Abstract
Background
In Britain, unprecedented restrictions on daily life associated with the Covid-19 pandemic included the suspension of professional sports events during the initial ‘lockdown’. This provides opportunities to observe changes in sports bettors’ behaviour when their primary form of activity is removed and assess the impact of Covid-19 related circumstances upon gambling.
Methods
In July 2020, we conducted an online cross-sectional survey of people who bet regularly (at least monthly) on sports before Covid-19 (n = 3866). Bi-variate analyses compared changes in gambling behaviours before and during the initial lockdown. Multi-variate logistic regression models explored associations between problem gambling (men) and moderate risk or problem gambling (MRPG) (women) with changes in Covid-19 related circumstances and changing gambling behaviours during Britain’s initial ‘lockdown’ (March-June 2020).
Results
29.8% of male sports bettors and 33.4% of female sports bettors stopped gambling altogether during the initial Covid-19 lockdown, though 17.3% of men and 16.5% of women started a new form of gambling during lockdown. Among men, adjusted odds ratios of problem gambling were higher among those starting a new gambling activity during lockdown (OR = 2.50 [95% CI 1.38–4.53]). Among women, adjusted odds ratios of MRPG were higher among those whose frequency of gambling on any activity increased during lockdown (OR = 4.21 [1.99–8.92] and among those shielding for health reasons. Poorer wellbeing was associated with problem gambling for men and MRPG for women.
Conclusions
Those changing gambling behaviours during the initial Covid-19 lockdown (e.g. increasing gambling frequency or starting a new gambling activity) are potentially vulnerable to gambling harms.
Keywords: Gambling, Covid-19, Sports betting, Gambling harms
1. Introduction
The outbreak of the Covid-19 pandemic in early 2020 precipitated unprecedented global restrictions on daily life. In Britain, a population-wide ‘lockdown’ began on March 23rd 2020, with the closure of educational establishments, non-essential workplaces, shops, leisure and recreational facilities. This included all gambling venues including bookmakers, casinos, arcades and bingo halls. As elsewhere, almost all professional sports events (including horse racing) were suspended. These lockdown conditions remained in place until mid-June when some restrictions began to ease: British horseracing and championship snooker were reintroduced in Britain on 1st June; English Premier League football on 17th June and 20th June for English championship football, with other sports following thereafter (the Scottish Football Premier League resumed in August).
When lockdown was first imposed, concerns were raised about the impact of Covid-19 upon gambling behaviours. The British All Party Parliamentary Group for Gambling Harms suggested lockdown might encourage more people to gamble and increase vulnerability to harms (APPG, 2020). The British regulator, the Gambling Commission (GC), identified transitions to different forms of gambling, especially faster (online) forms of gambling, as an important Covid-19 related risk (GC, 2020).
These concerns were echoed elsewhere and some governments, including Spain and Sweden, banned gambling advertising and set strict loss limits during the early phase of the pandemic. Since then an emerging evidence base has suggested variable impact of Covid-19 on gambling behaviours, with studies from Sweden and Nordic countries noting an overall decline in gambling participation (Håkansson, 2020, Lindner et al., 2020, Auer et al., 2020), but some highlighting changing gambling behaviours being associated with elevated risk of harms (Håkansson, 2020) and emphasising associations with broader personal, social, and economic circumstances with harmful gambling (Price, 2020).
The suspension of most sports events in Britain during the initial lockdown provides an unprecedented opportunity to explore if and how the gambling behaviours of sports bettors change when opportunities for sports betting are restricted. In addition, we can assess how broader personal, social and economic conditions relating to Covid-19 impacts behaviour. The analysis presented here is part of a broader study of the impact of Covid-19 on young people and regular sports bettors, which includes qualitative interviews to explore experiences and behaviours in depth and analyses of marketing strategies and spend to understand industry’s reactions (Hunt et al, 2020). Here we report analyses of a bespoke survey which investigated the extent to which regular sports bettors substituted one form of gambling (sports betting) for others during the initial lockdown and the extent to which changes in personal, social and economic circumstances relate to changing gambling behaviours and the experience of problem gambling. We aim to provide evidence to inform policy guidance in the continuing cycles of national lockdowns experienced in Great Britain and elsewhere.
Data reported here examine any differences in gambling behaviours in the three months prior to the initial Covid-19 lockdown (December 2019-February 2020) and during the initial lockdown in Britain (mid-March-mid June 2020). Our objectives were to a) understand any changes that regular sports bettors reported in their gambling behaviours during this lockdown period and b) conduct analysis to explore whether such changes were related to the experience of gambling harms.
Furthermore, as gambling is a maladaptive coping behaviour for some (Blaszczynski & Nower, 2002, Jauregui et al., 2017) and concerns about increases in maladaptive coping behaviours during Covid-19 (Ogueji et al., 2021; Piece et al, 2020), we hypothesised that changes in social and economic circumstances during the initial Covid-19 lockdown (e.g. becoming unemployed/furloughed, changes in financial or living circumstances) would be associated with the experience of gambling-related harms. As existing evidence suggests that life-stressors are also associated with heightened experience of gambling harms (Blaszczynski & Nower, 2002, Wang et al., 2020), we also hypothesised that changes in boredom, stress, conflict and free time would be associated with gambling-related harms (hypotheses were pre-registered: Hunt, Critchlow & Wardle, 2020).
2. Methods
2.1. Design
Data come from work package 1b of ‘The Betting and Gaming Covid-19 Impact Study’ (Hunt et al, 2020) and represent the first wave of longitudinal assessment of how the pandemic has impacted on ‘regular’ sports betters. A cohort of regular sports bettors from Britain completed an online survey between 6th and 19th July 2020 (n = 3,866) with follow-up waves scheduled for November/December 2020 and March/April 2021. This cross-sectional analysis uses data from the first survey on gambling behaviour, attitudes and circumstances during the initial Covid-19 lockdown (at the time of writing, follow-up data were not yet available). The survey was conducted by YouGov who recruited a cohort of sport bettors from their existing non-probabilistic panel of one million people. YouGov’s panel contains information on members, allowing us to target specific panellists (Twyman, 2008, Kennedy et al., 2016). All participants who completed the survey received YouGov points (equivalent to between £0.50 and £1.50) in return for participation. Remuneration is determined by YouGov, with higher points offered to under-represented groups.
3. Sample and eligibility
This study focuses on ‘regular’ sports bettors in Britain. To be eligible, panellists had to be aged 18 or over, living in Britain, and to have previously reported betting on sports (including horse racing) at least monthly (hereafter ‘regularly’). In total, YouGov identified 19,237 eligible panellists who were sent e-mail invitations. Of those who accessed the initial information and consent page using the e-mail link, 78% completed the survey. From this sample, only those who reported betting on sporting events (including horse racing) in the three months prior to the initial Covid-19 lockdown were included in the study. Survey data were weighted to reflect the age, sex, geographic and betting profile of the broader sample of regular sports bettors within the YouGov panel.
4. Measures
The survey covered topics relating to gambling behaviour before and during the initial Covid-19 lockdown, gambling attitudes, awareness of gambling marketing, experiences of gambling harm, and Covid-19 health and lifestyle experiences. It was devised by HW with input from research team members. Full details of the survey measures are available elsewhere (Hunt, Critchlow & Wardle, 2020); here we a provide a summary of key measures relevant to the current analyses.
4.1. Gambling behaviours before, and during, the initial Covid-19 pandemic
Regular gamblers take part in various activities and thus participants were asked to report engagement in 23 different gambling activities, covering all main subsectors (e.g. sports betting, casino and poker, gaming, lotteries etc) and formats (e.g. land-based, online etc) (see Table 1a /1b for full details). These questions were asked in two blocks, relating to the three months immediately prior to the initial Covid-19 lockdown (defined as December 2019 to February 2020) and then during this lockdown (defined as mid-March to mid-June 2020).
Table 1a.
