The use of personalized messages on wagering behavior of Swedish online gamblers: An empirical study

Highlights

• Personalized messages (PMs) are used to prevent online gamblers over-spending money.

• This study evaluated the efficacy of targeted PMs among 7134 online gamblers.

• Gamblers bet significantly less money on the day they read a PM.

• Gamblers bet significantly less money seven days after they read a PM.

• PMs appear an effective tool in reducing gambling expenditure at online websites.

Abstract

Over the past few years, online gambling has become a more commonplace leisure time activity. However, for a small minority, online gambling can become problematic. Consequently, the gambling industry has started to acknowledge their role in player protection and harm minimization and some online gambling companies have introduced responsible gambling tools such as targeted personalized messages as a way of helping players stay in control. The present study evaluated the effectiveness of targeted messages among 7134 Swedish online gamblers who played at one of five sites within the ComeOn Group between July 2019 and January 2020. The results showed that online gamblers receiving personalized feedback (i.e., feedback concerning their own actual gambling behavior in the form of text messages) wagered significantly less money on both the day they read a personalized message and seven days after they read a personalized message. The data support the results found by previous laboratory and real-world studies showing that targeted personalized information can be an effective tool for online gambling companies to reduce gambling expenditure among their clientele. The findings will also be of interest to other stakeholders including gambling regulators, policymakers, and researchers.

Introduction

Online gambling has become increasingly commonplace in many countries since its inception in the late 1990s. While most individuals gamble without any problems, a small minority within most populations have a gambling problem (Calado & Griffiths, 2016). For susceptible and vulnerable individuals, there are many situational and structural characteristics that can make gambling via the internet potentially risky including 24/7 accessibility, convenience, anonymity, and high event frequency (McCormack & Griffiths, 2013). Some types of online game appear to be more problematic for individuals than others such as online casino games and online sports betting (particularly in-play sports betting) (Killick & Griffiths, 2019; Wardle, Moody, Griffiths, Orford, & Volberg, 2011).

Studies have consistently shown that compared to land-based gambling, there is typically a higher prevalence rate of problem gambling among those that gamble via the internet (e.g., Gainsbury, Russell, Hing, Wood, & Blaszczynski, 2014a; Griffiths & Barnes, 2008; Griffiths, Wardle, Orford, Sproston, & Erens, 2009; Wood & Williams, 2011; Wood, Williams, & Lawton, 2007). However, most land-based gamblers also gamble online (Wardle et al., 2011). Furthermore, the severity of problem gambling is associated with overall gambling engagement. However, Philander and MacKay (2014) found that gambling via the internet is not a predictor of problem gambling when the volume of gambling is controlled for.

The present study was conducted with Swedish online gamblers. Therefore, a quick overview of the Swedish market is presented in this section. A report by Folkhälsomyndigheten (2015) noted that among the gamblers who called the national problem gambling helpline (and for whom information on the main form of gambling causing problems was recorded), 43% specifically had problems with online casino games, and a further 10% had problems with online poker, and 13% had problems with online sports betting. Abbott, Romild, and Volberg (2018) reported findings from a Swedish longitudinal study with a stratified random sample of 8165 participants (aged 16–84 years at baseline) and re-assessed a year later (n = 6021). They found that utilizing the Problem Gambling Severity Index (PGSI), combined current problem and moderate-risk gambling prevalence rates were 2.2% at baseline and 1.9% at follow-up. Combined incidence rates (IRs) were 1.0% (with the revised South Oaks Gambling Screen [SOGS-R]) and 1.4% (with the PGSI), with more than three-quarters being new cases. Widinghoff and Håkansson (2018) reported that the national prevalence of problem gambling in Sweden was 2%, including 0.4% meeting the criteria for gambling disorder. They also reported that incidence numbers were substantially higher due to the dynamic pattern of the disorder with individuals moving into and out of the problem gambling group. Rapid internet games, such as internet casino games and online sports betting, comprised a predominant share of problem gambling.

