Comparative analysis of sports practice by types of activities.

Garcia, Muniz ; Muniz, Cristina ; Rodriguez, Placido 等

Abstract

This paper makes a comparative analysis of the determinants of adult sports practice in different types of activities. Specifically, we analyze frequency of participation in walking, individual versus group sports, indoor versus outdoor sports, and sports that require facilities versus sports that do not require them. In the empirical analysis we use the Spanish Time Use Survey 2002-03 and we estimate zero-inflated negative binomial (ZINB) count data models to explain the frequency of sports participation in the previous four weeks. The covariates included are demographic and socioeconomic factors. We compute individual marginal effects of individual and family characteristics on the expected frequency of participation, on the probability of being a potential participant, and on the expected frequency conditioned on participation. The results show that gender and labor status are significant correlates of participation and frequency in all types of activities. We also find some interesting differences among types of sports.

Keywords: sports practice, ZINB count data models, sports classifications

Introduction

The Special Eurobarometer 412 (European Commission, 2014) shows that 41% of Europeans play sports at least once a week, and this figure has not significantly changed since 2009. In general, sports participation declines when going from north to southern Europe. Specifically, 70% of Swedish people play sports whereas the lowest figure is that of Bulgaria (11%). In the case of Spain, the percentage of people who practice sports weekly is above the EU28 average (46%). However, the percentage of Spaniards who never play sports is also slightly higher than the EU28 average: 44% and 42%, respectively.

Regarding the engagement in other physical activities, such as cycling, gardening, or dancing, we find the same pattern: Nordic countries are among the most likely to practice them and the lowest figures correspond to Southern Europe. On average, 48% of Europeans engage in physical activity at least once a week. The Spanish average is 10 percentage points below this figure. The difference is much higher when we look at the percentage of people who never engage in physical activity: the EU28 average is 30% and the Spanish one is 49%. However, Spain ranks second on the percentage of people who walk 10 minutes or more at least four days per week.

Concerning the type of sports, according to the last wave of the Survey on Sports Habits in Spain (Encuesta sobre Habitos Deportivos en Espana), carried out in 2015, fitness training is the most practiced sport on a weekly basis, followed by running and cycling. (1)

In past decades, there has been a growing literature on the determinants of sports practice. Usually, studies on this issue define sports in an aggregate way (adding up all activities performed by individuals) and/or they analyze specific sports. Previous literature on this topic has found relevant differences among individuals in the type and intensity of sports activities.

The goal of this study is to make a separate analysis of different sports to determine whether there are differences in individuals' behavior in relation to specific activities. The reason for disaggregating sports is that not all sports are similarly intensive in time nor require the same amount of expenditures for its practice (Taks et al., 1994). There are also differences with regard to the need of facilities or coordination with other people. Consequently, the aggregate analysis of physical activity may hide relevant differences in the correlates of different types of activities.

Our paper is in line with that of Humphreys and Ruseski (2007) in that we analyze the determinants of sports participation by classifying sports into different categories. However, Humphreys and Ruseski (2007) define an exclusive classification, whereas we make three different classifications according to different characteristics of sports, and each particular sport is included in all three classifications. In particular, we study sports participation and frequency in the following types of sports: individual versus group sports, indoor versus outdoor sports, activities that require the use of facilities versus those that do not require them, and walking. We study walking as a separate category because it cannot be classified as a sport but it is a physical activity very common in Spain.

In the empirical specification, we used a sample of adults from the Spanish Time Use Survey 2002-03. (2) All the dependent variables are defined as the number of times the individual has practiced the activity in the four weeks prior to the survey. Regarding the econometric methodology, we estimated zero-inflated negative binomial (ZINB) count data models. The covariates included to explain individual decisions are age, family composition, educational level, labor status, wage, non-labor income, seasonality, population size, health, and gender.

With the estimated coefficients, we computed for each individual in the sample the marginal effect of the socio-economic covariates on the expected frequency, the probability of being a potential participant, and the expected frequency conditioned to participate. To the best of our knowledge, this is the first research that computes a decomposition of marginal effects in count data models by types of sports.

Literature Overview

There is a broad literature examining the determinants of individual decisions about engagement in sports or other forms of physical activity for recreational purposes. Downward et al. (2009) and Cabane and Lechner (2015) survey the international evidence on the topic. As these authors point out, there are two main theories to explain sports participation: the neoclassical theory and heterodox approaches. Regarding the empirical evidence, there are some common results in the literature. Males are more likely to play sports than females. Other variables that tend to be significant are age, education, income, health status, and family-related factors. However, the effect of those variables on the frequency of participation is sometimes the opposite to their effect on the probability of playing sports.

The probability of practicing sports or physical activity is usually specified as a probit or logit model (e.g., Farrell & Shields, 2002; Downward, 2004, 2007; Stratton et al., 2005; Breuer et al., 2011). Other studies analyze both participation and frequency, assuming that the dependent variable is continuous and applying Tobit models (Ruseski et al., 2011), Heckman methods (Downward & Riordan, 2007; Humphreys & Ruseski, 2007; Garcia et al., 2011; Muniz et al., 2011) or double hurdle models (Buraimo et al., 2010; Humphreys & Ruseski, 2011, 2015). In other publications, the frequency of participation is defined as an ordinal variable, so that ordered probit or zero inflated ordered probit models are applied (Lera-Lopez & Rapun-Garate, 2005; Downward et al., 2011), or as a count variable, in which case count data models are used (Dawson & Downward, 2011; Muniz et al., 2014).

Regarding the definition of the dependent variable, most studies analyze sports in the aggregate, adding up all types of sports (sometimes, other kind of physical activities are also included). This is the case of Farrell and Shields (2002), Downward (2007), Downward and Riordan (2007), Buraimo et al. (2010), Breuer et al. (2011), Dawson and Downward (2011), Garcia et al. (2011), or Muniz et al. (2011, 2014). Some of these, as well as other studies (e.g., Downward, 2004; Humphreys & Ruseski, 2015), also examine specific sports, usually the most popular ones.

Instead, there is scarce literature that groups sports or physical activities into different categories according to their characteristics. It is worth mentioning the papers by Scheerder et al. (2005) and Humphreys and Ruseski (2007). Scheerder et al. (2005) study sports involvement in Belgium by estimating logit models. In their empirical analysis, these authors define sports as an aggregate but they also analyze different groups of sports, classified according to their organizational context (club participation, non-organized sports) or other characteristics (lifetime sports, individual-com-petitive sports, outdoor sports, racquet sports, martial arts, and team sports). Instead, Humphreys and Ruseski (2007) define five categories of sports or physical activities: outdoor recreation (e.g., snow skiing, surfing, fishing), household activities (e.g., carpentry, gardening, mowing the lawn), group sports (e.g., aerobics class, badminton, boxing, soccer), individual sports (e.g., bicycling, golf, judo, running), and walking. In their classification, each activity is only included in one category. They estimate participation and time spent in the activity applying the Heckman method and using a sample of U.S. adults.

As we explain in the next section, our classification is somewhat similar to the one by Humphreys and Ruseski (2007) but we define three different classifications--besides walking--and, unlike Humphreys and Ruseski (2007), in our case each sport is included in all three classifications.

