Evolving domestic tourism destination preferences post-apartheid

Abstract

Tourism is embedded within societal structures, and imbalances upheld through social structures, like systemic racism, leave symbolic boundaries where certain activities perceivably belong to designated population groups. Further, socio-economic factors impede travel behavior especially in emerging markets. Resultantly, marginalization becomes a barrier to inclusive tourism. Domestic tourism patterns in post-apartheid South Africa were explored to determine whether changes have occurred, and whether these changes are a function of race, ethnicity, or socio-economic status. Preferences for three different destinations were compared using two nationally representative samples from the 2006 and 2017 South African Social Attitude Survey. Results indicate that, although race and economic status remained significant, ethnicity was the main impacting variable and interactional effects between ethnicity, age, poverty status, and geotype explained most of the variance. While travel habitus and cultural distance influence preferences, the youth market within certain ethnic groups is most likely to transition to new tourism destinations.

Keywords:

• Race
• ethnicity
• marginalization
• apartheid
• destination preference
• domestic tourism

Tourism participation operates within contexts and systems, including systemic racism, and it must be understood as deeply embedded within local historical and societal, as well as global power structures (Alderman et al., 2022; Benjamin & Dillette, 2021; Stodolska et al., 2019). Tourism preferences, destination images, and travel behavior are considered to be historically loaded constructs, impacted by images and traces from the past (Grundlingh, 2006; Saarinen et al., 2017; Sixaba & Rogerson, 2019), and individuals’ participation in tourism are directed by these historical dynamics (Dillette & Benjamin, 2022; Musavengane & Leonard, 2019). Neglecting a historical perspective contributes to continued segregation and slow pace of transformation in tourism destination choice (Rogerson, 2020; Scheyvens & Biddulph, 2018). Tourism and leisure as global phenomena offer the opportunity to critically study the issue of race alongside other intersecting forms of oppression (Thakur et al., 2023).

Studies on the topic of destination choice tend to focus on race as homogenous groups (Benjamin et al., 2024). This approach has been criticized given that societies have been, and progressively become, more diversified due to greater mobility and migration (Lee & Scott, 2017; Saarinen et al., 2017; Whiting et al., 2017). Greater recognition of both racial and ethnic differences is necessary as marginalization and power dynamics change over time (Abdullah et al., 2023). Though race and ethnicity are tied together they are separate constructs and often confused with each other (Gracia, 2014; Mersha & Beck, 2020). Race is mostly determined by phenotypes (physical features) to create biological taxonomies, while ethnicity is not constrained by inheritable physical features, is more flexible, contextual and related to country of birth, cultural elements (language, beliefs, traditions, values), and geography (Bhadra, 2017; Du Preez & Govender, 2020; Gracia, 2014). Language and geographical location are often considered the most powerful ethnic group markers (Maleševic, 2004). However, both race and ethnicity are organic constructs occurring within a social context, representing personal and social biases (Gracia, 2014). The classifications are mostly used for public policy implementation and may serve systemic racism (Corlett, 2003, as cited in Gracia, 2014),

Race and ethnicity form part of a myriad of aspects related to marginalization in tourism, which in turn coincides with different degrees of discrimination (Abdullah et al., 2023; Krymkowski et al., 2014). Individuals may be physically pushed away through spatial boundaries, but also socially and economically (Abdullah et al., 2023), creating symbolic boundaries (Thakur et al., 2023). Certain destinations can be perceived as risky and inhospitable, for example in the United States (US) (Benjamin & Dillette, 2021; Davis, 2019; Hudson et al., 2020), South Africa (Smith, 2023; Rogerson et al., 2022) and Asia (Tjiptono & Yang, 2018). Even though discriminatory policies may have been removed, the symbolic effect of the impact of such discriminatory policies remains and affects destination and activity preferences. Not only would this perception impact a person directly but will also form part of a tourism consciousness that may be carried over from generation to generation (Musavengane & Leonard, 2019). This creates territorial tourism spaces (Saarinen et al., 2017), leisure cultural distance (Ahn & McKercher, 2015), and travel habitus (standard travel moment) (Lee & Scott, 2017; Musavengane & Leonard, 2019) that impact destination preferences. According to the travel habitus theory, differences among social groups created by social and historical circumstances, are eventually internalized and institutionalized by groups members and become norms or tendencies that guide behavior and thinking (Scott & Lee, 2018).

Neglecting distinctions between ethnic groups and measures of cultural, limits our understanding of more refined patterns of participation and marginalization in leisure (Aizlewood et al., 2006; Krymkowski et al., 2014; Shores et al., 2007) and tourism (Benjamin et al., 2016, 2024). However, comparative studies that investigate how groups evaluate, avoid, and decide on travel destinations are sparse (Benjamin & Dillette, 2021; Liu et al., 2018). These finer segmentations become important, as reflected in the increasing number of historical tourism studies encompassing terms such as ancestral tourism (Weaver et al., 2017) and diasporic tourism (Li et al., 2020). The term cultural distance has also become more common in tourism literature, referring to the extent of difference between the culture of tourists’ origins and that of the travel destinations (Ahn & McKercher, 2015). This increasing recognition of racial and ethnic differences happens as marginalization and power dynamics change over time (Abdullah et al., 2023).

Problem statement

South Africa is used as case study and offers a unique opportunity to investigate the dynamics around race in tourism given its historically diverse populations (four racial groups and nine ethnic groups) and discriminatory political structures reflected in apartheid. Here the White population was in the minority, while in other countries that also bare the effect of systemic racism on the tourism industry, for example in the (US) (Alderman et al., 2022; Benjamin & Dillette, 2021), the White population was in the majority. Though the political system was distinct from that of the US, South Africa experiences similar prevailing challenges in generating travel to certain local destinations (Hudson et al., 2020; Kruger & Douglas, 2015; Rogerson, 2017), consider distinctly different groups within the market to develop tourism product offerings (Benjamin et al., 2024; Lubbe et al., 2013), and to overcome stereotypical views on the travel choices of different racial and ethnic groups (Benjamin & Dillette, 2021; Smith, 2023; Hlatshwayo, 2016). In 1996 (apartheid was abolished in 1994), the newly established Department of Environmental Affairs and Tourism prioritized previously neglected population groups (based on race) as a key emerging domestic market segment in South Africa. Subsequently new marketing campaigns were developed to grow this potentially lucrative market (National Department of Tourism, 2011). However, apart from a lack of awareness, there was a mismatch between demand and supply across regional destinations (Lubbe et al., 2013, 2016). By 2017, the new National Tourism Sector Strategy (National Department of Tourism, 2017) reflected little change. Domestic tourism showed a decline ascribed to economics (finances, employment), seeing no reason to take a trip, and disliking travel. Despite the duration since, and these tourism development efforts made post-apartheid, tourism persistently reflects an exclusionary nature (Rogerson, 2020). Further, South African society continues to present systemic injustice, marginalized social groups (between and within racial groups), and spaces that maintain racial inequalities despite post-apartheid transformation (Pillay, 2022). Recent studies on marginalized groups within South African tourism relate to, for example, tourists with disabilities (Makuyana & Dube, 2023), gender issues (van Zyl & Bartis, 2024), and a wide range of studies on the involvement of marginalized rural communities in tourism development. To the best of the authors’ knowledge, no studies investigate tourist behavior by comparing the preferences between population groups from a marginalization perspective or considering multiple dimensions (ethnicity, race, economics, socio-demographics), while such an approach offers the most insight into the phenomenon (Krymkowski et al., 2014; Thakur et al., 2023).

Study aim and research questions

The study aim is to increase the theoretical and empirical understanding of destination preferences among emergent travelers by considering constraints brought about by marginalization based on race, ethnicity, and economic status. Washburne’s (1978) conceptualization of subculture theory is applied to test which of two hypotheses of marginalization, namely the “ethnicity hypothesis” (symbolic constraints around race and ethnicity) or “marginality hypothesis” (structural constraints around economic status), dominate with regards to destination preferences. These hypotheses encapsulate key dimensions of marginalization that limit individuals’ full participation in tourist experiences (Baah et al., 2019, as cited in Thakur et al., 2023). To the authors’ best knowledge, no leisure or tourism studies within South Africa have applied Washburne’s conceptualization as theoretical framework. Furthermore, industry transformation and inclusivity are investigated from a critical perspective in this study to better understand tourist decision making (Abdullah et al., 2023; Mura & Wijesinghe, 2023; Thakur et al., 2023). It also contributes to the underrepresentation of leisure studies from global South regions including Africa (Carnicelli & Uvinha, 2023; Singh et al., 2023).

