Problem Gambling Among UK Racing Punters: What the Data Actually Shows

The 2.8% Number That Quietly Sits in the Background

The first time I had to explain problem-gambling data to a non-racing audience was at a journalism event in 2022. The audience expected to hear that racing was a high-harm vertical; the data said almost the opposite. The headline figure from the most carefully constructed UK study – the 2018 Health Survey for England – placed the problem-gambling rate among regular UK racing bettors at approximately 2.8%, materially below the equivalent rates for slots and casino-style online gambling. The room genuinely did not believe me until I produced the table.

That figure remains the most-cited benchmark for racing-specific harm in 2026, even though the underlying study is now nearly eight years old. No more recent vertical-specific study has produced a competing number with the same methodological rigour, and the 2.8% figure sits behind almost every industry-side and academic conversation about how UK racing fits into the broader gambling-harm landscape.

What follows is a careful read of what that number actually means, what the methodology behind it permits us to say, what it does not permit us to say, and how the 2025 GSGB sentiment data sits against the 2018 baseline. The conclusion is not what either the regulated industry or the harm-reduction lobby would write on its own – the data tells a more complicated story than either side typically presents.

The Health Survey That Anchors the Figure

The 2.8% problem-gambling rate among UK racing punters comes from the 2018 Health Survey for England, which applied a combination of validated harm-measurement instruments – the Problem Gambling Severity Index and the DSM-IV gambling-disorder screen – to a representative UK adult sample with subgroup analysis by gambling activity. The methodology is the gold standard for population-level harm measurement.

The 2.8% headline applies specifically to UK adults who reported betting on horse races during the survey period. The figure measures the proportion of that subgroup who scored above the clinical threshold on at least one of the harm instruments. The base rate is meaningful – it is not a measure of the entire UK adult population, but of the population of active racing-betting customers, which makes it directly comparable to vertical-specific rates for other gambling activities.

The instruments themselves are worth understanding briefly. The PGSI uses nine items to identify problem-gambling behaviours over the previous twelve months, with scoring thresholds that distinguish low-risk, moderate-risk and problem-gambling cohorts. The DSM-IV instrument uses ten criteria derived from the clinical diagnostic literature on pathological gambling. The 2018 Health Survey applied both instruments and reported subgroup rates on each.

The 2.8% figure represents the higher of the two instrument-derived rates rather than the lower – the figure is, if anything, conservative against the underlying data. Industry-side commentary sometimes presents the figure as 2.5% or 2.3% based on a single-instrument reading, but the 2.8% number is the cleaner read because it captures the cohort identified by either instrument rather than only by one.

The 2018 study has not been formally replicated for racing-betting subgroup analysis in the years since. The Gambling Survey for Great Britain – the GSGB, now the Commission’s flagship instrument – uses different methodological choices that make direct comparison with the 2018 number difficult. The result is that the 2.8% figure remains the operational benchmark in 2026 even though its underlying data is increasingly historical.

Racing Against Slots, Football and Lottery

The vertical comparison is where the 2.8% figure is most informative. The 2018 Health Survey produced equivalent subgroup rates across the major UK gambling activities, allowing a direct comparison between racing and the rest of the landscape. The headline comparison: racing at 2.8%, sports betting in aggregate at approximately 3.6%, online slots at approximately 8.5%, online casino games at approximately 9.2%, and lottery products at approximately 1.0%.

The ordering is meaningful. Racing sits in the middle of the harm spectrum, well above lottery and instant-win products but well below the slot-and-casino vertical. Within sports betting in aggregate, racing-specific rates appear to sit at the lower end – the broader sports-betting figure of 3.6% includes football and other in-play sports betting, both of which carry higher subgroup rates than racing on its own.

The structural explanation for racing’s lower rate compared with slots is now reasonably well established in the harm-reduction literature. Racing’s event-to-event cycle is slow – typically 30 to 45 minutes between races – which limits the speed at which losses can compound. The product itself requires meaningful cognitive engagement (reading form, understanding place terms, evaluating jockey-trainer combinations) that creates a natural friction against compulsive re-betting. The combination of slow cycle time and high cognitive engagement is the structural defence against rapid harm progression.

Slots and online casino games sit at the opposite end. Cycle times are measured in seconds rather than minutes; cognitive engagement requirements are minimal; outcome variance is high and unpredictable in ways that drive repeat-play behaviour. The harm rates in those verticals reflect those structural features.

