1. Introduction
Tourism is a major force in global trade that plays a vital role in the social, cultural, and economic development of most nations (
Smith 1995). According to statistics compiled by the World Travel and Tourism Council, in 2019, the scale of the global tourism industry reached USD 8.9 trillion, with a contribution rate of 10.3% to the world’s gross domestic product. At the same time, the industry employed 330 million people worldwide, accounting for approximately 10% of global employment. A country’s tourism market generally consists of two markets with different customer sources, namely, inbound and domestic tourism. The domestic tourism market gradually expands with economic growth, increases in residents’ income, and adjustments to vacation arrangements. According to the World Tourism Organization, the scale of the domestic tourism market is 10 times that of the international market (
Page et al. 2001). Therefore, domestic tourism contributes significantly to a country’s tourism revenue.
If one considers the example of Taiwan, in 2019 the number of inbound tourists reached 11.86 million, of which 90% were from within Asia, and the tourism revenue amounted to USD 14.411 billion (
Tourism Bureau 2020). There were 169 million domestic travelers, 14.24 times the number of inbound tourists, although the tourism revenue was USD 12.698 billion, or 88 percent of that for the inbound tourism market. The key reason for the substantial disparity in the number of tourists despite identical revenue levels was the differences in tourist behavior between the two tourism markets. The average length of stay of inbound tourists was 6.20 nights, whereas that of domestic tourism was mainly 1.51 days, with 66% choosing to return the same day without staying in accommodation facilities. The low level of demand for accommodation was the main reason why the performance of domestic tourism failed to surpass that of inbound tourism. Therefore, understanding the factors influencing the demand for accommodation on the part of domestic tourists in order to increase the duration of stay is an important topic when it comes to expanding the domestic tourism market.
When establishing an econometric model to discuss the factors influencing the demand of domestic tourists for accommodation, the first issue is to deal with a large influx of tourists who do not spend any money on accommodation. The traditional least squares method assumes that dependent variables have continuity and can be measured. If this approach is used to estimate model parameters when observed values are constrained by censored data, it may result in such parameters being biased and inconsistent (
Maddala 1983;
Judge et al. 1988). As tourism is not necessary for livelihood, the phenomenon of zero expenditure widely exists in research on tourism spending (
Dardis et al. 1994;
Hong et al. 1996;
Cai 1999;
Lee 2001;
Zheng and Zhang 2013;
Weagley and Huh 2004;
Nicolau and Màs 2005;
Jang and Ham 2009;
Alegre et al. 2013;
Bernini and Cracolici 2015;
Sun et al. 2015). This fact makes the choice of appropriate econometric techniques crucial for the consistency of the empirical results (
Maddala 1983;
Amemiya 1984). With regards to zero expenditure in tourism, the models commonly used by scholars include the double-hurdle (DH) model (
Cragg 1971) and the Heckit model (
Heckman 1979). Unlike traditional economic models that consider the purchase and consumption decisions of consumers to occur simultaneously, these two models divide consumer behavior into two decision-making processes, i.e., whether to buy and how much to buy—also referred to as the two-stage decision model. According to the two-stage decision model that is in line with the theory of consumer behavior, consumers will collect information before purchasing products and will use that information as a reference to decide whether or not to buy, and then decide how much to spend once they have made their purchase decision.
Past studies on tourism expenditure reveal that a few of the discussions focus on the demand for tourist accommodation, for example,
Hong et al. (
1996) and
Cai (
1999). However, while both studies have adopted the Tobit model that considers zero expenditure as no consumption (
Su and Yen 1996), they neglect the fact that no consumption may be the result of a lack of willingness to participate. Thus, using the Tobit model to analyze tourist accommodation expenditure may have certain limitations, resulting in an inability to grasp different influencing factors between the intention to make use of and the decision to actually spend money on tourist accommodation. More recently, a few studies have discussed this issue by using a different approach. For example,
Masiero et al. (
2015) utilized a quantile regression model to analyze the relationship between key travel characteristics and the price paid to book the accommodation.
Ismail et al. (
2021) adopt a two-step Chi-square automatic interaction detection (CHAID) procedure to segment spending on accommodation for visitors according to demographic, trip-related, and psychographic factors.