Participation in different forms of gambling, pre Covid-19 and during initial lockdown, among male sports bettors (N = 3084).
| Covid-19-related experiences |
|
|
|||||||
|---|---|---|---|---|---|---|---|---|---|
| Pre Covid-19 lockdown | During initial Covid-19 lockdown | Starting rate (% of those who did not do this prior to lockdown who started during lockdown) | Stopping rate (% of those who did this prior to lockdown who stopped during lockdown) | p-value | |||||
| n | %(95% CI) | n | %(95% CI) | n | %(95% CI) | n | %(95% CI) | ||
| Key sports betting activities | |||||||||
| Online betting on horse/dog races | 1396 | 43.2 (41.4–45.1) | 373 | 11.6 (10.5–12.9) | 45 | 2.6 (1.8–3.4) | 1068 | 76.5 (74.3–78.8) | p < 0.001 |
| Online sports betting | 2327 | 78.7 (77.3–80.1) | 464 | 16.1 (14.8–17.6) | 20 | 2.9 (1.7–4.1) | 1883 | 80.3 (81.5–84.5) | p < 0.001 |
| Betting on horse/dog races in a bookmakers | 782 | 23.0 (22.0–25.0) | 20 | 0.6 (0.4–1.0) | 4 | 0.3 (0.1–0.5) | 766 | 98.1 (97.1–99.1) | p < 0.001 |
| Betting on sports in a bookmakers | 806 | 26.7 (25.1–28.4) | 27 | 1.1 (0.7–1.6) | 11 | 0.6 (0.3–0.9) | 790 | 97.5 (96.4–98.6) | p < 0.001 |
| Other gambling activities still available during lockdown | |||||||||
| Lottery | 1975 | 62.1 (60.3–63.9) | 1581 | 48.7 (46.8–50.6) | 59 | 5.9 (4.5–7.3) | 453 | 25.2 (23,4–27.2) | p < 0.001 |
| Scratchcards | 769 | 26.6 (25.0–28.3) | 435 | 14.7 (13.4–16.0) | 31 | 1.5 (1.0–2.0) | 365 | 49.0 (45.5–52.5) | p < 0.001 |
| Online betting on esports | 71 | 2.8 (2.2–3.6) | 90 | 3.4 (2.7–4.2) | 69 | 2.7 (2.1–3.3) | 50 | 71.5 (61.0–82.0) | p = 0.10 |
| Online betting on virtual sports/races | 82 | 3.0 (2.4–3.8) | 119 | 4.2 (3.6–5.2) | 94 | 3.5 (2.8–4.2) | 57 | 69.7 (69.8–79.6) | p < 0.01 |
| Online betting on other events | 252 | 8.8 (7.6–10.0) | 85 | 2.9 (2.3–3.6) | 42 | 1.5 (1.1–1.9) | 209 | 83.3 (77.2–89.4) | p < 0.001 |
| Online slot games | 314 | 10.9 (9.7–12.1) | 285 | 9.9 (8.8–11.2) | 62 | 2.6 (2.0–3.2) | 91 | 30.2 (24.0–36.4) | p < 0.05 |
| Online casino games (excluding poker) | 322 | 11.8 (10.6–13.1) | 234 | 8.5 (7.5–9.7) | 53 | 2.4 (1.8–3.0) | 146 | 45.6 (40.2 51.0) | p < 0.001 |
| Online bingo | 137 | 4.8 (4.0–5.7) | 114 | 4.1 (3.4–4.9) | 37 | 1.4 (1.0–1.8) | 60 | 43.0 (34.7–51.3) | p < 0.05 |
| Online poker | 198 | 7.0 (6.0–8.0) | 180 | 6.7 (5.8–7.8) | 71 | 3.1 (2.5–3.7) | 89 | 46.0 (39.1–52.9) | p = 0.18 |
| Private betting with friends/family | 217 | 7.9 (6.9–9.1) | 88 | 3.2 (2.5–3.9) | 26 | 1.0 (0.6–1.4) | 155 | 71.5 (65.5–77.5) | p < 0.001 |
| Other gambling activities mainly unavailable during lockdown | |||||||||
| Betting on esports in a bookmakers | 62 | 2.4 (1.8–3.2) | 9 | 0.4 (0.2–0.8) | 4 | 0.1 (0.05–0.37) | 57 | 89.8 (82.3–97.3) | p < 0.001 |
| Betting on virtual bets/races in a bookmakers | 90 | 3.3 (2.7–4.1) | 14 | 0.5 (0.3–0.9) | 11 | 0.4 (0.2–0.6) | 87 | 96.1 (88.8–98.7) | p < 0.001 |
| Betting on other events in a bookmakers | 142 | 4.7 (4.0–5.6) | 6 | 0.2 (0.0–0.5) | 5 | 0.2 (0.08–0.48) | 141 | 99.4 (96.6–99.9) | p < 0.001 |
| Slot machines | 259 | 8.9 (7.9–10.1) | 22 | 0.8 (0.5–1.3) | 7 | 0.2 (0.11–0.51) | 244 | 93.3 (90.3–96.3) | p < 0.001 |
| Machines in a bookmakers | 175 | 5.7 (4.9–6.6) | 3 | 0.1 (0.0–0.5) | 2 | 0.1 (0.03–0.51) | 174 | 99.5 (97.2–99.9) | p < 0.001 |
| Football pools | 252 | 8.6 (7.6–9.8) | 84 | 2.5 (2.0–3.2) | 8 | 0.3 (0.1–0.5) | 176 | 74.2 (68.9–79.5) | p < 0.001 |
| Casino games at a casino | 191 | 7.3 (6.3–8,4) | 3 | 0.1 (0.04–0.6) | 2 | 0.1 (0.03–0.55) | 190 | 99.6 (97.8–99.9) | p < 0.001 |
| Playing poker in pub/club/venue | 74 | 2.6 (2.1–3.3) | 4 | 0.2(0.06–0.42) | 1 | 0.03 (0.01–0.18) | 71 | 95.2 (86.5–98.4) | p < 0.001 |
| Bingo in hall/club/other venue | 105 | 3.7 (3.0–4.5) | 2 | 0.1 (0.03–0.57) | 2 | 0.1 (0.03–0.53) | 105 | 100 (N/A) | p < 0.001 |
*Where starting and stopping rates are close to 0 or 100, Wilson’s Confidence Intervals have been calculated.
Table 1b.
Participation in different forms of gambling, pre Covid-19 and during initial lockdown, among female sports bettors (N = 782).
| Covid-19-related experiences |
|
|
|||||||
|---|---|---|---|---|---|---|---|---|---|
| Pre Covid-19 lockdown | During initial Covid-19 lockdown | Starting rate (% of those who did not do this prior to lockdown who started during lockdown) | Stopping rate (% of those did this prior to Covid-19 who stopped during lockdown) | p-values | |||||
| n | % (95% CI) | n | % (95% CI) | n | % (95% CI) | n | % (95% CI) | ||
| Key sports betting activities | |||||||||
| Online betting on horse/dog races | 415 | 50.5 (46.7–54.2) | 80 | 9.2 (7.4–11.5) | 5 | 1.5 (0.3–2.7) | 340 | 83.2 (79.6–86.8) | p < 0.001 |
| Online sports betting | 443 | 61.4 (57.8–64.5) | 70 | 9.9 (7.7–12.6) | 3 | 0.8 (0.27–2.5) | 376 | 84.3 (80.9–87.7) | p < 0.001 |
| Betting on horse/dog races in a bookmakers | 191 | 22.0 (19.0–24.8) | 6 | 0.6 (0.3–1.4) | 2 | 0.3 (0.08–1.12) | 187 | 98.1 (95.2–99.3) | p < 0.001 |
| Betting on sports in a bookmakers | 126 | 16.1 (13.5–19.0) | 3 | 0.5 (0.2–1.6) | 1 | 0.1 (0.02–0.79) | 124 | 97.6 (91.7–99.4) | p < 0.001 |
| Other gambling activities still available during lockdown | |||||||||
| Lottery | 523 | 65.2 (61.5–68.8) | 380 | 45.8 (42.1–49.6) | 18 | 6.5 (3.5–9.5) | 161 | 33.2 (29.2–37.2) | p < 0.001 |
| Scratchcards | 316 | 42.7 (39.0–46.5) | 159 | 21.1 (18.1–24.4) | 11 | 2.7 (1.2–4.2) | 168 | 54.3 (48.8–59.8) | p < 0.001 |
| Online betting on esports | 16 | 2.4 (1.5–4.1) | 11 | 1.8 (1.0–3.2) | 8 | 1.4 (0.6–2.3) | 13 | 82.3 (54.8–94.7) | p = 0.36 |
| Online betting on virtual sports/races | 30 | 4.3 (2.9–6.0) | 55 | 8.7 (6.6–11.4) | 39 | 6.8 (5.0–8.6) | 14 | 49.0 (31.1–66.9) | p < 0.01 |
| Online betting on other events | 56 | 7.8 (5.9–10.1) | 17 | 2.3 (1.4–3.8) | 88 | 0.9 (0.2–1.6) | 47 | 81.2 (69.7–92.7) | p < 0.001 |
| Online slot games | 93 | 12.6 (10.3–15.4) | 67 | 8.2 (6.4–10.4) | 18 | 2.7 (1.5–3.9) | 44 | 53.9 (43.8–64.0) | p < 0.001 |
| Online casino games (excluding poker) | 81 | 8.5 (9.3–14.3) | 49 | 6.8 (5.1-0.9.0) | 13 | 2.2 (1.1–33) | 45 | 57.8 (47.0–68.6) | p < 0.001 |
| Online bingo | 100 | 13.5 (11.1–16.4) | 74 | 9.3 (7.3–11.6)) | 21 | 3.4 (2.0–4.8) | 47 | 53.4 (43.6–63.2) | p < 0.01 |
| Online poker | 27 | 4.1 (2.8–6.1) | 19 | 3.0 (1.9–4.8) | 6 | 1.1 (0.4–1.8) | 14 | 52.8 (34.0–71.6) | p = 0.12 |
| Private betting with friends/family | 42 | 6.7 (4.8–9.1) | 15 | 2.3 (1.3–4.0) | 9 | 1.5 (0.6–2.4) | 36 | 86.1 (75.6–96.6) | p < 0.001 |
| Other gambling activities mainly unavailable during lockdown | |||||||||
| Betting on esports in a bookmakers | 17 | 2.7 (1.6–4.4) | 2 | 0.4 (1.9–4.8) | 1 | 0.2 (0.04–1.1) | 16 | 93.4 (66.9–99.0) | p < 0.001 |
| Betting on virtual bets/races in a bookmakers | 19 | 2.7 (1.6–4.4) | 5 | 0.8 (0.3–2.0) | 3 | 0.5 (0.18–1.6) | 17 | 88.0 (62.2–97.0) | p < 0.01 |
| Betting on other events in a bookmakers | 44 | 5.2 (3.9–7.1) | 4 | 0.7 (0.3–1.9) | 3 | 0.6 (0.19–1.7) | 43 | 96.6 (82.4–99.4) | p < 0.001 |
| Slot machines | 67 | 8.6 (6.8–10.9) | 5 | 0.6 (0.3–1.6) | 3 | 0.3 (0.1–0.9) | 65 | 95.8 (85.6–98.9) | p < 0.001 |
| Machines in a bookmakers | 39 | 5.0 (3.6–6.8) | 1 | 0.2 (0.0–1.2) | – | – | 38 | 96.4 (81.5–99.4) | p < 0.001 |
| Football pools | 41 | 5.3 (3.8–7.3) | 12 | 1.6 (0.9–3.0) | 2 | 0.3 (0.09–1.2) | 31 | 75.5 (62.3–88.7) | p < 0.001 |
