Detecting Free Riders in Collective Brands through a Hierarchical Choice Process

Introduction

A collective brand is defined as a brand created by a public or governmental institution in order to endorse the product quality of the companies concerned, which can sell products with the legal backing and prestige of a specific geographical area and/or production method related to some level of quality (Fernández-Barcala and González-Díaz 2006). According to Vertinsky and Zhou (2000), a collective brand strategy involves producers of the same product category agreeing to use a label as the main brand, which is then tied to a code of practice that certifies the quality standards to be achieved. Only those firms meeting the quality standards can use the collective brand, thus rewarding them for investments contributing to the brand (Rangnekar 2004). Therefore, on the supply side, it allows service providers to have access to a collective reputation. In fact, trade associations very often manage collective reputation to defend their members’ interests with stakeholders such as regulators, employees, suppliers, and the media (Tucker 2008). On the demand side, a collective brand represents a quality signal (collective reputation) solving the problem of information asymmetries. In fact, the adoption of a collective brand is a very relevant marketing and communication tool for clusters since it transmits to the consumer a characteristic common to all the products or services included in the collective brand. This strategy is closely linked to the umbrella place branding, defined asa place brand spanning several functional contexts, thus favoring economies of scale since resources are focused on the determination of a consistent place brand (Therkelsen and Halkier 2008).

However, the reputation effect of the collective brand can be affected by the problem of free riding: If consumer perception of the quality of a collective brand is determined by their experiences of the quality offered by different members of the brand and if high-quality provision requires high investment, a given member will have the incentive to adopt a free-riding attitude with regard to the investments of the other collective brand members (Fishman et al. 2010). In a more succinct and intuitive way, free riding describes the situation where individuals or enterprises achieve benefits that they have not paid for (Lundtorp 2003). In this context, Bellini et al. (2010) claims that there is a clear need for political efforts of mediation, in order to reduce potential conflicts among the diverse stakeholders and harmonize innovative and traditional assets under a collective identity. Interestingly, Lundtorp (2003) detects a lack of academic interest in free riders in tourism, and more recently, Dwyer (2011, p. 245) suggests that this research line needs further exploration. Certainly, those tourism studies that deal with free riders focus mainly on specific facets of destinations, without considering collective brands that include different destinations (e.g., Angella and Go 2009; Hojman and Hiscock 2010). This is even more surprising on account of the growing interest in “interdestination ties” and network connections between destinations (Ness et al. 2013).

What is more, as collective brands base their essence on the transfer of quality perceptions between the individual brands and their collective brands (Wernerfelt 1988), in order to assess the impact of collective branding strategies on consumer choice behavior, Erdem (1998) suggests that researchers should study the process by which consumers draw inferences about these transferences and perceptions.

Accordingly, the objective of this study is to detect free-riding destinations by estimating the contribution of the different entities (individual destinations) that share the use of a collective brand from the point of view of consumer perception. Specifically, to estimate the individual brand’s contribution to the collective brand, and thus fill this gap in the literature, our analysis assumes that a collective brand should come from a range of relationships or associations among brands, in particular, from consumer impressions of the contributions of the individual brands to the attractiveness of the collective brand, in line with the Associative Network Theory (Collins and Loftus 1975) reviewed later in this article and with Keller (2003) who claims that collective brands exist to create a common set of associations. Moreover, we consider that collective brands could be better understood if they were studied through an approach that assesses the incremental effect of the individual brand through the various stages of a hierarchical choice process followed by the tourist, in line with Eymann and Ronning (1992) and Eymann (1995), in which the following choices are nested (not independent): whether to buy, what collective brand to buy, and what individual brand to buy.

In order to fulfill this objective, the remainder of the paper is arranged as follows: the second section reviews free riders in tourism, presents the context of the study, and justifies the estimation of the individual brands’ contribution to the attractiveness of their collective brands, based on a hierarchical choice process. The third section covers the design of the investigation, describing the methodology, sample, and variables used. The fourth section presents the results obtained and their discussion. Finally, the fifth section summarizes the conclusions.