To operationalize the study, three research questions are stated. The first question focuses on race as the foundation of the apartheid system Accordingly, all South Africans were firstly crudely classified as belonging to one of four main racial groups based on phenotypes, including Black African, Colored, Indian/Asian, or White with the White population considered superior. Throughout the paper the authors will refer to race groups as Black African, Colored, Indian/Asian, and White as per the Statistics South Africa terminology. Colored is the term used for individuals of mixed race. The authors acknowledge the sensitivities around these terms and that racial classification remains offensive. The term ethnicity is also often used as a synonym for race in South Africa and the authors acknowledge sensitivity around this practice. The South African White minority was considered superior and their social structures regarded as the desirable norm. The apartheid system established segregated leisure and tourism spaces (Rogerson & Rogerson, 2020), and the first question and corresponding research hypothesis tested whether race played a role in the formation of post-apartheid destination preferences.

RQ 1: Does race predict destination preferences when discrimination is no longer enforced through formal policies?

H₁: Race is a significant predictor of destination preference when discrimination is no longer enforced through formal policies.

The second research question with three corresponding hypotheses combined the three elements proposed by Washburne (1978) to compare the impacts of race (H_2a), ethnicity (H_2b), and socio-economics (H_2c) on the difference in destination preferences.

RQ 2: Does race, ethnicity and economic status contribute equally to differences in destination preference when discrimination is no longer enforced through formal policies?

As stated previously, discriminatory policies were firstly based on race and not ethnicity. However, to strengthen the government’s apartheid policy of separate development, government designated eight geographic areas, known as “Bantustans” or “Homelands,” to nine Black African ethnic groups (Zulu, Xhosa, Swati, Ndebele, Pedi, Sotho, Setswana, Venda, and Tsonga) under the Bantu Homelands Citizenship Act of 1970. The authors would like to acknowledge that these terms remain offensive. These separate territories were some of the most telling stories and symbols of domestic colonialism (Rogerson et al., 2022). Further segregation of the Black African population allowed the ethnic groups to sharpen their distinct identities (Du Preez & Govender, 2020; Hino et al., 2018). Determination of ethnicity in South Africa can therefore be considered as being dual in nature. While the ethnic groups evolved autonomously, they were also extensively shaped by colonial influences (Hino et al., 2018), the first two hypotheses stated under RQ2 (H_2a and H_2b) tested the “ethnicity hypothesis.”

H_2a: Compared to ethnicity and economic status, race is the most important predictor of destination preference when discrimination is no longer enforced through formal policies.

H_2b: Compared to race and economic status, ethnicity is the most important predictor of destination preference when discrimination is no longer enforced through formal policies.

A lasting legacy of apartheid is the high inequality levels between the White and nonwhite population groups, which remain and continue to bear a persistent racial undertone. The latest inequality trends indicate that, even two decades after apartheid, the White population still owns an unequal portion of the country’s income and wealth (Seekings, 2023). Other socio-economic factors that influence travel include employment as well as education attainment (Uvinha et al., 2017). The third hypothesis stated under RQ2 (H_2c) tested the “marginality hypothesis.”

H_2c: Compared to race and ethnicity, economic status is the most important predictor of destination preference when discrimination is no longer enforced through formal policies.

Apart from race, ethnicity and economic status, socio-demographic variables influence broader participation in tourism as well as specific travel preferences. These include age (Krymkowski et al., 2014; Tomić et al., 2019; Wangari, 2017), and cohabitation (Kasim et al., 2013; Krymkowski et al., 2014). The geotype a person resides in plays a significant role considering marginalization in leisure (Krymkowski et al., 2014), cultural distance (Ahn & McKercher, 2015) and travel habitus (Lee & Scott, 2017; Musavengane & Leonard, 2019). Geotype is especially influential in travel decisions when considering race and ethnicity in South Africa (Du Preez & Govender, 2020). The following research question is stated:

RQ 3: Which socio-demographic variables act as enablers for travelers to transition to destinations considered to be outside their travel habitus?

While studies in the field tend to be qualitative and descriptive in nature, this explanatory study determined the what and why through a quantitative research design. The authors compared secondary data from the 2006 and 2017 iterations of the South African Attitudes Survey (SASAS) to address the research questions. These two different time points allowed the authors to explore whether and on what grounds population groups differed regarding destination preferences after abolishment of discriminatory legislation in 1994. A UNIANOVA was performed to firstly determine if race predicted destination preference (RQ 1; H₁), and secondly to test which variable (race, ethnicity, or economic status) had the greatest influence on destination preferences (RQ2, H_{2a, b, c}). Economic status included poverty status, employment and education. The effect of other socio-demographic variables (age, gender, cohabitation, geotype) were explored by analyzing main and interactional effects of the UNIANOVAS (RQ 3). The paper ultimately contributes to the growing critical scholarship on the impact of systemic racism within a tourism context (Dillette et al., 2019) by specifically focusing on apartheid as a unique form of legislated racism (Rogerson & Rogerson, 2020).

Literature overview

A growing body of literature promulgates the significance of tourism’s past and contexts, especially from a racial perspective as race and tourism are “inextricably bound together in their practices” (Jamerson, 2016, p. 1043). Acknowledging and confronting the historical foundations of racial impact on tourism is essential to foster a more just and accessible travel environment (Thakur et al., 2023). Chio et al. (2020) present a collection of expert opinions on the issue of how race shapes tourism, and challenge scholars to view race as an ever-evolving constituent of the tourism landscape. They argued that more work is needed from a critical perspective to investigate the intersections of travel, leisure, economics, and race.

Different research streams studying race within tourism

In this article the authors identified six perspectives from which the topic has been studied within tourism and leisure. These include racial discrimination; race, cultural constraints and socio-economic constraints (respectively) as distinguisher between destination preferences; intersectionality of these different constraints; and studies that inform ways to encourage travel outside of habitus.

Critical examination of the impact of race on tourism mostly emanates from the US and underscores a complex historical narrative (Chio et al., 2020). Various facets of travel experiences in the US are permeated by racial discrimination rooted in the legacy of slavery and institutionalized segregation dating back to the Jim Crow era (Alderman et al., 2022; Foster, 1999). While legal barriers were dismantled by the Civil Rights Movement, contemporary issues persist (Dillette & Benjamin, 2022), including racial profiling and disparities in access to travel resources (Davis, 2019). Rogerson and Rogerson (2020) made a comparison between the legacy of apartheid in South African and the Jim Crow period in the US, while Rogerson (2020, 2022) analyzed the effect of the negative legacies left by the native reserves, established during apartheid, on rural tourism development. Interactions between tourism and racial dynamics remain complex and multifaceted within South Africa.

Scholars have investigated the differences in destination preferences between race groups in the US. Similar to racialized tourism (Benjamin et al., 2016; Duffy et al., 2019; Hudson et al., 2020) and leisure spaces (Davis, 2019; Pinckney et al., 2018, 2024; Scott & Lee, 2018), within that country, the apartheid system established racialized tourism spaces in South Africa (Smith, 2023; Rogerson & Rogerson, 2020). Entrenched domestic travel patterns present limits to distribution of domestic travelers across different regions (Du Preez & Govender, 2020; Rogerson, 2015). Destination avoidance and inertia is also evident among certain population groups (Butler & Richardson, 2015; Kruger & Douglas, 2015).

