The causal effect of professional sports on amateur sport participation--an instrumental variable approach.

Mutter, Felix ; Pawlowski, Tim

Introduction

In general, a better understanding of factors influencing amateur sport participation is of great importance. Amateur sport participation is associated with better physical (e.g., Humphreys, MacLeod, & Ruseski, 2011) and mental health (e.g., Downward & Rascuite, 2011; Pawlowski, Downward, & Rascuite, 2011; Rascuite & Downward, 2010). It contributes significantly to revenues of various industries (e.g., Pawlowski & Breuer, 2012b; Preuss, Alfs, & Ahlert, 2012) as well as tax income of public authorities (Pawlowski & Breuer, 2012a).

In this regard, an increasing number of studies have recently focused on analysing the so-called demonstration effect of professional sports (Weed, 2009), i.e., the possible effect of professional sports on amateur sport participation. These studies could detect that professional sports stimuli, such as star athletes (Schmidt & Hogele, 2011; Tribe Group, 2009), the success of national athletes (Hallmann, Breuer, & Kuhnreich, 2013; IpsosMORI, 2009), major sporting events (UK SPORT, 2011; Weed, Coren, & Fiore, 2009; Weed et al., 2012), or sport-related media behavior (Lines, 2007) motivates people to participate in sport. However, stated motivation and actual behavior are two distinct events (Heckhausen & Gollwitzer, 1987). A stated motivation is not subject to a budget and time constraint. Furthermore, estimating professional sports stimuli on the aggregate level, such as success of national athletes, neglects a possible source of heterogeneity between individuals that is connected to their perceived relevance of success (Pawlowski, Downward, & Rasciute, 2013).

Therefore, this paper offers two contributions: First, it is investigated whether professional sports stimuli indeed transfer into observable behavior, i.e., an increase in sports participation. Second, instead of making use of an aggregated professional sports stimulus, such as success of national athletes, a comprehensive measure is employed based on the concept of psychological involvement (Zaichkowsky, 1985) which expresses the personal relevance of professional sports to an individual and therefore allows for heterogeneity between individuals as described before.

The data used to analyze this effect was collected in a survey amongst N=863 triathletes in Germany. (1) Since (for instance) the individually perceived relevance of professional triathlon could increase with an extensive participation in amateur triathlon, a potential reverse causality bias aggravates the estimation of the causal effect of interest. To deal with this endogeneity issue, an instrumental variable approach is employed (Angrist & Krueger, 2001). Different test statistics provide evidence for the relevance and validity of the chosen instruments.

As indicated by the results, the individually perceived relevance of professional triathlon indeed causes amateur participation in our setting. According to the 2SLS results, amateur triathletes who place more importance on professional triathlon participate between 3% and 33% more in triathlons themselves. Therefore, this paper provides further evidence on the existence of a trickle-down effect in sports.

The paper is organized in the following way. Section 2 outlines the theoretical framework of the study and provides an overview of the related literature. In section 3, the identification strategy is presented in detail. Section 4 outlines the data collection and the empirical model specifications. The descriptive and analytic results are considered in Section 5. Section 6 concludes with a discussion of the results, their implications, and their limitations.

Background

Theoretical framework

From a neoclassical perspective, studies that have analyzed the demand for sport participation have mainly drawn on Becker's (1965) theory of household production (Dawson & Downward, 2013; Downward & Riordan, 2007) and on Cawley's (2004) SLOTH framework (Garcia, Lera-Lopez, & Surarez, 2011; Humphreys & Ruseski, 2011). Becker (1965) introduced time as an important budget constraint for the production of non-market goods, like sport participation. Cawley (2004) extended this framework and assumed that people allocate their time to sleeping, working and leisure activities in order to maximize their utility subject to a budget constraint including time, money and biology.

Other studies have taken a heterodox perspective on the decision to participate in sport (Downward, 2007). Approaches from social science, e.g., psychology (Scitovsky, 1976) or sociology (Bourdieu, 1988), have been incorporated in economic analyzes to explain sporting behavior on the basis of differing preferences. The framework of sporting role models is in line with this heterodox perspective. Incorporated into the economic demand analysis for participation in sports, it explains why the individually perceived relevance of professional sports might influence the demand for amateur sport participation (Mutter & Pawlowski, 2013). (2) The framework is based upon a modern understanding of role models, introduced by both Jung (1986) and Gibson (2003, 2004).

Based on their exceptional status and their media presence, professional athletes could be characterized as role models (Bandura, 1986). In contrast to the related concepts of mentors and behavioral models, professional athletes can be categorized as distant or symbolic models because, most of the time, no personal relationship between the model and the observer is required (Gibson, 2004).

Role models can have a motivational effect on behavioral tendencies (Gibson, 2003, 2004; Jung, 1986). In the sports context, they could potentially attract non- participants to sport, reactivate former participants, or motivate active participants to participate more (Weed, 2009). According to Lyle (2009), the complexity of motivations and barriers to sport participation heavily aggravates a role model's influence on non- participants. Indeed, Mutter and Pawlowski (2013) found that professional sports role models, i.e., the German national soccer team, attracted only few non- participants to soccer. Nevertheless, the theoretical framework of role models suggests that a motivational effect is strongest on already existing behavioral tendencies (Gibson, 2004; Jung, 1986). Indeed, sporting role models were found to motivate active participants to increase their participation (Mutter & Pawlowski, 2013). Similar results were obtained in various studies and reports (Breuer & Hallmann, 2011; IpsosMORI, 2009; UK SPORT, 2011; Weed, 2009).

Two conditions determine the motivational effect of sporting role models (Mutter & Pawlowski, 2013). (1) Availability of professional athletes is considered to be a necessary condition. Because of the missing personal relationship, the stimuli of symbolic role models appear more distant and on a less regular basis (Bandura, 1986). Therefore, to ensure availability, professional athletes have to be present via the media or as participants in live events. (2) Furthermore, the perceived proximity of an observer to a model positively influences the model's motivational effect (Mutter & Pawlowski, 2013). This proximity is based on demographic dimensions, such as the nationality, the age or the gender. A German professional tennis player, for example, is most likely to motivate a German amateur tennis player in his or her sporting activity. Furthermore, proximity could be based on the perceived capability of an individual to replicate the observed behavior (Lockwood & Kunda, 1997).