Descriptive Analysis

The survey used in our empirical analysis is the Spanish Time Use Survey (Encuesta de Empleo del Tiempo), conducted by the Spanish Statistical Office (Instituto Nacional de Estadistica - INE) in 2002-03. This is the first Spanish time use survey and, although it was carried out again in 2009-10, the questions on sports practice in previous weeks were dropped from the second wave, and this is the reason why we could not use it. When interpreting our results, we have to bear in mind that it has been more than 10 years since the survey was conducted. The Special Eurobarometers on sports practice published during these years (European Commission, 2004, 2010, 2014) show that the percentage of Spaniards who never practice sports has not substantially changed between 2004 and 2013 (47% in 2004 versus 44% in 2013) but the frequency of sports practice has increased (37% of Spanish people played sports at least once a week in 2004 whereas the figure increased to 46% in 2014). (3)

The aim of the survey is to gather information about individual allocation of time. Its sample size is about 60,000 individuals aged 10 or older from around 24,000 households. Data collection was evenly distributed from October 2002 to September 2003. Besides the activities diary, it also includes questions about personal and family characteristics, as well as participation in sports and other leisure activities. Our study focuses on the working-age population between 18 and 65 years old.

The dependent variables in our empirical analysis are defined from the following questions: "Have you done any of the following sports in the last four weeks? How many times in the last four weeks?" and the questionnaire provides a list of sports, including walking.

From this information, we make the following three classifications of sports, apart from walking, which is treated as a separate category:

Group versus individual sports: Group sports are those that require other participants and includes group ball games, tennis, golf and other ball or disc games, martial arts or boxing, and water sports. Individual sports include running, cycling, skiing, mountaineering, fitness, swimming, gymnastics, and skating. The practice of group sports involves more limitations than individual sports since they require the coordination with other people.

Sports requiring facilities versus those that do not require facilities: In the first group we include group ball games, martial arts or boxing, fitness, swimming, skating, gymnastics, skiing, tennis, and golf or other ball or disc games, whereas we consider that running, cycling, mountaineering and water sports do not require facilities. Sports that require facilities may be more time-intensive than others, since individuals need to move to the sports infrastructures. Moreover, these activities are usually more expensive if people have to pay a fee to use them.

Outdoor versus indoor sports: Outdoor activities encompass running, cycling, skiing, mountaineering, tennis, golf, or other ball or disc games, and water sports, whereas group ball games, martial arts or boxing, fitness, swimming, skating, and gymnastics are classified as indoor activities. Weather may affect the practice of activities that take place outdoors.

Note that a particular sport is included in all three classifications. For example, running is considered as an individual sport, a sport that does not require facilities, and an outdoor activity. Moreover, categories are not mutually exclusive within each classification, in the sense that there may be individuals who have practiced both types of sports in the previous weeks. For example, a person can be a participant in fitness (an individual sport) and basketball (a group sport) at the same time. It is also worth noting that the classification is sometimes arbitrary, since the survey information is not detailed enough in some sports. For instance, football or other group ball games can be played indoors or outdoors and the survey does not specify this information.

Table 1 provides information about the number and percentage of participants in the activity, and the mean frequency of participation, in all activities defined.

The number and percentage of participants are always higher for men than for women except in the case of walking. The data also reveal that there are greater gender differences in the participation rates than in the mean frequency of practice by participants. The highest differences in participation rates are found in the case of group sports, outdoor sports, and those that do not require facilities. Regarding frequency of sports practice by participants, female frequencies are slightly greater in indoor sports, sports that require facilities, and walking.

Model Specification

The analysis of sports practice can be modeled applying the neoclassical time allocation model. According to this approach, individuals consume goods and allocate time to labor and different leisure activities in order to maximize their utility, subject to budget and time constraints (Becker, 1965). From this point of view, the types of sports analyzed can be considered as leisure activities that increase individuals' welfare. Moreover, the division of sports is justified if the covariates have different effects on the time spent on each type of sport.

In this model, the demands of consumption and time allocated to each activity depend on wage and non-labor income--both of these variables determine the budget constraint--and other factors that affect individual preferences. The model does not allow us to anticipate the effect of changes in wages or non-labor income on the time spent on a specific leisure activity because there may be redistribution effects among different uses of time.

Our database provides information about the frequencies of participation in different sports, so that our dependent variables are count data and show a relatively high percentage of non-participants. Thus, in our empirical analysis we choose count data models and, after some preliminary estimates of different count data specifications, we conclude that the ZINB count data model is the most suitable for our data. (4)

The ZINB specification assumes two types of individuals: (1) those that would never participate in the activity and, consequently, whose observed frequency is zero (Always Zero group), and (2) those who may or may not participate, in which case their observed frequency may be zero or positive (Not Always Zero group). Moreover, the model assumes that these groups are unobserved; data only provide information about the number of counts, which may be zero or positive.

The probability that an individual belongs to the Always Zero group is assumed to follow a logit specification in which the dependent variable, [A.sub.i], is a binary variable that is equal to 1 if the individual i belongs to the Always Zero group and 0 otherwise. The probability of each count in the Not Always Zero group (i.e., for the individuals with positive levels of participation or zero participation due to corner solutions) is computed by a negative binomial regression. The model is estimated by maximum likelihood.

Since exclusion restrictions are not required for identification, we assume that the variables that determine the probability of belonging to the Always Zero group are the same as those that affect the probability of each count. In particular, the independent variables included are those that determine the budget constraint: logarithm of hourly earnings (this variable is equal to observed wage for workers and predicted wage for non-working people) (5) and non-labor income, and others that may reflect time constraints or affect preferences: gender, age and age squared, health, marital status, number of children aged 12 years or younger, a dummy equal to 1 if there are more than two adults living at home, and labor status. Finally, we also include dummies to control for the term and degree of urbanization. Tables A1 and A2 in the Appendix provide information about the definition of each variable and the descriptive statistics.

Given that the ZINB model is not linear, coefficients do not have a simple interpretation. Therefore, it is more useful to compute the marginal effects that show the response of the dependent variables to changes in the covariates. Specifically, we analyze the effect of the covariates on three variables: the expected number of counts, the probability of being a potential participant in each category of sports, and the expected number of counts conditioned to belonging to the Not Always Zero group.

To compute the marginal effects, we begin by defining the expected value of counts (y), which is equal to the product of the probability that the individual belongs to the Not Always Zero group (A = 0) and the expected value of counts in this group: (6)

E(y)=p(A=1)*0+(1)p(A=1)*E(y|A=0) (1)

Thus, the marginal effect of a covariate, [x.sub.k], on the expected number of counts when [x.sub.k] is a continuous covariate can be expressed as:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (2)

As shown in Equation (2), [x.sub.k] influences the expected value of y in two ways: on the one hand, it modifies the probability of belonging to the Not-Always Zero group and, on the other hand, it also affects the expected number of counts, conditioned to belonging to that group. (7) For each individual, we obtain the total effect of the covariate on the expected number of counts, as well as two partial effects: the change in the probability of being a potential participant (i.e., belonging to the Not Always Zero group) and the change of the expected frequency conditioned on participation.

In the case of categorical covariates, we compute the total marginal effect as the discrete difference between the expected value of the dependent variable when the covariate [d.sub.j] is equal to one and zero:

E(y|[d.sub.j] = 1)-E((y|[d.sub.j] = 0)(3)

The effects of [d.sub.j] on the probability of being a potential participant in a group (A = 0), and on the expected number of counts in the Not Always Zero group are respectively equal to:

p(A = 0|[d.sub.j] = 1) - p(A = 0|[d.sub.j] = 0) (4)

E(y|[d.sub.j] = 1, A = 0) - E(y|[d.sub.j] = 0, A = 0) (5)

Finally, in the case of monetary variables, wage, and non-labor income, we compute the elasticities from the marginal effects. (8) Wage elasticities are defined in the following way:

[eta.sub.w] = [delta]E(y)/[delta]w w/y (6)

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (7)

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (8)

In previous equations [[eta].sub.w,] [[eta].sup.P.sub.w]], and [[eta].sup.F]w stand for the total wage elasticity, the wage elasticity of being a potential participant, and the wage elasticity of the expected frequency conditioned on participation, respectively. Non-labor income elasticities are computed in a similar way.

