A typical problem gambler affects six others

Abstract

While the financial and psychological burden on problem gamblers can be severe, at least some of the ill effects are also passed on to family or other close social ties. The present study estimated the number of affected-others for the typical problem gambler. Australian members of an online panel with gambling problems (N = 3076) and panel members who indicated that they had been affected by someone else’s gambling (N = 2129) were asked to estimate the number of other people who were negatively affected by their gambling. Using robust statistics to analyse this data, the study found lower estimates made by problem gamblers (four affected people) compared to estimates made by affected others (six affected people, including the respondent). It was concluded that a point-estimate of six people affected is a more accurate figure since it does not suffer from self-presentation effects of problem gamblers. Low-risk and moderate-risk gamblers, unsurprisingly, affected far fewer other people (one and three, respectively). Both gamblers and affected-others most often identified close family members, including spouses and children, as the people impacted by others’ gambling problems. These results provide an approximate measure of the number of people affected, per problem gambler, to facilitate accurate accounting of the harms accruing from gambling problems.

Full text

Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=rigs20
International Gambling Studies
ISSN: 1445-9795 (Print) 1479-4276 (Online) Journal homepage: http://www.tandfonline.com/loi/rigs20
A typical problem gambler affects six others
Belinda C. Goodwin, Matthew Browne, Matthew Rockloff & Judy Rose
To cite this article: Belinda C. Goodwin, Matthew Browne, Matthew Rockloff & Judy Rose (2017)
A typical problem gambler affects six others, International Gambling Studies, 17:2, 276-289, DOI:
10.1080/14459795.2017.1331252
To link to this article: https://doi.org/10.1080/14459795.2017.1331252
Published online: 07 Jun 2017.
Submit your article to this journal
Article views: 270
View related articles
View Crossmark data

INTERNATIONAL GAMBLING STUDIES, 2017
VOL. 17, NO. 2, 276289
https://doi.org/10.1080/14459795.2017.1331252
A typical problem gambler aects six others
Belinda C. Goodwina, MatthewBrownea, MatthewRockloa and Judy Roseb
aSchool of Human, Health, and Social Sciences, Central Queensland University (CQU), Branyan, Australia;
bSchool of Social Science, University of Queensland, Brisbane, Australia
ABSTRACT
While the nancial and psychological burden on problem gamblers
can be severe, at least some of the ill eects are also passed on to family
or other close social ties. The present study estimated the number of
aected-others for the typical problem gambler. Australian members
of an online panel with gambling problems (N = 3076) and panel
members who indicated that they had been aected by someone
else’s gambling (N = 2129) were asked to estimate the number of
other people who were negatively aected by their gambling. Using
robust statistics to analyse this data, the study found lower estimates
made by problem gamblers (four aected people) compared to
estimates made by aected others (six aected people, including the
respondent). It was concluded that a point-estimate of six people
aected is a more accurate gure since it does not suer from self-
presentation eects of problem gamblers. Low-risk and moderate-risk
gamblers, unsurprisingly, aected far fewer other people (one and
three, respectively). Both gamblers and aected-others most often
identied close family members, including spouses and children, as
the people impacted by others’ gambling problems. These results
provide an approximate measure of the number of people aected,
per problem gambler, to facilitate accurate accounting of the harms
accruing from gambling problems.
Introduction
In evaluating the harm associated with problematic gambling behaviour, it is important
to consider how ‘aected others’ – including spouses, children, friends and associates of
the gambler – might be negatively impacted (Korn, Gibbins, & Azmier, 2003). Research
in various international settings has revealed the extent to which close friends and family
can be impacted by others’ gambling (Dowling, Rodda, Lubman, & Jackson, 2014; Ferland
et al., 2008; Hing, Tiyce, Holdsworth, & Nuske, 2013; Holdsworth, Nuske, Tiyce, & Hing,
2013; McComb, Lee, & Sprenkle, 2009; Orford, Templeton, Velleman, & Copello, 2005;
Salonen, Castrén, Alho, & Lahti, 2014; Wenzel, Øren, & Bakken, 2008). For example, a recent
cross-sectional Norwegian study found that concerned signicant others (CSOs) reported
elevated conict, nancial detriment and impaired mental and physical health as a result of
a partner’s gambling (Wenzel et al., 2008). Longitudinal evidence from a Swedish population
KEYWORDS
Gambling harms; gamblers;
affected others; PGSI;
problem gambling; family
ARTICLE HISTORY
Received 26 August 2016
Accepted9 May 2017
© 2017 Informa UK Limited, trading as Taylor & Francis Group
CONTACT Belinda C. Goodwin b.goodwin@cqu.edu.au

