Britain hosts one of the largest regulated gambling markets in the world — worth £14.2bn in 2019/20. A key component of this regulation involves the reduction of gambling harms. To be effective, harm reduction strategies need to be informed by an understanding of how gamblers really behave. However, much of the existing research is limited to insights from self-report methods that risk misrepresenting actual behaviour.
A benefit of bank transaction data is that it can offer a more objective sense of peoples’ gambling behaviour, including more accurate records of how much people deposit into online gambling accounts, and how frequently.
We were commissioned by GambleAware to further explore what bank transaction data can tell us about gambling behaviour — a core aim of the Gambling Commission’s (GC) National Strategy to Reduce Gambling Harms.
We partnered with HSBC UK, the UK’s largest bank and Monzo, one of the UK’s fastest growing digital only banks, to explore the relationship between customer transaction data, such as income, everyday spendings and savings records, and gambling habits.
Both banks used slightly different approaches to identify gambling customers from their transaction records. Monzo shared the daily transaction records for ~10,000 anonymised customers, spanning an average of 10 months’ worth of data per customer between May 2018 – November 2019. 11% made at least one gambling-related transaction during this period, with an average monthly gambling spend of £136. We categorised all customers into one of three groups: ‘non-gamblers’, ‘below-average gamblers’ (<£136 per month), and ‘above-average gamblers’ (>£136 per month).
Meanwhile, HSBC UK identified 1.5m customers who had made at least one gambling transaction between March 2019 and February 2020 (representing ~19% of their customer base). Transaction data for these customers was drawn from March 2016 – February 2020. HSBC UK categorised these customers into three groups according to gambling spend relative to disposable account income: ‘in-control’, ‘concerning’ and ‘very concerning’.
We compared the financial behaviour of the three Monzo and three HSBC UK groups across four research themes. Overall, we found that transaction data has a valuable role to play in predicting, identifying, and helping to mitigate gambling harms.
Our key findings
Theme 1: Gambling behaviour
Monzo’s data indicated that above average gamblers spent £684 per month and averaged almost one gambling transaction per day, compared to below-average gamblers’ who spent £18 per month, over an average of one transaction. For all gamblers, gambling activity tended to increase between Thursday and Saturday of each week.
Similar trends were found among HSBC UK customers. The ‘very concerning’ group spent the most on gambling, at an average of £2,202 per month (vs. ‘concerning’ = £574, ‘in-control’ = £17). This is reflected in the average number of monthly transactions (‘very concerning’ = 36, ‘concerning’ = 16, ‘in-control’ = 1) and the average number of operators used each month (‘very concerning’ = 6, ‘concerning’ = 4, ‘in-control’ = 2).
Gambling behaviour varied markedly across the customers in our samples. This casts doubt over the benefits of ‘one-size-fits-all’ approaches to gambling controls, and suggests that more targeted and flexible options might be beneficial.
Theme 2: Spending behaviour
Across Monzo’s groups, above-average gamblers had much higher monthly outgoings overall. Gambling transactions accounted for ~50% of their monthly spend (vs. only 3% for below-average customers).
From HSBC UK’s data, we found as disposable income and average monthly spend on non-essentials increased, the magnitude of gambling spend also increased (with all three highest among the ‘very concerning’ group).
The proportion of disposable income spent on gambling was also over 3 times higher among HSBC UK’s ‘very concerning’ (58%) vs. ‘concerning’ (19%) gamblers. Despite this, the ‘very concerning’ gamblers were the only group to end an average month with surplus income.
This raises additional considerations for how best to measure the impact of gambling on a customer’s financial wellbeing. Intuitive metrics, such as the ratio of gambling deposits to disposable income, may not necessarily be good predictors of gambling harm (see also theme 3.2).
Read the full reports here:
Part 1 (Monzo)
Part 2 (HSBC UK)
Theme 3.1: Saving behaviour
Monzo’s below-average gamblers contributed 42 times as much money into interest-earning savings accounts than they spent on gambling, while above-average gamblers contributed only 0.1 times as much as they spent on gambling.
Positive frictions could be leveraged to encourage saving behaviour among those who gamble. For example, when gambling transactions go above a given level, customers could be served a prompt asking if they would prefer to save the deposit instead.
Theme 3.2: Credit use
HSBC UK’s higher gambling groups were more likely to have arranged overdrafts, but the proportion spent on credit by each group in an average month was very similar.
Use of unsecured lending over time was a mixed picture between the groups. However, higher proportions of customers in the ‘very concerning’ segment showed movement upwards into higher bands of unsecured lending 12 months later, vs. ‘in control’, and ‘concerning’ customers.
In conjunction with our findings under Themes 1 and 2, this suggests that classifying the potential severity, or risk of gambling behaviour, may benefit from both a short- (e.g. a rolling monthly risk rating), and a longer-term (e.g. an 8 – 12 month rating) view.
Theme 4: Gambling block use
Within Monzo’s data, one third of above-average gamblers lifted their gambling block for longer than 30 days, compared to a quarter of below-average gamblers. During the week before gamblers activated the block, average daily gambling spend tripled, from an average of £6.90 to £22.90.
Gambling block functions currently offered by banks remain relatively simple to turn off, and offer little flexibility in how they operate. Our findings point to a role for more flexible block functionality, such as allowing customers to block gambling transactions for certain periods of time.
Our work is insightful, but must be considered in light of its limitations. For example, records from a single bank will not reflect an individual’s gambling spend if they gamble using another account, use other electronic payment methods to gamble (such as e-wallets), or cash. Further, analysis of data from a single bank reflects only the kinds of transactions an individual makes using that account. Nonetheless, we are encouraged by our findings and the potential role for bank transaction data as a tool for safer gambling research.
Read the full reports here: