International Gambling Studies

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/rigs20

An examination of the validity and reliability of
the Positive Play Scale: findings from a Canadian
national study
Nassim Tabri , Richard T. A. Wood , Kahlil Philander & Michael J. A. Wohl
To cite this article: Nassim Tabri , Richard T. A. Wood , Kahlil Philander & Michael J. A.
Wohl (2020) An examination of the validity and reliability of the Positive Play Scale: findings
from a Canadian national study, International Gambling Studies, 20:2, 282-295, DOI:
10.1080/14459795.2020.1732442
To link to this article: https://doi.org/10.1080/14459795.2020.1732442

Published online: 27 Feb 2020.

Submit your article to this journal

Article views: 127

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=rigs20

2020, VOL. 20, NO. 2, 282–295
https://doi.org/10.1080/14459795.2020.1732442

An examination of the validity and reliability of the Positive
Play Scale: findings from a Canadian national study
Nassim Tabri

a

, Richard T. A. Woodb, Kahlil Philander

c

and Michael J. A. Wohla

a
Department of Psychology, Carleton University, Ottawa, Canada; bGamRes Limited, Hawkesbury, Ontario,
Canada; cSchool of Hospitality Business Management, Carson College of Business, Washington State
University, Everett, WA, USA

ABSTRACT

ARTICLE HISTORY

The Positive Play Scale (PPS) is a self-report measure that assesses
responsible gambling beliefs and behaviours amongst players. The
PPS was shown to be a reliable and valid measure that consisted of
four subscales: Personal Responsibility, Gambling Literacy, Honesty
and Control, and Pre-commitment. However, the PPS development
research had limitations, including the use of an exploratory statistical
approach, treating the non-normally distributed PPS item-level data
as continuous, and the use of samples that consisted mainly of older
players living in British Columbia, Canada. Herein, we replicated and
extended the four-factor structure of the PPS using exploratory structural equation modelling with PPS item-data modelled as ordered
categorical in a large and demographically diverse sample of players
from across Canada (N = 5751). Once again, the four-factor structure
of the PPS provided an excellent fit to the data. PPS factors were all
internally consistent. Results also replicated and extended findings
from prior research. Specifically, all PPS subscales were negatively
correlated with measures of disordered gambling beliefs and behaviours, risk factors (e.g. impulsivity) and gambling motives (e.g. financial). Findings indicate that the PPS is a reliable and valid tool that
assesses responsible gambling beliefs and behaviours, which can be
used to inform responsible gambling initiatives.

Received 31 July 2019
Accepted 8 February 2020
KEYWORDS

Responsible gambling;
corporate social
responsibility; disordered
gambling; positive play;
structural equation
modelling

In the field of gambling studies, numerous self-report scales have been developed that assess
disordered beliefs and behaviours about gambling (e.g. Ferris & Wynne, 2001; Steenbergh,
Meyers, May, & Whelan, 2002; for a review of various measures, see Caler, Vargas Garcia, &
Nower, 2016). Although such scales have been shown to be a valid and reliable way to
understand disordered gambling (e.g. Currie, Hodgins, & Casey, 2013; Orford, Wardle,
Griffiths, Sproston, & Erens, 2010; Philander, Gainsbury, & Grattan, 2019), people with
past-year disordered gambling beliefs and behaviours only comprise between 0.1% to 5.8%
of the adult population (for a review of worldwide prevalence rates, see Calado & Griffiths,
2016). The focus on disordered gamblers, however, ignores the majority of people who play
largely without problems, but who may or may not be gambling responsibly. Since distorted
beliefs about gambling and excessive gambling behaviour are implicated in the aetiology of
CONTACT Nassim Tabri
nassim.tabri@carleton.ca
By Drive, Ottawa, Ontario K1S 5B6, Canada
© 2020 Informa UK Limited, trading as Taylor & Francis Group

Department of Psychology, Carleton University, 1125 Colonel

gambling disorders (Blaszczynski & Nower, 2002), understanding those characteristics may
be important to the prevention of harms. The recently developed Positive Play Scale (PPS;
Wood, Wohl, Tabri, & Philander, 2017) was the first measure to assess responsible gambling
beliefs and behaviours, and although good psychometric properties were reported, validation across a larger and more diverse population base is needed to understand the scale’s
generalizability. To this end, we assessed the validity and reliability of the PPS with a sample
of 5,751 players from across Canada.

The development and psychometric properties of the PPS
In consultation with players and experts, Wood et al. (2017; see Study 1) created an array
of items using the following definition of responsible gambling:
Responsible gambling means only spending what is affordable to lose and sticking to
personally allocated spend and time limits (formal or informal). Responsible play includes
honesty and openness with self and others about personal gambling habits. Belief in luck or
other superstitions may be present, but they do not have a significant negative impact on
play. There is recognition that gambling will always involve some degree of chance (p. 3).