INTERNATIONAL GAMBLING STUDIES 277
suggests that CSOs of gamblers experience poor mental health, risky alcohol use, nancial
hardship and strained relationships, although causality could not be established. Similar
ndings are true for Australian and Canadian samples, where problem gamblers’ spouses
typically report decrements to nancial security, social activity, emotional and physical
health, and family interaction (Ferland et al., 2008; Hing et al., 2013; Holdsworth et al., 2013).
ese issues are accompanied by increased problems at work, personal debt and marital
problems, as well as drug and alcohol use (Ferland et al., 2008; Hing et al., 2013; Holdsworth
et al., 2013). Negative physical and mental health outcomes are also reported for children of
gamblers, who may experience neglect due to diminished parental care or lack of resources
(Darbyshire, Oster, & Carrig, 2001; Shaw, Forbush, Schlinder, Rosenman, & Black, 2007).
Furthermore, nancial insecurity associated with ongoing gambling by parents can aect
more than one generation (Darbyshire et al., 2001), and can also aect extended relatives
and friends, and extend to the wider community (Clarke, Abbott, DeSouza, & Bellringer,
2007; Hing et al., 2013; Kalischuk, Nowatzki, Cardwell, Klein, & Solowoniuk, 2006).
ese ndings, largely derived from qualitative research, describe the experience of
gambling-related harm from the perspective of aected others, and highlight the way in
which harm is not limited only to those in close proximity to the gambler, but also aects
extended familial, social and community networks. Given the impact of problem gam-
bling on others, it is not surprising that eorts to quantify the social cost of gambling have
attempted to include this aspect of harm (Centre for Social & Health Outcomes Research
& Evaluation, 2008; Productivity Commission, 2010). For these calculations to be correct,
however, it is critically important to employ a reasonable estimate of the mean number of
aected others for every problem gambler. It is also helpful to know what demographic
groups are most aected by the gambling of others.
An Australian Productivity Commission report (1999) suggested that each gambler in
Australia aects between 5 and 10 other individuals. is gure has been cited oen, both
in the literature (Banks, 2007; Hinchlie, 2008; Leung, Wong, Lau, & Yeung, 2010) and in
non-academic communications (e.g. Responsible Gambling Fund Trustees, 2007); however,
no empirical evidence has been oered in support. A similarly non-precise estimate of ‘at
least 10 people’ has been attributed to Ladouceur (1993, as cited in Ferland et al., 2008).
Several population-based studies have estimated the incidence of others aected by gam-
bling. For example, Scandinavian studies, using a range of dierent measures, suggested
that between 2% and 19% of the population are gambling-related CSOs (Wenzel et al., 2008;
Salonen et al., 2014; Svensson, Romild, & Shepherdson, 2013). Abbott, Bellringer, Garrett,
and Mundy-McPherson (2014) estimated that 8% of New Zealanders were aected by the
gambling of someone close to them. Given that the prevalence of problem gambling in New
Zealand is estimated to be less than 1% (Devlin & Walton, 2012), their gure might be taken
to imply approximately eight CSOs per problem gambler, but this would ignore those who
are aected by individuals in Problem Gambling Severity Index (PGSI; Ferris & Wynne,
2001) categories other than ‘high risk/problem-gambling’. e varying estimates between
countries may be due to dierences in each survey’s wording and the terminologies used,
rather than an indication of international heterogeneity. For example, the term ‘concerned
signicant other’ infers a close family member (most likely a partner) who is showing con-
cern whereas the term ‘aected other’ does not imply relationship status or level of concern.

278 B. C. GOODWIN ET AL.
Study aims
Determining an estimate of the number of people aected by a typical problem gambler is
of signicant practical importance, particularly in understanding the aggregate harm caused
by problem gambling. Previous research has provided estimates of the proportion of people
aected by gambling (most commonly CSOs); however, to date no studies have provided a
precise estimate of the amount of people aected per gambler. is article presents ndings
from direct questioning of gamblers and aected others in terms of how many people are
inuenced by problem gambling harms, with the aim of providing a precise estimate of
the number and type of people aected per each low-, moderate- and high-risk gambler
according to the respective PGSI categories.
Limited research has shown that relationship status and type of gambling activity might be
associated with more negative impact from another’s gambling (Dowling, Rodda, Lubman,
& Jackson, 2014), but the likelihood of being an aected other based on these characteristics
has not been directly assessed. e current study also aims to identify the most aected
people (e.g. spouses, children, co-workers, etc.) and the types of gambling activities that
are impacting on the greatest number of aected others. is is important knowledge for
targeting intervention eorts for aected others towards the most vulnerable groups and
most risky products.
Method
Participants and procedure
We analysed data gathered as part of a large-scale survey of gambling-related harm to
gamblers and aected others (Browne et al., 2016). Research participants were invited
via email by a commercial online panel provider in Australia. Participants were compen-
sated for their time by points that could be accumulated and exchanged with the agency
for cash. e sample comprised 5205 participants (45.3% male) ranging from 18 to 89
years of age (M = 46.96, SD = 15.18) who reported either (a) that their own gambling
had caused them some degree of problems at some point in their lives (n = 3076), or (b)
having had a close relationship with a person whose gambling had caused them prob-
lems at some point in their lives (n = 2129). Participants were recruited in two phases. In
the rst phase, participants were rst asked ‘Has there been a time when your gambling
has caused problems in your life, no matter how minor?’ ose who answered ‘yes’ were
directed to a survey for gamblers. If they answered ‘no’, participants were asked ‘Have you
had a close relationship with a person whose gambling has caused problems in your life, no
matter how minor?’ In the second phase of recruitment the question order was reversed
and participants were preferentially directed to the aected others survey (see Appendix A
for a visual representation of the process). All participants were Australian residents, with
the majority (80.5%) born in Australia. Age, gender, income, education level and country
of birth are described separately for each group in Appendix B. Ethical approval for the
research was granted by the university’s internal review board and participants provided
informed consent before participating.

