shown to spend more money on their gambling activities compared to players low in financial
focus, despite receiving pop-up messages indicating that they were approaching their money
limit (Tabri et al., 2019).
Another reason for the association between a financially focused self-concept and
disordered gambling is because those who are more financially focused tend to have etiological
risk factors of gambling disorder compared to those who are less financially focused (Tabri,
Salmon et al., in press; Tabri, Werner et al., 2018; Tabri, Wohl et al., 2018). For example, among
players without gambling problems, Tabri, Wohl et al. (2018) showed that having a financial
focus was moderately and positively associated with etiological risk factors for disordered
gambling, including erroneous gambling beliefs, financial dissatisfaction, low perceived selfefficacy to cope with life adversity, and impulsivity. These findings dovetail with recent research
linking financial focus among disordered gamblers with etiological subtypes of disordered
gambling specified in the Pathways Model (Blaszczynski & Nower, 2002; also see Nower &
Blaszczynski, 2017). The Pathways Model describes three distinct subtypes of disordered
gambling that are defined by unique vulnerabilities. Behaviourally conditioned players do not
have any etiological risk factors (e.g. premorbid psychopathology) but develop gambling
problems due to the reinforcement contingencies of gambling. Emotionally vulnerable players
are vulnerable to stress and negative affect (e.g., mood disorders), and develop gambling
problems because they gamble to escape these aversive states. Anti-social impulsive players have
impulsivity and anti-social personality traits and develop gambling problems because they
gamble to enhance positive affect. Tabri, Salmon, et al. (in press) showed that having a financial
focus was greater among disordered players who were classified as emotionally vulnerable or
anti-social impulsive compared to disordered players classified as behaviorally conditioned.

Although prior research suggests that financially focused self-concept may proliferate
and maintain disordered gambling, the extant research relied solely on cross-sectional research
designs. As such, it is difficult to draw causal inferences about the nature of the relation between
financially focused self-concept and gambling outcomes. To help determine the nature of this
relation, longitudinal (e.g., do financial focus and gambling problems reinforce each other over
time?) and experimental (e.g., does manipulating beliefs about financial success affect gambling
behavior?) research is needed.
To facilitate longitudinal research, antecedent psychometric research is required to
confirm that the measurement of the financially focused self-concept is invariant over time. This
is important because when measurement properties of a given scale are not temporally invariant,
results from longitudinal data analyses may lead to ambiguous or misleading conclusions about
change (Little, 2013; Lord, 1958; Newsom, 2015). Likewise, before proceeding with
experimental or correlational research comparing different types of players, antecedent
psychometric research is needed to confirm that the same financially focused self-concept
construct is being measured among different types of players. This is important because when
measurement properties of a given scale are not invariant across groups, results from
comparative analyses may lead to ambiguous or misleading conclusions about differences
between groups.
Overview of the current research
The current research had two aims. The first aim was to examine whether the
measurement of the financially focused self-concept construct is invariant over time. The second
aim was to examine the measurement invariance of the financially focused self-concept construct
between different demographic groups of people who gamble. To accomplish the first aim, we

examined the longitudinal measurement invariance, stability, and concurrent validity of the
financially focused self-concept construct in two studies involving different demographic groups
of people who gamble.
In Study 1, we conducted an analysis of existing data from players who gamble with a
Canadian provincial gaming provider. The data were collected as part of a large research project
in which associations examined in the current research were not reported in the original study
(see Wood et al., 2017; Study 3). Participants completed the 4-item financially focused selfconcept scale (FFS; Tabri Wohl, et al., 2017) and the Problem Gambling Severity Index (PGSI;
Ferris & Wynne, 2001) twice, approximately one month apart. To understand whether the
financially focused self-concept construct is invariant over time, we examined the temporal
measurement invariance of the 4-item FFS as well as the stability and change in the 4-item FFS
over a one-month period. We also examined the concurrent validity of the 4-item FFS in relation
to the PGSI over time, and whether an increase in FFS is accompanied by an increase in PGSI.
In Study 2, the goal was to replicate and extend Study 1 with a sample of people who
gamble that are enrolled in university. They completed the FFS and PGSI measures on two
assessments approximately 6-7 months apart. To date, no published research has examined
financially focused self-concept among university students who gamble and doing so is
important for external validity of the financially focused self-concept construct. In addition,
examining financially focused self-concept among university students is important because
recent research has shown that college and university students who view money as an important
indicator of success, prestige, and power were more likely to gamble and to have severe
gambling problems (e.g., Lostutter et al. 2019).