The generated PPS items were then tested and refined across two studies in which four
subscales were identified. The Personal Responsibility subscale assesses the extent to which
players accept that they hold the ultimate responsibility for the amount of money and time
they spend gambling. The Gambling Literacy subscale assesses the extent to which players
hold an accurate understanding about their (low) odds of winning. The Honesty and
Control subscale assesses the extent to which players are open and truthful with others
about the amount of money and time they spend gambling and are in control of their
gambling behaviour. Lastly, the Pre-commitment subscale assesses to the extent players
consider how much money and time they should spend gambling.
Initial testing of the PPS showed it to have good psychometric properties. The
composed items of each subscale held together well and had good to excellent internal
consistency. Additionally, in terms of test-retest reliability over a one-month period, the
PPS subscales had moderate and positive autocorrelations, which is consistent with the
notion that PPS beliefs and behaviours are malleable. The subscales were also shown to
have convergent validity, as evidenced by small to moderate negative correlations
between the PPS subscales on the one hand and known risk factors for disordered
gambling (e.g. trait impulsivity and neuroticism) on the other hand. In contrast, the
PPS subscales had small to moderate positive correlations with conscientiousness, financial satisfaction, and a general sense to of self-efficacy to face difficulty. Together, these
initial results suggest that the PPS is a psychometrically sound self-report measure that is
both reliable and valid.
However, there were some statistical and methodological issues that limit the generalizability of the original findings. In the PPS development paper (Wood et al., 2017), the
construct validity of the PPS subscales was assessed using principal component analysis (PCA)
that were conducted for the items that assess responsible beliefs (personal responsibility and
gambling
literacy)
and
behaviours
(honesty
and
control,
and
precommitment), respectively. Within this testing strategy it is unclear whether the four-factor
structure of the PPS would replicate when the PPS beliefs and behaviour items are analysed

N. TABRI ET AL.

simultaneously. Furthermore, PCAs are exploratory and validation would be improved
through a confirmatory statistical approach, such as confirmatory factor analysis (CFA; see
Kline, 2016). Alternative structures of the PPS items were not tested and ruled-out relative to
the four-factor structure of the PPS identified via the PCAs.
Moreover, participants’ responses to the PPS items were non-normally distributed.
That is, responses were characterized by negative skew and kurtosis because most
participants agreed with each PPS item. The developers of the PPS treated the nonnormally distributed PPS item data as continuous in all analyses, which is potentially
problematic because doing so may lead to biased estimates for factor loadings and
internal consistency.
Perhaps most importantly, participants were residents of a single province in Canada
(British Columbia). As well, many participants were between 55 and 65 or over 65 years
old – an age group that has been shown to have less severe gambling problems relative to
younger players (Calado & Griffiths, 2016). Thus, it is unclear whether the four-factor
structure of the PPS identified via PCAs would replicate in a more demographically and
geographically diverse sample of players. Indeed, as noted in the limitations of the PPS
development paper, application of the PPS in a more generalizable sample of players
would improve confidence in the scale’s external validity.

The need for replication and extension: overview of the current study
The purpose of the current study was to replicate and extend the psychometric properties
of the PPS using statistical and methodological approaches that overcome the limitations
of prior PPS psychometric research. To do so, we used a large Canadian sample of players
from each gambling jurisdiction (i.e. province). We treated participants’ responses to the
PPS items as ordered categorical and tested the four-factor structure of the PPS by
analysing all PPS items (beliefs and behaviours) simultaneously.
Furthermore, we used exploratory structural equation modelling (ESEM; see Marsh,
Morin, Parker, & Kaur, 2014) to test the four-factor structure of the PPS. ESEM is
a combination of the best aspects of exploratory factor analysis (EFA) and confirmatory
factor analysis (CFA). In ESEM, questionnaire items are allowed to load on more than
one latent factor (as in EFA) and models with different numbers of latent factors can be
estimated and compared (as in CFA). In the analyses, we conducted an a priori confirmatory test of the four-factor PPS model. We also tested three alternative models
(single-factor, two-factor, and three-factor models).
To examine internal consistency of the PPS subscales, we used McDonald’s coefficient
ω (McDonald, 1970). Unlike coefficient α (see Cronbach, 1951), coefficient ω does not
assume that the factor loadings of items on a given scale are identical. Instead, coefficient
ω takes into account the varying factor loadings of the items. Coefficient ω is computed as
a ratio of true score variance among the items of a subscale (i.e. sum of the standardized
factor loadings squared) divided by the total variance of the items (i.e. sum of the
standardized factor loadings squared plus error variance). Simulation research has
showed that coefficient ω is more accurate relative to Cronbach’s α when factor loadings
on a given factor differ and when there are moderate to strong correlations between
factors (Şimşek & Noyan, 2013).

Lastly, we examined the convergent validity of the PPS subscales through associations
with constructs that have been linked to gambling in prior research. These constructs were
disordered gambling severity, erroneous gambling beliefs, financially focused self-concept,
and impulsivity. As in Wood et al. (2017), we expected that PPS scores would be negatively
associated with disordered gambling severity, erroneous gambling beliefs, financially
focused self-concept, and impulsivity. We also extended our analysis to examine how PPS
scores may relate to gambling motives (i.e. enhancement, coping, financial, and social).
Because these gambling motives have been shown to be positively associated with disordered gambling severity (e.g. Dechant, 2014; Schellenberg, McGrath, & Dechant, 2016;
Tabri, Wohl, Eddy, & Thomas, 2017), we expected that the PPS subscales would be
negatively associated with enhancement, coping, financial, and social motives for gambling.