In addition to race, ethnic variation in leisure and tourism preference exists, making it important to determine the role of ethnicity, cultural elements (Li et al., 2020; Pinckney et al., 2024; Whiting et al., 2017) and levels of acculturation (Du Preez & Govender, 2020) in travel decisions. Cultural norms, values and customs provide followers with a script and can both facilitate or constrain participation in different leisure activities (symbolic constraints). This forms cultural boundaries and boundary maintenance (maintaining the group’s distinctiveness), as well as travel habitus (standard travel moment); perpetuating ethnic and/or racial differences in leisure activities (Floyd & Stodolska, 2014; Scott & Lee, 2018) and tourism (Li et al., 2020). Destination images are culturally specific and important since ethnic groups may not participate in recreation in certain settings perceived to detract from their collective identity (Floyd & Stodolska, 2014). Even without travel barriers and restrictions, values and customs within groups provide followers with a script about the kinds of leisure and outdoor recreation behaviors to which they ought to conform (Whiting et al., 2017). Group members engage in boundary maintenance (maintaining distinctions between the group and others) to the extent that they adhere to cultural norms (Du Preez & Govender, 2020). This process of actively constructing differences in leisure activities partially explains differences among cultural groups (Scott & Lee, 2018).

On the other hand, differences in recreation behavior are attributed to socio-economic factors. According to the literature, socio-economic constraints are in fact one of the most important constraints that affect travel behavior in general (de Almeida & Kastenholz, 2019). The majority of scholars that have studied group comparisons in the US argue that these differences are due to socio-economic factors (Godbey et al., 2010; Sevilla et al., 2012; Scott, 2013). Regardless of race, background, or ethnicity, low-income Americans are far more constrained in their leisure compared to other Americans (Scott, 2013), while a lack of resources, specifically between ethnic and racial groups, leads to lower access and interest in leisure (Stodolska et al., 2019). Similar results have emerged in South African as a developing country (National Department of Tourism, 2017; Dzikiti & Leonard, 2016; Kruger & Douglas, 2015; Rogerson, 2020).

Researchers have explained travel differences as a multilayered phenomenon, looking into the intersectionality of ethnicity, class, vulnerability, and marginalization (Du Preez & Govender, 2020; Krymkowski et al., 2014; Shinew et al., 1996; Shores et al., 2007). Lastly, a small number of studies describe ways of ensuring minority participation in tourism given the existence of travel habitus (standard travel patterns), for example the role of social media (more specific within the Black Travel Movement) (Dillette et al., 2019).

Subculture theory

The concept of subculture was first used at the Chicago School of Sociology and initially linked with biology and psychology to define deviant behavior in both American and British literature. Durkheim broadened the conceptualization to include groups that reacted to objects that affected them by forming unity to confront anomie (Durkheim, 1901, as cited in Blackman, 2014), and Cohen subsequently theorized that subcultures form as a result of frustration with status levels (Cohen, 1956, as cited in Blackman, 2014). Subcultures are interpreted as social groupings that form within the wider social, political, and historical landscapes. Washburne (1978) was one of the first to explore subcultural theory in relation to constraints to outdoor recreation. He suggested that the constraints perceived by racial and ethnic minorities in the US may not only be a result of economic marginalization. He applied subculture theory to develop two hypotheses that explain differences in leisure participation, namely that non- or under-participation in leisure can either be explained as a function of racial/ethnic elements (the “ethnicity” hypothesis), or socio-economic elements (the “marginality” hypothesis). Washburne’s conceptualization of subculture theory is one of the most cited approaches to understand underrepresentation in leisure and despite criticism by some scholars preferring multiple hierarchy modeling techniques to explain constraints, this theory has recently made a resurgence (Stodolska et al., 2019). Since this ground-breaking article by Washburne (1978), a great number of researchers from the US have indicated that constraints experienced by marginalized racial and ethnic minority groups in the US are at least partially explained by historic discrimination, different cultural values, and personal or institutional forms of discrimination (Floyd & Stodolska, 2019; Krymkowski et al., 2014; Pinckney et al., 2024; Scott & Lee, 2018; Stodolska et al., 2019). Despite being well established, Washburne’s approach has been criticized because it uses the race and ethnicity terms interchangeably and does not distinguish between ethnic groups within race groups (Floyd, 1999; Johnson, 1997; Stodolska, 2018). The current study makes such a clear distinction based on the classifications of population groups delineated by apartheid policy.

Subcultures in the South African context

The South African Population Registration Act of 1950 provided the basic framework for what is known as apartheid – a political system whereby all South Africans were primarily crudely classified based on race (according to phenotype). Designated apartheid policies engineered the segregation of leisure in South Africa including The Land Act of 1910 and 1936; The Reservation of Separate Amenities Act of 1953; The Group Areas Act of 1955; The General Sea-Shore Regulations of 1962; and the Black Affairs Administration Act of 1971 (Magi & Nzama, 2002). As a result, the growing tourism economy was almost exclusively a privilege of the White population and tourism spaces remained firmly under the control of the apartheid government (Rogerson & Rogerson, 2020; Rogerson et al., 2022). Accordingly, nonwhite travelers could, for example, not visit certain game parks or beaches in South Africa (Smith, 2023; Kruger & Douglas, 2015; Rogerson, 2017). During the apartheid years, the tourism market consisted mainly of White domestic travelers as well as White travelers from neighboring countries such as Zimbabwe (then called Rhodesia) and Mozambique (Grundlingh, 2006; Rogerson & Rogerson, 2020). The most disadvantaged nonwhite community was the Black African majority who only had access to limited designated tourism and leisure spaces and facilities (Dzikiti & Leonard, 2016). The Colored group (which included people of mixed Black African, Khoisan, and European descent) and the Indian/Asian group (not indigent to South Africa but immigrated as sugar cane laborers in the latter half of the 19th century), were also subjected to apartheid policies. However, these two groups were relegated to a superior position vis-à-vis Black African people (Pirtle, 2021). To strengthen the government’s apartheid policy of separate development, government designated eight geographic areas, known as Bantustans or homelands, to nine Black African ethnic groups (Zulu, Xhosa, Swati, Ndebele, Pedi, Sotho, Setswana, Venda, and Tsonga) under the Bantu Homelands Citizenship Act (1970). This segregation allowed Black African ethnic groups to sharpen their distinct identities and these groups have both natural and constructed dimensions (Hino et al., 2018). The natural dimension is rooted in the ideas of bonds in biology and ancestry and is based on the premise that ethnic groups are extended kinship networks that serve as basic dividing lines within societies. Groups differentiated by color, language, religion, and race are embraced within these ethnic identities (Gracia, 2014; Mersha & Beck, 2020). In South Africa, ethnicity involves visible local communities, built on face-to-face signals of dialect, status, religion, and cultural practices, and the most common identifier of ethnicity remains language as was the case during apartheid (Hino et al., 2018). Due to complex structural colonial forces that played out around these identities during apartheid, ethnic groups were able to sharpen their distinct identities (Du Preez & Govender, 2020; Hino et al., 2018; Rogerson et al., 2022)

Methods

The study employed quantitative methodology to answer the research questions. This section presents the sampling and data collection techniques, followed by a description of the dependent variable along with the analysis to determine construct validity. This is followed by a description of the independent variables, before explaining the data analysis techniques.

Sampling and data collection

The study made use of secondary data obtained from the SASAS. The SASAS is a nationally representative, longitudinal cross-sectional survey conducted annually by the Human Sciences Research Council (HSRC). The survey was designed to yield a representative sample of 3,500 adults aged 16 and older living in South Africa. The SASAS sample design process included three stages. During the first stage, 500 geographical areas demarcated by Statistics South Africa (Small Area Layers) were selected, drawn probability proportionate to size, and stratified by geographical location (urban formal, urban informal, rural formal and rural informal) as well as by race (Black African, Colored, Indian/Asian and White). Seven households were then selected with equal probability from within each selected area. As the third sampling stage, data collectors were taken to the selected household where they had to list all persons 16 years and older living in the household. They then selected one random respondent using the Kish Grid method (Kish, 1949). Structured questionnaires, which were translated into the eleven official South African languages, were used as the research instruments. The mode of data collection was face-to-face and data collectors were recruited in each province to ensure that they would be able speak the home language of the selected respondents. The mode of face-to-face data collection was critical given the low literacy levels. After being collected and cleaned, the data was weighted to the latest mid-year population estimates (Lin et al., 2013) provided by Statistics South Africa to obtain a representative sample. In 2006 the sample size was 2904 (weighted = 31 136 800) and 3098 (weighted = 39 797 122) in 2017. Very few studies on the topic of marginalization and subculture theory are nationally representative (Krymkowski et al., 2014).