INTERNATIONAL GAMBLING STUDIES 279
Measures
At the outset of the survey, participants were directed to consider a 12-month period in
their life when gambling had been causing them the most problems. Questions were then
phrased retrospectively with respect to this period. irty-three percent of participants
reported this period to be in the most recent 12 months. As part of the larger research pro-
ject, participants also completed a checklist of gambling-related harms that had occurred
to them during this time (see Appendix C). Age, gender and relationship with gambler/
aected other was recorded for all participants as well as the preferred form of gambling
of the gambler.
Amount of aected others
Gamblers were asked to consider the checklist of gambling-related harms (Appendix C)
and report the number of people whom they believed had been aected by their gambling.
Aected others were asked to report the number of people whom they believed had been
aected by one gambler.1 Item wording was as follows: ‘Considering all the issues raised
earlier, how many other people would you estimate were aected by your gambling during
this period of time?’ (gambler) and ‘Considering all the issues raised earlier, how many other
people would you estimate were aected by this person’s gambling during this period of time?’
(aected other).
Problem gambling status
Gamblers responded to nine items on the Problem Gambling Severity Index (PGSI)
designed to measure problem gambling in the general population (Ferris & Wynne, 2001).
All items began with ‘At this time’ (this replaced the typical text ‘In the last 12 months’
for some items) to reect retrospective responding. For aected others, the PGSI was also
modied for second-person responding; that is, to describe the problem gambling status
of a gambler, rather than themselves (e.g. ‘At this time, did you feel that the person bet more
than they could really aord to lose?’). A detailed evaluation of the psychometric validity of
these modications has been reported elsewhere (Browne et al., 2016). In brief, as described
in Browne et al. (2016), the PGSI was shown to have measurement invariance for recent
retrospective reporting, and for reporting by self and aected others. One exception was
that self-reporters tended to provide lower mean scores than aected others – presumably
due to a tendency to minimize the negative impact of their behaviour. Cronbach’s alpha
values were α==.90 and α= =.78 for the rst and second-hand reported PGSI measures
respectively. Based on summed PGSI score, participants and gamblers nominated by aected
others were categorized as ‘low-risk’ (PGSI = 1–2), ‘moderate-risk’ (PGSI = 3–7) or ‘prob-
lem-gambler’ (PGSI>7) according to Ferris & Wynne (2001).2
Analysis
Descriptive analyses were conducted to examine the characteristics of those aected by
others’ gambling. In designing the main analysis, we assumed that variability in the number
of aected others reported could be aected by several factors:
(1) PGSI category of the gambler. More severe gambling problems should be related
to an increased number of aected others.

280 B. C. GOODWIN ET AL.
(2) First-hand versus second-hand reporting. We expected that self-reporters may be
more likely to minimize the impact of their gambling on others due to self-serving/
presentation bias (Greenberg, Pyszczynski, & Solomon, 1981/1982). Additionally,
aected others are a censored sample, in that at least one person (the respondent)
must have been aected in order for them to be eligible to complete the survey.
(3) Natural variation within PGSI category, including familial structure, and size of
social network.
(4) Individual dierences in response frame. Respondents would be expected to vary
with regard to an implicit threshold of what it means to be ‘aected’.
Because the response, a count of aected others, is bounded at zero, these sources of
variation have the potential to create an upward bias in the calculation of a simple mean.
erefore, our estimates and uncertainty estimates are based on bootstrapped trimmed
means using a stringent threshold for excluding extreme cases, with 25% of extreme values
excluded. Analyses were conducted using standard function in the R statistical program-
ming environment (R Core Team, 2014).
Results
e (25% trimmed) mean number of aected others was compared between PGSI status
for both the self-report and reports of the aected others. As shown in Figure 1, problem
gamblers reported the highest number of total aected others (self-report M =3.65; reported
by others M = 5.88), followed by moderate risk (self-report M = 0.73; reported by others M =
3.20) and low risk gamblers (self-report M = 0.03; reported by others M = 1.51). us, there
Figure 1. 25% trimmed means and bootstrapped 95% confidence intervals by gambler status and
reporting group.

