Determinants of individual consumption on sports attendance in Spain.

Lera-Lopez, Fernando ; Rapun-Garate, Manuel ; Suarez, Maria Jose 等

Introduction

Over the last 20 years, the sport market has significantly increased its role in the economy of many countries. On the one hand, somewhere in the range of 150-175 million Europeans participate in some form of sport on a regular basis (Amnyos Consultants, 2008). This contrasts with about 118 million Americans; that is, according to a 2005 survey of the U.S. (Humphreys & Ruseski, 2009) in which sales of sports goods increased by 79%--$50,725 million to $90,802 (U.S. Census Bureau, 2010)--between 1990 and 2007, while total attendance at sporting events in 2005 was estimated at just over 277 million. In addition, personal consumption expenditure on sporting events increased by 285% from $4.8-18.5 billion in the period between 1990-2007 (U.S. Census Bureau, 2010). New evidence obtained by Amnyos Consultants (2010) assessed the total E.U. market for all sports goods and services to be somewhere in the range of 550-600 billion [euro] in 2004.

In Spain, the importance given to sport has changed drastically over a period of only 30 years. One example is the striking increase in mass sports participation. In 1975, 22% of all Spanish adults practiced some kind of sport, while the figure had risen to 45% by 2010 (Spanish High Council for Sport, 2010). The last few decades have also seen a significant increase in the economic importance of sport in Spain. In 1990, the first estimations made by Alonso, Ruesga, Saez, and Vicens (1991) estimated the economic importance of sport at 1.2% of national GDP. Recent estimations made by Lera-Lopez, Rapun-Garate, and Aguirre (2008) put the figure at 2.38% of GDP in 2006--on par with estimations for other European countries where sports constitutes a significant economic activity, such as the U.K. (Sport Industry Research Centre, 2010).

Thus, there is clear evidence of the growing economic importance of the sports sector, with sport accounting for a significant percentage of household consumption spending (Sport Industry Research Centre, 2008). In many European countries, the financial structure of sport is heavily weighted toward household consumption, which, in 2005, accounted for 50% of total sports expenditure in Europe; local authorities and state contributions accounted for 24% and 12% respectively (Amnyos Consultants, 2008). A cross-country comparison of the structure of sports finance in Europe in 1990 (Andreff, 2006) and 2005 (Amnyos Consultants, 2008) reveals a significant increase in the importance of household spending in the structure of sport financing in countries such as Germany, Finland, France, Portugal, and Sweden. A rough breakdown shows that household consumption is more concentrated on sports services than sports goods (Andreff, 2006).

The importance of sports consumption as a share of overall household consumption highlights the need for a deeper understanding of the variables that influence sports participation and consumption, and it reveals the barriers and incentives involved in sports demand. Individual sports demand can be viewed from two perspectives: active and passive participation. Active sports participation is mainly associated with mass participation in sporting activities. Passive sports participation involves sports consumption, either in a live setting or via live or recorded broadcast media (Dawson & Downward, 2009). Generally speaking, passive sports participation embraces professional and amateur team sports and sporting events.

The aim of this paper is to analyze the key determinants of consumption on sports attendance in Spain and to contribute new empirical evidence about this relevant issue to sports economics literature. The data for our empirical analysis came from a national survey carried out in 2007 with the financial support of the Spanish High Council for Sport. The survey uses a sample of 640 participants stratified by gender and age; it is considered to be representative of the overall Spanish population aged 18 years and older. The survey results provide detailed information about active and passive sports participation and consumption data for both activities. In the empirical analysis, we consider two alternative specifications for the comparative analysis of the results. In both situations, we assume that the decision whether or not to attend an event and the subsequent consumption decision at the event follow a sequential process. The first specification assumes that the two decisions are independent, whereas the second allows for correlation between them, resulting in a double-hurdle model, also called the Tobit model with selectivity. As far as we know, sports attendance and consumption have never before been explored as two correlated decisions using a double-hurdle model. Also, against a backdrop of scant empirical evidence about sport-related consumption, this is the first study to address the subject of consumer expenditure on attending professional and amateur sporting events or to jointly analyze the key determinants of attendance and consumption decisions for a single country.

Individual Sport-Related Consumption

Theoretical Background

Individual sports consumption decisions could be analyzed within the framework of the neoclassical consumption theory or the new household economics approach developed by Gary Becker (1965; Downward, Dawson, & Dejonghe, 2009). Both approaches assume that individuals' consumption decisions are aimed at maximizing their utility subject to budget and time constraints. Implicitly, sport uses non-economic and non-obligated time--the demand for which reflects the trade-off between the utility derived from the consumption of goods and the opportunity cost of an hour of sport (e.g., participation, attendance, TV). Thus, monetary and time restrictions limit the individual's opportunities for active sport (Wicker, Breuer, & Pawlowski, 2010). In the basic neoclassical model, utility depends on the amount of goods consumed, whereas in Becker's approach, preferences are a function of commodities produced by the household using time and goods.

Another set of theories has considered other determinants of sports demand. The post-Keynesian approach emphasizes that individual behavior is linked to broader aspects of social behavior, such as the importance of social values; it also posits that sports consumption involves the habit of learning-by-doing and spillover effects and that the agent's preferences are endogenous (Lavoie, 2004). In the sociological analysis of sport, sporting styles and individual preferences are linked not only to individual feelings, but to social pressure and the influence of habitus as well. For example, Bourdieu's sociological approach (Bourdieu, 1984) emphasizes the role played by income and education in explaining differences in sporting habits and individual preferences. (1) The psychological approach (Mannell & Kleiber, 1997) places the emphasis on intrapersonal motivation, which changes over the life span. Sensation-seeking, emotional awakening, concern, pleasure, or anxiety can be potential sources of demand for sport and leisure activities (Downward, 2007). In consequence, this approach focuses on constraints on sports decisions at the individual level, minimizing the role played by social constraints. Finally, some ecological models (Beaton, Funk, & Alexandris, 2009; Henderson, 2009) consider that participation is conditioned by individual intrapersonal factors as well as cultural and environmental influences.

