The effects of advertising, previous experience and recommendation in tourist´s length of stay: a hurdle count data approach.
Authors and e-mails of them:
David Boto García
Department: ECONOMICS
University: University of Oviedo
Subject area: Tourism and territory
Abstract:This article analyses tourist´s length of stay in a particular destination (Asturias) using a Hurdle Negative Binomial model which allow us to firstly separate hikers from proper tourists and then explain the length of stay of those who actually stay for more than a day. Apart from sociodemographic characteristics of the individual, we are interested in the effect of distance, mode of transportation, type of accommodation and party size. Moreover, one of the relevant features this paper addresses is how advertising, recommendations from relatives or friends and previous experience at the destination affects both the probability of an overnight stay and the length of stay at the destination, given that these factors minimize the risk of uncertainty. Another prominent issue is that we control from regional differences in preferences for tourism within Asturias depending on geographical characteristics. Our estimations are based on a conditional demand function for tourism using a time series of cross-sectional data. The results indicate that having seen advertising of Asturias increase the probability of stopping over but it has no effect on the length of stay, whereas recommendation from friends or relatives and previous experience both positively affect the decision to sleep at the destination and the number of days.
Keywords: length of stay, tourist´s decision-making, advertising, conditional demand, Negative Binomial model
JEL codes: C35, D12, D81
- Introduction
Length of stay in a tourist destination is one of the most relevant issues in tourist decision-making process(Decrop and Snelders, 2004). The more he stays, the more is the knowledge about products, services and places to visit (Davies and Mangal, 1992; Gokovali et al., 2007; Martínez and Raya, 2008) and, consequently, the more will be the expenditure. In fact, some studies have found evidence about a strong correlation between length of stay and total expenditure (Leones et al., 1998; Agarwal and Yochum, 1999; Thrane, 2002; Laesser and Crouch, 2006; Mehmetoglu, 2007; Fredman, 2008), even though short stays are usually associated with higher levels of average daily expenditure (Cannon and Ford, 2002; Downward and Lumsdom, 2000, 2003; Hsieh et al., 1997, among others). Given the importance for destinations to have long-stay tourism, it seems necessary to identify which factors determine the length of stay.
In this article, we analyze the determinants of length of stay in a particular tourist destination (Asturias) using a two-step count data model. Specifically, our main interest lies in the role that information about the destination plays in tourist decision regarding how long to stay. It is worth to identify how the different sources of information affect length of stay for the proper design of marketing policies as to promote longer stays, associated with higher occupancy rates and revenue streams.
There are several papers in the literature which investigate the effects of sociodemographic characteristics such as age, income or nationality, among others, (Gokovali et al., 2007; Barros and Machado, 2010) on length of stay. Besides, other articles analyze the relationship between the number of days the tourist stays at the destination and the mode of transport, the type of accommodation selected or the purpose of the trip (Martínez-García and Raya, 2008; Alen et al., 2014). In this article, apart from controlling for this sort of factors, our study purpose is to determine how having seen advertisements, previous experience at the destination and recommendation from friends or relatives (word of mouth effect) affect length of stay. Another issue of interest is how duration is connected with the most valued attribute of the destination for each tourist, so that we can identify the most important pull factor, ceteris paribus.
This paper employs a pooled cross-section data for a sample of 33.461 tourists visiting Asturias in the period 2010-2016. For analyzing the effects of the different sources of information on tourist´s length of stay, we estimate a Hurdle Negative Binomial count data model (M ullahy, 1986). This methodology allows us to; firstly, identify the factors which determine the decision to stop over and, afterwards, to analyze the effect of a set of variables on the length of stay for those who actually sleep at least one night at the destination. From a methodological point of view, we employ a Truncated at Zero Negative Binomial P count data model for modelling the positive outcomes in the second step, which is considered as the best alternative (Greene, 2008).
The estimation results show that having seen advertising of Asturias increase the probability of stopping over but it has no effect on the length of stay. Recommendation from friends or relatives and previous experience both positively affect the decision to sleep at the destination and the number of days. Performing some active tourism activities or booking the trip through travel agencies also increase the length of stay. Moreover, the estimations also indicate that the relationship between distance to origin and length of stay is not linear. Foreign visitors tend to stay more than Spaniards whereas education is not significant for explaining the number of days a tourist stays. Climate, natural environment and looking for tranquility emerge as the main pull factors that increase length of stay at Asturias.
The paper is structured as follows. After this introductory section, we present a review of the economic literature regarding this topic. Later on, we present the theoretical model under which the estimations are based. Afterwards, in the fourth section we describe the database and the variables employed. Then, in the fifth section we present the empirical model. The sixth section outlines the main results. Last section concludes.
