Big data: Study suggests even a moderate gambling habit is linked to increased mortality and other bad stuff
R, me hearties: Boffins sift through 6.5 million individual dataset
Banking data from 6.5 million individuals over six years is providing evidence that even moderate gambling can be associated with financial distress, negative lifestyles, and an increased likelihood of death.
Research published in Nature Human Behaviour showed gambling is found alongside higher rates of future unemployment and physical disability and, at the highest levels, with substantially increased mortality.
The study is one of the first to document gambling in the United Kingdom with large-scale objective data.
Having worked for Lloyds Banking Group while holding an honouree (unpaid) affiliation with Warwick University, paper co-author Dr Naomi Muggleton gained access to the bank's data sets.
After working through ethical and legal processes, a data set of around 7 million anonymised records was selected from the bank, one of the largest UK financial institutions with 30 million customers and 65,000 employees.
Muggleton, now a research fellow at Oxford University, extracted the data from the bank's Teradata data warehouse using the SQL Assistant and performed the statistical analysis in R (those interested can have a poke around the scripts used here, though all commercially sensitive code has been redacted).
"Extracting the data in an efficient way that didn't break in spool was challenging," she told The Register. "Once the data was extracted it was all aggregated at a very high level, the analysis was pretty straightforward. We deliberately wanted to have quite simple statistical tests to be broad and not make the paper about the type of analysis."
The researchers identified gambling transactions by relying the pre-existing category in the bank's typography of transactions, which includes various forms of gambling such as offline and online bookmakers, casinos, lotteries and other providers.
It did not cover cash gambling transactions, or electronic transactions using third-party payment processors and as such might provide a conservative estimate. The study also included pre-existing markers for spending associated with health and lifestyle.
The approach differs from other gambling studies. Some have relied on social scientific surveys, which rely on people self-reporting their gambling behaviour, data that can be skewed by remembering biases. Gambling firms also study behaviour among gamblers, but only have access to their own data sets.
Not two types of gamblers: it's a continuum
Using a new approach, the study reveals some worrying trends with the top 10 per cent of gamblers spending more than £1,800 a year, nearly 8 per cent of their total spending. But it was evidence in the long tail of gamblers that struck the researchers.
"Traditionally we've thought that there is a small pot of individuals who are labelled as 'problem gamblers' and there's harm there, and for the vast majority of people there's no harm associated with gambling," Muggleton said. "But what we find here is that there's more of a continuum. We find that harm is associated with gambling and even [at] quite low levels. Rather than seeing the sort of two camps of gamblers, it looks as though there is a more of a kind of a gradient, which is something that I don't think we necessarily knew before we did the analysis but it was something that emerged from it."
The finding should be of interest to anyone trying to reduce the harm caused by gambling, said Neil Stewart, Warwick University professor of behavioural science and paper co-author. "Suppose one took the 1 per cent of people in the UK who gambled the most and wave the magic wand and cured them of their gambling, how much of the harms that we see in our data would be reduced?
"If it were the case that they really were just a very small fraction of problem gamblers and everybody else was fine, then you'd pretty much fix all of the harm. In fact, our data shows you wouldn't have reduced it by very much at all."
But the researchers cautioned against any conclusion suggesting gambling is a direct cause of the associated harms. "All we know is that the type of person who gambles this much is the type of person who has these other financial circumstances and wider circumstances around them, and that's important for policy decisions," Stewart said.
For example, banning gambling advertising during football matches might not reduce harm if gambling is not the root cause, he said. Other factors at play might be general attitudes towards risk-taking, or socio-economic class, but the study is silent on these factors.
Nonetheless, the researchers hope it will contribute to debate around the gambling industry, which is worth around £14bn annually. The Department for Digital, Culture, Media and Sport is currently conducting a review of the Gambling Act 2005, which saw a significant shift in attitudes towards the pastime.
"A lot of the work that's been done has been involves getting self-reports or using smaller samples in [surveys]," Muggleton said. "I think we really wanted to contribute a more objective view, less based on people's recollections, to contribute this to the evidence base and allow policy-makers to use as they see fit."
Stewart said another motivation was to show that data not collected for the express purpose of social science or psychology could have value in these fields.
"In the olden days, if I wanted to know stuff about you, we would have to get you on the phone or get you into the laboratory. That's just a really slow way of getting data: your bit-rate is going to be terrible unless you've got a huge interview team and that's incredibly costly. Whereas, now, we can get data for millions of people that is just sitting there. It's intrinsically interesting to see how we do behavioural science and psychology with this newly available data." ®