How can we prevent gambling harms?

An interactive chart of strategies to prevent gambling harms
Gambling
Author

Rob Heirene

Published

September 18, 2023

Planning ahead

I (hopefully) have a long career ahead of me in addiction research, and I don’t want it to be guided solely by the ever-increasing pressure to publish. I want to do research that adds real value to the world.

With this in mind, I’m thinking about problem gambling prevention a lot at the moment and where to place my focus. To help with this, I’ve started mapping all the existing harm prevention strategies used in practice or proposed in the gambling literature.

Prevention strategies visualised

To make this process a little more interesting for me and anyone who reads this, I’ve been playing around with the plotly R package to present my catalogue of prevention strategies in a visual and interactive way.

Work in progress

Please note that I’ve only just started making the chart and I’m conscious that it currently provides a very incomplete picture of the strategies available. I will be reading around and adding more strategies and information over the next few months (starting July 2023).

That said, please feel free to suggest some strategies or edits using the comments section at the end (or you can send me an email).

Below is what I have so far, visualised as a sunburst chart. Hover over each category or strategy to reveal descriptions of each, and click on the bars to see more. It’s best viewed on a PC/mac or tablet (if you are on mobile, turn it to landscape mode).

Show set-up code 
# Install and load the groundhog package to ensure consistency of the package versions used here:
# install.packages("groundhog") # Install

library(groundhog) # Load

# List desired packages:
packages <- c('tidyverse', # Clean, organise, and visualise data
              'plotly', # Add interactive elements to figures
              'sysfonts', # Special fonts for figures
              'showtext', # Special fonts for figures
              'htmlwidgets') # Make plotly plots HTML format for rendering in Quarto
              

# Load desired package with versions specific to project start date:
groundhog.library(packages, "2023-06-05") # Set Slightly earlier than start date as some of the plotly dependency packages wouldn't load properly with date as 07/07/2023

# Load new font for figures/graphs
font_add_google("Reem Kufi", "Reem Kufi")
showtext_auto(enable = TRUE)
Show code used to create chart 
# Load manually input data:
# gam_prevention_data<- read.csv("gamb_prevention_data.csv") %>% select(Layer:Description) # Warning, something as simple as an apostrophe and this dataset can cause it to not play nice

#  gam_prevention_data<- read.csv("posts/2023-gambling-harms-chart/gamb_prevention_data.csv") %>% select(Layer:Description) # Load in session
# 
# 
# # Make sunburst chart:
#  gam_prevention_sunburst<- plot_ly(gam_prevention_data,
#                                    ids = ~IDs,
#                                    labels = ~Label,
#                                    parents = ~Parents,
#                                    hoverinfo = "text",  # Needed for below argument to work
#                                    hovertext = ~Description,
#                                    type = 'sunburst',
#                                    maxdepth = 3,
#                                    insidetextorientation = 'radial',
#                                    marker = list(line = list(width = 5, color = '#202B2A')),  # Adjust border size + color
#                                    sort = TRUE) %>%
#    plotly::layout(colorway = c("#289998", "#288D98", "#287699", "#7E6F98", "#BFB3C0", "#FFF7E7", "#E6C97F", "#8C8033", "#7E9F6E", "#65BBA3", "#94C8C0", "#998F28", "#62A899", "#986289", "#289970"),
#          font = list(
#            size = 14,
#            family = "Reem Kufi", # change main font
#                      color = "#FFF7E7"), # Set main font so the central layer text can be seen
#          paper_bgcolor = '#202B2A', # background consistent with site
#          plot_bgcolor = '#202B2A',
#          hoverlabel = list(font = list(size = 12, family = "Reem Kufi")))  # Increase the font size and family for hover text)
# 
# # The standard way of changing the font in ggplotly doesn't seem to work for me, and it seems like other people having the same issue. I found this workaround online (https://github.com/plotly/plotly.R/issues/2117) which I now use below to load and use the correct font in our ggplotly:
# 
# # Get the URL for the "Reem Kufi" font from Google Fonts:
# reem_kufi_file <- showtextdb::google_fonts("Reem Kufi")$regular_url
# 
# # Create custom CSS:
# reem_kufi_css <- paste0(
#   "<style type = 'text/css'>",
#     "@font-face { ",
#       "font-family: 'Reem Kufi'; ",
#       "src: url('", reem_kufi_file, "'); ",
#     "}",
#   "</style>"
# )
# 
# 
# # Add the CSS as a dependency for the plotly plot:
# gam_prevention_sunburst$dependencies <- c(
#   gam_prevention_sunburst$dependencies,
#   list(
#     htmltools::htmlDependency(
#       name = "reem-kufi-font",
#       version = "0",
#       src = "",
#       head = reem_kufi_css)))
# # ****Uncomment to update****
# # Save plot as a HTML file (it doesn't render well with quarto otherwise):
# saveWidget(gam_prevention_sunburst, 'posts/2023-gambling-harms-chart/sunburst_fig.html')
Recreating the chart

I’ve included the R code used to create the chart in the drop-down sections just above it. Feel free to recreate it (the data are on the Github repo for this site) or adapt the code to your own chart, but please credit me if you do :)

The evidence

As I (re-)read the literature on this topic, I’m also creating a record of relevant studies and resources relating to each prevention strategy. This section is even less developed than the chart, so the same call for suggestions applies.