To accomplish the second aim, we examined whether the measurement invariance of the
financially focused self-concept between different demographic groups of people who gamble.
More specifically, we tested whether the FFS indicators at time 1 are invariant across
participants in Study 1 (an older community sample of people who gamble) and participants in
Study 2 (a younger student sample of people who gamble). We examined measurement
invariance of the FFS indicators across both groups at time 1 only because we expected the FFS
indicators at times 1 and 2 to be invariant within each group.
All research materials, statistical input and output files, and the analyzed data (for Study
2 only) reported herein can be accessed via the Open Science Framework (OSF):
https://osf.io/c49ft/?view_only=b1cca60c236944ab96a174b803490cbd
Study 1
Method
Participants, procedure, and materials. Participants were 147 (49 men and 51women)
community player who were customers of the British Columbia Lottery Corporation (see Study 3
in Wood et al., 2017). In terms of self-reported age, 43.5% of participants were +65 years old,
27.2% were in the 55-64 age band, 16.3% were in the 45-54 age band, 7.5% were in the 35-44
age band, 4.1% were in the 25-34 age band, and 1.4% were in the 19-24 age band. They
completed the 4-item FFS twice about one month apart. The items were “How I feel about
myself is largely based on the amount of money I have,” “My moods are influenced by the
amount of money I have,” “People will think less of me if I don’t have a lot of money,” and “The
opportunities that are available to me depend on the amount of money I have.” Responses were
anchored at 0 (not at all) and 4 (extremely). The items at each time point were internally

consistent (time 1 α = 0.804, time 2 α = 0.821). An average at each time point was calculated
with higher number reflecting greater financial focus.
To assess severity of disordered gambling, participants completed the Problem Gambling
Severity Index (PGSI; Ferris & Wynne, 2001) at both assessments. The PGSI includes nine items
that measure the extent of problem gambling behaviors (e.g., “Have you gone back another day
to try and win back the money you lost?”) and the consequences of engaging in problem
gambling behaviors (e.g., “Has your gambling caused any financial problems for you or your
household?”). Participants responded by indicating how frequently they engaged in problem
gambling behaviors and experienced consequences due to their gambling behavior. Responses
were anchored at 0 (never) and 3 (almost always). At time 1, the PGSI assessed severity of
disordered gambling over the last 12 months. At time 2, the PGSI assessed severity of disordered
gambling over the last month. Items at each time point were internally consistent (time 1 α =
0.966, time 2 α = 0.831). Total scores were calculated at each time point with higher numbers
reflecting greater disordered gambling severity.
Data analytic plan. The longitudinal measurement invariance of the 4-item FFS scale at
times 1 and 2 was assessed using structural equation modelling (SEM; see Kline 2016). This
involves conducting a series of sequential models that test the equality of different model
parameters. In the first model, known as the configural invariance model, the factor
configuration of the FFS is examined simultaneously across the two measurement occasions,
which involves testing a confirmatory two-factor model of FFS (with four observed indicators)
on two occasions. In the model, residual covariances between the same indicators across
measurement occasions are included and no equality constraints are included on any model
parameters across the two occasions. For model identification purposes, factor means were fixed

to zero and factor variances were fixed to one. A well-fitting configural model is necessary
before testing the equality of model parameters across measurement occasions. The likelihood
ratio χ2 test of model fit, comparative fit index (CFI), and root mean square error of
approximation (RMSEA) and its 95% confidence interval were used to adjudicate fit of the
configural model. An excellent model fit would be reflected by a statistically nonsignificant χ2, a
CFI close to 1, and RMSEA of .05 or less with a 95% confidence interval that includes zero (see
Kline 2016).
With a well-fitting configural model, a series of sequential models were conducted to test
for metric invariance (item factor loading invariance), scalar invariance (item intercept
invariance), and residual variance invariance (item uniqueness invariance) across the two
measurement occasions. When testing for scalar and residual invariance, the mean and variance
of the FFS latent factor at time 2 were freely estimated in the model because factor means are
partly a function of the item intercepts (see Newsom, 2015). The nested models were compared
using chi-square difference tests (Δχ2). Strict temporal measurement invariance is established
when factor loadings, intercepts, and residual variances for identical items across measurement
occasions are equal. After establishing strict temporal invariance, we tested whether there was a
difference between the FFS latent factor means and variances at times 1 and 2 while maintaining
measurement invariance equality constraints. This was accomplished by testing whether an
equality constraint on the latent factor means and variances worsens model fit.
Concurrent validity of the FFS was assessed by examining associations between FFS and
PGSI at times 1 and 2. In addition, if FFS and PGSI scores changed from times 1 to 2, then we
examined whether change in FFS correlates with change in PGSI using a latent difference score
model (see McArdle, 2009). The latent difference score model is akin to conducting a paired

