The analysis of the individual spending behavior of visitors to specific destinations is increasingly becoming a topic of interest, according to Brida, Disegna, and Osti (2013), Craggs and Schofield (2009), Dolnicar et al. (2008), Hung, Shang, and Wang (2012) and Nicolau and Más (2005). Tourism demand is mainly analyzed using a macro approach, in which the unit of analysis is a set of aggregated data. These kinds of studies use economic indicators such as the influence of the tourism sector on GDP, GVA or employment at the national or regional level. Individual spending behavior and the sociodemographic and economic factors that affect spending patterns (i.e., the micro approach) have been studied to a lesser extent (Brida et al., 2013; Fredman, 2008; Laesser & Crouch, 2006).

Reviews of the previous literature on tourism demand including the works of Lim (1997, 2006) and Crouch (1994) reflect that there is a lack of studies employing micro-economic analysis. Lim (2006) explains that only 8 out of the 124 articles studied employed this type of analysis. Wang and Davidson (2010) indicate that the first studies concerning demand at a micro level were carried out by Mak, Moncur, and Yonamine (1977). Subsequently, this type of work aroused the interest of researchers once again in the 1990s. As of then, interest increased rapidly, especially at the turn of the century.

In the case of sports tourism, the same situation is observed. Studies have neglected the specific aspects of products and services and have focused on economic impact and the use of aggregated data. It therefore now seems reasonable to also focus on understanding the spending behavior of sports tourists who come as spectators (Brida, Schubert, Osti, & Barquet 2011; Cannon & Ford, 2002; Yusof, Shah, & Geok, 2012) and participants (Dixon, Backman, Backman, & Norman, 2012; Downward, Lumsdon, & Weston, 2009; Gibson, 2005).

Mok and Iverson (2000) explain that the growth of international tourism and tourist spending has extended the interest in carrying out research on consumer behavior. Understanding the patterns and activities of tourist spending when visiting a particular destination is a key issue in strategic facility planning. Knowing the factors influencing tourist behavior can lead to the better planning of marketing and sales management and improved industry opportunities (Laesser & Crouch, 2006; Oh, Cheng, Lehto, & O’Leary, 2004; Saayman & Saayman, 2009; Saayman & Saayman, 2012).

In light of the above, this paper aims to identify the factors influencing the expenditure of those attending the matches of a top-level professional league basketball team. The team Rio Natura Monbus Obradoiro (commonly known as Obradoiro), which plays in the top Spanish league (Liga Endesa), was taken as the case example for this study. The Liga Endesa is Spain's professional basketball league. It is a relevant sporting event that has gained both national and international importance. It consists of 17 teams that regularly compete. The top eight finishers at the end of the regular season face a second round of play-offs, after which the League champion is proclaimed. The paper additionally analyses the existence of differences in the spending behavior pattern of attendees according to how much they spend.

The contribution of this work lies in the way it analyses the determinants of spending for a type of sporting event that, as yet, has not been discussed in the literature (for a professional team over a whole season). This kind of empirical study is becoming increasingly relevant, given that leagues differ from occasional events (which have been more widely studied). Events that take place every season are capable of attracting attendees to the territory regularly. Besides, attendees often repeat their visit throughout the season. It is thus increasingly important to collect reliable information about attendee spending behavior. Event organizers will be better equipped to attract attendees end develop their loyalty, while policy makers could use the results to help maximize the spending impact on the local economy. Another contribution of this work rests on the use of sports factors in addition to factors most commonly employed in previous models: sociodemographic, economic and tourist variables.

The paper is organized as follows. The next section presents the literature review relating to expenditure behaviors in tourism and sport events. This is followed by a description of the sample and variables used in the empirical analysis, before an explanation of the methodology. The fifth section presents the results, with the final sections containing the discussion and conclusions.

The studies on tourist spending behavior by Brida and Scuderi (2013), Lehto, O’Leary, and Morrison (2002) and Wang, Rompf, Severt, and Peerapatdit (2006) affirm that economic and sociodemographic variables, as well as characteristics related to travel, are the most frequently used in studies on spending determinants. As these works explain, they use economic variables such as income, assets or the existence of economic difficulties. In terms of the sociodemographic variables, current models have used age, education, sex, marital status, residence, occupation or profession and race. For their part, travel characteristics are usually represented by accommodation, length of stay, activities undertaken during the trip, destinations visited, source of travel information, transportation, purpose for traveling, previous experience at the destination or group size (Brida & Scuderi, 2013; Wang & Davidson, 2010).

