Visitors’ heritage location choices in Amsterdam in times of mass tourism: a latent class analysis

ABSTRACT

The popularity of must-see tourism destinations raises concerns about negative impacts including pollution, vandalism, and overcrowding in cultural heritage sites and the surrounding landscape. To address these challenges, understanding visitors’ choice behavior and motivations is essential. This study aims to identify visitor segments based on their location choice and formulate policy recommendations to address overtourism challenges by examining visitors’ choice behavior in a stated choice experiment. It provides evidence-based policy suggestions for mitigating mass tourism effects. The choice experiment was implemented in an online survey which was completed by a sample of 437 individuals who had paid a visit to Amsterdam in recent years. The results shed light on how individuals trade-off location attributes in making a choice between heritage destinations. A latent class analysis revealed three segments which can be labeled as cultural-attraction seekers, selective sightseers, and city-life lovers. The results show that crowdedness and entrance fees for additional experiences significantly influence visitors’ location choices across all segments.

KEYWORDS:

• Choice modeling
• overtourism
• urban heritage
• latent class model
• visitor segmentation

Introduction

Heritage sites are essential parts of the tourism industry in the urban context (Jung & Han, 2014). Urban heritages are a subcategory of touristic places that attract visitors to the unique historical places in urban settings. They contribute to the promotion of cultural tourism defined as a ‘type of tourism activity in which the visitor’s essential motivation is to learn, discover, experience and consume the tangible and intangible cultural attractions/products in a tourism destination. These attractions/products relate to the set of distinctive material, intellectual, spiritual, and emotional features of a society that encompasses arts and architecture, historical and cultural heritage, culinary heritage, literature, music, creative industries and the living cultures with their lifestyles, value systems, beliefs and traditions’ (UNWTO, 2023).

Urban heritage sites and artifacts are valuable sources for cultural tourism in European cities (García-Hernández et al., 2017). Historical cities such as Venice, Barcelona, and Amsterdam have harnessed the inherent value of their irreplaceable mix of tangible and intangible heritage. They have used cultural tourism as a medium to drive economic prosperity. Although tourism activities in historical cities create economic benefits, mass tourism creates problems in cities such as excessive use of public spaces by visitors and deterioration of the urban landscape (García-Hernández et al., 2017), traffic congestion (McCool & Martin, 1994), increasing crime activities (Látková & Vogt, 2012), the destruction of the natural habitat or ecosystem and the degradation of residents’ welfare (Giannoni, 2009; Jurowski & Gursoy, 2004). Moreover, although urban heritages are considered tourist magnets, the popularity of must-see tourism points of interests (i.e. places or objects with heritage value) has led to the deterioration of historic areas.

Overtourism, caused by the excessive visitor numbers in historic areas, has gained attention in recent years due to the rapid growth of tourism and has resulted in the emergence of countermeasures. Although overtourism issues have been discussed in the tourism literature since the mid-1960s (Koens et al., 2018), the growing magnitude of tourism since 2017, and the emerging counter activities such as anti-tourism and tourism phobia have led to more awareness of overtourism (Mihalic, 2020). As examples of measures to tackle overcrowdedness, the Amsterdam municipality introduced some precautions such as recommending less-crowded places and relocating a moveable tourist attraction (‘I Amsterdam’ sign) to less-known places with the aim of distributing people from Central Amsterdam to the vicinity of the city. In Barcelona, the number of visitors to certain attractions like Park Güell’s Monument area is limited, because an entrance fee was introduced for tourists, while residents enjoy free access to this location. In general, city administrators are looking for ways to find a balance between tourism growth, cultural heritage preservation, and the well-being of both visitors and locals.

To address the negative impacts of tourism in historic cities, several actions are usually implemented, including dispersing visitors throughout the city, offering incentives through creating itineraries, crafting city experiences, and engaging in effective communication with local stakeholders (Postma & Schmuecker, 2017), as well as providing tourists with information about less-known attractions (Mitas et al., 2023). Although current measures have shown some level of effectiveness, they can be inadequate in terms of providing a more personalized approach to reduce mass tourism and avoid congestion. This highlights a critical gap in the urban heritage tourism field, and there is still limited understanding of how to tailor these approaches to distinct the specific needs of different visitor segments.

In order to respond to current mass tourism challenges properly, identifying visitor segments can be helpful in attracting certain groups of visitors to areas that align with their interests. The approach of giving personalized feedback and recommendations about possible visitation locations can support distributing visitors’ groups in the city to avoid overcrowdedness at tourist attractions. Therefore, the primary objective of this research is to distinguish visitors’ segments based on their cultural tourism destination choice and to identify the primary characteristics that define these segments. In this research, segmentation refers to the process of clustering a heterogeneous visitor population into specific homogenous groups based on shared characteristics, behaviors, and preferences. It allows us to gain a deeper understanding of different visitor segments to design more tailored policy strategies.

To identify the segments, it is necessary to look into several variables such as personal characteristics (i.e.; age, gender) (Altınay Özdemir, 2022; Dane et al., 2020; Martínez Suárez et al., 2021); visitors’ behavior (i.e. motivations) (Andruliene et al., 2018; Katsikari et al., 2020; Tu, 2020) and individual preferences related to location choices (i.e.; type of attractions; available facilities; accessibility of services) (Bravi & Gasca, 2014; Chiabai et al., 2014; Lupu et al., 2021; Neuts & Vanneste, 2020). Segments can be distinguished based on differences in preferences (tastes) and related to individuals’ personal characteristics and behavior regarding destination choice. For example, segments may include family-oriented visitors looking for high safety, adventure-seekers prioritizing outdoor heritage locations, and cultural attraction seekers favoring immersive culture experiences. Adopting the visitors’ perspective is important for identifying interventions, such as finding potential historical places that could be recommended to visitors as substitution of overly crowded popular locations, to combat overtourism and avoid congestion.

Expanding on the discussion of overcrowdedness and the need for interventions in historical cities for distributing the crowds, it is important to understand the factors that attract and motivate visitors to these destinations. Accordingly, the purpose of the present study is twofold: (i) to identify visitor segments based on their urban heritage destination location choice and (ii) to formulate policy recommendations to mitigate mass tourism and avoid congestion. For these purposes, a survey including a stated choice experiment is conducted to measure visitors’ location preferences. The stated choice experiment method is widely used in tourism research. For instance, Birenboim et al. (2022) generated intervention strategies for sustainable tourism development, Altinay Ozdemir (2022) focused on Turkish travelers’ preferences for destination choice, Neuts and Vanneste (2020) analyzed urban redevelopment in tourism destinations and Lupu et al. (2021) explored information-seeking and variety-seeking behavior of tourists considering their choice of heritage attractions. The results support possible suggestions to mitigate the negative effects of mass tourism in historical cities.

The remainder of this paper is structured as follows. The next section reviews related work focused on personal characteristics, visitors’ behavior, and individual preferences related to location choice in heritage tourism. The methodology section focuses on the experiment design and research method used in this study. This is followed by a discussion of the results and policy recommendations. Finally, a discussion of the conclusions, remaining problems for future research, and limitations completes the paper.

