Who bets on sports? Some further empirical evidence using German data

Abstract

Given the rapid expansion of the German sports-betting market and recent changes in market regulations, it is interesting to reexamine the socioeconomic profile of German sports bettors: Who bets on sports? In order to analyse this question, this study used an online survey to collect data on sports-betting behaviour (N=634). It modelled participation in sports betting by means of a logit model that recovers determinants of a person’s propensity to participate in sports betting. Results show that the typical sports-bettor is 32 years old and male, has a low household income, is highly interested in sports, and is willing to take risks.

Keywords

gambling gambling market games of chance logit regression sports bets

Introduction

Elias and Dunning (2003: 172) argued that games of chance in general and sports betting in particular represent leisure-time activities that offer a quest for excitement insofar as they trigger the recurrence of ‘alternating short moments of fear with happy moments of hope’ (Elias and Dunning, 2003: 195). Historically, sports betting with fixed win quotas originated in Great Britain and Scandinavia. Systematic betting on soccer started in Great Britain as early as 1921 (Mintas, 2009: 26), while betting on soccer was advanced in Germany only after the end of World War II, when sports associations initiated the establishment of state soccer pools companies in 1948 (Endes and Feldner, 2012; Mintas, 2009). The German state soccer pools had been the only sports-betting suppliers for a long time. Addiction prophylaxis was the official justification for the rigid regulation of the German sports-betting market; the so-called Oddset sports bet was introduced in Germany only in 1990. The Oddset sports bet, organized by the German lottery and pools block (DLTB), was mostly offered by private organizers from abroad via the Internet and also by the owners of an old GDR (German Democratic Republic) permit (Mintas, 2009: 27), which was still valid after German reunification.

While in the period 2008–2011 the state had a monopoly on sports betting, an amendment to law (Glücksspielstaatsänderungsvertrag) made in 2012 opened up the market for private providers, but up to now no licenses have been assigned (Rebeggiani and Breuer, 2015). National and international private providers with old GDR and foreign licenses currently offer sport bets and pay a voluntary tax rate of 5 per cent on the betting amount, which was implemented as a transitional system, and private providers accept this system to avoid future disadvantages in case licenses will be distributed (Rebeggiani and Breuer, 2015).

In the more recent past, the introduction of new media technologies and the widespread adoption of the Internet laid the foundation for a strong expansion of the market for sports bets in Germany (Endes and Feldner, 2012). Impressive revenue growth rates of almost 300% between 2004 and 2005 document this rapid growth of the German sports betting market (Bendixen, 2008), and technological progress may lead to a further growth in the coming years (Spöring, 2012).¹

Given the rapid expansion and regulatory changes of the German sports betting market, it is natural to reexamine what is known about the socioeconomic profile of bettors: Who bets on sports? In order to study this question, this study surveys the literature on (sports) betting markets in Section 2. Section 3 describes the data on sports betting behaviour that was collected by means of an online survey. The results of a quantitative analysis of the data are documented in Section 4, and the study’s main results are summarized in Section 5, which also offers some concluding remarks.

Literature review

Researchers have analysed: (1) the regulation, fiscal effects, and distributional effects of betting markets (on the monopoly on games of chance in Germany, see Albers, 1993: Chapter 3); (2) the implications of sports betting for the propensity to get addicted to sports betting and gambling (see Bloch, 1951; Kearney, 2005; and for Germany: Beckert and Lutter, 2007; 2008); and (3) the determinants of players’ decision to participate in betting

As for the regulation of betting markets, Reiche (2013) compares cultural with social motives that led to the imposition of state regulations on games of chance in both the United States and Germany. Beckert and Lutter (2008, 2009) argue that lotteries have three fiscal and distributional effects (2009: 475): (1) by participating in a lottery, players agree that there will be few winners and many losers, i.e. a certain proportion of the stake of a player is paid out to other players following the principle of chance; (2) only part of the gaming turnover (approximately 48%) is paid out to winning players, where part of turnover covers operating costs (approximately 13%) and taxes or concession levy (39%) that have to be paid to the state; and (3) state lotteries are an example of regressive taxation (Beckert and Lutter, 2009: 2483), as the lower classes contribute a disproportionately large amount to tax income from lotteries while tax revenues are used to finance projects often benefitting the upper classes (Beckert and Lutter, 2009: 482). Blalock et al. (2007), Borg et al. (1991) and McKinney and Swain (1993) also note the regressive-taxation effect of lotteries.²

Abarbanel (2012) compares 64 countries with respect to sports-betting policies, where the focus is on the influence of cultural predictors as measured by Hofstede index scores (Uncertainty Avoidance Index, Individualism, Power Distance Index and Masculinity). Abarbanel differentiates between governments that actively allow, passively allow, actively prohibit, and passively prohibit sports bets. The Uncertainty Avoidance Index and Individualism were the best predictors of policy types. The Uncertainty Avoidance Index helped to discriminate between ‘allow’ and ‘prohibit’ policies. The Individualism Index, in turn, was higher for the ‘actively allow policy’ countries than the ‘actively prohibit’ countries.

In order to measure addiction effects, the proportion of players in a population and the frequency with which they bet are commonly used as indicators for addictive behaviour. LaBrie and Shaffer (2011) identify patterns of sports gambling that distinguish ‘sports bettors with self-reported gambling-related problems from sports bettors without such difficulties’ (2011: 56). The authors use a ‘Bwin’ longitudinal dataset that includes data on betting behaviour of 47,134 individuals for a 2-years period. Their results show that half of the account closers with self reported gambling-related problems show a ‘homogeneous and distinct pattern of sports-betting behavior’ (LaBrie and Shaffer, 2011). These bettors made more and larger bets, bet more frequently, and were more likely to exhibit intense betting.

