Abstract
Background, aims and design
The increase in mobile telephone‐only households may be a source of bias for traditional landline gambling prevalence surveys. Aims were to: (1) identify Australian gambling participation and problem gambling prevalence using a dual‐frame (50% landline and 50% mobile telephone) computer‐assisted telephone interviewing methodology; (2) explore the predictors of sample frame and telephone status; and (3) explore the degree to which sample frame and telephone status moderate the relationships between respondent characteristics and problem gambling.
Setting and participants
A total of 2000 adult respondents residing in Australia were interviewed from March to April 2013.
Measurements
Participation in multiple gambling activities and Problem Gambling Severity Index (PGSI).
Findings
Estimates were: gambling participation [63.9%, 95% confidence interval (CI) = 61.4–66.3], problem gambling (0.4%, 95% CI = 0.2–0.8), moderate‐risk gambling (1.9%, 95% CI = 1.3–2.6) and low‐risk gambling (3.0%, 95% CI = 2.2–4.0). Relative to the landline frame, the mobile frame was more likely to gamble on horse/greyhound races [odds ratio (OR) = 1.4], casino table games (OR = 5.0), sporting events (OR = 2.2), private games (OR = 1.9) and the internet (OR = 6.5); less likely to gamble on lotteries (OR = 0.6); and more likely to gamble on five or more activities (OR = 2.4), display problem gambling (OR = 6.4) and endorse PGSI items (OR = 2.4‐6.1). Only casino table gambling (OR = 2.9) and internet gambling (OR = 3.5) independently predicted mobile frame membership. Telephone status (landline frame versus mobile dual users and mobile‐only users) displayed similar findings. Finally, sample frame and/or telephone status moderated the relationship between gender, relationship status, health and problem gambling (OR = 2.9–7.6).
Conclusion
Given expected future increases in the mobile telephone‐only population, best practice in population gambling research should use dual frame sampling methodologies (at least 50% landline and 50% mobile telephone) for telephone interviewing.
Keywords: Cellphones, dual‐frame, gambling, mobile telephone, prevalence, problem gambling, sampling, surveys
Introduction
The standardized past‐year prevalence of problem gambling averages 2.3% across countries 1. In Australia, problem gambling estimates have plateaued, with the most recent state/territory gambling surveys 3, 4, 5, 6, 7, 8, 9 identifying rates of 61.2–73.8% for gambling participation, 0.5–0.8% for problem gambling, 1.5–2.9% for moderate risk gambling and 3.4–8.4% for low‐risk gambling using the Problem Gambling Severity Index (PGSI) 2. Designed to explore interactive gambling, the only national prevalence survey conducted in the last 15 years 10 revealed a gambling participation estimate of 64.3%. Using the PGSI on a subsample of 1768 respondents, the rate of problem gambling was 0.6%, moderate‐risk gambling was 3.7% and low‐risk gambling was 7.7%.
The available studies are limited by an over‐reliance on state/territory estimates, modifications to the PGSI response options 5, 6, 8, and the administration of the PGSI to purposive samples 4, 9, 10. Moreover, most of these computer‐assisted telephone interviewing (CATI) surveys have employed single sample frames of landline telephone numbers 7, 8, 10. An emerging issue facing telephone survey researchers across all fields of study, however, is the increasing proportion of the population residing in ‘mobile telephone‐only’ households. Current estimates suggest that 29% of Australian adults reside in these households 11. The non‐coverage of these households by traditional landline sampling methods is therefore a probable source of bias. Mobile telephone respondents are consistently more likely to be male, young, of indigenous status, Australian born, full‐time workers, high wage‐earners, employed, living in group households, living in metropolitan regions and residentially mobile than landline respondents 3, 6, 12, 13, 14, 15. They are also more likely to report psychosocial problems, such as anxiety/depression, psychological distress, alcohol use problems, binge drinking, illegal drug use, smoking, social isolation, financial stress or hardship, being a victim of crime and ethnic or religious discrimination 6, 12, 14, 16.
There is also now growing evidence that significant differences exist in gambling behaviour between landline and mobile sample frames. This issue was first explored in two Australian dual‐frame surveys (43–50% mobile frame) 12, 14. In these studies, landline frame respondents displayed significantly higher gambling participation and regular gambling rates than mobile telephone‐only respondents 12, but the mobile telephone and mobile telephone‐only samples both reported higher rates of problem gambling 12, 14 and endorsement of specific items on problem gambling measures 12 than the landline samples. Since the publication of this research, several Australian states (South Australia and Tasmania) have adopted dual‐frame methodologies in their most recent prevalence surveys (25–30% mobile frame) 3, 6. In both studies, the mobile frames displayed higher overall gambling participation and participation on specific gambling activities (horse or greyhound racing, lotteries, keno, casino table games and sporting events) than the landline frames 3, 6. There was, however, no difference in PGSI categories between sample frames 3.
Although these findings suggest that traditional landline and mobile telephone sampling approaches derive different gambling estimates, they provided relatively limited exploration of the impact of these approaches 3, 6, 12, 14, generally failed to explore the impact of telephone status (e.g. mobile dual users and mobile‐only users) 3, 6, 14 and did not explore the effect of sample frames on relationships between respondent characteristics and problem gambling 3, 6, 14. Several also included relatively small proportions of mobile samples 3, 6, only employed data from individual states/territories 3, 6, employed brief gambling measures [e.g. single items, the National Opinion Research Center at the University of Chicago (NORC) Diagnostic Screen for Gambling Disorders–Control, Lying and Preoccupation (NODS‐CLiP)] 12, 14 or were conducted when the proportion of mobile telephone households was much lower 6, 12, 14. Given the increasing adoption of dual‐frame methodologies by Australian state/territory governments 3, 6 and the increasing proportion of mobile telephone‐only households 11, there is a need for a more thorough exploration of the role of sample frame and telephone status in gambling estimates to determine whether dual‐frame methodologies, which are more expensive than landline telephone samples 12, 17, are necessary in future population gambling surveys.
The study aims are therefore to: (1) identify national gambling participation and problem gambling prevalence estimates using a dual‐frame methodology; (2) explore the univariate and multivariate predictors of sample frame and telephone status; and (3) explore the degree to which sample frame and telephone status moderate the relationships between respondent characteristics and problem gambling. Although it is expected that mobile telephone respondents will display a distinct demographic profile and a higher likelihood of gambling participation, gambling problems and psychosocial problems than landline telephone respondents, the moderation aim was exploratory.
Method
Study design
A dual‐frame sample design using CATI random digit dialling (RDD) aimed to obtain a nationally representative sample (Australia does not have a national telephone register). The 50% mobile frame, which is similar to that used by leading survey research organizations in the United States 18, was selected to minimize costs, provide a large enough sample to explore differences between telephone user groups and provide lower variance than more skewed blends.
Both the landline and mobile samples were generated by the commercial sample vendor SampleWorx on 22 March 2013. In Australia, the Australian Communications and Media Authority (ACMA) allocates blocks of landline and mobile telephone numbers to telecommunication carriers by way of allocating telephone number prefixes to each carrier. The numbers generated are provided to commercial sample vendors, who test them to identify working and non‐working numbers. For the landline component, sample members were selected from each household using a random allocation to the ‘next birthday method’ or an alternative quasi‐random method based on the number of eligible sample members in the household 19. No within‐unit selection routine was required for the mobile frame, as the mobile telephone was treated as an individual, rather than a household or shared, device.
The landline frame used probability proportional to size quotas for 15 geographic strata (all capital cities and non‐capital city regions of each state/territory, with the exception of the Australian Capital Territory which was treated as one stratum). The population distribution for these quotas ranged from 0.4% (rest of Northern Territory) to 20.6% (Sydney) 20. A simple random sample was used for the mobile frame as these numbers have no geographic identifiers. Both sample frames produced geographic distributions similar to that of the population, although the mobile sample displayed a slight under‐representation of areas outside capital cities, which is aligned with the slightly lower take‐up of mobile telephones outside major cities 21.
Setting
After piloting, fieldwork was conducted from 25 March to 23 April 2013. The average interview length was 14.3 minutes. The study was approved by the University of Queensland's Behavioural and Social Sciences Ethical Review Committee (2011001133).