The dependent variables

In 2001 a South African domestic tourism survey was undertaken by South African Tourism (a national tourism agency responsible for marketing South Africa internationally and domestically) to determine domestic travel patterns in South Africa (Rule et al., 2001). In 2006 and 2017 the HSRC researchers included a scale of 13 items from this survey in its SASAS survey to determine how destination preferences had changed over time. The selected items measured the most popular leisure activities and would be familiar to the wider South African public. These items were all presented in the form of statements and respondents had to indicate their level of agreement (from 1 = strongly agreed to 5 = strongly disagreed).

To determine the linear combinations of the 13 items, two initial principal component analyses (PCAs) were conducted for the 2006 and 2017 data respectively. For both the 2006 and 2017 data, the KMO was .6 and the Bartlett’s Test of Sphericity was significant (p < 0.000), implying sufficient underlying correlations to do a meaningful PCA. All the items had communalities of .3 and higher which is considered acceptable (Tabachnick et al., 2007). The PCAs were followed by a Varimax orthogonal rotation, yielding five factors for both the 2006 and 2017 data; however, only three of these factors that comprised of identical items for both 2006 and 2017 were retained (these are shown in Table 1).

Table 1. Factor analysis of tourist preferences.

	2006			2017
Item	Game Park enthusiasts	Beach enthusiasts	VFR enthusiasts	Game Park enthusiasts	Beach enthusiasts	VFR enthusiasts
Cronbach Alpha	.637	.512	.144	.675	.531	.251
My favorite holiday destination is a Game Park	.787			.836
I would rather go to the Kruger Park than the beach	.857			.629
I would love to spend time in the desert	.526			.800
The best place for a holiday is at the beach		.823			.697
My favorite holiday destination is Cape Town		.753			.706
Foreign destinations are better than local ones		.395			.566
Most of my holidays are spent at the homes of family members or relatives			.688			.682
Traditional African food should be available in hotels			.639			.692

Item 3 (I would love to spend time in the desert) was removed from the first factor (for both 2006 and 2017) since this item had a negative impact on the overall reliability (Cronbach’s alpha) of the factor. Removing this item improved the alpha value from .637 to .739 in 2006 and .675 to .742 in 2017. An alpha of >.6 is regarded as acceptable (Kline, 2011). Following the same reasoning, item 6 (Foreign destinations are better than local ones) was deleted from the second factor, improving the alpha from .512 to .556 in 2006. Although it did not improve the alpha in 2017 (.531 to .493), it was considered appropriate to remove this item since it was not clear how it directly related to the new factor. All of the factor alphas for these attitudinal scales were considered sufficient (Hinton et al., 2014). The third dimension’s Cronbach alpha was very low (.144 in 2006 and .251 in 2017) and for this reason, only one item (Item 7) was used. The item was: Most of my holidays are spent at the homes of friends, family members or relatives This item was selected since it tapped more directly into VFR travel than item 8 (Traditional African food should be available in hotels). Dolnicar (2019) shows that a single statement can be used to capture the essence of a construct.

The first factor consisted of the statements: My favorite holiday destination is a game park and I would rather go to Kruger Park than to the beach. This factor was coined Game Park and included statements that represented specific locations and well-known top attractions for both the international and domestic markets. These destinations hold significant historically meaning from a marginalization perspective (Kruger & Douglas, 2015; Smith, 2023). The second factor consisted of the statements: The best place for a holiday is the beach; My favorite holiday destination is Cape Town and was coined Beach. It also included statements that represented specific locations and well-known top attractions for both the international and domestic markets. Similar to Game Park, these destinations hold significant historically meaning from a marginalization perspective (Rogerson, 2017). The third factor was coined VFR given that this is a commonly accepted term for visiting friends and relatives (Backer et al., 2017). This is a popular form of travel, especially given the migratory travel between places of origin and work created by apartheid policies. (Rogerson, 2015) and remains as the main form of domestic travel (NDT, 2017). The Indian population, for example, travel to the province where they originated from based on their ethnic and cultural identities (du Preez & Govender, 2020). These three factors became the dependent variables. Composite scores were created for the three dependent variables by converting the scores to a 0–100 scale with 0 indicating no preference and 100 indicating extreme preference.

Independent variables

The race variable consisted of the broader classification of race groups on which the apartheid policies were built, namely Black African, Indian/Asian, Colored and White. These discriminatory terms remain commonly used to group South Africans (Hino et al., 2018). For the ethnicity variable, race (the aforementioned four) and language were combined to create an ethnicity variable across the four race groups and 11 official languages, resulting in 16 ethnic groups. According to the 2017 Report of the South African Reconciliation Barometer Survey (Potgieter, 2017), language and race are by far the most prominent identities of South Africans and approximately one-half of South Africans consider language and race as either their primary or secondary identities. Eleven of the ethnic groups were African–Zulu, Xhosa, Swati, Ndebele (which form part of the bigger Nguni group); Pedi, Sotho, Setswana (part of the Sotho group), Venda, and Tsonga. Black African respondents who indicated Afrikaans or English as their main language were categorized as separate distinct categories. The minority non-African groups were grouped as Colored English, Colored Afrikaans, and Indian/Asian English. European descents were grouped into White Afrikaans and White English. Both of these variables were asked as categorical variables.

Table 3. Preferences for Game Park (mean scores by ethnic grouping).

2006								2017
	N	Mean	Mean	Mean	Mean	Mean	Mean		N	Mean	Mean	Mean	Mean	Mean	Mean	Mean	Mean
Colored English	82	2.4						African IsiXhosa	402	2.3
African Afrikaans	23	2.5						African Siswati	50	2.4	2.4
Colored Afrikaans	317	2.6	2.6					Colored Afrikaans	364	2.4	2.4
Indian English	266	2.7	2.7	2.7				Colored English	120	2.5	2.5	2.5
African IsiZulu	374	2.7	2.7	2.7	2.7			African IsiZulu	423	2.5	2.5	2.5	2.5
African IsiXhosa	300		2.9	2.9	2.9	2.9		Indian English	308	2.6	2.6	2.6	2.6	2.6
African Siswati	89		2.9	2.9	2.9	2.9		African Sesotho	225		2.7	2.7	2.7	2.7
White English	88		3.0	3.0	3.0	3.0	3.0	African Afrikaans	28		2.7	2.7	2.7	2.7	2.7
African Sesotho	230			3.0	3.0	3.0	3.0	African English	41			2.8	2.8	2.8	2.8	2.8
African Setswana	233			3.0	3.0	3.0	3.0	African Setswana	224			2.8	2.8	2.8	2.8	2.8
White Afrikaans	198				3.1	3.1	3.1	African Sepedi	233				2.8	2.8	2.8	2.8
African English	17				3.1	3.1	3.1	White English	93					2.9	2.9	2.9
African Sepedi	266					3.2	3.2	African IsiNdebele	34					2.9	2.9	2.9
African IsiNdebele	37					3.3	3.3	White Afrikaans	203						3.0	3.0	3.0
African Xitsonga	92						3.3	African Xitsonga	74							3.1	3.1
African Venda	46						3.3	African Venda	43								3.4
Significance		0.127	0.57	0.119	0.056	0.071	0.081	Significance		0.099	0.150	0.053	0.067	0.106	0.060	0.059	0.053

Dunnett’s Test (means comparison).

Data source: South African Social Attitude Survey (SASAS) (2006; 2017). *p <.00.

Table 4. Preferences for Beach (mean scores by ethnic grouping).