There will eventually be a lot of information presented here, so I’ve hidden the text for each prevention strategy inside expandable boxes for easier navigation.

Consumer protection tools

  • The authors in this study collected retrospective data from a European gambling site that offers their customers the ability to estimate their recent gambling behaviour and expenditure (“Recent gambling self assessment” in the chart), including amount wagered, deposited, and withdrawn, as well as play duration in hours (all over the last 30 days). They compared peoples’ (n = 639) deposit amount before and after this self-assessment and found a significant reduction that wasn’t affected by whether people underestimated, overestimated, or correctly estimated their past 30 day deposit amount. While the reduction was quite substantial (averaged between -$85 to $-300 across the three estimation groups), there was no control/comparator group and so we do not know whether this change would have occurred in the absence of this self-assessment (just regression to the mean?). Also, as the authors note, the group of people who voluntarily and actively took time to complete this self-assessment may have already made the decision to decrease their gambling before taking the self-assessment.

Restrictions

  • A survey (n = 2352) of Norsk Tipping (a Norwegian Government-owned gambling operator) customers’ perspectives on the recently implemented global loss limit of NOK 20,000 per month (~$2450USD; Auer et al., 2018). It’s slightly unclear, but it appears that players also had to set a daily and monthly loss limit specific to game types (e.g., online casino, bingo), but the all-game global limit of 20k NOK per month still applied. Most found the new limits easy to understand (76%) and had positive attitudes towards them (79%). Only 25% agreed entirely or in part that a maximum loss limit was relevant to the, and 6% of green players (i.e., low risk of problem gambling) and 16% of red players (i.e., high risk of problem gambling) said they continued to gamble with other companies after reaching their limit.

  • This UK Gambling Commission report from 2020 discusses consumer perspectives on stake limits (n = 32, weekly gamblers). Low-risk participants (i.e., PGSI score 0-2) were in favour of stake limits, but wary of losing the fun of gambling and preferred to limit stakes on non-skill games. There was more disagreement within moderate risk participants (i.e., PGSI score 3-7), but they mostly agreed with the idea of stake limits, arguing that multiple variables need to be considered when determining the limit amount (e.g., household income & state of mind) and that players should be able to overrule limits. High risk gamblers (i.e., PGSI score ≥8) were positive about stake limits believed the responsibility to protect players layed firmly on the shoulders of betting firms. Overall, participants were more positive about loss limits than stake limits.

  • It seems that a lot of online gambling sites indirectly impose stake limits by having maximum payout amounts (e.g., £100,000). Payout limits vary from site to site and across activities, but these operator-protecting restrictions (which are sometimes reduced for big winners…) can limit customer stakes. Read more here.

Operator communications

  • This UK Gambling Commission report from 2020 discusses consumer perspectives on safer gambling communications (n = 32, weekly gamblers). Low-risk participants (i.e., PGSI score 0-2) were likely to say that safer gambling messaging was effective, but saw it as not relevant for them and therefore they opted out of such messaging. Moderate risk gamblers (i.e., PGSI score 3-7) opted out of messaging also as they didn’t feel it was for them and didn’t like to be reminded to “gamble responsibly” because they felt in control of their gambling. High risk gamblers (i.e., PGSI score ≥8) stated that highlighting losses could push people to spend more money and that such communications may need to be replaced by limits to be effective. In general, safer gambling messages were viewed as being easy to opt out of and easy to ignore, warranting actual barriers to risky play (i.e., limits).

  • This 2022 study including 3 pre-registered randomized experiments (n ≈ 500 per study) found no evidence for a protective effect (on spending behaviour) of the “When the fun stops, stop” warning message commonly used by UK gambling companies. Participants were UK adults who had previously gambled. They were given a monetary endowment with which they were allowed to bet on either an online roulette game (Study 2) or real soccer events (Study 1 & 3). The presence and presentation of the warning message varied between participants in each study.

    • Newall, P. W. S., Weiss-Cohen, L., Singmann, H., Walasek, L., & Ludvig, E. A. (2022). Impact of the ‘When the Fun Stops, Stop’ gambling message on online gambling behaviour: A randomised, online experimental study. The Lancet Public Health, 7(5), e437–e446. https://doi.org/10.1016/S2468-2667(21)00279-6.