Response rate
The American Association of Public Opinion Research (AAPOR) response rate 3 29 was 19.5% (21.7% landline; 17.8% mobile), the cooperation rate was 43.1% (38.0% landline; 49.7% mobile) and the refusal rate was 33.0% (42.0% landline; 24.4% mobile). An extended call cycle was employed for the landline frame, while a truncated call cycle, in which no more than four unanswered calls were made per week, was employed for the mobile frame. Most interviews (90% landline, 100% mobile) were obtained within 10 calls. Calls for both samples were made at different times and on different days, including weekends. Other response maximization strategies included the operation of a 1800 number, leaving messages on answering machines/voicemail, refusal conversion interviewing, and interviewing in languages other than English.
Participants
The in‐scope population was Australian residents aged 18 years and over, contactable by either a landline or mobile telephone. The sample comprised 2000 respondents (1000 landline, 1000 mobile) across all states/territories.
Measures
Past‐year gambling participation was measured using single items for multiple gambling activities and internet gambling (to play casino games or poker for money) 6. The nine‐item PGSI 22 using the standard scoring 23 measured past‐year problem gambling severity. Scores were classified into non‐problem (score of 0), low‐risk (scores of 1 or 2), moderate‐risk (scores between 3 and 7) or problem (scores of 8 or higher) gambling. The PGSI has good psychometric properties 2, 22, with very good internal consistency in the current sample (α = 0.88). Consistent with the binge drinking definition in the Australian National Health and Medical Research Council guidelines 24, respondents were required to indicate how frequently they drank five or more alcoholic standard drinks in the previous 3 months. Respondents were also asked to indicate their smoking status. The six‐item Kessler‐6 (K6 25) was employed to measure psychological distress in the past 4 weeks. Scores were classified into low (scores of 6–11), moderate (scores of 12–19) or high (scores of 20–30) psychological distress 26. The K6 is effective at screening for serious mental disorders 27, 28, 29 and had excellent reliability in the current study (α = 0.81). Finally, the Short Form general health status question (SF11 30), which is an accurate predictor of future health care use, mortality and morbidity 30, was employed as a general indicator of physical and mental health.
Statistical weights
A two‐stage weighting approach involved the calculation of: (a) a post‐stratification weight using a raking approach to adjust for the disproportionate nature of the sample and differential survey response rates across age, gender, educational attainment 31, country of birth 32 geographical location 33 and telephone status 21 using independent population benchmarks, and (b) a design weight for each frame that included typical adjustments relating to the number of in‐scope people in each household and the number of landline telephone connections per household. An additional pre‐weight was calculated to adjust for the overlapping chances of selection for persons with both landline and mobile telephones into both sample frames by adjusting for the telephone status of sample members to population parameters 34 (19% mobile telephone‐only users, 72% dual users, 9% landline telephone‐only users 21).
Data analysis
Due to the negligible amount of missing data, single items and PGSI/K6 items with more than 30% missing data were deleted pairwise. For respondents with less than 30% missing PGSI data, a single imputation approach was employed using an Expectation Maximization Likelihood algorithm 35 based on their available PGSI data. All analyses were conducted in Stata version 13. The unweighted demographic sample frame estimates were compared to population benchmarks using single‐sample proportion tests. The remainder of the analyses employed weighted data, which was analysed using robust linerarized standard errors implemented in Stata's ‘svy’ toolbox. Univariate predictors (respondent characteristics) of sample frame (landline frame versus mobile frame) and telephone status (landline frame [reference category] versus mobile dual users and mobile‐only users) were examined using binary and multinomial logistic regression analyses, respectively. Given the large number of characteristics, three separate multivariate models were estimated in which all the significant univariate demographic, psychosocial and gambling characteristics (except individual PGSI items) (P < 0.05) were entered as predictors, respectively. All significant predictors (P < 0.05) in these models were retained as predictors in the final binary and multinomial multivariate models. Finally, separate ordinal logistic regression models with simple slopes analyses were employed to explore the degree to which sample frame (landline frame versus mobile frame) and telephone status (landline frame versus mobile‐only users) moderate the relationships between each respondent characteristic and problem gambling category.
Results
Dual‐frame national gambling estimates
The dual‐frame national gambling estimates (Table 1) reveal a rate of 63.9% [95% confidence interval (CI) = 61.4–66.3] for gambling participation, 0.4% (95% CI = 0.2–0.8) for problem gambling, 1.9% (95% CI = 1.3–2.6) for moderate‐risk gambling and 3.0% (95% CI = 2.2–4.0) for low‐risk gambling.
Table 1.
Estimates of gambling participation and problem gambling by sample frame and telephone status.a
| Gambling characteristic | Dual‐frame | Landline frame | Mobile telephone frame | Sample frame (binary Logistic regression results) b | Telephone status (multinomial logistic regression result b | |||||
|---|---|---|---|---|---|---|---|---|---|---|
| (n = 2000) % | Total (n = 1000) % | Dual‐user (n = 817) % | Landline only (n = 183) % | Total (n = 1000) % | Dual‐user (n = 647) % | Mobile telephone only (n = 380) % | Landline Total versus mobile Total (OR) | Landline Total versus mobile dual (RRR) | Landline Total versus mobile telephone only (RRR) | |