Although these approaches have usually been applied to explain consumer spending on active participation in sport, they could also be applied to consumer spending on sports attendance.

Our theoretical framework is the basic neoclassical theory of consumption. According to this theory, people choose the amount of goods and services that maximize their utility--subject to the standard budget constraints. Also, the demand equations are a function of prices, income, and a vector of socio-demographic factors that may influence individual preferences. However, our empirical specification does not estimate a structural model of demand for sporting events because our database is cross-sectional and lacks price data. Therefore, we estimate a reduced-form model of consumer spending on sporting events.

Moreover, following Downward et al. (2009), we assume that sports goods and service-related demand and consumption (e.g., equipment, clothes, facilities, travel) are demands derived from the frequency and intensity of sports participation. Therefore, our hypothesis is that the individual decision is sequential: One first decides whether to attend and then how much to spend. (2)

Empirical Evidence

At the micro-economic level, the empirical evidence about the determinants of sports consumption is based mainly on mass participation in sport. (3) Studies such as Michon, Ohl, and Faber (1987); Lamb, Asturias, Roberts, and Brodie (1992); and Taks, Vanreusel, and Renson (1994, 1995, 1999) emphasize the positive influence of income variables and the significant role of some individual and household characteristics, such as age, gender, educational level, and household size, the degree of urbanization, and sport involvement factors.

The empirical evidence on consumer spending on attending sporting events is more limited than for mass participation in sport. Some studies, such as Taks and Kesenne (2000), are descriptive and quantify the amount spent on attending sporting events as a component of total household sports consumption in Flanders. Taks et al. (1994) limited their analysis to two sports: soccer and golf. Davies (2002), focusing on the population of Sheffield (U.K.), estimated consumer spending on sporting events and showed that it was not correlated with participation in sport. After surveying visitors to the 1995 and 1999 Alamo Bowl college football games, Cannon and Ford (2002) reported that the only factors showing positive correlation with spending at sporting events were high-income level, previous visitation, and out-of-state travel. Other variables--such as educational level, marital status, and ethnicity--were not significant.

Pawlowski and Breuer (2011) have recently analyzed the determinants of expenditure shares in 17 different sports and recreational services for the German continuous household budget survey. They reported a negative relationship between the expenditure share on sporting events and the degree of urbanization and level of education. In addition, households with teenagers and young adults tended to spend proportionally more money on sporting events. Other socio-demographic variables--such as age, occupation, marital status, and size of the household--were not statistically significant. In another paper, Pawlowski and Breuer (2010) estimated expenditure elasticities for sporting events, classifying them as luxury goods.

Finally, other authors have jointly analyzed consumption on active sports participation and attending sporting events. In Spain, Lera-Lopez and Rapun-Garate (2005, 2007) quantified and analyzed--using Tobit and ordered probit models, respectively--the determinants of sports consumption in the region of Navarre. In their earlier study, the authors reported a positive relationship between income and sport expenditure, differentiating between active and passive sports participation and classifying sports goods as a normal good. They identified gender, size of population, and age (Ushaped function) as determinants of sports consumption. In their later study (LeraLopez & Rapun-Garate, 2007), the authors compared the determinants of mass sports participation and related expenditure using ordered probit models in two different samples: the population as a whole and regular participants. Among the former, young and more highly-educated people who reported high-frequency sports participation were more likely to spend more money on sport. Among the sport participants, sports consumption was higher in men than in women; it also varied by social status and occupation and increased with income.

On the other hand, Loyland and Ringstad (2009) applied the Almost Ideal Demand System to the household demand for sport (i.e., sports activities and events) in Norway for the period between 1986-2002. They reported a price elasticity that was not significantly greater than 0 and an income elasticity close to 1. They found substantial variation in demand across household types, with single person households showing the proportionally highest sports consumption levels; households with small children showed the lowest. They also found that the demand for sport was lower in more urbanized settings and in female- and elderly-headed households.

Database and Descriptive Analysis

The Dataset

The data for the study came from a survey designed to provide a detailed database of information on Spanish sporting habits and sport-related consumption; it was funded by the Spanish High Council for Sport, a public institution in charge of matters relating to sport. This survey was conducted by phone in 2007 and covered a sample of 640 individuals between the ages of 18 and 75 years. Subjects from random households were selected from a stratified sample of municipalities, maintaining gender and age quotas consistent with the general population. The confidence level was estimated at 95.5% with a sampling error of +/-3.95%.

The questionnaire was divided into four parts. The first part contained questions to determine the frequency of active participation in sport or any other physical activity, the type of sport practiced, and whether the individual has participated in the organization of sports activities. The second part included questions about different categories of spending on active sports (e.g., equipment, clothes, tickets and fees, travel) and was measured in intervals. The third part of the questionnaire focused on passive sports participation, defined as attendance at amateur or professional sporting events and viewing sporting events on television. The questions were designed to probe the frequency of practice of both types of participation and individual annual spending on sporting events; transport to and from such events; meals, drinks, and promotional products; pay-per-view systems to watch sporting events on television; newspapers, CDs, or DVDs with sports content; and, finally, sports lotteries and betting. All the requested data referred to 2006. The fourth part of the questionnaire was intended to gather socio-demographic data, such as gender, age, marital status, education level, employment status, socioeconomic status, number of adults and children in the household, and annual household income.