Results

The ZINB models are separately estimated for each category of sports. Tables 2-5 show the estimated coefficients and t-statistics in each case.

Although the coefficients do not show the marginal effects, their signs indicate whether the relationship is direct or inverse. Moreover, in the participation equations, we must note that a positive sign implies an increase in the probability of belonging to the Always Zero group (i.e., in the probability of being a non-participant).

Beginning with participation decisions, gender is significant for all types of sports. Furthermore, this variable has very high levels of significance. The coefficients show that women are less likely to practice any kind of sport. Walking is the only exception to the previous statement.

Age is usually negatively related to participation decisions because, although it sometimes displays a U-shaped relationship, the minimum values are reached at levels above the upper age limit of our sample--65 years. The only activities in which age does not affect the probability of participation are sports that do not require facilities. Moreover, it is worth noting the case of walking: This is the only activity in which younger people are less likely to practice. This may be because walking does not require great physical effort, enabling older people to do it.

The educational level always has a positive association with the probability of participating in all sports activities. On the contrary, marital status usually has a negative influence on the probability of practicing sports, except for group sports and walking, for which the variable is not significant. The number of children at home is also negatively associated with participation in all sports except group sports, and the presence of more than two adults in the household has a negative effect too. Therefore, the general conclusion is that family responsibilities tend to reduce the likelihood of practicing sports. Regarding health, individuals who have a chronic illness are less likely to play sports.

Participation in sports activities is also related to season. In particular, the probability of participating in some sports decreases during winter and spring, whereas in summer the probability of practicing sports increases (whenever these variables are significant). Sports have an urban character, since participation generally decreases in smaller population areas. However, it is not significant in sports that do not require infrastructure. The influence of the variables related to the degree of urbanization may be explained because the best infrastructures are generally located in large cities.

Having a job generally decreases or does not affect the probability of practicing sports. It seems that the availability of time may be a barrier to practicing some sports. Finally, hourly wages and non-labor income are always significant. Both covariates are positively linked to the probability of participating.

Focusing now on the effect of the covariates on frequency decisions, gender is always significant. The signs of the coefficients indicate that males participate more frequently than females in individual, group, and outdoor sports, and sports that do not require facilities, whereas females participate more frequently in the rest of categories. Age, when significant, displays a U-shaped relationship with respect to frequency. (9) Regarding education, the frequency of walking and playing indoor sports decreases with the educational level.

Family variables are not generally significant in sports frequency. The season is also less important in frequency than in participation decisions. However, we have obtained some interesting results. The frequency of walking takes the lowest value in the last quarter of the year and the practice of indoor sports and sports that require infrastructure increases in summer. While the population size variables were generally significant for participation, they are not significant in frequency of practice.

The number of times individuals practice any sports or walk is lower when they are working. With regard to economic variables, wage is positively related to the frequency of practicing individual, group, and indoor sports, as well as sports that require facilities. In general, indoor sports and sports that require facilities may be relatively more expensive than other sports since individuals usually have to pay a fee to practice them. Non-labor income diminishes the frequency of walking, playing outdoor sports, and sports that do not require facilities. The negative effect of non-labor income on the frequency of practicing certain sports might reveal the presence of proletarian sports (Wilson, 2002). In other words, although participation increases with non-labor income for all sports, some individuals might want to differentiate themselves by moving away from the practice of certain sports associated with lower classes (Bourdieu, 1984; Peterson, 1997).

Finally, the illness covariate also yields interesting results. While this variable generally reduces sports participation, the results on frequency are mixed. It only has a negative relationship with the frequency of playing group sports, whereas it has a positive relationship with the frequency of individual and indoor sports, and sports that require facilities. Moreover, individuals with illnesses also walk more frequently. It may be that ill people play sports that have a therapeutic effect or can be practiced without substantial physical effort such as swimming or yoga.

Although Humphreys and Ruseski (2007) make a different classification and use different data and methodology, the general results about the correlates of the probability of participation are quite similar in comparable categories (i.e., walking, individual, group, and outdoor sports). In the case of frequency, the results are mixed, except for the effect of gender.

Apart from analyzing the positive or negative influence of the variables included, it is worthy to compute the marginal effects in order to check whether there are relevant differences in the response of the dependent variables to changes in the covariates. In doing so, we will gain a better understanding of the importance of the different variables in individual decisions about sports practice.

These effects have been calculated for all individual and family characteristics: gender, age, educational level, number of children aged 12 or less, cohabitation with other adults in the household, health, labor, and marital status. Marginal effects are different for each individual.

Tables 6-9 offer information about the mean values of the total and partial marginal effects of personal and family covariates in each category of physical activity. Table 10 provides information about the mean sample elasticities with respect to wage and non-labor income as well as the standard deviations. (10)

We start by analyzing the marginal effects on walking. Workers and males are less likely to practice this activity, and with lower expected frequency. In particular, the expected frequency of workers who walk is 4.19 times less than that of non-workers. However, the influence of educational variables is much lower. Children and adults in the household decrease the probability of participation but do not affect the frequency of practice. There are two variables that have an opposite effect on the probability of participation and on frequency: age and health status. Age increases the likelihood of walking but reduces the frequency of practice whereas the effect of a bad health status is the opposite. The wage and non-labor income elasticities are very small--smaller than for the rest of defined sports categories. Therefore, the decisions to walk and how often are barely affected by the individual economic situation. This is a sensible result since walking does not require major expenditures on equipment or facilities.

When comparing the marginal effects of the covariates on individual and group sports, we find that educational variables have a total positive effect via their influence on the probability of participation, being the marginal effects on individual sports considerably higher than on group sports. On the contrary, labor status has a negative total effect, which is almost four times greater for individual sports than for group sports. Another noteworthy result is the influence of gender; the marginal effects of being male are positive and higher for group sports than for individual sports. In particular, the probability of participating in group sports is 0.15 times greater for men than women, whereas for individual sports this difference is around 0.03. Another interesting result is that a bad health status reduces the frequency of practicing group sports but increases the frequency of individual sports, though the conditional marginal effects are not high. Regarding the influence of economic variables, they have a positive and greater effect on the probability of participation than on frequency. All elasticities are smaller than 1, although the response to changes in own earnings is higher than that of changes in non-labor income. Moreover, there are not great differences between the elasticities of individual versus group sports.

When we compare the total marginal effects of sports that require facilities and those that do not require facilities, we observe that gender is the variable with the greatest--and most positive--impact on sports that do not require the use of infrastructure, and its effect is much lower on sports that do require facilities. Conversely, family variables, education, and labor status have a much higher absolute total marginal effect on sports played at facilities than on the rest. The partial effects corroborate these results. On the one hand, the probability of participating in sports that do not require facilities is 10 points higher for males than for females and the expected frequency of practice of males conditioned on being a participant increases by almost two times over the previous four weeks. However, in the case of sports requiring facilities, the positive effect of gender on the probability of participation is lower and its effect on the conditional expected frequency is negative. On the other hand, the educational variables also have a greater positive impact on the probability of practicing sports requiring infrastructures. Poor health is negatively associated with the probability of playing sports that require facilities but positively related to the conditional frequency, although the marginal effects are small. As for the elasticities, the most distinctive result is the low--and sometimes not significant--impact of the economic variables on the conditional expected frequency.