| Casino games at a casino | 44 | 6.4 (4.8–8.7) | 1 | 0.2 (0.0–1.4) | – | – | 43 | 96.8 (83.4–99.5) | p < 0.001 |
| Playing poker in pub/club/venue | 13 | 1.7 (1.1–3.0) | 1 | 0.1 (0.0–1.4) | – | – | 12 | 95.2 (72.6–99.3) | p < 0.001 |
| Bingo in hall/club/other venue | 90 | 12.2 (9.8–15.0) | 2 | 0.2 (0.0–0.8) | 2 | 0.2 (0.06–0.86) | 90 | 100 | p < 0.001 |
*Where starting and stopping rates are close to 0 or 100, Wilson’s Confidence Intervals have been calculated.
For each gambling activity reported, within each timeframe (pre/during lockdown), we collected three further metrics: frequency of participation recorded on an eight-point scale (1 = Several times a day-8 = Never); expenditure on each activity (estimated to nearest £GBP), and, with the exception of lotteries and scratchcards, the amount of time spent gambling on an eight-point scale (1 = <30 min per day to 8 = 8 + hours per day). For betting activities, participants were asked to estimate how long they spent planning and placing their bet.
Three main variables were derived from the gambling participation questions. First, we computed ‘starting and stopping’ rates for each activity. The ‘starting’ rate captures data on activities in which a participant reported no participation during the three months prior to the initial lockdown, but reported some participation during this lockdown, whereas the ‘stopping rate’ is the inverse. From these variables, we also identified participants who: started any new form of gambling during the lockdown; stopped completely; and continued to gamble on some or all of their pre-lockdown activities. Second, we computed total expenditure on all gambling activities before and during the lockdown, and categorised participants as ‘spent more during lockdown’, ‘spent less during lockdown’, or ‘spent the about the same’ (i.e., spend deviated ≤£10 compared to pre-lockdown). Finally, for gambling frequency, participants were binary coded based on those who reported increasing frequency on any gambling activity during lockdown versus those whose frequency decreased or stayed the same.
4.2. Problem gambling
Participants completed the 9-item Problem Gambling Severity Index (PGSI) (Ferris & Wynne, 2001) scored on a four-point scale (0 = Never to 3 = Almost always). PGSI scores range from 0 to 27 where a score of 0 indicates non-problem gambling or non-gambling; 1–2 is low risk gambling; 3–7 is moderate risk gambling; and a score of 8 or more is indicative of problem gambling. In our sample, the composite score had strong internal consistency (Cronbach's α = 0.91). Although PGSI usually uses a 12-month reference period, we purposively used a three month reference period to match the initial Covid-19 pandemic timeframe and reported involvement in gambling activities during the three month lockdown. Previous research has shown the utility of using a shorter PGSI timeframe when assessing the impact of interventions (Abbott et al., 2012, Kushnir et al., 2018).
4.3. Changes in circumstances due to initial Covid-19 lockdown
Several measures were adapted from the Centre for Longitudinal Studies Covid-19 Surveys (UCL, 2020) to capture changes in personal and financial circumstances resulting from the initial lockdown.
Concerning personal circumstances, participants reported changes during lockdown in the amount of spare time, boredom, stress and conflict with others they experienced. Responses were coded as more than before; same as before; less than before.
Questions asked about participants’ financial circumstances during the initial lockdown on a five-point scale ranging from “much better off” to “much worse off” (because of base sizes collapsed to “about the same”, “worse off”, “better off”). Changes in job status were captured, specifically if participants had lost their job or were furloughed (i.e., undertaken government-supported leave of absence) during lockdown. Questions were asked about changes in living arrangements during lockdown. Due to base sizes for categories, participants were coded as ‘did not cohabit pre/post lockdown’, ‘cohabited with partner pre/post lockdown’ and ‘cohabitation status with spouse/partner changed during lockdown’. Finally, participants were asked whether they had been asked to ‘shield’ during the lockdown (Yes/No); ‘shielding’ was the official UK Government term used to advise those with underlying health conditions/vulnerabilities to stay home and avoid contact with others. It differs from self-isolating, which is what people are asked to do for a time-limited period if displaying symptoms or after contact with someone with the virus.
4.4. Wellbeing during the lockdown
Wellbeing was measured using the Short Warwick and Edinburgh Mental Wellbeing Scale (WEMWEBS) (Tennant et al., 2007) . Scores range between 7 and 35, with those scoring between 7 and 20 identified as having lower wellbeing and those scoring 21 + having higher wellbeing. The composite score had acceptable internal consistency in our sample (Cronbach's α = 0.85).
4.5. Demographics
Information on age (coded: 18–34, 35–44, or 55 + years), gender (men, women), area level deprivation (based on postcode, divided into quintiles), and educational attainment (described in results) were provided by YouGov from existing information on panellists.
5. Ethics
Approval for the study was granted by the University of Stirling’s General University Ethics Panel (ref: GUEP (19 20) 930).
5.1. Statistical analysis
Data were obtained from 4016 participants. We excluded 89 participants who completed the survey in ≤ 4/5 minutes, depending on their level of gambling involvement, who were judged to have not given proper consideration to questions. Timing thresholds were established by reviewing the first 250 responses. A further 61 participants were excluded on the basis of implausible answers about gambling (e.g. saying they gambled ‘bi-weekly’ at closed venues such as bookmakers during lockdown). Hence, 3866 participants (3084 men, 782 women) were included in the main analyses. Missing data within this final sample were minimal and are excluded unless explicitly stated.
As gambling behaviour is highly gendered, with patterns of engagement and the experience of harms differing for men and women (Wardle, 2015, McCarthy et al., 2018, McCarthy et al., 2019, Williams et al., 2021), we have followed recommendations (McCarthy et al., 2018, Forrest and McHale, 2020) to conduct analyses separately for men and women.
Frequencies examined the proportion of regular sports betters who participated in each of the 23 gambling activities before and/or during the initial Covid-19 lockdown and the ‘stopping’ and ‘starting’ rates for each activity. McNemar tests for paired nominal data were used to assess changes in participation prior to and during lockdown. Further bivariate analyses examined the extent to which personal and financial circumstances during the lockdown (e.g. furlough, cohabitation change) varied by PGSI status. A Walds-F test examined the extent to which problem gambling status varied by each independent variable presented in Table 2a, Table 2b . All bi-variate analyses were performed using the complex survey module in SPSS v25.