| Gambling participation | ||||||||||
| Electronic gaming machines | 20.7 (18.8, 22.7) | 18.9 (16.2, 22.0) | 19.8 (17.2, 22.7) | 17.6 (12.4, 24.3) | 22.5 (19.9, 25.3) | 20.7 (17.7, 24.1) | 24.6 (20.4, 29.4) | 1.2 (1.0, 1.6) | 1.1 (0.9, 1.5) | 1.4* (1.0, 1.9) |
| Horse or greyhound races | 15.9 (14.2, 17.7) | 13.9 (11.6, 16.5) | 16.4 (13.9, 19.1) | 9.9 (5.9, 15.9) | 18.0 (15.7, 20.6) | 17.6 (14.9, 20.8) | 18.4 (14.7, 22.8) | 1.4* (1.0, 1.8) | 1.3 (1.0, 1.8) | 1.4 (1.0, 2.0) |
| Instant scratch tickets | 22.0 (20.1, 24.1) | 23.8 (20.9, 27.0) | 26.6 (23.6, 29.8) | 19.4 (14.1, 26.1) | 20.1 (17.7, 22.8) | 21.7 (18.7, 25.1) | 18.1 (14.4, 22.5) | 0.8 (0.6, 1.0) | 0.9 (0.7, 1.1) | 0.7* (0.5, 1.0) |
| Lottery tickets | 49.2 (46.7, 51.7) | 54.7 (50.9, 58.5) | 56.2 (52.7, 59.6) | 52.3 (44.0, 60.5) | 43.3 (40.2, 46.5) | 48.8 (44.9, 52.7) | 36.5 (31.7, 41.7) | 0.6** (0.5, 0.8) | 0.8* (0.6, 1.0) | 0.5** (0.4, 0.6) |
| Keno | 7.2 (6.0, 8.6) | 7.5 (5.7, 9.8) | 7.7 (6.1, 9.8) | 7.1 (3.7, 13.1) | 6.8 (5.4, 8.6) | 6.6 (4.9, 8.8) | 7.1 (4.8, 10.3) | 0.9 (0.6, 1.3) | 0.9 (0.6, 1.3) | 0.9 (0.6, 1.6) |
| Table games at a casino | 5.9 (5.0, 7.1) | 2.2 (1.4, 3.3) | 3.1 (2.1, 4.6) | 0.6 (0.1, 4.3) | 10.0 (8.2, 12.1) | 9.5 (7.4, 12.1) | 10.5 (7.7, 14.1) | 5.0** (3.1, 8.1) | 4.8** (2.9, 7.9) | 5.3** (3.1, 9.2) |
| Bingo at a club or hall | 3.0 (2.4, 3.8) | 2.8 (2.0, 4.0) | 3.6 (2.5, 5.2) | 1.5 (0.6, 3.7) | 3.2 (2.3, 4.5) | 3.3 (2.2, 5.0) | 3.1 (1.7, 5.5) | 1.1 (0.7, 1.9) | 1.2 (0.7, 2.0) | 1.1 (0.6, 2.2) |
| Sporting event | 5.7 (4.8, 6.8) | 3.8 (2.8, 5) | 5.7 (4.3, 7.6) | 0.6 (0.2, 2.4) | 7.8 (6.3, 9.7) | 7.1 (5.3, 9.3) | 8.8 (6.2, 12.2) | 2.2** (1.5, 3.2) | 1.9** (1.3, 3.0) | 2.5** (1.5, 4.0) |
| Private games | 3.1 (2.4, 4.0) | 2.2 (1.5, 3.3) | 3 (2.0, 4.4) | 0.9 (0.2, 3.9) | 4.1 (3.0, 5.6) | 3.3 (2.1, 5.1) | 5.1 (3.2, 8.0) | 1.9* (1.1, 3.3) | 1.5 (0.8, 2.8) | 2.4** (1.3, 4.5) |
| Internet including mobile devices | 1.5 (1.1, 2.2) | 0.4 (0.2, 1.0) | 0.7(0.3, 1.7) | 0.0 (0.0, 0.0) | 2.7 (1.8, .4.0) | 2.2 (1.3, 3.6) | 3.4 (1.9, 5.9) | 6.5** (2.5, 17.1) | 5.1** (1.8, 14.5) | 8.2** (2.8, 23.5) |
| Any gambling activity | 63.9 (61.4, 66.3) | 65.1 (61.3, 68.7) | 67.3 (64.0, 70.5) | 61.6 (53.1, 69.3) | 62.6 (59.4, 65.6) | 64.1 (60.3, 67.8) | 60.6 (55.4, 65.6) | 0.9 (0.7, 1.1) | 1.0 (0.8, 1.2) | 0.8 (0.6, 1.1) |
| Number of gambling activities | ||||||||||
| 0 | 36.1 (33.7, 38.6) | 34.9 (31.3, 38.7) | 32.7 (29.5, 36.0) | 38.4 (30.7, 46.9) | 37.4 (34.4, 40.6) | 35.9 (32.2, 39.7) | 39.4 (34.4, 44.6) | 1.1 (0.9, 1.4) | 1.0 (0.8, 1.3) | 1.2 (0.9, 1.6) |
| 1 | 25.9 (23.7, 28.2) | 27.0 (23.6, 30.7) | 24.3 (21.5, 27.4) | 31.4 (24.1, 39.7) | 24.7 (22.1, 27.5) | 24.7 (21.5, 28.2) | 24.6 (20.4, 29.4) | 0.9 (0.7, 1.1) | 0.9 (0.7, 1.1) | 0.9 (0.7, 1.2) |
| 2–4 | 34.9 (32.6, 37.3) | 36.2 (32.7, 39.9) | 40.0 (36.6, 43.4) | 30.2 (23.4, 38) | 33.6 (30.6, 36.6) | 35.7 (32.1, 39.5) | 30.9 (26.3, 35.9) | 0.9 (0.7, 1.1) | 1.0 (0.8, 1.2) | 0.8 (0.6, 1.0) |
| 5+ | 3.0 (2.4, 3.9) | 1.8 (1.2, 2.7) | 3.0 (2.0, 4.4) | 0.0 (0.0, 0.0) | 4.3 (3.2, 5.8) | 3.7 (2.4, 5.5) | 5.1 (3.2, 8.0) | 2.4** (1.4, 4.0) | 2.0* (1.1, 3.7) | 2.9** (1.5, 5.4) |
| PGSI category | ||||||||||
| Non‐gambling/non‐problem gambling | 94.7 (93.5, 95.7) | 95.6 (93.7, 96.9) | 95.7 (94.0, 96.9) | 95.4 (90.7, 97.8) | 93.9 (92.1, 95.3) | 94.3 (92.1, 95.8) | 93.4 (90.3, 95.6) | 0.7 (0.5, 1.1) | 0.8 (0.5, 1.3) | 0.7 (0.4, 1.2) |
| Low‐risk gambling | 3.0 (2.2, 4.0) | 2.8 (1.8, 4.4) | 2.4 (1.5,3.8) | 3.4 (1.4, 7.7) | 3.1 (2.2, 4.5) | 2.7 (1.6, 4.3) | 3.7 (2.2, 6.3) | 1.1 (0.6, 2.1) | 1.0 (0.5, 1.9) | 1.3 (0.6, 2.8) |
| Moderate‐risk gambling | 1.9 (1.3, 2.6) | 1.5 (0.9, 2.7) | 1.7 (1.0, 2.9) | 1.2 (0.3, 4.8) | 2.2 (1.4, 3.4) | 2.4 (1.4, 3.9) | 2.0 (1.0, 4.1) | 1.5 (0.7, 3.0) | 1.6 (0.7, 3.5) | 1.3 (0.5, 3.4) |
| Problem gambling | 0.4 (0.2, 0.8) | 0.1 (0.0, 0.5) | 0.2 (0.0, 0.8) | 0.0 (0.0, 0.0) | 0.8 (0.4, 1.6) | 0.7 (0.3, 1.7) | 0.9 (0.3, 2.6) | 6.4* (1.3, 30.8) | 5.8* (1.1, 30.2) | 7.2* (1.2, 43.5) |
| PGSI item endorsementc | ||||||||||
| Bet more than could really afford to lose | 3.8 (2.8, 5.1) | 2.3 (1.3, 4.1) | 1.9 (1.0, 3.4) | 3.0 (1.1, 8.2) | 5.4 (3.8, 7.5) | 5.5 (3.6, 8.2) | 5.3 (2.9, 9.3) | 2.4* (1.2, 4.8) | 2.5* (1.2, 5.1) | 2.4* (1.0, 5.5) |
| Gamble with larger amounts for same excitement | 2.7 (1.9, 3.8) | 1.7 (0.9, 3.1) | 2.0 (1.1, 3.6) | 1.0 (0.1, 6.8) | 3.9 (2.6, 5.9) | 3.7 (2.2, 6.1) | 4.3 (2.2, 8.0) | 2.4* (1.1, 5.3) | 2.3 (1.0, 5.2) | 2.7* (1.1, 6.8) |
| Gone back another day to win back lost money | 3.2 (2.3, 4.4) | 1.9 (1.0, 3.6) | 1.8 (1.0, 3.3) | 2.0 (0.5, 7.7) | 4.6 (3.2, 6.7) | 3.8 (2.3, 6.1) | 5.7 (3.3, 9.8) | 2.5* (1.2, 5.5) | 2.1 (0.9, 4.8) | 3.2* (1.3, 7.7) |
| Borrowed or sold something to get money to gamble | 1.0 (0.6, 1.8) | 0.3 (0.1, 1.1) | 0.5 (0.2, 1.7) | 0.0 (0.0, 0.0) | 1.8 (1.0, 3.3) | 1.4 (0.6, 3.1) | 2.4 (1.0, 5.6) | 5.4* (1.5, 19.9) | 4.1 (1.0, 16.8) | 7.2** (1.7, 30.7) |
| Have a problem with gambling | 2.5 (1.7, 3.6) | 1.8 (0.9, 3.7) | 1.9 (1.1, 3.5) | 1.5 (0.2, 10.0) | 3.3 (2.1, 5.0) | 3.6 (2.1, 5.9) | 2.9 (1.3, 6.2) | 1.8 (0.8, 4.4) | 2.0 (0.8, 5.0) | 1.6 (0.5, 4.8) |
| Gambling caused health problems | 2.2 (1.4, 3.3) | 1.6 (0.7, 3.6) | 0.8 (0.3, 2.1) | 3.0 (1.0, 9.0) | 2.8 (1.8, 4.5) | 3.6 (2.1, 5.9) | 1.9 (0.7, 5.0) | 1.8 (0.7, 4.8) | 2.3 (0.8, 6.2) | 1.2 (0.3, 4.4) |
| People criticized gambling | 2.1 (1.4, 3.1) | 1.3 (0.6, 2.8) | 1.3 (0.6, 2.7) | 1.5 (0.3, 6.3) | 3.0 (1.9, 4.7) | 3.0 (1.7, 5.2) | 2.9 (1.3, 6.2) | 2.2 (0.9, 5.4) | 2.3 (0.9, 5.9) | 2.2 (0.7, 6.5) |
| Gambling caused financial problems | 1.4 (0.9, 2.2) | 0.4 (0.2, 1.1) | 0.6 (0.2, 1.7) | 0.0 (0.0, 0.0) | 2.5 (1.5, 4.0) | 2.5 (1.4, 4.5) | 2.4 (1.0, 5.6) | 6.1** (2.0, 18.8) | 6.2** (1.9, 20.2) | 5.9** (1.5, 22.5) |
| Felt guilty about gambling | 4.5 (3.5, 5.9) | 3.7 (2.5, 5.7) | 4.5 (3.0, 6.7) | 2.5 (0.8, 7.8) | 5.4 (3.9, 7.6) | 5.9 (4.0, 8.7) | 4.8 (2.6, 8.6) | 1.5 (0.8, 2.6) | 1.6 (0.9, 3.0) | 1.3 (0.6, 2.8) |
Weighted data.