Free Riders, Context of Study, and Decision Process

Research Design

Methodology

In order to detect potential free riders, we follow a three-phase method: (1) we test and compare a two-stage hierarchical choice process of whether to buy and what individual brand to buy and a three-stage process of whether to buy, what collective brand to buy, and what individual brand to buy, using random coefficient multinomial logit models; (2) we then calculate each tourist’s utilities of collective and individual brands to estimate the contribution of the latter on the former by regressing the collective utility over the individual utilities; and (3) within each collective brand, we compare the individual brand’s contributions among them and to an average value to identify potential free riders.

First phase

We test two alternative hierarchical multistage choice processes: First, a two-stage model of the whether to buy and what individual brand to buy decisions. The choice in the first stage is between buying and not buying. People who choose to buy in the first stage go on to a second stage in which they decide among different individual brands. Second, a three-stage model of whether to buy, what collective brand to buy, and what individual brand to buy. The choice in the first stage is between buying and not buying. People who choose to buy in the first stage go on to a second stage in which they decide among different collective brands. Those who choose a collective brand in the second stage go on to a third stage choice among different individual brands of the previously chosen collective brand. A random coefficient multinomial logit model (RCL) estimates the earlier specified two- and three-stage models. Specifically, RCL models are an alternative to the more traditional multinomial logit models due to (Train 2009) (1) their ability to deal with the unobserved heterogeneity of tourists, by assuming that the coefficients of the variables vary among tourists (empirical research shows that heterogeneity in tourism is a fundamental aspect to consider [Barros, Butler, and Correia 2008; Correia, Santos, and Barros 2007]); and (2) their flexibility, which allows representation of different correlation patterns among alternatives, which is all important to this study as it allows us to reflect the decision structure of collective and individual brands.

With regard to the first point, unobserved heterogeneity of tourists in parameter estimations exists in the tourism market and it must be modeled. Hence, the utility of alternative brand i for tourist t is defined as U i t = X i t β t + ε i t , where X_it is a vector that represents the attributes of the brand and the characteristics of tourists; β_t is the vector of coefficients of these attributes of brands and characteristics for each individual t which represent personal tastes; and ε_it is a random term that is iid extreme value. This specification of the RCL model allows coefficients β_t to vary over tourists with density g(β), which means that it differs from the traditional Logit model in which β is fixed. Thus, the nonconditional probability is the integral of P_t(i/β_t) over all the possible values of β_t:

P i = ∫ β t exp { ∑ h = 1 H x i h β t h } ∑ j = 1 J exp { ∑ h = 1 H x j h β t h } g ( β t | θ ) d β t ,

where J is the number of alternatives and g is the density function of β_t, and θ are the parameters of this distribution (mean and variance).

As for the flexibility of the RCL model, it allows us to represent different correlation patterns among nonindependent alternatives. Certainly, apart from avoiding the assumption of Independence from Irrelevant Alternatives (IIA), the fact that RCL models do not have the restrictive substitution patterns of traditional Logit models allows representation of any random utility model (McFadden and Train 2000). In particular, an RCL model can approximate a Nested Logit (NL), which is appropriate for nonindependent and nested choice alternatives. Following Browstone and Train (1999), the RCL model is analogous to an NL model in that it groups the alternatives into nests by including a dummy variable in the utility function that indicates which nest an alternative belongs to. Technically, the presence of a common random parameter for alternatives in the same nest allows us to obtain a covariance matrix with elements distinct from zero outside the diagonal, obtaining a similar correlation pattern to that of an LN model. For the case of beach destinations versus World Heritage destinations, let us assume that the utility function of alternative i is U_it =βx_t + µ_tz_i + ε_it, where µ is a vector of random terms with zero mean and variance σ²_µ,, and ε_it is independently and identically distributed extreme value with variance σ²_ε. The nonobserved random part of the utility is η_i = µ_tz_i + ε_it, which can be correlated with other alternatives depending on the specification of z_i. For example, assume that four alternatives beach destination A, beach destination B, World Heritage destination C, and World Heritage destination D have the following utility functions:

U A t = β x t + μ t + ε A t

U B t = β x t + μ t + ε B t

U C t = β x t + ε C t

U D t = β x t + ε D t

If two alternatives A and B are truly correlated, their covariance is Cov(η_A,η_B) = E(µ_t + ε_At)(µ_t + ε_Bt) = σ²_µ, which permits identification of correlated nonindependent alternatives. Therefore, if the parameter of the variance σ²_µ is significantly different from zero, it implies that the alternatives are correlated and must be “closer to each other” and even at the same level of decision. In the context of this study, it means that the two destinations belong to the same “nest,” i.e., the same collective brand.

A further interpretation can be given to this common random parameter, as it represents the attractiveness of the nest. In fact, according to Train (2009), it plays an analogous role to the inclusive value of McFadden (1981) or category attractiveness of Guadagni and Little (1998) of NL models. In the context of collective brands, this coefficient indicates the utility that the tourist receives no matter which alternative individual brand they choose in the nest; and it represents a measure of the value of the collective brand to the consumer (Kamakura and Russell 1993). In line with these authors, we associate the parameter α_k with nest k in order to reflect the information that describe collective brands, so that these variables differ over collective brands but not over alternatives within each collective brand. Thus, the utility of alternative individual brand i for tourist t is defined as:

U i t = α k t + β i t + χ t p i + δ t m t p i + ε i t ,

where α_kt is the k-nest-specific parameter (collective brand parameter) for tourist t, β_it is the alternative-specific constant (individual i-brand parameter) for tourist t, χ _t is the parameter of the attribute price p_i for each tourist t, which represents personal sensitivity, and δ_t is the parameter of the interaction motivation × price, which indicates the moderating effect of motivations m_t on the influence of price. This approach follows that of Kamakura and Russell (1993) whose choice model includes the constant and situational variables as explanatory variables; analogously, we use these independent variables for illustrative purposes and to show how to detect free riders. Especially important, though, is the constant, as it defines the intrinsic utility of the brand and “represents the value of the brand to the tourist, after the effect of the situational factors” (Kamakura and Russell 1993, p. 12). Therefore, the equation (1) becomes:

P ( i / p i , m t , θ ) = ∫ P ( i / p i , m t , α , β , χ , δ ) g ( α , β , χ , δ / θ ) d β

Note that g(p_i, α, β, χ /θ) is the distribution of the random parameter vector β in the population of all tourists, and θ are the parameters of this distribution (mean and variance). We can derive the distribution h(α, β, χ/i, p_i, θ) of the sensitivities of tourists selecting alternative i, by applying Bayes’ rule:

h ( α , β , χ , δ / i , p i , m t , θ ) ⋅ P ( i / p i , m t , θ ) = P ( i / p i , m t , α , β , χ , δ ) ⋅ g ( α , β , χ , δ / θ )

And rearranging,

h ( α , β , χ , δ / i , p i , m t , θ ) = P ( i / p i , m t , α , β , χ , δ ) g ( α , β , χ , δ / θ ) P ( i / p i , m t , θ )

From this equation, we can obtain the collective brand parameter α_t and the individual brand parameter β_t of tourist t through the expressions

α ¯ t = ∫ α ⋅ h ( α , β , χ , δ / i , p i , m t , θ ) = ∫ α ⋅ P ( i / p i , m t , α , β , χ , δ ) g ( α , β , χ , δ / θ ) d β P ( i / p i , m t , θ ) = ∫ α ⋅ P ( i / p i , m t , α , β , χ , δ ) g ( α , β , χ , δ / θ ) d β ∫ P ( i / p i , m t , β , χ , δ ) g ( α , β , χ , δ / θ ) d β