Method
Participants
In 2017, a third-party survey company (Research Now) recruited 7,980 Canadians who
gamble from their established online panel of 500,000 people. Research Now was responsible for identifying and contacting participants and remunerating participants in the form
of points redeemable for gift cards. A total of 15,991 surveys were started and 7,980 were
completed, which corresponds to a completion rate of 49.9%. Approximately 1,000 participants were recruited from most regions in Canada: British Columbia, Alberta,
Saskatchewan, Manitoba, Ontario, Quebec, a conglomerate of Atlantic Provinces that
share some gambling-related operations (i.e. New Brunswick, Prince Edwards Island, and
Newfoundland and Labrador), and Nova Scotia. Within each of these regions, quota
sampling was used to recruit an equal number of men and women of which 80% gambled
in the last month and all had gambled in the last year. This sampling strategy helped ensure
that we obtained a sample of players who recently gambled.
We excluded participants recruited from Quebec (n = 1,000) to reduce possible noise
created from including a translated version of the scale (participants completed the scale in
French). We also excluded participants who had missing data (n = 97)1 and who failed one
or more attention checks (n = 1,132; e.g. ‘please respond 1 to this question’). Participants
who fail attention checks have been shown to provide lower quality data compared to
participants who pass attention checks (e.g. Gummer, Roẞmann, & Silber, in press). The
final sample consisted of 5,751 participants (see Table 1 for demographic information).
Formal ethics approval for a secondary analysis of the data was granted by the first
author’s institution Human Research Ethics Committee.
Procedure and measures
After providing informed consent, participants completed a questionnaire battery that
included most of the questionnaires used by Wood et al. (2017) to examine the validity
of the PPS. The questionnaires were completed online and presented in a single order that
began with basic demographic questions (e.g. age, gender) followed by questions about the
type of games played and frequency of play. Questions about income, education level and
ethnicity were presented at the end of the survey. The main body of the survey contained

N. TABRI ET AL.

Table 1. Distributions of participants’ age, gender, ethnicity, education, and income.
Mean age in years
Gender
Male
Female
Other
Ethnic/racial background
Asian (e.g. Chinese, Japanese, Korean)
South Asian (e.g. East Indian, Pakistani, Punjabi, Sri Lankan)
South East Asian (e.g. Cambodian, Indonesian, Laotian)
Arab/West Asian (e.g. Armenian, Egyptian, Iranian, Lebanese, Moroccan)
Black (e.g. African, Haitian, Jamaican, Somali)
Latin American/Hispanic
Aboriginal
White/Euro-Caucasian
Other
Prefer not to answer
Education
Elementary school
High School
College
Vocational training
University (undergraduate)
University (graduate)
University (post graduate)
Prefer not to answer
Income
Under $25,000
$25,000 to under $40,000
$40,000 to under $60,000
$60,000 to under $80,000
$80,000 to under $100,000
$100,000 to under $150,000
$150,000 or more
Prefer not to answer

47.59 (range: 18–89)
48.3%
51.6%
0.1%
4.2%
1.1%
0.8%
0.3%
1.0%
0.5%
3.2%
85.6%
1.9%
1.3%
1.0%
25.8%
21.7%
14.8%
15.9%
13.0%
7.0%
0.7%
6.7%
12.1%
15.4%
16.4%
14.0%
17.4%
8.0%
10.0%

the PPS followed by the remaining behavioural and attitudinal scales of which the following
were examined in the current research:
Positive Play Scale (PPS; Wood et al., 2017)
The PPS consists of 14 items divided across four subscales: Personal responsibility,
gambling literacy honesty and control, and pre-commitment (see Table 2 for a list of
items). Items assessing personal responsibility and gambling literacy were anchored at 1
(strongly disagree) and 7 (strongly agree), whereas items assessing honesty and control
and pre-commitment were anchored at 1 (never) and 7 (always). Participants were asked
to respond in relation to the gambling in the last month.
Disordered gambling severity
The Problem Gambling Severity Index (PGSI; Ferris & Wynne, 2001) was used to assess
disordered gambling severity. The PGSI consists of nine questions (α = .79) that measure
the extent of problem gambling behaviours and the consequences of engaging in problem
gambling behaviours, over the last 12 months. Response options were anchored at 0
(never) and 3 (almost always). The responses were summed with higher scores indicating
greater problem gambling severity.

Table 2. Descriptive statistics of all PPS items.
Response scale proportions (%)
Item
I felt in control of my gambling behaviour
I was honest with my family and/or friends
about the amount of MONEY I spent
gambling
I was honest with my family and/or friends
about the amount of TIME I spent
gambling
I only gambled with MONEY that I could
afford to lose
I only spent TIME gambling that I could
afford to spend
I considered the amount of MONEY I was
willing to lose BEFORE I gambled
I considered the amount of TIME I was
willing to spend BEFORE I gambled
I should be able to walk away from
gambling at any time.
I should be aware of how much MONEY
I spend when I gamble.
It’s my responsibility to spend only money
that I can afford to lose.
I should only gamble when I have enough
money to cover all my bills first.
Gambling is not a good way to make
money.
My chances of winning get better after
I have lost [reversed]
If I gamble more often, it will help me to
win more than I lose [reversed]