It should also be noted that the Swedish gambling market changed at the start of January 2019 because the Swedish monopoly changed into a license-based market. There were several reasons for this but according to a government report (Staten Offentliga Utredningar, 2017), a major reason was that international online gambling companies had been operating in Sweden to the extent that the monopoly system had eroded. Furthermore, the Swedish Gambling Act is one of very few acts of legislation which specifically requires licensed operators to counteract problematic gambling through continuous monitoring of gambling behavior (Ministry of Finance, 2018).

Over the past decade, responsible gambling tools have become more utilized by gambling operators in an attempt to help their online clientele gamble more responsibly (Harris & Griffiths, 2017). Such tools include various types of direct messaging to gamblers which can include general messaging concerning on how to gamble more responsibly, information about the gambler's actual gambling behavior in-session and/or over time, or information about erroneous perceptions and common misbeliefs about gambling (Auer, Hopfgartner, & Griffiths, 2018). However, studies investigating the efficacy of such tools have been mixed, especially those concerning messaging that attempts to correct or change erroneous beliefs (Dixon, 2000; Focal Research, 2004; Ladouceur and Sevigny, 2003; Hing, 2003; Williams & Connolly, 2006).

Some empirical studies have shown that education and prevention programs targeting erroneous gambling beliefs can help both adult and adolescent gamblers (e.g., Calado, Alexandre, Rosenfeld, Pereira, & Griffiths, 2019; Wohl, Christie, Matheson, & Anisman, 2010; Wulfert, Blanchard, Freidenberg, & Martell, 2006). Empirical studies (mainly experimental laboratory-based research) highlight that the way targeted messages are presented can also influence gamblers’ behavior and thinking. For instance, animated, interactive and/or pop-up messaging and information appear to be more effective in changing both irrational belief patterns and gambling behavior than static messaging (e.g., Cloutier, Ladouceur, & Sevigny, 2006; Ladouceur & Sevigny, 2003; Monaghan & Blaszczynski, 2007 & 2010a; Monaghan, Blaszczynski, & Nower, 2009), Schellink & Schrans, 2002; Stewart & Wohl, 2013; Wohl, Gainsbury, Stewart, & Sztainert, 2013) and messaging on slot machines that includes the capacity for gamblers to engage in self-appraisal and self-regulation help change gambling thoughts and behavior (Monaghan & Blaszczynski, 2010a, b). Use of graphic messaging has also been reported as being more effective for gamblers than static messaging in adhering to gambling warning signs (Munoz, Chebat, & Borges, 2013). Wohl, Parush, Kim, and Warren (2014) found that messaging systems employing Human Computer Interaction (HCI) and Persuasive System Design (PSD) led to a pop-up tool (for monetary limit-setting) being significantly more effective than a tool that not incorporating HCI and PSD principles.

However, two recent experimental studies by Hollingshead, Wohl, and Santesso (2019) with slot machine gamblers (n = 124; n = 109) who played on a virtual slot machine at a local Canadian gambling venue found that players did not adhere to a pre-determined limit more often when they received a pop-up message about their monetary loss. Additionally, approximately 50% of players were unable to recall the content of the pop-up message, even when the pop-up message remained on the slot machines for a 10-s period.

In addition to experimental research showing that messaging can effectively change thoughts about gambling and the gambling behavior itself, research has also suggested that the content of messages is important (Monaghan & Blaszczynski, 2010a, b). In a focus group study with young adults (18–24 years), seniors (60+ years), frequent (weekly) gamblers, and gamblers of skill-based games (poker, sports betting), Gainsbury, Abarbanel, Philander, and Butler (2018) found that the wording of message content also influences the effectiveness of messages. Findings showed that seniors preferred messages concerning limit setting, whereas young adults and frequent gamblers responded more positively to messages concerning their own play and expertise. Skill-based game gamblers were more interested in the odds of winning and their own outcomes over time.

One of the limitations of all the aforementioned studies is that they were experiments comprising very small sample sizes. However, there have now been a number of real world studies examining the effects of messaging with real gamblers, in real time, on real gambling sites. For example, two studies evaluated the efficacy of pop-up messaging. The first was by Auer, Malischnig, and Griffiths (2014) who examined the efficacy of a pop-up message among players at the win2day website who played online slot machine games. The pop-ups only appeared if gamblers had played 1000 consecutive games during a single gambling session (approximately 1 h of slot machine playing). The study examined 200,000 playing sessions prior to the pop-up's introduction and 200,000 playing sessions after the pop-up's introduction. Auer et al. reported that less than 1% of players stopped gambling after seeing the message and concluded that pop-up messaging has a limited effect among a small minority of players.