Table 2a.
Problem Gambling Severity Index Score, by Covid-19-related experiences among men.
| Covid-19-related experiences |
Problem gambling status (Problem Gambling Severity Index Score) |
||||||||
|---|---|---|---|---|---|---|---|---|---|
| Non-problem gambling (0) (n = 2096) |
Low risk gambling (1–2)(n = 564) |
Moderate risk gambling (3–7) (n = 289) |
Problem gambling (8 + )(n = 135) |
p-values* |
|||||
| Men | n | % (95% CI) | n | % (95% CI) | n | % (95% CI) | n | % (95% CI) | |
| All male regular sports bettors | 64.7 | 19.3 | 10.7 | 5.4 | |||||
| Employment changes | |||||||||
| Became unemployed | 79 | 62.8 (53.1–71.6) | 18 | 14.3 (9.0–21.8) | 21 | 20.4 (13.3–30.1) | 3 | 2.5 (0.8–7.5) | p < 0.05 |
| Furloughed | 142 | 57.6 (50.5–64.4) | 43 | 21.3 (16.0–27.8) | 20 | 10.4 (6.6–16.0) | 22 | 10.8 (6.9–16.4) | p < 0.05 |
| Change in finances | |||||||||
| Worse off | 489 | 57.7 (54.0–61.3) | 164 | 21.4 (18.5–24.6) | 90 | 12.3 (10.0–15.5) | 57 | 8.6 (6.6–11.1) | p < 0.01 |
| About the same | 1057 | 68.8 (66.1–71.3) | 245 | 17.6 (15.6–19.8) | 120 | 9. 4 (7.8–11.3) | 55 | 4.3 (3.2–5.6) | |
| Better off | 550 | 64.2 (60.5–67.8) | 155 | 20.3 (17.5–23.5) | 79 | 11.3 (9.8–14.1) | 23 | 4.2 (2.7–6.4) | |
| Change in stress levels | |||||||||
| More than before | 620 | 58.1 (54.8–61.3) | 206 | 21.1 (18.5–23.9) | 125 | 13.2 (11.0–15.6) | 67 | 7.7 (6.0–9.8) | p < 0.01 |
| About the same | 1212 | 70.2 (67.8–72.6) | 282 | 18.0 (16.1–20.1) | 122 | 8.7 (7.2–10.4) | 42 | 3.1 (2.2–4.2) | |
| Less than before | 264 | 60.0 (54.7–65.1) | 76 | 19.7 (15.8–24.3) | 42 | 12.1 (8.8–16.2) | 20 | 8.2 (5.5–12.1) | |
| Change in levels of boredom | |||||||||
| More than before | 860 | 59.6 (56.7–62.4) | 280 | 21.4 (19.1–23.9) | 155 | 12.6 (10.7–14.8) | 74 | 6.4 (5.0–8.0) | p = 0.313 |
| About the same | 1109 | 70.8 (68.2–73.2) | 237 | 16.6 (14.7–18.7) | 112 | 8.3 (6.9–10.0) | 51 | 4.3 (3.2–5.8) | |
| Less than before | 127 | 58.0 (50.7–65.0 | 44 | 22.9 (17.5–29.4) | 22 | 13.5 (8.8–20.2) | 10 | 5.6 (2.9–10.4) | |
| Change in amount of free time | p = 0.058 | ||||||||
| More than before | 929 | 60.2 (57.4–62.9) | 288 | 21.1 (18.9–23.5) | 155 | 12.3 (10.4–14.4) | 76 | 6.4 (5.1–8.1) | |
| About the same | 1029 | 70.3 (67.6–72.6) | 232 | 16.9 (15.0–19.1) | 114 | 8.9 (7.4–10.7) | 44 | 3.8 (2.8–5.2) | |
| Less than before | 138 | 61.8 (54.8–68.4) | 44 | 20.5 (15.5–26.7) | 20 | 10.4 (6.6–15.9) | 15 | 7.3 (4.4–11.9) | |
| Change in amount of conflict with others | |||||||||
| More than before | 204 | 46.4 (41.4–51.5) | 107 | 25.5 (21.3–30.1) | 64 | 17.4 (13.6–21.9) | 41 | 10.8 (7.9–14.5) | p < 0.05 |
| About the same | 1527 | 70.5 (68.3–72.6) | 342 | 17.5 (15.8–19,4) | 151 | 8.1 (6.8–9.5) | 67 | 3.9 (3.1–5.0) | |
| Less than before | 365 | 58.6 (54.2–62.8) | 115 | 20.5 (17.2–24.3) | 74 | 14.7 (11.7–18.3) | 27 | 6.2 (4.2–9.2) | |
| Whether shielding | |||||||||
| Yes | 175 | 60.3 (54.0–66.3) | 53 | 19.4 (14.9–24.9) | 31 | 11.8 (8.3–16.4) | 17 | 8.5 (5.2–13.6) | p < 0.05 |
| No | 1907 | 65.3 (63.4–67.2) | 506 | 19.3 (17.8–20.9) | 255 | 10.6 (9.3–12.0) | 112 | 4.8 (4.0–5.8) | |
| Change in cohabitation status with spouse | |||||||||
| Did not cohabit pre/post lockdown | 631 | 60.8 (57.3–64.2) | 177 | 19.6 (17.0–22.6) | 105 | 13.0 (10.7–15.8) | 49 | 6.6 (4.9–8.8) | p = 0.054 |
| Cohabitation status with partner changed during lockdown | 25 | 45.7 (32.1–60.0) | 14 | 27.1 (16.2–41.6) | 7 | 12.5 (5.9–24.6) | 8 | 14.7 (7.3–27.3) | |
| Cohabited with partner/spouse pre and post lockdown | 1440 | 67.3 (65.1–69.4) | 373 | 18.8 (17.1–20.7)) | 177 | 9.4 (8.1–10.9) | 78 | 4.4 (3.5–5.6) | |
*P-values compare rates of problem gambling within each variable. For employment changes, this compares becoming unemployed with not becoming unemployed or being furloughed with not being furloughed. Data not shown for each comparator group.
Table 2b.
Problem Gambling Severity Index Score, by Covid-19-related experiences among women.