Related work

Our paper focuses on uncovering individual preferences that influence urban heritage destination location choices and identifying visitor segments. Understanding visitors’ behavior and motivations for destination choice enables segmentation based on shared aspects including personal characteristics (Altınay Özdemir, 2022; Dane et al., 2020; Martínez Suárez et al., 2021), visitors’ behavior (Andruliene et al., 2018; Katsikari et al., 2020; Tu, 2020), and individual preferences related to location (Bravi & Gasca, 2014; Neuts & Vanneste, 2020). This section provides a literature review of these aspects.

Personal characteristics

Sociodemographic variables play an important role in explaining individual differences in visitor behavior and preferences. These variables including age, gender, education, income, and occupation are fundamental components for gaining understanding about visitors. More homogeneous segments can often be identified by defining target groups that refer to a specific subset of visitors sharing certain sociodemographic characteristic (Altınay Özdemir, 2022). For instance, in the study of Martines Suarez (2021), significant differences were found between visitor groups, especially based on their socio-demographic characteristics. It was found that older adults who intend to visit touristic areas plan to incur less expenses compared to other age groups. Determining and understanding the differences between visitor groups based on their socio-demographic characteristics can thus offer a basis for providing more tailored recommendations for visitors.

Visitors’ motivation

Travel motivations have been generally studied in the context of specific heritage settings such as museums, religious sites, and waterfront sites (Poria et al., 2004; Su et al., 2020). Su et al. (2020), describe heritage travel motivation based on three dimensions: emotional experiences, recreational experiences, and educational-cultural experiences. Common motivations for heritage visits are related to recreational experiences such as being in a calm atmosphere, having a good time with a friend, and relaxation contributing to a sense of leisure and well-being during heritage visits. Emotional connections as highlighted by Poria et al. (2004), can contribute to the sense of belonging and attachment to a heritage. Furthermore, educational and cultural experiences are related to learning about the physical nature, history of the heritage and its surroundings, and authentic experiences (Kerstetter et al., 2001).

Previous studies have used different methods and data to measure visitors’ travel motivations. For instance, Su et al. (2020) found that heritage travel motivation is related to visitor engagement, visitor experience, and heritage destination image, based on an empirical study using structural equation modeling on data from a large sample of visitors. Oh et al. (1995) conducted a correlation analysis of 30 factors including reasons related to knowledge and intellectual interests, kinship and social interaction, novelty and adventure, entertainment and prestige, sports, escape and rest. It was found that destination choice is motivated by visitor experience and emotions such as safety, cleanliness and comfort-related destination attributes.

The study by Tu (2020), focused on heritage tourism motivations to understand what encourages visitors to travel. The author identified 18 items related to heritage tourism and applied exploratory factor analysis. The initial exploratory factor analysis revealed three primary factors derived from 18 items related to heritage tourism. Upon further analysis, the final selection narrowed down to eight items, which were categorized into two distinct and themes as recreational benefits and long-term values. Recreational benefits include novel and fun experiences, and social and emotional benefits. In addition, cultural inheritance was included within this group since heritage site visits might be related to emotional bonding. The long-term value consisted of health, the sense of achievement, a good life, and work motivation.

Based on the above findings, we can conclude that visitors’ travel motivations can be categorized into four main groups. First, there is an emphasis on knowledge acquisition and cultural learning, aligning with the educational-cultural experiences highlighted in previous studies. Second, the importance of personal bonds and sense of belonging emerges as other motivators, reflecting the emotional experiences described by Poria et al. (2004). Third, the appeal of authenticity and popularity is central in travel motivations, as indicated by Su et al. (2020). Lastly, the desire for pleasure and entertainment, characterized by novel and fun experiences, emerges as a key factor which is also consistent with the recreational experiences.

Individual preferences related to location choice

In this section, previous studies that focused on visitors’ location choices in the domain of heritage tourism are reviewed. Existing studies have focused on various topics including economic theories to explain visitors’ location choices (Bravi & Gasca, 2014), residents’ preferences related to urban development plans in a touristic historical city (Neuts & Vanneste, 2020), visitors’ choices for and interests in coastal tourism (Lacher et al., 2013), the design of a travel recommender system that take into account visitor preferences (Arentze et al., 2018), and demands and needs of different groups of off-season visitors in mass tourism destinations (Figini & Vici, 2012).

Individual preferences considered in the literature related to heritage location choice can be grouped into the following six main categories of location attributes: place-based attributes, supporting products, monetary costs, physical accessibility, crowdedness, and time-related attributes. An overview of attributes that are extracted from this literature is given in Appendix A.

Relevant place-based attributes of locations may include layout (i.e. street layout (Neuts & Vanneste, 2020)), accommodation availability and price (Birenboim et al., 2022), esthetic design (Ferretti & Gandino, 2018), nature (i.e. presence of green areas (Neuts & Vanneste, 2020), environmental quality (i.e. footfall (Lupu et al., 2021), litter level (Birenboim et al., 2022)), heritage type (Lupu et al., 2021), and attraction value (Arentze et al., 2018)). These attributes are used in these studies to measure individuals’ preferences of destination-related elements that can explain their location choice.

Supporting products include attributes such as the quality of shops (Neuts & Vanneste, 2020), opening hours of shops (Figini & Vici, 2012), dining facilities (i.e. food and wine (Bravi & Gasca, 2014), presence of nearby bar/café/restaurant facilities (Dane et al., 2019), restaurant quality (Lacher et al., 2013)), hospitality package (i.e. accommodation and food, and optional entertainment (Altınay Özdemir, 2022)), contextual activities related to a desire for novel experiences and/or entertainment (Arentze et al., 2018), and availability of activities (Lacher et al., 2013). In general, the attractiveness of heritage sites is influenced by factors such as the availability of supporting products like accommodation, souvenir shops, and restaurants. In addition, supporting products can be offered in the context of experience-related activities such as tasting special wines (Capitello et al., 2017).

Monetary costs are considered in relation to individuals’ preferences for saving on expenditures in participating in some activities during the touristic visit. Cost factors considered are entrance fees (Alexandros & Jaffry, 2005; Apostolakis & Jaffry, 2005a, 2005b; Capitello et al., 2017; Ferretti & Gandino, 2018), trip costs (Lacher et al., 2013), and activity costs (Arentze et al., 2018).

Physical accessibility concerns the ease with which a destination can be reached, and the availability of destination-related information. Accessibility attributes that have been considered are, for instance, supported modes of transportation (i.e. public transportation (Neuts & Vanneste, 2020), bike sharing (Bravi & Gasca, 2014)), availability of pre-trip information (i.e. GPS based journey planner, booking planner, e-forum and virtual tour) (Chiabai et al., 2014), presenting information using simple text on a card or use of Audio/Video materials (Alexandros & Jaffry, 2005), and travel time between origin and destination (Arentze et al., 2018; Lupu et al., 2021).

Crowdedness represents crowd-related and congestion-based attributes. Crowdedness emerging from overtourism has an impact on destination choice. Crowdedness considerations include concerns such as crowdedness at the street level (Birenboim et al., 2022), positive crowd utility value (Neuts & Vanneste, 2020), and congestion level (Alexandros & Jaffry, 2005; Apostolakis & Jaffry, 2005a, 2005b). Moreover, time-related attributes include season (Altınay Özdemir, 2022), the duration of the visit (Dane et al., 2020; Lupu et al., 2021), and arrival time.