samples t-test (see Coman et al., 2013) for each of FFS and PGSI simultaneously. In the model,
we regressed the latent difference score for each of FFS and PGSI on its time 1 counterpart to
examine the correlation between the residualized difference scores of FFS and PGSI. Doing so
facilitates the interpretation of the association between FFS and PGSI difference scores as being
due to variability that is unexplained by FFS and PGSI at time 1. Model fit was adjudicated using
the likelihood ratio χ2 test, CFI, and RMSEA, but with robust standard errors because PGSI
scores are not normally distributed.
Results
Temporal measurement invariance. Descriptive statistics and correlations between the
four FFS items on both measurement occasions are reported above the diagonal in Table 1. The
configural invariance model provided an excellent fit to the data, χ2(15) = 20.04, p = 0.17, CFI =
0.99, and RMSEA = 0.05 [.00, .10]. The metric invariance model did not fit worse than the
configural invariance model, Δχ2(4) = 4.34, p = 0.36. Likewise, the scalar invariance model did
not fit significantly worse than the metric invariance model, Δχ2(4) = 0.16, p = 0.99. Further, the
residual variance invariance model did not fit significantly worse than the scalar invariance
model, Δχ2(2) = 4.85, p = 0.09. The findings show that the 4-item FFS has strict measurement
invariance over time. Standardized factor loadings as well as unstandardized intercepts and
residual variances are reported in Table 2. There was a moderate-to-large positive correlation
between the latent FFS factors at times 1 and 2, r = .67, z = 11.42, p < .001.
Next, we tested whether there was a difference between the latent factor means and
variances of FFS over time in the context of strict temporal invariance. Imposing equality
constraints on the means and variances at times 1 and 2 did not worsen model fit relative to the

strict invariance model, Δχ2(2) = 0.38, p = 0.83. Thus, there was no change in FFS over a onemonth period.
Concurrent validity. Descriptive statistics and zero-order correlations between FFS and
the PGSI total score at times 1 and 2 are reported in the top diagonal of Table 3. FFS and PGSI
were moderately and positively correlated at each time point and across time points. However,
like FFS, there was no change in the PGSI total score from time 1 to time 2, t(146) = -1.15, p =
.25. Indeed, based on the scoring criteria of the PGSI, at time 1, almost all participants were nonproblem gamblers (n = 139; 94.6%; i.e., a PGSI total score of zero), few were at low risk for
gambling problems (n = 5; 3.4%; i.e., a PGSI total score between 1 and 2), even fewer were at
moderate risk for gambling problems (n = 4; 1.4%; i.e., a PGSI total score between 3 and 7), and
only one was at high risk for gambling problems (.7%; i.e., a PGSI total score above 8). At time
2, almost all participants were non-problem gamblers (n = 142; 95.5%), few were at low risk for
gambling problems (n = 3; 2%), one was at moderate risk for gambling problems (.7%), and only
one was at high risk for gambling problems (.7%). Thus, we did not estimate the latent difference
score model to examine whether change in FFS is accompanied by change in PGSI.
Discussion
The findings of Study 1 are significant because they demonstrate that the same
financially focused self-concept construct is being measured over time. In terms of stability, we
observed a moderate-to-strong positive correlation between financial focus at time 1 and 2,
which suggests that the relative rank-order of participants was stable over time. These findings
were observed over a one-month period and there was no mean difference in financial focus over
time.

In terms of concurrent validity, financial focus and disordered gambling severity were
positively correlated at each assessment and across assessments. However, because there was
little or no change in financial focus and disordered gambling severity, we could not assess
whether change in financial focus accompanies change in disordered gambling severity. The lack
of change is not surprising given the short duration (one month) between assessments. As such,
we examined the longitudinal measurement invariance, stability, and concurrent validity of the
financially focused self-concept construct over a longer duration in Study 2.
Study 2
Method
Participants, procedure, and materials. Participants were recruited from a large pool of
newly enrolled undergraduate university students at the beginning of the 2017-2018 academic
year. Students were eligible to participate in our study if they indicated that they have engaged in
gambling activities (e.g., slot machines, poker, blackjack, roulette, sports betting) and had done
so within the past 12 months. A total of 308 students (94 men and 214 women) were eligible.
They ranged in age from 17 to 63 (M = 21.29, SD = 6.72) years.
All 308 participants initially completed a brief survey that included the 4-item FFS (α =
0.776) and PGSI (α = 0.902). We re-contacted all 308 participants from the initial session
approximately 6-7 months later to complete a brief follow-up survey. The follow-up survey
included the 4-item FFS (α = 0.728) and PGSI (α = 0.845), among other questionnaires (see
OSF). Participants were instructed to complete the PGSI for the period from Fall 2017 (i.e., the
initial time point) to the present. A total of 135 students (45 men and 90 women) completed the
follow-up survey, but four participants did not complete the PGSI. Participants with missing data
on the PGSI were included in the analyses using maximum likelihood (see Enders, 2010).