The Variables

The estimated models sought to explain the decision to attend sporting events and related consumption. Our specifications, therefore, featured two dependent variables. Participation was a dummy variable equal to 1 if the individual had attended any sporting events during 2006. Consumption was the total annual individual consumer spending on sporting event attendance. This variable was measured in hundreds of euros, and it was defined as the sum of two different expenditures: fees and tickets for sporting events (professional and amateur), and travel expenses to the venue (including parking fees and accommodation).

The variables explaining sporting event attendance and/or sports consumption were selected on the basis of theoretical models and previous empirical evidence. Age and gender are the two main demographic determinants commonly considered (Breuer, Hallmann, Wicker, & Feiler, 2010). Some cross-sectional studies have concluded that consumer expenditure declines significantly with age (Lamb et al., 1992; Lera-Lopez & Rapun-Garate, 2005, 2007; Thrane, 2001b), whereas others found the relationship to be non-significant (Pawlowski & Breuer, 2011; Wicker et al., 2010). Furthermore, age is positively associated with sporting event attendance, particularly if only professional sports are considered (Zhang, Lam, Bennet, & Connaughton, 2003). Generally speaking, studies have found that men are more frequent attendees (Montgomery & Robinson, 2010; Thrane, 2001a) and spend more money on sports than women (Lamb et al., 1992; Lera-Lopez & Rapun-Garate, 2005, 2007; Thrane, 2001b; Wicker et al., 2010). In our empirical analysis, age was a continuous variable measured in years, and gender was a binary variable equal to 1 for males.

A similar consensus exists in relation to the positive influence of educational level on sporting event attendance (Wilson, 2002; Zhang et al., 2003), although, Pawlowski and Breuer (2011) reported a negative relationship with spending at sporting events. We considered two dummy variables for secondary education and higher education.

Traditionally, studies have shown that size of household is negatively associated with sporting event attendance (Zhang et al., 2003) and sports consumption (Lera-Lopez & Rapun-Garate, 2005). Married couples and large families can be expected to have less time for sports activities, although these characteristics could be positively related with consumer spending on sporting event attendance (Pawlowski & Breuer, 2011). As household characteristics, we included two additional variables: marital status (a dummy variable equal to 1 if the individual is married or cohabiting) and number of minors in the household.

Household income has been reported as a key determinant of expenditures on sports (Lera-Lopez & Rapun-Garate, 2005, 2007; Pawlowski & Breuer, 2010; Taks et al., 1999; Wicker et al., 2010). Our database provided information on annual household income, but it was recorded in group format and the non-response rate was very high (almost half of the sample refused to provide income data). We therefore presented two different sets of estimates. In one, the income variable was replaced with employment status (a dummy variable equal to 1 if the individual is working) as a proxy for economic status. In the other, we estimated the logarithm of income, using interval regression, to compute predicted family income, which was included as an explanatory variable in the expenditure equation.

Finally, degree of urbanization has been used as a proxy for sports facility provision, which is expected to have a positive relationship with attendance at sporting events (Coates & Humphreys, 2007). In our study, degree of urbanization is defined as the number of people living in the locality.

Descriptive Analysis

The average expenditure of Spanish adults on sports-related activities was estimated at 595 [euro] per year--75% of which was related with active sports practice and 25% with passive participation in sport. Attendance at sporting events accounted for over one-third of passive sport spending.

We obtained sports attendance data for 627 individuals. The summary statistics of their individual and socio-economic characteristics are displayed in Table 1. Forty-four percent of the respondents were male, the average age was 43 years old, and most respondents were married or cohabiting (70%). In regard to education, 39% of the respondents completed high school and 26% were university graduates; the sample was fairly evenly split between workers and non-workers (57% and 43%, respectively). Finally, in relation to sports and physical activity, 55% of the sample had attended sporting events in 2006 and 86% had practiced some form of sport or physical activity at some time in 2006. Of the 627 people whose descriptive statistics are shown in Table 1, 42 have attended sporting events but reported no spending in relation to this.

Table 2 shows the distribution of the 585 persons who provided attendance and spending data. As shown in Table 2, 52% of the 585 individuals had attended sporting events during 2006, while 283 respondents had not attended any. Among the attendees, 161 reported zero expenditure. Of those reporting positive expenditure (141 respondents), the average amount spent on attending sporting events in 2006--including tickets or entry fees, travel, and accommodation---was 211 [euro] and ranged from 30 [euro] to 1,800[euro].

Econometric Specification and Empirical Results

As the database does not allow for the estimation of a structural model of the demand for sporting events, we estimated a reduced-form model of sports attendance consumption. Following Downward et al. (2009), we used a two-stage specification of the individual decision. The first decision was whether to attend sporting events. A negative decision meant zero expenditure, and a positive one could have meant either zero or positive expenditure. The second choice was how much to spend on this activity. If participation and consumption are independent decisions, we can obtain separate estimates using Probit and Tobit models respectively.