- Literature Review
The economic relevance of tourism has aroused an increasing interest in analyzing the determinants of the length of stay (LOS) in the economic literature. Several studies have involved mainly descriptive analysis of differences in length of stay given tourist’s socio-demographic and/or trip-related characteristics, including Oppermann (1995, 1997), Seaton and Palmer (1997), Sung et al. (2001) and Lew and McKercher (2002), among others. These studies show how length of stay varies with nationality, age, occupation status, repeat visit behavior, stage in the family life cycle and physical distance between place of origin and destination, among other variables. However, their descriptive nature hinders formal inference tests on the causal relationships between individual socio-demographic profiles and actual trip experiences and length of stay. However, in the last decade, there has been a widespread empirical application of regression models for explaining length of stay, which allow the researcher to study the pure effect of a covariate on the dependent variable, ceteris paribus.
Alegre and Pou (2006) analyze German and Britain tourists’ length of stay at the Balearic Islands (Spain) in terms of a Logit model. Their empirical results show that tourists who are over 60 present a higher probability to stay for longer than the rest of age groups whereas the higher is the education level, the least is length of stay. Gokovali et al. (2007) investigate the determinants of length of stay for a sample of sun and beach tourists visiting Bodrum (Turkey) using survival models. They indicate that Russian tourists tend to stay for the longest duration, followed by tourists from Germany and the Netherlands.Income, party size, previous visits and tourist experience (taste for travelling) positively affect LOS meanwhile daily spending and educational level are negatively related with the probability of staying for longer.
Martínez-García and Raya (2008) model the length of stay for low-cost travelers. Being over 50, having only primary education, travelling in the high peak season and staying in campsites or private accommodation are the factors which further increase the number of days stayed at the destination. Barros et al. (2008) analyze length of stay of Portuguese tourists travelling to South America on charter flights. Controlling for sociodemographic characteristics, budget and temporal constraints, their main interest relies on how brochures and the degree of advanced booking affect length of stay. The econometric estimations prove that the time spam a tourist stays at a destination is positively related to having booked in advance, having seen advertisements, previous visits and the frequency of travel. Barros and Machado (2010)examine the determinants of length of stay in the Madeira Island (Portugal) using a sample of foreign tourists departing from Funchal Airport. The main conclusions of their study are that age, gender, education and hotel quality increase length of stay but expenditure reduces it. Besides, Germans stay longer than British, Dutch and French tourists do. Gomes de Menezes and Moniz (2011) account for the connection between trip experiences and length of stay.Their main interest here is to disclose how travel motive, alternative destinations considered, repeat visitation rate, overall satisfaction and revisit intention condition the stay. Repeat visitors, taking charter flights and those who visit friends or relatives tend to exhibit longer stays.
With the purpose of distinguishing different groups of tourists with homogeneous preferences, Alegre et al. (2011) estimate a latent class count data model for length of stay which endogenously assigns individuals one of the two considered existing classes. Their estimations indicate that age, profession, nationality and the total tourist expenditure are all statistically significant in defining preferences. For both segments, the price per day´s stay has a negative effect on the length of stay, being the magnitude higher for the shorter-stay segment.
Grigolon et al. (2014) develop a dynamic model of choice of length of stay, discriminating among going on holidays for short, medium or long periods. They point out that the effect of a particular vacation length made in the past affects travelers’ choice of a future vacation with the same length.Oliveira-Santos et al. (2015) employ a shared heterogeneity duration model to Brazil visitors‘ length of stay. The estimations indicate that income does not have a significant effect on expected length of stay; tourists visiting two destinations stay shorter than those who visit a single destination; the effect of party size is negative following a non-monotic path; summer season is associated with the longest stays and those who lodge at hotels stay for short periods.
More recently, Thrane (2016) scrutinizes Norwegian students´ length of stay at summer vacation destinations. The novel aspect about this research is that the interest is focused on the differences between those who decide the return date before taking the trip and those who take the decision along the way. “Open-returners” have the longest length of stay, suggesting that “pre-fixed” returners may face more economic and available time constraints. Another salient result is that females tend to stay longer than males.
Finally, Nicolau et al. (2016) conduct a research in which they assess the relation between distance and previous visits and length of stay. They employ a Truncated Negative Binomial model whose estimations indicate that as distance increases, length of stay increases too. This may be caused by the fact that tourists want to compensate for the spent time and costs to reach to the destination and spread the fixed costs.
As for the methodologies employed, different modelling strategies can be identified in the literature: duration models (Gokovali, et al. 2007; Martinez-Garcia and Raya 2008; Gomes de Menezes et al., 2008; Barros et al. 2010; Barros and Machado 2010; Machado, 2010; Peypoch et al. 2012; Thrane 2012; Wang et al., 2012), tobit (Mak and Moncur, 1979; Fleischer and Pizam, 2002), OLS regression (Mak and Nishimura, 1979; Thrane and Farstad, 2012, 2015), panel data (Martínez-Roget and Rodríguez, 2006; Grigolon et al., 2014),ordered logit (Ferrer-Rosell et al., 2014), binomial logit (Alegre and Pou, 2006), multinomial logit (Grigolon et al. 2014), count data models (Hellström, 2006; Brida et al., 2013; Nicolau et al., 2016), latent class (Alegre et al. 2011; Yang and Zhang, 2015) or nested logit model (Nicolau and Más, 2009). We will discuss and justify our count data approach for studying length of stay in section five.