Auer and Griffiths (2015a) carried out a follow-up study (again using data from gamblers at the win2day website), and examined the efficacy of a newly designed pop-up message which included normative and self-appraisal information. This enhanced message was compared to the simple (i.e., non-enhanced) pop-up message that was examined in the previous study by Auer et al. (2014). The follow-up study was much larger and examined 1.6 million online slot machine sessions and compared two conditions (the enhanced pop-up message vs. the simple pop-up message) comprising two representative random samples of 800,000 gambling sessions. Auer and Griffiths reported that the newly designed (‘enhanced’) pop-up message was twice as effective in getting gamblers to cease their online slot machine playing (1.39% vs. 0.67%). However, like the previous study, the efficacy of pop-up messaging was limited and only facilitated a minority of online slot machine gamblers to stop their in-session gambling.

A study by the Behavioural Insights Team (2018) tested the effect of electronic messages on the frequency of using RG tools in a sample of online gamblers. They found that messages which contained the link of the RG tool increased the number of players who used the RG tool. However, normative feedback, which compares a player's time and money spent with other players did not change the usage of RG tools.

In another study, Auer and Griffiths (2015b) evaluated the efficacy of mentor (i.e., a behavioral tracking tool that provides personalized messages to players based on their actual gambling behavior). The study investigated 1358 gamblers who had voluntarily signed up to use mentor at an unnamed European online gambling website. Using a matched pairs design they compared players who used mentor with players who did not use it. The study found that gamblers who used mentor and received personalized messages spent significantly less time and money gambling compared to the gamblers who did not.

The only experimental study regarding the effects of personalized feedback in a real-world setting was conducted by Auer and Griffiths (2016) with players from the state owned Norwegian gambling operator Norsk Tipping. A total of 5528 online players participated in an experiment and received a combination of personalized and normative feedback about the amount of money that they had recently lost gambling. They found that personalized behavioral feedback enabled behavioral change in gambling but that normative feedback did not change gambling behavior significantly more than personalized feedback.

The present study examined the efficacy of personalized feedback (i.e., feedback concerning their own actual gambling behavior in the form of digital text messages) given to players after they logged on to a gambling session via a pop-up window. More specifically, the aim was to investigate the effects of personalized feedback about past gambling behavior on future gambling. It was hypothesized that gamblers receiving targeted personalized feedback about their online gambling behavior would be more likely to change (i.e., reduce) their behavior (as measured by the amount of money wagered) compared to before receiving the feedback. This study also aimed to confirm the findings of previous similar studies (e.g., Auer and Griffiths [2015b; 2016]) which found that personalized feedback led to a significant reduction in gambling expenditure.

Except for the two studies by Auer and Griffiths (2015b, 2016), no other study has investigated the effect of text messages informing players about their own behavior and providing specific recommendations. Very few previous studies have been conducted with real-world players on real gambling sites. Two previous studies with real-world players conducted by Auer and Griffiths (2014, 2015a) investigated pop-up messages which appeared after 60 min of consecutive play and informed players that they had played 1000 consecutive slot machine games (equating to approximately 1 h's continuous play). In the present study, players were informed about specific aspects of their own gambling behavior and were provided with recommendations that could help change their behavior.

The present study used the amount wagered as a proxy for gambling intensity. However, problem gambling was not directly measured. Braverman, LaPlante, Nelson, and Shaffer (2013) compared problem and non-problem gamblers’ online wagering and found that problem gamblers had a higher average wager amount. Several other studies have found correlations between amount wagered and problem gambling (e.g. Boldero, Bell, & Moore, 2010; Clarke, 2008). The present study was conducted with gamblers from Sweden. As far as the present authors are aware, no previous studies examining personalized messaging have ever examined Swedish gamblers. It is important to test responsible gaming tools across different cohorts in different locations and languages to evaluate whether laboratory results also hold true in real-world settings.