Risk assessment

  • This 2023 study collected PGSI responses from nearly 10,000 Canadians who held an account with Loto-Québec (a provincial Crown Corporation) in Quebec, Canada. They also collected their preceding 12 months of behavioural account data from Loto-Québec and used this data to try and predict PGSI scores (5+ or 8+) using six different machine learning models: Decision trees, K-nearest neighbours, Support vector machines, logistic regression, artificial neural networks, Random forests. The Random Forest model had the highest predictive accuracy (AUC = 84.3% for 5+; 82.5% for 8+). Unlike most studies that try to do something similar, this one actually shares the underlying code and a wealth of supplemental detail regarding methods and results (well done to the authors!).

    • Murch, WS, Kairouz, S, Dauphinais, S, Picard, E, Costes, J-M, French, M. Using machine learning to retrospectively predict self-reported gambling problems in Quebec. Addiction. 2023; 118( 8): 1569– 1578. https://doi.org/10.1111/add.16179

I can’t find any published research on this topic (let me know if you know of something!), but this letter from the UK Gambling Commission clarifies what affordability checks would look like in practice in the UK after they were recommended in the 2023 Gambling White Paper:

Public health campaigns

  • This 2023 study discusses the ‘Odds Are: They Win’ messaging campaign implemented in the UK (Manchester) towards the end of 2022 ( to coincide with the Football World Cup). The authors call this campaign a “disruptive innovation”, given the focus of messages on the gambling industry, rather than individual responsibility. Several different advertisements were displayed over a three-month period on social media and in Greater Manchester’s tram system and coffee shops. Look at the paper to see how the advertisements were displayed, but they included things like “Gambling harm is not only financial, it can impact your health, well-being, and relationships” and “The main purpose of gambling companies to maximise profit, generated through customer losses”. A service evaluation of the programme (the results of which can be found here) suggested the campaign had a wide reach (16,000+ unique clicks to landing page). This all sounds great and I support such messages, but there’s no comparison/control so we can’t draw much from this article about this type of messaging just yet.

    • Mills, T., Grimes, K., Caddick, E., Jenkins, C.l., Evans, J., Moss, A., Wills, J., & Sykes, S. (2023). ‘Odds Are: They Win’: a disruptive messaging innovation for challenging harmful products and practices of the gambling industry. Public Health, 224, 41-44, https://doi.org/10.1016/j.puhe.2023.08.009

Blocking tools

This 2023 report from GamCare summarises the outcomes and recommendations from a virtual workshop on bank gambling block involving researchers, representatives from the financial and gambling industries, gambling support services, debt and money advice services, and individuals with lived experience. They make a number of recommendations for how banks can improve their blocking tools not discussed here. Key findings:

  • Many UK banks now offer gambling blocks, but awareness of these tools is low.

  • The ease of accessing and using the blocks varies across banks.

  • Most banks allow the customers to self deactivate the blocks using their mobile app, although there is to be gay 48 or 72 hour cooling off period before this takes effect. Some banks require customers to speak to their customer support team before turning off the block, increasing friction.

  • There are many loopholes which allow customers with a bank gambling block to continue gambling (e.g., e-wallets; overseas gambling).

Product design

  • This 2021 report by the UK Behavioural Insights Team involved a trial of different deposit limit anchors (i.e., potential limit values presented in a drop down menu to consumers when first setting their limit) with three arms: [1] a business as usual condition (with anchors ranging from £5 to £100,000), [2] anchors only as high as £250, and [3] a free text box to enter a desired limit with no visible anchors. Existing bet 365 customers about deposit limits were randomly selected to be sent prompts encouraging them to set limits between January–April 2020. 1,731 participants split roughly evenly across the 3 arms set limits. Customers in the control (arm 1) group set the highest limits (Mdn = £14.3, M = £1601), followed by those in the reduced anchor group (arm 2; Mdn = £8.3, M = £230.8), and the free-text box group (arm 3; Mdn = £7.1, M = £308.78). Customers in the 2nd and 3rd arm were slightly more likely to remove their limit in the trial period, but the difference was not statistically significant.

Coming soon!

Citation

BibTeX citation:
@online{heirene2023,
  author = {Heirene, Rob},
  title = {How Can We Prevent Gambling Harms?},
  date = {2023-09-18},
  url = {https://robheirene.netlify.app/posts/2023-gambling-harms-chart},
  langid = {en}
}
For attribution, please cite this work as:
Heirene, Rob. 2023. “How Can We Prevent Gambling Harms? .” September 18, 2023. https://robheirene.netlify.app/posts/2023-gambling-harms-chart.