As reported by the German Federal Centre for Health Education (BZgA), 70% of the German population played some kind of lottery in their lives, where 11.1% of citizens stated that they had bet on sporting events at least once in their life and 3.4% stated that they had taken part in bets during the last year. Bets with the state office Oddset were most frequent and only a small proportion of respondents answered that they had bet using the Internet (BZgA, 2012). In contrast, Daumann et al. (2011) report results of a questionnaire study that suggest that betting using the Internet is quite common in Germany, and that many players have reservations about Oddset’s bet products. According to Lutter (2010a: 22), 40% of adult Germans participate in a lottery at least once a year, while the 12-month prevalence of games of chance using casinos, slot machines, sports bets, and Internet casinos varies between 3 and 5.8% (see Beckert and Lutter, 2009: 237).

LaBrie et al. (2007) show, using data records from the sporting bet provider ‘Bwin’ (N = 40,499), that players made an average of 2.5 bets with an average stake of €4 every four days in four months.³ According to Lutter (2010a: 22), 22% of those players who made sports bets at least once a year (1.3% of the total population) tended to bet every week.

As for the factors that determine betting participation, Browne and Brown (1994: 344) find that American college students play lotteries more frequently and spend more money if their parents or members of their peer group play lotteries, whereby the influence of the peer group is stronger for males than for females.⁴ Beckert and Lutter (2008) also consider the possibility that the integration into a peer group in which betting is common goes along with a greater possibility that a person participates in a lottery. Moreover, they report that three additional factors help explain why people play lotteries (Beckert and Lutter, 2008: 240f.). First, players may estimate incorrectly their chances to win a positive amount of money. In this respect, players’ level of education may matter (also see Eadington (1988: 270) and Blalock et al. (2007: 547) on games of chance and rational behavior). Second, even with the expected payoff being negative betting can still be rational because there is a small chance of winning a large amount of money with only a small amount of money being at stake (Beckert and Lutter, 2008: 250ff.; see Albers, 1993: 21; Blalock et al., 2007: 551, Eadington, 1987: 267, on Friedman-Savage paradoxes).Third, upon buying a lottery ticket, players buy the possibility of building pipe-dreams (Beckert and Lutter, 2008: 254). Pipe-dreams may be the result of poor economic conditions. According to Blalock et al. (2007: 546), poor economic conditions lead to acts of desperation and can lead to participation in games of chance as a possibility to escape from economic constraints. Using macroeconomic data for US states, they find that sales of lottery tickets increase as the rate of poverty rises.⁵ Similarly, Kearney uses data for 21 US states to show that household expenditure on lottery tickets comes from expenditure that is not gambling related, and that expenditure on lottery tickets is particularly high for low-income households (2005: 2295).

Breuer et al. (2009) analyse sports-betting financial products. Usually, sports betting products consist of a fixed base interest rate and an uncertain bonus payment, which depends on the outcome of a predefined sporting event. According to the authors, there are ‘theoretically several potential reasons which may make this instrument an attractive investment’ (Breuer et al., 2009: 2252). They performed a computer-based experiment with 385 students and found that ‘non-monetary utility components may be only of indirect importance by influencing probability assessments’, and conclude that sport-betting financial products are mere marketing gimmicks.

Lin and Lu (2015) explore psychological and sociodemographic factors influencing participation in sports lotteries. They select 100 sports-lottery-betting stations in Taiwan and perform face-to-face interviews with bettors (N = 1032). The authors report evidence of “herding behavior”⁶ and a gender difference in sports lottery betting (Lin and Lu, 2015: 118). They show that bettors with neuroticism have lower risk tolerance.

Paul and Weinbach (2010) analyse the determinants of betting volume of the National Basketball Association and National Hockey League for the 2008–2009 season. According to these authors: ‘Bettor behavior appears closely tied to fan behavior’ (Paul and Weinbach, 2010: 137). Furthermore, the quality of teams and the availability of television coverage have a positive effect on betting volume. Paul and Weinbach conclude ‘that consumption plays a major role in the decision to gamble on sports’ (2010: 128).

Welte et al. (2002) use a national telephone survey with 2,630 representative US residents to describe demographic patterns of gambling participation. Possibly due to a small sample size, they aggregate data on betting on cards, games of skill (e.g. pool and golf), and sports betting into the category games/sports. They report that, particularly with regard to sports betting and games, male bettors gamble more frequently and have larger wins and losses than female, where males are more interested in sports than females. They also find that it is more likely that individuals with higher socioeconomic status bet than individuals with a low socioeconomic status. Blacks have a lower participation rate, but those blacks who do bet have a higher level of involvement (Welte et al., 2002: 328f).

Results reported by Wicker and Soebbing (2012) for German data collected by means of an online questionnaire indicate that gamblers who participated in sports betting in 2010 are male with a high income, have a low level of education, are of non-German origin, have an affinity for other games of chance, and do not do own sporting activities.

In an international comparison of data for Great Britain and Canada, Humphreys and Soebbing (2012) show that the probability of participating in sports bets decreases with increasing age. Males are more likely to participate in sports bets than females, and the probability of participating in sports bets increases when household income is low and household members are employed. The average age of betting participants is approximately 44 in Great Britain and approximately 35 in Canada.

LaBrie et al. (2007) characterize the typical online sports gambler as male, about thirty years old, and of European descent. In a comparison of active athletes, former athletes and non-athletes, Weiss and Loubier (2010) show that current and former athletes participate to a greater extent in games of chance and especially in sports bets than persons that have not been active in sports.⁷ The tendency to play games of chance seems to be higher for former athletes than for other groups. Moreover, former athletes seem to prefer to play games of chance that they think they can influence by their own skills (e.g. sports bets).