The comparison should be made with a single caveat: the cohort that engages with each vertical is not the same cohort. Customers who choose to bet on horses are demographically and behaviourally different from those who choose to play online slots, and some of the rate difference is explained by self-selection rather than by vertical structure alone. A customer at high pre-existing risk of gambling harm is more likely to choose slots than racing as their primary activity, biasing the headline rates downward for racing and upward for slots in ways that pure vertical structure does not fully explain.

How 2025 GSGB Sentiment Sits Against the 2018 Figure

The 2025 GSGB does not produce a directly comparable harm rate for racing because its instrument design differs from the 2018 Health Survey. What it does produce is sentiment data about specific recent gambling experiences. Across the 2025 racing-betting cohort, 42% described their last gambling experience as positive, with the balance distributed across neutral and negative responses.

The 42% positive figure is consistent with the lower harm rate identified in the 2018 study. If racing-betting were a high-harm vertical at the slots-and-casino level, we would expect the sentiment distribution to skew much more negatively than it does. A 42% positive response rate on a recent gambling experience is structurally inconsistent with a cohort that is, in the main, experiencing harm – even allowing for the partial cohort overlap between recreational and harm-experiencing customers.

The 2025 sentiment data also captures a meaningful neutral cohort of approximately 35%, which is the harder cohort to interpret. Neutrality could indicate either contentment or disengagement, and the survey instrument cannot reliably distinguish between the two. What it can say is that explicit negative responses are a minority – somewhere in the low-twenties percent of the racing-betting cohort – which is broadly consistent with the 2018 harm-rate finding.

The 2025 data does not refute the 2.8% problem-gambling figure; it provides indirect corroboration that the underlying harm rate has not shifted dramatically since 2018. The composition of the harming cohort may have changed – different demographics, different specific behaviours – but the aggregate rate appears stable at the population level. This is consistent with the broader stability of UK racing-betting participation rates over the same period.

What the 2025 data does suggest is that the wider regulatory environment – affordability checks, enhanced verification, deposit-limit prompts – has not measurably reduced the harm rate even while it has measurably increased the friction in the regulated customer experience. The trade-off between consumer protection and customer friction is not delivering proportionate reductions in measured harm, at least on the racing-betting subgroup data currently available.

The Caveats No Punter Should Skip

The 2.8% headline figure has limitations that any honest analysis has to acknowledge. The most important caveat is that the 2018 data is now eight years old, and meaningful changes in the structure of UK racing betting have occurred since – the shift from shop to phone has accelerated, the in-play market has expanded, and the operator landscape has consolidated around fewer, larger players. Whether the underlying harm rate has moved through those structural shifts is genuinely uncertain.

The second caveat is the population-level versus individual-level distinction. The 2.8% rate is a population-level statistic that says nothing about any individual customer’s risk profile. A casual once-a-year Grand National punter and a multi-meeting daily customer have very different risk profiles, and the headline figure averages across both. Individual risk depends on personal factors that population-level data cannot capture.

The third caveat is the subgroup definition itself. The 2018 study defined the racing-betting subgroup as anyone who reported betting on horses during the survey period, which captures both regular customers and one-off casual customers. The harm rate among regular daily customers is likely materially higher than among casual annual customers, and the headline figure obscures that internal heterogeneity.

The fourth caveat is the structural feature of each-way and place-only betting. The harm-rate data does not distinguish between win-only customers and customers who primarily place each-way slips. The structural argument that each-way and place-only betting carries a different harm profile from straight win betting is plausible – the lower variance of place outcomes may reduce the chasing-loss dynamic that drives harm progression – but the data to confirm or refute the hypothesis is not currently available at the necessary granularity.

The fifth and final caveat is the methodological gap between PGSI and DSM-IV instruments. The two instruments do not always identify the same individuals as problem gamblers, and the choice of which instrument to use materially shapes the headline rate. The 2018 study reported both rates separately and the 2.8% headline reflects the higher of the two; users of the data should be aware that single-instrument readings produce slightly different numbers.

The honest summary for the regular racing punter is that the population-level harm rate is lower than most public commentary suggests, but the data does not give individual customers a clean read on their personal risk. The structural defences in racing – slow cycle time, high cognitive engagement, varied outcome distributions – work at the population level but cannot substitute for individual self-awareness and account-control discipline. The tools that customers can actually use to manage their own risk are worth understanding in detail, and I’ve worked through the practical account-control walkthrough in the piece on setting betting limits in the UK.