2006					2017
	N	Mean	Mean	Mean		N	Mean	Mean	Mean
African Venda	46	3.0			African Xitsonga	67	3.0
African Xitsonga	92	3.0	3.0		African Sesotho	215	3.0
African IsiZulu	377	3.1	3.1	3.1	African Venda	42	3.0
African Siswati	85	3.2	3.2	3.2	African Siswati	45	3.0	3.0
African English	18	3.3	3.3	3.3	African IsiXhosa	397	3.1	3.1
White English	93	3.3	3.3	3.3	African IsiNdebele	33	3.1	3.1
African IsiXhosa	313	3.3	3.3	3.3	African IsiZulu	398	3.1	3.1
African Setswana	237		3.4	3.4	African Sepedi	222	3.2	3.2	3.2
African Sepedi	271		3.4	3.4	African Setswana	214	3.2	3.2	3.2
Indian English	267			3.4	White Afrikaans	201	3.2	3.2	3.2
Colored English	82			3.4	Indian English	301	3.2	3.2	3.2
Afrikaans White	206			3.4	White English	94	3.3	3.3	3.3
African Sesotho	236			3.4	African Afrikaans	27		3.3	3.3
African Afrikaans	26			3.5	Colored Afrikaans	353		3.3	3.3
African IsiNdebele	36			3.5	Colored English	118			3.4
Colored Afrikaans	358			3.5	African English	41			3.5
Significance		0.054	0.063	0.052	Significance		0.072	0.054	0.060

Dunnett’s Test (means comparison).

Data source: South African Social Attitude Survey (SASAS) (2006; 2017). *p <.001.

Table 5. Preferences for VFR (mean scores by ethnic grouping).

2006							2017
	N	Mean	Mean	Mean	Mean	Mean		N	Mean	Mean	Mean	Mean	Mean	Mean
English White	93	2.9					English White	94	3.5
White Afrikaans	207	3.1					Afrikaans White	206	3.6	3.6
African Xitsonga	94		3.4				African English	41	3.6	3.6	3.6
Colored English	83		3.5				African Siswati	51		3.9	3.9	3.9
African Siswati	89		3.5	3.5			Indian English	313			3.9	3.9
Indian English	269		3.5	3.5			Colored English	123			3.9	3.9	3.9
African IsiNdebele	38		3.6	3.6	3.6		African Sesotho	235				4.0	4.0
African IsiZulu	389		3.7	3.7	3.7	3.7	African Afrikaans	28				4.1	4.1	4.1
African English	19		3.7	3.7	3.7	3.7	African IsiNdebele	35				4.1	4.1	4.1
African Sesotho	243		3.8	3.8	3.8	3.8	Colored Afrikaans	380				4.1	4.1	4.1
Colored Afrikaans	378		3.8	3.8	3.8	3.8	African IsiXhosa	420				4.2	4.2	4.2
African Venda	46		3.8	3.8	3.8	3.8	African Xitsonga	75				4.2	4.2	4.2
African Sepedi	276			3.9	3.9	3.9	African Sepedi	238				4.2	4.2	4.2
African IsiXhosa	327				4.0	4.0	African Setswana	231				4.2	4.2	4.2
African Setswana	259					4.0	African IsiZulu	438					4.3	4.3
African Afrikaans	29					4.1	African Venda	43						4.3
Significance		0.301	0.053	0.063	0.089	0.108	Significance		0.591	0.560	0.580	0.670	0.610	0.169

Dunnett’s Test (means comparison).

Data source: South African Social Attitude Survey (SASAS) (2006; 2017). *p <.001.

Three socio-economic variables were used to represent “marginality” (economic marginalization). Firstly, economic status was measured by a subjective statement of personal and family wealth, labeled as subjective poverty (Wang et al., 2020). The decision to do this was mostly based on respondents refusing to give their actual income and rather than attempting to impute income, the subjective poverty question was used. Floyd et al. (1994) established that subjective social class provides a more accurate reflection of the impact of socio-economic factors on recreation choices and although there are differences between subjective and objective income measures, Wang et al. (2020) showed that income and subjective poverty measures are associated. To measure subjective poverty respondents had to classify themselves and their family as being either wealthy, very comfortable, reasonably comfortable, just getting by, poor or very poor. (categorical variable). Two other socio-economic variables considered important in influencing travel destination preference (Uvinha et al., 2017), included employment (categorized as employed, unemployed) and educational attainment (categorized as junior primary and below, senior primary, incomplete secondary, complete secondary, and tertiary).

To explore the relevance of socio-demographics, age has proven to predict tourism preferences in general (Tomić et al., 2019), but also in developing countries such as South Africa (Dzikiti & Leonard, 2016) and Kenya (Wangari, 2017), and was included as a continuous variable. Cohabitation has also proven to be important in destination/activity preferences (Kasim et al., 2013) and was included as a categorical variable (married, widowed, divorced/separated, never married). A variable measuring in which geotype a person resides (categorized as urban formal, urban informal, traditional authority areas, and rural farms) was also included since geography is strongly associated with ethnic identity (du Preez & Govender, 2020). All of the independent variables were present in both the 2006 and 2017 datasets and could therefore be included in the analysis. Three separate models were constructed for preferences for Game Parks, Beaches, and VFRs. Since there was only one dependent variable in each model, a UNIANOVA was performed which provides a regression analysis and analysis of variance for one dependent variable by the selected independent variables. The test determines the equality of a composite of the means (optimized to yield the maximum possible F-ratio) across groups (UCLA Statistical Consulting Group, 2020).

After obtaining a statistically significant result for a specific main effect or interaction, the univariate F tests for each variable were examined to interpret their respective individual effects. It is critical to measure the size of the effects of each of the variables since this quantifies the proportion of variance that is explained by the different socio-demographic variables in terms of the respective destination preferences. Partial eta squared statistics were calculated and presented to illustrate the effect size (Pallant, 2016). In all models, all variables were found to be statistically significant (p < .001). This is typically found with larger sample sizes and in this type of study where the data is weighted to the population (Lin et al., 2013). Because of this, the partial eta squared (ηp²), was interpreted and an effect size of .01 considered as small, .06 as medium, and .14 as large effect size (Foster et al., 2018). In the sections below, the discussion will first consider the main effect variables and thereafter the interaction effects.

Results and discussion

The section first presents the main effects for race and ethnicity and provides more detailed description for these two variables through the Dunnett T3 test results. This is followed by a discussion of the main effects for the socio-economic variables, before discussing all the interactional effects. Lastly, socio-demographic variables that encourage travelers to transition to destinations outside their travel habitus are illustrated using two ethnic groups with the most opposing preferences, as examples.

Main effects

Race and destination preference (H₁, H_2a)

From Table 2, it is evident that race was a significant predictor in all three models (p < .001) but its contribution was minimal in discriminating between destination preferences. The contribution of race in Model 1 (preference for Game Parks) in 2006 was almost non-existent (ηp² = .001) and remained weak in 2017 (ηp² = .002). Considering race in Model 2 (preference for Beach destinations), it was again evident that race was significant in terms of differentiating between the races but as in the previous model, the contribution to the model in 2006 was again very weak (ηp² = .001); it increased slightly in 2017 (ηp² = .009). Results for Model 3 similarly showed that the contribution of race was weak in explaining preference for VFR in 2006 (ηp² = .001) and remained weak in 2017 (ηp² = .002). Despite its contribution being weak, the results show that race is a significant predictor and confirms the literature on race and tourism which concludes that race plays a role in determining preferences and impacts tourism choices (Stodolska et al., 2019). H₁ is therefore supported.

Table 2. UNIANOVA model to determine preferences for Game Park, Beach and VFR.