Regression significance:
P < 0.01;
P < 0.05.
Of those who endorsed any gambling activity (n = 1287). OR = odds ratio; RRR = relative risk ratio; PGSI = Problem Gambling Severity Index.
Univariate predictors of sample frame and telephone status
Demographic characteristics (Table 2)
Table 2.
Sample demographic profile by sample frame and telephone status.a
| Demographic characteristic | Population Benchmarks | Dual frame | Landline frame | Mobile telephone frame | Sample frame (binary logistic regression results) c | Telephone status (multinomial logistic regression results) c | |||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| (n = 2000) % | Total (n = 1000) % | Dual‐user (n = 817) % | Landline only (n = 183) % | Total (n = 1000) % | Dual‐user (n = 647) % | Mobile telephone only (n = 353) % | Landline Total versus Mobile Total (OR) | Landline Total versus mobile dual (RRR) | Landline Total versus Mobile Telephone only (RRR) | ||
| Gender (male) | 48.9 | 44.5b (42.3, 46.7) | 36.5b (33.6, 39.5) | 37.2 (34, 40.6) | 33.3 (26.9, 40.5) | 52.5b (49.4, 55.6) | 50.9 (47.0, 54.7) | 55.5 (50.3, 60.6) | 1.9** (1.6, 2.3) | 1.8** (1.5, 2.2) | 2.2** (1.7, 2.8) |
| Age group (years) | |||||||||||
| 18–24 | 12.2 | 11.8 (10.4, 13.3) | 4.9b (3.7, 6.5) | 5.4 (4.1, 7.2) | 2.7 (1.1, 6.4) | 18.6b (16.3, 21.1) | 16.0 (13.3, 19.0) | 23.4 (19.3, 28.2) | 4.4** (3.2, 6.1) | 3.7** (2.6, 5.2) | 5.9** (4.0, 8.6) |
| 25–39 | 27.2 | 23.6b (21.8, 25.5) | 16.2b (14.0, 18.6) | 18.3 (15.8, 21.2) | 6.6 (3.8, 11.3) | 31.0b (28.2, 33.9) | 24.2 (21, 27.6) | 43.4 (38.3, 48.7) | 2.3** (1.9, 2.9) | 1.7** (1.3, 2.1) | 4.0** (3.0, 5.2) |
| 40–49 | 18.5 | 17.3 (15.7, 19.0) | 17.4 (15.2, 19.9) | 20.0 (17.3, 22.8) | 6.0 (3.4, 10.6) | 17.2 (15.0, 19.7) | 18.4 (15.6, 21.6) | 14.9 (11.5, 19.0) | 1.0 (0.8, 1.2) | 1.1 (0.8, 1.4) | 0.8 (0.6, 1.2) |
| 50–64 | 23.9 | 27.2b (25.3, 29.2) | 30.3b (27.5, 33.2) | 30.9 (27.8, 34.2) | 27.5 (21.5, 34.4) | 24.1 (21.6, 26.9) | 29.5 (26.1, 33.1) | 14.3 (11.0, 18.4) | 0.7** (0.6, 0.9) | 1.0 (0.8, 1.2) | 0.4** (0.3, 0.5) |
| 65+ | 18.2 | 20.2 b (18.5, 22.0) | 31.2 b (28.4, 34.1) | 25.4 (22.5, 28.5) | 57.1 (49.8, 64.2) | 9.1 b (7.5, 11.1) | 11.9 (9.7, 14.7) | 4.0 (2.4, 6.6) | 0.2** (0.2, 0.3) | 0.3** (0.2, 0.4) | 0.1** (0.1, 0.2) |
| Region (capital city) | 66.1 | 68.9b (66.8, 70.8) | 63.8 (60.8, 66.7) | 63.9 (60.5, 67.1) | 63.4 (56.1, 70.1) | 73.9b (71.1, 76.5) | 74.5 (71.0, 77.7) | 72.8 (67.9, 77.2) | 1.6** (1.3, 1.9) | 1.7** (1.3, 2.1) | 1.5** (1.2, 2.0) |
| Aboriginal & Torres Strait Islander status | 1.9 | 1.2b (0.8, 1.8) | 1.3b (0.8, 2.2) | 1.1 (0.6, 2.1) | 2.2 (0.8, 5.7) | 1.1b (0.6, 2.0) | 0.9 (0.4, 2.1) | 1.4 (0.6, 3.4) | 0.8 (0.4, 1.9) | 0.7 (0.3, 1.9) | 1.1 (0.4, 3.1) |
| Birthplace (Australia) | 64.6 | 65.2 (63.1, 67.3) | 69.6b (66.7, 72.4) | 72.7 (69.5, 75.6) | 56.3 (49.0, 63.3) | 60.7b (57.6, 63.7) | 61.2 (57.3, 64.9) | 59.9 (54.6, 64.9) | 0.7** (0.6, 0.8) | 0.7** (0.6, 0.8) | 0.7** (0.5, 0.8) |
| Time in neighbourhood (more than 5 years) | 60.2 (58.0, 62.3) | 72.7 (69.9, 75.4) | 70.1 (66.8, 73.1) | 84.6 (78.6, 89.2) | 47.6 (44.5, 50.7) | 59.6 (55.7, 63.3) | 25.6 (21.3, 30.4) | 0.3** (0.3, 0.4) | 0.6** (0.4, 0.7) | 0.1** (0.1, 0.2) | |
| Home ownership | |||||||||||
| Own home | 36.6 (34.3, 38.8) | 47.7 (44.5, 50.9) | 44.0 (40.5, 47.6) | 64.2 (56.6, 71.2) | 24.8 (22.1, 27.8) | 34.3 (30.4, 38.4) | 8.3 (5.7, 11.9) | 0.4** (0.3, 0.4) | 0.6** (0.5, 0.7) | 0.1** (0.1, 0.2) | |
| Paying mortgage | 30.8 (28.7, 33.0) | 32.5 (29.5, 35.6) | 37.1 (33.7, 40.6) | 11.5 (7.5, 17.4) | 29.0 (26.1, 32.1) | 34.1 (30.3, 38.2) | 20.0 (15.9, 24.8) | 0.8 (0.7, 1.0) | 1.1 (0.9, 1.3) | 0.5** (0.4, 0.7) | |
| Paying rent | 32.7 (30.5, 34.9) | 19.8 (17.4, 22.6) | 18.9 (16.2, 21.8) | 24.2 (18.3, 31.4) | 46.2 (42.9, 49.5) | 31.6 (27.8, 35.6) | 71.7 (66.5, 76.5) | 3.5** (2.8, 4.3) | 1.9** (1.5, 2.4) | 10.3** (7.6, 13.8) | |
| Living arrangements | |||||||||||
| With parents/guardians | 7.9 (6.8, 9.2) | 4.4 (3.3, 5.9) | 4.6 (3.4, 6.3) | 3.4 (1.5, 7.4) | 11.5 (9.6, 13.7) | 14.3 (11.7, 17.2) | 6.3 (4.1, 9.5) | 2.8** (1.9, 4.0) | 3.6** (2.5, 5.3) | 1.4 (0.8, 2.5) | |
| Couple only | 26 (24.0, 28.0) | 28.3 (25.5, 31.2) | 28.5 (25.5, 31.7) | 27.4 (21.3, 34.5) | 23.6 (21.0, 26.4) | 23.7 (20.5, 27.2) | 23.4 (19.1, 28.2) | 0.8* (0.6, 1.0) | 0.8* (0.6, 1.0) | 0.8 (0.6, 1.0) | |
| Couple with children | 31.8 (29.8, 33.9) | 33.2 (30.3, 36.3) | 35.9 (32.6, 39.3) | 21.1 (15.7, 27.8) | 30.4 (27.5, 33.4) | 35.4 (31.8, 39.3) | 21.0 (16.9, 25.7) | 0.9 (0.7, 1.1) | 1.1 (0.9, 1.4) | 0.5** (0.4, 0.7) | |
| One parent family | 6.8 (5.8, 8.1) | 6.6 (5.2, 8.3) | 6.5 (5.0, 8.5) | 6.9 (3.9, 11.7) | 7.1 (5.6, 8.9) | 7.2 (5.4, 9.5) | 6.9 (4.6, 10.2) | 1.1 (0.8, 1.5) | 1.1 (0.7, 1.6) | 1.0 (0.6, 1.7) | |
| Group household | 9.1 (7.9, 10.5) | 3.4 (2.4, 4.7) | 3.4 (2.3, 4.9) | 3.4 (1.5, 7.4) | 14.9 (12.8, 17.3) | 10.3 (8.1, 12.9) | 23.7 (19.4, 28.5) | 5.0** (3.4, 7.4) | 3.3** (2.1, 5.0) | 8.8** (5.7, 13.5) | |
| One person household | 18.3 (16.7, 20.1) | 24.1 (21.5, 26.9) | 3.4 (2.3, 4.9) | 3.4 (1.5, 7.4) | 12.5 (10.6, 14.8) | 10.3 (8.1, 12.9) | 23.7 (19.4, 28.5) | 0.5** (0.4, 0.6) | 0.3** (0.2, 0.4) | 0.7 (0.5, 1.0) | |