INTERNATIONAL GAMBLING STUDIES 281
was a consistent discrepancy between reporting by aected others and gamblers, whereby
aected others estimated a greater number of people aected by the gamblers’ behaviour.
Table 1 details the relationship status and preferred product of gamblers who aected
others. e table also shows the percentage of gamblers who reported aecting at least one
spouse (or partner), close friend, parent, sibling, child, other family member or colleague/
co-worker. As shown in Table 2, almost half of the aected others’ sample were aected by
someone who primarily played electronic gaming machines (EGMs) (47.5%), followed by
race bettors (23.5%). Aected others were most oen spouses (38.0%), children (19.2%)
and close friends (14.8%). In support of these direct results from aected others, over half
(51.7%) of gamblers who reported aecting at least one other person indicated they had
aected a spouse or a partner. Children (19.2%), close friends (18.9%) and parents (19.2%)
were also commonly reported to have been aected by the gamblers in their self-reports.
We analysed the potential impact of demographic and gambling status variables on the
number of aected others reported. Given that the response is a non-negative integer count,
and subject to overdispersion, we applied negative-binomial count multiple regression with
a log link. While this model specication is well suited to handle to a reasonable degree
of overdispersion, very large outliers may still create problems with estimation. erefore
we tested the model using outlier rejection thresholds, with counts of greater than 10, 20
and 30 being excluded. Estimated beta coecients appeared stable across these dierent
thresholds. Table 3 presents beta coecients and standard errors for the model for cases
reporting 30 or fewer aected others (N = 4520). e number of aected others reported
Table 1.Relationship statuses reported by gamblers affected others.
^Gamblers could select more than one relationship that is affected.
N (%)
Gambler^(n = 2076) Aected other(n = 2069)
The person is my …
Spouse, de facto or romantic partner 1589 (51.7) 809 (38.0)
Son/daughter 590 (19.2) 408 (19.2)
Close friend 580 (18.9) 315 (14.8)
Other family 387 (12.6) 264 (12.4)
Sibling 289 (9.4) 142 (6.7)
Colleague/co-worker 289 (9.4) 73 (3.4)
Parent 573 (18.6) 58 (2.7)
Other N/A 60 (2.8)
Table 2.Preferred product of gamblers reported by gamblers and affected others.
^Only gamblers who reported affecting at least one other person included.
N (%)
Gambler^(n = 2120) Aected other(n = 2129)
The gambler’s preferred product is …
Electronic gaming machines (EGM) 1094 (51.6) 1014 (47.6)
Race betting 325 (15.3) 501 (23.5)
Casino table games 141 (6.7) 146 (6.9)
Sports betting 236 (11.1) 134 (6.3)
Poker 125 (5.9) 123 (5.8)
Lottery 172 (8.1) 63 (3.0)
Keno 27 (1.3) 19 (0.9)
Don’t know N/A 129 (6.1)

282 B. C. GOODWIN ET AL.
signicantly increased by problem gambler category, and when the nominating party was
an aected other (rather than a gambler). Controlling for gambling risk-status, sports and
racing gamblers aected signicantly more people than EGM players, while keno and lotto
players aected fewer people than EGM players. When controlling for gambling character-
istics, gamblers who were older and female tended to aect fewer others. However, when
the reporting was done by an aected other, those respondents who were older and female
tended to report more people were aected.
Discussion
e current study aimed to estimate the typical number of people aected by a problem
gambler, and to identify the most aected people and the types of gambling activities that
have the most impact. Although gures on the typical number are oen quoted in the lit-
erature, to our knowledge this is the rst study to directly investigate this point estimate. A
key feature of the study is that we surveyed both gamblers and aected others.
Best estimates for the number of aected others
We found that a typical problem gambler reported aecting about four others, whereas
those who were aected by a problem gambler on average estimated this gure to be six –
including themselves. As mentioned above, this discrepancy is probably primarily due to
(a) under-reporting by gamblers due to the self-presentation bias that is more common in
self-report data compared to data reported by others (Nederhof, 1985), and (b) censoring of
the sample of aected others. Again, censoring bias occurs because the survey response of
Table 3.Negative binomial regression predicting number of affected others reported.
*p<0.05; **p<0.01; ***p<0.005.
Dependent variable:
Count # Aected
r(SE)
Age gambler −0.010*** (0.001)
Affected other reporting 0.222 (0.129)
Female gambler −0.117*** (0.038)
PGSI low risk −0.565*** (0.125)
PGSI moderate risk 0.384*** (0.084)
PGSI problem gambler 1.035*** (0.079)
Preferred gambling activity (EGM vs.)
Sports 0.105* (0.052)
Race 0.114*** (0.036)
Poker 0.081 (0.063)
Casino −0.050 (0.056)
Keno −0.269* (0.138)
Lotto −0.176*** (0.059)
Other −0.004 (0.077)
Age (Affected other) 0.006*** (0.002)
Female (Affected other) 0.116* (0.056)
Intercept 0.704*** (0.119)
Observations 4520
Log likelihood −10,114.200
theta 2.219*** (0.086)
Akaike Inf. Crit. 20,260.410