The review of previous studies on visitors’ location choices in heritage tourism indicates several factors that influence individuals’ decisions. These studies utilize place-based characteristics to measure individuals’ inclinations toward various destination-related aspects including heritage category, attraction value, and overall environmental quality which can elucidate individuals’ location choice. Moreover, heritage attractiveness is affected by the presence of supporting products (i.e. shops). Monetary costs are weighted concerning individuals’ preferences for economizing on costs while engaging in certain activities such as guided tours during their tourist visit. Accessibility explains the ease of reaching a tourist destination and the accessibility of pertinent information about the destination. Congestion-related attributes including crowd density and congestion levels are also taken into account because crowdedness stemming from overtourism impacts individuals’ location choices. Understanding these factors is important to interpret what attracts visitors to urban heritage locations.

Methodology

In this part, we explain the experiment design, survey design, and model estimation method used to reveal individual preferences and identify visitor segments in urban heritage destinations.

Experiment design

For this study, the stated choice experiment (SCE) was implemented in a survey to measure visitors’ preferences for attributes of heritage locations and their surroundings. In a SCE, participants are presented hypothetical alternatives, in this case, heritage locations and their characteristics, and asked to indicate their choice. By varying the attributes of alternatives presented independently of each other, individuals’ preferences for varying characteristics of a heritage destination and its surroundings can be identified by statistical analysis of the choice data thus obtained (Arentze et al., 2018; Kemperman, 2021; Kemperman et al., 2019).

Based on findings from the literature reviewed above, eight different attributes were selected and included in the experiment. All attributes have three levels. The attributes and levels are represented in Table 1. The chosen attributes in this study capture various aspects of heritage locations and visitor experiences. These include heritage category, historical urban landscape value, entrance fee, the availability of pre-visit information, the availability of other heritages and facilities within walking distance, perceived attractiveness by other visitors, overall evaluation of other visitors, and perceived average crowdedness level by other visitors. The attributes and levels were selected to comprehensively analyze visitor preferences and their impact on heritage location choice.

Table 1. Attributes and levels.

Attribute	Description	Level
Heritage category	It refers to historical buildings such as churches, museums, restaurants, shops, and transportation buildings; historical gardens such as parks, zoos, and historical squares in urban areas.	Commercial or governmental heritage buildings and areas (governmental buildings, industrial buildings, shopping facilities, storage, transportation buildings, and office buildings) Cultural or educational heritage buildings and areas (church, museum, cultural buildings, and education buildings) Recreational heritage buildings and areas (caterings, garden, park)
Historic Urban Landscape Value (HUL)	HUL value can be defined as an approach that includes the historic character of the heritage locations’ environment. Heritages are evaluated not only for the existence of sole value, also for their spatial settings including historic layering of cultural and natural values (Rey-Pérez & Pereira Roders, 2020).	Architecture (the area is characterized by buildings with special architectural value) Urban (the area is characterized by historical buildings and public spaces such as plazas and streets) Nature (the area is characterized by natural elements such as gardens with historical value)
Entrance fee	It refers to the fact that some heritage buildings provide a guided tour with an entrance fee such as in Rijksmuseum and Heineken Experience. It is assumed that this fee includes drink, photograph service or experience.	0 Euro (Free) 20 Euro 40 Euro
Availability of pre-visit information	It indicates that information may be available in advance to check the schedule (i.e. opening time, activities), the location (map-based info), and booking info.	A mobile application that can be downloaded on smart phone A website where the information can be found There is no available information
Other heritages and facilities within walking distance	It refers to other possibilities that visitors might have taken advantage of while they are performing touristic activities.	Other heritages Hotel-restaurant-café and shopping areas No available heritages or other facilities
Perceived heritage attractiveness by other visitors	It refers to the degree of heritage attractiveness value that is assigned by other visitors.	3-star (very good) 2-star (good) 1-star (average)
Overall evaluation of other visitors	It means what other visitors think of the given heritage location and its surroundings with regard to safety, cleanliness and maintenance	Very good Good Average
Perceived average crowdedness level by other visitors	It refers to the crowdedness that can be described as too many people within limited spaces. The attribute means how other visitors evaluated crowdedness of the place.	Not crowded Moderate crowd Very crowded

Given the number of eight attributes all with three levels, a full-factorial design consists of $3^8$ ( =  6.561) possible attribute profiles. To reduce the size of the design, a more efficient fractional-factorial design was used which includes 27 profiles. This fraction is an orthogonal design with balanced attribute levels that allows the identification of the main effects of the attributes. Choice sets were created by randomly drawing two attribute profiles from this design. In addition, each choice set includes a null alternative ‘none of these options.’¹

Figure 1 shows an example of a choice set. The description of attributes and levels that are presented in Table 1, was also shown for explaining the attributes and levels to the respondents in the questionnaire. Thus, respondents were able to see the same description before starting the Stated Choice Experiment. In addition, the text ‘Imagine that you would visit heritage in Amsterdam. It is expected that the congestion level will be high in the city center (within UNESCO boundary) due to tourist arrivals during the time you will visit. Keeping this in mind, which of the below heritage buildings or site options would you prefer to visit? You can also choose “none of these” if you find none of the two options attractive.’ was presented in a choice set example that respondents received before starting the experiment (Figure 1). The respondents were asked to indicate their choices considering the given attributes of each location alternative. Each respondent received 9 such choice tasks.

Figure 1. Choice set example.

Survey design

The survey consists of four sections, including the SCE. First, questions about socio-demographic characteristics were asked to capture the distribution of respondents on personal background variables. Second, questions about respondents’ last visit to Amsterdam with an urban heritage tourism purpose were queried. An urban heritage tourism purpose refers to visiting cities for sightseeing (i.e. architecture, monuments, parks) and visiting cultural amenities (i.e. museums), possibly combined with enjoying entertainment facilities. In this part, the travel season, travel party size, initial mode used to travel to Amsterdam, and the travel mode used within Amsterdam were asked for the last trip made. Also, the respondents’ awareness of heritage was measured to understand how much they value heritage for a touristic visit. This question was used to assess whether visitors are intentionally seeking heritage sites or, at least, recognized the heritage value of places during their visit. Third, four groups of statements regarding visitors’ motivation that are related to different types of benefits that heritage visits could offer, were presented. The four groups were labeled as: knowledge and learning, personal bond, popularity, and pleasure and entertainment. In the context of these groups, a total of 22 statements were presented to respondents and they were asked to rate them on a five-point Likert Scale ranging from 1 (strongly disagree) to 5 (strongly agree). This approach aimed to gain insights into respondents’ attitudes and motivations. Finally, the fourth part of the survey contains the SCE experiment that was explained in the previous section.

Model estimation

The SCE is based on random utility theory which states that individuals make choices that maximize their utility given their preferences for the attributes. The Multinomial Logit (MNL) model is the most commonly applied framework to estimate the utility function. In this model, the utility function is defined as (Hensher et al., 2015): (1) $U_i=V_i+{\epsilon }_i$(1) where $U_i=$ Utility of alternative $i$, $V_i=$ Structural utility component of alternative $i$, and ${\epsilon }_i=$Random utility component of alternative $i$.