| Relationship status (in ongoing relationship) | 25.7 (22.9, 28.7) | 19.6 (16, 23.8) | 24.0 (19.4, 29.2) | 6.2 (2.8, 13.2) | 30.8 (26.8, 35.1) | 29.9 (24.7, 35.6) | 32.0 (25.9, 38.8) | 1.8** (1.3, 2.5) | 1.7** (1.2, 2.5) | 1.9** (1.3, 2.8) | |
| Education (Bachelor's degree or higher) | 19.8 | 32.6b (30.5, 34.7) | 26.9b (24.2, 29.8) | 30.2 (27.2, 33.5) | 11.4 (7.5, 17.1) | 38.3b (35.3, 41.4) | 38.9 (35.2, 42.7) | 37.2 (32.2, 42.5) | 1.7** (1.4, 2.0) | 1.7** (1.4, 2.1) | 1.6** (1.2, 2.1) |
| Household income | |||||||||||
| Less than $60 000 | 49.0 (46.5, 51.5) | 52.8 (49.2, 56.3) | 47.4 (43.5, 51.2) | 79.8 (72.0, 85.9) | 45.1 (41.7, 48.7) | 41.7 (37.5, 46.1) | 51.3 (45.4, 57.1) | 0.7** (0.6, 0.9) | 0.6** (0.5, 0.8) | 0.9 (0.7, 1.2) | |
| $60 000 to less than $100 000 | 25.1 (23.0, 27.3) | 22.4 (19.6, 25.5) | 24.5 (21.4, 28.0) | 11.6 (7.1, 18.4) | 27.8 (24.8, 31.1) | 29.0 (25.2, 33.2) | 25.6 (20.8, 31.1) | 1.3* (1.1, 1.7) | 1.4** (1.1, 1.8) | 1.2 (0.9, 1.6) | |
| $100 000 or more | 25.9 (23.8, 28.2) | 24.8 (21.9, 28.0) | 28.1 (24.8, 31.7) | 8.5 (4.8, 14.8) | 27.0 (24.0, 30.3) | 29.2 (25.4, 33.4) | 23.1 (18.5, 28.4) | 1.1 (0.9, 1.4) | 1.3 (1.0, 1.6) | 0.9 (0.7, 1.3) | |
Unweighted data.
Unweighted proportion is significantly different from population benchmark using single sample proportion tests (P < 0.05).
Regression significance:
P < 0.01;
P < 0.05. OR = odds ratio; RRR = relative risk ratio.
Comparison to population benchmarks
Compared to population benchmarks, the landline frame under‐represented male, aged 18–39 years, and Aboriginal and Torres Strait Islander (ATSI) respondents and over‐represented 50+ year‐old, Australian‐born and university‐educated respondents. In contrast, the mobile frame under‐represented 65+ year‐old, ATSI and Australian‐born respondents and over‐represented male, 18–39‐year‐old, capital city‐residing and university‐educated respondents. The dual frame sample under‐represented male, 25–39‐year‐old and ATSI respondents and over‐represented 50+ year‐old, capital city‐residing and university‐educated respondents.
Sample frame
Relative to landline frame, the mobile frame was significantly more likely to be male, aged 18–39 years, residing in capital cities, renting, living with parents, living in a group household, in an ongoing relationship, university‐educated and earning $60 000–100 000; and significantly less likely to be aged 50+ years, Australian‐born, residentially stable, home owners, living as a couple, living alone and earning less than $60 000.
Telephone status
Relative to the landline frame, both mobile dual users and mobile‐only users were significantly more likely to be male, aged 18–39 years, residing in capital cities, renting, living in a group household, in an ongoing relationship and university‐educated; and significantly less likely to be 65+ years, Australian‐born, residentially stable and home owners. However, only the mobile‐only users were significantly less likely to be aged 50–64 years, paying a mortgage and living as a couple with children. In contrast, only the mobile dual users were significantly more likely to be living with parents, and earning $60 000–100 000; and significantly less likely to be living as a couple without children, living alone and earning less than $60 000.
Psychosocial characteristics (Table 3)
Table 3.
Estimates of psychosocial characteristics by sample frame and telephone status.a
| Psychosocial characteristic | Dual‐frame | Landline frame | Mobile telephone frame | Sample frame (binary logistic regression results) b | Telephone status (multinomial logistic regression results) b | |||||
|---|---|---|---|---|---|---|---|---|---|---|
| (n = 2000) % | Total (n = 1000) % | Dual‐user (n = 817) % | Landline only (n = 183) % | Total (n = 1000) % | Dual‐user (n = 647) % | Mobile telephone only (n = 380) % | Landline Total versus Mobile Total (OR) | Landline Total versus Mobile dual (RRR) | Landline Total versus mobile telephone only (RRR) | |
| K6 psychological distress | ||||||||||
| Low | 73.4 (71.2, 75.6) | 75.8 (72.3, 79.0) | 77.1 (74.1, 79.9) | 73.6 (65.6, 80.2) | 71.0 (68.0, 73.8) | 72.1 (68.4, 75.4) | 69.7 (64.7, 74.3) | 0.8* (0.6, 1.0) | 0.8 (0.6, 1.1) | 0.7* (0.5, 1.0) |
| Moderate | 22.7 (20.7, 24.9) | 20.5 (17.6, 23.9) | 19.8 (17.1, 22.7) | 21.8 (15.7, 29.4) | 25.0 (22.4, 27.9) | 24.9 (21.7, 28.5) | 25.2 (21.0, 30.0) | 1.3* (1.0, 1.6) | 1.3 (1.0, 1.7) | 1.3 (1.0, 1.8) |
| High | 3.8 (2.9, 5.0) | 3.7 (2.4, 5.6) | 3.1 (2.1, 4.5) | 4.6 (2.1, 9.9) | 4.0 (2.9, 5.4) | 3.0 (2.0, 4.7) | 5.1 (3.2, 8.0) | 1.1 (0.6, 1.9) | 0.8 (0.4, 1.5) | 1.4 (0.7, 2.7) |
| Drank 5 or more standard drinks in one session at least twice per week in past 3 months | 9.5 [8.2,11.0] | 7.1 (5.5, 9.1) | 7.9 (6.2, 10.0) | 5.9 (3.1, 10.7) | 12.1 (10.1, 14.3) | 9.1 (7.1, 11.6) | 15.7 (12.3, 19.9) | 1.8** (1.3, 2.5) | 1.3 (0.9, 1.9) | 2.4** (1.6, 3.6) |
| Daily smoking | 13.7 [12.1,15.4] | 10.6 (8.6, 12.9) | 11.1 (9.1, 13.5) | 9.8 (6.2, 15.1) | 16.9 (14.7, 19.5) | 13.9 (11.4, 16.8) | 20.7 (16.8, 25.2) | 1.7** (1.3, 2.3) | 1.4 (1, 1.9) | 2.2** (1.6, 3.1) |
| SF‐1 health (excellent or very good) | 54.5 [52.0,57.0] | 50.4 (46.5, 54.2) | 56.6 (53.2, 60.0) | 40.4 (32.6, 48.8) | 58.9 (55.8, 62.0) | 57.0 (53.1, 60.8) | 61.3 (56.0, 66.2) | 1.4** (1.2, 1.7) | 1.3* (1.0, 1.6) | 1.6** (1.2, 2.0) |
Weighted data.