INTERNATIONAL GAMBLING STUDIES 283
the ‘aected other’ necessarily includes the respondent themselves as ‘one’ of those aected.
With respect to censoring, this bias is likely to be signicant (e.g. between 0.5 and 1.0) in
the case of the low-risk category, where it is plausible that a large proportion of associated
gamblers truly do not aect even one person. However, for problem gamblers, it is likely
that only a negligible proportion of gamblers truly do not aect anybody, and therefore the
censorship bias is likely to be slight. With respect to under-reporting by gamblers, there are
multiple psychological explanations for the minimization of self-reported impact on others,
such as common tendency to present oneself in a positive light (Greenberg et al., 1981/1982),
attribute negative outcomes to external forces (Rotter, 1966), or positive memory biases
(Walker, Skowronski, & ompson, 2003). erefore, our interpretation is that the gure of
six aected others per problem gambler is the most valid since it is least aected by under-re-
porting. is is within, but at the lower end of, the range of gures commonly quoted
(Ferland et al., 2008; Productivity Commission, 1999). Taking into account censoring, and
therefore rounding-down our point-estimates, we conclude that a typical moderate-risk
gambler aects about three people, and a low-risk gambler aects one person. at is,
both of these latter gures are on the low side of our estimated ranges to account for the
attenuating eects of censoring. Our separate estimates for number of aected others per
gambler at each level of PGSI risk is useful as it can be weighted according to the specic
population prevalence statistics for low-, moderate- and high-risk (problem) gamblers to
produce accurate estimates of total aected others in the population, and potentially be
applied to international settings.
Who is most likely to be harmed by another person’s gambling?
In terms of demographic characteristics of aected others, current ndings were similar
to those from previous research. For example, Dowling et al. (2014) found that over 60%
of aected others seeking counselling were the spouse or partner of a problem gambler,
and almost 20% were the children of gamblers. is suggests people who live in the closest
proximity and are dependent nancially and emotionally on a problem gambler are most
likely to be aected by their behaviour.
What gambling games are most likely to harm aected others?
Dowling et al. (2014) also reported that over 40% of the concerned signicant others in
their sample were related to gamblers who were primarily EGM players. In the current
study almost half of the aected others’ sample fell into this category. is likely reects the
high proportion of gamblers who play EGMs. EGMs feature a combination of risky struc-
tural characteristics such as rapid playing speeds and payout intervals, multiplier potential,
reinforcing payout schedules, and attractive audiovisual eects. ese features are more
amenable to risky and problematic play than many other gambling products (Blaszczynski,
Walker, & Sharpe, 2001; Jackson, omas, omason, Holt, & McCormack, 2000; Smith
& Wynne, 2004); therefore we might also expect that EGM players are more likely than
other gamblers to export gambling-associated harms to others (Breen & Zimmerman, 2002;
Doughney, 2002).
Sports and race betting were associated with greater (gambler-reported) estimates for
aected others harmed, whereas lotto and keno were conversely associated with lower

284 B. C. GOODWIN ET AL.
estimates (see Table 3). Both sports and race-betting are dominated by male gamblers,
whereas lotto and keno attract proportionately more female gamblers. Given the relative
rarity of female sports and race bettors, it is dicult to determine whether gender or type
of game dominates in our analysis. Future research may illuminate whether the type of
preferred games or alternatively gambler-demographic factors are more inuential in deter-
mining the dispersion and severity of harm that impacts aected others. Nevertheless, the
nding that certain games, such as EGMs, sports and race-betting, are associated with a
greater number of others being harmed is important in estimating population-level harm.
ese games not only cause harm to problem gamblers, but also export harm to a greater
number of aected others, magnifying their eects on the whole community. Moreover,
although we expect most western countries are likely to have broadly similar numbers of
aected others to those estimated in this article, jurisdictions outside Australia may have
fewer or more aected others for every gambler based on dierent mixes of preferred
gambling products.
Limitations
Any consideration of a numeric gure depends heavily on the threshold one uses to dene
being ‘aected’. Our operational denition of ‘aected’ is the occurrence of at least one of
the items on the gambling harms checklist presented in Appendix C which represents a
variety of nancial, relationship, work- or study-related, emotional and health-related harms
experienced by gamblers (Browne et al., 2016). e downside of this approach is that we are
unable to estimate the degree to which individuals, other than the participants themselves,
are aected by the gambler.
In addition to the potential bias caused by gamblers under-reporting harms, it must
also be acknowledged that aected others’ reports may also be subject to similar issues. It
is dicult, however, to predict whether aected others are susceptible to under-reporting
harms due to lack of knowledge of the full extent of the gamble’s impact on others, or to
overestimating the proportion of harm due to negative emotions regarding the person with
the behaviour. Given that judging the quantity and extent of harm is intrinsically subjective,
there is not an obvious solution to this issue. Nevertheless, the dual perspectives presented
in the present study go some way to addressing uncertainty due to reporting bias.
Finally, recruitment was done through an online panel, which is not representative of
the general population of gamblers, and in which problem gamblers are over-represented.
However, our results are provided with respect to, or control for, gambler risk category.
Although an eort was made to allocate aected others and gamblers to their respective
surveys in an unbiased manner (see Appendix A), the smaller amount of participants taking
part in the second phase of recruitment meant that gamblers who also identied as aected
others may have been slightly under-represented overall.
Conclusion
Accurate understanding of how many aected others are impacted by gambling problems,
as well as a better understanding of who is aected, is helpful in eorts to reduce com-
munity harm. e costs of problem gambling are not limited to the immediate eects on
the nancial and emotional well-being of the problem gambler, but also extend to people