Accommodation is a major component of tourist expenditure (
Laesser and Crouch 2006). However, in the case of domestic tourism, accommodation may not be made use of by everyone, i.e., not all individuals participate in this expenditure activity, thus reporting values of expenditure equal to zero. Therefore, the analytical tool should be adequate to account for a large proportion of observations with a value of accommodation expenditure equal to zero. This study considers a data-oriented approach, employs the nonnested test method and selects an appropriate two-stage decision model to discuss the factors influencing the consumer behavior of domestic tourists in regard to accommodation. By estimating the double-hurdle model, the effects of the associated determinants on the intention to use tourist accommodation and expenditure decisions can be identified. Furthermore, despite numerous empirical studies that examine the determinant factors of total tourism expenses, a particular determinant factor may have varying impacts on a specific expenditure type. The research results may help to improve the economic benefits of the domestic tourism market and serve as valuable reference for relevant businesses in developing marketing strategies.
3. Methodology
3.1. Two-Stage Decision Model
The two-stage decision model is comprised of limited dependent variable models of the participation decision and consumption decision, primarily the DH model (
Cragg 1971) and the Heckit model (
Heckman 1979).
Cragg (
1971) recognized that zero expenditure may be caused by consumers choosing not to participate in the decision-making stage or choosing to participate in the first stage, but not actually spending due to certain factors when it comes to the consumption decision. In other words, the observed values for zero expenditure in the DH model not only exist in the participation decision stage but also in the consumption decision stage. According to
Heckman (
1979), zero spending occurs predominantly during the participation stage, with positive consumption expenditure occurring once consumers make a purchase decision.
3.1.1. DH Model
The idea behind the DH model is that a consumer has to overcome two hurdles before recording a positive expenditure. These two hurdles are: (1) the participation market (potential consumers), and (2) actual consumption (
Angulo et al. 2001). A complete DH model consists of the participation and consumption decisions, with equations set as follows (
Jones 1989;
Aristei et al. 2008):
In Equation (2), a value of larger than 0 and a value of of 1 indicates that consumers decide to participate in the consumption. A value of equal to or less than 0 and a value of of 0 indicates that consumers will decide not to participate in the consumption. is a variable influencing the participation decision.
In Equation (3), is the latent consumption variable and is the variable influencing consumption expenditure. It can be clearly observed from Equations (2) and (3) that zero expenditure can appear in the participation decision stage when consumers choose not to participate or else choose to participate but do not have actual consumption expenditure.
Assuming that the error terms of the participation decision and consumption decision equations are mutually independent, the log-likelihood function of the independent DH model can be expressed as follows (
Moffatt 2005;
Aristei et al. 2008):
In Equation (4), is the cumulative distribution function, is the standard normal density function, 0 means zero consumption, and + means that the consumption value is positive.
Assuming that the error terms of the participation and consumption decision equations are correlated and that simultaneous participation and consumption decisions are possible, the bivariate normal distribution of the error terms of the two equations of the DDH model is as follows:
In Equation (5),
is the degree of correlation between the error terms of the participation and consumption decision equations. After adding the correlation coefficient, the log-likelihood function of the DDH model is as follows (
Jones 1992):
The data distribution of limited dependent variables often reveals a significant positive skew, which is therefore unable to fulfill the hypothesis of a normal distribution of error terms. Therefore, if the maximum likelihood method is used to estimate the model, it is not possible to maintain parameter consistency. Through the inverse hyperbolic sine (IHS), dependent variables can generate consistent parameter estimates for model estimation (
Newman et al. 2001). The IHS conversion function is as follows:
After the dependent variables are converted through the IHS, the log-likelihood function of the DH model can be expressed as follows:
3.1.2. Heckit Model
Heckman (
1979) proposed a two-step estimation method to resolve the problem of sample selection bias caused by using observable sample data. The two-step estimation method first uses the probit method to estimate the coefficients of all observed values and calculates the inverse Mills ratio (IMR). It has subsequently used the ordinary least squares method to estimate nonzero observed values, to include the IMR as an explanatory variable, and to estimate the coefficients of the model. The Heckit model mainly comprises a selection equation and an outcome equation:
In Equation (9),
is the latent variable,
is the explanatory variable influencing participation and consumption, and
is the corresponding coefficient. Equation (9) reflects the relationship between
, the latent variable of the selection mechanism, and
, the dichotomous dummy variable actually observed (
Huang and Wang 2016).