	2006			2017
Main effect	Model 1 Game Park enthusiasts	Model 2 Beach enthusiasts	Model 3 VFR enthusiasts	Model 1 Game Park enthusiasts	Model 2 Beach enthusiasts	Model VFR 3 enthusiasts
	ηp²	ηp²	ηp²	ηp²	ηp²	ηp²
Corrected model	.347*	.335*	.334*	.423*	.442*	.449*
Intercept	.913*	.938*	.944*	.918*	.961*	.966*
Education	.018*	.005*	.007*	.001*	.004*	.017*
Poverty status	.005*	.005*	.007*	.001*	.006*	.005*
Geotype	.009*	.008*	.003*	.010*	.001*	.013*
Race	.001*	.001*	.000*	.002*	.009*	.002*
Ethnicity	.070*	.041*	.018*	.065*	.033*	.056*
Employment	.006*	.006*	.010*	.008*	.004*	.013*
Cohabitation	.003*	.001*	.001*	.004*	.002*	.000*
Age	.002*	.031*	.004*	.002*	.052*	.001*
Interactional effect	Model 1 Game Park enthusiasts	Model 2 Beach enthusiasts	Model 3 VFR enthusiasts	Model 1 Game Park enthusiasts	Model 2 Beach enthusiasts	Model VFR 3 enthusiasts
Age * Poverty status	.016*	.014*	.017*	.030*	.035*	.034*
Age * Ethnicity	.040*	.035*	.047*	.092*	.052*	.053*
Age * Employment	.009*	.008*	.011*	.024*	.016*	.025*
Age * race	.002*	.004*	.002*	.004*	.001*	.002*
Education * Poverty status	.007*	.010*	.016*	.034*	.027*	.022*
Education *Ethnicity	.036*	.036*	.042*	.032*	.029*	.044*
Poverty status * Ethnicity	.043*	.046*	.042*	.061*	.079*	.063*
Poverty status * race	.001*	.001*	.000*	.003*	.004*	.000*
Ethnicity*Employment	.011*	.013*	.011*	.028*	.039*	.042*
Age * Poverty status * Ethnicity	.079*	.086*	.099*	.129*	.102*	.147*
R Squared	.407	.367	.428	.455	.491	.475

Data source: South African Social Attitude Survey (SASAS) (2006; 2017). *p <.00.

However, when controlling for other variables, it is clear from the data that race is not the most important predictor of destination references. Hypothesis 2_a is therefore not supported.

Despite being weak, the differences between race groups and preference for destinations remain statistically significant. To test the differences between race groups and destination preference in more detail, the Dunnett T3 test was used to determine which specific race groups differed from each other (results not displayed). In model 1 (Game Parks), results showed that Whites had the highest mean preference score both in 2006 and 2017, and was significantly different from all other race groups. The Black African and Indian/Asian groups had slightly lower mean scores than the White group. The Colored race group was significantly different from all other race groups, having the lowest mean score in terms of preference for Game Parks. In 2017, the preference score from the Black African group had decreased, widening the gap between preference for Game Parks between the White group and other race groups. In 2017, the preference score from the Black African group had decreased, widening the gap between preference for Game Parks between the White group and other race groups. Continued limited visitation of nonwhite groups to South African national parks has been found in other studies (Kruger & Douglas, 2015; Musavengane & Leonard, 2019; Smith, 2023).

For Model 2 (Beach), in both 2006 and 2017, the Colored group was most interested in Beach activities, followed by the Indian/Asians, Black African, and White groups. This finding is not surprising given that the majority of the Colored community resides in coastal areas in the Western Cape Province, as well as the Northern and Eastern Cape that have coastlines (Alexander, 2021). The Colored people also have an intimate historical relationship with the sea (Buchanan & Hurwitz, 1950). As was the case with interest in Game Parks, the Black African groups became less interested in Beach activities over time and in 2017, this race group had significantly lower preference scores than all other race groups. This is contrary to what would have been expected namely that interest in these destinations would have increased post-apartheid given the abolishment of discriminatory legislation.

Model 3 (VFR) indicated that preference for VFR was highest among Black African respondents, followed by Colored respondents, Indian/Asian, and then White respondents. This positioning remained similar between 2006 and 2017. Results confirm that VFR was historically not severely curtailed by discriminatory policies in South Africa (Grundlingh, 2006; Rogerson, 2015).

Ethnicity and destination preference (H_2b)

From Table 2 it is evident that when considering ethnicity alongside other main variables, ethnicity had the most explanatory power in terms of distinguishing preference in all three Models, both in 2006 and 2017. In Model 1, the partial eta squared for ethnicity in 2006 was medium (ηp² = .07), and in 2017 it was similar (ηp² =.065). In Model 2, ethnicity as a main variable had the largest effect size in 2006 (ηp² = .041) and (ηp² = .033) in 2017, which was second only to age. This is in line with other studies that show preference for Beach destinations is linked to age, specifically with younger age cohorts (Wu et al., 2019).

As in Models 1 and 2, ethnicity was the most important contributor as a main variable in distinguishing levels of preference for VFR (Model 3). In 2006 the effect size of ethnicity was large (ηp² = .018) and it became even more pronounced in 2017 (ηp² = .056), The data show that among the main effect variables or contributors, ethnicity had the largest influence in all three models and therefore had the biggest discriminatory power in explaining preferences for Game Parks, Beach, and VFR in the analysis. Hypothesis 2_b is therefore supported

Given that ethnicity has the largest contribution to preferences in terms of a main variable, it is worth considering the ethnicity variable in more detail for each of the models in 2006 and 2017 (results displayed in Tables 3–5). The Dunnett’s mean comparison test between 2006 and 2017 shows that the ethnic groups in the various mean clusters have basically remained the same.

In the case of Game Parks (Table 3), in both 2006 and 2017, the Colored minority remained in the lowest cluster. Likewise, apart from the IsiNdebele group, the Nguni language groups had low scores in both 2006 and 2017. Sotho-speaking groups fell in the middle clusters in both 2006 and 2017. Nullifying popular perceptions that Game Parks are mostly preferred by White Afrikaans speakers, Venda and Xitsonga African ethnic groups had a higher mean score than the White Afrikaans-speaking minority in both 2006 and 2017. Although this may go against the popular notion, this is not surprising given that Xitsonga and Venda speaking Africans are culturally attached to nature and have an intimate knowledge of their natural surroundings and local fauna and flora (Anthony, 2006; Constant & Tshisikhawe, 2018; Fairer-Wessels, 2008). From the results, it seems evident that geographic proximity, a factor often considered in ethnic classification, plays a central role in destination preferences. The Colored minority is a group that resides mostly in Cape Town, which is geographically far removed from Game Parks (specifically the Kruger National Park) (approximately 1100 miles), and can therefore be considered as outside of the cultural boundary or travel habitus of this group. Conversely, Tsonga and Venda groups are adjacent to the biggest game park in South Africa, namely the Kruger Park. Geographical proximity is therefore an element that potentially plays a role when considering cultural distance factors for tourism preferences (McKercher, 2018; Waters, 2018).

Looking at the different ethnic groups and their preference for Beach destinations (Table 4), it was evident that preferences remained relatively similar between 2006 and 2017. In contrast to Game Parks, Venda and Xitsonga ethnic groups had the lowest mean scores for Beach. These groups are also physically removed from Beach areas and proximity again appears to play a role). Based on Dunnett’s Test of mean comparisons, the Nguni and Sotho groups were in the middle groups in both 2006 and 2017, indicating neither high nor low preference for Beach destinations. Non-Black African minorities had the highest mean score on preference for Beach destinations, which is not unexpected given that these groups have a higher per capita income which is typically associated with preferences for Beach destinations (Gradin, 2015; Hino et al., 2018). Contrary to Game Park preferences, Colored Afrikaans and English speakers had the highest mean scores when considering preference for the Beach. Again, geographic proximity seems to heighten preferences, given that the majority of the Colored community resides in coastal areas in the Western Cape Province (Alexander, 2021).

Contrary to Models 1 and Models 2, Black African ethnic groups tended to have higher mean scores on than other groups Model 3 – indicating higher preference for VFR relative to the other minority groups (Table 5). Conversely, White English and Afrikaans speakers scored lowest, thus demonstrating more disassociation with VFR than other ethnic groups. Although VFR is typically associated with lower income groups (Kasim et al., 2013; Rogerson, 2015), a history of migratory work patterns among Black Africans also potentially entrench patterns of VFR (Rogerson, 2015).

Socio-economic status and destination preference (H_2c)

As mentioned previously, literature has shown that economic status plays a significant role in preferences for destinations (Aizlewood et al., 2006; Sevilla et al., 2012; Stodolska et al., 2019). Other socio-economic variables such as education and employment has also been proven to impact destination choice and as such these variables were included in the multivariate analysis in an attempt to test the marginality hypothesis. Results show (Table 2) that despite being statistically significant, the contribution of wealth (measured by the subjective poverty question), was weak both in 2006 and 2017. Education and employment also had very weak explanatory power. Despite the weak contribution, it is worth mentioning that in Model 1, education contributed positively (ηp² = .018); implying that higher education was associated with Model 1 (preference for Game Park destinations). At the same time, education (ηp² = .017) had a negative association with VFR as well as employment both in 2006 VFR (ηp² = .01) and 2017 (ηp² = .01). Despite being significant, it is however evident that the effect size of socio-economic variables as a main effect variable was negligible, with regards to all three models. Hypothesis 2_c is therefore not supported.