In sum, results of earlier studies suggest that the propensity to participate depends on several sociodemographic factors: males tend to bet more frequently than females, and gender, income, and age appear to play a role for betting behaviour. Moreover, bettors’ interest in sports alongside their own sporting activities has also been analysed. Attitudes towards risk and a tendency to bet (Lutter, 2010a) may also play a role. Taken together, however, the available empirical evidence on how exactly the likelihood that an individual participates in sports bets is linked to these sociodemographic factors is at least in part mixed. For example, against the background of the available empirical evidence, it is unclear as to whether income has a positive or a negative effect on the propensity to participate in sports bets. Some researchers report evidence that the desire to escape from economic constraints induces individuals with a low income to participate in sports betting, while other researchers find that participation in sports bets is positively linked to income. Similarly, while sports betting appears to be more prevalent among males and young persons the effect of socioeconomic status remains unclear. Moreover, some results reported in earlier literature indicate that own sports activities increase the likelihood to participate in sports betting, while other researchers report that sports bettors are rather passive with regard to their own sports activities.

As far as results for Germany are concerned a further concern is that empirical evidence is rather scarce and based on relatively small datasets (Daumann et al., 2011; Wicker and Soebbing, 2012). The fact that researchers could analyse only small datasets should not be viewed per se as a drawback of earlier research on sports betting in Germany given that collecting data on sports betting is a difficult task. At the same time, however, it is important to check the robustness of results of earlier research using novel data; this is what is done in the current research. Moreover, given that questionnaire data used in earlier research were collected during short periods of time in 2010/2011 (Daumann et al., 2011; Wicker and Soebbing, 2012), an interesting question is whether two sports mega events, the European Football Championship and the Olympic Summer Games of 2012, have had an effect on sports betting in Germany. Because this study collected its data in 2013, these data should reflect such effects.

Finally, given recent media coverage of corruption related to sports bets, an interesting question neglected in earlier empirical literature is whether corruption cases and/or betting manipulation affect betting behaviour. As reported by, for example, Ashelm (2014), sports bets are an ‘El Dorado for criminals’. However, researchers have hardly investigated whether the integrity of sports games is threatened by manipulation for betting. A study by Keating in 2003 (quoted by McKelvey, 2004) shows that 52% of US citizens believe that manipulation took place in some sports. Koch and Maennig (2006) show that there is no evidence of a systematic threat to sports from cases of corruption linked to betting. Notwithstanding, they argue that a decisive anti-corruption policy is needed to ward off damage to sporting norms from corruption linked to betting.⁸

Data and methods

The method

This study conducted an online survey from 2 July to 31 July 2013. Based on a standardized questionnaire, the online survey was used to collect data on sociodemographic variables (gender, age, etc.) and data on the experience of respondents with sports bets (how often they had participated). More than 86% of the respondents answered that they participate irregularly in sports bets. For this reason, this study used as the dependent variable in its empirical analysis the answer to the question whether respondents ever had participated in sports betting. This question leaves open when exactly respondents had participated in sports betting (‘last week’, ‘last month’, etc.). Overall, 27% of respondents (N = 528) stated that they had participated in sports betting.

This study also asked questions concerning the respondents’ attitudes towards risk, their sports behaviour, and their media consumption. In total, 1,490 persons visited the Internet page of the online survey, and 634 persons answered the questions, where 77% answered all questions. Because this study’s aim was to reach young adults (20–40 years old; the majority of players participating in sports bets are young adults, see LaBrie et al., 2007), it mainly used social networks to distribute the online survey. Specifically, it disseminated the link to the online survey via Facebook,⁹ and asked sports students to distribute the link among their friends. In addition, this study used an e-mail distributor. Participation in the survey was possible independently of age, gender, sporting interest, and tendency to bet.

The sample

Men are somewhat underrepresented in this study’s data as 42% of respondents are males and 58% are females. The average age of respondents at the time of the survey was approximately 31 years (standard deviation: 11 years) and is skewed to the right. In comparison to the German population, mainly young people participated in the survey, which is what was expected.¹⁰ The majority of respondents were employed at the time of the survey (40%) or studied at a university or technical college (46%). A small difference was observed between males and females as a larger proportion of females were students (51%; men: 39%). Overall, there were more males (46%) than females (35%) in employment, a difference that perhaps can be explained by differences between the age distributions of males and females in this study’s data.

The level of education of respondents was comparatively high, with 39% holding a university degree, 12% holding a technical college qualification, and 22% having completed professional training. Approximately 32% of respondents were still in training at the time of the survey. Only 4% of respondents answered that they had no education qualification and were not in training. With regard to the education level, there are differences between this study’s data and the German population. In 2012, only 7.8% of the population held a university degree average independently of age (Statistisches Bundesamt [Federal Statistics Office], 2013) – in the 25–29 age group 11% had a university degree, 6% a degree from a technical college, 51% completed a professional training, and 25% had not finished an education qualification (Genesis-Online Datenbank, 2013a). Highly educated young adults, thus, are overrepresented in this study’s data.

As was expected given the age distribution, respondents had less money available than a member of a representative German household. Slightly more than a third of respondents had available a monthly household income of less than €1,000 and approximately a quarter had available between €1,000 and less than €2,250 per month. 15% of respondents answered that they had available €2,250 and less than €3,500 per month and approximately the same proportion answered that they had available a household income of between €3,500 and less than €4,500 per month.

As compared to the average German household, respondents lived in below average financial conditions. Two-thirds of respondents answered that they had available a household income of €2,250 or less per month, which is substantially lower than the national average of €2,988 (Federal Statistics Office, 2011)¹¹. and higher than the national average individual net income of €1,618 in the 18–24¹² age group (Genesis-Online Datenbank, 2013b). Only 5% of both males and females had available between €4,500 and less than €5,500 or more per month. At the same time, 42% of respondents judged their financial situation to be good or very good at the time of the survey, with approximately 20% of respondents judging it to be bad or very bad. Females had significantly less money available on average than males.