INTERNATIONAL GAMBLING STUDIES 285
intimately connected to the gambler through family and other social ties. ese connec-
tions must be considered to understand the wider costs of problem gambling, and provide
a foundational knowledge for the investment in appropriate interventions. is research
provides an important rst step in accounting for who and how many are aected by gam-
bling problems. Interventions in the UK and US have successfully assisted aected others
in dealing with the consequences of others’ problem gambling (GamCare, 2003; Winters,
Benston, & Stincheld, 1996). In the future, it may be possible to tailor such campaigns
to provide the most relevant assistance and advice to the people most at risk and to those
posing the most risk to others.
Conict of interest
Funding sources
is work was supported by the Victorian Responsible Gambling Foundation (VRGF)
under Grant No. VRGF 1–13.
Competing interests
e authors have no competing interests to declare. e VRGF had no involvement in the
research design, methodology, conduct, analysis or write-up.
Constraints on publishing
Some results and methodology from this manuscript form part of the research report written
for the VRGF which underwent their review. No changes were made as a result of this review.
Notes
1. In calculating total aected others, the respondent was included by adding 1 to each response.
2. Non-gamblers (PGSI = 0) were categorized as low risk as all study targets had experienced
and/or caused some form of gambling harm and very few participants recorded a PGSI score
of zero (n = 140).
Notes on contributors
Belinda C. Goodwin is a PhD candidate, casual lecturer and research worker for the Experimental
Gambling Research Laboratory (EGRL) at Central Queensland University (CQU). She has a strong
background in gambling and addictive behaviour research.
Matthew Browne is a senior lecturer and a senior researcher in the EGRL. He is an expert statistician
and has been chief or co-investigator on several large-scale, prominent gambling-related research
projects since 2009.
Matthew Rocklo leads the EGRL and serves as the head of the Population Research Laboratory
(PRL) at CQU. He has a long history of successfully managing large-scale, prominent gambling-re-
lated research projects.
Judy Rose has completed a Bachelor of Arts, a Masters in applied linguistics and a PhD in social
science. She is a key member of the ERGL and has extensive research experience.

286 B. C. GOODWIN ET AL.
ORCID
Matthew Browne http://orcid.org/0000-0002-2668-6229
Matthew Rocklo http://orcid.org/0000-0002-0080-2690
References
Abbott, M., Bellringer, M., Garrett, N., & Mundy-McPherson, S. (2014). New Zealand 2012 national
gambling study: Gambling harm and problem gambling (No. 2). Auckland: Auckland University of
Technology, Gambling and Addictions Research Centre.
Banks, G. (2007, August). Gambling in Australia: Are we balancing the equation. In Australian
Gambling Expo Conference, Sydney.
Blaszczynski, A., Walker, M., & Sharpe, L. (2001). e assessment of the impact of the reconguration
on electronic gaming machines as harm minimisation strategies for problem gambling. Sydney:
University of Sydney Gambling Research Unit.
Breen, R. B., & Zimmerman, M. (2002). Rapid onset of pathological gambling in machine gamblers.
Journal of Gambling Studies, 18, 31–43. doi:10.1023/A:1014580112648
Browne, M., Langham, E., Rawat, V., Greer, N., Li, E., Rose, J., & Best, T. (2016). Assessing gambling-
related harm in Victoria: A public health perspective. North Melbourne: Victorian Responsible
Gambling Foundation.
Centre for Social and Health Outcomes Research and Evaluation. (2008). Assessment of the social
impacts of gambling in New Zealand. Auckland: Ministry of Health.
Clarke, D., Abbott, M., DeSouza, R., & Bellringer, M. (2007). An overview of help seeking by problem
gamblers and their families including barriers to and relevance of services. International Journal
of Mental Health and Addiction, 5, 292–306. doi:10.1007/s11469-007-9063-y
Darbyshire, P., Oster, C., & Carrig, H. (2001). e experience of pervasive loss: Children and young
people living in a family where parental gambling is a problem. Journal of gambling studies / co-
sponsored by the National Council on Problem Gambling and Institute for the Study of Gambling
and Commercial Gaming, 17, 23–45. http://www.ncbi.nlm.nih.gov/pubmed/11705015
Devlin, M. E., & Walton, D. (2012). e prevalence of problem gambling in New Zealand as measured
by the PGSI: Adjusting prevalence estimates using meta-analysis. International Gambling Studies,
12, 177–197. doi:10.1080/14459795.2011.653384
Doughney, J. (2002). Socioeconomic banditry: Poker machines and income redistribution in Victoria.
Competing Visions, 137. Retrieved from https://www.sprc.unsw.edu.au/media-/SPRCFile/NSPC01_
RefereedProceedings.pdf#page=141
Dowling, N. A., Rodda, S. N., Lubman, D. I., & Jackson, A. C. (2014). e impacts of problem
gambling on concerned signicant others accessing web-based counselling. Addictive Behaviors,
39, 1253–1257. doi: 10.1016/j.addbeh.2014.04.011
Ferland, F., Fournier, P.-M., Ladouceur, R., Brochu, P., B ouchard, M., & Pâquet, L. (2008). Consequences
of pathological gambling on the gambler and his spouse. Journal of Gambling Issues, 219–229. doi:
10.4309/jgi.2008.22.5
Ferris, J. A., & Wynne, H. J. (2001). e Canadian problem gambling index: User manual. ON: Canadian
Centre on Substance Abuse Toronto.
GamCare. (2003). GamCare care services report. London: National Association for Gambling Care,
Education and Resources.
Greenberg, J., Pyszczynski, T., & Solomon, S. (1981/1982). e self-serving attributional bias: Beyond
self-presentation. Journal of Experimental Social Psychology, 18, 56–67.
Hinchlie, S. (2008). Gambling-an update on duty of care in Victoria. Macquarie Journal of Business,
5(87), 396–408.
Hing, N., Tiyce, M., Holdsworth, L., & Nuske, E. (2013). All in the family: Help-seeking by signicant
others of problem gamblers. International Journal of Mental Health and Addiction, 11, 396–408.
doi:10.1007/s11469-012-9423-0
Holdsworth, L., Nuske, E., Tiyce, M., & Hing, N. (2013). Impacts of gambling problems on partners:
Partners’ interpretations. Asian Journal of Gambling Issues and Public Health, 3(1), 1–14. Retrieved
from http://link.springer.com/article/10.1186/2195-3007-3-11