Interactional effects

Overall, the interactional effects between variables accounted for more of the variance than main effect variables. The R² for main effects for Game Parks was 0,132 in 2006 and 0,112 in 2017. For interactional effects it was 0,339 in 2006 and 0,415 in 2017. For Beach, the R² for main effects was 0,109 in 2006 and 0,159 in 2007 while it was 0,329 for interactional effects in 2006 and 0,440 in 2017. For VFR the R² for main effects 2006 was 0,080 and 0,125 in 2017. It was 0,334 in 2006 and 0,442 in 2017 for interactional effects. Reporting the traditional ANOVA source table and discussing the associated significance levels is merely the beginning of an analysis, since there is much more to be gained from analyzing, detailing, and considering the interactional effects between variables (Brown, 2008). In the next section, the interactional effects for each of the different preferences or models will be discussed.

Preference for Game Parks (model 1)

Preferences for travel and tourism activities are complex and preferences cannot merely be explained by a single domain factor but rather by the interplay of demographics variables. The interactional effect of age and poverty status in Model 1 was significant in both 2006 (F(20,28440832) = 20008.647, p < .001, ηp² = .016) and 2017 (F(20,20573154) = 20291.465, p < .001, ηp² = .03). In 2006, the 16–24 age group regardless of poverty status had the highest scores but in 2017 the trend shifted with poverty status rather than age dictating interest. Other socio-economic status variables such as poverty status, educational attainment, and employment had a significant (and positive) association with preference for Game Parks. Higher education, income, and better employment when interacting with age and ethnicity showed positive associations. This pattern was observed in 2006 but was even more pronounced in 2017. The effect size of the interactional effect of age and ethnicity was small in 2006 (F(51,28440832) = 19426.719, p < .001, ηp² = .04) but medium in 2017 (F(49,20573154) = 27301.264, p < .001, ηp² = 0.092). Descriptive analysis of this finding showed that regardless of age group, Xitsonga, Tshivenda, and White Afrikaans respondents had the highest mean scores for Game Park preference.

The biggest interaction and explanatory power for Game Parks for both 2006 and 2017 was found for the interaction between age, poverty status, and ethnic group. In 2006 this contribution was large (F(151,28440832) = 13700.449, p < 0.001, ηp² = .079) and in 2017 even larger (F(132,20573154) = 14802.309, p < .001, ηp² = .129). When further analyzing the interactional effect of these variables descriptively, checking mean difference using the Dunnett’s Test, the data show that the change over time has not brought distinctions between ethnic groups but rather between the youngest age groups within ethnic groups. Regardless of poverty status (from wealthy to very poor), the youngest cohort (between 16–24 years of age) of the Sotho, Nguni, White Afrikaans, White English, and Indian groups had the highest increase in scoring over time; indicating an increased preference for Game Parks. This suggests a generational effect, with the younger cohorts of most ethnic groups showing an increase in this preference.

Preference for Beach (model 2)

The interactional effect of age and poverty status for Model 2 was significant in both 2006 (F(20,29116602) = 15713.777, p < .001, ηp² = .014) and 2017 (F(20,19740697) = 14693.46, p < .001, ηp² = .035). As was the case in Model 1, the 50+ age cohort was statistically different from other age groups in both 2006 and 2017 showing an inverse association between age and interest in Beach. The interactional effect between age and ethnicity showed that among all ethnic groups, younger respondents showed highest preference for this activity. The youngest cohort (16–24 years), regardless of poverty status or employment, was positively associated with the dependent variable.

In 2006, the poor or very poor, those just getting along, as well as the wealthy significantly differed from each other with those incrementally wealthier and younger more interested in Beach preferences. In 2017, a form of confluence took place between those just getting along and the wealthy who became equally interested in Beach destinations. Education and being self-employed were positively associated with a preference for the Beach, with the young cohorts (16–34 years) with a matric or post-matric qualification exhibiting highest mean score for interest in Beach destinations.

Interaction between age cohort, poverty status, and ethnicity again proved to be the strongest predictor of preference for Beach destinations with a large effect size (F(152,29116602) = 13392.348, p < .001, ηp² = .086) in 2006 and even larger (F(127,19740697) = 7285.15, p < .001, ηp² = .102) in 2017. The results show an inverse relationship between interest in Beach destinations and age. The youngest cohorts (16–34 years old) among all ethnic groups had higher mean scores on this dimension than older age cohorts.

Preference for VFR (model 3)

The interaction between age and poverty status was significant in Model 3 in both 2006 (F(20,30087140) = 21364.912, p < .001, ηp² = .017) and in 2017 (F(20,21100000) = 20816.226, p < .001, ηp² = .034). This interaction yielded the opposite effect than illustrated for Beach destinations, with the poor or very poor as well as older individuals most interested in VFR. The socio-economic variables education and employment were also negatively associated with the dependent. The interaction between the socio-economic variables such as poverty status and employment status had significant but nominal effect sizes and it was mostly the unemployed, discouraged work seekers, the disabled, and pensioners who measured high on this dimension. An inverse relationship between education and preference for VFR was evident, with people with no or only a primary school education having a greater preference for VFR than people with higher education levels.

In 2006 the poverty status-ethnicity interrelationship (F(51,30087140) = 21680.314, p < .001. ηp² = .042) revealed that those that regarded themselves as very poor or poor had higher mean scores and thus a preference for VFR, the only exceptions being the Venda and English-speaking groups. In 2017 the relationship between poverty status and ethnicity (F(50,21100000) = 16159.528, p < .001, ηp² = .063) showed that for all ethnic groups, the tendency was to be less interested in VFR as poverty status increased to being less poor. As with model 1 and model 2, the three-way interaction between age, poverty status, and ethnicity explained and contributed the most to the model. In 2006 the contribution was large (F(155,30087140) = 17734.818, p < .001, ηp² = .099) and in 2017 even larger F(133,21100000) = 15582.124, p < .001, ηp² = .147. However, contrary to Model 1 and Model 2, the relationship between age, poverty status, and interest were in different directions. Those interested in VFR were typically associated with the older age cohort (50 + years) and those whose considered themselves to be very poor or poor. The youngest age cohort (16–24 years) who considered themselves to be not very poor or poor were less interested in VFR.

Transition to destinations considered to be outside their travel habitus

The aim of RQ 3 was to determine which socio-demographic variables act as enablers for travelers to transition to destinations considered to be outside their travel habitus. In order to answer this research question, preferences for the different destinations were analyzed by the socio demographics in detail and it was found that the youngest cohorts (16–34 years) and the non-poor were more interested in destinations geographically far removed from the group’s location and outside the typical travel habitus of the group. Older and poor respondents were more interested in destinations that were closer in terms of geographical proximity and popular within the travel habitus. To illustrate this, the Tsonga and Colored groups are examined. These two groups were chosen given that they have opposing preferences for Game Parks and the Beach. Of all ethnic groups, the Colored group had the lowest mean score for preference for Game Parks and the highest for the Beach. Contrary, the Tsonga group had the highest mean preference score for Game Parks and the lowest mean score for the Beach. These groups are thus distinct in their preferences.

The Tsonga ethnic group (Figure 1) is geographically located inland, near the largest national park in South Africa namely the Kruger National Park (KNP). This ethnic group had the highest mean preference scores for Game Parks which can be attributed the fact that Game Parks form part of their cultural boundary or typical habitus. At the same time their physical location is far removed from Beach areas, which could explain the lack of interest in Beach destinations, given that these areas are outside their cultural boundary or typical habitus. In the following two graphs, the intersectionality between age, wealth and interest in Game Parks (considered to be within cultural habitus of the Tsonga group) and Beach (considered to be outside the cultural habitus of the Tsonga group) are illustrated. From the first graph it is evident that poorer Tsonga respondents had greater interest in Game Parks compared to wealthy Tsonga respondent. Contrary, wealthy Tsonga respondents had a greater interest in Beach destinations compared to poorer respondents. When considering age in the second graph, it is evident that those younger than 50 years showed higher interest in both destinations but among those 50 years and older, a preference for Game Parks rather than Beach areas was apparent. The data suggest that the wealthy and the young show more interest in destinations that can be considered as outside their cultural boundary or typical travel habitus (Figure 1).