Approximately two-thirds of respondents did not live alone but lived with at least one person. Overall, more females than men lived alone, which probably also partly explains their lower household income.

This study observed only small differences between males and females with regard to sporting activity. 91% of respondents stated that they played sports. Two-thirds of the 9% who were currently inactive used to play sports and thus only 3% of respondents never had played sports. Respondents, thus, were on average young, highly educated, did play or had played sports, and had a low income.¹³

The empirical model

This study modelled respondents’ propensity to participate in sports betting by means of a logistic (logit) model to differentiate between respondents who bet and respondents who do not bet.

As for the explanatory variable age, cubic splines were used to model the relationship between age and the propensity to bet (see, e.g. Harrell, 2001: 20ff; Lohmann, 2010: 678). Harrell (2001) recommends three to five knots to estimate a spline, which is generally set automatically based on the age distribution.

This study also considered gender effects, and it used household income as an explanatory variable. Because data on household income were collected in groups in the survey group-specific dummy variables were formed to add household income to the study’s model.

As additional explanatory variables, a measure of corruption risk, the influence of interest in sports, and a measure of attitudes towards risk were used. As for corruption risk, respondents were asked to classify how much threat they guess for sports in general on a five-level scale from 1 ‘no threat at all’ to 5 ‘extreme threat’ (median 4). Overall, the distribution of the answers on corruption risk was slightly skewed to the left but almost symmetric. This explanatory variable was interpreted as a metric variable.¹⁴

As for interest in sports, respondents were asked to indicate how frequently they collected information on sports from newspapers, the Internet, radio, TV, and other media. This information was used to compute a metric indicator by calculating their unweighted average value, which is a summary measure of the frequency of media use.¹⁵ The assumption here is that respondents who use different media tend to be more interested in sports than respondents who only gather information on sports from a single newspaper. The variable ‘interest in sports’ assumes a minimum of zero for respondent who never gather information on sports and a maximum of four for respondents who gather information on sports from five different kinds of media. The mean of this study’s media-consumption indicator is two, i.e. respondents were moderately interested in gathering information on sports from the media.

As for the measure of attitudes towards risk used in this research,¹⁶ respondents were asked to choose, on a five-point scale, the extent to which the statement ‘safety is always the most important thing to me’ was applicable to them, from 1 ‘not applicable’ to 5 ‘absolutely and fully applicable’. This variable was interpreted metrically because it was assumed that the distances between the scale values are identical. On average, respondents entered a value of 3, whereby most respondents chose the value 4, so that the variable is only slightly skewed to the left. Respondents, thus, appear to be rather risk averse.

Empirical results

Several models were estimated (see Table 1). The coefficients of the models were y-standardized. The fully (y- and x-) standardized regression coefficients in the multivariate Model 5b allow the strength of the influence of the various variables to be compared. Results show that age and gender have the strongest effects followed by ‘interest in sports’. Income plays a moderate role. Both risk aversion and respondents’ subjective belief of a threat of corruption in sports also exert a weak effect, with the former having a somewhat stronger influence than the latter. The effect of subjective belief of a threat of corruption in sports is significant at the 6% level.

Table 1.

Results of the logistic regression model (y-standardized β-coefficients).

	Model 1	Model 2	Model 3	Model 4	Model 5a	Model 5b*
Sex:
Male	1.053^***	0.924^***	0.790^***	0.768^***	0.746^***	0.369^***
	(7.84)	(7.61)	(6.60)	(6.43)	(6.25)	(6.25)
Income:
< €1,000	Ref.	Ref.	Ref.	Ref.	Ref.	Ref.
€1,000 <	−0.423^**	−0.370^*	−0.427^**	−0.423^**	−0.436^**	−0.195^**
€2,250	(–2.70)	(–2.45)	(–2.78)	(–2.77)	(–2.85)	(–2.85)
€2,250 <	−0.549^**	−0.407^*	−0.331⁺	−0.329⁺	−0.307⁺	−0.111⁺
€3,500	(–2.83)	(–2.18)	(–1.82)	(–1.82)	(–1.68)	(–1.68)
€3,500 <	−0.527^**	−0.412^*	−0.456^*	−0.471^*	−0.507^**	−0.169^**
€4,500	(–2.60)	(–2.10)	(–2.36)	(–2.45)	(–2.60)	(–2.60)
€4,500 <	−0.284	0.062	−0.052	−0.088	−0.081	−0.019
€5,500	(–1.14)	(0.24)	(–0.21)	(–0.34)	(–0.32)	(–0.32)
>€5,500	−0.479⁺	−0.216	−0.167	−0.113	−0.080	−0.018
	(–1.76)	(–0.83)	(–0.65)	(–0.44)	(–0.32)	(–0.32)
Age
Spline 1		0.069^**	0.066^**	0.073^**	0.073^**	0.819^**
		(2.81)	(2.69)	(2.96)	(2.93)	(2.93)
Spline 2		−0.296^***	−0.281^***	−0.300^***	−0.288^***	−0.133^***
		(–3.64)	(–3.52)	(–3.71)	(–3.59)	(–3.59)
Interest in sports			0.258^***	0.253^***	0.259^***	0.317^***
			(5.13)	(5.01)	(5.08)	(5.08)
Threat of corruption				−0.117⁺	−0.116⁺	−0.104⁺
				(–1.90)	(–1.89)	(–1.89)
Risk aversion					−0.133^*	−0.140^*
					(–2.46)	(–2.46)
Constants	−1.881^***	−5.914^***	−7.368^***	−6.680^***	−5.667^***	−5.667^***
	(–7.30)	(–4.04)	(–4.57)	(–4.07)	(–3.32)	(–3.32)
N	433	433	433	433	433	433
Mc Fadden Pseudo R²	0.183	0.245	0.311	0.319	0.333	0.333
Nagelkerke Pseudo R²	0.270	0.350	0.430	0.440	0.455	0.455

The β-coefficients are y-standardized; t-statistic in parenthese;* the β-coefficients in the overall model 5b are x- and y-standardized; ⁺p < 0.10; ^*p < 0.05; ^**p < 0.01; ^***p < 0.001.