In Equation (11), is the latent consumption expenditure variable, is the observed consumption expenditure variable, is the variable influencing consumption expenditure, and is the corresponding coefficient. The Heckit model assumes that the error terms () of the selection equation and the outcome equation are correlated, with the degree of correlation being expressed by . The normal distribution of the error terms of the two equations is represented in Equation (5).
Apart from the two-step estimation method, the Heckit model can also adopt the maximum likelihood method to estimate the parameters, and its log-likelihood function is as follows (
Aristei et al. 2008;
Wodjao 2007):
3.2. Description of Data and Variables
This study employs domestic tourism data from the “Survey of Travel by R.O.C Citizens” conducted by the Tourism Bureau of the Ministry of Transportation and Communications of Taiwan from 2014 to 2018. The sample covers 60,817 individuals, with 26,085 having tourist accommodation and an average accommodation expenditure of NTD 1824. As for the dependent variables, the discrete nature of the decision “having accommodation” is represented as a dichotomous variable, in such a way that it takes a value of 1 if tourists have accommodation, and 0 if otherwise. This variable, related to accommodation expenditure, is found by a quantitative variable that represents the cost incurred during the accommodation. The six categories of explanatory variables are described as follows.
1. Economic factor: The individual’s average monthly income. This variable is divided into six categories: no income, under NTD 30,000, NTD 30,001–50,000, NTD 50,001–70,000, NTD 70,001–100,000, and over NTD 100,001 (
Table 1). The group with less than NTD 30,000 in average monthly income accounts for the largest proportion at 39.0%, followed by NTD 30,001–50,000 at 27.62%.
2. Social stratum: Education level and occupation. The education level is divided into five categories, namely, elementary (junior) high school and below, senior high (vocational) school, college, university, and postgraduate school or above, with the level of elementary (junior) high school and below as the benchmark for comparison. Among the five categories of education level, university accounts for the largest proportion at 31.38%. Occupation is divided into five categories as follows: white-collar worker, blue-collar worker, housewife, retiree, and others, with the blue-collar worker as the benchmark. Among the five categories of occupation, blue-collar workers account for the largest proportion at 45.33%.
3. Family life cycle: Includes variables, such as gender, traveling companions between the ages of 7 and 11, traveling companions between the ages of 0 and 6, marital status, and age. In terms of gender, females make up the majority, accounting for 56.67%. Marital status is divided into three categories, namely, unmarried, married, divorced/ separated, or widowed, among which the married group accounts for the largest proportion at 71.49%. Age is divided into seven categories, with 20–29 as the benchmark, and the 40–49 age group accounts for the largest proportion at 22.0%. The average number of children is 0.2 for the groups “traveling with children between the ages of 0 and 6” and “traveling with children between the ages of 7 and 11.”
4. Residential area: This study classifies the residential area of respondents into five regions, namely, northern, central, southern, eastern, and other regions. Among them, the northern region accounts for the largest proportion at 43.45%, with other regions being used as the benchmark.
5. Tourism behavior: Includes the days of the trip, travel season, travel date, and favorite activity during the trip. The average days for domestic trips are 1.72 days. There are four travel seasons, and individuals primarily travel in the first season, which accounts for 27.5%. The travel date is divided into national holidays, workdays, weekends, and Sundays; most individuals travel during weekends and Sundays, which accounts for 54.25%. Favorite activities during the trip include sightseeing, cultural experience, sports, visiting amusement parks, tasting food and snacks, visiting family and friends, and others. Among them, sightseeing accounts for the largest proportion at 40.45% and visiting amusement parks accounts for the smallest proportion at 2.04%.
6. Vacation policy: The Taiwanese government has implemented a leave policy that enforces a five-day work week with “one fixed day off and one flexible rest day” since December 2016.
4. Results and Discussions
This study uses four two-stage decision models, namely, the Heckit model, DH model, DDH model, and IHS DH model. Moreover, it adopts the nonnested Vuong testing method to select models suitable for the demand for accommodation in domestic tourism.