When analyzing attendance, the dependent variable (d) is equal to 1 if the individual attended some professional or amateur sporting event in 2006, and it is 0 otherwise. We estimated the probability of attendance using a Probit model, which assumes that there is an unobservable variable (d*) that measures the difference between the utility derived from attending sporting events and the utility received from not attending them. This utility index depends linearly on a set of explanatory variables ([X.sup.d]) and has a random component ([[epsilon].sup.d]) that is normally distributed with 0 mean and variance 1:

[d.sup.*.sub.i] = [alpha][X.sup.d.sub.i] + [[epsilon].sup.d.sub.i] (4.1)

In the above equation, i indexes the individual observations.

The observed variable, d, is equal to 1 when the utility index is positive, and it is 0 otherwise:

[d.sub.i] = 1 if [d.sup.*.sub.i] [greater than or equal to] 0

[d.sub.i] = 0 if [d.sup.*.sub.i] < 0 (4.2)

The vector of independent variables contains individual and household characteristics that might determine individual preferences towards sporting events: gender, age, educational level, marital status, and the number of minors in the household. Finally, we included the degree of urbanization as a proxy of the supply of sporting events.

Table 3 displays the Probit estimates for attendance at sporting events. According to our results, males and the higher-educated were more likely to participate in these activities, whereas age was negatively correlated and household type had no impact. Finally, contrary to expectations, the higher the degree of urbanization, the lower the probability was of attending sporting events. (4)

The second stage of the model estimates individual spending on attending sporting events for those reporting attendance in 2006. At this stage, we estimated Tobit models (Wooldridge, 2002) because the dependent variable is censored at 0, and there was a high percentage of individuals who attended sporting events without spending any money. The general formulation starts from an unobservable dependent variable, [g.sup.*]. It depends linearly on a set of explanatory variables ([X.sup.g]) and has a normally distributed random disturbance term ([[epsilon].sup.g] ), with 0 mean and standard deviation [sigma]. The observed variable, g, is zero when g* [less than or equal to] 0 and is equal to the latent index when this is positive:

[g.sup.*.sub.i] = [beta][X.sup.g] + [[epsilon].sup.g] (4.3)

[g.sub.i] = [g.sup.*.sub.i] if [g.sup.*.sub.i] [greater than or equal to] 0 [g.sub.i] = 0 if [g.sup.*.sub.i] < 0 (4.4)

As previously mentioned, we ran two separate estimates that differ in the included explanatory variables. In both cases, the vector of independent variables contains demographic variables; these were included because they may determine either individual preferences or purchasing opportunities. Specifically, we included gender, educational level, and household variables (i.e., marital status and number of minors). In the first case, we did not include family income because of the high percentage of missing values. Instead, we considered a dummy for employment status. However, since economic theory suggests that income is a relevant explanatory factor of individual spending on goods and services, we incorporated predicted income--obtained from the interval estimation of the logarithm of income using the subsample of people who provided data--in our second estimates. (5)

The Tobit estimates are shown in Table 4. Having dropped observations with zero attendance at sporting events in 2006, we had a smaller sample size than for the Probit. (6) The scarcity of significant coefficients might be partly due to the small sample size. The only statistically significant variables in the first estimate were the education dummies: the highest levels of consumption on sporting events being reported by people with secondary education, followed by university graduates. However, when family income was included, it affected consumption positively and educational level lost its significance. Thus, it seems that the influence of education on sport attendance expenditure was indirect. That is, higher educated people lived in households with higher family income, and the higher the income, the more was spent on sporting events. Age also was positively related to sports consumption.

Previous estimates assume that attendance and consumption associated with sporting events are independent decisions. If there are unobserved factors simultaneously influencing attendance and consumption decisions, then we should estimate a Tobit model with selectivity--this can be seen as a special case of double-hurdle model (Flood & Grasjo, 2001).

Double-hurdle models were originally developed by Cragg (1971) and have been applied mainly to the estimation of demand functions for goods with high proportions of zero consumption. (7) The common feature of double-hurdle models is the assumption that individuals make two decisions regarding the acquisition of goods: whether to buy and how much to buy. These decisions may be determined by various explanatory factors, which may be correlated.

The double-hurdle model estimated in this study has also been applied by Flood and Grasjo (2001). The model specification consists of two equations: one used to model attendance at sporting events, and the other to model consumption:

[d.sup.*.sub.i] = [alpha][X.sup.d.sub.i] + [[epsilon].sup.d.sub.i] Attendance equation (4.5)

[g.sup.*.sub.i] = [beta][X.sup.g.sub.l] + [[epsilon].sup.g.sub.l] Consumption equation (4.6)

[[epsilon].sup.d.sub.i], [[epsilon].sup.g.sub.l] > BVN(0,0,1, [sigma]2, [rho])

In the equations (4.5) and (4.6), d* and g* are latent variables. In the case of attendance, what we observe is a binary variable, d, indicating whether or not the individual has attended sporting events. In the case of consumer spending, the observed variable, g, only takes positive values when the individual participates (d = 1) and actually spends money, as shown in the following equations:

[d.sub.i] = 1 if [d.sup.*.sub.i] [greater than or equal to] 0

[d.sub.i] = 0 if [d.sup.*.sub.i] < 0 (4.7)

[g.sub.i] = [g.sup.*.sub.i] if [d.sup.*.sub.i] > 0 & [g.sup.*.sub.i] > 0

[g.sub.i] = 0 otherwise (4.8)

In this specification, a person may attend sporting events (d = 1) but spend nothing (g = 0). Unlike the usual double-hurdle model, which does not separate non-spending participants from non-participants, we could identify three groups of people: non-attendees, non-spending attendees, and spending attendees. (8) The contribution to the likelihood function of each of these groups is:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (4.9)

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (4.10)

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (4.11)

Where [phi] is the density function of the standard normal distribution, [phi] is the cumulative distribution function, and [[phi].sub.2] is the distribution function of a bivariate normal. Consequently, the likelihood function to be maximized, expressed in logarithms, takes the following form:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (4.12)

For identification reasons, we included the degree of urbanization only in the participation equation and added a dummy for employment status or the predicted household income in the consumption equation.