Literature review

Previous studies relevant to this paper have analyzed either the effect of national sporting success or hosting major sporting events on sport participation rates or sports club membership figures. These studies offer mixed findings. For instance, while the membership figures of curling in Scotland increased after the Olympic success of the British women in 2002 (Sportscotland, 2004), the membership figures of tennis in Germany decreased as a result of the successful era of national athletes from 1986 to 1996 (Feddersen, Jacobsen, & Meannig, 2009). Looking at general levels of sports participation, Humphreys, Maresova, and Ruseski (2012) found a negative correlation between past Olympic success and mass sport participation in a country. Based on the Olympic results of Australia between 1976 and 1996, Hogan and Norton (2000) identified no correlation between national sporting achievements and sedentariness rates of the Australian population. Furthermore, Frawley and Cush (2011) found an increase in the membership figures of organized rugby in Australia after the hosting of the 2003 Rugby World Cup. Based on a literature review, McCartney et al. (2010) could neither support nor refute the expectations about health benefits for the host population of major sporting events.

Other studies have previously analyzed variables of professional sports as possible determinants of the time spent in physical activity. For instance, following Dawson and Downward (2011, 2013) and Downward, Lera-Lopez, and Rasciute (2011), watching sports on television has a positive effect on the frequency of sport participation. Analysing the DCMS Taking Part Survey, Downward and Dawson (2011) revealed that people who watched sports on television (live and non-live) reported 67 minutes more sport participation per week. In addition, they showed that attendees of a live sport event in the four weeks before the interview reported 44 minutes more weekly sport participation.

However, the causality of these findings is hard to interpret because the studies have not accounted for the potential endogeneity of the analyzed professional sports stimuli. If the time spent on sport participation influences the probability of attending events or watching sport on television, a reverse causality bias could occur. Both Lera-Lopez, Ollo-Lopez, and Rapun-Garate (2012) and Thrane (2001) found such an effect. According to their results, the frequency of sport participation positively affects the probability of attending sporting events as a spectator. Obviously, professional sports stimuli cannot be both endogenous and exogenous, so the causality of the revealed effects remains unclear.

Therefore, it is the objective of this paper to analyze the causal effect of the individually perceived relevance of professional sports on the time spent in amateur sport participation. To do so, an identification strategy is developed as explained in the following.

Identification

A reverse causality bias in our setting might be triggered through three channels: (1) More amateur participation could increase a superior affinity to sport in general. According to Kenyon and McPherson (1973), such a superior affinity can influence both professional sports interest and active sport participation. Similarly, Dawson and Downward (2011, 2013) concluded that a latent form of sport consumption influences both manifestations: spectatorship of professional sports and amateur participation. (2) More amateur participation leaves less time for professional sports activities, such as following events on the media. Consequently, the available net time for leisure activities could influence both variables simultaneously. Amateur sport participation and professional sports activities could then be interpreted as substitution goods (Lera-Lopez et al., 2012). (3) More amateur participation could increase the perceived proximity of an amateur to professional sports. On average, more participation leads to higher skills and an attendance of amateur competitions of higher quality. According to the framework of sporting role models, the perceived proximity to a role model increases the model's motivational effect (Mutter & Pawlowski, 2013).

As indicated by these theoretical considerations, a simple OLS estimation of amateur sport participation on the perceived relevance of professional sports and controls could either overestimate (latent sport affinity; proximity) or underestimate (available net time) the causal effect of interest. To avoid such estimation bias, the use of (exogenous) instrumental variables is an acknowledged method (Angrist & Krueger, 2001). In this study we make use of two distinct instruments. The choice of these instruments as well as the underlying assumptions are discussed in the following paragraphs.

The first instrument is related to the residence of the surveyed amateurs. Its relevance, i.e., its partial effect on the endogenous regressor, is derived from the framework of sporting role models. Availability of a model is a necessary condition for its effectiveness. However, compared with other sports, the nationwide media coverage of triathlon in Germany is not very distinct (Ruhle, 2003; Zubayr & Gerhard, 2008). Nevertheless, there are professional triathlon events located in Germany, notably in the cities of Hamburg and Roth. An annual event of the ITU World Triathlon Series, held in the city of Hamburg since 2007 (International Triathlon Union [ITU], 2012). In addition, since 2002, one traditional long-distance triathlon of the "Challenge Family" has been held in the city of Roth (Challenge Roth, 2012). A large number of visitors are inhabitants of the city or region in which the event takes place (Preufi, Seguin, & O Reilly, 2007; Siegfried & Zimbalist, 2000). Therefore, the probability of attending a professional triathlon event as a spectator is higher the closer it is located to the residence. According to Zanger and Schweitzer (2004), professional triathlon events are fascinating and deliver experiences of movement and fitness. In addition, before, during and after the event, it is present in the regional and local media, providing the availability of professional triathlon for residents. Consequently, it is assumed that the residence of amateur triathletes near one of these two events positively influences their perceived relevance of professional triathlon on average.

In general, an instrument must be exogenous, i.e., independent of potential outcomes of the dependent variable. For the most part, the decision of the residence was determined before the decision about the time spent on amateur triathlon participation. (3) Temporal order of the determination of variables is often a good hint for exogeneity (Angrist & Pischke, 2009). In addition, residence should not influence the time spent on amateur participation, other than by influencing the relevance of professional triathlon. Previous studies have shown that the availability of sport infrastructure influences participation in physical activity (Huang & Humphreys, 2012). Consequently, living in the city with over one million inhabitants, such as Hamburg, could influence the decision to participate. To deal with this issue, the estimated models also control for the degree of urbanisation. Finally, it could be argued that people living near Hamburg and Roth have more opportunities to participate in amateur triathlon because the professional events in Hamburg and Roth are surrounded by additional races for amateurs. "Proximity to home" is the main determinant for the choice of races in which to participate among amateur triathletes (Tribe Group, 2009). However, according to the German Triathlon Union (DTU, 2012), over 1000 amateur events for triathletes are organized in Germany every year. Because these events are well distributed over Germany (Triathlon-Deutschland, 2012), it is assumed that Hamburg and Roth do not provide extraordinary opportunity to become engaged in amateur triathlon.