Specifically, studies of the determinants of spending at sporting events have constructed models for analysis from a combination of socioeconomic variables and the demographic characteristics. Every sporting event is observed to have different characteristics influencing the choice of variables used to build the model. Bilgic, Florkowski, Yoder, and Schreiner (2008), Taks, Green, Chalip, Kesenne, and Martyn (2013) and Downward et al. (2009) have analyzed participant spending. Meanwhile, Yusof et al. (2012), Brida et al. (2011) and Cannon and Ford (2002) have focused on spectator spending.

In relation to economic factors, Brida and Scuderi (2013) and Wang and Davidson (2010) explain, individual income is one of the most important and commonly used determinants in these types of studies. Yet, respondents are more reluctant to share this type of data. Thus, in cases where it is not possible to obtain personal income for the dataset, other variables are used to represent the level of income. Abbruzzo, Brida, and Scuderi (2014), Aguiló Perez and Juaneda (2000), Medina-Muñoz and Medina-Muñoz (2012) and Svensson, Moreno, and Martín (2011) employ variables such as occupation and/or profession to represent the level of income.

Taking into consideration other kind of factors, Abbruzzo et al. (2014) explain that, it is interesting to consider the variable of other activities, that is, if attendee performs other activities besides attending the event, such as, tourism, shopping or leisure activities. The objective of this variable is to identify whether there are activities that are contributing to increase attendee spending in the geographic area where the study is taking place. Another case is represented by variables that characterize the event. Authors such as Bilgic et al. (2008), Dixon et al. (2012), Shani, Wang, Hutchinson, and Lai (2010) and Sato, Jordan, Kaplanidou, and Funk (2014) include this kind of variables. However, most papers do not do this.

Economic and sociodemographic attributes are insufficient to understand decisions concerning tourist destination. Therefore, the use of psychographic variables is proposed to justify decisions regarding tourism destination development (Lehto et al., 2002; Wang et al., 2006). Psychographic variables include all of the characteristics that may influence consumer response to products and services (Lehto et al., 2002) such as interests, opinions, attitudes and lifestyles (Brida & Scuderi, 2013). Thrane (2002), Chen and Chang (2012), Henthorne (2000), Aguiló Perez and Juaneda (2000) and Abbruzzo et al. (2014) include psychographic variables when studying the determinants of expenditure. Psychographic variables have proven to be powerful predictors of travellers’ decisions in tourist destinations (Lehto, Morrison, & O’Leary, 2001; Lehto et al., 2002) and, by association, they can also contribute to a better understanding of spending patterns (Wang et al., 2006).

Metric and categorical response models are used to study the determinants of cost models. In the former, the most common method is the classical linear regression model by ordinary least squares (OLS). The most usual alternative estimation procedures to the OLS are Tobit regression model, Double-Hurdle Model and the Negative Binomial Model (Brida & Scuderi, 2013). Other alternative to OLS model is the quantile regression approach which makes possible to assess local behavior in specific portions of the empirical distribution according to the different quantiles rather than the mean values. The universal Logit, Probit and Logit ordinal models are among the categorical response models (Brida & Scuderi, 2013).

In the case of sports tourism, there are studies that identify the determinants of event attendance using quantile regression. This is the case of papers by Jane, Kuo, Wu, and Chen (2010), that it investigates the determinants of game-day attendance in the Chinese Professional Baseball League (CPBL) from 2001 to 2007. Other example is the paper by Serrano, García-Bernal, Fernández-Olmos, and Espitia-Escuer (2015). It focuses on the relationship between the expected quality of the event and attendance at the European football stadiums. Both works found that the outcome uncertainty and the quality of the contestant teams have positive effect on attendance demand. However, no studies have been conducted using quantile regression to analyze the spending patterns of those attending a sporting event. This study addresses this issue by providing evidence concerning the influence of a variety of proposed factors on different spending levels.