Vuong (
1989) used the log-likelihood function value as the basis, applied simple conversion equations, and proposed modified likelihood ratio testing for the nonnested maximum likelihood estimation. This study uses
STATA software to perform the maximum likelihood estimation for limited dependent variable models, namely, the Heckit model, DH model, DDH model, and IHS DH model. The final log-likelihood function values of various models are depicted in
Table 2, and these figures are further tested via nonnested specification tests. In terms of the nonnested test for the Heckit model vs. the DH model, the Vuong value is 3.21 (
Table 3), indicating that the Heckit model is significantly better than the DH model. In terms of the nonnested test for the Heckit model vs. the IHS DH model, the Vuong value is 24.18, indicating that the Heckit model is better than the IHS DH model. In terms of the nonnested test for the Heckit model vs. the DDH model, the Vuong value is −102.78, indicating that the DDH model is better than the Heckit model. It can be determined through a series of nonnested tests that the DDH model is significantly better than the Heckit model, DH model, and IHS DH model. Based on the above results of the specification tests, of the four limited dependent variable models, this study suggests that the DDH model is more appropriate for explaining the decision-making behaviors in relation to the intention to use and the expenditure on accommodation in domestic tourism.
4.1. Results of Participation Decision
Table 4 depicts the estimated coefficients of the DDH model with regard to the decisions on the intention to use and the expenditure on accommodation in domestic tourism. The Wald test (
Table 5) and
Table 4 reveal that the variables for the social stratum, family life cycle, tourism behavior, residential area, and vacation policy have a significant impact on people’s intention to use accommodation in domestic tourism, supporting hypotheses H1a, H1b, H1c, H1d, and H1e.
As regards to the individual variables, we first observed the impact of the variables for the social stratum on the intention to use tourist accommodation. There is a positive relationship between the education level and the intention to use tourist accommodation with the coefficients of the variables for the four education levels being significantly different from 0, of which the group with a university level education (EDU4) has the highest intention to use tourist accommodation in domestic tourism, followed by the group with a postgraduate school or above education level (EDU5). As for the occupation variables, the occupation of students and unemployed (OCU5) is used as the benchmark, and the variable coefficients for white-collar workers (OCU1), blue-collar workers (OCU2), and retirees (OCU3) are significantly different from 0. Through observing the estimated coefficients of the occupation variables, the white-collar group has the highest intention to use tourist accommodation, followed by the blue-collar group, indicating that employed workers have a relatively high demand for vacation and tourism quality beyond their busy schedules, whereas the group of students and unemployed has the lowest intention to use tourist accommodation. The results related to education level and occupation variables are consistent with previous studies (
Nicolau and Màs 2005;
Jang and Ham 2009;
Alegre et al. 2013;
Bernini and Cracolici 2015).
With respect to the family life cycle, females have a significantly higher intention to use tourist accommodation compared to males. The numbers of traveling companions between the ages of 0 and 6 (A06) and 7 and 11 (A711) have a significant positive impact on the intention to use tourist accommodation. In terms of the marital status variables, the married group (MAR2) has the highest intention to use tourist accommodation with a significant estimated coefficient; the unmarried group (MAR1) has the lowest intention to use tourist accommodation with an insignificant estimated coefficient. In terms of the age variables, the 12–19 age group (AGE1) has the highest intention to use tourist accommodation and the over 70 age group (AGE7) has the lowest intention to use tourist accommodation, with a coefficient that is significantly different from 0. As age increases, the intention to use tourist accommodation declines (
Figure 2). With regard to the residential area, the eastern region (RE) is used as the benchmark, and among the four residential areas, only the variable coefficient for other regions (RO) reaches the significance level. From the perspective of the estimated coefficients, tourists residing in the southern region (RS) have the highest intention to use accommodation, and those residing in other regions have the lowest intention to use accommodation. The results provide proof for the argument of
Jang and Ham (
2009) and
Bernini and Cracolici (
2015) that the family life cycle, and in particular, having children in the household, is a determinant of the travel decision and, as a result, of the accommodation decision.
In terms of the tourism behavior variables, the variable coefficients for the three travel seasons are significantly different from 0, and the third season (SEA3) witnesses the highest intention to use tourist accommodation, whereas the first season (SEA1) witnesses the lowest. Workdays (TD3) witness the highest intention to use tourist accommodation, whereas national holidays (TD1) witness the lowest intention to use tourist accommodation; the estimated coefficients of the two variables are significantly different from 0. Regarding the variables for the favorite activity during the trip, except for visiting amusement parks (ACT4), other variables are significantly different from 0; individuals who prefer sports (ACT3) and visiting amusement parks (ACT4) have a higher intention to use accommodation, whereas those who prefer visiting families and friends (ACT7) and cultural experience (ACT2) have a lower intention to use accommodation. Days of the trip (TDS) reveal a significant positive impact on the intention to use tourist accommodation. The implementation of the “one fixed day off and one flexible rest day” policy has a significant positive impact on the intention of Taiwanese to use tourist accommodation. Therefore, the vacation policy variable is a determinant of the accommodation decision, in line with
Zhang et al. (
2016).