The second instrument is derived from a measure of previously stated relevance of professional triathlon. The instrument is a variable that describes whether professional triathlon, for example through event spectatorship or the success of national athletes, had influenced the decision of the amateur triathletes to initiate their participation. According to Kenyon and McPherson (1973), the attitude towards professional sports is a persistent one. Therefore, a partial effect of previously stated relevance on contemporaneous relevance is assumed.

It is assumed that this instrument is exogenous. Again, the temporal order of decisions is crucial. The reverse causality bias develops through the amateur participation. Consequently, any decision that was taken before the participation cannot be affected by that participation.

Method

Data

Prior empirical evidence has suggested that professional sports stimuli mostly affect active participants (Mutter & Pawlowski, 2013). Moreover, the framework of sporting role models suggests that relevance of professional triathlon most likely affects other active triathlon participants. Consequently, German amateur triathletes were surveyed. Because no systematic data of the approximately 150,000 German triathletes were available, a random sampling was not feasible. In a cluster sampling approach, a constructed and pretested questionnaire was mailed to the 16 regional triathlon associations in Germany, who were asked to forward the questionnaire to their members. Between July and September 2012, quantitative data on N=863 German amateur triathletes were collected online. It is crucial that the amateur triathletes do not participate in the professional events of Hamburg and Roth themselves. This would damage the validity of the instrument related to the residence of the amateur triathletes. To separate distinctly between professional and amateur athletes, the skill level of the triathletes was surveyed. Respondents that reported a high skill level, i.e., "national level" and "near national level," were excluded from further analysis. This leaves a final sample of n=759 amateur triathletes.

The dependent variable of this study is the frequency of triathlon participation. It was surveyed by the total hours of triathlon activity per week, i.e., the time for training and races [AMATEUR_ ACTIVITY].

The regressor of primary interest is the individually perceived relevance of professional triathlon, which was measured with the Sports Involvement Inventory (SII), a scale developed by Shank and Beasley (1998). The SII is an eight-item semantic differential with seven response categories, which is displayed in Table 1. It was designed to capture an individual's personal relevance, i.e., the involvement, of a sports object and it has already been employed in other studies (Ko, Kim, & Claussen, 2008; Tokuyama & Greenwell, 2011). Adapted to triathlon, each respondent had to assess professional triathlon on eight item pairs. After recoding, this leads to a quasi-metric score [PRO_RELEVANCE] that displays the individually perceived relevance of professional triathlon.

For the first instrument, two binary variables were generated based on the postal codes indicating whether an individual lives within 60 minutes' travel time of the city centre of Hamburg or the city centre of Roth. Combining these, the variable [HAMB_ROTH] indicated whether a person lives within 60 minutes' travel time of Hamburg or Roth. For the second instrument, the respondents were asked on a binary variable whether professional triathlon had influenced their decision to start their triathlon activity [START].

For control purposes, the age [AGE], gender [FEMI], marital status [MARRIED], occupational status [FULLTIME], and the size of the households [FAMILY] were surveyed. In addition, the degree of urbanisation [URBAN] and the years of amateur triathlon experience [EXPER] were surveyed. Table 2 summarizes all variables.

Model specification and estimation A standard linear model is considered:

In y = [alpha] + yd + [beta]x + [epsilon], (1)

with logarithmized hours of triathlon activity as dependent variable, standardized relevance of professional triathlon as the regressor of primary interest, a K x 1 vector of observables, and an error term e. The dependent variable was included in logarithmic form because it is assumed that [PRO_RELEVANCE] influences [AMATEUR_ACTIVITY] non-linearly: amateur triathletes with higher time investments were assumed to receive a stronger motivational effect of professional triathlon (Mutter & Pawlowski, 2013). The transformation is straightforward because the variable takes on only positive values above zero (Wooldridge, 2010).

In line with prior studies that have modeled the frequency of sport participation, the observables contained gender, age and age squared, experience and experience squared, marital status, size of household, occupational status, and degree of urbanization.

Two different models were estimated. Model 1 is based on an OLS estimation with the potentially endogenous regressor [PRO_RELEVANCE]. Model 2 is based on a 2SLS estimation with the residence of the amateur triathletes [HAMB_ROTH] and their previously stated relevance of professional triathlon [START] as instruments for regressor [PRO_RELEVANCE]. Instead of using both variables as an instrument each, the 2SLS estimator builds a linear combination of both variables, which usually obtains the most efficient results (Angrist & Krueger, 2001). To test the relevance of both instruments, the first stage results and F-statics of excluded instruments were reported. In addition, a Hansen J Test was employed to test the overidentification restriction. All estimations were executed with the STATA command ivreg2 (Baum, Schaffer, & Stillman, 2007). The reported standard errors are robust to heteroscasticity, which is likely in cross-section data.

Results

Table 3 shows the descriptive statistics of the total sample and conditional on the two instruments.

Available data for selected sample characteristics from Wicker, Prinz, and Weimar (2012) who surveyed German triathletes in 2010 and 2011, has similar mean values: one quarter of the censored sample is female (21.7%), and the average triathlete is approximately 38 years old (36.69). (4) On average, the respondents have participated in triathlon for 7.44 years (7.37) and spend 9.48 hours per week participating in triathlon (9.14). Forty-five percent are married, 71% work fulltime, and the average household consists of approximately three persons. The distribution of the degree of urbanisation is rather u-shaped, with more amateurs living in cities with under 20,000 inhabitants and over 200,000 inhabitants, respectively. Fifteen percent of the triathletes live within 60 minutes' travel time of the city centres of Hamburg or Roth, and approximately 18% report an influence of professional triathlon on their decision to initially start their own amateur triathlon activity.