The estimated coefficients are shown in Table 5. The first thing to note is the high degree of correlation between attendance and spending, confirming that this specification is more appropriate than the previous one, which assumed two independent decisions.

The results for the explanatory variables in the decision to attend sporting events were similar to those of the Probit model presented in Table 3. Thus, gender, age, and educational level were significant determinants of attendance at sporting events in Spain. The main difference is that the degree of urbanization played no part in explaining the probability of attendance.

With regard to consumer spending when attending sporting events, in addition to the positive influence of the educational level, gender was significant: males spent more than females. Moreover, when family income was included, it was found to have a positive influence on individual expenditure on sporting events, while educational level was no longer significant--leading to conclude, again, that education played a largely indirect role in passive sports consumption.

In general, the results were consistent with previous findings for sports consumption, particularly in view of the lack of evidence about attendance and consumer spending at sporting events. Generally speaking, when it comes to sport involvement, there was a sharp distinction between men and women, with men being more likely to attend sporting events and to spend more money when attending. The existing research also reflected a consensus regarding the positive influence of educational level on the decision to attend and spend money at sporting events; however, the relationship with secondary education had an inverted U-shape, which revealed it to be the strongest influence in both decisions, followed by higher education. Consequently, attendance at sporting events seemed to be more closely associated with middle-class and middle-education groups than with the lower social classes.

Age appeared to be another key determinant of the decision to attend sporting events, with young people being more likely to attend both amateur and professional sporting events. Household characteristics seemed to be non-significant in explaining attendance and consumption. This unexpected result could be the consequence of simultaneous consideration of amateur and professional sporting events. In future research, it might be interesting to perform a deeper analysis of the influence of household characteristics on sport attendance, considering amateur and professional sporting events separately. Likewise, neither occupation nor degree of urbanization seemed to be significant.

In addition, it is important to note the significance of household income on spending on sports attendance, as economic theory suggests. This result was common in the empirical literature (Cannon & Ford, 2002; Lera-Lopez & Rapun-Garate, 2005, 2007; Loyland & Ringstad, 2009; Pawlowski & Breuer, 2010).

Finally, we have estimated a multinomial logit model to test the robustness of our previous results. We considered three categories: zero participation, positive participation with zero spending, and positive participation with positive spending. The model is shown in Table A2 in the Appendix, and the conclusions were again similar to the double-hurdle model.

Summary and Conclusions

In this paper, we studied the determinants of consumer expenditure on attendance at sports events by estimating a reduced-form model with two different specifications. First, we assumed that the decision to attend a sporting event and the attendance-related consumption decision were independent and could, therefore, be estimated separately. Second, the assumption that attendance and consumption were correlated turned our model into a double-hurdle model that jointly estimated the attendance decision and the spending decision. This model enabled us to distinguish three types of behavior: non-attendance, attendance with zero expenditure, and attendance with positive expenditure. To our knowledge, this is the first empirical study to apply this methodology to sports consumption.

The reported survey was conducted in 2007 on a sample of Spanish adults. The results showed that the probability of attending sporting events was higher among young men with a high school or college education. Household composition did not seem to affect this decision. As for the degree of expenditure on sporting events, consumption tended to be higher among men and people with secondary and higher education. Household income was also a positive determinant of consumption at sporting events, which, when included, caused higher education to lose its significance. This suggests that the latter plays an indirect role in sports expenditure. The results strongly corroborate the importance of economic variables suggested by the neoclassical theory of consumption. They also indicate the relevance of social values and cultural influences (gender, education level), which might be explained by their influence on individual preferences in the context of the neoclassical theory. However, it might also support the post-Keynesian and sociological approaches. On the other hand, although sports attendance is a time-consuming good, time constraints--measured through variables such as occupation, marital status, and number of children--did not seem to affect expenditure on sports attendance. Finally, we found a strong positive correlation between attendance and consumption, which enabled us to conclude that there are unobservable variables with a simultaneous effect on both decisions. This finding supports the use of double-hurdle models to analyze sports attendance and consumption decisions jointly.

In the light of this study, future research should be aimed at comparing sports attendance consumption across different types of sports and different countries. A cross-country analysis would be a first step towards evaluating country- and culture-specific effects in this context. Future research might also explore the relationship between sports attendance and attendance-related consumption and other leisure and sporting activities, such as active participation in sports, sports pay-per-view purchase rates, and online sports betting.