The McFadden pseudo-R² of the model assumes a value of 33%.¹⁷ The quality of the model was also checked by comparing the predicted and actual values (Table 2). Sensitivity and specificity assume values of 54.64% and 94.05% and indicate a good model performance. Further analysis showed that the area under the ROC curve is 0.87. Overall, the model correctly classifies 53 gamblers and 316 non-gamblers (out of 433 respondents). The model, therefore, correctly predicts in 85% of all cases.

Table 2.

Predictive power of the model.

		True value
		Gambler	Non-gambler	Total
Predicted value	Betting	53	20	73
	Not betting	44	316	360
	Total	97	336	433
	Sensitivity		54.64%
	Specificity		94.05%
	Correctly predicted		85.22%

The threshold is set at 0.5.

Gender effects and the probability of betting

The gender effect is significant across all models. The probability that a respondent had participated in sports bets is higher for males than for females, consistent with results reported in earlier research (Welte et al., 2002; Wicker and Soebbing, 2012). The overall model (Table 1, Model 5a) implies that a 26-year old male bettor who earns less than €1,000 and whose socioeconomic profile is identical to the profile of an average respondent had participated in sports betting with a probability of 58%. For a female bettor with the same characteristics, the probability is only 16% (see Table 3); the chance that a male respondent had already participated in a sports bet is approximately seven times higher¹⁸ than for a female respondent.

Table 3.

Predicted probability of already having participated in sports bets, separated into males and females, estimated from the overall regression model (Table 2, Model 5a, N=472).

Ideal type (age: 26 (median), income: under €1,000, remainder at means)	Predicted probability	95%-confidence interval
Males	58%	[44%; 72%]
Females	16%	[8%; 24%]

Age and the probability of betting

Age has an inverted u-shaped effect on the probability that a respondent had participated in sports bets (Figure 1).¹⁹ The probability of betting first increases, attains a maximum at the age of 32, and then starts decreasing. Thus, for a 32-year old respondent the probability is approximately four times larger than that for a 20-year old respondent (the reference group);²⁰ Figure 1 further shows that the effect of age is only significant for a certain range of ages but not for middle-aged respondents possibly because this study has only available a few data for middle-aged respondents. Estimates, however, regain reliability for respondents of approximately 56 years or older. The probability that a 56-year old had participated in sports bets is only 16%.²¹

Figure 1.

Logits and confidence interval of betting, estimated from the overall regression model (Table 2, Model 5a, N=472).

Income and the probability of betting

A respondent who had available a middle household income per month (categories: €1,000 < €2,250 and €2,250 < €3,500) has a lower participation probability than a respondent who had less than €1,000 available per month (Table 4). Overall male respondents in the middle-income categories have the lowest probability of having had bet on sports. Hence, this study’s results on the effect of income on the propensity to participate in sports bets corroborate results of earlier research that find that sports bets are more prevalent among individuals with a low income (Humphreys and Soebbing, 2012; see, however, Wicker and Soebbing, 2012, who find that sports bettors tend to have a high income). However, confidence intervals are rather wide (the insignificance of the upper income-categories reflects the composition of the sample) and, therefore, the interpretation of differences in probabilities across income groups should not be stretched too far.

Table 4.

Predicted probability of already having participated in sports bets, separate for males, estimated from the overall regression model (Table 2, Model 5a, N=472).

Average type (age: 26 (median), male, other variables at means)	Predicted probability	95% confidence interval
< €1,000	58%	[44%; 72%]
€1,000 < €2,250	30%	[16%; 44%]
€2,250 < €3,500	38%	[18%; 58%]
€3,500 < €4,500	26%	[9%; 44%]
€4,500 < €5,500	53%	[21%; 84%]
> €5,500	53%	[22%; 84%]

Interest in sports and the probability of betting

Figure 2 shows (for males and females) the effect of interest in sports on the participation probability for a respondent with an income under €1,000, an age of 26 years (median age), and an otherwise average socioeconomic profile. The participation probability of bets increases with increasing interest in sports. Males who stated that they were very interested in sports have a participation probability of 84%. The probability that females already had participated in sports betting, if they are not interested in sports, is only 4%. Females with a strong interest in sports have approximately the same participation probability (27%) as males with a low interest in sports (1 on the scale) (26%).²²

Figure 2.

Predicted probability of betting for the variable interest in sports, separated by sex, estimated from the overall regression model (Table 2, Model 5a, N=472).

Perceived threat to sports from corruption and the probability of betting

The probability that a respondent had participated in sports bets decreases in the perceived threat of corruption (Figure 3). The confidence interval is rather wide for females who do not perceive any threat to sports from corruption, but in general the confidence interval is somewhat narrower than for males. Females who do not perceive any threat to sports from corruption also tend not to bet. Males who do not perceive any threat of corruption have an 80% participation probability.²³

Figure 3.

Predicted probability of betting for the variable ‘threat to sports from corruption’, separated by sex, estimated from the overall regression model (Table 2, Model 5a, N=472).