With 5.57 scale points (on a scale from 1 to 7), the average relevance of professional triathlon in the sample is rather strong. The conditional statistics deliver a first insight in the correlation between the endogenous regressor and the instruments. Both amateurs living next to Hamburg or Roth and amateurs whose starting decision had been influenced by professional triathlon report a higher relevance of professional triathlon on average. According to a U-Test, both correlations are statistically significant.

Table 4 displays the results of both models, including the first stage regression results of the 2SLS estimation. In both models, variance inflation factors are below 5, indicating no signs of serious multicollinearity. (5)

Overall, the impact of [PRO_RELEVANCE] on [AMATEUR_ACTIVITY] is positive and significant for both models; according to the OLS results, amateurs who report a higher relevance of professional triathlon participate approximately 11% more in amateur triathlon. The 95% confidence interval is tight, ranging from 7.56% to 14.86%. At the mean of 9.48 hours, this equals between 44 and 87 minutes more amateur triathlon each week. According to the 2SLS results, the size of [PRO_RELEVANCE] is even larger. Amateurs who report a higher relevance of professional triathlon participate approximately 18% more in amateur triathlon themselves.

The variance of an instrumental variable estimation is always larger than the variance of an OLS estimation (Wooldridge, 2010). Indeed, the standard error of the coefficient of [PRO_RELEVANCE] from the 2SLS regression is approximately four times larger than the corresponding OLS standard error. Consequently, the 95% confidence interval is also much larger, ranging from 2.5% to 33.2%. It even contains the corresponding interval of the OLS regression. At the mean of 9.48 hours, this equals between 15 and 195 minutes more amateur triathlon each week.

It is worth noting that both instruments had a significant effect on [PRO_RELEVANCE] with the effect of [START] being approximately twice as large as the effect of [HAMB_ROTH]. According to the Kleibergen-Paap statistics, the model is assumed to be identified (Baum et al., 2007). Finally, the Sargan-Hansen statistic testing the overidentification restrictions is very low (p-value: 0.4067). Based on this result, the null hypothesis of valid instruments cannot be rejected (Wooldridge, 2010).

Regarding the other variables, the effect of experience on hours of amateur participation, which is jointly significant, seems to be inverse u-shaped. All time constraints, namely being married, the size of the family and the occupational status, are negatively correlated with hours of triathlon activity. However, only being married is statistically significant (Dawson & Downward, 2011, 2013; Eberth & Smith, 2010; Garcia et al., 2011; Humphreys & Ruseski, 2006, 2007, 2009).

Discussion

This study analyzes whether the individually perceived relevance of professional triathlon increases the demand for amateur triathlon participation. The OLS estimation reveals that relevance of professional triathlon is correlated with 44 to 87 minutes more amateur participation each week. This result is in line with the study of Dawson and Downward (2013), who revealed that watching professional sports on television was correlated with 67 minutes more sport participation per week. Due to the large standard error of the 2SLS estimation, the 95% confidence interval of the coefficient of interest increases to 15 to 195 minutes more triathlon participation per week. The results of the 2SLS estimation demonstrate evidence for a causal effect of professional sports on amateur sport participation in our setting. This effect is in line with the theoretical framework of sporting role models.

Both employed instruments have a significant effect on the perceived relevance of professional triathlon. However, the effect of the residence of the triathletes is comparatively low. In combination with the finite sample, the results could be seriously biased and have large standard errors (Bound, Jaeger, & Baker, 1995). The instrument relevance could be challenged by the new media. Although television coverage of professional triathlon is less common in Germany, nowadays nearly everybody has the opportunity to follow professional triathlon on the Internet. (6) Furthermore, besides Hamburg and Roth, there are cities in Germany where smaller events for professionals (and amateurs) take place. Examples are Darmstadt, Dusseldorf, Gladbeck and Hannover (DTU, 2012). On these occasions, high-level triathlon could also be available. However, the regional media coverage, which conveys availability to non- attendants, is not comparable with the coverage of the two international events in Hamburg and Roth. Regarding the second instrument, the exogeneity of the previously stated relevance of professional triathlon could be challenged. The measurement of the variable may be biased due to the cross-sectional nature of the data. Although the variable should display the relevance of professional triathlon at the starting point of their own sport participation, all variables were measured at the same point in time. Therefore, it is possible that the previously stated relevance of professional triathlon has accidentally been influenced by their own participation. Nevertheless, according to the very low Sargan-Hansen statistic of the 2SLS estimation, both instruments are assumed to be valid. This statistic is especially convincing because both instruments were theoretically differently derived.

Nevertheless, the quite similar estimates of OLS and 2SLS might also indicate that the reverse causal influence (identified to potentially cause bias in the OLS estimation) is not severe. According to Hill and Robinson (1991), triathlon participation can be categorized as a form of fanatic consumer behavior. Fanatic consumers demonstrate a high obligation and devotion to their activity, which is apparent in the high time and money investments of triathletes (Wicker et al., 2012). Consequently, the latent sport affinity is probably very high for all amateur triathletes, regardless of their frequency of participation. In addition, triathletes often participate in one of the three individual sports prior to their triathlon activity (Tribe Group, 2009). Therefore, they might feel a certain similarity to professional sports based on their acquired skill level in the individual sport regardless of their frequency of triathlon participation. On the other hand, the substitution effect between professional triathlon and triathlon participation could be substantial. The average triathlete spends over nine hours per week participating. In addition, he is working fulltime in a well-paid job. Consequently, the scarce time available has to be allocated carefully. The more time spent for triathlon participation, the less time is available for professional triathlon interest (Lera-Lopez et al., 2012). The causality of such a substitution effect is important for sport managers and politicians. In this setting, it seems that amateur participation potentially suppresses time spent on professional triathlon activities. Such an effect would be alerting for managers of professional sports who are concerned with attendance rates for spectatorship or numbers of television viewers.