| Covid-19-related experiences | Problem gambling status (Problem Gambling Severity Index Score) | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Non-problem gambling (0) (n = 574) | Low risk gambling (1–2)(n = 124) | Moderate risk gambling (3–7) (n = 59) | Problem gambling (8 + )(n = 25) | p-values* | |||||
| Women | n | % (95% CI) | n | % (95% CI) | n | % (95% CI) | n | % (95% CI) | |
| All female regular sports bettors | 71.2 (67.6–74.6) | 16.6 (14.0–19.7) | 8.3 (6.4–10.7) | 3.9 (2.5–5.9) | |||||
| Employment changes | |||||||||
| Became unemployed | 23 | 72.4 (53.8–85.9) | 2 | 4.9 (1.2–17.8) | 7 | 22.7 (10.4–42.6) | 0 | – | n/a |
| Furloughed | 54 | 59.9 (47.6–71.1) | 14 | 19.9 (11.6–32.0) | 10 | 14.5 (7.8–25.5) | 3 | 5.6 (1.6–17.9) | p < 0.05 |
| Change in finances | |||||||||
| Worse off | 167 | 63.7 (56.7–69.8) | 45 | 19.0 (14.3–25.0) | 26 | 12.4 (8.4–17.9) | 12 | 4.9 (2.8–8.6) | p = 0.30 |
| About the same | 291 | 75.2 (70.0–79.7) | 54 | 14.7 (11.2–18.9) | 21 | 5.9 (3.8–9.1) | 11 | 4.3 (2.2–8.1) | |
| Better off | 116 | 74.0 (65.7–80.8) | 25 | 17.2 (11.5–25.0) | 12 | 7.4 (4.2–12.9) | 2 | 1.4 (0.3–5.8) | |
| Change in stress levels | |||||||||
| More than before | 249 | 67.4 (61.9–72.4) | 64 | 18.8 (14.8–23.6) | 29 | 8.9 (6.1–12.8) | 14 | 4.9 (2.8–8.6) | p = 0.65 |
| About the same | 237 | 76.3 (70.6–81.2) | 39 | 13.0 (9.4–17.6) | 20 | 7.5 (4.8–11.5) | 8 | 3.3 (1.5–6.8) | |
| Less than before | 88 | 70.7 (61.2–78.7) | 21 | 18.4 (12.0–27.4) | 10 | 8.5 (4.5–15.4) | 3 | 2.3 (0.8–7.1) | |
| Change in levels of boredom | |||||||||
| More than before | 263 | 69.2 (63.8–74.1) | 68 | 18.5 (14.6–23.2) | 31 | 9.0 (6.3–12.8) | 11 | 3.3 (1.7–6.3) | p = 0.457 |
| About the same | 252 | 75.2 (69.9–80.0) | 42 | 14.0 (10.4–18.7) | 21 | 7.5 (4.8–11.3) | 10 | 3.3 (1.8–6.1) | |
| Less than before | 53 | 66.9 (54.7–77.1) | 14 | 17.0 (9.9–27.5) | 7 | 8.1 (3.8–16.6) | 4 | 8.0 (2.9–20.6) | |
| Change in amount of free time | |||||||||
| More than before | 300 | 70.6 (65.4–75.3) | 61 | 16.8 (13.0–21.3) | 31 | 8.7 (6.1–12.3) | 15 | 4.0 (2.3–6.8) | n/a |
| About the same | 226 | 71.8 (66.1–76.9) | 52 | 16.0 (12.3–20.6) | 21 | 7.4 (4.9–11.2) | 10 | 4.7 (2.4–9.1) | |
| Less than before | 48 | 72.7 (60.0–82.5) | 11 | 17.9 (10.0–29.9) | 6 | 9.4 (4.2–19.7) | 0 | – | n/a |
| Change in amount of conflict with others | |||||||||
| More than before | 90 | 65.5 (56.4–73.6) | 19 | 14.4 (9.2–21.7) | 18 | 15.1 (9.5–23.1) | 6 | 5.0 (2.1–11.6) | p = 0.65 |
| About the same | 396 | 76.6 (72.3–80.5) | 70 | 14.3 (11.2–18.0) | 25 | 5.2 (3.5–7.7) | 15 | 3.8 (2.1–6.6)) | |
| Less than before | 88 | 58.3 (49.3–66.7) | 35 | 26.7 (19.3–35.7) | 16 | 12.1 (7.4–19.1) | 4 | 2.9 (1.1–7.6) | |
| Whether shielding | |||||||||
| Yes | 44 | 62.3 (47.9–74.7) | 8 | 13.6 (6.4–26.6) | 4 | 5.6 (2.1–14.5) | 8 | 18.4 (9.0–34.0) | p < 0.01 |
| No | 527 | 72.0 (68.2–75.4) | 116 | 16.9 (14.1–20.1) | 53 | 8.6 (6.6–11.1) | 16 | 2.5 (1.5–4.3) | |
| Change in cohabitation status with spouse | |||||||||
| Did not cohabit pre/post lockdown | 192 | 70.5 (63.9–76.4) | 42 | 18.1 (13.2–24.2) | 21 | 8.7 (5.6–13.2) | 6 | 2.7 (1.1–6.5) | p = 0.15 |
| Cohabitation status with partner changed during lockdown | 17 | 63.1 (39.5–81.7) | 2 | 6.6 (1.6–23.5) | 3 | 16.5 (5.2–41.3) | 2 | 13.8 (3.3–42.5) | |
| Cohabited with partner/spouse pre and post lockdown | 365 | 72.1 (67.7–76.1) | 80 | 16.5 (13.3–20.2) | 35 | 7.6 (5.4–10.5) | 17 | 3.8 (2.3–6.3) | |
*P-values compare rates of problem gambling within each variable. For employment changes, this compares becoming unemployed with not becoming unemployed or being furloughed with not being furloughed. Data not shown for each comparator group.
Binary logistic regression models examined which factors were associated with PGSI status. For men, the dependent variable was problem gambling (a PGSI score: >=8). For women, due to the smaller sample size, the number experiencing problem gambling was low (n = 25). Consequently, the dependent variable modelled was experience of moderate risk or problem gambling (a PGSI score ≥ 3). For comparability with women, an additional model was run for men using moderate risk or problem gambling as the dependent variable (see supplementary table S1). In each model, the key independent variables were changes in gambling behaviour or circumstances during the initial Covid-19 pandemic, which were selected for inclusion if the association with problem gambling according to the Wald’s F test was<0.1. Additional covariates were included based on known associations with gambling harms (age, area deprivation, education levels, and wellbeing). All variables included in the models were categorical and were compared using the simple indicator contrasts. The reference categories are reported in the results. Missing data were excluded, except for shielding status among men (n = 28) which was coded as a dummy category.
Diagnostic checks on multi-collinearity were conducted by calculating the variance inflation factors (VIF) of all independent variables (Mansfield & Helms, 1982). Becoming unemployed and being furloughed were highly correlated, with VIF values > 5 in all models. Becoming unemployed was thus removed from the model. Results were checked using all permutations and this did not affect results. All other variables had VIF values of < 2, indicating they were not too closely correlated. Regression models were performed using the complex survey function in Stata v15 to adjust for the weighted stratified survey design (Rao and Scott, 1984). All estimates were weighted to match the age, sex, regional and sports betting profile of the broader sample of regular sports bettors within the YouGov panel. True (unweighted) bases and sample sizes are presented.
6. Results
6.1. Bivariate analyses: Changes in gambling behaviours
Table 1a, Table 1b (column A) shows the percentage of male (n = 3084) and female (n = 782) regular sports bettors who participated in each of the 23 gambling activities in the three month periods prior to, and during, the initial Covid-19 lockdown. Prior to lockdown online sports betting was the most commonly reported gambling activity (78.7% men, 61.4% women). Other common gambling activities pre-lockdown included: the lottery (62.1% men, 65.2% women); online betting on horse/dog races (43.2% men, 50.5% women); betting at a bookmakers on sports (26.7% men, 16.1% women) or horse/dog races (23.0% men, 22.0% women); and scratchcards (26.6% men, 42.7% women). For these activities, and most other gambling activities, the percent participating was lower post lockdown. For example, 78.7% of male participants bet on online sports events prior to the initial Covid-19 lockdown, falling to 16.1% during this lockdown. Only betting online on virtual sports/races saw a significant increase in participation during lockdown.
Table 1a, Table 1b (Column B) also show starting and stopping rates for each of the 23 gambling activities and use the McNemar test for paired data to assess changes in participation prior to and during the initial Covid-19 lockdown The vast majority of male and female sports bettors stopped participating in key sports betting activities during lockdown, as did those taking part in other activities which were largely unavailable during lockdown. However, 2.9% (McNemar’s X2 = 1975, p < 0.01) of male sports bettors and 0.8% (X2 = 472, p < 0.01) of female sports who had not previously bet on sports online started to do so during the initial Covid-19 lockdown. Among men, highest starting rates were observed for lotteries (5.9%, X2 = 301.65, p < 0.01) and online betting on virtual sports/races (3.5%, X2 = 8.58, p < 0.01). Among women highest rates were observed for online betting on virtual sports/races (6.8%, X2 = 10.87, p < 0.01); lotteries (6.5%, X2 = 112.65, p < 0.01) and online bingo (3.4%, X2 = 9.19, p < 0.01). While the proportion starting new individual activities in these three months was relatively low, cumulatively 17.3% of men and 16.5% of women started gambling on at least one new form of gambling activity during the initial Covid-19 lockdown; 29.8% of men and 33.4% of women stopped all gambling during the initial Covid-19 lockdown; and 31.3% of male sports bettors and 30.3% of female sports bettors increased their frequency of gambling on at least one activity during the initial Covid-19 lockdown.
6.2. Problematic gambling and Covid-19 related experiences
Of the 3084 male and 782 female participants, 5.4% of male sports bettors (Table 2a) and 3.9% of female sports bettors (Table 2b) experienced problem gambling (PGSI score ≥ 8) during the initial Covid-19 lockdown in Britain. A further 10.7% of men and 8.3% of women experienced moderate risk gambling (PGSI score of 3–7).
Bivariate results showed that for both men and women, problem gambling rates were higher among those who were: furloughed (10.8% men (Walds F (5.9, 22971) = 2.69, p < 0.05); 5.6% women (F (5.7, 21923) = 1.7, p < 0.05) or shielding for health reasons (8.5% men (F (2.9, 11263) = 2.1, p < 0.05); 18.4% women, (F (2.9–11113) = 10.8, p < 0.01).; Additionally, for men, problem gambling rates were higher among those who reported more conflict with others during lockdown (10.8%, F (5.9, 22973) = 15.9, p < 0.05), who were financially worse off during lockdown (8.6%, (F (5.9–22814) = 4.9, p < 0.01) and who reported increased levels of stress (7.7%, F (5.9–23003) = 8.2, p < 0.01).