The structural utility component is defined as: (2) $V_i={\mathrm{\Sigma }}_n{\beta }_n{X}_{\mathrm{in}}$(2) where ${\beta }_n=$Utility parameter of attribute $n$ and $X_{\mathrm{in}}=$Value of alternative $i$ on attribute $n$.

The MNL model defines the probability that an individual chooses alternative i in a given choice set as: (3) $P_i={\displaystyle \frac{\exp (V_i)}{{\mathrm{\Sigma }}_{i^{\mathrm{\prime }}\in S}\exp (V_{i^{\mathrm{\prime }}})}}$(3) where $P_i=$Probability that alternative $i$ is chosen and $S=$ Choice set of alternatives.

For estimating the utility parameters, ${\beta }_n$, we used effect coding of attribute levels where consistently the highest level was taken as a base. Using effect coding, the two binary variables used to define each 3-level attribute are coded as; level 1: [1,0], level 2: [0,1], and level 3: [−1, −1].

To take possible heterogeneity into account and identify classes, the Latent Class (LC) model is used. In the LC model, it is assumed that individuals are grouped into a set of $K$ classes. Each class (i.e. segment) has one set of parameters. Thus, the structural utility is defined as: (4) $V_{\mathrm{ic}}={\mathrm{\Sigma }}_n{\beta }_{\mathrm{nc}}{X}_{\mathrm{in}}$(4) where ${\beta }_{\mathrm{nc}}=$Utility parameter of attribute $n$ for class $c$, $X_{\mathrm{in}}=$ Value of alternative $i$ on attribute $n$, and $c_{}=1,2,\dots K$.

The choice probabilities are then defined as: (5) $P_{\mathrm{ic}}={\displaystyle \frac{\exp (V_{\mathrm{ic}})}{{\mathrm{\Sigma }}_{i^{\mathrm{\prime }}\in S}\exp (V_{i^{\mathrm{\prime }}c})}}$(5) $P_{\mathrm{iqc}}=$ Probability that an individual of class $c$ choices alternative $i$

In a latent-class analysis, the class membership of each individual is estimated simultaneously with the utility parameter values within each class. The number of classes $K$ is pre-defined. To find the optimal value of $K$ the minimum Akaike Information Criterion (AIC) is used (Kamakura & Russell, 1989).

After identifying the classes, a Multinomial Logistic Regression analysis is conducted to predict the class of an individual based on sociodemographic characteristics, choices made in the last visit to Amsterdam, and benefits sought in visiting heritage. A class membership function can be estimated simultaneously with the estimation of preference values. However, the estimation did not converge in our case. For that reason, we chose to use the two-step approach which provides more robust results in this case.

Results

In this section, we explain the data collection procedure, the sample characteristics, descriptive statistics of respondents’ attitudes and motivations related to previous urban heritage visits and the results of the stated choice experiment.

Data collection and sample

The survey was applied via an online platform called LimeSurvey. Invitations to participate in the survey were sent to a random sample of an existing national panel in December 2022 in the Netherlands. The survey was examined by the ethical committee at Eindhoven University of Technology (ERB2022BE040) and approved for data collection. Only respondents who have visited Amsterdam in the last five years were allowed to proceed to the survey. The survey underwent a preliminary evaluation through a soft launch involving 36 participants. The outcomes of the soft launch were positive, with panel users expressing that the survey was meaningful and easily interpretable. Responses from a total of 546 respondents were collected. Unfinished surveys were excluded from the analysis. Furthermore, respondents who had completed the questionnaire in less than five minutes were excluded assuming that it was not possible to finish the survey within that short time. After cleaning, the final sample includes 437 respondents.

Descriptive statistics

Sample characteristics

Table 2 presents some descriptive statistics of the sample. The gender distribution of respondents is balanced, and the majority of the respondents are between 18 and 34 years old, have university of applied science degree, have a net income of 20.000€ – 50.000€ in a year and have a full-time job.

Table 2. Sample characteristics.

Variable	Category	Sample percentage
Gender	Male	49.2
	Female	50.8
Age	18–34 year	36.6
	35–54 year	35.9
	55 year or older	27.5
Education	Primary education	29.5
	Basic education	10.5
	Secondary education	12.0
	University of applied sciences	32.7
	University/scientific education	15.3
Income	<20.000€	11.5
	20.000€ –50.000€	47.8
	>50.000€	31.1
	Prefer not to answer	9.6
Occupation	Full-time (>32 hours)	51.9
	No paid employment	5.5
	Part-time (<32 hours)	18.3
	Retired	12.4
	Student	6.9
	Unemployed	5.0

Characteristics of the last visit to Amsterdam

Table 3 shows descriptive statistics of trips related to respondents’ last visit in Amsterdam within the past five years. 73% of the respondents visited the city for a one-day trip and 27% made a multiple-day trip. Respondents’ main purpose was a city trip (81.2%). Most respondents traveled together with others (87.9%) and mainly with family (only adults, 51.5%). The other travel parties were family members including children (22.4%), colleagues (5.5%), friends (28.6%), or others (3%). Most respondents traveled together with two people including themselves (52.2%).

Table 3. The distribution of the last visit.

Variable	Category	Sample percentage
Number of day(s)	Multiple days	27.0
	One-day	73.0
Purpose	City trip	81.2
	Other	18.8
Travel party	Yes	87.9
Family only adult	Yes	51.5
Family including children	Yes	22.4
Colleagues	Yes	5.5
Friends	Yes	28.6
Others	Yes	3.0
Number of people (including respondent)	Alone	7.8
	2	52.2
	3	16.9
	4+	23.1
Travel days	Full week	4.6
	Weekdays	43.5
	Weekends	51.9
Travel season	Fall	19.4
	Spring	29.1
	Summer	42.8
	Winter	8.7
Initial travel mode to Amsterdam	Car	54.9
	Other	2.3
	Public transportation	42.8
Travel mode within Amsterdam	Car	13.0
	Other	3.4
	Public transportation	27.5
	Walking	56.1
Heritage awareness	Didn't visit (no aware)	27.4
	No (aware but did not visit)	24.3
	Yes (aware and visit)	48.3

As for travel days, respondents are distributed quite evenly across weekends (51.9%) and weekdays (43.5%). In terms of season, a considerable number of respondents chose summer (42.8%), and spring (29.1%). This might be affected by the time (December) when the survey was conducted since we asked respondents about their last visit. Most people indicate that car (54.9%) or public transportation (42.9%) is used for the trip to Amsterdam. For traveling within Amsterdam, walking is chosen most often (56.1%) followed by public transport (27.5%). 48.3% of respondents indicate that they visit heritage and are aware of its heritage value.

Respondents’ attitudes and motivations toward heritage visit

The sum scores of four groups of statements are used as a measure of the sought benefits of visiting heritage. Responses were coded as 1 = strongly disagree and 5 = strongly agree. Cronbach’s alpha analysis is conducted to assess the internal consistency of the measures. Results of the analysis are shown in Table 4. For all groups, Cronbach’s alpha values indicate sufficient internal consistency (Cronbach’s α > 0.7). Therefore, all statements are kept and sum scores per group are used in the further analyses.

Table 4. Statistics of the attitude items scores.