Implications

The results of this study could be utilized by sport managers in different areas. Managers of amateur triathlon events could engage professional athletes to make a special appearance in order to increase the demand for participation. In addition, the triathlon goods and services industries could incorporate professional sports in their marketing strategies, using the motivational power to increase their own revenues. The running brand Asics, for example, currently sponsors the German professional triathletes Justus Steffen and Maik Petzold. To confirm further this idea, the choice of triathlon brands for American amateurs was found to be influenced by brands that sponsor both events and professional athletes (Tribe Group, 2009). Furthermore, policy interventions targeting active sports participants could utilize professional sports in programs or promotions to increase the frequency of participation. Making professional sports more available, especially in sports with lower media coverage, is of crucial importance. Finally, compared with other sports, amateur participants in triathlon and marathon events are offered the possibility to line-up alongside of professional athletes, which increases the similarity between professional and amateur athletes. Applying this concept to other sporting events could utilize the motivational effect of professional sports and potentially increase attendance rates of amateur participants.

Limitations

Conclusions about the population of all triathletes in Germany have to be drawn cautiously. The sample was non-randomly drawn and only consists of organized triathletes. According to the DTU (2012), there exist three times as many unorganized triathletes in Germany. Therefore, the results might not be representative of the population of all amateur triathletes in Germany. Nevertheless, the sample characteristics matched very well with the samples of prior studies that also survey amateur triathletes (Tribe Group, 2009; Wicker et al., 2012). Another limitation is concerned with the rather low explained variance of the estimated models. Comparable with other cross-sectional studies in sport participation research, the R squared of the models ranged from 7% to 9% (Breuer & Wicker, 2008). The decisions of fanatic consumers cannot always be explained by rational choice (Holbrook & Hirschman, 1982). Therefore, the observable covariates could not capture all of the variance of the frequency of participation properly. In addition, Wicker et al. (2012) found that the population of amateur triathletes is very heterogeneous. They identified three segments of triathletes (serious pursuiters, sport lovers and socializes) with different behavioral patterns and socio-economic characteristics. According to these results, treating all triathletes as a homogeneous group might aggravate behavioral analysis.

Conclusion

The present study provides evidence that an individually perceived relevance of professional triathlon causally increases the time spent on amateur triathlon participation. However, the transfer of these results to other countries and/or other sports is limited due to special characteristics of triathlon, such as the limited number of participants in general or the high physiological requirements.

Nevertheless, further research in this area is promising. Increasing sport participation is still a major (sports) policy concern and, in addition, of highest relevance for sport managers of different industries. However, it has to be mentioned that the methodological approach to evaluate an isolated causal effect of a professional sports stimulus faces some serious difficulties. Advanced research designs and longitudinal data will be necessary to deliver further insights.

Authors' Note

Special credit is due to Rob Simmons and two anonymous referees, who significantly helped to improve the article.

References

Angrist, J. D., & Krueger, A. B. (2001). Instrumental variables and the search for identification: From supply and demand to natural experiments. Journal of Economic Perspectives, 15(4), 69-85.

Angrist, J. D., & Pischke, J.-S. (2009). Mostly harmless econometrics: An empiricist's companion. Princeton: Princeton University Press.

Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory. Englewood Cliffs, NJ: Prentice Hall.

Baum, Ch. F., Schaffer, M. E., & Stillman, S. (2007). Enhanced routines for instrumental variables/GMM estimation and testing. The Stata Journal, 7(4), 465-506.

Becker, G.S. (1965). A theory of the allocation of time. The Economic Journal, 75, 493-513.

Bound, J., Jaeger, D. A., & Baker, R. M. (1995). Problems with instrumental variables estimation when the correlation between the instruments and the endogeneous explanatory variable is weak. Journal of the American Statistical Association, 90(430), 443-450.

Bourdieu, P. (1988). Program for a sociology of sport. Sociology of Sport Journal, 5, 53-161.

Breuer, C. & Hallmann, K. (2011). Diegesellschaftliche Relevanz des Spitzensports in Deutschland. Koln: Sportverlag Straufi.

Breuer, C., & Wicker, P. (2008). Demographic and economic factors concerning the inclusion in the German sport system--A micro-analysis of the years 1985 to 2005. European Journal for Sport and Society, 5(1), 33-42.

Cawley, J. (2004). An economic framework for understanding physical activity and eating behaviors. American Journal of Preventive Medicine, 27(3), 117-125.

Challenge Roth. (2012). Retrieved from http://www.challenge- roth.com/de/index.html

Dawson, A., & Downward, P. (2011). Participation, spectatorship and media coverage in sport: some initial insights. In W. Andreff (Ed.), Contemporary issues in sports economics. Participation and professional team sports (pp.15-42). Cheltenham: Northampton, Mass: Edward Elgar.

Dawson, A., & Downward, P. (2013). The relationship between participation in sport and sport volunteering: An economic analysis. International Journal of Sport Finance, 8(1), 75-92.

Deutsche Triathlon Union. (2012). Retrieved from http://www.dtu- info.de/amateursport.html

Deutscher Olympischer Sportbund. (2012). Bestandserhebung.

Downward, P. (2007). Exploring the economic choice to participate in sport: Results from the 2002 General Household Survey. International Review of Applied Economics, 21, 633-653.

Downward, P., Lera-Lopez, F., & Rasciute, S. (2011). The zero-inflated ordered probit approach to modelling sports participation. Economic Modelling, 28, 2469-2477.

Downward, P., & Rasciute, S. (2011). Does sport make you happy?: An analysis of the well-being derived from sports participation. International Review of Applied Economics, 25(3), 331-348.

Downward, P., & Riordan, J. (2007). Social interaction and the demand for sport: An economic analysis. Contemporary Economic Policy, 25(4), 518-537.

Eberth, B., & Smith, M. D. (2010). Modelling the participation decision and duration of sporting activity in Scotland. Economic Modelling, 27, 822-834.

Feddersen, A., Jacobsen, S. &. Maennig, W. (2009). Sport heroes and mass participation: The (double) paradox of the "German tennis boom" (Working paper No. 29). University of Hamburg.

Frawley, S., & Cush, A. (2011). Major sporting events and participation legacy: The case of the 2003 Rugby World Cup. Managing Leisure, 16(1), 65-76.

Garcia, J., Lera-Lopez, F., & Suarez, M. J. (2011). Estimation of a structural model of the determinants of the time spent on physical activity and sport: Evidence for Spain. Journal of Sports Economics, 12(5), 515-537.