Regression significance:
P < 0.01;
P < 0.05. OR = odds ratio; RRR = relative risk ratio.
Sample frame
Relative to the landline frame, the mobile frame was significantly more likely to report moderate psychological distress, binge drinking, daily smoking and excellent/good health; and significantly less likely to report low psychological distress.
Telephone status
Relative to the landline frame respondents, both mobile dual users and mobile‐only users were significantly more likely to report excellent/good health. However, only the mobile‐only users were significantly more likely to report binge drinking and daily smoking and significantly less likely to report low psychological distress.
Gambling characteristics (Table 1)
Sample frame
Relative to the landline frame, the mobile frame was significantly more likely to gamble on horse or greyhound races, casino table games, sporting events, private games and the internet; significantly less likely to gamble on lotteries; and significantly more likely to gamble on five or more activities, be classified as problem gamblers and to endorse several PGSI items.
Telephone status
Relative to the landline frame, both mobile dual users and mobile‐only users were significantly more likely to gamble on casino table games, sporting events and the internet; significantly less likely to gamble on lotteries; and significantly more likely to gamble on five or more activities, be classified as problem gamblers and to endorse several PGSI items. However, only the mobile‐only users were significantly more likely to gamble on EGMs and private games and endorse several PGSI items; and significantly less likely to gamble on instant scratch tickets.
Multivariate predictors of sample frame and telephone status
Sample frame
The final multivariate model (Table 4) revealed that younger age, residential mobility, renting, casino table gambling, internet gambling and having excellent/good health were significant independent predictors of the mobile frame relative to the landline frame.
Table 4.
Multivariate predictors of sample frame and telephone status.a
| Sample frame (binary logistic model) | Telephone status (multinomial logistic model) | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Landline telephone frame (ref) versus mobile telephone frame | Landline telephone frame respondents (ref) versus mobile telephone dual users | Landline telephone frame respondents (ref) versus mobile telephone only users | |||||||
| Odds ratio | 95% CI | P | Odds ratio | 95% CI | P | Odds ratio | 95% CI | P | |
| Age | 0.96 | (0.95, 0.97) | .000 | 0.97 | (0.96, 0.97) | .000 | 0.95 | (0.94, 0.96) | .000 |
| Time in neighbourhood (+5 years) | |||||||||
| No | 1 (ref) | 1 (ref) | 1 (ref) | ||||||
| Yes | 0.61 | (0.47, 0.8) | .000 | 0.79 | (0.6, 1.06) | .115 | 0.40 | (0.27, 0.58) | .000 |
| Home ownership | |||||||||
| Own home | 1 (ref)2 | 1 (ref) | 1 (ref) | ||||||
| Paying mortgage | 0.96 | (0.7, 1.31) | .791 | 0.91 | (0.66, 1.26) | .580 | 1.52 | (0.87, 2.63) | .138 |
| Paying rent | 2.06 | (1.47, 2.89) | .000 | 1.19 | (0.83, 1.71) | .348 | 5.84 | (3.38, 10.07) | .000 |
| Electronic gaming machines | |||||||||
| No | NAb | 1 (ref) | |||||||
| Yes | 1.00 | (0.69, 1.46) | .989 | 1.29 | (0.77, 2.15) | .329 | |||
| Lottery tickets | |||||||||
| No | 1 (ref) | 1 (ref) | 1 (ref) | ||||||
| Yes | 0.71 | (0.49, 1.03) | .074 | 0.90 | (0.59, 1.38) | .639 | 0.51 | (0.3, 0.85) | .009 |
| Table games at a casino | |||||||||
| No | 1 (ref) | 1 (ref) | 1 (ref) | ||||||
| Yes | 2.94 | (1.62, 5.31) | .000 | 3.03 | (1.64, 5.58) | .000 | 2.57 | (1.24, 5.35) | .012 |
| Sporting event | |||||||||
| No | 1 (ref) | 1 (ref) | 1 (ref) | ||||||
| Yes | 1.02 | (0.59, 1.77) | .935 | 1.08 | (0.6, 1.94) | .805 | 1.03 | (0.52, 2.06) | .931 |
| Internet including mobile devices | |||||||||
| No | 1 (ref) | 1 (ref) | 1 (ref) | ||||||
| Yes | 3.47 | (1.19, 10.09) | .022 | 3.57 | (1.14, 11.21) | .029 | 3.36 | (0.91, 12.38) | .069 |
| Number of gambling activities | |||||||||
| 0 | 1 (ref) | 1 (ref) | 1 (ref) | ||||||
| 1 | 1.11 | (0.74, 1.66) | .612 | 0.95 | (0.6, 1.49) | .820 | 1.29 | (0.75, 2.22) | .366 |
| 2–4 | 1.10 | (0.71, 1.71) | .674 | 0.98 | (0.57, 1.68) | .931 | 1.08 | (0.54, 2.16) | .834 |
| 5+ | 0.68 | (0.28, 1.62) | .383 | 0.54 | (0.18, 1.6) | .262 | 0.67 | (0.19, 2.43) | .546 |
| Drank 5+ standard drinks in one session at least twice per week in past 3 months | |||||||||
| No | 1 (ref) | 1 (ref) | 1 (ref) | ||||||
| Yes | 1.43 | [0.97, 2.12] | .073 | 1.18 | [0.77, 1.82] | .447 | 1.98 | [1.2, 3.27] | .008 |
| Daily smoking | |||||||||
| No | 1 (ref) | 1 (ref) | 1 (ref) | ||||||
| Yes | 1.39 | [0.99, 1.95] | .057 | 1.29 | [0.89, 1.88] | .182 | 1.54 | [0.98, 2.4] | .060 |
| SF‐1 health (excellent or very good) | |||||||||
| No | 1 (ref) | 1 (ref) | 1 (ref) | ||||||
| Yes | 1.33 | [1.04, 1.69] | .022 | 1.22 | [0.95, 1.58] | .118 | 1.53 | [1.09, 2.16] | .014 |
Weighted data.
Not significant in preliminary logistic models and thus not used as predictor in final logistic model. CI = confidence interval.
Telephone status
The final multivariate model (Table 4) revealed that only younger age and casino table gambling were significant predictors of both mobile dual users and mobile‐only users relative to the landline frame. However, residential mobility, renting, lower lottery participation, binge drinking and excellent/good health were significant independent predictors of only mobile‐only users; and internet gambling was the only significant independent predictor of only mobile dual users.
Moderation by sample frame and telephone status
Sample frame
Sample frame had a significant interaction with gender [b = 1.05, standard error (SE) = 0.47, P = 0.027, 95% CI = 0.12–2.00, odds ratio (OR) = 2.86] in predicting PGSI category (see Fig. 1 for simple slopes illustration). Males had a lower probability of being classified as a non‐problem gambler relative to the higher‐risk categories than females, but only for those in the mobile frame.
Figure 1.
Interaction effect of gender and sample frame (landline telephone frame versus mobile telephone frame) in predicting Problem Gambling Severity Index (PGSI) category. Data are presented only for the probability associated with non‐problem gambling status for males and females across sample frame. The simple slope for the mobile telephone frame was significant [b = 0.97, standard error (SE) = 0.30, P = 0.002, 95% confidence interval (CI) = 0.37–1.57], but the simple slope for the landline telephone frame was not significant (b = –0.08, SE = 0.36, P = 0.817, 95% CI = –0.80–0.63)
Telephone status
Telephone status had a significant interaction with gender (b = 1.52, SE = 0.67, P = 0.023, 95% CI = 0.21–2.84, OR = 4.59), relationship status (b = –2.03, SE = 0.95, P = 0.032, 95% CI = 0.17–3.89, OR = 7.63) and excellent or good health (b = 1.31, SE = 0.61, P = 0.032, 95% CI = 0.11–2.51, OR = 3.71) in predicting PGSI category (see Figs 2, 3, 4 for simple slopes illustrations). Males and respondents in a relationship had a lower probability of being classified as a non‐problem gambler relative to the higher‐risk categories than females and those not in a relationship, but only for those in the mobile‐only user group (P = 0.010 and P = 0.039, respectively). In contrast, excellent/good physical health was associated with a higher probability of being classified a non‐problem gambler than poor/fair/good health, but only for those in the landline frame (P = 0.003).
Figure 2.