Regarding indoor/outdoor sports, gender is the variable with the greatest impact on outdoor sports; its effect is much lower in the case of indoor sports. When we analyze the partial effects, we find that males have a probability of practicing outdoor sports 0.13 times higher than females, compared to 0.04 for indoor sports. In contrast, men who play indoor sports practice them less often than women, contrary to what happens in the case of outdoor sports. With respect to the labor situation, workers have an expected frequency 1.6 times lower than non-workers in both categories. Marital status is the family variable with the highest negative effect on participation in indoor sports and frequency, whereas the presence of adults in the household has the highest impact on participation and frequency of practicing outdoor sports among all variables related to household composition. Finally, the probability of participation is more sensitive to wage changes in the case of outdoor sports, though the elasticity value is below 1.

As discussed previously in this article, the classification of some sports may be troublesome. Therefore, we have made some robustness checks re-estimating the models after changing the classification of some sports that could be controversial. In general, the results are similar, except when we consider group ball games--which have one of the highest participation rates in the sample--as outdoor sports. In this case, we find changes in the effect of some covariates. Some variables become significant (e.g., age and wage in the case of outdoor sports), and others lose their significance, such as term in the probability of practicing outdoor sports. The most remarkable change is the effect of gender; when group ball games are included as indoor sports, males have a higher probability of practicing them and lower frequency of participation, whereas when group ball games are excluded from indoor sports, men are less likely to practice them and there are no gender differences in the frequency of practice. Thus, the relationship between gender and indoor sports is driven by the inclusion of group ball games. (11)

Conclusions

In this study we conducted a disaggregated analysis of sports practice, studying differences and similarities between types of activities. In particular, we analyzed individual participation and frequency during the previous four weeks in walking, individual/group sports, indoor/outdoor sports, and sports that require and do not require facilities.

The database used is the Spanish Time Use Survey 2002-03 and, given the high rates of non-participation, individuals' behavior has been modeled through ZINB count data models. In the empirical specification, we assume that individual decisions about participation and frequency of participation in each type of activities depend on gender, other personal and family characteristics, economic variables, place of residence, and season.

With the estimated coefficients we performed a decomposition of the marginal effects of individual and family characteristics to quantify their influence on the participation and frequency decisions. As far as we are aware, this is the first time a decomposition of the marginal effects of socio-economic covariates has been carried out to discuss frequency and participation habits of individuals in different types of sports activities.

We find differences in the values of marginal effects and, in some cases, differences in the sign of the marginal effect. For example, workers are less likely to participate in sports that require facilities while they are more likely to practice sports that do not require them. Males practice indoor sports and sports at facilities less frequently than females, and the opposite happens with outdoor sports and those that do not require facilities.

Gender tends to be the most relevant variable, although there are some interesting differences among sports as previously above. Sports participation is predominantly male, with the only exception being walking. In frequency decisions, labor status has a high negative association, and this result may be explained by the time-intensive nature of sports. Concerning family variables, they tend to decrease the probability of participation in all sports categories but, in general, they are not significant in the frequency of practice.

Wage and non-labor income elasticities show that the probability of participation is generally more sensitive to changes in the economic variables than the frequency of participation. In addition, wage elasticities are usually greater than non-labor income elasticities. The least sensitive activity to wage changes is walking.

In summary, the aggregation of activities may hide relevant differences in the effect of the covariates on the probabilities of participation and on the frequency of participation in different types of sports. Thus, public authorities should take into account these differences if they want to promote certain activities.

A limitation of this study is that the database is not very recent. It would be interesting to replicate the analysis with more recent data to check whether there have been changes in the effect of the covariates on the practice of different sports.

References

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 judgement of taste. Cambridge, MA: Harvard University Press.

Breuer, C., Hallmann, K., & Wicker, P. (2011). Determinants of sport participation in different sports. Managing Leisure, 16, 269-286.

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

Breuer, C., & Wicker, P. (2009). Decreasing sports activity with increasing age? Findings from a 20-year longitudinal and cohort sequence analysis. Research Quarterly for Exercise and Sport, 80, 22-31.

Buraimo, B., Humphreys, B., & Simmons, R. (2010). Participation and engagement in sport: A double hurdle approach for the United Kingdom. The selected works of Dr. Babatunde Buraimo.

Cabane, C., & Lechner, M. (2015). Physical activity of adults: A survey of correlates, determinants, and effects. Journal of Economics and Statistics, 235, 376-402.

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

Downward, P. (2004). On leisure demand: A post Keynesian critique of neoclassical theory. Journal of Post Keynesian Economics, 26, 371-394.

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

Downward, P., Dawson, A., & Dejonghe, T. (2009). Sports economics: Theory, evidence and policy. London, UK: Butterworth-Heinemann.

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

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

European Commission. (2004). The citizens of the European Union and sport. Special Eurobarometer 213. Retrieved from http://ec.europa.eu/public_opinion/archives/ebs/ebs_213_report_en.pdf

European Commission. (2009). Sport and physical activity. Special Eurobarometer 334. Retrieved from http://ec.europa.eu/public_opinion/archives/ebs/ebs_334_en.pdf

European Commission. (2014). Sport and physical activity. Special Eurobarometer 412. Retrieved from http://ec.europa.eu/public_opinion/archives/ebs/ebs_412_en.pdf

Farrell, L., & Shields, M. (2002). Investigating the economic and demographic determinants of sporting participation in England. Journal of the Royal Statistical Society, 165, 335-348.

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, 515-537.

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

Humphreys, B., & Ruseski, J. (2011). An economic analysis of participation and time spent in physical activity. The BE Journal of Economic Analysis and Policy, 11, 1-47.

Humphreys, B., & Ruseski, J. (2015). The economic choice of participation and time spent in physical activity and sport in Canada. International Journal of Sport Finance, 10, 138-159.

Lera-Lopez, F., & Rapun-Garate, M. (2005). Sports participation versus consumer expenditure on sport: Different determinants and strategies in sports management. European Sports Management Quarterly, 5, 167-186.

INE. (2004). Encuesta de Empleo del Tiempo 2002-2003. Tomo I. Metodologia y Resultados Nacionales. Madrid, Spain: Instituto Nacional de Estadistica.

Long, J., & Freese, J. (2006). Regression models for categorical dependent variables using Stata (2nd ed.). College Station, TX: Stata Press.

McDonald, J., & Moffitt, R. (1980). The uses of tobit analysis. The Review of Economics and Statistics, 62, 318-321.

MECD. (2015). Encuesta de Habitos Deportivos 2015. Madrid, Spain: Ministerio de Educacion, Cultura y Deporte.

Muniz, C., Rodriguez, P., & Suarez, M. J. (2011). The allocation of time to sports and cultural activities: An analysis of individual decisions. International Journal of Sport Finance, 6, 246-264.

Muniz, C., Rodriguez, P., & Suarez, M.J. (2014). Sports and cultural habits by gender: An application using count data models. Economic Modelling, 36, 288-297.

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

Peterson, R. (1997). The rise and fall of highbrow snobbery as a status marker. Poetics, 25, 75-92.

Ruseski, J., Humphreys, B., Hallmann, K., & Breuer, C. (2011). Family structure, time constraints, and sport participation. European Review of Aging and Physical Activity, 8, 57-66.