Appendix A

Table A1. Consumer Expenditure on Attendance at Sports Events
(Tobit Models for the Entire Sample)

 Specification 1 Specification 2

Independent Coefficients f-Student Coefficients f-Student
variables

Constant 5.5414 5.26 8.4847 4.80
Male 1.3982 2.88 1.3948 2.96
Age 0.0090 0.50 0.0082 0.41
Secondary 2.7286 4.44 1.8785 2.62
 education
Higher 2.3642 3.55 0.5105 0.48
 education
Married 0.0813 0.15 0.6844 1.11
Number of 0.1860 0.73 0.1584 0.63
 children
 0-17
Worker 0.5707 1.14 --- ---
Predicted --- --- 0.0002 2.32
 family
 income

Log L 564.0087 561.8921

Pseudo-R2 0.0387 0.0423

Sample size 585

Appendix B

Table A2. Multinomial Logit Model

Independent variables Specification 1

 Coefficients f-Student

Positive participation and zero consumption

 Constant 0.0266 0.06
 Male 0.7767 3.49
 Age 0.0281 3.54
 Secondary education 0.5762 2.28
 Higher education 0.6408 2.26
 Married 0.0531 0.21
 Number of children 0-17 0.0530 0.44
 Degree of urbanization 0.0002 1.49
 (inhabitants/1000)
 Worker 0.0097 0.04
 Predicted family income --- ---

Positive participation and positive consumption

 Constant 1.3033 2.74
 Male 0.9125 3.88
 Age 0.0166 1.94
 Secondary education 1.3213 4.56
 Higher education 1.3802 4.36
 Married 0.0172 0.06
 Number of children 0-17 0.1379 1.13
 Degree of urbanization 0.0002 1.59
 (inhabitants/1000)
 Worker 0.0450 0.19
 Predicted family income --- ---

Log L 569.0097

Pseudo-R2 0.0730

Sample size 585

Base outcome: Non-participation

Independent variables Specification 2

 Coefficients f-Student

Positive participation and zero consumption

 Constant 0.1543 0.21
 Male 0.7902 3.68
 Age 0.0288 3.34
 Secondary education 0.6341 2.09
 Higher education 0.7661 1.73
 Married 0.0938 0.34
 Number of children 0-17 0.0574 0.48
 Degree of urbanization 0.0002 1.51
 (inhabitants/1000)
 Worker --- ---
 Predicted family income 0.000 0.35

Positive participation and positive consumption

 Constant 2.5308 3.10
 Male 0.8717 3.82
 Age 0.0093 0.97
 Secondary education 0.9795 2.87
 Higher education 0.6521 1.31
 Married 0.2521 0.86
 Number of children 0-17 0.1182 0.96
 Degree of urbanization 0.0002 1.57
 (inhabitants/1000)
 Worker --- ---
 Predicted family income 0.0001 1.89

Log L 566.4559

Pseudo-R2 0.0772

Sample size 585

Base outcome: Non-participation

References

Alonso, J., Ruesga, S., Saez; F., & Vicens, J. (1991). Impacto economico del deporte en Espana. [The economic impact of sports in Spain]. Revista de Investigaciones y Documentacion sobre las Ciencias de la Educacion Fisica y el Deporte, 18, 22-35.

Amnyos Consultants (2008). Study of public and private financing in sport in Europe. Paris: Minister for Health, Youth, Sport, and the Voluntary Sector.

Amnyos Consultants. (2010, February). The different funding models for grassroot sports in the EU. Paper presented at the Brussels Conference What Future Sustainable Funding Model(s) for Grassroots Sports in the Internal Market? Brussels, Belgium.

Andreff, W. (2006). Sport and financing. In W. Andreff & S. Szymanski (Eds.), Handbook of the economics of sport (pp. 271-281). Cheltenham, United Kingdom: Edward Elgar.

Beaton, A. A., Funk, D. C., & Alexandris, K. (2009). Operationalizing a theory of participation in physically active leisure. Journal of Leisure Research, 41, 177-203.

Becker, G. S. (1965). A theory of the allocation of time. The Economic Journal, 75, 493-517.

Bourdieu, P. (1984). Distinction: A social critique of the judgment of taste. Cambridge, MA: Harvard University Press.

Breuer, C., Hallmann, K., Wicker, P., & Feiler, S. (2010). Socio-economic patterns of sport demand and ageing. European Review of Aging and Physical Activity, 7, 61-70.

Cannon, T., & Ford, J. (2002). Relationship of demographic and trip characteristics to visitor spending: An analysis of sports travel visitors across time. Tourism Economics, 8, 263-271.

Coates, D., & Humphreys, B. (2007). Ticket prices, concessions and attendance at professional sporting events. International Journal of Sport Finance, 2, 161-170.

Cragg, J. G. (1971). Some statistical models for limited dependent variables with application to the demand for durable goods. Econometrica, 39, 829-844.

Davies, L. E. (2002). Consumer's expenditure on sport in the United Kingdom: Increased spending or underestimation? Managing Leisure, 7, 83-102.

Dawson, P., & Downward, P. (2009). Participation, spectatorship and media coverage in sport Working paper, University of Bath, United Kingdom.

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., Dawson, A., & Dejonghe, T. (2009). Sports economics: Theory, evidence and policy. Oxford, United Kingdom: Elsevier.

Flood, L., & Grasjo, U. (2001). A Monte Carlo simulation study of Tobit models. Applied Economics Letters, 8, 581-584.

Henderson, K. A. (2009). A paradox of sport management and physical activity interventions. Sport Management Review, 12, 57-65.

Humphreys, B. R., & Ruseski, J. E. (2009). Estimates of the dimensions of the sports market in the U.S. International Journal of Sport Finance, 4, 94-113.

Humphreys, B. R., & Ruseski, J. E. (2010). The economic choice of participation and time spent in physical activity and sport in Canada. Working Paper, University of Alberta, Canada.

Lamb, L. L., Asturias, L. P., Roberts, K., & Brodie, D. A. (1992). Sports participation: How much does it cost? Leisure Studies, 11, 19-29.

Lavoie, M. (2004). Post Keynesian consumer theory: Potential synergies with consumer research and economic psychology. Journal of Economic Psychology, 25, 639-649.