Other properties

M(SD)
Skew Kurtosis
1.60 0.60 1.40 2.40 5.00 10.20 78.90 6.55(1.12) −3.16 10.57
1.80 1.30 1.90 2.70 5.20 8.30 78.80 6.48(1.24) −2.84
7.83
1.70 1.10 1.50 2.70 5.00

7.80 80.20 6.52(1.20) −2.99

8.92

1.70 0.80 2.00 3.30 5.60

9.60 77.00 6.47(1.21) −2.75

7.57

1.40 0.80 1.50 2.60 5.10

9.90 78.70 6.53(1.14) −3.03

9.52

1.50 0.90 1.50 2.80 5.70 11.10 76.60 6.50(1.15) −2.90

8.79

2.50 1.10 2.60 5.00 7.20 11.40 70.20 6.28(1.39) −2.23

4.52

0.50 0.30 0.80 1.80 3.80

8.60 84.20 6.71(0.83) −3.79

16.88

0.40 0.10 0.30 1.10 2.80

8.30 87.00 6.79(0.68) −4.59

26.92

0.30 0.10 0.20 1.00 2.10

7.30 89.0

6.82(0.63) −5.19

34.92

0.60 0.10 0.40 1.00 2.30

5.80 89.80 6.81(0.71) −5.16

91.67

2.10 1.10 2.30 5.90 7.40 11.60 69.60 6.28(1.36) −2.18

4.39

3.30 1.50 3.40 7.70 5.70 11.00 67.50 6.14(1.54) −1.90

2.76

2.90 1.00 2.60 6.30 6.20 11.40 69.60 6.24(1.44) −2.14

4.04

N = 5751.

Erroneous gambling beliefs
The Gambling Beliefs Questionnaire (GBQ; Steenbergh et al., 2002) consists of 21
items that are divided into two subscales. The first subscale consists of eight items
that measure control beliefs about gambling (e.g. ‘My knowledge and skill in
gambling contribute to the likelihood that I will make money’). The second subscale
consists of 13 items that measure beliefs in luck (e.g. ‘Where I get money to gamble
doesn’t matter because I will win and pay it back’). Response options were anchored
at 1 (strongly disagree) and 7 (strongly agree). Items for each subscale were averaged
such that they reflected greater beliefs in control over gambling outcomes (α = .87)
and greater beliefs in luck (α = .94).
Gambling motives
The Gambling Motives Questionnaire – Financial (GMQ-F; Dechant, 2014) consists of
four items (α = .78) that assess financial motives (e.g. ‘To win money’), four items
(α = .87) that assess coping motives (e.g. ‘To forget your worries’), four items (α = .91)
that assess enhancement motives (e.g. ‘Because it makes you feel good’), and four items
(α = .84) that assess social motives (e.g. ‘To be sociable’). Response options were
anchored at 1 (never or almost never) and 4 (almost always or always). Responses for
each motive were averaged and coded such that higher scores indicate greater endorsement of the motive.

N. TABRI ET AL.

Financially focused self-concept
The 4-item Financially Focused Self-concept scale (FFS; Tabri et al., 2017) was used to
measure the extent to which a player’s self-concept is focused on financial success. These
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 options were anchored at 0 (not at all) and 4 (extremely).
Responses were averaged such that higher scores reflect greater FFS (α = .83).
Impulsivity
The short form Barratt Impulsiveness Scale (BIS; Spinella, 2007), consists of the 5-items
that assesses the extent to which people engage in non-planning (e.g. ‘I say things without
thinking’), 5-items that assess motor impulsivity (e.g. ‘I act on impulse’), and 5-items that
assess attentional impulsivity (e.g. ‘I am restless at lectures or talks’). Response options
were anchored at 1 (rarely/never) and 4 (almost always/always). We averaged responses
across subscales such that higher scores reflect greater impulsivity (α = .77).
Data analytic approach
In the ESEM analyses, all PPS items were modelled as ordered categorical variables using
the robust diagonally weighted least squares (WLSMV) method of estimation. To
adjudicate model fit, the mean and variance adjusted likelihood ratio chi-square test of
model fit (χ2), comparative fit index (CFI), root mean square error of approximation
(RMSEA), and weighted root mean square residual (WRMR) were used. An excellent
model fit would be reflected by a statistically nonsignificant χ2 and a CFI close to 1 as well
as an RMSEA of .05 or less with zero in its 95% confidence interval, and a WRMR value
less than 1 (see Kline, 2016). We compared the fit of the hypothesized four-factor model
to the fit of the alternative models using a WLSMV chi-square difference test (Δχ2). In the
ESEM analyses, we also corrected the standard errors of all parameter estimates based on
how the data were nested within provinces. The ESEM analyses were conducted using
Mplus version 8 (Muthén & Muthén, 1998–2017).
In the ESEM analyses, we also included two residual correlations to statistically control
for method variance due to similarity in item wording (see Podsakoff, MacKenzie, Lee, &
Podsakoff, 2003). The first residual correlation was between the item ‘I was honest with my
family and/or friends about the amount of MONEY I spent gambling’ and the item ‘I was
honest with my family and/or friends about the amount of TIME I spent gambling.’
The second residual correlation was between the item ‘I considered the amount of
MONEY I was willing to lose BEFORE I gambled’ and the item ‘I considered the amount
of TIME I was willing to spend BEFORE I gambled.’ Both residual correlations were
included in all models.
Internal consistency of the PPS subscales was estimated using McDonald’s coefficient ω
based on the results of the ESEM analysis. As such, McDonald’s ω reflects the reliability of
the underlying latent continuous factor of a given subscale – not the reliability of the
observed items on a given subscale. We also report estimates of Cronbach’s α for each PPS
subscale because it is the most popular measure of internal consistency. However, note that
Cronbach’s α will be underestimated when items on a given scale do not have identical