INTERNATIONAL GAMBLING STUDIES 287
Jackson, A. C., omas, S. A., omason, N., Holt, T. A., & McCormack, J. (2000). Analysis of clients
presenting to break even problem gambling services, July 1 1998 to June 20 1999. Melbourne, Victoria:
Victorian Department of Human Services.
Kalischuk, R. G., Nowatzki, N., Cardwell, K., Klein, K., & Solowoniuk, J. (2006). Problem gambling
and its impact on families: A literature review. International Gambling Studies, 6, 31–60.
doi:10.1080/14459790600644176
Korn, D., Gibbins, R., & Azmier, J. (2003). Framing public policy towards a public health paradigm
for gambling. Journal of Gambling Studies / co-sponsored by the National Council on Problem
Gambling and Institute for the Study of Gambling and Commercial Gaming, 19, 235–256. Retrieved
from http://www.ncbi.nlm.nih.gov/pubmed/12766471
Ladouceur, R. (1993). Jeu pathologique [Pathological gambling]. In R. Ladouceur, O. Fontaine, &
J. Cottraux (Eds.), érapie Cognitive et Comportementale (pp. 123–128). Paris: Masson.
Leung, K. C., Wong, I. L. K., Lau, K. M., & Yeung, S. C. (2010). Stress, health and coping resources
of Chinese pathological gamblers’ spouses. Asian Journal of Gambling Issues and Public Health, 1,
61–74. doi:10.1186/BF03342119
McComb, J. L., Lee, B. K., & Sprenkle, D. H. (2009). Conceptualizing and treating problem
gambling as a family issue. Journal of Marital and Family erapy, 35, 415–431. doi:10.1111
/j.1752-0606.2009.00146
Nederhof, A. J. (1985). Methods of coping with social desirability bias: A review. European Journal
of Social Psychology, 15, 263–280.
Orford, J., Templeton, L., Velleman, R., & Copello, A. (2005). Family members of relatives with
alcohol, drug and gambling problems: A set of standardized questionnaires for assessing stress,
coping and strain. Addiction, 100, 1611–1624.
Productivity Commission. (1999). Australia’s gambling industries. Report No. 10. Canberra: AusInfo.
Productivity Commission. (2010). Gambling (No. Report no. 50). Canberra.
R Core Team. (2014). R: A language and environment for statistical computing. R Foundation for
Statistical Computing, Vienna, Austria. Retrieved from http://www.R-project.org/
Responsible Gambling Fund Trustees. (2007). Review of the registered clubs industry in NSW. Retrieved
from http://www.ipart.nsw.gov.au/les/sharedassets/website/trim-holdingbay/submission_-_
registered_clubs_review__responsible_gambling_fund_trustees_-_rev_harry_herbert_-_apd_
web.pdf
Rotter, J. B. (1966). Generalized expectancies for internal versus external control of reinforcement.
Psychological Monographs: General & Applied., 80(1), 1–28. doi:10.1037/h0092976
Salonen, A. H., Castrén, S., Alho, H., & Lahti, T. (2014). Concerned signicant others of people
with gambling problems in Finland: A cross-sectional population study. BMC Public Health, 14,
398–407. doi:10.1186/1471-2458-14-398
Shaw, M. C., Forbush, K. T., Schlinder, J., Rosenman, E., & Black, D. W. (2007). e eect of
pathological gambling on families, marriages, and children. CNS Spectrums, 12, 615–622.
Smith, G. J., & Wynne, H. J. (2004). VLT gambling in Alberta: A preliminary analysis: nal report.
Alberta: Alberta Gaming Research Institute.
Svensson J., Romild U., & Shepherdson E. (2013). e concerned signicant others of people with
gambling problems in a national representative sample in Sweden–a 1year follow-up study. BMC
Public Health, 13, 1087–1098. doi:10.1186/1471-2458-13-1087
Walker, W. R., Skowronski J. S., ompson Charles P. (2003). “Life is pleasant – and memory helps
to keep it that way!” (PDF). Review of General Psychology, Educational Publishing Foundation. 7,
203–210. doi:10.1037/1089-2680.7.2.203
Wenzel, H. G., Øren, A., & Bakken, I. J. (2008). Gambling problems in the family–A stratied
probability sample study of prevalence and reported consequences. BMC Public Health, 8, 412–417.
doi:10.1186/1471-2458-8-412
Winters, K. C., Benston, P. L., & Stincheld, R. D. (1996). Finding from a follow-up study of callers to
the Minnesota problem gambling hotline. Menneapolis, MN: University of Minnesota.