Lera-Lopez, F., & Rapun-Garate, M. (2005). The determinants of consumer expenditure on sport: A Tobit model. In G. T. Papanikos (Ed.), International research on sports economics and production (pp. 61-78). Athens, Greece: Athens Institute for Education and Research.

Lera-Lopez, F., & Rapun-Garate, M. (2007). The demand for sport: Sport consumption and participation models. Journal of Sport Management, 21, 103-122.

Lera-Lopez, F., Rapun-Garate, M., & Aguirre, J. (2008). Analisis y evaluation economica de la participation deportiva en Espana. [Economic analysis and evaluation of sports participation in Spain.] Madrid, Spain: Consejo Superior de Deportes.

Loyland, K., & Ringstad, V. (2009). On the price and income sensitivity of the demand for sports: Has Linder's disease become more serious? Journal of Sports Economics, 10, 601-616.

Mannell, R. C., & Kleiber, D. A. (1997) A social psychology of leisure. State College, PA: Venture Publishing.

Michon, B., Ohl, F., & Faber, C. (1987). Les prix de la pratique sportive pour le consommateur. [Prices of sports for the consumer]. Strasbourg, France: Universite des Sciences Humaines.

Montgomery, S., & Robinson, M. (2010). Empirical evidence of the effects of marriage on male and female attendance at sports and arts. Social Science Quarterly, 9(1), 99-116.

Pawlowski, T., & Breuer, C. (2010). Expenditure elasticities of the demand for leisure services. Lisbon, Portugal: New Economics Papers Sports and Economics.

Pawlowski, T., & Breuer, C. (2011). The demand for sports and recreational services: Empirical evidence from Germany. European Sport Management Quarterly, 11(1), 5-34.

Spanish High Council for Sport. (2011). Encuesta sobre los hdbitos deportivos en Espana 2010: Avance de resultados. [Survey about sports habits in Spain 2010: Breakthrough results.] Madrid, Spain: Author. Retrieved May 15, 2011, from http://www.csd.gob.es/

Sport Industry Research Centre. (2008). Sport market: Forecasts 2008-2010. Sheffield, United Kingdom: Author.

Sport Industry Research Centre. (2010). 2004 Sport Satellite Account for the UK. Sheffield, United Kingdom: Author. Retrieved October 2, 2010, from http://www.culture.gov.uk

Taks, M., & Kesenne, S. (2000). The economic significance of sport in Flanders. Journal of Sport Management, 14, 342-365.

Taks, M., Renson, R., & Vanreusel, B. (1995). Social stratification in sport: A matter of money or taste? European Journal for Sport Management, 2, 4-14.

Taks, M., Renson, R., & Vanreusel, B. (1999). Consumer expenses in sport: A marketing tool for sports. European Journal for Sport Management, 6, 4-18.

Taks, M., Vanreusel, B., & Renson, R. (1994). What does sport really cost? A micro-economic study of the consumer cost of golf and soccer. European Journal for Sport Management, 1, 2234.

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

Thrane, C. (2001b). The differentiation of personal sport expenditures: The Norwegian case. International Journal of Sport Management, 2, 237-251.

U.S. Census Bureau (2010). The 2010 Statistical Abstract, Arts, Recreation, and Travel. Washington, DC: Author. Retrieved September 28, 2010, from http://www.census.gov/com pendia/statab/cats/arts_recreation_travel.html

Weagley, R., & Hugh, E. (2004). Leisure expenditures of retired and near-retired households. Journal of Leisure Research, 36, 101-127.

Wicker, P., Breuer, C., & Pawlowski, T. (2010). Are sports club members big spenders? Findings from sport specific analyses in Germany. Sport Management Review, 13, 214-224.

Wilson, T. C. (2002) The paradox of social class and sports involvement: The roles of cultural and economic capital. International Review for the Sociology of Sport, 37, 5-16.

Wooldridge, J. M. (2002). Econometric analysis of cross section and panel data. Cambridge, MA: The MIT Press.

Zhang, J. J., Lam, E. T. C., Bennett, G., & Connaughton, D. P. (2003) Confirmatory factor analysis of the spectator decision-making inventory. Measurement in Physical Education and Exercise Science, 7, 57-70.

Authors' Note

The authors wish to acknowledge the financial support from the Spanish High Council for Sport (32/UPB20/06).

Fernando Lera-Lopez [1], Manuel Rapun-Garate [1], and Maria Jose Suarez [2]

[1] Public University of Navarra, Campus Arrosadia, Spain

[2] University of Oviedo, Campus del Cristo, Spain

Fernando Lera-Lopez is senior lecturer of economics at the Department of Economics of the Public University of Navarra (Spain). His research interests include sports economics and sport participation.

Manuel Rapun-Garate is professor of economics at the Department of Economics of the Public University of Navarra (Spain). His research interests include sports economics and regional studies.

Maria Jose Suarez is senior lecturer of economics at the Department of Economics of the University of Oviedo (Spain). Her research interests include labor and sports economics.

Endnotes

(1) Education is likely to have a significant influence on the sports decision. More highly-educated people appear to give more importance to sport than less-educated people. On the other hand, education can be considered as a proxy for income since a high educational level is closely associated with higher levels of income.

(2) As a referee suggested, this assumption may not be suitable in some cases; however, we believe that, in general, it can be appropriate when attempting to explain individual decision making.

(3) See Breuer et al. (2010) and Wicker et al. (2010) for a review of the literature.

(4) Loyland and Ringstad (2009), using a Norwegian database, also found a negative effect of population size on the budget share allocated to active and passive sport activities. They concluded that the demand for sport is higher in rural areas.