Tourist accommodation, to a certain extent, reflects the importance attached by individuals to tour quality, and the single-day tour approach often sacrifices tour quality due to time constraints. The above analyses can be summarized as follows: females, people with a university level of education, white-collar workers, tourists traveling with children between the ages of 0 and 6 and 7 and 11, married people, people aged 12–19, residents of the southern region, people traveling during the third season, people traveling during normal days, and people preferring sports and visiting amusement parks are those with a high intention to use accommodation in domestic tourism.
4.2. Results of Consumption Decision
As for the consumption decision regarding expenditure on accommodation in domestic tourism, the economic factor, social stratum, family life cycle, residential area, tourism behavior, and vacation policy are variables with significant influence (see
Table 4 and
Table 5). Research hypotheses H2a, H2b, H2c, H2d, H2e, and H2f are all supported. In terms of the economic factor, an individual’s average monthly income has a significant positive correlation with the tourist accommodation expenditure; in other words, with an increase in income, the amount of money a family spends on tourist accommodation during domestic trips also increases. This research result is in line with the research findings by
Thompson and Tinsley (
1978),
Dardis et al. (
1981),
Davies and Mangan (
1992),
Dardis et al. (
1994),
Hong et al. (
1996),
Fish and Waggle (
1996),
Cai (
1999),
Weagley and Huh (
2004),
Alegre et al. (
2013), and
Sun et al. (
2015), i.e., there is a positive correlation between income and tourism expenditure.
In terms of the social stratum, among the education level variables, only EDU4 and EDU5 reach the significance level, indicating that there is a positive correlation between the education level and accommodation expenditure in domestic tourism. As the education level increases, the accommodation expenditure in domestic tourism also increases. Studies conducted by
Dardis et al. (
1981),
Dardis et al. (
1994),
Hong et al. (
1996),
Cai (
1999),
Weagley and Huh (
2004),
Alegre et al. (
2013),
Bernini and Cracolici (
2015), and
Sun et al. (
2015) also obtained the same result. In terms of occupation, the coefficients for retirees and housewives are significantly different from 0; housewives have the highest tourist accommodation expenditure, and blue-collar workers have the lowest tourist accommodation expenditure.
With regard to the family life cycle, the accommodation expenditure of females is higher than that of males, with a coefficient significantly different from 0. In terms of marital status, the coefficients are all insignificant; the unmarried group has the highest tourist accommodation expenditure, followed by the married group, and the divorced/separated or widowed group has the lowest expenditure. There is a significant negative correlation between the numbers of traveling companions between the ages of 0 and 6 and 7 and 11 and tourist accommodation expenditure, mainly because the higher the number of traveling companions between the ages of 0 and 6 and 7 and 11, the higher the tourism expenditure, and thus the accommodation budget needs to be reduced. In terms of the age variables, only AGE1 is insignificant, and the other age groups are all significantly different from 0, with individuals over the age of 70 having the highest tourist accommodation expenditure. Among those over the age of 40, as age increases, the tourist accommodation expenditure also increases (
Figure 2). Compared with the study by
Nicolau and Màs (
2005), we obtained similar results in terms of age and marital status, showing their effect on the level of accommodation/tourism expenditure. Unlike
Alegre et al. (
2013) who found evidence of a positive and increasing relationship with the household’s tourism expenditure, we found that the accommodation expenditure behavior in Taiwan is negatively affected by the presence of children in the household.
In terms of tourism behavior, the first season witnesses the highest tourist accommodation expenditure with a coefficient significantly different from 0. The reason for this is that the first season coincides with the school winter vacation and the Lunar New Year festival, which is the peak tourism season in Taiwan, and the demand for accommodation significantly rises, thereby increasing tourist accommodation expenses. The fourth season witnesses the lowest tourist accommodation expenditure, with an insignificant coefficient. In terms of the travel date, the two variables are both significantly different from 0; national holidays witness the highest tourist accommodation expenditure, followed by workdays, and then weekends and Sundays. In terms of the favorite activity during the trip, the coefficients of all six variables reach the significance level. Individuals visiting family and friends and those visiting amusement parks have the highest tourist accommodation expenditure, whereas those engaging in cultural experience and sightseeing activities have the lowest accommodation expenditure. There is a significant positive correlation between the days of the trip and tourist accommodation expenditure, in line with the finding from
Nicolau and Màs (
2005), indicating that longer stays lead to higher spending levels.