Before ending this section, it is important to note that, as the different empirical studies regarding length of stay at a particular destination that we have presented above refer to different countries, temporal periods and type of tourists, conclusions should be drawn with caution. Nonetheless, a general conclusion is that length of stay at a destination can be explained by opportunity, possibility and preference. In other words, time and economic constraints (daily prices of accommodation, travel costs, income, etc.), travel characteristics (motive of the trip, mode of transportation, who you travel with, party size, etc.), sociodemographic characteristics (age, gender, labor status, etc.) and the level of information about the place emerge as the key determinants. By controlling for the specific circumstances of each tourist, an important source of heterogeneity among them can be accommodated (Heckman, 2001).
- Theoretical framework
Lancaster´s consumer theory (1966) indicates that consumer´s utility is generated by certain attributes or characteristics which the consumption or possession of physical entities produce. However, a tourist (traveler) does not derive utility from possessing or consuming travel destinations but from being in the particular destination for a certain period of time (Rugg, 1983).
The theoretical framework of this paper is based on the discrete choice or random utility models proposed by McFadden (1974) and Manski (1977) – firstly applied to length of stay byAlegre and Pou (2006) – where consumers compare the utility of different alternatives and choose that which maximizes his utility subject to time and economic constraints, bearing in mind that they do not demand quantities but time (length of stay).
Following Dubin and McFadden (1984), tourist´s length of stay is the result of a utility maximization process subject to budget restrictions and time constraints so that:
Max U (q, Z, ttrans, ttur, n, ε)
q, z, ttrans, ttur,
s.a
p´q + ptrans+ p tur t tur≤ Y
t trans + t tur ≤ T
q, z, t trans, p, p trans, p tur ≥0(1)
where q is a vector of consumer goods except tourist ones; Z refers to the characteristics of the trip (accommodation, transport, etc.,); t is the total length of the trip, disaggregated into the necessary travel time for reaching to the destination (ttrans) and the properly length of stay there (ttur); n represents sociodemographic characteristics and preferences of the tourist and ε is a random error term for non-observable factors (McFadden, 1981). Moreover, p is the price vector of goods other than tourist, and ptrans and ptur are the daily prices of transport and accommodation respectively.
The individual chooses a destination j among a choice set S under a utility maximization criterion. As the number of alternatives in the choice set is unlimited, our analysis is conditional on the election of the observable destination j. The information about the tourist good is restricted to the final election, not having information about the alternatives he might have considered. Consequently,assuming that thetourist has previously elected a particular destination j as it provides him the highest level of conditioned utility, the conditional demand function for length of stay at destination j given the characteristics of the trip can be expressed as:
t j-tur= ttur (p, pj-tur, zj, Y – pj-trans, T – tj-trans, n, ε)(2)
This conditional demand function (Pollak, 1969, 1971) allows us to estimate length of stay taking pre-fixed values of the selected destination and trip characteristics so that length of stay (ttur) explicitly depends on these arguments.
Under the assumption that the utility function of the individual is weakly separable[1] the conditional demand function for length of stay can be written as:
t j-tur= ttur (pj-tur, zj, Y – pq– pj-trans, T – tj-trans, n, ε)(3)
The theoretical model presented above states that the tourist´s choice of length of stay at a previously selected destination depends on the trip characteristics, the daily price, income, available time, consumer´s preferences and a random error term. Following this approach and as it is usual in the literature (see Oliveira-Santos et al., 2015 for a wide review of the determinants of length of stay), we can decompose tourist´s preferences among sociodemographic characteristics (Soc), destination attributes (Attrib), and trip-related characteristics (Travel).
Under the assumption that the individual´s utility also depends on the expected quality of the destination chosen, information availability must also be incorporated in the model. When deciding how long to stay – and especially for first-timers –, tourists face a high risk of making a bad decision as the specific characteristics of a destination are unknown until the individual reaches there (i.e., intangibility). Due to this uncertainty, tourist´s choice regarding the number of days to stay strongly depends on the available both external and internal information about the destination and its characteristics (Moutinho 1987). This “experience good” nature of tourism (Mill and Morrison, 2009) induces travelers to carry out extensive information search strategies (Klein, 1998; Roehl and Fesenmaier, 1992). Information searchhas been widely studied in tourism marketing (Fodness and Murray 1997, 1999; Vogt and Fesenmaier 1998, Gursoy and McCleary, 2004). Consequently, the selected LOS for each individual may be strongly related with the awareness of the destination´s characteristics in presence of risk aversion. We distinguish three main sources of information: advertisement, previous experience and worth-of-mouth effect.
Advertisement
Stigler (1961) identified that advertising reduces consumer's search costs as it provides proper, critical and useful information to potential and current consumers. In the same vein, Woodside and Dubelaar (2002) conclude that advertising helps the individual to get positive perceptions of the destination and then increases expenditures there. Nowadays, tourism advertising is regarded as one of the most influential information sources for prospective and current visitors (Gretzel et al., 2000; Woodside and King, 2001; Kim et al., 2005; Park and Nicolau, 2015).