The data is derived from a survey of attendees at Obradoiro matches. A total of 2797 surveys were conducted at 11 of the team's 17 home matches during the 2012/2013 season and at the play-off game. Non-residents account for 64% of the sample, while residents account for 36%.

The survey was conducted on attendants with residence in the city of Santiago de Compostela (where the event takes place) as well as non-residents in the city. To ensure the randomness of the sample, interviewers followed a pattern of conduct addressing one out of every three persons (but only one person per group) in each area of the venue during different moments throughout the event. Nevertheless, this study only used data from the non-resident attendants whose sole purpose for traveling to the city was to attend the basketball match. The definition of economic impact indicates that only this type of attendee injects cash into the territory (Preuss, Könecke, & Schütte, 2010). Preliminary works on spending determinants have followed the same premise. Such is also the case of Dixon et al. (2012) and Sato et al. (2014).

The dependent variable is measured by the average individual expenditure per match attendant (only for non-residents). Attendee expenditure refers the total spending by each visitor in of Santiago de Compostela. It includes accommodation, food, shopping, leisure and other expenses. Attendee expenditure does not include transport expenditure to Santiago de Compostela from their city of origin. The surveyor first asks about the different expenditure items. If the answers are partial or null, the surveyor enquires about aggregate spending. If the respondents do not give a figure, they are asked to choose from one of the proposed intervals (less than 50€, 51€–100€, 101€–200€, 201€–300€, 301€–400€, 401€–500€ or over 500€). The average individual spending of the attendees is obtained from the data provided. The distribution of expenditure is normalized using its natural logarithm.

Expenditure on tickets (admission fees) is computed from the information provided by the event organizers. Attendees have not been asked about the money spent on tickets. In order to know a particular attendee's total expenditure, the average price of the ticket for each match has been added to the rest of the expenses. Ticket price varies from match to match.

The explanatory variables are classified into sociodemographic, economic, tourist, psychographic and sports factors. A list and description of these variables are introduced in Table 1. The sociodemographic factors employed are age, gender and offspring. According to Wang and Davidson (2010), there are no common empirical findings concerning the effect of sociodemographic variables. Existing studies are very diverse and use different units of analysis. Marrocu, Paci, and Zara (2015) use sociodemographic variables, such as gender, as control variables in their empirical models.

The variable offspring is defined as a dummy variable indicating whether attendants are accompanied by their children. The survey is only addressed to people over 18 years of age. It considers that people who travel with their children take on their expenses. In studies of determinants of spending at tourist destinations, this variable usually refers to the number of household members or the number of minors (Brida & Scuderi, 2013). Here, this variable indicates that spectators have traveled to the match accompanied by their children. This data is more significant for the study, given that a person may have children and not take them along to the match.

Economic factors are measured by a categorical variable indicating whether the individual is employed or unemployed. The personal income of the attendants is not available in dataset. In this way, an alternative variable is used such as occupation. Regarding tourist factors, the origin of those attending and those undertaking other activities in the city in addition to attending the match (dummy variable) are included. In this paper, the origin of the attendees is classified into attendees from the rest of the province of A Coruña, the rest of Galicia and the rest of Spain.

A set of psychographic variables is proposed in order to check the extent to which they can explain the spending behavior of individuals attending a match. The set of psychographic variables consists of a dummy variable considering whether the espectator is attending the match for the first time or is repeating their attendance. In addition, three Likert scale variables express the attendee's assessment of the event, the city and the site of the venue (1=very negative to 5=very positive). We have created three dummy variables from these Likert type variables. They reflect if the image of the city, the event or the city as host is good (when the value in the response in the Likert scale is 4 or 5). These variables are those that we include in the model.

Finally, we consider a number of factors related to the type of event. Specifically, for this study, variables such as rival team, difference of matches won with the rival team, time of the match, broadcast on TV and price of the ticket have been included.

Certain characteristics of the event under study can influence attendee spending. The match schedule is not always the same. Matches may be held in the morning or afternoon. Similarly, the admission price varies depending on the type of match and the opponents. These data have been computed from the total ticket revenue and the attendants at each match. Thus, it is interesting to know how both variables influence individual attendee spending.