Interaction effect of gender and telephone status (landline telephone frame respondents versus mobile telephone‐only users) in predicting Problem Gambling Severity Index (PGSI) category. Data are presented only for the probability associated with non‐problem gambling status for males and females across telephone status. The simple slope for the mobile telephone‐only users was significant [b = 1.44, standard error (SE) = 0.56, P = 0.010, 95% confidence interval (CI) = 0.34–2.54], but the simple slope for the landline telephone frame respondents was not significant (b = –0.08, SE = 0.37, P = 0.816, 95% CI = –0.80–0.63)
Figure 3.
Interaction effect of relationship status and telephone status (landline telephone frame respondents versus mobile telephone‐only users) in predicting Problem Gambling Severity Index (PGSI) category. Data are presented only for the probability associated with non‐problem gambling status for respondents in and not in a relationship across telephone status. The simple slope for the mobile telephone‐only users was significant [b = 1.04, standard error (SE) = 0.50, P = 0.039, 95% confidence interval (CI) = 0.06–2.02], but the simple slope for the landline telephone frame respondents was not significant (b = 0.99, SE = 0.80, P = 0.216, 95% CI = –2.56–0.58)
Figure 4.
Interaction effect of health category and telephone status (landline telephone frame respondents versus mobile telephone‐only users) in predicting Problem Gambling Severity Index (PGSI) category. Data are presented only for the probability associated with non‐problem gambling status for respondents reporting low health (poor/fair/good) and high health (very good/excellent) across telephone status. The simple slope for the landline telephone frame respondents was significant [b = –1.19, standard error (SE) = 0.40, P = 0.003, 95% confidence interval (CI) = 1.98–0.40], but the simple slope for the mobile telephone‐only users was not significant (b = 0.12, SE = 0.46, P = 0.787, 95% CI = –0.78–1.03)
Discussion
In this first Australian national dual‐frame gambling study, the gambling participation and problem gambling rates are at the low end of recent estimates 10, 20. Although the extant dual‐frame literature reports somewhat inconsistent findings with respect to specific gambling behaviours, the higher participation on many gambling activities 3, 6, problem gambling 12, 14 and PGSI item endorsement 12 by mobile frame respondents compared to landline frame respondents is consistent with previous research. The finding in relation to higher gambling involvement is new, and the finding in relation to higher internet gambling is inconsistent with previous literature 6, 12, due probably to differences in the measurement of internet gambling across studies. The findings therefore add to the growing evidence that there are differences in gambling behaviour between landline and mobile telephone sample frames. Only casino table gambling and internet gambling, however, were associated independently with an increased odds of mobile frame membership, suggesting that many of the gambling behaviour differences between sample frames are accounted for by respondent characteristics, such as younger age, residential mobility, living in a rental property and good health.
There is concern that traditional landline sampling approaches exclude the growing mobile telephone‐only population. Importantly, the mobile telephone‐only sample seems to include respondents characterized by characteristics deemed typically ‘hard to reach’ by conventional landline sampling approaches. They also display demographic 10, 20, 36 and psychosocial 10, 12, 14, 16, 37, 38 profiles that are associated typically with problem gambling. Indeed, the finding that the mobile‐only sample reported higher rates of problem gambling 12, 14 and endorsement of problem gambling measure items 12 suggest that the gambling field has been omitting the exact population most likely to endorse gambling problems. Only casino table gambling and lower lottery participation, however, were significant independent predictors of mobile telephone‐only users, suggesting that many of the gambling behaviour differences between these telephone users are accounted for by other respondent characteristics, such as younger age, residential mobility, living in a rental property, binge drinking and good health.
There is also some concern about the differential non‐response by dual users who predominantly use their mobile telephones. The findings of this study confirm that compared to respondents in the traditional landline frame, mobile telephone dual users were more likely to report a distinct demographic profile and higher gambling participation and problem gambling. Again, casino table gambling and internet gambling were significant independent predictors of mobile telephone dual users. These findings suggest that profile differences produced from the sample frames are not wholly explained by the differences in sample coverage and that dual‐frame methodologies have the added benefit of contacting dual users who predominantly use their mobile telephones.
Males and younger respondents were under‐represented in the landline frame and over‐represented in the mobile frame, while older and Australian‐born respondents were over‐represented in the landline frame and under‐represented in the mobile frame. These findings imply that the inclusion of a mobile frame will probably increase the representativeness of a population sample, particularly in relation to gender, age and country of birth. Despite these promising findings, both sample frames under‐represented ATSI respondents and over‐represented university educated respondents, suggesting that there remain other ‘hard‐to‐reach’ subpopulations, even with the use of dual‐frame methodologies. Moreover, the 50/50 dual‐frame sample still under‐represented males and younger respondents and over‐represented older respondents. It is therefore of concern that the recent dual‐frame Australian gambling surveys have adopted relatively low proportions of mobile telephone sampling (25–30%) 3, 6. Future gambling surveys could achieve additional sample representativeness by adopting a higher proportion of mobile telephone sampling, particularly given expected increases in the size of the mobile telephone‐only population.
The moderation analyses revealed that being female 39 and not in a relationship was protective for the development of gambling problems, but only for respondents in the mobile frame or mobile‐only user group. In contrast, good health was protective only for landline frame respondents. These buffering effects extend the implications of the findings beyond gambling prevalence estimates by indicating that the sample frame employed and the telephone status of respondents can affect the ability of particular characteristics to predict problem gambling development. It is therefore recommended that sample frame and telephone status be taken into consideration in any future examination of these factors in the general population.
Limitations of the current study include a relatively small sample size, moderate response rates, a cross‐sectional design and a reliance upon self‐report measures. This study was, however, the first to explore comprehensively the impact of dual‐frame sampling approaches in a nationally representative sample with standard measures of gambling participation and problems. The results suggest that traditional landline sampling approaches create biases for gambling surveys due to their non‐coverage of mobile telephone‐only users and a differential non‐response by dual users who predominantly use their mobile telephones. While it is of concern that the gambling research field has been excluding the one‐third of the population that is most likely to display the greatest rate of gambling problems, it is unlikely that the existing body of gambling knowledge needs to be reconsidered, because the growth of this subgroup has been only during the last several years 11. Given the expected future increases in this population, however, dual‐frame population sampling methodologies are recommended despite their increased expense.
Declaration of interests
The authors declare that the material has not been published in whole or in part elsewhere; the paper is not currently being considered for publication elsewhere; all authors have been personally and actively involved in substantive work leading to the report, and will hold themselves jointly and individually responsible for its content; all relevant ethical safeguards have been met in relation to patient or subject protection, including, in the case of all clinical and experimental studies review by an appropriate ethical review committee and written informed patient consent. The Problem Gambling Research and Treatment Centre at the University of Melbourne, with which several co‐authors were affiliated when the data were collected, was previously supported by the Victorian Responsible Gambling Foundation. The Victorian Responsible Gambling Foundation had no involvement in the current publication.
Acknowledgements
The authors would like to acknowledge the Social Research Centre Dual‐frame Omnibus Survey as the source of the data presented in this manuscript.
Dowling, N. A. , Youssef, G. J. , Jackson, A. C. , Pennay, D. W. , Francis, K. L. , Pennay, A. , and Lubman, D. I. (2016) National estimates of Australian gambling prevalence: f indings from a dual‐frame omnibus survey. Addiction, 111: 420–435. doi: 10.1111/add.13176.