β ¯ t = ∫ β ⋅ h ( α , β , χ , δ / i , p i , m t , θ ) = ∫ β ⋅ P ( i / p i , m t , α , β , χ , δ ) g ( α , β , χ , δ / θ ) d β P ( i / p i , m t , θ ) = ∫ β ⋅ P ( i / p i , m t , α , β , χ , δ ) g ( α , β , χ , δ / θ ) d β ∫ P ( i / p i , m t , β , χ , δ ) g ( α , β , χ , δ / θ ) d β

Second phase

We estimate the specific contribution of each individual brand to their collective brand by regressing the collective brand utility on each brand utility. The final utility of an alternative depends on the value of both the collective and individual brands. Thus, the decision is influenced by the addition of the two values. The parameter α is an average value of the collective brand; if we add the parameter β, the result would show the intrinsic global value that the tourist gives to the final choice (i.e., β would indicate how higher or lower the individual brand stays compared with the rest of the individual brands within the same collective brand). However, the central question is: how is the collective brand value generated? The value of a collective brand is created by the components in it, so if we estimate the value of its individual brands we can gauge their contributions to the collective brand they belong to. That is, this analysis allows us to know the positive or negative effect of the individual brand value on the average collective brand value. Even though the utility of an alternative is formed by the addition α + β, the individual brand value β exerts an effect on the collective brand value α. The contributions are obtained by estimating the following system of equations with one equation for each collective brand:

α K 1 , t = λ 0 + ∑ i = 1 n 1 λ i β i + μ K 1 ,

α K 2 , t = θ 0 + ∑ i = 1 n 2 θ i β i + μ K 2 ,

where λ_i ∀i∈{1, …, n₁} and θ_i ∀i∈{1, …, n₂} are the contributions of the individual brand utility on the collective brand utility and µ_k1 and µ_k2 are the respective disturbances for each collective brand equation. According to Klapper, Ebling, and Temme (2005), as the variables in the regression are estimated values, we weight them by their standard error.

3rd phase

As β_i would indicate how higher or lower the contribution of an individual brand is than the rest of the individual brands within the same collective brand, we consider the average magnitude of the contribution of individual brands to the collective brand as a free-riding threshold. Specifically, an individual brand below this average contribution would reflect that tourists’ perceptions lead it to be regarded as a “free-rider” in a collective brand.

Results and Discussion

First, we choose the best hierarchical multistage process with the different motivations. The results obtained (see Table 1) show that the likelihood function calculated for nested structure with three stages (first stage: whether to buy [going on holiday]; second stage: what collective brand to buy [Sun, Sea, and Sand vs. World Heritage]; third stage: what individual brand to buy) has a better fit than that of a nested structure with two stages (first stage: whether to buy [going on holiday], second stage: what individual brand to buy). This better fit is significant for all the motivations (see likelihood ratio tests in Table 1). Therefore, this result indicates that the structure to better represent the tourist decision sequence is a nested structure with a first stage in which individuals decide whether or not to go on holiday; a second stage in which those who decide to go on holiday choose between the Sun, Sea, and Sand and World Heritage collective brands; and a third stage that decides the individual brand of the previously chosen collective brand. It supports the idea that the decision process on tourist brands (what to buy) is nested.

OPEN IN VIEWER

Table 1.

Performance of Alternative Hierarchical Choice Processes with Different Motivations (Log Likelihoods).

	Motivation
Structure	Climate	Tranquility	Visiting Friends and Relatives
Whether to buy and what to individual brand to buy	−2857.27	−2909.22	−2930.33
Whether to buy, what collective brand and what to individual brand to buy	−2853.10	−2902.20	−2926.32
Likelihood ratio test	8.32 ***	14.03 *	8.02 ***

*

p < 0.1%; ***p < 5%.

Within this three-stage process, search for climate is identified as the motivation that reaches the highest likelihood function (see Table 1); thus, Table 2 shows the estimates for this sequential model with climate motivation. With regard to the coefficients estimated, it is important to stress that the significance of parameter b indicates the average effect of the dimension analyzed, and that the significance of the parameter of standard deviation SD(β) shows that the effect of this dimension is different for each tourist (which shows the existence of heterogeneity and the superiority of the RCL model over the standard Logit). The results obtained show the following:

OPEN IN VIEWER

Table 2.