With regard to residential areas, other regions witness the highest tourist accommodation expenditure, followed by the northern region, and the coefficients of both reach the significance level, with tourists residing in the eastern region having the lowest accommodation expenditure. The days of the trip (TDS) have a significant positive impact on tourist accommodation expenditure. The implementation of the “one fixed day off and one flexible rest day” policy has a significant negative impact on tourist accommodation expenditure. This might be because, following the implementation of the policy, employees of private enterprises have more vacations and more opportunities to travel overseas, thereby reducing the accommodation expenditure in domestic tourism.
Zhang et al. (
2016) obtained a similar finding: as China implemented a new vacation policy, the domestic tourism demand was substituted by an increasingly large outbound tourism market.
Based on the above analyses, it can be determined that females, those in high income groups, people with a postgraduate school or above education level, housewives, people traveling with fewer children between the ages of 0 and 6 and 7 and 11, people over the age of 70, people traveling during the first season, people traveling during national holidays, people who prefer visiting family members and friends and visiting amusement parks, and residents of other regions are those with higher accommodation expenditure in domestic tourism.
5. Conclusions and Implications
Increasing the demand for accommodation in domestic tourism is currently an important topic for developing the tourism industry, in particular when international tourism is faced with the difficulties brought about by the COVID-19 pandemic. As tourism products are not necessities for livelihood, situations where there is zero consumption and accommodation expenditure in tourism frequently occur. When conducting relevant research on tourism expenditure using cross-sectional survey data, it is necessary to incorporate zero consumption expenditure into the demand estimation model. In the discussion of tourism expenditure, it is necessary to face and deal with the issues of using appropriate analytical models, understanding the selection process of consumption, and analyzing the factors influencing participation and consumption decisions.
This study employs a two-stage decision model to discuss the factors influencing tourist accommodation expenditure in domestic tourism in Taiwan. It considers a data-oriented approach, uses the nonnested test method and selects the DDH model as the analytical model. According to the empirical results, the participation decision to make use of accommodation in domestic tourism is influenced by five categories of variables, namely, the social stratum, family life cycle, tourism behavior, residential area, and vacation policy. The decision to engage in tourist accommodation expenditure is influenced by six categories of variables, namely, the economic factor, social stratum, family life cycle, tourism behavior, residential area, and vacation policy. The variables in the two decision equations have different degrees and directions of impact on the intention to use accommodation and to spend money on it. Therefore, it is inappropriate to use single-equation analysis consisting of zero consumption expenditure data and to assume that the same variables influence the participation and consumption decisions. This study contributes to the existing literature by being the first to attempt to apply a two-stage model specification to the accommodation decision process, that is, whether or not to use accommodation and how much to spend.
In terms of the individual variables, there is a significant positive correlation between an individual’s average monthly income and tourist accommodation expenditure. There is a significant positive correlation between an individual’s education level and intention to use accommodation in domestic tourism. People usually have higher-paying occupations when they have a higher education level (
Nicolau and Màs 2005). With the increase in education level, the intention to use accommodation in domestic tourism increases, thereby increasing the accommodation expenditure. White-collar workers have the highest intention to use accommodation in domestic tourism, whereas students and unemployed people have the lowest intention. In terms of accommodation expenditure, housewives have the highest expenditure, followed by retirees, then students and unemployed people. Females have a higher intention to use and higher expenditure on accommodation in domestic tourism compared to males. The number of traveling companions between the ages of 0 and 6 and 7 and 11 has a significant positive impact on the intention to use accommodation in domestic tourism, but a negative impact on accommodation expenditure. While this does not mean that the number of traveling companions between the ages of 0 and 6 and 7 and 11 acts as a hindrance to accommodation in domestic tourism, in considering the limitations of their overall travel budget, those tourists may have to reduce their accommodation expenditure.