Scheerder, J., Vanreusel, B., & Taks, M. (2005). Stratification patterns of active sport involvement among adults. International Review for the Sociology of Sport, 40, 139-162.

Stratton, M., Conn, L., Liaw, C., & Conolly, L. (2005). Sport and related physical activity: The social correlates of participation and non-participation by adults. Sport Management Association of Australia and New Zealnd Conference, November 25-26, Canberra, Australia.

Taks, M., Renson, R., & Vanreusel, B. (1994). Of sport, time and money. International Review for the Sociology of Sport, 29, 367-380.

Wilson, T. (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.

Endnotes

(1) See MECD (2015).

(2) See INE (2004).

(3) See Breuer and Wicker (2008, 2009) for an analysis of the effect of time on sports practice in Germany.

(4) See Long and Freese (2006) for more details about the ZINB count data specification and the comparison among count data models.

(5) Labor earnings are only observed for workers. In order to obtain a measure of non-worker wages, we estimated a log wage equation using the subsample of workers and applying a Heckman two-step procedure to correct for sample selection bias. Wage equation estimates are available upon request.

(6) Individual subscripts are omitted for notational convenience.

(7) McDonald and Moffitt (1980) define this decomposition of marginal effects for tobit models.

(8) Pawlowski and Breuer (2012) estimate expenditure elasticities of leisure demand.

(9) The age values that minimize the expected frequency of practice of potential participants range from 36.5 years for indoor sports to 38.1 for sports that require facilities. In the case of walking, the minimum is reached at 23.3 years.

(10) Given that wage is included in logarithms, wage elasticities of the conditional frequency of practice are constant for all individuals and equal to the wage coefficient in frequency equations. This is the reason why the standard deviations are 0 in Table 10.

(11) Estimates are available upon request.

Appendix Table A1. Descriptive Statistics of the Dependent Variables (N = 26,980) Variable Nwalking Number of times individuals have walked in the last four weeks Nindsport Number of times individuals practiced individual sports in the last four weeks Nteamsport Number of times individuals practiced team sports in the last four weeks Noutsport Number of times individuals practiced outdoor sports in the last four weeks Ninsport Number of times individuals practiced indoor sports in the last four weeks Nfacilsport Number of times individuals practiced sports that require facilities in the last four weeks Nnofacilsport Number of times individuals practiced sports that no require facilities in the last four weeks Walking 1 if individuals have walked in the last four weeks, 0 otherwise Indsport 1 if individuals practiced individual sports in the last four weeks, 0 otherwise Teamsport 1 if individuals practiced team sports in the last four weeks, 0 otherwise Outsport 1 if individuals practiced outdoor sports in the last four weeks, 0 otherwise Insport 1 if individuals practiced indoor sports in the last four weeks, 0 otherwise Facilsport 1 if individuals practiced sports that require facilities in the last four weeks, 0 otherwise Nofacilsport 1 if individuals practiced sports that no require facilities in the last four weeks, 0 otherwise Variable Mean Std. Dev. Min. Max. Nwalking 8.8345 11.2137 0 84 Nindsport 2.9429 7.4189 0 80 Nteamsport 0.6880 2.8595 0 60 Noutsport 1.2518 4.3896 0 66 Ninsport 2.3790 6.2475 0 80 Nfacilsport 2.6455 6.6117 0 87 Nnofacilsport 0.9854 3.9444 0 61 Walking 0.5635 0.4960 0 1 Indsport 0.2361 0.4247 0 1 Teamsport 0.1024 0.3032 0 1 Outsport 0.1408 0.3479 0 1 Insport 0.2139 0.4101 0 1 Facilsport 0.2388 0.4264 0 1 Nofacilsport 0.1066 0.3086 0 1 Table A2. Descriptive Statistics of the Explanatory Variables (N = 26,980) Variable Mean Male 1 if 1 if the respondent is a man, 0 otherwise Age Age of respondent Married 1 if respondent is married, 0 otherwise No.Childsl2 Number of children aged 12 years or younger Adult3 1 if respondent lives in a household with more than 2 adults, 0 otherwise Educl (*) 1 if respondent has primary education, 0 otherwise Educ2 (*) 1 if respondent has high school education or vocational training, 0 otherwise Educ3 (*) 1 if respondent has college education, 0 otherwise The reference category is uneducated individual Ill 1 if respondent is ill, unfit or has a disability, 0 otherwise Worker 1 if the respondent has a job, 0 otherwise Log W Logarithm of observed hourly earnings for workers Logarithm of hourly predicted earnings for non-workers, computed from a wage equation through Heckman's two-stage method Nlabinc Non-labor individual income, calculated as income from other household members Quartl 1 if month is January, February or March; 0 otherwise Quart2 1 if month is April, May or June; 0 otherwise Quart3 1 if month is July, August or September; 0 otherwise Urb2 (*) 1 if respondent lives in a township with more than 100,000 inhabitants (and no provincial capital), 0 otherwise Urb3 (*) 1 if respondent lives in a township with fewer than 100,000 inhabitants (and no provincial capital), 0 otherwise Variable Std. Dev. Min. Max. Male 0.4568 0.4981 0 1 Age 41.4183 13.3467 18 65 Married 0.6378 0.4806 0 1 No.Childsl2 0.3724 0.7136 0 6 Adult3 0.4570 0.4982 0 1 Educl (*) 0.3431 0.4747 0 1 Educ2 (*) 0.2679 0.4429 0 1 Educ3 (*) 0.1454 0.3526 0 1 Ill 0.1728 0.3780 0 1 Worker 0.5332 0.4989 0 1 Log W 1.4078 0.5694 -0.6177 5.7599 Nlabinc 1258.7860101 1.6850 0 6000 Quartl 0.2746 0.4463 0 1 Quart2 0.2602 0.4388 0 1 Quart3 0.2263 0.4184 0 1 Urb2 (*) 0.0822 0.2747 0 1 Urb3 (*) 0.5393 0.4985 0 1 (*) The reference category is the provincial capitals.

Jaume Garcia (1), Cristina Muniz (1), Placido Rodriguez (2), and Maria Jose Suarez (2)

(1) Pompeu Fabra University, Spain

(2) University of Oviedo, Spain

Jaume Garcia is professor of applied economics in the Department of Economics and Business. His research interests include sport economics, applied microecono-metrics, housing economics, and individual behavior.

Cristina Muniz is an assistant professor in the Department of Economics. Her research interests focus specifically on various aspects related to the participation of individuals in leisure activities.

Placido Rodriguez is a professor in the Department of Economics and Director of the Sports Economics Observatory Foundation. He is also honorary president of the International Association of Sports Economists.