Parameter Estimates: Hierarchical Choice Process: Whether to Buy, What Collective Brand, and What Individual Brand to Buy.

Independent Variables	B	Standard Error (b)	SD of β	Standard Error (SD of β)
Whether to buy: decision to go on holidays
Prices	−0.322 *	0.049	0.1661 **	0.053
Prices × Climate motivation	0.530 *	0.069	0.091 *	0.016
What collective brand to buy
Nest “Spain, sun, sea, and sand” collective brand	0.597 *	0.133	0.627 *	0.147
Nest “World Heritage” collective brand	−1.841 *	0.132	2.754 ****	1.567
What individual brand to buy
Constant 1, Costa Blanca	0.980 *	0.170	2.293 **	0.874
Constant 2, Costa Brava	−0.153	0.103	0.773 *	0.202
Constant 3 Costa del Sol	0.126	0.217	1.401 *	0.263
Constant 4, Costa de la Luz	0.027	0.119	0.824 *	0.170
Constant 5, Santiago de Compostela	−0.794 ****	0.424	4.726 *	1.357
Constant 6, Cáceres	0.316 *	0.082	0.578 *	0.169
Constant 7, Ávila	0.285 ***	0.135	1.990 **	0.724
Constant 8, Salamanca	−0.199	0.346	2.298 ****	1.179
Constant 9, Cuenca	−0.165	0.165	2.532 **	0.848
Constant 10, Segovia	−0.003	0.125	0.674 *	0.148
Constant 11, Toledo	−0.385 *	0.113	0.972 **	0.358
Constant 12, Córdoba	−0.407 *	0.128	0.290 *	0.067
Maximum likelihood	–2853.1088

*

p < 0.1%; **p < 1%; ***p < 5%; ****p < 10%.

In the first stage of the choice process (“whether to buy”: decision to go on holiday), price shows a significant and negative parameter. This suggests that tourists tend not to go on holidays with higher prices; in line with Smith (1995). Therefore, price is a dissuasive element. Standard deviation of the coefficient of price is significant, which indicates that its effect is not homogeneous for all individuals. For the interaction price × search for climate, the estimation of the interactive coefficient presents a positive sign, significant at the 0.001 level, which implies that an individual motivated by climate as an important attribute of a holiday is prepared to accept higher prices or, in other words, the negative effect of prices is lower when an individual is searching for climate as an important attribute of their holidays. The interactions price × search for tranquility and price × visiting friend and relatives also show positive and significant results in line with the expectations explained before. For the sake of space, we do not portray the estimates of the models including them, but they are available from the authors on request.

Regarding the nest parameters, a positive coefficient is associated with the Sun, Sea, and Sand collective brand and a negative coefficient with the World Heritage collective brand. As they represent the attractiveness of the nest, their values show the collective brand attractiveness derived from the contribution of the individual brands contained in each nest. These different signs might be explained by individual interests and the level of maturity of each collective brand. As for individual interests, it is important to note that the most sought type of tourism in Spain is beach holidays, which in turn, is in line with the aforementioned motivations of climate and tranquility. As climate and relaxation are the main motivations for holidays (Familitur Reports 1999–2009), beach destinations are generally linked to this interest. Concerning the maturity argument, note that the maturity of the World Heritage collective brand is by no means that of the Sun, Sea, and Sand collective and individual brands: coastal brands were promoted in the sixties while the promotion of World Heritage cities collective and individual brands started in the midnineties. The different degree of maturity affects the manner in which tourists behave, as they do not know what they are going to find with World Heritage brands (at least, to lesser extent) as compared to coastal brands, because the former are not as popular (more than half of Spanish tourists [about 60%] travel to beach destinations (Familitur Reports 1999–2009)).