References
- 1. Williams R., Volberg R., Stevens R. The Population Prevalence of Problem Gambling: Methodological Influences, Standardized Rates, Jurisdictional Differences, and Worldwide Trends In: Report prepared for the Ontario Problem Gambling Research Centre and the Ontario Ministry of Health and Long Term Care. Ontario: Ontario Problem Gambling Research Centre and the Ontario Ministry of Health and Long Term Care; 2012. [Google Scholar]
- 2. Neal P., Delfabbro P., O'Neil M. Problem Gambling and Harm: A National Definition. Adelaide, Australia: South Australia Centre for Economic Studies; 2005. [Google Scholar]
- 3. ACIL Allen Consulting , Social Research Centre , Problem Gambling Research and Treatment Centre Third Social and Economic Impact Study of Gambling in Tasmania. vol. 2, 2013. Tasmania: Tasmanian Gambling Prevalence Survey; (prepared for the Tasmanian Department of Treasury and Finance); 2014. [Google Scholar]
- 4. Davidson T., Rodgers B. 2009 Survey of the Nature and Extent of Gambling, and Problem Gambling, in the Australian Capital Territory. Canberra: Australian National University; 2010. [Google Scholar]
- 5. Queensland Government. Queensland Household Gambling vSurvey 2011–12. State of Queensland: Department of Justice and Attorney‐General; 2012. [Google Scholar]
- 6. Social Research Centre . Gambling Prevalence in South Australia 2012 Office for Problem Gambling, Adelaide, SA (Aust): Department for Communications and Social Inclusion; 2013. [Google Scholar]
- 7. Sproston K., Hing N., Palankay C. Prevalence of Gambling and Problem Gambling in New South Wales. Sydney: NSW Office of Liquor, Gaming and Racing; 2012. [Google Scholar]
- 8. Victorian Department of Justice . A Study of Gambling in Victoria: Problem Gambling From a Public Health Perspective. Victoria: Department of Justice; 2009. [Google Scholar]
- 9. Young M., Abu‐Duhou I., Barnes T., Creed E., Morris M., Stevens M. et al Northern Territory Gambling Prevalence Survey 2005. Darwin: School for Social and Policy Research, Charles Darwin University; 2005. [Google Scholar]
- 10. Gainsbury S. M., Russell A., Hing N., Wood R., Lubman D. I., Blaszczynski A., et al. The prevalence and determinants of problem gambling in Australia: assessing the impact of interactive gambling and new technologies. Psychol Addict Behav J Soc Psychol Addict Behav 2014; 28: 269–279. [DOI] [PubMed] [Google Scholar]
- 11. Australian Communications and Media Authority . Australians get mobile. Australian Communications and Media Authority, 2015. Available at: http://www.acma.gov.au/theACMA/engage‐blogs/engage‐blogs/Research‐snapshots/Australians‐get‐mobile) (accessed 19 October 2015). http://www.webcitation.org/6cPIS69B8 (Archived at: http://www.webcitation.org/6cPIS69B8)
- 12. Jackson A. C., Pennay D., Dowling N. A., Coles‐Janess B., Christensen D. R. Improving gambling survey research using dual‐frame sampling of landline and mobile phone numbers. J Gambl Stud 2014; 30: 291–307. [DOI] [PubMed] [Google Scholar]
- 13. McBride O., Morgan K., McGee H. Recruitment using mobile telephones in an Irish general population sexual health survey: challenges and practical solutions. BMC Med Res Methodol 2012; 12: 45. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Pennay D. 2010. Profiling the ‘mobile phone only’ population: results from a dual‐frame telephone survey using a landline and mobile phone sample frame. Presented at the ACSPRI Social Science Methodology Conference, Sydney. December 1–3, 2010
- 15. Vicente P., Reis E. The mobile phone only population in Portugal and its impact in a dual‐frame telephone survey. Survey Research Methods 2009; 3: 105–111. [Google Scholar]
- 16. Pennay D., Vickers N. J. Social Research Centre Dual Frame Omnibus Survey: Technical and Methodological Report. Social Research Centre: Melbourne; 2013. [Google Scholar]
- 17. Hu S. S., Balluz L., Battaglia M. P., Frankel M. R. Improving public health surveillance using a dual‐frame survey of landline and cell phone numbers. Am J Epidemiol 2011; 173: 703–11. [DOI] [PubMed] [Google Scholar]
- 18. McGeeney K., Keeter S. Pew Research increases share of interviews conducted by cellphone. Pew Research Center. 2015. Available at: http://www.pewresearch.org/fact‐tank/2014/01/15/pew‐research‐increases‐share‐of‐interviews‐conducted‐by‐cellphone/ (accessed 19 October 2015). http://www.webcitation.org/6cPIpm4Vl (Archived at: http://www.webcitation.org/6cPIpm4Vl). July 11, 2012
- 19. Lavrakas P. J., Tompson T. N., Benford R., Fleury C. Within‐unit respondent selection errors in landline RDD surveys. 2012 RC33 Conference, Sydney, Australia, 2012.
- 20. Australian Bureau of Statistics (ABS) Australian Bureau of Statistics, 2010 Estimated Residential PopulationCustomized Report, 2011. Melbourne: Australian Bureau of Statistics; 2011. [Google Scholar]
- 21. Australian Communications and Media Authority . Communications Report, 2010–11 series. Report 2—Converging Communications Channels: Preferences and Behaviours of Australian Communications Users Australian Communications and Media Authority: Melbourne, VIC (Aust); 2011. [Google Scholar]
- 22. Ferris J., Wynne H. J. The Canadian Problem Gambling Index. Ottawa: Canadian Centre on Substance Abuse; 2001. [Google Scholar]
- 23. Jackson A. C., Wynne H., Dowling N. A., Tomnay J. E., Thomas S. A. Using the CPGI to determine problem gambling prevalence in Australia: measurement issues. Int J Mental Health Addict 2010; 8: 570–82. [Google Scholar]
- 24. National Health and Medical Research Council Australian Guidelines to Reduce Health Risks from Drinking Alcohol. Canberra: Commonwealth of Australia; 2009. [Google Scholar]
- 25. Kessler R. C., Andrews G., Colpe L. J., Hiripi E., Mroczek D. K., Normand S. L. et al. Short screening scales to monitor population prevalences and trends in nonspecific psychological distress. Psychol Med 2002; 32: 959–76. [DOI] [PubMed] [Google Scholar]
- 26. Andrews G., Slade T. Interpreting scores on the Kessler psychological distress scale (K10). Aust NZ J Public Health 2001; 25: 494–7. [DOI] [PubMed] [Google Scholar]
- 27. Cairney J., Veldhuizen S., Wade T., Kurdyak P., Streiner D. L. Evaluation of 2 measures of psychological distress as screeners for depression in the general population. Canadian Journal of Psychiatry 2007; 52: 111–120. [DOI] [PubMed] [Google Scholar]
- 28. Furukawa T. A., Kessler R. C., Slade T., Andrews G. The performance of the K6 and K10 screening scales for psychological distress in the Australian National Survey of Mental Health and Well‐Being. Psychol Med 2003; 33: 357–62. [DOI] [PubMed] [Google Scholar]
- 29. Kessler R. C., Barker P. R., Colpe L. J., Epstein J. F., Gfroerer J. C., Hiripi E., et al. Screening for serious mental illness in the general population. Arch Gen Psychiatry 2003; 60: 184–89. [DOI] [PubMed] [Google Scholar]
- 30. Bowling A. Just one question: if one question works, why ask several? J Epidemiol Community Health 2005; 59: 342–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Australian Bureau of Statistics (ABS) Table Builder: 2011 Census of Population and Housing. Non‐school qualification (QALLP). Melbourne: Australian Bureau of Statistics; 2011. [Google Scholar]
- 32. Australian Bureau of Statistics (ABS) Table Builder: 2011 Census of Population and Housing (BPLP). Melbourne: Australian Bureau of Statistics; 2011. [Google Scholar]
- 33. Australian Bureau of Statistics (ABS) Table Builder: 2011 Census of Population and Housing. Non‐school qualification (GCCSA). Melbourne: Australian Bureau of Statistics; 2011. [Google Scholar]
- 34. Best J. First‐stage weights for overlapping dual frame telephone surveys. Presented at the AAPOR 65th Annual Conference, Chicago, IL, 2010.
- 35. Little R. J., Rubin D. B. Statistical Analysis With Missing Data. New York: John Wiley and Sons; 1987. [Google Scholar]
- 36. Productivity Commission . Gambling. Report no. 50. Australia: Productivity Commission; 2010. May 13–16, 2010 [Google Scholar]
- 37. Dowling N., Brown M. Commonalities in the psychological factors associated with problem gambling and Internet dependence. Cyberpsychol Behav Soc Netw 2010; 13: 437–41. [DOI] [PubMed] [Google Scholar]
- 38. Scholes‐Balog K. E., Hemphill S. A., Dowling N. A., Toumbourou J. W. A prospective study of adolescent risk and protective factors for problem gambling among young adults. J Adolesc 2014; 37: 215–24. [DOI] [PubMed] [Google Scholar]
- 39. Dowling N. A. The cognitive‐behavioural treatment of female problem gambling In: Richard D. C. S., Blaszczynski A., Nower L., editors. The Wiley‐Blackwell Handbook of Disordered Gambling. West Sussex: John Wiley and Sons; 2013, pp. 225–50. [Google Scholar]