Multivariate logistic regressions are presented in Table 3a (men) and Table 3b (women). Men were more likely to experience problem gambling if they were younger (under 35 compared with over 55, adjusted odds ratios (AOR) 5.24 (95% CI: 2.62–10.49)); had lower rather than higher wellbeing scores (AOR: 2.23 (1.39–3.60)), had started a new form of gambling during lockdown (AOR: 2.50 (1.38–4.53)) compared with those who had not, or had changed their level of spending on gambling. Shielding status was associated with problem gambling but none of individual categories varied significantly from the reference group of not-shielding.
Table 3a.
Odd ratios (OR) for problem gambling among regular male sports bettors.
|
Problem gambling |
||||
|---|---|---|---|---|
| N (%) | OR | 95% CI (lower) | 95% CI (upper) | |
| Age group (p > 0.01) | ||||
| 55+ | 1387 (35.5%) | 1 (ref) | ||
| 54–35 | 1296 (42.3%) | 2.11 | 1.18 | 3.77 |
| 35 and under | 401 (35.5%) | 5.24 | 2.62 | 10.49 |
| Highest Educational attainment (p = 0.08) | ||||
| Post-secondary education (including degree or higher) | 1565 (52.1%) | 1 (ref) | ||
| A Level or equivalent | 663 (22.2%) | 1.28 | 0.76 | 2.15 |
| GCSE Level or equivalent | 566 (17.1%) | 1.74 | 1.00 | 3.02 |
| CSE or equivalent/other/none | 290 (8.6%) | 2.44 | 1.14 | 5.24 |
| Area deprivation quintile (p = 0.68) | ||||
| Least deprived | 739 (23.5%) | 1 (ref) | ||
| 2nd | 636 (20.7%) | 1.16 | 0.61 | 2.24 |
| 3rd | 605 (19.8%) | 1.24 | 0.63 | 2.43 |
| 4th | 574 (18.8%) | 1.23 | 0.64 | 2.36 |
| Most deprived | 530 (17.2%) | 1.60 | 0.84 | 3.02 |
| Wellbeing (p < 0.01) | ||||
| Moderate/high wellbeing scores (>=20) | 2025 (63.5%) | 1 (ref) | ||
| Lower wellbeing scores (<=20) | 1059 (36.5%) | 2.23 | 1.39 | 3.60 |
| Change in employment status (p = 0.48) | ||||
| Employed pre and post lockdown | 1856 (63.1%) | 1 (ref) | ||
| Employed pre covid, furloughed during lockdown | 227 (7.8%) | 1.45 | 0.72 | 2.91 |
| Other employment status | 1001 (29.0%) | 0.87 | 0.48 | 1.55 |
| Subjective change in finances (p < 0.05) | ||||
| About the same | 1477 (66.4%) | 1 (ref) | ||
| Financially worse off | 800 (14.9%) | 1.50 | 0.93 | 2.41 |
| Financially better off | 807 (18.7% | 0.73 | 0.43 | 1.23 |
| Change in stress levels (p = 0.07) | ||||
| About the same | 1658 (52.0%) | 1 (ref) | ||
| More than before | 1018 (34.1%) | 1.41 | 0.87 | 2.28 |
| Less than before | 408 (13.8% | 2.15 | 1.11 | 4.18 |
| Change in free time (p = 0.90) | ||||
| About the same | 1419 (49.4%) | 1 (ref) | ||
| More than before | 1448 (43.1%) | 1.00 | 0.62 | 1.63 |
| Less than before | 217 (7.5%) | 1.16 | 0.58 | 2.35 |
| Change in levels of conflict with others (p = 0.27) | ||||
| About the same | 2087 (14.9%) | 1 (ref) | ||
| More than before | 416 (6.6%) | 1.54 | 0.90 | 2.62 |
| Less than before | 581 (18.7%) | 1.09 | 0.60 | 1.98 |
| Whether shielding (p < 0.01) | ||||
| No | 2780 (90.1%) | 1 (ref) | ||
| Yes | 276 (8.3%) | 1.98 | 0.97 | 4.01 |
| Missing | 28 (1.2%) | 6.80 | 2.48 | 18.69 |
| Change in cohabitation with spouse/partner (p = 0.68) | ||||
| Did not cohabit pre/post lockdown | 962 (33.9%) | 1 (ref) | ||
| Cohabitation status with partner changed during lockdown | 54 (2.1%) | 1.52 | 0.52 | 4.45 |
| Cohabited with partner/spouse pre and post lockdown | 2068 (66.0%) | 0.95 | 0.61 | 1.47 |
| Change in gambling activities (p < 0.01) | ||||
| Continued gambling during lockdown | 1693 (53.0%) | 1 (ref) | ||
| Started a new form of gambling during lockdown | 490 (17.3%) | 2.50 | 1.38 | 4.53 |
| Stopped all gambling during lockdown | 901 (29.8%) | 0.81 | 0.44 | 1.51 |
| Change in gambling expenditure (p < 0.05) | ||||
| Spent about the same during lockdown | 363 (11.5%) | 1 (ref) | ||
| Spent less during lockdown | 2554 (82.5%) | 3.92 | 1.36 | 11.33 |
| Spent more during lockdown | 167 (6.0%) | 3.55 | 1.07 | 11.82 |
| Change in gambling frequency (p = 0.21) | ||||
| No change/frequency decreased | 2167 (68.7%) | 1 (ref) | ||
| Frequency of gambling on any activity increased | 917 (31.3%) | 1.53 | 0.78 | 2.98 |
Table 3b.
Odds ratios (OR) for moderate risk/problem gambling among regular female sports bettors.
| N (%) | OR | 95% CI (lower) | 95% CI (upper) | |
|---|---|---|---|---|
| Age group (p < 0.01) | ||||
| 55+ | 174 (26.1%) | 1 (ref) | ||
| 54–35 | 235 (38.7%) | 4.46 | 1.72 | 11.55 |
| 35 and under | 283 (35.2%) | 4.06 | 1.63 | 10.15 |
| Highest Educational attainment (p = 0.98) | ||||
| Post-secondary education (including degree or higher) | 442 (58.2%) | 1 (ref) | ||
| A Level or equivalent | 148 (20.3%) | 0.93 | 0.45 | 1.94 |
| GCSE Level or equivalent | 129 (14.1%) | 1.04 | 0.48 | 2.22 |
| CSE or equivalent/other/none | 63 (6.3%) | 1.23 | 0.36 | 4.17 |
| Area deprivation quintile (p = 0.49) | ||||
| Least deprived | 168 (23.1%) | 1 (ref) | ||
| 2nd | 165 (21.0%) | 2.11 | 0.90 | 4.94 |
| 3rd | 174 (20.2%) | 1.77 | 0.71 | 4.43 |
| 4th | 145 (18.5% | 1.32 | 0.51 | 3.44 |
| Most deprived | 130 (17.2%) | 1.52 | 0.63 | 3.65 |
| Wellbeing (p < 0.01) | ||||
| Moderate/high wellbeing scores (>=20) | 487 (59.9%) | 1 (ref) | ||
| Lower wellbeing scores (<=20) | 295 (40.1%) | 3.48 | 1.97 | 6.16 |
| Change in employment status (p = 0.48) | ||||
| Employed pre and post lockdown | 453 (60.1%) | 1 (ref) | ||
| Employed pre covid, furloughed during lockdown | 81 (11.5%) | 1.24 | 0.57 | 2.72 |
| Other pre/post lockdown | 248 (27.6%) | 0.90 | 0.44 | 1.83 |
| Subjective change in finances (p = 0.09) | ||||
| About the same | 377 (63.0%) | 1 (ref) | ||
| Financially worse off | 250 (18.7%) | 1.79 | 0.99 | 3.22 |
| Financially better off | 155 (18.3%) | 0.89 | 0.39 | 2.05 |
| Change in levels of conflict with others (p = 0.31) | ||||
| About the same | 506 (65.6%) | 1 (ref) | ||
| More than before | 133 (15.8%) | 0.75 | 0.39 | 1.42 |
| Less than before | 143 (18.6%) | 1.27 | 0.55 | 2.95 |
| Whether shielding (p < 0.01) | ||||
| No | 718 (92.3%) | 1 (ref) | ||
| Yes | 64 (7.7%) | 4.08 | 1.84 | 9.06 |
| Missing | N/A | N/A | ||
| Change in gambling activities (p = 0.13) | ||||
| Continued gambling during lockdown | 416 (50.5%) | 1 (ref) | ||
| Started a new form of gambling during lockdown | 118 (16.5%) | 1.83 | 0.91 | 3.68 |
| Stopped all gambling during lockdown | 248 (33.4%) | 0.62 | 0.27 | 1.44 |
| Change in gambling expenditure (p = 0.47) | ||||
| Spent about the same during lockdown | 627 (79.8%) | 1 (ref) | ||
| Spent less during lockdown | 119 (15.9%) | 1.58 | 0.76 | 3.28 |
| Spent more during lockdown | 36 (4.4%) | 1.38 | 0.44 | 4.33 |
| Change in gambling frequency (p < 0.01) | ||||
| No change/frequency decreased | 559 (69.7%) | 1 (ref) | ||
| Frequency of gambling on any activity increased | 223 (30.3%) | 4.21 | 1.99 | 8.92 |
Multivariate logistic regressions for women showed that age, wellbeing, shielding status and increases in gambling frequency were all associated with moderate risk or problem gambling. AOR were: 4.06 (95% CI: 1.63–10.15) times higher among women under 35 than those over 55; 3.48 (1.97–6.16) times higher among those with lower rather than higher wellbeing scores; 4.08 (1.84–9.06) times higher among those who were shielding than those who were not; and 4.21 (1.99–8.92) times higher among those whose gambling frequency on any activity increased during lockdown.