Participants ranged in age from 17 to 63 years (M = 21.56, SD = 6.63). Thus, the data of 135
participants who completed the survey on both measurement occasions were included in the
analyses. The data analytic plan was identical to that of Study 1.
All participants provided informed consent and received course credit for participating in
the second wave of data collection. The first authors’ Research Ethics Board approved Study 2.
Results
Temporal measurement invariance. Descriptive statistics and correlations between all
FFS items on both measurement occasions are reported in the top diagonal of Table 2. The
configural invariance model provided a marginal fit to the data, χ2(15) = 30.38, p = 0.01, CFI =
0.96, and RMSEA = 0.09 [.04, .13]. Modification indices suggested that the inclusion of three
residual covariances would enhance model fit. These covariances were between “The
opportunities that are available to me depend on the amount of money I have” at time 2 and
“People will think less of me if I don’t have a lot of money” at times 1 and 2 as well as between
“The opportunities that are available to me depend on the amount of money I have” and “People
will think less of me if I don’t have a lot of money” both at time 1. The inclusion of the three
residual covariances improved model fit, Δχ2(3) = 12.56, p = 0.005. The revised configural
model provided an excellent fit to the data, χ2(12) = 17.44, p = 0.12, CFI = 0.98, and RMSEA =
0.06 [.00, .11]. The metric invariance model did not fit worse than the configural invariance
model, Δχ2(4) = 2.51, p = 0.64. Likewise, the scalar invariance model did not fit significantly
worse than the metric invariance model, Δχ2(8) = 1.80, p = 0.99. Further, the residual variance
invariance model did not fit significantly worse than the scalar invariance model, Δχ2(4) = 2.20,
p = 0.70. Standardized factor loadings as well unstandardized intercepts and residual variances

are reported in Table 3. There was a moderate-to-large positive correlation between the latent
factors of FFS at times 1 and 2, r = .69, z = 8.89, p < .001.
Next, we examined whether there was a difference in the mean and variance of the FFS
latent factors at time 1 and time 2 in the context of strict measurement invariance. Constraining
the mean and variance of each latent factor to be equal to at times and 2 worsened model fit,
Δχ2(2) = 16.05, p = 0.0003. The reason is that there was a small-to-moderate increase in the
latent mean of FFS from time 1 to time 2, B = .32, z = 3.91, p < .001.
Concurrent validity. Descriptive statistics and correlations between the observed
variables for FFS and PGSI total score at times 1 and 2 are reported in the bottom diagonal of
Table 1. Four participants did not complete the PGSI at time 2. There were moderate and
positive correlations between FFS and PGSI at each time point and across time points. Like FFS,
there was an increase in the PGSI total score from time 1 to time 2, t(130) = 6.87, p < .001.
Based on the scoring criteria of the PGSI, at time 1, 61.5% of participants were non-problem
gamblers (n = 83), 30.4% were at low risk for gambling problems (n = 41), 5.2% were at
moderate risk for gambling problems (n = 7), and 3% were at high risk for gambling problems (n
= 4). A time 2, 31.1% were non-problem gamblers (n = 42), 37% were at low risk for gambling
problems (n = 50), 21.5% were at moderate risk for gambling problems (n = 29), and 7.4% were
at high risk for gambling problems (n = 10).
Thus, we examined whether an increase in FFS from time 1 to time 2 correlates
positively with an increase in the PGSI total score from time 1 to time 2 using a latent difference
score model. The model provided an excellent fit to the data, Δχ2(2) = 2.74, p = 0.25, CFI = 1,
and RMSEA = .05 [.00, .19]. Results showed that the increases in FFS and PGSI from time 1 to
time 2 were positively associated, covariance = .23, z = 2.044, p = .04, r = .18.

Discussion
The results of Study 2 replicate and extend Study 1 with a sample of university students
who gamble. More specifically, we observed that the 4-item FFS has strong temporal
measurement invariance over 6-7 months. As in Study 1, we observed a moderate-to-strong
positive correlation between the financial focus latent factors at times 1 and 2. This finding
suggests that the relative rank-order of participants remains the same over a one-month period
(Study 1) and a 6-7 month period (Study 2). Unlike Study 1, we observed a small-to-moderate
increase in the latent factor means of financial focus. Of note, this analysis statistically controlled
for measurement error in the observed FFS indicators using latent variables. As such, Study 2 is
the first to show that individuals’ financial focus can change over time.
In terms of concurrent validity, like Study 1, the observed measures of financial focus
and disordered gambling severity were positively correlated within and across assessments.
However, unlike Study 1, there was a small-to-moderate increase in the observed measures of
financial focus and disordered gambling severity over the 6-7-month period. Of note, the
increase in financial focus was positively correlated with the increase in disordered gambling
severity.
Studies 1 and 2 measurement invariance of the 4-item FFS across community and
student players
Thus far, we showed that the same financially focused self-concept construct is being
measured over time among community (Study 1; N = 147) and student (Study 2; N = 135)
players. To examine whether the same financially focused self-concept construct is being
measured among different types of players, we tested whether the 4-item FFS is invariant across
community and student gamblers at time 1. To accomplish this, we first estimated a multi-group