This pattern is also manifested through the individual brand parameters. Costa Blanca shows a positive utility with respect to the reference alternative “not going on holiday” and the other three brands (Costa Brava, Costa del Sol, and Costa de la Luz) remain neutral (none of them are negative). Regarding the World Heritage destinations, two out of eight (Cáceres and Ávila) present significant and positive utilities, three (Salamanca, Cuenca, and Segovia) are neutral, and three show negative parameters (Santiago, Toledo, and Córdoba). Note that the parameter of standard deviation SD(β) is significant in all of the variables, showing the existence of heterogeneity.

Finally, we estimate the specific contribution of each individual brand to the collective brand utility by regressing each destination collective brand utility on every individual destination brand utility for each individual, that is, the Sun, Sea, and Sand collective brand over Costa Blanca, Costa Brava, Costa del Sol, and Costa de la Luz; and the World Heritage collective brand over Santiago de Compostela, Cáceres, Ávila, Salamanca, Cuenca, Segovia, Toledo, and Córdoba (see Table 3).

OPEN IN VIEWER

Table 3.

Individual Brand Contribution to Collective Brand (Standard Deviation in Parenthesis).

System of Equations
Equation 1Individual Contribution to Sun, Sea, and Sand Collective Brand		Equation 2Individual Contribution to World Heritage Collective Brand
	Coefficients λi		Coefficients θi
Costa Blanca	0.4185* (0.0181)	Santiago de Compostela	0.2510* (0.0157)
Costa Brava	0.3581* (0.0175)	Cáceres	0.2388* (0.0157)
Costa del Sol	0.3515* (0.0173)	Ávila	0.2484* (0.0157)
Costa de la Luz	0.3872* (0.0179)	Salamanca	0.1989* (0.0156)
		Cuenca	0.2153* (0.0157)
		Segovia	0.155* (0.0156)
		Toledo	0.1551* (0.0156)
		Córdoba	0.1015* (0.0155)
Spain, sun, sea, and sand Constant	4.3667* (0.0476)	World Heritage Constant	−3.4967* (0.1630)
Adjusted R-squared	0.3424		0.3341

*

p < 0.1.

Equation (1) shows positive significant coefficients for every individual destination under the Sun, Sea, and Sand collective brand, explaining 34.24% of variation. Furthermore, Costa Blanca contributes the most to this collective brand, followed by Costa de la Luz, Costa Brava, and Costa del Sol. Equation (2) depicts positive significant coefficients for all the destinations under the World Heritage collective brand, with an ability to explain 33.41% of variation. We observe that the ordered contributions are Santiago de Compostela, Ávila, Cáceres, Cuenca, Salamanca, Toledo, Segovia, and Córdoba.

In order to identify potential free-riding behaviors, we can make comparisons between each individual brand’s contribution to its collective brand and the average contribution within this collective brand. This allows the analyst to see which individual brand contributes above or below average (see Table 4). For example, we observe that in the Sun, Sea, and Sand collective brand, every individual brand is around the average contribution (0.3788) except for Costa Blanca, which has a higher value. Although the analysis and the reasons for the position occupied by each brand is beyond the purpose of this article, anecdotally one might consider that the most renowned Spanish beach holiday destination, that is, Benidorm, is located in Costa Blanca.

OPEN IN VIEWER

Table 4.

Differences in Contributions.