factor loadings, when there are correlations between the error variances of items on a given
scale, or both (see McNeish, 2018). To examine convergent validity, the observed items of
each PPS subscale were averaged to form their respective subscales. We then computed
Pearson correlations between each PPS subscale and all other substantive variables.

Results
Descriptive analyses
Descriptive statistics for the PPS items are reported in Table 2. All PPS items were
negatively skewed and kurtotic (see Table 2) – most participants had a high score on each
PPS item. These findings confirm that treating participants’ PPS responses at the itemlevel as ordered categorical in the ESEM analyses is appropriate.
Structure of the PPS
The hypothesized four-factor ESEM of the PPS with two residual correlations provided
a marginal fit to the data (see Table 3). Inspection of the residual matrix indicated that
adding two residual correlations would improve model fit. The first residual was between
the item ‘I should be able to walk away from gambling at any time’ and the item ‘I should
be aware of how much MONEY I spend when I gamble.’ The second residual correlation
was between the item ‘I only gambled with MONEY that I could afford to lose’ and the
item ‘I considered the amount of MONEY I was willing to lose BEFORE I gambled.’ The
model with the two additional correlated residuals provided an excellent fit to the data
(see Table 3). The magnitude of the two residual correlations were positive and very small
(rs < .10), but including them in the model improved model fit, Δχ2(2) = 55.15, p < .0001.
A single factor model provided a poor fit to the data (see Table 3). Reducing the
number of factors from four to one worsened model fit, Δχ2(36) = 7607.89, p < .0001.
Likewise, a two-factor alternative model provided a poor fit to the data (see Table 3) and
reducing the number of factors from four to two worsened model fit, Δχ2(23) = 1175.05,
p < .0001. Similarly, a three-factor alternative model provided a poor fit to the data (see
Table 3) and reducing the number of factors from four to three worsened model fit, Δχ2
(11) = 314.20, p < .0001. Taken together, the hypothesized four-factor model was the
superior model and reflected the factor structure very well (see Table 3).
Table 3. Test of model fit and fit indices for hypothesized and alternative models of the PPS.
Model
Hypothesized model
Four factors
Four factors + two residual correlations
Alternative models
One factor
Two factors
Three factors

χ2 (df)

CFI

RMSEA
and 95% CI

WRMR

99.75(39)**
50.99 (37)

.99

.02 [.01, .02]
.01 [.00, .01]

.47
.28

8738.92 (73)**
1202.18 (60)**
295.41 (48)**

.86
.98

.14 [.14, .15]
.06 [.06, .06]
.03 [.03, .03]

8.65
2.43
.96

CFI = Comparative Fit Index; RMSEA = Root Mean Square Error of Approximation; CI = Confidence Interval;
WRMR = Weighted Root Mean Square Residual.
N = 5751
*p < .05; **p < .001.

N. TABRI ET AL.

Standardized factor loadings from the ESEM analysis of the four-factor model are
reported in Table 4. As expected, items measuring honesty and control, precommitment, personal responsibility, and gambling literacy loaded moderately-tostrongly on separate factors with few and small cross-loadings on other factors. The latent
factor for honesty and control was strongly correlated with the latent factor for precommitment (r = .82, z = 42.22, p < .001) and moderately correlated with the latent factors
for personal responsibility (r = .66, z = 25.66, p < .001) and gambling literacy (r = .32,
z = 14.63, p < .001). Likewise, the latent factor for pre-commitment was strongly correlated
with the latent factors for personal responsibility (r = .71, z = 38.40, p < .001) and gambling
literacy (r = .38, z = 24.09, p < .001). The latent factors for personal responsibility and
gambling literacy were moderately correlated (r = .55, z = 33.75, p < .001). Although the
associations between the PPS subscale factors were positive and ranged from moderate to
strong, the ESEM analyses indicated that the PPS factors are empirically distinct.

Internal consistency of the PPS subscales
Participants, on average, scored high on personal responsibility (M = 6.78, SD = 0.60),
gambling literacy (M = 6.22, SD = 1.22), honesty and control (M = 6.52, SD = 1.06), and
pre-commitment (M = 6.45, SD = 1.04). Internal consistency estimates for personal
responsibility (ω = .94, α = .86), gambling literacy (ω = .83, α = .68), honesty and control
(ω = .87, α = .87), and pre-commitment (ω = .90, α = .87) were good to excellent reliability.