By visiting heritage, I can benefit from	Mean	SD	Cronbach’s α
Knowledge and learning (Su et al., 2020; Tu, 2020)			0.80
… learning about different cultural inheritance	3.83	0.74
… sense of achievement	3.16	0.88
… self-growth	3.76	0.79
… learning about the site	3.94	0.65
… learning about historical background	3.98	0.70
… learning about physical nature of the site	3.81	0.69
Personal bond (Su et al., 2020; Tu, 2020)			0.81
… histories that is precious for me	3.68	0.79
… knowledge to understand part of my own heritage	3.43	0.89
… feeling the sense of belonging to the site	3.46	0.85
… a place for satisfying spiritual needs	2.76	1.02
… visiting places that I want to feel emotionally involved	3.20	0.90
Popularity (Li & Lo, 2004; Su et al., 2020)			0.79
… places that I feel obliged to visit heritage building/sites during my touristic visit	2.93	1.01
… places that I worry about missing out	2.95	1.00
… world-famous sites that I should see once in my life	3.52	0.92
Pleasure and entertainment (Su et al., 2020; Tu, 2020)			0.84
… a novel and fun experience	4.00	0.69
… travel experiences to share with family and friends	3.89	0.78
… interaction with family and friends	3.62	0.84
… positive mood improvement	3.56	0.80
… subjective wellbeing improvement	3.53	0.82
… fun and learning experience	4.00	0.68
… a day out	4.21	0.69
… some entertainment	3.95	0.70

As the statistics in Table 4 show, the benefits respondents mostly seek in cultural heritage trips within each group include learning about the historical background (knowledge and learning), histories that are precious to them (personal bond), world-famous sites that they should see once in their life (popularity) and, a day out by visiting heritage (pleasure and entertainment). In comparison to existing studies, the findings align with the study conducted by Poria et al (2004) which identified similar reasons for cultural heritage trips including having a day out and entertainment, wanting to see a world-famous site, and learning about the physical nature of the site. Additionally, the results correspond with the study by Oh et al. (1995) which highlighted fun and entertainment were reasons for heritage visits.

Stated choice experiment

Of the total number of choice observations (437 respondents’ times 9 choices), the null alternative (the ‘none of these’ option) was chosen 14.2% of the times. Hence, most of the time respondents chose to visit a location from the options offered to them.

Based on the SCE data, the LC model is estimated using NLOGIT Version 6. To find the optimal number of classes K, an LC model is estimated for two, three, four, and five classes. Table 5 shows the goodness-of-fit statistics for the five models. The three-classes model achieves a considerable improvement in fit compared to the $K=1$ (MNL) and $K=2$ models. Although four- and five-classes models give a further improved AIC value, these models are not considered useful as they include classes that are too small for reliable estimation of the parameters within the classes. Hence, the three-classes model was chosen. An adjusted Rho-square of 0.28 indicates a satisfactory fit of the model.

Table 5. Model comparison.

LC model statistics	1 class	2 classes	3 classes	4 classes	5 classes
Log-likelihood of the estimated parameters $(\mathrm{LL\beta })$	−3418.30	−3173.37	−3080.14	−3043.74	−3023.18
Log-likelihood of the null model ($\mathrm{LL}0$)	−4320.84	−4320.84	−4320.84	−4320.84	−4320.84
McFadden's Rho-squared (${\rho }^2$)	0.21	0.27	0.29	0.30	0.30
Adjusted Rho-square	0.21	0.26	0.28	0.28	0.28
Number of observations	3933	3933	3933	3933	3933
Akaike Information Criterion (AIC)	6870.60	6416.70	6266.30	6229.50	6224.40
Bayesian Information Criterion (BIC)	−66617.68	−144882.66	−222844.16	−300692.16	−378504.40
Number of parameters	17	35	53	71	89

Table 6 represents the detailed estimation results for the base MNL model and the three-classes LC model. The MNL estimation results show average preference values across all segments. In the MNL model, a significant value of the constant shows that there is a positive base preference for a heritage location (compared to the null alternative). Looking at the significance of the attributes, only the availability of pre-visit information did not have significant effects. Considering the utility values for the attribute levels, commercial or governmental categories of heritage are less preferred compared to recreational heritage. On average, individuals prefer to visit a location that is in an area characterized by natural elements such as gardens with historical value, with no (0€) entrance fee, with web-based pre-visit information, the availability of hotel-restaurant-café and shopping facilities within walking distance of the heritage location, a 3-star location, a very good overall evaluation by others, and not crowded heritage areas.

Table 6. Estimation results for the MNL and LC models.

		Base model (MNL)		3-classes
				LC1		LC2		LC3
Attributes	Level	Utility Value	sign.	Utility Value	sign.	Utility Value	sign.	Utility Value	sign.
Constant	Base Location alternative	0.579	***	1.240	***	−0.266	***	1.459	***
	Null alternative	−0.579		−1.240		0.266		−1.459
Heritage category	Commercial or governmental	−0.225	***	−0.242	***	−0.339	***	−0.237
	Cultural or educational	0.099	***	0.111	**	0.122		0.195
	Recreational	0.127		0.131		0.217		0.041
Historic Urban Landscape Value (HUL)	Architecture	−0.099	***	−0.094	**	−0.158	**	0.055
	Urban	0.039		0.093	*	0.099		−0.302	*
	Nature	0.060		0.001		0.059		0.248
Entrance fee	0€	0.722	***	0.156	**	1.493	***	2.481	***
	20€	0.032		0.133	***	−0.016		0.103
	40€	−0.754		−0.289		−1.477		−2.584
Availability of pre-visit information	Mobile application	0.010		−0.029		0.078		0.134
	Website	0.031		0.077	*	0.063		−0.141
	No available information	−0.041		−0.048		−0.140		0.007
Other heritages and facilities within walking distance	Other heritages	−0.014		0.005		0.101		−0.167
	Hotel, restaurant, cafe and shops	0.152	***	0.111	**	0.132		0.599	***
	Not available	−0.138		−0.116		−0.232		−0.432
Perceived heritage attractiveness by other visitors	3-star	0.185	***	0.203	***	0.271	***	−0.004
	2-star	0.070	**	0.044		0.317	***	−0.147
	1-star	−0.255		−0.247		−0.588		0.152
Overall evaluation of other visitors	Very good	0.074	**	0.079	*	0.095		0.244	*
	Good	0.027		0.013		0.085		0.107
	Average	−0.101		−0.092		−0.180		−0.351
Perceived average crowdedness level by other visitors	Not crowded	0.208	***	0.155	***	0.412	***	0.177
	Moderate crowded	0.108	***	0.117	**	0.261	***	−0.022
	Very crowded	−0.315		−0.272		−0.673		−0.155

***, **, * Significant at 1%, 5%, 10% level; bold indicates reference level.

The size of the utility difference between the most and least preferred level indicates the attribute importance in the choice of a location (see Figure 2). The importance of an attribute depends on the range used. To indicate preferences displayed by the classes, the relative importance assigned to attributes in each class, as derived from the utility estimates is shown in Table 6. Regarding entrance fee, it should be noted that the range include ‘no fee.’ This means that there are qualitative differences between the levels of this attribute which may explain the relatively large importance value.

Figure 2. Attribute importance in case of the LC model.