Attitude towards risk and the probability of betting

Figure 4 shows that the probability that a respondent had participated in sports bets is lower the higher is the degree of risk aversion. The effect is approximately the same (on a somewhat different level) for males and females, where males have a stronger tendency to bet than females, independent of risk aversion. The predicted probability that males for whom risk-aversion is high would bet is 43% and, thus, still above the corresponding probability for females. For males for whom risk aversion is low, the probability is 30%. The influence of gender, thus, is stronger than the influence of risk aversion.

Figure 4.

Predicted probability of betting for the variable risk aversion, separated by sex, estimated from the overall regression model (Table 2, Model 5a, N=472).

Summary and concluding remarks

Against the mixed and in part conflicting evidence reported in earlier research on the characteristics of sports bettors, the purpose of this study was to reexamine the determinants of the propensity to participate in sports bets using a novel dataset. Part of this study’s results confirm earlier evidence, other results contradict earlier evidence, and still other results have no counterpart in earlier research (the effect of, for example, the perceived threat to sports of corruption has been largely neglected in earlier research).

A total of 634 primarily relatively young respondents were surveyed in the summer of 2013 using an online questionnaire. Respondents in this study’s non-representative sample were younger and better educated overall and had less money available than the average German citizen. Keeping this in mind, this study’s results show that the probability that a respondent had already participated in sports betting depends on a gender effect. The probability that a male respondent had already participated in sports betting is seven times larger than for a female respondent, corroborating findings of earlier research (Humphreys and Soebbing, 2012; LaBrie et al., 2007; Wicker and Soebbing, 2012).

In contrast to Wicker and Soebbing (2012), this study found an inverted u-shaped link between age and the probability that a respondent had already participated in sports betting. Because some earlier researchers (Humphreys et al., 2011) report evidence of a nonlinear age effect this study analysed the age effect with splines in more detail and could confirm the presence of a nonlinear effect.

In comparison to results documented in earlier research (Humphreys and Soebbing, 2012; Wicker and Soebbing, 2012), respondents in this study study were younger and had available a lower household income. In this respect, it should be kept in mind that this study’s sample is not representative of the German population.

This study’s results further suggest that the probability that a respondent had participated in sports betting is higher for respondents with a low household income. The probability is lowest for respondents who belong to the middle-income group, suggesting the possibility that participation in sports bets is linked to income in a nonlinear way (the effect is not significant for the top income groups in this study’s analysis). This possibility should be further explored in future research.

A high probability that a respondent had participated in sports betting goes along with increased interest in sports. A higher value of this study’s measure of attitudes towards risk tends to reduce the probability of having already participated in sports betting, but a gender effect dominates the effect of risk aversion. The participation probability also decreases when the perceived threat of corruption in sports increases, an aspect that has not been documented in earlier empirical research on sports betting in Germany. Given the selective nature of this study’s sample and the fact that the effect of the perceived threat of corruption is not overly significant, it would be interesting to analyse in future research in more detail the link between the threat of corruption in sports and sports betting.

As more empirical evidence on the characteristics of sports bettors becomes available, it is also interesting to develop in future research a foundation for a sociological theory of sports betting. Before such a full-fledged and detailed sociological theory of sports betting can be developed, however, as many empirical results as possible on the characteristics of sports bettors should be assembled. Given that empirical research on the characteristics of sports bettors typically is based on relatively small samples of data (and this study’s research is no exception in this respect), it continues to be important to reexamine in future research the question ‘Who bets on sports?’.

Footnotes

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Abarbanel

(2012) Cultural indicators of internet sports betting policy. UNLV Gaming Research and Review Journal 16(2012): 77–85.

Albers

(1993) Ökonomie des Glücksspielmarktes in der Bundesrepublik Deutschland [Economy of Gambling Market in the Federal Republic of Germany]. Berlin, Germany: Duncker & Humblot.

Ashelm

(2014) Studie: Sportwetten sind ein Eldorado für Kriminelle [Study: Sports betting are an El Dorado for criminals]. FAZ-online, 5 May. Available at: http://www.faz.net/aktuell/wirtschaft/wirtschaftspolitik/studie-sportwetten-sind-eldorado-fuer-kriminelle-12941238.html (accessed 8 July 2014).

Atkinson

Flint

(2004) Snowball sampling. In: Lewis-Beck

Bryman

Liao

(eds) The SAGE Encyclopedia of Social Science Research Methods, vol. 3. Thousand Oaks, CA: SAGE, pp.1043–1044.

Beckert

Lutter

(2007) Wer spielt, hat schon verloren? Zur Erklärung des Nachfrageverhaltens auf dem Lottomarkt [The player always loses? Toward an explanation of demand behavior in lottery markets]. Kölner Zeitschrift für Soziologie und Sozialpsychologie [Cologne Journal of Sociology and Social Psychology] 59(2): 240–270.

Beckert

Lutter

(2008) Wer spielt Lotto? Umverteilungswirkungen und sozialstrukturelle Inzidenz staatlicher Lotteriemärkte [Who plays the lottery. Distribution effects and social-structural incidence of state lottery markets]. Kölner Zeitschrift für Soziologie und Sozialpsychologie [Cologne Journal of Sociology and Social Psychology] 60(2): 233–264.

Beckert

Lutter

(2009) The inequality of fair play: Lottery gambling and social stratification in Germany. European Sociological Review 25(4): 475–488.

Bendixen

(2008) Ökonomie des Glücksspiels [Economy of gambling]. In: Gebhardt

Grusser-Sinopoli

(eds) Glücksspiel in Deutschland. Ökonomie, Recht, Sucht [Gambling in Germany. Economy, Law, Addiction]. Berlin, Germany: Walter de Gruyter, pp.30–52.

Blalock

Just

Simon

(2007) Hitting the jackpot or hitting the skids: Entertainment, poverty, and the demand for state lotteries. American Journal of Economics and Sociology 66(3): 545–570.

10.