7. Discussion
Among previously regular male and female sports bettors, the main impact of the initial Covid-19 lockdown was either a reduction in gambling activity or, for around a third of the sample, stopping gambling altogether. This is to be expected, given that the primary segment of sports bettors’ gambling repertoires was generally unavailable to them during this time. Other studies have noted similar reductions during the initial Covid-19 outbreak (Håkansson, 2020, Lindner et al., 2020, Auer et al., 2020). However, when examining online sports betting, a minority continued to bet as some horse races and sports were still available in other jurisdictions. Furthermore, a minority initiated gambling on forms of activity they had not previously engaged in.
Whilst the starting rates for most individual activities were low, taken together around one in six sports bettors started a new form of gambling activity during the initial Covid-19 lockdown and around one in three increased their frequency of gambling on at least one activity. Thus, reductions in overall participation rates mask changes in gambling behaviours for some individuals. Importantly, our study, like Håkansson (2020), found that changes in gambling activity during lockdown were associated with elevated rates of problem gambling. For men, starting a new form of gambling during lockdown was associated with the experience of problem gambling. Notably change in gambling frequency was not associated with problem gambling for men, suggesting it may be the act of switching activity type, rather than increasing gambling frequency, that is a driver of harm.
This has implications for regulators and industry. Concerns were raised at the outset of the initial lockdown period that gambling operators may attempt to “cross-sell” sports bettors into other, potentially more risky, activities (All Party Parliamentary Group on Gambling Harms, 2020). The results presented here suggest that male sports bettors who did switch activity, regardless of level of risk, are more likely to experience gambling harms and should be viewed as vulnerable to harms. Relatedly, a review of the Gambling Act 2005 is currently being undertaken by the UK government, with debates exploring whether placing greater controls or restrictions on certain forms of gambling may generate changes in behaviour. The initial Covid-19 lockdown provides a quasi-experimental opportunity to look at what happens when certain forms of gambling are restricted. Our results show that among regular sports bettors, restrictions in supply generated changes in behaviour, including reductions in gambling for the majority, who did not appear to seek out other ways to gamble, but arguably stimulated engagement in new gambling activities among a minority. This stimulation may have been enhanced by industry actions, such as advertising and marketing.
Interestingly, among male sports bettors, both increased and decreased spending on gambling during lockdown were associated with problem gambling. This may be consistent with different behaviour trajectories among those experiencing gambling harms, with some increasing spending as behaviours become less controlled and others attempting to cut back. Among women, increases in gambling frequency, rather than changes in gambling spend or activity, were associated with moderate risk or problem gambling, although the sample may have been underpowered to detect changes in spend or type of gambling.
Whilst bivariate analyses showed some associations between problem gambling for men and moderate risk or problem gambling for women with Covid-19 related personal and financial circumstances, these were not borne out in the multivariate models. Thus none of our hypotheses about the impact of Covid-related personal circumstances upon gambling harms were supported in this initial assessment, though it is plausible that increased stress, anxiety or conflict may be associated with changes in gambling behaviours, which in turn are associated with gambling harms. Future waves of this study should assess this.
Among women, only ‘shielding’ status was significantly and substantially associated with moderate risk or problem gambling (AOR 4.08; 95% CI 1.84–9.06). Shielding people were advised to avoid all social contact for a long duration of time because of their underlying health problems. This evidence is commensurate with the noted association between poorer health and gambling harms (Cowlishaw and Kessler, 2016). Whilst general health status was not accounted for in the models, wellbeing was, and other mechanisms may contribute to this association – such as increased social isolation.
Finally, for both men and women, lower levels of wellbeing were strongly associated with gambling harms. This association has been documented previously (Farrell, 2018). However, studies have suggested that mental distress has increased significantly during Covid-19 (Pierce et al, 2020), suggesting that gambling behaviours should be considered within the nexus of Covid-19 related behaviours and experiences that may impact upon mental health and wellbeing (Price, 2020).
As with all research, this study has limitations. Results are cross-sectional, with attendant issues relating to inferring causality. Participants were recruited from a non-probability online panel, with consequent issues for generalisability. However, regular sports bettors are a hard to reach and niche group, and the YouGov panel allowed us to identify people who were regular (at least monthly) sports bettors prior to the Covid-19 pandemic and to sample from that pool. Non-probability panel surveys tend to produce higher estimates of risk-taking behaviours than random probability surveys (Callegaro et al 2014). However, analysis shows that closed non-probability online panels tend to produce similar conclusions to probability methods when focusing on multi-variate analyses and when exploring the relationship between variables (Callegaro et al, 2014). Whilst data quality was generally high, there was some attrition between logging on to the survey and completion and a few participants (3.7%) were removed from the sample data on the basis of ‘seriousness’ checks. However, a strength of sampling from YouGov’s panel of known sports bettors is that we could weight data for non-response by age, sex, geography and gambling behaviour. All data are self-reported and may be subject to a range of biases, although, where possible, we used well-validated instruments. Participants were first asked to recall their behaviours pre-Covid-19, and then during lockdown. This may have been cognitively challenging for some participants. However, key outcome measures like problem gambling had reference periods explicitly defined (e.g. in the last three months) so that we could be sure that these were people’s concurrent experiences during the initial gambling lockdown. Because regular gamblers take part in a range of activities, mapping changes in measures like frequency was complex and data were condensed for simplicity but may mask further patterns. Finally, analyses for women could only look at moderate and problem gambling combined and not examine problem gambling separately as base sizes were too small. However, additional models were run for men looking at the factors associated with moderate risk and problem gambling combined, giving broadly similar results (with the addition of any change in levels of conflict and gambling frequency being associated with MRPG among men).
8. Conclusions
Whilst a reduction in gambling was the norm for most regular sports bettors during the initial Covid-19 lockdown in Britain, some started new forms of gambling or increased their frequency of gambling on other activities. . Among men, those who switch forms of gambling under lockdown conditions and experience poorer wellbeing should be considered vulnerable to harms, as should women who increased their frequency of gambling on any activity. Among women, shielding status and poorer wellbeing were also associated with the experience of gambling harms during Britain’s initial Covid-19 lockdown. Hence, those facing these challenges should be considered potentially vulnerable to gambling harms. Regulators and industry should take action to further protect these emerging vulnerable groups.
Funding
This work is supported by a grant from UK Research and Innovation (Economic and Social Research Council), grant number ES/V004549/1.
Conflicts of Interest
HW was Deputy Chair of the Advisory Board for Safer Gambling between May 2015 and March 2020. She was remunerated by the Gambling Commission (the industry regulator) for this. She is a member of the WHO panel on gambling and in 2019 worked on a study looking at gambling and suicide funded by GambleAware. GR, FD, NC, RP and MS have worked together on one study funded by GambleAware, which examined the impact of gambling marketing/advertising on young people and vulnerable groups. The funders have no role in the design of the study; in the collection, analyses or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.
CRediT authorship contribution statement
Heather Wardle: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Supervision, Validation, Writing - review & editing. Craig Donnachie: Conceptualization, Data curation, Investigation, Validation. Nathan Critchlow: Writing - review & editing. Ashley Brown: . Christopher Bunn: Conceptualization. Fiona Dobbie: Conceptualization. Cindy Gray: Conceptualization. Danielle Mitchell: Writing - review & editing. Richard Purves: Writing - review & editing. Gerda Reith: Conceptualization. Martine Stead: . Kate Hunt: Conceptualization, Funding acquisition, Methodology, Project administration, Supervision.