Convergent validity of the PPS
As expected, higher scores on each of the PPS subscales were distinct from and negatively
correlated with scores on the PGSI as well as with scores on the GBQ luck and control
subscales and all the gambling motives (see Table 5). Furthermore, higher scores on the
Table 4. Standardized factor loadings for the PPS items from the exploratory structural equation
model with four factors.
Item
I felt in control of my gambling behaviour
I was honest with my family and/or friends about the amount of MONEY
I spent gambling
I was honest with my family and/or friends about the amount of TIME I spent
gambling
I only gambled with MONEY that I could afford to lose
I only spent TIME gambling that I could afford to spend
I considered the amount of MONEY I was willing to lose BEFORE I gambled
I considered the amount of TIME I was willing to spend BEFORE I gambled
I should be able to walk away from gambling at any time.
I should be aware of how much MONEY I spend when I gamble.
It’s my responsibility to spend only money that I can afford to lose.
I should only gamble when I have enough money to cover all my bills first.
Gambling is not a good way to make money.
My chances of winning get better after I have lost [reversed]
If I gamble more often, it will help me to win more than I lose [reversed]
N = 5751.
*p < .05; **p < .001.

Factor 1 Factor 2 Factor 3 Factor 4
0.80** 0.04
0.09**
0.04**
0.66** 0.20*
0.04*
0.06**
0.55**

0.31**

0.05*

0.07**

0.17*
0.70** 0.05
0.02
−0.06
1.00** 0.01
0.02*
0.24** 0.50** 0.17**
0.02
0.16*
0.66** 0.05* −0.02*
0.27** −0.02
0.66** 0.01
0.13** −0.04*
0.85** 0.01
−0.05
0.03
0.96** 0.01
−0.11**
0.06*
0.91** 0.03
−0.04
0.10*
0.26** 0.32**
0.04
−0.04**
0.01
0.85**
−0.02
0.01
−0.01
1.00**

Table 5. Pearson correlations between the PPS subscales and other measured constructs.

PGSI (total score)
Illusion of control
Belief in luck
Financial motive
Coping motive
Enhancement
motive
Social motive
FFS
Impulsivity

M(SD)
1.73(3.86)
21.52(10.49)
24.41(14.28)
2.35(0.81)
1.44(0.62)
2.41(0.87)

Personal
Responsibility
−0.436**
−0.241**
−0.385**
−0.108**
−0.320**
−0.136**

Gambling
Literacy
−0.337**
−0.500**
−0.624**
−0.233**
−0.386**
−0.226**

Honesty and
Control
−0.565**
−0.215**
−0.360**
−0.139**
−0.379**
−0.216**

Precommitment
−0.535**
−0.231**
−0.368**
−0.163**
−0.356**
−0.197**

1.74(0.69)
1.47(1.01)
1.93(0.46)

−0.188**
−0.197**
−0.247**

−0.295**
−0.272**
−0.216**

−0.177**
−0.242**
−0.281**

−0.189**
−0.285**
−0.342**

PGSI = Problem Gambling Severity Index; FFS = Financially Focused Self-Concept. Sample size for the correlations ranged
between 5,751 and 5,708 because of missing data on the non-PPS variables.
*p < .05; **p < .01.

PPS subscales were each associated with lower impulsivity and FFS (see Table 5). The
magnitude of these correlations ranged from small to moderate.

Discussion
The current research replicates and extends the robust psychometric properties of the
PPS using a more rigorous statistical approach and testing strategy with a large demographically and geographically diverse sample of players. Indeed, we confirmed that the
PPS consists of four empirically distinct and correlated subscales: Personal
Responsibility, Gambling Literacy, Honesty and Control, and Pre-Commitment. The
four-factor model provided an excellent fit to the data compared to three alternative and
more parsimonious models (i.e. single factor, two-factor, and three-factor models). Thus,
the current research provided a rigorous confirmation of the four-factor structure of
the PPS.
In terms of reliability, we showed that the PPS subscales each have good
to excellent internal consistency. We also found that participants, on average, had high
scores on all of the PPS subscales. That is, most participants reported having responsible
gambling beliefs and engaging in responsible gambling behaviours. These findings
replicate and extend the results observed in the PPS development paper (see Wood
et al., 2017) with a much larger demographically and geographically diverse sample of
players. Thus, the PPS has good external validity.
We also showed that the PPS subscales have good convergent validity. More specifically, players with higher scores on the PPS subscales had less severe disordered gambling
beliefs, tendencies, and symptoms. The magnitude of these associations were moderate,
which suggests that the PPS is not simply measuring the absence of disordered gambling.
In addition, players with higher scores on the PPS subscales were less impulsive and
financially focused. These results replicate prior PPS research (see Wood et al., 2017) in
a larger and more demographically diverse sample of players from across Canada.
Furthermore, as expected, players with higher scores on the PPS subscales were less
likely to gamble for financial gain, to cope with negative affect, to enhance positive
affect, and for social reasons – four gambling motives that have been shown to be
linked with disordered gambling in prior research (e.g. Dechant, 2014; Schellenberg