Bloch

(1951) The sociology of gambling. American Journal of Sociology 57(3): 215–221.

11.

Borg

Mason

Shapiro

(1991) The Economic Consequences of State Lotteries. New York, NY: Praeger.

12.

Breuer

Hauten

Kreuz

(2009) Financial instruments with sports betting components: Marketing gimmick or a domain for behavioral finance? Journal of Banking & Finance 33(12): 2241–2252.

13.

Browne

Brown

(1994) Predictors of lottery gambling among American college students. Journal of Social Psychology 134(3): 339–347.

14.

Bundeszentrale für gesundheitliche Aufklärung [Federal Centre for Health Education] (eds) (2012) Glücksspielverhalten und Glücksspielsucht in Deutschland. Ergebnisse aus drei repräsentativen Bevölkerungsbefragungen 2007, 2009 und 2011 [Gambling Behavior and Gambling Addiction in Germany. Results of Three Representative Population Surveys 2007, 2009 and 2011]. Köln, Germany: Bundeszentrale für gesundheitliche Aufklärung (BZgA).

15.

Chimi

Russell

(2009) The Likert scale: A proposal for improvement using quasi-continuous variables. The Proceedings of the Information Systems Education Conference 26: 1–10.

16.

Daumann

Römmelt

Heinze

(2011) Forschungsreport - Der Markt für Sportwetten in Deutschland aus Bevölkerungsperspektive [Research report. The market for sports betting in Germany from the point of view of population]. Sciamus – Sport und Management 3: 28–31.

17.

Eadington

(1987) Economic perceptions of gambling behavior. Journal of Gambling Studies 3(4): 264–273.

18.

Elias

Dunning

(2003) Sport und Spannung im Prozess der Zivilisation [Quest for Excitement: Sport and Leisure in the Civilising Process]. Frankfurt am Main, Germany: Suhrkamp.

19.

Endes

Feldner

(2012) Sportwetten - zu jeder Zeit aktuell [Sports betting - at all times up-to-date]. In: Kainz

Scherrer

Werner

(eds) Sportfinanzierung und Sportwetten. Reflexionen zu Phänomenen, Möglichkeiten und Gefahren im kommerziellen Sport [Sports Financing and Sports Betting. Reflections on Phenomenons, Possibilities, and Threats in Commercial Sports]. Stuttgart, Germany; Wien, Germany; Zürich, Switzerland: Boorberg; Sramek; Schulthess, pp.79–86.

20.

Eurostat (2011) Demography Report 2010. Older, More Numerous and Diverse Europeans. Available at: http://epp.eurostat.ec.europa.eu/cache/ITY_OFFPUB/KE-ET-10-001/EN/KE-ET-10-001-EN.PDF (accessed 28 January 2014).

21.

Forrest

Simmons

(2003) Sports and gambling. Oxford Review of Economic Policy 19(4): 598–611.

22.

Gassmann

Emrich

Meyer

(2013) Die wirtschaftliche Lage und die Beschäftigungssituation saarländischer Studierender: Bildungsinvestitionen und Konsumverhalten der Studierenden [The economic situation and employment situation of Saarland students: Education investments and consumer behavior of students]. In: Emrich

Meyer

Rampeltshammer

(eds) Die Universität des Saarlandes in sozio-ökonomischer Perspektive. Ausgewählte Analysen sozialer und wirtschaftlicher Effekte [Saarland University in Socio-Economic Perspective. Selected Analysis of Social and Economic Effects]. Saarbrücken: Universaar, pp.103–134.

23.

Genesis-Online Datenbank (2013a) Mikrozensus Tabelle 12211-0041. Available at: https://www-genesis.destatis.de/genesis/online/data;jsessionid=19D4AE52B9EB77E17AE754A95FDD32C1.tomcat_GO_2_2?operation=statistikenVerzeichnis (accessed 27 May 2015).

24.

Genesis-Online Datenbank (2013b) EU-SILC Tabelle 63400-0001. Available at: https://www-genesis.destatis.de/genesis/online/data;jsessionid=19D4AE52B9EB77E17AE754A95FDD32C1.tomcat_GO_2_2?operation=statistikenVerzeichnis (accessed 27 May 2015).

25.

Glass

. (1972) Consequences of failure to meet assumptions underlying the analyses of variance and covariance. Review of Educational Research 42: 237–288.

26.

Goldmedia (2013) Studie zum deutschen Sportwettenmarkt: Die neue Glücksspielregulierung könnte ihre Ziele verfehlen [Study on German sports betting market: The new regulation of the gaming sector can fail to achieve its objectives]. Available at: http://www.goldmedia.com/presse/newsroom/gluecksspielmarkt-deutschland-2017.html (accessed 27 September 2013).

27.

Harrell

(2001) Regression Modeling Strategies: With Applications to Linear Models, Logistic Regression, and Survival Analysis. New York, NY: Springer-Verlag.

28.

Humphreys

Soebbing

(2012) Sports betting, sports bettors and sports gambling policy. In: Büch

M-P

Maennig

Schulke

H-J

(eds) Sport und Sportgroßveranstaltungen in Europa – zwischen Zentralstaat und Regionen [Sport and Large Sporting Events in Europe – Between Central Government and Regions]. Hamburg, Germany: Hamburg University Press (Verlag der Staats- und Universitätsbibliothek Hamburg), pp.15–38.

29.

Humphreys

Lee

Soebbing

(2011) ‘Modeling consumers’ participation in gambling markets and frequency of gambling. The Journal of Gambling Business and Economics 5(1): 1–22.

30.

Kearney

(2005) State lotteries and consumer behavior. Journal of Public Economics 89: 2269–2299.

31.