Acknowledgments
We would like to thank our Study Steering Committee for comments on the design and progress of the research and thank all participants for giving up their time to take part.
Footnotes
Supplementary data to this article can be found online at https://doi.org/10.1016/j.addbeh.2021.106876.
Appendix A. Supplementary data
The following are the Supplementary data to this article:
References
- Abbott, M., Bellringer, M., Vandal, A., Hodgins, D., Palmer Du Preez, K. Landon, J., Sullivan, S., Feigin, V. (2012). Effectiveness of problem gambling brief intervention telephone calls: a Randomised Control Trial. Gambling and Addictions Research Centre, Auckland University of Technology. Available at https://www.researchgate.net/publication/269875396_EFFECTIVENESS_OF_PROBLEM_GAMBLING_BRIEF_TELEPHONE_INTERVENTIONS_A_RANDOMISED_CONTROLLED_TRIAL. Accessed 02.01.2020.
- All Party Parliamentary Group on Gambling Harms (2020) Call for Tighter Restrictions on Online Gaming During the COVID-19 Pandemic. Letter to Rt Hon Nigel Huddlestone. Available at: http://www.grh-appg.com/wp-content/uploads/2020/04/Letter-to-Nigel-Huddleston-MP-COVID.pdf. Accessed 10.09.2020.
- Auer M., Malischnig D., Griffiths M.D. Gambling Before and During the COVID-19 Pandemic Among European Regular Sports Bettors: An Empirical Study Using Behavioral Tracking Data. Internatinal Journal of Mental Health Addiction. 2020 doi: 10.1007/s11469-020-00327-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- A. Blaszczynski, L. Nower 2002. A pathways model of problem and pathological gambling. Addiction, 97 (5), pp. 487-499. [DOI] [PubMed]
- Callegaro M., Villar A., Krosnick J., Yeager D. In: Online Panel Research: A Data Quality Perspective. Callegaro M., Baker R., Bethlehem J., Goritz A., Krosnick J., Lavrakas P., editors. John Wiley & Sons; UK: 2014. A Critical Review of Studies Investigating the Quality of Data Obtained With Online Panels. [Google Scholar]
- Cowlishaw S., Kessler D. Problem gambling in the UK: Implications for Health, Psychosocial Adjustment and Health Care Utilisation. Eur. Addict. Res. 2016;22:90–98. doi: 10.1159/000437260. [DOI] [PubMed] [Google Scholar]
- Farrell L. Understanding the Relationship Between Subjective Wellbeing and Gambling Behavior. J. Gambl. Studies. 2018;34:55–71. doi: 10.1007/s10899-017-9692-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Forrest D., McHale I.G. Transmission of Problem Gambling Between Adjacent Generations. J. Gambl. Studies. 2020 doi: 10.1007/s10899-020-09977-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Ferris J., Wynne H. Canadian Centre on Substance Abuse; Ottawa: 2001. The Canadian Problem Gambling Index: Final report. [Google Scholar]
- Gambling Commission Risks arising from Covid-19 and our response Available at: https://www.gamblingcommission.gov.uk/news-action-and-statistics/Statistics-and-research/Covid-19-research/Risks-arising-from-Covid-19-and-our-response.aspx 2020 Accessed 30.10.2020.
- Håkansson A. Changes in Gambling Behavior during the COVID-19 Pandemic—A Web Survey Study in Sweden. Int. J. Environ. Res. Public Health. 2020;17:4013. doi: 10.3390/ijerph17114013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Hunt K., Critchlow N., Brown A., Bunn C., Dobbie F., Donnachie C., et al. Protocol for a Mixed-Method Investigation of the Impact of the COVID-19 Pandemic and Gambling Practices, Experiences and Marketing in the UK: The “Betting and Gaming COVID-19 Impact Study. Int. J. Environ. Res. Public Health. 2020;17(22):8449. doi: 10.3390/ijerph17228449. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Hunt, K., Critchlow, N., & Wardle, H. (2020, August 6). Betting and gaming: the Covid-19 impact study. Retrieved from osf.io/f4zgw.
- Jauregui P., Onaindia J., Estévez A. Adaptive and maladaptive coping strategies in adult pathological gamblers and their mediating role with anxious-depressive symptomatology. J. Gamb. Stud. 2017;33(4):1081–1097. doi: 10.1007/s10899-017-9675-5. [DOI] [PubMed] [Google Scholar]
- Kennedy C, Mercer A, Keeter S, Hatley N, McGeeney K, Gimenez A. (2016) Evaluating Online Nonprobability Surveys. Pew Research. Available at https://www.pewresearch.org/methods/2016/05/02/evaluating-online-nonprobability-surveys/ Accessed 10.03.2020.
- Kushnir V., Godinho A., Hodgins D.C., et al. Self-Directed Gambling Changes: Trajectory of Problem Gambling Severity in Absence of Treatment. J. Gambl. Stud. 2018;34:1407–1421. doi: 10.1007/s10899-018-9769-8. [DOI] [PubMed] [Google Scholar]
- Lindner P., Forsström D., Jonsson J., Berman A.H., Carlbring P. Transitioning Between Online Gambling Modalities and Decrease in Total Gambling Activity, but No Indication of Increase in Problematic Online Gambling Intensity During the First Phase of the COVID-19 Outbreak in Sweden: A Time Series Forecast Study. Front. Public Health. 2020;8 doi: 10.3389/fpubh.2020.554542. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Mansfield E.R., Helms B.P. Detecting multicollinearity. The American Statistician. 1982;36(3a):158. [Google Scholar]
- McCarthy S., Thomas S.L., Randle M., et al. Women’s gambling behaviour, product preferences, and perceptions of product harm: Differences by age and gambling risk status. Harm Reduction J. 2018;15:22. doi: 10.1186/s12954-018-0227-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- McCarthy S., Thomas S.L., Bellringer M.E., et al. Women and gambling-related harm: A narrative literature review and implications for research, policy, and practice. Harm Reduction Journal. 2019;16:18. doi: 10.1186/s12954-019-0284-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Ogueji IA, Okoloba MM, Ceccaldi, BMD (2021). Coping Strategies of Individuals in the UK during Covid-19. Curr. Psychol. 2021 Jan 3 : 1–7. doi: 10.1007/s12144-020-01318-7 [Epub ahead of print]. [DOI] [PMC free article] [PubMed]
- Pierce M., Hope H., Ford T., Hatch S., Hotopf M., John A., et al. Mental health before and during the COVID-19 pandemic: A longitudinal probability sample survey of the UK population. Lancet Psych. 2020;7(10):883–892. doi: 10.1016/S2215-0366(20)30308-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Price A. Online gambling in the midst of COVID-19: a nexus of mental health concerns, substance use and financial stress. Int. J. Mental Health. 2020 doi: 10.1007/s11469-020-00366-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Rao J.N.K., Scott A.J. On chi-squared test for multi-contingency tables with cell proportions estimated from survey data. Annals Statist. 1984;12:46–60. [Google Scholar]
- Tennant R., Hiller L., Fishwick R., Platt S., Joseph S., Weich S., et al. The Warwick-Edinburgh Mental Well-being Scale (WEMWBS): Development and UK validation. Health and Quality of Life Outcomes. 2007;5:63. doi: 10.1186/1477-7525-5-63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Twyman J. Getting it right: YouGov and online survey research in Britain. J. Elect., Public Opinion Parties. 2008;18(4):343–354. [Google Scholar]
- University College London COVID-19 Online Survey Questionnaire - Wave 1 - April 2020 (Version 2) Available at: https://cls.ucl.ac.uk/wp-content/uploads/2020/07/UCL-Centre-for-Longitudinal-Studies-COVID-19-Online-Survey-Questionnaire-Wave-1-April-2020.pdf 2020 Accessed 10.06.2020.
- Wang C., Cunningham-Erdogdu P., Steers M., Weinstein A.P., Neighbors C. Stressful life events and gambling: The roles of coping and impulsivity among college students. Addictive Behaviors. 2020;107 doi: 10.1016/j.addbeh.2020.106386. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Wardle H. (2015) Female gambling behaviour: a case study of realist description [PhD thesis]: University of Glasgow; Available from: http://theses.gla.ac.uk/6117/. Accessed 20 Dec 2020.
- Williams R.J., Leonard C.A., Belanger Y.D., et al. Predictors of gambling and problem gambling in Canada. Canad. J. Public Health. 2021 doi: 10.17269/s41997-020-00443-x. 10.17269/s41997-020-00443-x [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.