Maria Jose Suarez is an associate professor in the Department of Economics. Her research interests include labor, cultural, and sport economics. Table 1. Participation and Frequency of Participation Males #Obs. %Particip. Mean (participants) Walking 6194 50.30 14.93 Individual sports 3202 25.98 12.89 Group sports 2203 17.87 6.95 No facilities 2046 16.60 9.70 Facilities 3416 27.72 10.75 Outdoor sports 2668 21.65 9.38 Indoor sports 2944 23.89 10.72 Females #Obs. %Particip. Mean (participants) Walking 9008 61.47 16.19 Individual sports 3192 21.78 12.04 Group sports 575 3.92 5.84 No facilities 846 5.77 8.18 Facilities 3051 20.82 11.43 Outdoor sports 1150 7.85 7.81 Indoor sports 2851 19.45 11.51 Table 2. ZINB Estimates: Walking Frequency Coef. t Male -0.03644 -2.51 Age -0.00782 -2.32 [Age.sup.2]/100 0.01676 4.39 Married -0.00855 -0.55 No. child. [less than or equal to] 12 -0.00107 -0.10 Adult3 0.00913 0.71 Ill 0.05025 3.48 Quart1 0.03201 1.94 Quart2 0.04502 2.75 Quart3 0.09675 5.89 Urb2 -0.03206 -1.42 Urb3 -0.01302 -1.05 Educ1 -0.02673 -1.83 Educ2 -0.03885 -2.06 Educ3 -0.03963 -1.62 Worker -0.27762 - 18.71 Log W -0.00302 -0.18 Nlabinc -0.00004 -6.13 P(A=1) Coef. t Male 0.38186 12.86 Age -0.01567 -2.02 [Age.sup.2]/100 -0.00827 -0.90 Married 0.04460 1.23 No. child. [less than or equal to] 12 0.14163 6.47 Adult3 0.20823 7.08 Ill 0.16532 4.55 Quart1 0.14895 4.12 Quart2 0.16375 4.48 Quart3 0.15265 4.03 Urb2 0.10321 2.09 Urb3 0.04487 1.60 Educ1 -0.13674 -3.85 Educ2 -0.23440 -5.66 Educ3 -0.42235 -8.02 Worker 0.42913 13.35 Log W -0.19168 -6.12 Nlabinc -0.00010 -6.43 N 26980 Log L -72979.57 Table 3. ZINB Estimates: Individual and Group Sports Individual Sports Frequency Coef. t Male 0.07490 2.86 Age -0.00784 -1.14 [Age.sup.2]/100 0.00854 1.03 Married -0.04833 -1.52 No. child. [less than or equal to] 12 -0.00088 -0.04 Adult3 0.05716 2.17 1ll 0.05982 1.77 Quart1 -0.01548 -0.47 Quart2 0.01319 0.40 Quart3 0.04047 1.26 Urb2 -0.03086 -0.69 Urb3 -0.03562 -1.49 Educ1 -0.01329 -0.34 Educ2 -0.00739 -0.18 Educ3 -0.04186 -0.89 Worker -0.15477 -5.32 Log W 0.06488 2.25 Nlabinc -0.00001 -0.90 P(A=1) Coef. t Male -0.16195 -4.73 Age 0.02846 3.07 [Age.sup.2]/100 -0.01037 -0.93 Married 0.31486 7.50 No. child. [less than or equal to] 12 0.15960 5.94 Adult3 0.26334 7.47 Ill 0.14508 3.12 Quart1 0.10275 2.36 Quart2 0.03000 0.69 Quart3 -0.22215 -5.01 Urb2 0.04744 0.80 Urb3 0.10811 3.28 Educ1 -0.49613 -9.85 Educ2 -0.86324 -16.07 Educ3 -1.09848 -17.55 Worker 0.06664 1.74 Log W -0.34250 -9.22 Nlabinc -0.00018 -10.80 N 26980 Log L -35612.61 Group Sports Frequency Coef. t Male 0.18150 3.28 Age -0.05936 -4.31 [Age.sup.2]/100 0.07938 4.49 Married -0.07950 -1.13 No. child. [less than or equal to] 12 -0.03368 -0.84 Adult3 0.00434 0.09 1ll -0.14444 -1.95 Quart1 0.08133 1.46 Quart2 -0.00402 -0.07 Quart3 0.03450 0.59 Urb2 -0.06301 -0.86 Urb3 0.02264 0.54 Educ1 0.02356 0.26 Educ2 -0.08413 -0.93 Educ3 -0.08644 -0.89 Worker -0.18394 -3.66 Log W 0.12113 2.63 Nlabinc -0.00003 -1.29 P(A=1) Coef. t Male -1.85784 -31.52 Age 0.12193 8.05 [Age.sup.2]/100 -0.06908 -3.65 Married 0.08765 1.23 No. child. [less than or equal to] 12 0.05180 1.24 Adult3 0.26280 4.76 Ill 0.31132 3.82 Quart1 0.14211 2.17 Quart2 -0.05877 -0.90 Quart3 0.04984 0.72 Urb2 0.00979 0.11 Urb3 0.18935 3.75 Educ1 -0.46927 -5.44 Educ2 -0.90963 -10.38 Educ3 -1.29421 -12.84 Worker 0.05203 0.89 Log W -0.24132 -4.20 Nlabinc -0.00017 -6.92 N 26980 Log L -14670.07 Table 4. ZINB Estimates: Sports That Do Not Require/Require Facilities No Facilities Frequency Coef. t Male 0.21047 4.66 Age -0.00639 -0.54 [Age.sup.2]/100 0.00875 0.60 Married 0.04366 0.78 No. child. [less than or equal to] 12 -0.05668 -1.65 Adult3 0.05313 1.24 Ill -0.03111 -0.47 Quart1 0.07308 1.41 Quart2 0.06635 1.28 Quart3 0.03010 0.58 Urb2 -0.08396 -1.14 Urb3 -0.02963 -0.76 Educ1 -0.03240 -0.46 Educ2 -0.06425 -0.91 Educ3 -0.07205 -0.93 Worker -0.19283 -3.95 Log W 0.01325 0.32 Nlabinc -0.00006 -3.38 P(A=1) Coef. t Male -1.11343 -22.86 Age 0.01330 0.96 [Age.sup.2]/100 0.02430 1.41 Married 0.34443 5.71 No. child. [less than or equal to] 12 0.06954 1.90 Adult3 0.20897 4.19 Ill 0.34349 4.67 Quart1 0.20416 3.40 Quart2 0.13732 2.29 Quart3 -0.00451 -0.07 Urb2 -0.00021 -0.01 Urb3 0.04395 0.96 Educ1 -0.33618 -4.38 Educ2 -0.75860 -9.75 Educ3 -0.96639 -10.88 Worker -0.08863 -1.65 Log W -0.35899 -7.10 Nlabinc -0.00012 -5.14 N 26980 Log L -17114.3 Facilities Frequency Coef. t Male -0.06956 -2.62 Age -0.02074 -2.95 [Age.sup.2]/100 0.02725 3.21 Married -0.07542 -2.26 No. child. [less than or equal to] 12 -0.00774 -0.37 Adult3 0.03637 1.39 Ill 0.06621 1.92 Quart1 -0.03613 -1.12 Quart2 -0.00317 -0.10 Quart3 0.08100 2.50 Urb2 -0.02851 -0.64 Urb3 -0.02614 -1.10 Educ1 0.00520 0.13 Educ2 -0.01528 -0.36 Educ3 -0.06163 -1.29 Worker -0.17505 -6.10 Log W 0.09963 3.45 Nlabinc 0.00001 0.47 P(A=1) Coef. t Male -0.37598 -10.74 Age 0.06882 7.37 [Age.sup.2]/100 -0.04560 -4.01 Married 0.27275 6.31 No. child. [less than or equal to] 12 0.15598 5.73 Adult3 0.26779 7.42 Ill 0.14465 3.05 Quart1 0.05274 1.19 Quart2 -0.02034 -0.46 Quart3 -0.17412 -3.83 Urb2 0.04056 0.68 Urb3 0.16226 4.84 Educ1 -0.52427 -10.17 Educ2 -0.87598 -16.03 Educ3 -1.15070 -18.00 Worker 0.12399 3.21 Log W -0.32688 -8.69 Nlabinc -0.00021 -12.31 N 26980 Log L -34707.44 Table 5. ZINB Estimates: Outdoor and Indoor Sports Outdoor Sports Frequency Coef. t Male 0.23598 5.67 Age -0.00779 -0.74 [Age.sup.2]/100 0.01164 0.89 Married 0.01178 0.23 No. child. [less than or equal to] 12 -0.05403 -1.79 Adult3 0.07617 1.97 Ill -0.07329 -1.21 Quart1 0.02583 0.54 Quart2 0.04773 1.00 Quart3 0.03413 0.72 Urb2 -0.10750 -1.62 Urb3 0.00338 0.09 Educ1 -0.03015 -0.46 Educ2 -0.06522 -1.00 Educ3 -0.03229 -0.45 Worker -0.20266 -4.57 Log W 0.01683 0.44 Nlabinc -0.00005 -3.23 P(A=1) Coef. t Male -1.15774 -25.95 Age 0.02857 2.26 [Age.sup.2]/100 0.00641 0.41 Married 0.28846 5.22 No. child. [less than or equal to] 12 0.07886 2.38 Adult3 0.31416 6.77 Ill 0.34146 5.23 Quart1 0.10343 1.89 Quart2 0.11951 2.15 Quart3 0.01382 0.24 Urb2 0.01595 0.21 Urb3 0.09355 2.21 Educ1 -0.38710 -5.56 Educ2 -0.86299 -12.14 Educ3 -1.14175 -14.09 Worker -0.06987 -1.40 Log W -0.42575 -9.02 Nlabinc -0.00019 -8.76 N 26980 Log L -21300.1 Indoor Sports Frequency Coef. t Male -0.09037 -3.30 Age -0.01924 -2.68 [Age.sup.2]/100 0.02637 3.04 Married -0.07419 -2.18 No. child. [less than or equal to] 12 0.00564 0.26 Adult3 0.04435 1.66 Ill 0.07705 2.21 Quart1 -0.02369 -0.72 Quart2 -0.01091 -0.32 Quart3 0.05693 1.73 Urb2 -0.02516 -0.55 Urb3 -0.02223 -0.91 Educ1 0.00576 0.14 Educ2 -0.02538 -0.60 Educ3 -0.09922 -2.07 Worker -0.15106 -5.12 Log W 0.11695 3.95 Nlabinc 0.00001 0.19 P(A=1) Coef. t Male -0.25637 -7.16 Age 0.06749 7.11 [Age.sup.2]/100 -0.04364 -3.77 Married 0.28427 6.39 No. child. [less than or equal to] 12 0.16360 5.75 Adult3 0.23416 6.38 Ill 0.09906 2.03 Quart1 0.09645 2.12 Quart2 -0.03865 -0.85 Quart3 -0.23112 -4.98 Urb2 0.01787 0.30 Urb3 0.16668 4.85 Educ1 -0.51687 -9.69 Educ2 -0.83587- 14.77 Educ3 -1.05745- 16.01 Worker 0.14832 3.77 Log W -0.26967 -7.03 Nlabinc -0.00018- 10.71 N 26980 Log L -31900.1 Table 6. Marginal Effects: Walking [delta]E(y)/x Mean Std. Dev. Male -1.661 0.32 Age -0.011 0.02 Married -0.231 0.05 No.child[less than or equal to]12 -0.502 0.09 Adult3 -0.645 0.11 Ill -0.146 0.12 Educ1 0.256 0.06 Educ2 0.228 0.04 Educ3 0.616 0.09 Worker -3.948 0.66 Walking [delta]Pr(A=0)/ x Mean Std. Dev. Male -0.090 (***) 0.01 Age 0.004 (**) 0.00 Married -0.010 0.00 No.child[less than or equal to]12 -0.033 (***) 0.00 Adult3 -0.048 (***) 0.00 Ill -0.039 (***) 0.00 Educ1 0.032 (***) 0.00 Educ2 0.023 (***) 0.00 Educ3 0.043 (***) 0.00 Worker -0.101 (***) 0.01 [delta]E(y/A=0)/ x Mean Std. Dev. Male -0.549 0.11 Age -0.113 (**) 0.02 Married -0.129 0.03 No.child[less than or equal to]12 -0.016 0.00 Adult3 0.138 0.03 Ill 0.771 (***) 0.15 Educ1 -0.408 (*) 0.08 Educ2 -0.182 (**) 0.04 Educ3 -0.012 0.00 Worker -4.192 (***) 0.50 Note: (*)p<0.1, (**) p<0.05, (***) p<0.01 Table 7. Marginal Effects: Individual and Group Sports [delta]E(y)/[delta]x Mean Std. Dev. Male 0.547 0.24 Age -0.079 0.03 Married -0.793 0.28 No.child[less than or equal to]12 -0.322 0.12 Adult3 -0.357 0.11 Ill -0.117 0.04 Educl 0.831 0.28 Educ2 0.810 0.22 Educ3 0.435 0.08 Worker -0.598 0.31 Individual Sports [delta]Pr(A=0)/[delta]x Mean Std. Dev. Male 0.027 (***) 0.01 Age -0.005 (***) 0.00 Married -0.054 (***) 0.02 No.child[less than or equal to]12 -0.027 (***) 0.01 Adult3 -0.044 (***) 0.02 Ill -0.024 (***) 0.01 Educl 0.072 (***) 0.02 Educ2 0.066 (***) 0.01 Educ3 0.048 (***) 0.01 Worker -0.011 (*) 0.00 [delta]E(y)/[delta]x Mean Std. Dev. Male 0.891 (***) 0.08 Age -0.091 0.01 Married -0.577 0.05 No.child[less than or equal to]12 -0.010 0.00 Adult3 0.679 (**) 0.06 Ill 0.723 (*) 0.07 Educl -0.159 0.02 Educ2 0.070 0.01 Educ3 -0.404 0.04 Worker -1.844 (***) 0.15 Group Sports [delta]E(y/A=0)/[delta]x Mean Std. Dev. Male 0.966 0.84 Age -0.097 0.11 Married -0.096 0.11 No.child[less than or equal to]12 -0.048 0.06 Adult3 -0.122 0.12 Ill -0.219 0.26 Educl 0.196 0.23 Educ2 0.141 0.12 Educ3 0.215 0.18 Worker -0.156 0.20 [delta]Pr(A=0)/[delta]x Mean Std. Dev. Male 0.150 (***) 0.12 Age -0.010 (***) 0.01 Married -0.007 0.01 No.child[less than or equal to]12 -0.004 0.00 Adult3 -0.021 (***) 0.02 Ill -0.024 (***) 0.02 Educl 0.029 (***) 0.03 Educ2 0.036 (***) 0.03 Educ3 0.039 (***) 0.03 Worker -0.004 0.00 [delta]E(y/A=0)/[delta]x Mean Std. Dev. Male 1.062 (***) 0.24 Age -0.336 (***) 0.08 Married -0.468 0.12 No.child[less than or equal to]12 -0.196 0.05 Adult3 0.025 0.01 Ill -0.804 (*) 0.21 Educl 0.142 0.03 Educ2 -0.624 0.15 Educ3 -0.013 0.00 Worker -1.076 (***) 0.25 Note: (*)p<0.1, (**) p<0.05, (***) p<0.01 Table 8. Marginal Effects: Sports That Do Not Require/Require Facilities [delta]E(y)/[delta]x