Table 6 also presents the estimation results of the three-classes LC model. The membership probabilities for the classes are 49%, 27%, and 24% respectively. A first observation is that individuals in class 1 and class 3 display a base preference for the given location options (accepters), whereas class 2 individuals have a base preference for not choosing a given location (rejecters). Considering the values of the attribute preferences, the classes can be characterized as follows.

In LC1, all attributes have a significant effect on utility. Individuals of this class prefer to visit recreational or cultural heritage over commercial heritage. The utility of HUL is found significant but relatively small, the environment that is characterized by the presence of historical buildings and public spaces is preferred over a sole historical building. Only individuals of this class are relatively insensitive for paying entrance fee and are indifferent between paying the lower entrance fee of 20€ and paying no fee. LC1 is the only class in which the availability of pre-visit information is significant with a positive value assigned to website information. Furthermore, the availability of hotel-restaurant-café and shops within walking distance, highly attractive locations with 3 stars, safe, clean, and well-maintained locations, and no-crowd or moderate-crowd locations are preferred by these individuals.

In LC2, heritage category, HUL, entrance fee, perceived heritage attractiveness, and perceived average crowdedness attributes have significant effects on utility. Individuals prefer to visit heritage in an urban landscape as HUL category. Regarding the entrance fee, individuals of LC2 are more sensitive to price compared to LC1 and have a stronger preference for no entrance fee. It turns out that the entrance fee is the most important attribute for these individuals, as can be seen in Figure 2. The presence of other heritage within walking distance does not influence the utility of heritage locations in this class. However, individuals of this class do prefer to visit attractive and highly attractive locations. Most importantly, they prefer no-crowd or moderate-crowd locations. For LC2 individuals, perceived crowdedness level by other visitors is the second most important attribute after the entrance fee (see Figure 2).

In LC3, HUL, entrance fee, other heritages and facilities within walking distance, and overall evaluation of attractiveness by other visitors and crowdedness have significant effects on utility. Individuals of this class prefer to visit heritages that are located in a nature area (i.e. area characterized by natural elements such as gardens with historical value). This class is most sensitive to price; the utility range for entrance fee has the highest value among all classes. The availability of hotel-restaurant-cafe and shops within walking distance is strongly preferred. Moreover, they prefer highly safe, clean, and well-maintained locations based on the evaluation of other visitors.

As shown in Figure 2, the entrance fee is the most important determinant of location choice for LC2 and LC3. For LC1 heritage attractiveness, entrance fee, crowdedness level, and heritage category have approximately the same importance in the choice of location. For LC2, crowdedness level and heritage attractiveness have nearly the same importance. The availability of pre-visit information, HUL, and overall evaluation are found less important than the other attributes. For LC3, heritages and other facilities within walking distance and overall evaluation have an impact on individuals’ location choice. On the other hand, the availability of pre-visit information, heritage attractiveness, and crowdedness level do not have a significant influence on their choice.

In sum, the three classes found in this analysis show different preference patterns. Class 1 (LC1) members can be described as people who prefer to visit cultural attractions but wish to avoid crowds. Hence, these respondents can be called ‘cultural-attraction seekers.’ Heritage attractiveness plays a significant role in their location choice, and they are willing to pay for experiences and attractions. Class 2 (LC2) members too wish to avoid crowds and are attracted to attractive cultural heritage. They are however more reluctant to pay for experience and entrance fees. They assign relatively high value to other people’s opinion about attractiveness. Moreover, they have a relatively high base reference for not choosing a location from available options. Hence, these respondents can be labeled as ‘selective sightseers.’ The class 3 (LC3), members can be described as people who assign more value to city life and experiences. Crowdedness is not important for these individuals; they give relatively high value to their experiences such as shopping, drinking, and eating. Hence, this class can be called ‘city-life lovers.’

The prediction of class membership

Having identified the classes, a Multinomial Logistic Regression analysis is conducted to predict the class of an individual based on sociodemographic characteristics, choices made in the last visit to Amsterdam, and benefits sought in visiting heritage. The class membership of each respondent is determined by assigning the person to the class that has the highest membership probability. This analysis offers insight in the relationships between attitudes, behavior and socio-demographic group, on the one hand, and heritage location preferences, on the other. The backward elimination method is used for model selection. The reference category is set arbitrarily to LC3. The detailed regression analysis results can be found in Appendix B. Table 7 summarizes the significant variables on which the classes differ and an indication of how they differ.

Table 7. Results of ML regression analysis of class membership.

Heritage location preferences
Class 1. cultural-attraction seekers	Class 2. selective sightseers	Class 3. city-life lovers
Socio-demographic characteristics
Less likely to be between 35 and 54 years old	more likely to be 55+	Reference category
Last visit in Amsterdam
More likely to travel with children	More likely to travel with children	Reference category
Not significant	More likely to travel with colleagues	Reference category
Less likely to travel with 4 + people	Less likely to travel with 4 + people	Reference category
More likely to travel by car	Not significant	Reference category
Not significant	Less likely to travel by car	Reference category
Less likely to visit heritages in their last trip	Not significant	Reference category
Benefits of visiting heritage
More likely to seek benefit about personal bond	Not significant	Reference category

As for the sociodemographic characteristics, only age group is found to be significant. Cultural attraction seekers are less likely to be between 34 and 55 years old, whereas selective sightseers are more likely to be 55 years and older compared to city lovers. This finding could be related to generational differences since young individuals may have greater interest in cultural attraction and historical sites, which align with the cultural-attraction seekers. On the other hand, older individuals are more selective in their location choice. They may prioritize destinations and experiences that correspond to their specific interests and preferences.

As for choices made during the last visit in Amsterdam, cultural attraction seekers and selective sightseers are more likely to travel with children and less likely to travel with four or more people compared to city lovers. Selective sightseers more often travel with colleagues. This divergence could be due to the nature of the locations these members prefer. These locations might be more appealing to smaller groups and families. Furthermore, cultural attraction seekers are more likely to travel by car. The reason could be the flexibility and convenience since they are more likely to travel with children. Selective sightseers are less likely to travel by car. They could prioritize other transportation modes for a local travel experience.

As for heritage awareness, cultural attraction seekers are less likely to visit heritage. The reason could be that this group has already been several times before to Amsterdam and have already seen the cultural heritage sites. For the benefits sought in visiting heritage, cultural attraction seekers are more likely to seek benefits about the personal bond in their last visit to Amsterdam. They might prioritize engagements and heritage locations that enable them to establish a personal connection with local culture, possibly driven by familiarity.

Policy recommendations

Based on the results presented in the previous section, the following policy suggestions can be drawn to address the challenges associated with mass tourism in urban heritage destinations. According to Neuts & Vannaste (2020), an overcrowded public space is an undesirable aspect. Once a certain threshold of crowdedness is exceeded, the adverse impacts of overcrowding in public spaces outweigh any potential benefits. These suggestions seek to create a more balanced distribution of and an enjoyable experience for visitors.