Gibson, D. E. (2003). Developing the professional self-concept: Role model construals in early, middle, and late career stages. Organizational Science, 14(5), 591-610.

Gibson, D. E. (2004). Role models in career development: New directions for theory and research. Journal of Vocational Behavior, 65, 134-156.

Hallmann, K., Breuer, C. & Kuhnreich, B. (2013). Happiness, pride and elite sporting success: What population segments gain most from national athletic achievements? Sport Management Review, 16(2), 226-235.

Heckhausen, H., & Gollwitzer, P. M. (1987). Thought contents and cognitive functioning in motivational versus volitional states of mind. Motivation and Emotion, 11(2), 101-120.

Hill, R. P., & Robinson, H. (1991). Fanatic consumer behavior: Athletics as a consumption experience. Psychology & Marketing, 8(2), 79-99.

Hogan, K., & Norton, K. (2000). The "price" of Olympic gold. Journal of Science and Medicine in Sport, 3(2), 203-218.

Holbrook, M., & Hirschman, E. C. (1982). The experiential aspects of consumption: Consumer fantasies, feelings, and fun. Journal of Consumer Research, 9, 132-140.

Huang, H., & Humphreys, B. R. (2012). Sports participation and happiness: Evidence form US microdata. Journal of Economic Psychology, 33(4), 776-793.

Humphreys, B. R., MacLeod, L., & Ruseski, J. (2011). Physical activity and health outcome: Evidence from Canada (Working Paper No. 2011-06). University of Alberta Department of Economics.

Humphreys, B. R., Maresova, K., & Ruseski, J. (2012). Institutional factors, sport policy, and individual sport participation: An international comparison (Working Paper No. 2012-01). University of Alberta Department of Economics

Humphreys, B. R. & Ruseski, J. (2006). Economic determinants of participation in physical activity and sport (Working Paper series No.06-13). International Association of Sports Economists.

Humphreys, B. R., & Ruseski, J. (2007). Participation in physical activity and government spending on parks and recreation. Contemporary Economic Policy, 25(4), 538-552.

Humphreys, B. R. & Ruseski, J. (2009). The economics of participation and time spent in physical activity (Working Paper No.2009-09). University of Alberta Department of Economics.

Humphreys, B. R. & Ruseski, J. (2011). An economic analysis of participation and time spent in physical activity. B.E. Journal of Economic Analysis and Policy, 11(1), 1-36.

International Triathlon Union. (2012). Retrieved from http://www.hamburg- triathlon.org/

IpsosMORI. (2009). Satisfaction with the quality of the sport experience (SQSE) Survey. London: Sport England.

Jung, J. (1986). How useful is the concept of role model? A critical analysis. Journal of Social Behavior and Personality, 1(4), 525-536.

Kassing, J. W., & Sanderson, J. (2010). Fan-athlete interaction and Twitter tweeting through the Giro: A case study. International Journal of Sport Communication, 3(1), 113-128.

Kenyon, G. S., & McPherson, B. D. (1973). Becoming involved in physical activity and sport: A process of socialization. In L. G. Rarick (Ed.), Physical activity. Human growth and development (pp. 303-332). New York: Academic Press.

Ko, Y. J., Kim, K., & Claussen, C. L. (2008). The effects of sport involvement, sponsor awareness and corporate image on intention to purchase sponsors' products. International Journal of Sports Marketing & Sponsorship, 9(2), 81-94.

Lera-Lopez, F., Ollo-Lopez, A., & Rapun-Garate, M. (2012). Sports spectatorship in Spain: Attendance and consumption. European Sport Management Quarterly, 12(3), 265-289.

Lines, G. (2007). The impact of media sport events on the active participation of young people and some implications for PE pedagogy. Sport, Education and Society, 12(4), 349- 366.

Lockwood, P., & Kunda, Z. (1997). Superstars and me: Predicting the impact of role models on the self. Journal of Personality and Social Psychology, 73(1), 91-103.

Lyle, J. (2009). Sporting success, role models and participation: A policy related review (Research Report no.101). Edinburgh: Sportscotland.

McCarthy, B. (2011). From shanfan to gymnastike: How online fan texts are affecting access to gymnastics media coverage. International Journal of Sport Communication, 4(3), 265-283.

McCartney, G., Thomas, S., Thomson, H., Scott, J., Hamilton, V., Hanlon, P., & Bond, L. (2010). The health and socioeconomic impacts of major multi-sport events: Systematic review (1978-2008). British Medical Journal, 340, c2369.

Mutter, F., & Pawlowski, T. (2013). Role models in sports--Can success in professional sports increase the demand for amateur sport participation. Sport Management Review, doi: 10.1016/j.smr.2013.07.003.

Pawlowski, T., & Breuer, C. (2012a). Die finanzpolitische Bedeutung des Sports in Deutschland. Wiesbaden: Springer-Gabler Research.

Pawlowski, T. & Breuer, C. (2012b). Expenditure elasticities of the demand for leisure services. Applied Economics, 44(26), 3461-3477.

Pawlowski, T., Downward, P., & Rasciute, S. (2011). Subjective well-being in European countries --On the age-specific impact of physical activity. European Review of Ageing and Physical Activity, 8(2), 93-102.

Pawlowski, T., Downward, P. & Rasciute, S. (2013). Does national pride from international sporting success contribute to well-being? An international investigation. Sport Management Review, doi: 10.1016/j.smr.2013.06.007.

Preuss, H., Alfs, Ch., & Ahlert, G. (2012). Sport als Wirtschaftbranche. Der Sportkonsum privater Haushalte in Deutschland. Wiesbaden: Springer-Gabler Research.

Preuss, H., Seguin, B., & O'reilly, N. (2007). Profiling major sport event visitors: The 2002 Commonwealth Games. Journal of Sport & Tourism, 12(1), 5-23.

Rasciute, S., & Downward, P. (2010). Health or happiness?: What is the impact of physical activity on the individual? Kyklos, 63(2), 256-279.

Ruhle, A. (2003). Sportprofile deutscher Fernsehsender 2002. Media Perspektiven, 5, 216-230.