configural model to examine whether the same factor configuration exists among community and
student gamblers, respectively. As determined in Study 2, a residual correlation was included
between the item “The opportunities that are available to me depend on the amount of money I
have” and “People will think less of me if I don’t have a lot of money” for student gamblers.
With a well-fitting configural model, we then tested a sequential series of models that
constrained model parameters to be equal across community and student gamblers, which were
the metric invariance model (item factor loadings are constrained to be equal) and scalar
invariance model (item intercepts are constrained to be equal).
Although item residual invariance is not a prerequisite for testing mean differences
between groups on a latent factor (see Vandenberg & Lance, 2000), we examined the residual
invariance model (item residual variances are constrained to be equal) because it is included in
many tests of measurement invariance. As such, with at least partial scalar invariance, we
proceeded to explore differences in factor variances and means between community and student
gamblers. Model estimation, fit, and comparisons were conducted as in Studies 1 and 2.
The configural model provided a good fit, χ2(3) = 7.712, p = 0.05, CFI = 0.99, and
RMSEA = 0.12 [.00, .20]. The metric invariance model did not fit worse than the configural
invariance model, Δχ2(3) = 2.03, p = 0.57. However, the scalar invariance model did fit
significantly worse than the metric invariance model, Δχ2(3) = 17.04, p = 0.0007. Modification
indices indicated that freeing the equality constraint on the intercept for item “The opportunities
that are available to me depend on the amount of money I have” would enhance model fit.
Indeed, the scalar invariance model did not fit worse than the metric invariance model after
freeing the equality constraint, Δχ2(2) = 2.04, p = 0.36. Community players had a lower intercept
value on the non-invariant item compared to student players (see Table 4). Thus, the results

indicate that the 4-item FFS has partial measurement invariance across community and student
players.
Furthermore, the residual variance invariance model fit significantly worse than the
partial scalar invariance model, Δχ2(4) = 86.86 p < 0.001. Modification indices indicated that all
equality constraints on the residual variances would need to be removed to enhance model fit.
As well, the model in which factor means were held equal across groups fit worse than the partial
scalar invariance model, Δχ2(1) = 13.35, p = 0.003. Likewise, the model in which factor
variances were held equal across groups fit worse than the partial scalar invariance model, Δχ2(1)
= 5.71, p = 0.02. Results indicated that the factor mean for student players was lower than the
factor mean for community players, B = -0.39, z = -2.38, p = 0.02. As well, the variance of the
factor mean for student players was larger than the variance of the factor mean for community
players, B = 2.09, z = 5.03, p < 0.001. Taken together, community players were more financially
focused than student players, and there was greater variability in financial focus among student
players relative to community players.
General Discussion
Financially focused self-concept is theorized to play a role in the etiology and
maintenance of disordered gambling (e.g., Tabri, Salmon, et al., in press). However, almost all
published research on financially focused self-concept relied on cross-sectional research designs,
which limits the ability to make causal inferences. Thus, longitudinal and experimental research
designs are needed to understand the nature of the relation between financially focused selfconcept and gambling outcomes. However, some of the key assumptions of longitudinal and
experimental research are that constructs are measured in the same way over time and across
comparison groups. Herein, we addressed this issue and found that the financially focused self-

concept construct has strong temporal measurement invariance among community members
(Study 1) and among university students (Study 2) who gamble.
We also found that the financially focused self-concept construct has partial measurement
invariance across community and university students who gamble. Together, these findings
provide strong prerequisite support for some of the key assumptions of longitudinal and
experimental research designs. As such, researchers interested in examining financially focused
self-concept over time, among different groups of people who gamble, or both can proceed with
more confidence in the psychometric properties of the 4-item FFS scale.
Implications and directions for future research
An important contribution of the current research was the moderate-to-strong correlation
between financial focus at times 1 and 2 in Studies 1 and 2. This correlation is analogous to a test
retest correlation coefficient. What is remarkable is that the correlation coefficient was virtually
identical in Studies 1 and 2 despite the different durations of time between assessments in Study
1 (one month) and Study 2 (6-7 months). Moreover, the magnitude of the correlation between
financial focus at times 1 and 2 in Studies 1 and 2 is consistent with longitudinal research on
having a focused self-concept in the domain of appearance measured three weeks apart among
university students (see Spangler & Stice, 2001) and measured 10 and 20 months apart among
adolescent women (Spangler, 2002). Together, these findings suggest that the relative rank order
of participants in terms of their financial focus is stable and this degree of stability may be
generalizable to having a focused self-concept in other domains.
The current findings also add to the literature by demonstrating that people’s financial
focus can change, on average, when observed for more than a month. Indeed, in Study 1, the
mean values for financial focus over a one-month period were similar. In contrast, in Study 2, we