Aiming at the cultural-attraction seekers class, strategies to distribute visitors throughout the city to less-known heritage locations and creating itineraries focused on cultural and recreational heritage most likely can reduce overcrowding. This class of visitors display a distinct preference for educational and recreational heritage. Promoting the access to pre-visit information through websites can enhance their experience. Tailored cultural tours that appeal to the age group between 35 and 54, which is overrepresented in this class, and incorporating less-known attractions can be offered to attract them. Given their inclination to travel with children, family-friendly packages can be considered. This is in line with the findings of the study by Molinillo and Japutra (2017) that popular cultural locations often attract people who travel with family. Also, since they are more likely to use cars for transportation, convenient and well-located car parking facilities near the less-known heritage sites can enhance the attractiveness of these destinations. Lastly, the introduction of interactive experiences in these less-known heritage sites, facilitating a personal connection with the local culture can be a valuable addition.

Selective sightseers, who seek attractive cultural heritage and prioritize cost-effectiveness, can be attracted through affordable or no entrance fees for the less-known heritage sites. This class involves visitors who are more likely to be aged 55+. Given the older age, offering seating areas and clear signages can enhance their experience. Furthermore, providing discounted entrance fees to older people is a possibly effective measure. As they are less likely to travel by car, enhancing public transportation options to the less-known heritage sites is important. Improving transportation options and easy connections between the point of interests within tourist districts and urban destinations is a critical factor (Le-Klähn & Hall, 2015). Providing real-time information on crowd levels can help them to choose less crowded times to visit.

For the city-life lovers, who are attracted to heritages in natural settings and value safety and cleanliness more strongly, preserving natural elements within the cityscape is important. This includes the maintenance of historical gardens and green spaces. To cater to their high price sensitivity, reasonable entrance fees can be considered. Providing information about facilities such as hotel-restaurant-café and shopping options within walking distance of their chosen heritage type, can enhance their urban experience.

The results offer valuable insights by assigning weights to different attributes, which can guide policymakers in formulating heritage tourism strategies for specific segments. Our analysis indicates that the entrance fee has the most significant impact on visitors’ choices. Implementation of diverse pricing schemes, such as introducing higher fees during peak hours at popular attractions, can help to more evenly distribute the influx of visitors throughout the day, and encourage the discovery of alternative heritage sites.

Conclusion and discussion

The study had two primary objectives (i) to identify visitors’ segments in urban heritage tourism based on their location choice, (ii) to formulate policy recommendations to mitigate mass tourism and avoid congestion. The analysis was conducted using a sample of individuals who had paid a visit to Amsterdam within the past five years.

The first finding of this study is that individuals see several benefits of heritage visits, including pleasure and entertainment, knowledge and learning, personal bond, and popularity. Similar to past studies, motivations for visiting heritage such as a desire to have a day out and entertainment, wanting to see a world-famous site (Poria et al., 2004), learning about the physical nature of the site (Tu, 2020), fun and entertainment (Oh et al., 1995) are reasons for urban heritage visits. Although Tu (2020) found that a sense of achievement shows long-term intrinsic motivation and should be considered for satisfaction, our results suggest that a sense of achievement has the lowest mean score within the knowledge and learning group. The reason could be that we framed this item as ‘knowledge and learning’ instead of ‘long-term value’; therefore, framed in that way, the importance of the perception of the ‘sense of achievement’ is not confirmed in our study. Regarding respondents’ recent visit to Amsterdam, most of them chose a one-day city trip, often accompanied by someone, on which they visited urban heritage sites.

The research aims were achieved through a stated choice experiment involving key attributes in urban heritage location choice. The MNL model estimated on the collected choice data, shows that entrance fee is the most important attribute in heritage location choice followed by crowdedness level and heritage attractiveness. The results are in line with the study by Alexandros and Jaffry (2005) which indicated that congestion levels are significant for people’s location choice and a high congestion level results in dissatisfaction. On the other hand, this is in contrast to the study by Li and Lo (2004) concluding that heritage type is a critical factor for visitors.

To identify segments based on preference, we conducted a latent class analysis. Based on the resulting classes, we categorized heritage tourists as ‘cultural attraction seekers,’ ‘selective sightseers’ and ‘city life lovers.’ It shows that visitors are not homogenous in their location choice behavior. The results of a multinomial logit regression analysis to predict membership indicates that age group, travel party size, mode of transportation, and awareness of heritage are significant factors in predicting to which segments of heritage location visitors a person belongs. Based on the results, policy recommendations were formulated to mitigate mass tourism and avoid congestion.

Crowdedness level emerges as a critical factor as was also emphasized in previous research (e.g. Neuts & Vanneste, 2020). To mitigate congestion, well-designed crowd management strategies, including measures such as visitor capacity limits, and reservation systems can be considered. Furthermore, our results show that visitors to Amsterdam often use car or public transportation to access the city. One approach to reduce crowding could involve offering combined tickets for heritage sites (popular and less-known heritage sites), along with discounted public transportation fares during off-peak hours. Moreover, since walking is the most common travel mode within Amsterdam, the installation of signage providing information on real-time crowdedness levels and distances between heritage areas can support route planning between popular and less-known heritage areas and also on-route activities such as dining, sightseeing, and shopping.

Limitations

Although the results gave insights into urban heritage location choice per visitor segment, a number of limitations should be mentioned. First, participants were selected based on whether or not they had visited Amsterdam one or more times in the past five years. Although this should offer a sample of potential cultural heritage visitors, the sample might not fully represent visitors from all demographic groups proportionally. Second, the chosen attributes were selected from relevant literature. However, only a limited number of attributes can be included in a SCE considering respondent burden. Third, this study used Amsterdam as a case area and the results might differ in other touristic cities. Therefore, the applicability of the policy suggestions to other regions is an issue.

The results of the current study can serve as valuable input for subsequent research, aimed at understanding interactions between urban heritage features and attraction of visitors, based on revealed data on choice behavior. Future studies may make use of location-based data such as geo-tagged photographs (Karayazi et al., 2022), geotagged textual data (Lupu et al., 2021), or GPS-based data of visitor flows (Dane et al., 2020) to enrich data on choice behavior. Furthermore, external factors such as pandemics, economic conditions, political conditions, and cultural events (Bogari et al., 2003; Hanqin & Lam, 1999; Uysal et al., 2009) that can affect visitor choices, were not explored in this study. Further research may consider the integration of such factors.

Acknowledgements

The authors want to thank Mandy van de Sande-van Kasteren (Education/Research Officer) for LimeSurvey administration.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This research was funded by The Republic of Turkiye Ministry of National Education, Directorate General for Higher and Overseas Education, under 2015–1 YLSY program.

Notes on contributors

Sevim Sezi Karayazi

Sevim Sezi Karayazi is a Ph.D. Candidate at the Department of Built Environment of Eindhoven University of Technology. Her research interests lie in data-driven solutions for managing tourism's negative impacts in historical cities. She works to integrate urban big data analysis with advanced methodologies to identify attractive heritage areas and analyze visitors’ patterns contributing to sustainable societies.

Gamze Dane

Dr. Gamze Dane is a tenured Assistant Professor at the Department of Built Environment of Eindhoven University of Technology. Her areas of expertise include decision-support systems in urban planning, human-environment interaction, GIS, urban informatics and data analytics. Her research integrates citizens into the decision-making process of urban transformations via digital tools and (big) data.