N. TABRI ET AL.

et al., 2016; Tabri et al., 2017). Importantly, the magnitude of associations between the
PPS subscales and the gambling motives were small-to-moderate. This suggests that
players with higher PPS scores are not gambling for reasons that may lead them to
develop gambling problems.
Limitations and implications
The PPS is a self-report measure and thus may not equate with how players gamble.
Therefore, an area of future research would be to assess the relation between how players
respond with the PPS and how they play using, for example, player-account data. In the
current research, we did not examine the temporal measurement invariance of the PPS and
so it is unclear whether the factor-structure and meaning of the factors transcend time. As
with other self-report measures, responses may also be biased by social desirability.
However, participants were anonymous and were unaware of how other participants
were responding. As such, the impact of social desirability bias is likely minimal.
Nevertheless, future PPS-related research should measure this potential bias in players
who score high on the PPS. Lastly, although the sample was demographically and geographically diverse, it was not a probability sample of Canadian gamblers. Future validation
work can consider representative and non-Canadian samples.
The PPS can also be used to inform and guide responsible gambling initiatives. For
example, if an operator finds that young adult players have low scores on the gambling
literacy (compared to older players), it may behoove them to develop a responsible
gambling awareness campaign that includes age-appropriate messaging about how to
play positively and responsibly (e.g. gambling for entertainment, winning at gambling is
not predictable). Follow-up PPS testing may later be utilized to see whether gambling
literacy scores improve following exposure to the campaign.
Conclusions
The current research provides support for the originally reported factor structure of the
PPS using a confirmatory statistical approach with a large demographically and geographically diverse sample of players. We also modelled the PPS data at the item-level in
these analyses to obtain appropriate estimates of reliability for each PPS subscale and
provide convergent validity for the measure. Together, the results indicate that the PPS is
a reliable and valid self-report measure of responsible gambling attitudes and behaviour
that has robust psychometric properties.

Note
1. Results remained virtually unchanged after including participants with missing data in the
analyses using Full Information Maximum Likelihood.

Compliance with ethical standards
Ethical review and approval was not required for this study as per the institutional and national
requirements. All participants were responding to a survey that was part of a prior agreement with

Research Now. Nevertheless, participants were fully informed about the nature of the study, provided
their written consent to take part and were free to withdraw at any point. All data was anonymized
and the study was carried out in accordance with the ethical standards as laid down in the 1964
Declaration of Helsinki and its later amendments. As well, the secondary analysis of the data was
reviewed and approved by Carleton University’s Research Ethics Board – B.

Conflict of interest
Funding sources
The research was led by GamRes Limited—a research consultancy service that implements and
evaluates responsible gambling strategies—and funded by the Canadian Responsible Gambling
Association (CRGA).

Competing interests
In the interest of transparency, all authors have directly or indirectly conducted previous consultancy
work for CRGA. Importantly, however, in conducting the current study, the authors were given full
consent to investigate, analyze, and report all findings that the authors perceived to be relevant for
understanding the Positive Play Scale—CRGA did not have a role in determining the aims and
outcomes of the current study. Therefore, we do not consider there to be a conflict of interest
between the authors of this manuscript and the funders of the research.

Constraints on publishing
All authors reported no constraints on publishing the submitted manuscript.

Disclosure statement
Dr. Nassim Tabri has received consulting fees from the gambling industry in Canada, New
Zealand, the US, and the UK via GamRes Limited—a research and consultancy service that
designs, implements, and evaluates responsible gambling strategies. He has also received research
funds from Gambling Exchange Ontario (GREO).
Over the last 20 years, Dr. Richard Wood has been funded to undertake responsible gambling
projects for more than 50 gaming companies, regulators and government organizations worldwide. These projects have included; research funds to develop responsible gambling tools, funds to
evaluate existing responsible gambling tools and consultancy fees to apply research evidence to
assist the gambling industry to develop responsible gambling initiatives and/or to assess the
possible risk of proposed gambling products for players who may be at an increased risk for
developing a gambling problem.
Since 2015, Dr. KahlilPhilander has received research funds from the Washington State Gaming
Commission, Manitoba Gambling Research Program, UNLV International Centre for Gaming
Regulation, U.S.-Japan Business Council, Wynn Resorts, Board of Regents of the Nevada System of
Higher Education. He has received consulting payments from British Columbia Lottery
Corporation, Responsible Gambling Council of Canada, the Commonwealth of The Bahamas,
West Virginia Lottery, Indiana Gaming Commission, Intralot, and iDevelopment and Economic
Association. He has received reimbursement for travel from the Northeast Asia Economic
Association Forum, National Council for Problem Gambling, International Association of
Gaming Advisors, National Centre for Responsible Gambling, North American State and
Provincial Lottery Association, Evergreen Council on Problem Gambling, Global Gaming Expo
Asia, and Alberta Gambling Research Institute.

N. TABRI ET AL.

Dr. Michael J. A. Wohl has received research funding from federal granting agencies in Canada
and Australia unconnected to his gambling research. In relation to his gambling research, he has
received research funds from provincial granting agencies in Canada. He has also received direct
and indirect research funds from the gambling industry in Canada, United States of America,
United Kingdom, Australia, and Sweden. Additionally, he has served as a consultant for the
gambling industry in Canada, United States of America, United Kingdom, and New Zealand. A
detailed list can be found on his curriculum vitae (http://carleton.ca/bettermentlabs/wp-content/
uploads/CV.pdf).