The variables Euroleague and Eurocup, which define whether the rival team usually plays in the Euroleague or the Eurocup, reflect the sporting level of the opponent team. The variable DIF represents the difference in matches won previously in the season with the opposing team. This figure is estimated by deducting the number of matches won by the away team from the matches won by Obradoiro at that time in the season (it can take positive and negative values). The variable TVG represents the matches broadcast by the regional television channel, Galician Television.

These variables seem to have an obvious influence on attendance. If the opposing team is at the top of the league table, if there is little difference between the teams in terms of points or if it is a game not being aired on TV, it seems reasonable to think that these factors will incentivize attendance. Although their influence on individual spending does not seem to be so perceptible, these variables affect the attendance of non-regular viewers (who are attracted by the opponent or the competitiveness of the game) whose spending behavior may differ from that of regular attendees. The significant effect of sports factors can be explained by the intangible effect generated by the attraction of spectators who are not regular viewers, who may take advantage of their visit to carry out other activities such as cultural tourism and shopping. Quick (2000) explains that attendees at sporting events are not homogeneous. Therefore, it has become increasingly relevant to identify how these variables affect the spending behavior of attendees.

The determinants of expenditure are analyzed by estimating an OLS model and a quantile regression model. The first econometric tool is useful for detecting the central tendency of the data. It estimates the average spending response of those attending matches, to changes in the independent variables. The second analysis estimates the response in specific levels of spending; thereby, it identifies differences in the spending patterns of spectators according to how much they spend.

Ever since the seminal work of Koenker and Basset (1978) explaining its different scientific functionalities, quantile regression has been used to analyze determinants in different areas, including the economy and tourism. The work developed by Buchinsky (1994) is an example that applies quantile regression to study the wage structure of the United States in the period 1963–1987. A more recent paper by Koenker and Hallock (2001) explains the use of quantile regression in different areas and empirical works. Several authors emphasize that the additional information it provides facilitates understanding and explains tourist spending patterns (Chen & Chang, 2012; Hung et al., 2012; Lew & Ng, 2012; Marrocu et al., 2015), therefore making it suitable for analysing spending determinants.

As explained by Marrocu et al. (2015), when there is significant heterogeneity in the effects of variables, quantile regression is expected to provide a more complete picture of spending behavior, given that the effect of the explanatory variables can be observed across the entire distribution. Attendance at basketball matches involves a common minimum expense associated with the payment of an admission fee. From this expense, it can be produced expenses in accommodation, restaurant or shopping that depends on the different activities carried out by the assistant in Santiago de Compostela. Thus, a quantile regression provides additional information on the key factors influencing spending and reveals the marginal effects of the explanatory variables for different quantiles of tourism expenditure.

This paper analyses the expenditure of match attendees. This expenditure is no censored because there are no restrictions. It is different when the dependent variable is the demand for tickets (attendance), where some observations may be censored. On the other hand, the population under study here is people attending the matches. Those attending each match are random samples, because everybody wanting to attend the match can do so due to the fact that none of the matches were sold out. In that sense, there is no problem of selection bias.

The OLS model is formulated as follows:

• -

  E: economic variables

• -

  SD: sociodemographic variables

• -

  S: sports variables

• -

  T: tourist variables

• -

  PS: psychographic variables

The model is expressed as follows:

The quantile regression developed by Koenker and Basset (1978) seeks to model the relationship between X and Y for different quantiles of the distribution of the dependent variable Y.

The estimation of the parameters, in the case of the quantile regression, is performed by minimizing weighted absolute deviations with asymmetrical weights. In this case, the dependent variable Yi is represented – as it was in the linear regression – by the average individual expenditure of match attendees. Yi is transformed through its natural logarithm. The variables Ei, SDi, Si, Ti and PSi represent the set of independent variables (Xi) listed in the previous section.

“One of the main advantages of a quantile regression as opposed to an OLS is the use of deviations in absolute values rather than the squared deviations to estimate the parameters βi. The estimates provided by the quantile regression are substantially unaltered by extreme values given that it ‘penalizes’ errors linearly; whereas, the OLS regression gives greater importance to raising errors to the square.” (Vicéns & Sánchez, 2012, p. 8).