Contributions Different from Average
Sun, Sea, and Sand Collective Brand(Average Contribution = 0.3788)			Wald Test	World Heritage Collective Brand(Average Contribution = 0.1955)			Wald Test
Costa Blanca	0.4185		4.77 ***	Santiago de Compostela		0.251	12.37 *
Costa Brava	0.3581		1.38	Cáceres		0.2388	7.62 *
Costa del Sol	0.3515		2.46	Ávila		0.2484	11.35 *
Costa de la Luz	0.3872		0.22	Salamanca		0.1989	0.04
				Cuenca		0.2153	1.59
				Segovia		0.155	6.43 ***
				Toledo		0.1551	6.63 ***
				Córdoba		0.1015	36.46 *
Differences in Individual Contributions Sun, Sea, and Sand (Wald test)
Individual Tests	Costa Brava			Costa del Sol			Costa de a Luz
Costa Blanca	9.59 **			11.82 *			2.63
Costa Brava				0.10			2.25
Costa del Sol							3.26 ****
Global Test: Costa Blanca = Costa Brava = Costa del Sol = Costa de la Luz							15.07 **
Differences in Individual Contributions World Heritage (Wald Test)
Individual Tests	Cáceres	Ávila	Salamanca	Cuenca	Segovia	Toledo	Córdoba
Santiago de Compostela	0.32	0.01	5.88 ***	2.79 ****	19.82 *	20.11 *	48.70 *
Cáceres		0.20	3.50 ****	1.19	14.62 *	15.41 *	39.89 *
Ávila			5.35 ***	2.41	18.13 *	18.56 *	45.31 *
Salamanca				0.56	4.085 ***	4.16 ***	19.75 *
Cuenca					7.595 **	8.12 **	27.92 *
Segovia						0.001	6.26 ***
Toledo							6.04 ***
Global Test: Santiago de Compostela = Cáceres = Ávila = Salamanca = Cuenca = Segovia = Toledo = Córdoba							85.39 *

*

p < 0.1%; **p < 1%; ***p < 5%; ****p < 10%.

Note that we have used the average contribution in this example, but the researcher’s criterion might change according to the circumstances of each situation analysis. The middle and bottom sections in Table 4 show differences in contributions among individual brands within their collective brand so that the analyst can establish a ranking and reach a more refined analysis. Note that as global tests confirm that the contributions of individual brands are significantly different within each collective brand, individual tests permit detection of significant differences among individual brands in a collective brand, thereby allowing for ranking.

Conclusions

Free-riding behaviors exist in tourism and they should be analyzed from a general perspective; while the firms operating in a destination might have incentives to free ride, the destinations themselves might also free ride if we consider collective brands that include different individual destinations. In this line, we have assessed the incremental effect of individual brands to their collective brands through various stages of a hierarchical choice process followed by the tourist (whether to buy, what collective brand to buy, and what individual brand to buy). The empirical analysis carried out on the sample reaches the following conclusions:

The joint modelization reveals the nested and nonindependent character of the tourist decisions of whether to buy and what to buy (both collective and individual brands) and also reveals a multistage nature of the decision-making process, on account of the fact that tourists would first structure various brands into a multilevel hierarchy. The structure that best represents the tourist decision sequence is that with a first stage in which tourists decide whether to buy (whether or not to go on holiday); a second stage in which those who decide to buy (go on holiday) choose between the Sun, Sea, and Sand and World Heritage collective brands and a third stage that decides a individual brand of the previously selected collective brand. The fact that the hierarchical has proven to be superior represents an important theoretical implication: according to the results, people tend to use a hierarchical decision process. Therefore, if the analyst wants to mimic as much as possible the process people follow to make decisions, more than one stage should be considered as it reflects more accurately how the information is processed in their mind. In terms of the estimation of the idiosyncratic utility of each collective and individual brand for each tourist and the posterior regression of the former on the latter, we have been able to evaluate the contribution of each individual brand to the collective brand it belongs to; and in this way we have revealed the importance of the problem of free riding, whose implications for managers and policy makers are as follows:

To conclude, the extent to which the free rider phenomenon represents a problem in tourism is contingent on the region or the country, but it is something not to be override. As tourism policies should include measures to prevent individual destinations from free riding, and since every plan needs evaluation and control, the method presented in this article offers a device to carry out these tasks.

Among the limitations of this study are the following: First is its static character, as it is only based on cross-section data. Alternatively, collective brand value may be inferred from studying tourist switching patterns over time. Second, the field of study is Spain and it would be better if the results were reinforced by applications on other geographical areas in order to be able to generalize the conclusions.

For further research remains the analysis of the factors controlled by a destination decision makers that determine the tourist’s utility; in other words, knowing the utilities—individual by individual—we can observe the effect of specific marketing variables on these utilities. Therefore, we would be able to see the brand-related dimensions that can be manipulated by brand managers in order to have a positive influence on their brand reputation.