found that financial focus moderately increased over a 6-7-month period even after controlling
for measurement error, which effectively rules out regression towards the mean as an alternative
explanation of the observed change. Thus, the observed change in financial focus over the 6-7
months period in Study 2 reflects “true” change. Of note, the increase in financial focus observed
in Study 2 was positively associated with an increase in disordered gambling severity, which
supports the concurrent validity of change in financial focus.
However, given that financial focus and disordered gambling were measured on two
assessments only, it is unclear to know whether an increase in financial focus predicts a
subsequent increase in disordered gambling, whether an increase in disordered gambling predicts
a subsequent increase in financial focus, or both. As such, a direction for future research is to
examine the shape and rate of change in financial focus over time using three or more
assessments and whether change in financial focus has prognostic importance for understanding
the onset and maintenance of gambling disorder.
Furthermore, the current research contributes to a small body of research that has
examined the association between change in a focused self-concept and change in psychological
impairment and distress. For example, Cooper and Fairburn (1993) showed that an increase in
appearance focused self-concept over 10-12 weeks was accompanied by a decrease in selfesteem and an increase in depressive symptoms severity among people diagnosed with bulimia
nervosa. Masheb and Grilo (2003) replicated and extended these results in a sample of people
diagnosed with binge eating disorder over a 4-week period. The current research extends this
body of work to the realm of behavioral addiction, namely disordered gambling. Together, these
results suggest that increases in a focused self-concept are linked to increases in severity of
psychiatric problems.

From a psychometric standpoint, the current research helped establish that three of the
four FFS items were invariant across community and student players with the item “The
opportunities that are available to me depend on the amount of money I have” being only
partially invariant (i.e., the factor loading was invariant, but the intercept was not invariant). Of
note, community players were more financially focused than student players in terms of the
latent factor mean, but student players scored higher on the partially non-invariant observed item
compared to community players. This finding suggests that student players agree more with the
view that having more money facilitates opportunities compared to community players. This
difference may be, partly, due to student players being much younger and thus less financially
established compared to the community players who were older (with most being at least the age
of retirement) and thus more financially established. Although no prior research has directly
compared the money attitudes of student and community members who gamble, published
research suggests that college students who gamble have more anxious attitudes towards money
(e.g., I worry that I will not be financially secure; see Lostutter et al., 2019) compared to
community members who gamble (Blaszczynski & Nower, 2010). Because age and financial
standing are likely positively correlated, it would behoove researchers to examine whether the
measurement of the financially focused self-concept construct is invariant across a range of
income groups balanced by age.
Furthermore, replicating and extending prior research, we observed that financial focus
was positively associated with disordered gambling severity in Studies 1 and 2. Of note, a tacit
assumption of prior and current research is that the financially focused self-concept construct is
being measured in the same way across people who gamble without problems and people who
gamble with varying degrees of problems. As such, researchers should investigate and confirm

that the measurement of the financially focused self-concept construct is invariant across
different groups of players defined by having no gambling problems as well as different
thresholds of gambling problems.
Conclusions
The current research is the first to demonstrate that the measurement properties of the 4item FFS scale used to measure the financially focused self-concept construct transcend time and
demographic groups of people who gamble (younger students and older community members).
We also observed that an increase in financial focus was accompanied by an increase in
disordered gambling severity. Together, these findings bolster the psychometric properties of the
FFS scale and promote its use in longitudinal and comparative gambling research.

Reference
Blaszczynski, A., & Nower, L. (2010). Instrumental tool or drug: relationship between attitudes
to money and problem gambling. Addiction Research & Theory, 18, 681–691.
Blaszczynski A., Nower L. 2002. A pathways model of problem and pathological gambling.
Addiction, 97, 487–499.
Clarke, D., Tse, S., Abbott, M., Townsend, S., Kingi, P., & Manaia, W. (2006). Key indicators of
the transition from social to problem gambling. International Journal of Mental Health
and Addiction, 4, 247–264. doi:10.1007/s11469-006-9024-x
Coman, E. N., Picho, K., McArdle, J. J., Villagra, V., Dierker, L., & Iordache, E. (2013). The
paired t-test as a simple latent change score model. Frontiers in Psychology, 4, 738. doi:
10.3389/fpsyg.2013.00738
Cooper, P. J., & Fairburn, C. G. (1993). Confusion over the core psychopathology of bulimia
nervosa. International Journal of Eating Disorders, 13, 385-389.
Crocker, J., & Knight, K. M. (2005). Contingencies of self-worth. Current Directions in
Psychological Science, 14, 200–203. doi:10.1111/j.0963-7214.2005.00364.x.
Crocker, J., & Park, L. E. (2004). The costly pursuit of self-esteem. Psychological Bulletin, 130,
392–414. doi:10.1037/0033-2909.130.3.392.
Ferris, J., & Wynne, H. (2001). The Canadian problem gambling index. Ottawa: Canadian
Centre on Substance Abuse.
Hahn, C., Wilson, T. D., McRae, K., & Gilbert, D. T. (2013). ‘show me the money’:
Vulnerability to gambling moderates the attractiveness of money versus suspense.
Personality and Social Psychology Bulletin, 39, 1259–1267. doi:
10.1177/0146167213492431