288 B. C. GOODWIN ET AL.
Appendix A: Recruitment process
Appendix Figure 1. Two-phase participant recruitment process.
Phase 1: Phase 2:
N = 5597
participants
Has there been a time
when your gambling
has caused problems
in your life, no matter
how minor?
Yes No
Completed
survey for
gamblers
N = 2458
Have you had a close
relationship with a
person whose
gambling has caused
problems in your life,
no matter how
minor?
Yes No
Completed
survey for
affected
others
N = 1678
Discontinued
N = 1461
N = 1294
participants
Have you had a close
relationship with a person
whose gambling has caused
problems in your life, no
matter how minor?
Yes No
Completed
survey for
affected
others
N= 451
Has there been a time
when your gambling
has caused problems
in your life, no matter
how minor?
Yes No
Completed
survey for
gamblers
N=618
Discontinued
N=225
Appendix B: Analysis of group dierences
Appendix B Table 1 details the gender, age, country of birth, education and income dierences
between the participants in the gamblers group versus those in the aected others group. A small
signicant eect was found for gender (r = .25) with more males appearing in the gamblers group
(55.7%) than in the aected others group (30.2%). Some very weak associations were also found for
tertiary education and income level, with slightly higher income earners in the gamblers group (r =
-.05) and slightly more tertiary educated participants in the aected others group (r=-.04).
Appendix B Table 1. Breakdown of gender, age, country of birth, education and income according
to group.

INTERNATIONAL GAMBLING STUDIES 289
Group A Group B Pearson’s r
(Gamblers) (Aected others)
Male 55.7% 30.2% .25*
Age 45.05 (14.98) 44.83 (15.48) −.01
Australian born 79.9% 81.3% −.02
Tertiary educated 46.7% 51.0% .04*
Personal income level^ 6.30 (2.87) 5.98 (2.97) −.05*
* = p < .05. Categorical variables displayed as proportion of group. Means and SDs displayed for continuous variables. ^
Income level was recorded in even brackets Level 6 (the mean for both Group A & B) represents the ‘$31,200–$41,599 per
year’ bracket.
Appendix C: 2015 Victorian harms checklist
Participants were asked ‘Have you experienced any of these issues as a result of your gambling in the
last 12 months?’ and were presented with a checklist of harms that covered the following six domains.
For full list of items, see Browne et al. (2016).
1. FINANCIAL IMPACT FROM GAMBLING
e.g. Reduction of my savings, Less spending on recreational expenses such as eating out, going to the
movies or other entertainment.
2. RELATIONSHIP IMPACTS FROM GAMBLING
e.g. Spent less time with people I care about, Neglected my relationship responsibilities.
3. EMOTIONAL OR PSYCHOLOGICAL IMPACT OF GAMBLING
e.g. Felt distressed about my gambling, Felt insecure or vulnerable.
4. HEALTH IMPACTS FROM GAMBLING
e.g. Reduced physical activity due to my gambling, Neglected my hygiene and self-care.
5. WORK/STUDY IMPACTS FROM GAMBLING
e.g. Was late for work or study, Used my work or study time to gamble.
6. OTHER PROBLEMS FROM GAMBLING
e.g. Le children unsupervised, Felt compelled or forced to commit a crime or steal to fund gambling
or pay debts.