Schmidt, S. L., & Hogele, D. (2011). Inspiringpersonlities. Deutschland braucht den Superstar. Die gesellschaftliche Bedeutung von Vorbildern im ProfifuBball. ISBS Research Series, Issue 4. EBS Business School. Oestrich-Winkel.

Schoenstedt, L. J., & Reau, J. (2010). Running a social-media newsroom: A case study of the Cincinnati Flying Pig Marathon. International Journal of Sport Communication, 3(3), 377-386.

Scitovsky, T. (1976). The joyless economy. New York, NY: Oxford University Press.

Shank, M., & Beasley, F. (1998). Fan or fanatic: Refining a measure of sports involvement. Journal of Sport Behavior, 21(4), 435-444.

Siegfried, J., & Zimbalist, A. (2000). The economics of sports facilities and their communities. Journal of Economic Perspectives, 14(3), 95-114.

Sportscotland. (2004). Curling success and its impact on participation (Research Report no. 92). Retrieved from

http://www.sportscotland.org.uk/sportscotland/Documents/Resources/CurlingStudyRe searchR eportNo92.pdf

Stigler, G., & Becker, G. S. (1977). De gustibus non est disputandum. The American Economic Review, 67(2), 76-90.

Thrane, C. (2001). Sport spectatorship in Scandinavia: A class phenomenon? International Review for the Sociology of Sport, 36(2), 149-163.

Tokuyama, S., & Greenwell, C. T. (2011). Examining similarities and differences in consumer motivation for playing and watching soccer. Sport Marketing Quarterly, 20(3), 148-156.

Triathlon-Deutschland (2012). Retrieved from http://www.triathlon- deutschland.de/triathlonsuche.html

Tribe Group. (2009). The mind of the triathlete: Market research report. Colorado Springs: USA Triathlon.

UK SPORT (2011). The inspirational effect of major sporting events (Research Report). Retrieved from http://www.eventimpacts.com/pdfs/the_inspirational_impact_of_major_sporting_ events.pdf

USA Triahtlon. (2012). About USAT. Retrieved from http://www.usatriathlon.org/aboutusat/about.aspx

Weed, M. (2009). The potential of the demonstration effect to grow and sustain participation in sport (Review Paper for Sport England from the SPEAR). Canterbury, Kent, UK: Canterbury Christ Church University. Retrieved from http://sm.avito.nl/wp-content/uploads/ 2013/02/Weed-M.-2009.-The-potential-of-the-demonstration-effect-to-grow-and- sustainparticipation-in-sport.-UK-Canterbury-Canterbury-Christ-Church- University-SPEAR.pdf

Weed, M., Coren, E., & Fiore, J. (2009). A systematic review of the evidence base for developing a physical activity and health legacy from the London 2012 Olympic and Paralympic Games. London: Department of Health.

Weed, M., Coren, E., Fiore, J., Wellard, I., Mansfield, L., Chatziefstathiou, D., & Dowse, S. (2012). Olympic and Paralympic Games: a policy-led systematic review. Perspectives in Public Health, 132 (2), 75-80.

Wicker, P., Hallmann, K., Prinz, J., & Weimar, D. (2012). Who takes part in triathlon? An application of lifestyle segmentation to triathlon participants. International Journal of Sport Management and Marketing, 12(1), 1-24.

Wicker, P., Prinz, J., & Weimar, D. (2012). Big spenders in a booming sport: Consumption capital as a key driver of triathletes sport-related expenditure (Working Paper No. 2011-06 No. 004). Lehrstuhl fur Allgemeine Betriebswirtschaftslehre. Universitat Duisburg- Essen.

Wooldridge, J. (2010). Econometric analysis of cross section and panel data. Cambridge, MA: MIT Press.

Zaichkowsky, J. L. (1985). Measuring the involvement construct. Journal of Consumer Research, 12, 341-352.

Zanger, C., & Schweizer, J. (2004). Das Eventpotential der 150 wichtigsten Sportarten: Ein wissenschaftliches Marktforschungsprojekt (1st ed.). Munchen: Schweizer GmbH.

Zubayr, C. & Gerhard, H. (2008). Tendenzen im Zuschauerverhalten. Fernsehgewohnheiten und Fernsehreichweiten im Jahr 2007. Media Perspektiven.

Felix Mutter [1] and Tim Pawlowski [2]

[1] Pricewaterhouse Coopers AG WPG

[2] University of Tubingen

Endnotes

(1) In Germany, more than 150,000 people actively participate in triathlon (Preuss, Alfs, & Ahlert, 2012). Triathlon is an increasingly popular sport worldwide. For instance, in the last five years the number of organized triathletes increased by over 80% (Deutscher Olympischer Sportbund [DOSB], 2012). Similar trends are visible in the USA (USA Triathlon, 2012). Unlike in other sports, the distinction between professional triathletes and amateur triathletes is not selective. On the one hand, amateur triathletes participate on a recreational basis and attend none or few amateur races per season. On the other hand, professional triathletes attend high-level competitions like the ITU World Series or the IOC Olympic Games. Other types of triathletes with different behavioural patterns could be arranged along this continuum. Throughout this paper, the term "professional triathlon" refers to the former, high-level triathlon. The term "amateur triathlon" refers to all other triathletes.

(2) Neoclassical economic frameworks are also capable to generate predictions about the effects of professional sports on amateur sport participation. One example would be the concept of consumption capital, derived by Stigler and Becker (1977).

(3) Indeed, 90.07 % of the total sample reported no major change of residence within the last four years. According to a chi square test, there is no correlation between the residence near Hamburg or Roth and the change of residence within the last four years (p-value: .3010).

(4) According to the Tribe Group (2009), the average American triathlete is 38 years old, and approximately one third of American triathletes are female.

(5) Exceptions are the variance inflation factors of [EXPER] and [EXPER2]. This is not a concern because these variables are related to each other by the construction of the model. Furthermore, these variables are only included in the model for control purposes. The coefficient of interest is not affected by the mulitcollinearity between them (Wooldridge, 2010).