Kline, R. B. (2016). Principles and practice of structural equation modeling, 4th. New York,
NY: The Guilford Press.
Lister, J. J., Nower, L., & Wohl, M. J. A. (2016). Gambling goals predict chasing behavior
during slot machine play. Addictive Behaviors, 62, 129-34. doi:
10.1016/j.addbeh.2016.06.018
Little, T. D. (2013). Longitudinal structural equation modeling. New York, NY: The Guilford
Press.
Lord, F. M. (1958). Further problem in the measurement of growth. Educational and
Psychological Measurement, XVIII, 437-451.
Lostutter, T. W., Enkema, M., Schwebel, F., Cronce, J. M., Garberson, L. A., Ou, B., Lewis, M.
A., & Larimer, M. E. (2019). Doing it for the money: the relationship between gambling
and money attitudes among college students. Journal of Gambling Studies, 35,143–153.
MacLaren, V. V., Ellery, M., & Knoll, T. (2015). Personality, gambling motives and cognitive
distortions in electronic gambling machine players. Personality and Individual
Differences, 73, 24–28. doi:10.1016/j.paid.2014.09.019
Masheb, R. M., & Grilo, C. M. (2003). The nature of body image disturbance in patients with
binge eating disorder. International Journal of Eating Disorders, 33, 333-341. doi:
10.1002/eat.10139
McArdle, J. J. (2009). Latent variable modeling of differences and changes with longitudinal
data. Annual Review of Psychology, 60, 577-605. doi:
10.1146/annurev.psych.60.110707.163612
Newsom, J. T. (2015). Longitudinal structural equation modeling. New York: Routledge.

Nower, L., Blaszczynski, A. (2017). Development and validation of the gambling pathways
questionnaire (GPQ). Psychology of Addictive Behavior, 31, 95–109.
Park, L. E., Ward, D. E., & Naragon-Gainey, K. (2017). It’s all about the money (for some):
Consequences of financially contingent self-worth. Personality and Social Psychology
Bulletin, 43, 601-622. doi: 10.1177/0146167216689080.
Spangler, D. L. (2002). Testing the cognitive model of eating disorders: The role of
dysfunctional beliefs about appearance. Behavior Therapy, 33, 87-105.
Spangler, D. L., & Stice, E. (2001). Validation of the beliefs about appearance scale. Cognitive
Therapy and Research, 25, 813-827.
Tabri, N., Dupuis, D. R., Kim, H. S., & Wohl, M. J. A. (2015). Economic mobility moderates the
effect of relative deprivation on financial gambling motives and disordered gambling.
International Gambling Studies, 15(2), 309-323. doi: 10.1080/14459795.2015.1046468
Tabri, N., Hollingshead, S. J., & Wohl, M. J. A. (2019). A limit approaching pop-up message
reduces gambling expenditures, except among players with a financially focused selfconcept. International Gambling Studies, 19, 327-338. doi:
10.1080/14459795.2019.1567806
Tabri, N., Salmon, M. M., & Wohl, M. J. A. (in press). Financially focused self-concept in
disordered gambling: Comparisons between non-disordered and disordered gambling
subtypes. Addiction Research & Theory. doi: 10.1080/16066359.2020.1714039
Tabri, N., Shead, N. W., & Wohl, M. J. A. (2017). Me, myself, and money II: Relative
deprivation predicts disordered gambling severity via delay discounting among gamblers
who have a financially focused self-concept. Journal of Gambling Studies, 33(4), 1201 –
1211. doi: 10.1007/s10899-017-9673-7

Tabri, N., Werner, K. M., Milyavskaya, M., & Wohl, M. J. A. (2018). Perfectionism predicts
disordered gambling via financially focused self-concept. Journal of Gambling Issues,
38, 252 – 267. doi: 10.4309/jgi.2018.38.13
Tabri, N., Wohl, M. J. A., Eddy, K. T., & Thomas, J. J. (2017). Me, myself, and money: Having
a financially focused self-concept and its consequences for disordered gambling.
International Gambling Studies, 17(1), 30 – 50. doi: 10.1080/14459795.2016.1252414
Tabri, N., Wohl, M. J. A., Wood, R. T. A., & Philander, K. (2018). Financially focused selfconcept is associated with etiological and maintenance factors of gambling disorder
among non-problem gamblers. Journal of Gambling Issues, 39, 308 – 313. doi:
10.4309/jgi.2018.39.12
Tabri, N., Wood, R. T. A., Philander, K., & Wohl, M. J. A (in press). An examination of the
validity and reliability of the Positive Play Scale: Findings from a Canadian national
study. International Gambling Studies. doi: 10.1080/14459795.2020.1732442
Wood, R. T. A., Wohl, M. J. A., Tabri, N., & Philander, K. (2017). Measuring responsible gambling
amongst players: Development of the Positive Play Scale. Frontiers in Psychology, 8, 227.
doi: 10.3389/fpsyg.2017.00227
Vandenberg, R. J., & Lance, C. E. (2000). A review and synthesis of the measurement invariance
literature: Suggestions, Practices, and recommendations for organizational research.
Organizational Research Methods, 3, 4-70. doi: 10.1177/109442810031002