Koch

Maennig

(2006) Spiel- und Wettmanipulationen – und der Anti-Korruptionskampf im Fußball. Macht und Versuchung auf dem grünen Rasen [Game and betting manipulation – And the fight against corruption in soccer]. Der Bürger im Staat 56(1): 50–56.

32.

LaBrie

Shaffer

(2011) Identifying behavioral markers of disordered Internet sports gambling. Addiction Research & Theory 19(1): 56–65.

33.

LaBrie

. (2007) Assessing the playing field: A prospective longitudinal study of Internet sports gambling behavior. Journal of Gambling Studies 23(3): 347–362.

34.

Lin

(2015) Elucidating the association of sports lottery bettors’ socio-demographics, personality traits, risk tolerance and behavioural biases. Personality and Individual Differences 73: 118–126.

35.

Lohmann

(2010) Nicht-Linearität und Nicht-Additivität in der multiplen Regression: Interaktionseffekte, Polynome und Splines [Non-linearity and non-additivity in the multiple regression: Interaction terms, polynomials, and splines]. In: Wolf

Best

(eds) Handbuch der sozialwissenschaftlichen Datenanalyse [Handbook of Social Science Data Analysis]. Wiesbaden, Germany: VS Verlag für Sozialwissenschaften, pp.677–706.

36.

Lubke

Muthen

(2004) Applying multigroup confirmatory factor models for continuous outcomes to Likert scale data complicates meaningful group comparisons. Structural Equation Modeling 11(4): 514–534.

37.

Lutter

(2010a) Märkte für Träume: Die Soziologie des Lottospiels [Markets for Dreams: The Sociology of Lottery]. Frankfurt am Main, Germany; New York, NY: Campus Verlag.

38.

Lutter

(2010b) Zur Erklärung von Diffusionsprozessen. Das Beispiel der Einführung staatlicher Lotterien in den USA [Explaining diffusion: The introduction of state lotteries in the USA]. Zeitschrift für Soziologie [Journal of Sociology] 39(5): 363–381.

39.

McFadden

(1978) Quantitative methods for analyzing travel behavior of individuals: Some recent developments. In: Hensher

Stopher

(eds) Behavioral Travel Modeling. London, UK: Croom Helm, pp.279–318.

40.

McKelvey

(2004) The growth in marketing alliances between US professional sport and legalised gambling entities: Are we putting sport consumers at risk? Sport Management Review 7(2): 193–210.

41.

McKinney

Swain

(1993) State lotteries: Explaining their popularity. International Journal of Public Administration 16(7): 1015–1033.

42.

Mintas

(2009) Glücksspiele im Internet. Insbesondere Sportwetten mit festen Gewinnquoten (Oddset-Wetten) unter strafrechtlichen, verwaltungsrechtechtlichen und europarechtlichen Gesichtspunkten (Schriften zum Strafrecht, 204) [Games of Chance on the Internet. Especially Sports Betting with Fixed Odds (Oddset-Betting) from the Point of View of Criminal, Administrative, and European Law, 204]. Berlin, Germany: Duncker & Humblot.

43.

Orsini

(2009) xbrcspline – Differences in predicted responses after restricted cubic spline models. Available at: http://www.stata.com/meeting/sweden09/se09_orsini.pdf (accessed 24 January 2014).

44.

Paul

Weinbach

(2010) The determinants of betting volume for sports in North America: Evidence of sports betting as consumption in the NBA and NHL. International Journal of Sport Finance 5(2): 128–140.

45.

Rebeggiani

Breuer

(2015) Neue Ordnung, neues Glück? Ordnungspolitische und fiskalische Aspekte des deutschen Sportwettenmarktes [New order, new luck? Regulatory and fiscal aspects of of the German sports betting market]. In: 19 Jahrestagung des Arbeitskreises Sportökonomie [19th annual meeting of the Arbeitskreis Sportökonomie], Leipzig, 23–25 April. Leipzig, Germany: Sportwissenschaftliche Fakultät, Universität Leipzig, pp.5–6.

46.

Reiche

(2013) The prohibition of online sports betting: A comparative analysis of Germany and the United States. European Sport Management Quarterly 13(3): 293–314.

47.

Spöring

(2012) Von “Asian Handicap” bis “Zweierkombi”: Einführung in die Theorie der Sportwetten [Form “Asian Handicap” to “Zweierkombi”: Introduction into the theory of sports betting]. In: Kainz

Scherrer

Werner

48.

Statistisches Bundesamt [Federal Statistics Office] (2013) Bildungsstand. Bevölkerung nach Bildungsabschluss in Deutschland. Available at: https://www.destatis.de/DE/ZahlenFakten/GesellschaftStaat/BildungForschungKultur/Bildungsstand/Tabellen/Bildungsabschluss.html(accessed 24 September 2014).

49.

Statistisches Bundesamt [Federal Statistics Office] (2011). Einkommen, Einnahmen und Ausgaben. Einkommen, Einnahmen und Ausgaben privater Haushalte 2011 in den Gebietsständen. Available at: https://www.destatis.de/DE/ZahlenFakten/GesellschaftStaat/EinkommenKonsumLebensbedingungen/EinkommenEinnahmenAusgaben/Tabellen/Gebietsstaende_LWR.html (accessed 24 September 2014).

50.

Veblen

(1899) The Theory of the Leisure Class: An Economic Study of Institutions. New York, NY: Macmillan.

51.

Weiss

Loubier

(2010) Gambling habits of athletes and nonathletes classified as disordered gamblers. The Journal of Psychology 144(6): 507–521.

52.

Welte

. (2002) Gambling participation in the US—Results from a national survey. Journal of Gambling Studies 18(4): 313–337.

53.

Wicker

Soebbing

(2012) Examining participation in sports betting in Germany. The Journal of Gambling Business and Economics 6(3): 17–33.