(6) Examples of the use of the Internet for media coverage of niche sports can be found in Kassing and Sanderson (2010) for cycling, McCarthy (2011) for gymnastics, and Schoenstedt and Reau (2010) for running.

Felix Mutter is a former PhD student at the Institute of Sport Economics and Sport Management at the German Sport University Cologne. His research interests include tax and transfer pricing issues in sport industries and analysis of sports demand.

Tim Pawlowski is a professor of sport economics, sport management, and sport media research in the Faculty of Economics and Social Science. His research interests include the analysis of sports demand, the financing of sport systems, and the economics of league competitions.

Table 1: The Sports Involvement Inventory, adapted to triathlon (Shank
& Beasley, 1998). Respondents answered the question: "For me,
professional triathlon is..."

1              2    3    4    5    6    7

Boring                                  Exciting
Interesting                             Uninteresting
Valuable                                Worthless
Appealing                               Unappealing
Useless                                 Useful
Not needed                              Needed
Irrelevant                              Relevant
Important                               Unimportant

Table 2: Relevant Variables

Variable            Description                             Scale

AMATEUR_ ACTIVITY   Hours of amateur triathlon activity     Metric
                      per week
PRO_RELEVANCE       Individually perceived relevance of     Ordinal
                      professional triathlon
HAMB_ROTH           Residence within 60 minutes travel      Binary
                      time of the city centres of
                      Hamburg or Roth (1 = yes)
START               Decision to start triathlon             Binary
                      activity was influenced by
                      professional triathlon (1 = yes)
AGE                 Years of age                            Metric
FEMI                Feminine (1 = yes)                      Binary
FULLTIME            Fulltime employment (1 = yes)           Binary
MARRIED             Married (1 = yes)                       Binary
FAMILY              Household size in numbers               Metric
EXPER               Years of active triathlon               Metric
                      participation
URBAN1              Residence: Under 20.000 inhabitants     Binary
                      (1 = yes)
URBAN2              Residence: 20.000-49.999                Binary
                      inhabitants (1 = yes)
URBAN3              Residence: 50.000-99.999                Binary
                      inhabitants (1 = yes)
URBAN4              Residence: 100.000-199.999              Binary
                      inhabitants (1 = yes)
URBAN5              Residence: 200.000 inhabitants and      Binary
                      more (1 = yes)

Table 3: Descriptive statistics with mean and standard deviation for
metric and ordinal variables and percentage yes for binary variables
for the total sample and conditional on both instruments.

                    Total sample    HAMB_ROTH = 1   START = 1

PRO_RELEVANCE       05.57 (01.04)   05.78 (0.838)   05.98 (0.764)
AMATEUR_ ACTIVITY   09.48 (04.19)   09.76 (04.16)   10.25 (04.01)
EXPER               07.44 (06.91)   06.65 (05.97)   06.29 (05.84)
AGE                 37.81 (11.55)   38.43 (10.44)   36.20 (10.74)
FEMI                25.36 %         25.22 %         17.07 %
FULLTIME            71.00 %         78.26 %         72.14 %
MARRIED             44.75 %         46.49 %         29.08 %
FAMILY              02.73 (02.02)   02.53 (01.22)   02.51 (01.32)
URBAN1              26.80 %         25.22 %         25.53 %
URBAN2              18.95 %         08.70 %         14.18 %
URBAN3              08.89 %         01.74 %         06.36 %
URBAN4              06.27 %         03.48 %         07.09 %
URBAN5              36.99 %         60.87 %         46.81 %
HAMB_ROTH           15.03 %                         18.44 %
START               18.43 %         22.61 %
Observations        N=765           N=115           N=141

Table 4: Coefficient of the OLS and 2SLS estimations including the
first stage results. Logarithmic hours of weekly amateur triathlon
participation as dependent variable. Robust standard errors in
parentheses.

                             Model 1 (OLS)        Model 2 (2SLS)
Variable                                          First stage

PRO_RELEVANCE                0.112 (.019) ***
FEMI                         - 0.060 (.038)       0.067 (.080)
AGE                          - 0.001 (.003)       - 0.003 (.006)
[AGE.sup.2]                  0.000 (.000)         0.000 (.000)
EXPER                        0.017 (.009) *       0.027 (.018)
[EXPER.sup.2]                - 0.001 (.000) *     - 0.001 (.001)
FULLTIME                     - 0.022 (.042)       0.064 (.100)
MARRIED                      - 0.109 (.041) ***   - 0.051 (.084)
FAMILY                       - 0.009 (.009)       0.015 (.010)
URBAN1                       Reference            Reference
URBAN2                       0.004 (.049)         - 0.051 (.115)
URBAN3                       0.049 (.065)         0.020 (.129)
URBAN4                       - 0.014 (.070)       0.201 (.163)
URBAN5                       - 0.008 (.040)       0.032 (.085)
HAMB_ROTH                                         0.217 (.087) **
START                                             0.478 (.078) ***
Constant                     2.251 (.084) ***     - 0.274 (.160) *

Observations                 759                  759
Centred [R.sup.2]            0.0882               0.050
Kleibergen-Paap rk LM
Kleibergen-Paap rk Wald F
Sargan-Hansen

Variable                     Second stage

PRO_RELEVANCE                0.179 (.078) **
FEMI                         - 0.061 (.039)
AGE                          - 0.001 (.003)
[AGE.sup.2]                  0.000 (.000)
EXPER                        0.016 (.010)
[EXPER.sup.2]                - 0.001 (.000) *
FULLTIME                     - 0.029 (.044)
MARRIED                      - 0.104 (.042) **
FAMILY                       - 0.009 (.010)
URBAN1                       Reference
URBAN2                       0.010 (.051)
URBAN3                       0.050 (.065)
URBAN4                       - 0.027 (.074)
URBAN5                       - 0.014 (.041)
HAMB_ROTH
START
Constant                     2.258 (.085) ***

Observations                 759
Centred [R.sup.2]            0.0685
Kleibergen-Paap rk LM        35.538
Kleibergen-Paap rk Wald F    15.415
Sargan-Hansen                0.688

* 10% Significance
** 5% Significance
*** 1% Significance