Theo Arentze

Prof. Dr. Theo Arentze is a Full Professor of Real Estate and Urban Systems and chair of the Real Estate and Urban Management group at the Eindhoven University of Technology. His research interests include choice behavior of users of the built environment, multi-actor decision making, decision support systems and large-scale simulation of urban and transport systems with applications in urban planning and real estate management.

Notes

1 The package that we use for implementing the online survey (LimeSurvey) does not allow us to randomly generate choice sets at the individual level. Therefore, a sufficient number of unique choice sets were generated once for the whole sample and from that set nine choice sets were randomly drawn (without replacement) for each respondent. The method %choiceeff in SAS was used to generate a set of 81 choice sets. This method generates an efficient design of choice sets based on the 27 alternatives supplied. An efficient design is a design for which the standard errors of the beta parameters to be estimated are minimized.

Appendices

A. Attribute distribution

Table

Attributes	Place-based attributes (PB)					Supporting products (S)			Monetary costs (M)
Author	1	2	3	4	5	1	2	3	1	2
Birenboim et al. (2022)	X		X						X
Altınay Özdemir (2022)	X							X
Neuts and Vanneste (2020)	X	X				X			X
Lupu et al. (2021)			X	X						X
Ferretti and Gandino (2018)	X									X
Arentze et al. (2018)		X			X			X		X
Capitello et al. (2017)							X	X		X
Bravi and Gasca (2014)							X	X
Chiabai et al. (2014)								X		X
Lacher et al. (2013)	X						X	X		X
Figini and Vici (2012)	X		X			X		X
Draper et al. (2012)						X		X
Dane et al. (2019)							X
Apostolakis and Jaffry (2005a)							X			X
Apostolakis and Jaffry (2005b)							X			X
Alexandros and Jaffry (2005)							X			X

Table

Attributes	Physical accessibility (PA)		Crowdedness (C)		Time-related subject (T)
Author	1	2	1	2	1	2	3
Birenboim et al. (2022)			X
Altınay Özdemir (2022)					X	X
Neuts and Vanneste (2020)	X		X
Lupu et al. (2021)						X	X
Ferretti and Gandino (2018)						X	X
Arentze et al. (2018)	X
Capitello et al. (2017)
Bravi and Gasca (2014)	X
Chiabai et al. (2014)		X
Lacher et al. (2013)
Figini and Vici (2012)						X
Draper et al. (2012)
Dane et al. (2019)
Apostolakis and Jaffry (2005a)		X		X
Apostolakis and Jaffry (2005b)		X		X
Alexandros and Jaffry (2005)		X		X

Notations are as follows: PB1 Layout, PB2 Nature, PB3 Location, PB4 environmental quality, PB5, heritage type, PB6 attraction value; S1 Shops, S2 dining, S3 experience; M1 trip/activity cost, M2 entrance fee; PA1 accessibility, PA2 information; C1 crowd, C2 congestion; T1 season, T2 duration of visit, T3 travel time.

B. The details of multinomial logistic regression

Model fitting information

Table

	AIC	BIC	−2LL	Chi-Sq	df	Sig.
Intercept only	915.359	923.519	911.359
Final	889.3	1019.858	825.3	86.06	30	<.001

Parameter estimates

Table

LC^a		B	Std. error	df	Sig.	Exp(B)	95% CI for Exp(B)
							Lower bound	Upper bound
LC1	Intercept	1.027	1.246	1	0.41
	Age = 18–34	0.191	0.345	1	0.58	1.211	0.615	2.383
	Age = 35–54	−0.71	0.34	1	0.037	0.492	0.253	0.957
	Age = 55 year or 55+	0^b	.	0	.	.	.	.
	Family including children = No	−0.858	0.402	1	0.033	0.424	0.193	0.933
	Family including children = Yes	0^b	.	0	.	.	.	.
	Colleagues = No	−0.606	0.72	1	0.4	0.545	0.133	2.237
	Colleagues = Yes	0^b	.	0	.	.	.	.
	Travel party = Alone	0.079	0.468	1	0.865	1.083	0.432	2.712
	Travel party = Four and more	−0.689	0.363	1	0.058	0.502	0.246	1.023
	Travel party = Three	0.553	0.441	1	0.21	1.739	0.732	4.13
	Travel party = Two	0^b	.	0	.	.	.	.
	Initial travel mode = Car	0.515	0.276	1	0.062	1.673	0.975	2.873
	Initial travel mode = other	0.763	1.056	1	0.47	2.145	0.271	16.981
	Initial travel mode = Public Transportation	0^b	.	0	.	.	.	.
	Travel mode within Amsterdam = Car	0.242	0.403	1	0.548	1.274	0.578	2.809
	Travel mode within Amsterdam = other	−0.713	0.8	1	0.373	0.49	0.102	2.354
	Travel mode within Amsterdam = Public Transportation	0.107	0.299	1	0.722	1.112	0.619	1.999
	Travel mode within Amsterdam = Walk	0^b	.	0	.	.	.	.
	Heritage awareness = didn't visit	−0.59	0.309	1	0.056	0.554	0.303	1.015
	Heritage awareness = No	−0.44	0.323	1	0.173	0.644	0.342	1.213
	Heritage awareness = Yes	0^b	.	0	.	.	.	.
	Personal bond	0.072	0.039	1	0.069	1.074	0.994	1.16
LC2	Intercept	4.677	1.369	1	<.001
	Age = 18–34	−0.701	0.382	1	0.066	0.496	0.235	1.048
	Age = 35–54	−1.203	0.378	1	0.001	0.3	0.143	0.63
	Age = 55 year or 55+	0^b	.	0	.	.	.	.
	Family including children = No	−1.068	0.458	1	0.02	0.344	0.14	0.844
	Family including children = Yes	0^b	.	0	.	.	.	.
	Colleagues = No	−1.699	0.774	1	0.028	0.183	0.04	0.833
	Colleagues = Yes	0^b	.	0	.	.	.	.
	Travel party = Alone	−0.32	0.531	1	0.548	0.727	0.256	2.059
	Travel party = Four and more	−1.017	0.421	1	0.016	0.362	0.158	0.826
	Travel party = Three	0.726	0.476	1	0.127	2.067	0.814	5.249
	Travel party = Two	0^b	.	0	.	.	.	.
	Initial travel mode = Car	−0.27	0.308	1	0.38	0.763	0.418	1.395
	Initial travel mode = other	1.067	1.03	1	0.3	2.907	0.386	21.871
	Initial travel mode = Public Transportation	0^b	.	0	.	.	.	.
	Travel mode within Amsterdam = Car	−1.054	0.55	1	0.055	0.348	0.119	1.023
	Travel mode within Amsterdam = other	−0.175	0.785	1	0.823	0.839	0.18	3.91
	Travel mode within Amsterdam = Public Transportation	−0.312	0.342	1	0.361	0.732	0.374	1.43
	Travel mode within Amsterdam = Walk	0^b	.	0	.	.	.	.
	Heritage awareness = didn't visit	0.311	0.347	1	0.369	1.365	0.692	2.692
	Heritage awareness = No	0.301	0.361	1	0.405	1.351	0.666	2.743
	Heritage awareness = Yes	0^b	.	0	.	.	.	.
	Personal bond	−0.07	0.044	1	0.115	0.932	0.855	1.017

^a The reference category is: LC3.

^b This parameter is set to zero because it is redundant.