Mean Std. Dev. Male 1.040 0.59 Age -0.017 0.01 Married -0.231 0.16 No.child[less than or equal to]12 -0.110 0.09 Adult3 -0.110 0.07 111 -0.273 0.21 Educl 0.189 0.14 Educ2 0.306 0.20 Educ3 0.188 0.11 Worker -0.123 0.12 Sports that do Not Require Facilities [delta]Pr(A=0)/[delta]x Mean Std. Dev. Male 0.103 (***) 0.06 Age -0.001 0.00 Married -0.033 (***) 0.02 No.child[less than or equal to]12 -0.006 (*) 0.00 Adult3 -0.019 (***) 0.01 111 -0.029 (***) 0.02 Educl 0.024 (***) 0.02 Educ2 0.040 (***) 0.02 Educ3 0.024 (***) 0.01 Worker 0.008 (*) 0.01 [delta]E(y/A=0)/[delta]x Mean Std. Dev. Male 1.789 (***) 0.24 Age -0.052 0.01 Married 0.366 0.06 No.child[less than or equal to]12 -0.478 (*) 0.08 Adult3 0.449 0.07 Ill -0.260 0.04 Educl -0.279 0.04 Educ2 -0.266 0.04 Educ3 -0.064 0.01 Worker -1.639 (***) 0.25 [delta]E(y)/[delta]x Mean Std. Dev. Male 0.467 0.16 Age -0.170 0.08 Married -0.686 0.29 No.child[less than or equal to]12 -0.289 0.12 Adult3 -0.362 0.13 Ill -0.077 0.05 Educl 0.809 0.33 Educ2 0.601 0.18 Educ3 0.407 0.08 Worker -0.688 0.39 Sports that Require Facilities [delta]Pr(A=0)/[delta]x Mean Std. Dev. Male 0.062 (***) 0.02 Age -0.011 (***) 0.00 Married -0.046 (***) 0.02 No.child[less than or equal to]12 -0.026 (***) 0.01 Adult3 -0.043 (***) 0.02 Ill -0.023 (***) 0.01 Educl 0.074 (***) 0.03 Educ2 0.062 (***) 0.02 Educ3 0.054 (***) 0.02 Worker -0.020 (***) 0.01 [delta]E(y/A=0)/[delta]x Mean Std. Dev. Male -0.735 (***) 0.10 Age -0.214 (***) 0.03 Married -0.808 (**) 0.11 No.child[less than or equal to]12 -0.082 0.01 Adult3 0.386 0.05 Ill 0.717 (*) 0.10 Educl 0.056 0.01 Educ2 -0.219 0.04 Educ3 -0.479 0.07 Worker -1.860 (***) 0.18 Note: (*)p<0.1, (**) p<0.05, (***) p<0.01 Table 9. Marginal Effects: Outdoor and Indoor Sports Outdoor Sports [delta]E(y)/[delta]x Mean Std. Dev. Male 1.315 0.73 Age -0.035 0.03 Married -0.251 0.16 No.child[less than or equal to]l2 -0.139 0.11 Adult3 -0.188 0.11 Ill -0.364 0.27 Educl 0.258 0.18 Educ2 0.399 0.24 Educ3 0.367 0.21 Worker -0.195 0.20 [delta]Pr(A=0)/[delta]x Mean Std. Dev. Male 0.133 (***) 0.07 Age -0.003 (**) 0.00 Married -0.034 (***) 0.02 No.child[less than or equal to]l2 -0.009 (**) 0.01 Adult3 -0.035 (***) 0.02 Ill -0.037 (***) 0.02 Educl 0.035 (***) 0.02 Educ2 0.056 (***) 0.03 Educ3 0.040 (***) 0.02 Worker 0.008 0.01 [delta]E(y/A=0)/[delta]x Mean Std. Dev. Male 1.886 (***) 0.26 Age -0.060 0.01 Married 0.093 0.02 No.child[less than or equal to]l2 -0.428 (*) 0.07 Adult3 0.606 (**) 0.10 Ill -0.568 0.10 Educl -0.243 0.04 Educ2 -0.274 0.04 Educ3 0.257 0.04 Worker -1.619 (***) 0.26 Indoor Sports [delta]E(y)/[delta]x Mean Std. Dev. Male 0.207 0.07 Age -0.154 0.07 Married -0.660 0.27 No.child[less than or equal to]l2 -0.255 0.10 Adult3 -0.277 0.10 Ill 0.019 0.06 Educl 0.759 0.32 Educ2 0.491 0.15 Educ3 0.211 0.05 Worker -0.612 0.33 [delta]Pr(A=0)/[delta]x Mean Std. Dev. Male 0.040 (***) 0.02 Age -0.010 (***) 0.00 Married -0.045 (***) 0.02 No.child[less than or equal to]l2 -0.025 (***) 0.01 Adult3 -0.036 (***) 0.01 Ill -0.015 (**) 0.01 Educl 0.068 (***) 0.03 Educ2 0.052 (***) 0.02 Educ3 0.041 (***) 0.01 Worker -0.023 (***) 0.01 [delta]E(y/A=0)/[delta]x Mean Std. Dev. Male -0.970 (***) 0.12 Age -0.202 (***) 0.03 Married -0.809 (**) 0.11 No.child[less than or equal to]l2 0.061 0.01 Adult3 0.479 0.06 Ill 0.851 (**) 0.11 Educl 0.064 0.01 Educ2 -0.339 0.04 Educ3 -0.763 (**) 0.10 Worker -1.633 (***) 0.16 Note: (*) p<0.1, (**) p<0.05, (***) p<0.01 Table 10. Wage and Non-Labor Income Elasticities WAGE ELASTICITIES Total Pr. Elasticity Potential Participant Mean St. Dv. Mean St. Dv. Walking 0.079 0.022 0.082 (***) 0.022 Individual Sports 0.323 0.045 0.258 (***) 0.045 Group Sports 0.335 0.034 0.214 (***) 0.034 No Facilities Sports 0.330 0.330 0.317 (***) 0.036 Facilities Sports 0.345 0.050 0.245 (***) 0.050 Outdoor Sports 0.376 0.058 0.359 (***) 0.058 Indoor Sports 0.327 0.037 0.210 (***) 0.037 NON-LABOR INCOME ELASTICITIES Total Pr. Elasticity Potential Participant Mean St. Dv. Mean St. Dv. Walking -0.002 0.018 0.049 (***) 0.038 Individual Sports 0.149 0.110 0.162 (***) 0.120 Group Sports 0.157 0.126 0.189 (***) 0.150 No Facilities Sports 0.050 0.042 0.129 (***) 0.101 Facilities Sports 0.194 0.144 0.187 (***) 0.140 Outdoor Sports 0.125 0.100 0.192 (***) 0.149 Indoor Sports 0.173 0.129 0.170 (***) 0.127 WAGE ELASTICITIES Expected Number of Counts Mean St. Dv. Walking -0.003 0 Individual Sports 0.065 (**) 0 Group Sports 0.121 (***) 0 No Facilities Sports 0.013 0 Facilities Sports 0.100 (***) 0 Outdoor Sports 0.017 0 Indoor Sports 0.117 (***) 0 NON-LABOR INCOME ELASTICITIES Expected Number of Counts Mean St. Dv. Walking -0.051 (***) 0.041 Individual Sports -0.013 0.011 Group Sports -0.032 0.026 No Facilities Sports -0.079 (***) 0.064 Facilities Sports 0.007 0.005 Outdoor Sports -0.068 (***) 0.054 Indoor Sports 0.003 0.002 Note: (*) p<0.1, (**) p<0.05, (***) p<0.01