Figure 1. Travel preference among Tsonga people by age and poverty status (0–5 scale).

In the next two figures (Figure 2), similar analysis is undertaken but based on a different ethnic group (Colored group) with a different geographic location and habitus. The Colored ethnic group is located near beach areas (Cape Town and surrounding area) and have a close association with the Beach. Geographically they are far removed from the Kruger National Park (1100 miles) and Game Park areas are considered outside the typical habitus of Colored respondents. The following graphs show that among poorer Colored respondents, interest in Beach areas was high relative to wealthy Colored respondents while the wealthy Colored respondents had a heightened interest in Game Parks. Young Colored respondents (16–24 years) showed higher interest in Game Parks compared to the other age groups. Again, the data show that the wealthy and the young tend to show more interest in destinations that can be considered as outside the travel habitus (Figure 2).

Figure 2. Travel preference among Colored people by age and poverty status (0–5 scale).

Given the evidence in this section, certain characteristics (being wealthy and young) seem to facilitate venturer notions (Plog, 1974). On the other hand, the older (50+ years) and less affluent travelers exhibited tendencies associated with dependables (Plog, 1974). They tend to prefer destinations that could be considered culturally familiar and part of their habitus. This study therefore contributes to the theory of Plog (1974) by explicating its applicability in terms of the movement of social groups (and not just individuals) into new and culturally distant destinations; and contributes to understanding how groups potentially migrate to new travel destinations. Although this finding is based on descriptive findings, it is worth considering for future interrogation and testing.

Conclusion

The dynamics associated with the travel destination choice of adults in an emerging market, with a history of legislative discrimination and high levels of material inequality, were investigated in this study. Despite the abolishment of legislated racial discrimination, differences in racial and ethnic destination preference were evident. Symbolic boundaries, which might have been established due to segregation policies; creating a form of travel habitus, remain despite the removal of legislative barriers. Changes in destination preferences should therefore not be assumed as a mere consequence of the abolishment of legislated segregation leisure policies.

Given the history of South Africa, the authors were particularly interested in testing tourism preferences against Washburne’s (1978) interpretation of subculture theory. The most significant finding pertaining to this theory was that the more refined ethnicity classification, rather than the overarching race variable, was a better predictor of preference for the types of destination preferences addressed in the study. This finding supports findings from other researchers that caution against the homogenization of racial groups (Floyd & Stodolska, 2014; Garcia, 2014; Stodolska, 2018). Another finding from this paper shows that unidimensional, single, or main effect variables only had small effect sizes while medium effect sizes were noted for interactional effects. Neglecting distinctions between ethnic groups and measures of cultural, limits our understanding of more refined patterns of participation and marginalization in leisure (Aizlewood et al., 2006; Krymkowski et al., 2014; Shores et al., 2007) and tourism (Benjamin et al., 2016, 2024).

Some of the findings from this study reiterate existing knowledge about specific destination preferences. In line with many other studies such as Wu et al. (2019), preference for Beach was associated with age. Further, the two populations groups with greater interest were those that live in provinces with coastlines (Coloureds) (Alexander, 2021), or with strong cultural ties (Indian) (du Preez & Govender, 2020), confirming the role of travel habitus in destination choice. Socio-economic realities dictated VFR preferences, where it was mostly associated with lower income (as also stated by Kasim et al., 2013 and Rogerson, 2015) as well as higher age.

In terms of Game Parks, national and international findings are supported that, as a race group, White travelers tend to dominate interest in Game Parks (Butler & Richardson, 2015; Kruger & Douglas, 2015; Musavengane & Leonard, 2019). However, certain Black ethnic groups (those residing adjacent to the largest National Game Park – Kruger Park) had the same levels of preference for Game Parks as White travelers. The history of Kruger National Park in particular, but also Game Park destinations in general, perpetually underrepresent Black Africans’ indigenous knowledge of, and cultural affinity toward nature conservation (Smith, 2023). This corroborates Pinckney et al.’s (2024) who state that environmentalism needs to be viewed from a cultural perspective. Interest in Game Parks was also associated with age (the younger generation), poverty status, and education. Growing interest in these natural attractions among younger members of the emerging domestic market indicates that the visitor profile can be diversified to sustain these natural areas into the future. Krymkowski et al. (2014) similarly found evidence that disparities in visitations by different ethnic group to national parks in the US showed a decline among younger respondents.

Travel habitus is critical in understanding the tourism consciousness (Musavengane & Leonard, 2019) and factors such as geographic location and cultural distance determine the extent to which new destinations are considered by racial and ethnic group members. Acculturation and ethnocultural identities among the youth could help facilitate diversification of the domestic tourism market (Du Preez & Govender, 2020). South Africa is a youthful nation with the post-1994 generation (born frees) comprising almost a third of the South African population (Statista, 2024) and removing barriers to participation among the South African youth post-apartheid (Dzikiti & Leonard, 2016) is key to tourism growth and development. This change and potential for diversification of the domestic tourism landscape is facilitated by changes to marginalization and power dynamics that have changed over time (Abdullah et al., 2023).

Limitations and recommendations for future research

The study has several limitations that also pose opportunities for future research. The study was limited to Game Parks, beach visits and VFR because they would be best known to all members of the South African population. While the subculture theory was appropriate in defining the formation of interracial groups within the South African context, it presents a limited perspective on the dynamic process of racial and ethnic identities. Future studies should consider other relevant theories such as critical race theory (Mura & Wijesinghe, 2023) and perspectives on marginalization where individuals’ experiences are recognized (Abdullah et al., 2023; Dillette et al., 2019; Thakur et al., 2023). Closer scrutiny should also be given to linking the severity of historical place-based discriminatory policies to destination preferences. For instance, while certain beaches were opened to people of color during apartheid, Game Parks s were not. The experiences of Black Africans when using public transportation for travel, leisure and recreational purposes left historical images that challenge tourism and leisure participation post-apartheid (Smith, 2023). The study is limited by quantitative methodology and qualitative methods will provide deeper insight into the trends observed. In addition, though the chosen data analysis technique allowed comparisons between groups, limited combinations and permutations of the explanatory variables were possible. It is recommended to apply choice models such as multiple hierarchy stratification to overcome this limitation. The study used secondary data from only two cross-sectional surveys while trends are more reliably identified through longitudinal studies with more iterations. Furthermore, the time periods (2006, 2017) do not reflect preferences immediately post-apartheid in 1994, nor current preferences, that may present a more holistic perspective given the extent of change in the social structure of the country right across the post-apartheid era.

Ethics statement

A HSRC Research Ethics Committee (REC) was established on 27 November 2002. The HSRC REC is registered with the South African National Health Research Ethics Council of the SA National Department of Health (NHREC No 290808-015) and also has US OHRP Federal-wide Assurance (FWA) accreditation (FWA 00006347, IRB No. 00003962). From 2003 all HSRC researchers are required to submit all research applications for REC approval before commencement of research. Ethics approval is given to project proposals for a period of a year, after which the PI must apply for a renewal/recertification. The South African Social Attitude Survey initially applied and received ethical clearance (Protocol No REC 6/22/09/10) on an annual basis up to 2010. In 2010 the ethics committee suggested that the project apply for a class approval. This was suggested given that the SASAS study is a repeat investigation which has already been approved for a couple of years, the participants are of similar demography and vulnerability status and the methodology and content barely change. The class approval was accepted (Protocol No 5/17/08/11). Since this paper was part of a doctoral thesis, the study also received ethical clearance from the tertiary institution.

Acknowledgments

Thank you to Dr. Marthi Pohl for statistical support. Thank you to the HSRC for collaboration on the research.

Disclosure statement

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