(5) The sample size of the income equation is 339 observations, and the independent variables are age and its square, number of adults in the household, and dummies for educational level and marital and employment status. In addition, we include three dummies for the employment status of the head of household: self-employed, employed in occupations requiring a university degree, and employed in other occupations (the omitted category is unemployed). The estimation method is interval regression and, from the estimated coefficients, we computed the predicted income. As income was recorded in group format, we decided to use predicted income for all members of the sample. The estimated coefficients had the expected signs and the majority were significant. Results are available upon request.

(6) We had also run Tobit estimates for the entire sample, as is usual in the empirical literature on sports expenditures. The results are shown in the Appendix.

(7) Double-hurdle models have been used, among other applications, to estimate the demand for leisure activities (Weagley & Hugh, 2004) and participation/non-participation and the amount of time actively spent on sports or physical activity (Humphreys & Ruseski, 2010).

(8) We tried to estimate the standard double-hurdle model in order to compare the results, but we were unable to obtain convergence. However, Flood and Grasjo (2001) compared both versions of the double-hurdle model using Monte Carlo simulations and concluded that it is better to use Tobit selection (d = 1 and g = 0) if known.

Table 1. Summary Statistics

Variable Mean Standard Minimum Maximum
 deviation

Attending 0.549 0.498 0.000 1.000
 sporting
 events
Age 43.273 14.921 18.000 75.000
Male 0.439 0.497 0.000 1.000
Secondary 0.386 0.487 0.000 1.000
 education
Higher 0.262 0.440 0.000 1.000
 education
Worker 0.569 0.496 0.000 1.000
Married 0.699 0.459 0.000 1.000
Number of 0.668 0.960 0.000 9.000
 children
 0-17
Sports 0.863 0.344 0.000 1.000
 practice
Degree of 344.619 758.022 0.135 2881.506
 urbanization
 (inhabitants/
 1000)

Table 2. Distribution of the Sample Based on Their Attendance and
Consumption at Sporting Events

 Consumption = 0 Consumption > 0

Attendees of 161 respondents 141 respondents
 sporting events Mean consumption = 211.38
 Minimum consumption = 30
 Maximum consumption = 1,800
Non-attendees of sporting events283 0 respondents
respondents

Table 3. Probit Estimates for the Probability
of Attending Sporting Events

Independent Coefficients f-Student
variables

Constant 0.0789 0.36
Male 0.5159 4.61
Age 0.0140 3.45
Secondary 0.5523 4.15
 education
Higher 0.5851 3.97
 education
Married 0.0102 0.08
Number of 0.0535 0.90
 children
 0-17
Degree of 0.0001 1.90
 urbanization
 (inhabitants/
 1000)
Sample size 585
Log L 364.9304
Pseudo-[R.sup.2] 0.0993

Table 4. Consumer Expenditure on Attendance at Sports Events
(Tobit Models)

 Specification 1 Specification 2

Independent Coefficients f-Student Coefficients f-Student
variables

Constant 3.3935 3.36 5.9192 3.48
Male 0.3681 0.75 0.3521 0.73
Age 0.0236 1.26 0.0380 1.87
Secondary 1.8855 2.95 1.1539 1.56
 education
Higher 1.3587 2.00 0.2972 0.27
 education
Married 0.2483 0.43 0.8239 1.25
Number of 0.0145 0.05 0.0031 0.01
 children
 0-17
Worker 0.6740 1.30 --- ---
Predicted --- --- 0.0002 2.04
 family
 income

Log L 475.8082 474.5427
Pseudo-R2 0.0142 0.0169

Sample size 302

Table 5. Double-Hurdle Model for Attendance at
Sporting Events and Related Consumption

Independent variables Specification 1

 Coefficients f-Student

Pr(attendance at
sporting events)

 Constant 0.0139 0.06
 Male 0.5009 4.29
 Age 0.0129 3.25
 Secondary 0.5580 4.18
 education
 Higher 0.5483 3.70
 education
 Married 0.0008 0.01
 Number of 0.0433 0.69
 children 0-17
 Degree of 0.0001 1.32
 urbanization
 (inhabitants/
 1000)

Consumer expenditure
on attendance at
sports events/100

 Constant 5.4038 5.34
 Male 1.4495 2.76
 Age 0.0120 0.64
 Secondary 2.7522 4.60
 education
 Higher 2.3828 3.42
 education
 Married 0.0408 0.07
 Number of 0.1523 0.52
 children 0-17
 Worker 0.4605 0.96
 Predicted --- ---
 family income

Rho 0.9635 2.14
Log L 833.4725
Sample size 585

Independent variables Specification 2

 Coefficients f-Student

Pr(attendance at
sporting events)

 Constant 0.0255 0.11
 Male 0.4992 4.27
 Age 0.0130 3.30
 Secondary 0.5605 4.21
 education
 Higher 0.5370 3.61
 education
 Married 0.0027 0.02
 Number of 0.0466 0.75
 children 0-17
 Degree of 0.0001 1.45
 urbanization
 (inhabitants/
 1000)

Consumer expenditure
on attendance at
sports events/100

 Constant 7.7536 4.58
 Male 1.4025 2.81
 Age 0.0021 0.10
 Secondary 0.0021 2.93
 education
 Higher 2.0814 0.80
 education
 Married 0.9106 0.82
 Number of 0.5572 0.40
 children 0-17
 Worker 0.1141 ---
 Predicted 0.002 1.85
 family income

Rho 0.9699 2.09
Log L 831.5291
Sample size