Notes on contributors
Dr. Nassim Tabri is an Assistant Professor in the Department of Psychology at Carleton
University. His research is focused on etiological and maintenance factors underlying addictive
and other health-compromising behaviors.
Dr. Richard T. A. Wood is the President and founder of Gamres Limited where he undertakes
responsible gambling research and consultancy services.
Kahlil Philander is an Assistant Professor in the School of Hospitality Business Management. He
also holds an appointment as an Honorary Lecturer in the School of Psychology at the University
of Sydney. His research interests are in the socio-economic impacts of gambling.
Dr. Michael J. A. Wohl (Professor of Psychology) examines factors that lead to excessive play (e.g.,
erroneous cognitions, craving) and means to increase responsible gambling (e.g., setting a limit on
play). Recently, attention has been paid to the promotion of positive play and the influence loyalty
program membership can have on gambling behavior. Ultimately, he is interested processes that
promote positive behavioral change.

ORCID
Nassim Tabri
http://orcid.org/0000-0002-7085-9350
http://orcid.org/0000-0002-0747-0772
Kahlil Philander

References
Blaszczynski, A., & Nower, L. (2002). A pathways model of problem and pathological gambling.
Addiction, 97, 487–499.
Calado, F., & Griffiths, M. D. (2016). Problem gambling worldwide: An update and systematic
review of empirical research (2000–2015). Journal of Behavioral Addictions, 5, 592–613.
Caler, K., & Vargas Garcia, J. R., & Nower, L. (2016). Assessing problem gambling: A review of
classic and specialized measures. Current Addiction Reports, 3, 437–444. doi:10.1007/s40429016-0118-7
Cronbach, L. J. (1951). Coefficient alpha and the internal structure of tests. Psycbometrika, 16,
297–334.
Currie, S. R., Hodgins, D. C., & Casey, D. M. (2013). Validity of the problem gambling severity
index interpretive categories. Journal of Gambling Studies, 29, 311–327.
Dechant, K. (2014). Show me the money: Incorporating financial motives into the gambling
motives questionnaire. Journal of Gambling Studies, 30, 949–965.
Ferris, J. A., & Wynne, H. J. (2001). The Canadian problem gambling index (pp. 1–59). Ottawa,
ON: Canadian Centre on Substance Abuse.
Gummer, T., Roẞmann, J., & Silber, H. (in press). Using instructed response items as attention
checks in web surveys: Properties and implementation. Sociological Methods & Research, 1–27.
doi:10.1177/0049124118769083

Kline, R. B. (2016). Principles and practice of structural equation modeling, 4th. New York, NY: The
Guilford Press.
Marsh, H. W., Morin, A. J., Parker, P. D., & Kaur, G. (2014). Exploratory structural equation
modeling: An integration of the best features of exploratory and confirmatory factor analysis.
Annual Review of Clinical Psychology, 10, 85–110.
McDonald, R. P. (1970). Theoretical foundations of principal factor analysis and alpha factor
analysis. British Journal of Mathematical and Statistical Psychology, 23, 1–21.
McNeish, D. (2018). Thanks coefficient alpha, we’ll take it from here. Psychological Methods, 23,
412–433.
Muthén, L. K., & Muthén, B. O. (1998–2017). Mplus user’s guide (8th ed.). Los Angeles, CA:
Muthén & Muthén.
Orford, J., Wardle, H., Griffiths, M., Sproston, K., & Erens, B. (2010). PGSI and DSM-IV in the
2007 British gambling prevalence survey: Reliability, item response, factor structure and
inter-scale agreement. International Gambling Studies, 10, 31–44.
Philander, K. S., Gainsbury, S. M., & Grattan, G. (2019). An assessment of the validity of the
gamblers belief questionnaire. Addictive Behaviors, 97, 104–110.
Podsakoff, P. M., MacKenzie, S. B., Lee, J.-Y., & Podsakoff, N. P. (2003). Common method biases
in behavioral research: A critical review of the literature and recommended remedies. Journal of
Applied Psychology, 88, 879–903.
Schellenberg, B. J., McGrath, D. S., & Dechant, K. (2016). The gambling motives questionnaire
financial: Factor structure, measurement invariance, and relationships with gambling
behaviour. International Gambling Studies, 16, 1–16.
Şimşek, G. G., & Noyan, F. (2013). McDonald’s ωt, Cronbach’s α, and Generalized θ for composite
reliability of common factors structures. Communications in Statistics- Simulation and
Computation, 42, 2008–2025.
Spinella, M. (2007). Normative data and a short form of the Barratt Impulsiveness Scale.
International Journal of Neuroscience, 117, 359–368.
Steenbergh, T. A., Meyers, A. W., May, R. K., & Whelan, J. P. (2002). Development and validation
of the gamblers’ beliefs questionnaire. Psychology of Addictive Behaviors, 16, 143–149.
Tabri, N., Wohl, M. J., 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, 30–50.
Wood, R. T., Wohl, M. J., Tabri, N., & Philander, K. (2017). Measuring responsible gambling
amongst players: Development of the Positive Play Scale. Frontiers in Psychology, 8, 227.