Table 1.
Descriptive statistics and inter-correlations for all items in the FFS at both measurement occasions in Studies 1 and 2.
M(SD)
M(SD)
1. FFS1 Time 1
1.35(1.14)
—
0.74**
0.53**
0.50**
0.47**
0.55** 0.35** 0.34** 1.65(0.76)
2. FFS2 Time 1
1.57(1.16)
0.69**
—
0.45**
0.44**
0.41**
0.58** 0.28** 0.26** 1.73(0.77)
3. FFS3 Time 1
1.21(1.18)
0.49**
0.41**
—
0.47**
0.32**
0.42** 0.62** 0.36** 1.44(0.71)
4. FFS4 Time 1
2.46(1.26)
0.40**
0.40**
0.42**
—
0.27**
0.41** 0.37** 0.57** 2.19(0.99)
5. FFS1 Time 2
1.61(1.13)
0.55**
0.51**
0.29**
0.27**
—
0.69** 0.54** 0.52** 1.69(0.75)
6. FFS2 Time 2
1.93(1.10)
0.56**
0.53**
0.21*
0.28**
0.70**
—
0.53** 0.57** 1.75(0.77)
7. FFS3 Time 2
1.49(1.13)
0.25**
0.24**
0.47**
0.28**
0.43**
0.36**
—
0.36** 1.45(0.65)
8. FFS4 Time 2
2.60(1.13)
0.33**
0.31**
0.23**
0.47**
0.25**
0.31** 0.36**
—
2.23(1.01)
Note. FFS = financially focused self-concept. FFS1 = How I feel about myself is largely based on the amount of money I have;
FFS2 = My moods are influenced by the amount of money I have; FFS3 = People will think less of me if I don’t have a lot of
money; FFS4 = The opportunities that are available to me depend on the amount of money I have. Descriptive statistics and
correlations above the diagonal are for Study 1 and descriptive statistics and correlations below the diagonal are for Study 2.
*p < .05; **p < .01.
Study 1 N = 135 and Study 2 N = 147.

Table 2
Factor loadings, intercepts, and residual variance of the FFS items constrained to be equal across time in Studies 1 and 2
Study 1
Study 2
Item

FL

INT

RES

FL

INT

RES

1. How I feel about myself is largely based on the amount of money I have

0.64

1.67

0.15

1.00

1.32

0.32

2. My moods are influenced by the amount of money I have

0.63

1.74

0.19

0.94

1.60

0.44

3. People will think less of me if I don’t have a lot of money

0.42

1.44

0.29

0.65

1.24

0.98

4. The opportunities that are available to me depend on the amount of money I have 0.61 2.21 0.60 0.49
2.45 1.18
Note. FL = factor loading; INT = intercept; RES = residual variance. Reported factor loadings are standardized for ease of
comparison across Studies 1 and 2, whereas intercepts and residual variances are unstandardized. All estimates are statistically
significant, p < .05.
Study 1 N = 147 and Study 2 N = 135.

Table 3.
Descriptive statistics and inter-correlations between FFS and PGSI in Studies 1 and 2
Variable
M(SD)
M(SD)
1. FFS Time 1
1.65(0.92)
—
.28**
.63**
.29**
1.75(0.64)
2. PGSI total score Time 1
1.13(2.80) 0.19*
—
.25**
.95**
0.19(1.45)
3. FFS Time 2
1.91(0.83) 0.63** 0.07
—
.23**
1.78(0.65)
†
4. PGSI total score Time 2
2.36(3.36) 0.21*
0.80**
0.17
—
0.13(0.92)
Note. FFS = financially focused self-concept; PGSI = problem gambling severity index. Descriptive
statistics and correlations above the diagonal are for Study 1 and descriptive statistics and correlations
below the diagonal are for Study 2. Sample size for descriptive statistics and correlations in Study 2 for
PGSI at time 2 is 131 because four participants had missing PGSI data at T2.
†

p = .05; *p < .05; **p < .01.

Study 1 N = 147 and Study 2 N = 135.

Table 4
Measurement invariance parameters of the FFS items across community (Study 1) and student players (Study 2)
Community

Student

Item

FL

INT

RES

FL

INT

RES

1. How I feel about myself is largely based on the amount of money I have

0.69

1.64

0.10

0.69

1.64

0.30

2. My moods are influenced by the amount of money I have

0.62

1.75

0.21

0.62

1.75

0.52

3. People will think less of me if I don’t have a lot of money

0.44

1.42

0.33

0.44

1.42

0.97

4. The opportunities that are available to me depend on the amount of money I have 0.48 2.19 0.63 0.48
2.66 1.22
Note. FL = factor loading; INT = intercept; RES = residual variance. Reported factor loadings, intercepts and residual variances are
unstandardized. All estimates are statistically significant, p < .05.
Study 1 N = 147 and Study 2 N = 135.