As for marital status, married people have the highest intention to use accommodation in domestic tourism, whereas unmarried people have the highest accommodation expenditure. People in the 12–19 age group have a higher intention to use accommodation in domestic tourism. As for expenditure on accommodation, for the over 40 age groups, accommodation expenditure increases with age and reaches a peak with the over 70 age group. Every year, the third season witnesses the highest intention to use accommodation in domestic tourism. With regard to accommodation expenditure, the highest amount recorded is in the first season, reflecting the seasonal features and characteristics of the domestic tourism market. In terms of the travel date, workdays witness the highest intention to use accommodation in domestic tourism, whereas national holidays witness the lowest intention to use accommodation. This could be caused by the limited accommodation supply coupled with higher expenses compared with workdays, thereby reducing the demand for accommodation. In practice, national holidays witness the highest accommodation expenditure.
In terms of favorite activities during domestic trips, the two activities of sports and visiting amusement parks have the highest intention to use accommodation in domestic tourism. By contrast, the two activities of visiting family and friends and visiting amusement parks exhibit relatively high expenditure. As for residential areas, tourists residing in the southern region of Taiwan have the highest intention to use accommodation, whereas tourists in other regions incur the highest expenditure. The “one fixed day off and one flexible rest day” policy has a significant positive impact on the intention to use tourist accommodation, but a negative impact on accommodation expenditure.
To sum up, the results of this study indicate that accommodation expenditure models should allow for the existence of a correlation between the participation decision and the expenditure that is conditional on the participation decision. The effects of the above variables on accommodation expenditure are, however, not totally consistent with previous studies on tourism expenditure. These differences may result from the datasets, or the samples being obtained from people of different nationalities. The reasons for the differences need more investigation in future studies. Two variables, namely, tourism behavior and vacation policy, which were previously seldom included in the model’s estimation, were examined in this study for their effects on the accommodation/expenditure decision. Despite the significant effects, it is necessary to more accurately understand the divergent results by performing further investigations.
Based on the analysis of the factors influencing the participation and consumption decisions in relation to domestic tourist accommodation using the two-stage decision model, the results of this research might influence the managerial direction in relation to market segmentation. Such information regarding the demand for accommodation under different economic and demographic conditions is useful to hotel managers in that it provides an alternative perspective for market segmentation. Due to the joint effect or differentiated effect of the variable, hotel managers should reconsider characterizing the profile of tourists with the greatest propensity to use accommodation and to find their expenditure patterns. This is fundamental for the development of marketing strategies. The research results lead to the following specific implications: (1) Attention could be paid to expanding the accommodation market targeted at family travelers who may consider taking children on domestic trips during the summer vacation and will choose accommodation. Therefore, entertainment and leisure space, facilities, and activities for children could be improved to develop business opportunities. (2) Faced with an aging society, there is a strong market potential for tourism for the elderly. This group has the lowest intention to use tourist accommodation but has relatively high tourist accommodation expenditure. The planning of a hospitable environment and travel itinerary for elderly travelers could be strengthened to increase accommodation incentives.
This research has some limitations. First, the model was developed and validated with data from one area. The research should be replicated to test the proposed model and hypotheses of the present research using samples from other regions and other datasets. The second limitation is that the list of variables may not be exhaustive, and thus further exploration should be encouraged. According to
Isık et al. (
2020), policy-related economic uncertainty plays a significant role in tourists’ vacation plans. Thus, the EPU index could be included as a predictor of tourism demand. Third, the impact of the COVID-19 pandemic on travel should be a topic for further research. Tourism and travel demand were reduced to a minimum level during the period of the pandemic and domestic tourism has been the first to recover as the lockdown gradually ended. A detailed analysis of the variations in the intention to use accommodation and accommodation expenditure may be a valuable topic for future research. Finally, some researchers have broadened the knowledge of tourism expenditure by adopting a new analytical approach (e.g.,
Alfarhan et al. 2022;
Chulaphan and Barahona 2021;
Pellegrini et al. 2021). With regard to the different levels of service and nature of accommodation, many facets of accommodation expenditure decisions may need to be considered, because accommodation expenditure is not a single product but rather a number of interrelated subproducts. Tourists may additionally arrange several subset decisions within accommodation expense types, such as dining, recreational activities, and travel itineraries. In referring to
Park et al. (
2020), the analyses of accommodation expenditure across and within expense types could be addressed in future research. A multi-perspective view of modeling is important for gaining an enhanced understanding of tourism/accommodation expenditure patterns.