Rural tourism demand: duration modeling for drive tourists' length of stay in rural areas of the United States.
Rural communities in the United States have experienced increasing economic challenges in recent decades and many have looked to increase tourism in their destinations as a means to support economic development. Tourism in these rural destinations tends to be undertaken by independent travellers using private vehicles. The current research explores the factors that have contributed to an increased focus on rural tourism and examines two key challenges experienced in the development of tourism in general, and rural tourism specifically. The first challenge is determining which statistical techniques best explain the factors contributing to tourism demand. The second challenge addressed in the current research is identifying which factors contribute to tourism demand.
Tourism and rural renewal
The continuing economic and social decline of rural areas has been recognized as a crucial factor threatening the sustainability of rural regions in the United States. Since the 1970s, the decline in rural areas has made a negative impact on the economic, social, and cultural aspects of rural communities. During the 1980s the "farm crisis" in the Midwest of the United States had harmful effects on rural society (Wilson et al., 2001). Since then, rural communities have sought alternative ways to address the rural exodus and the economic challenges, many of them having turned to tourism to support economic development. Rural tourism can, thus, be a major element of rural and regional development as well as a driving force of rural community's social revitalization (Skuras, 2006). Additionally, rural tourism has attracted the attention of destination marketers, policy makers, and scholars as a means of facilitating rural society's revitalization (Lane, 1994; Gartner, 2004; Fleischer & Tchetchik, 2005; Hegarty & Przezborska, 2005; Loureiro & Jervell, 2005).
Successful rural tourism development requires essential factors to provide an appealing consumer experience. The factors include basic tourism attractions in rural areas, marketing activities by the local Destination Marketing Organization (DMO), tourism infrastructure like transportation, tourism related services (e.g. lodging, restaurants, and retail services), and hospitality by host community's residents (Wilson et al., 2001). One factor that is particularly important to rural tourism development is road access to the destination as most rural attractions are isolated from public transportation and rural tourists tend to use private modes of transportation (Dickinson and Robbins, 2008). As such, the drive tourism market or "self drive" rural tourists can be regarded as one of the main target markets in rural tourism destinations.
Increased tourism demand may exert both positive and negative effects on a rural destination. Tourism destination and economic development managers should carefully plan to maximize the benefits of their activities. Dickinson and Robbins (2008) said car travel has contributed to tourism development in rural areas, but such travel also may cause global environmental issues and local travel problems, and that increased tourism demand, which is due to drive tourism, could aggravate quality of life in rural destinations. On the contrary, it is reported that drive tourists' activity is an important source to improve rural destination's vitality and economic development (Taylor & Carson, 2010). Most of drive tourism studies agreed that rural destinations can gain economic and social benefit from drive tourism as long as rural destinations can deal with adverse impact from car-based trips. Nevertheless, it is important for researchers, policy makers, and practitioners to understand the demand factors of rural tourism for effective planning purposes.
Drive tourism and rural tourism development
Drive tourism can be defined as "tourism that centres on travelling from an origin point to a destination by car that is either privately owned or rented, and engaging in tourism-related activities during the journey" (Prideaux et al., 2001). Drive tourism has been a significant form of leisure travel in many developed geographical areas, such as the United States and Europe. Increasing personal income, vehicle availability and continuous infrastructure development (e.g. the construction of highways) have enabled drive tourism to flourish. According to the U.S. National Household Travel Survey 2001 (NHTS), almost 90 % of all long-distance trips in the United States are mare by using a personal vehicle (Bureau of Transportation Statistics, 2003).
The relationship between rural tourism and drive tourism is firmly established. Rural tourism in the United States relies on the interstate/state highway system and private auto vehicles because of the lack of public transportation in rural areas. Self-drive tourists have been recognized as an important rural tourism target market because self-drive tourists are not subject to geographical limitations of rural destinations (Howat et al., 2006; Lane & Waitt, 2007; Dickinson & Robbins, 2008; Taylor & Carson, 2010). For example, an automobile has provided reliable transportation to link most rural communities and destinations and rural tourists can select tourism destinations without geographical limitations (English et al., 2000; Wilson et al., 2001; Gartner, 2004;). As many tourists participate in self-drive trips to a rural destination, rural tourism demand increases (Pennington-Gray, 2003; Prideaux et al., 2005; Carson & Taylor, 2008; Connell & Page, 2008). Drive tourism overcomes the inherent geographical limitations of rural destinations and has the potential to send visitors to communities and revitalize rural economies (Gartner, 2004). Therefore, one must understand self-drive tourists' characteristics and trip-related behaviour in rural destinations so that rural destination marketers can effectively program to attract rural tourists to the destination.
As rural drive tourism spans a number of disciplines, including tourism, transportation, lodging, and rural development, research relevant to the present study comes from a number of topics. These topics include destination marketing to drive tourists (Prideaux, 2000; Pennington-Gray, 2003), transportation related issues in drive tourism (Mallett & McGuckin, 2000; Prideaux, 2000; MacKay et al., 2002; Pennington-Gray, 2003), effects of drive tourism on rural destinations (Downward & Lumsdon, 2004), travel information searching behaviour in rural drive tourism (Pennington-Gray & Vogt, 2003), and the relationship between the external change of drive tourism market and rural tourism (Trent & Pollard, 1983; Prideaux et al., 2001). There have also been some exploratory attempts to examine rural drive tourists' travel behaviour, which includes drive tourists' travel preference, the effect of drive tourists' socio-demographic variables on trips, and rural tourism demand studies (Mallett & McGuckin, 2000; Scott, 2002; Pennington-Gray, 2003; Shih, 2006; Howat et al., 2006, Skuras, 2006; Lane & Waitt, 2007).
However, while the extant research provides a foundation for understanding the rural drive market, there is still much to learn about rural drive tourists and the rural drive tourism phenomenon. Despite drive tourism's popularity and importance in rural destinations, little academic research has been done on self-drive tourist market in rural destinations, and few scholars have explored the relationships between the drive tourist market and rural tourism (Prideaux & Carson, 2003).
Consumer demand studies provide a good starting point for consumer-oriented tourism planning and extensive academic research has been devoted to the study of tourism demand. Consumer demand research can be divided into two main approaches: macroscopic and microscopic (Hu, 2002). Even though there are two separate tourism demand research trends, these are based on the same traditional economic theory. Stone (1954) first estimated the demand function for six commodities with consumer theory and focused on the effect of price change in demand function. In tourism demand studies, Stone's economic theory has been widely used as a theoretical foundation for estimating the tourism demand function. The difference between macroeconomic and microeconomic approaches is in terms of the research objectives, the data attributes, and the estimating methods of the studies.
Macroeconomic consumption theory provides a theoretical basis to analyze the relationship of income, expectations, total consumption, and travel demand. Macroeconomics focuses on examining travel demand at aggregate data levels to forecast future demand (Smeral, 1988; Hu, 2002). Clawson and Knetsch (1966) attempted to measure economic benefits with an empirical study which investigated the economic relationship between trip cost and recreational participation rates for regional destinations, by using aggregated zonal data and employing shadow price estimation for tourism demand. The basic assumption of the shadow pricing method is that travel costs are associated with the rate of visitation to a tourism destination, and it represents a visitor's willingness to pay for the destination. With respect to Clawson's work, Brown and Nawas (1973) noted that "The most commonly employed approach for estimating outdoor recreation demand is based upon the pioneering research of Clawson", but they argued that this approach has a potential flaw because it assumes that tourists are essentially homogeneous. In essence, travel-cost methods for estimating the demand curve with aggregated data tend to cause multicollinearity. Therefore, there is difficulty in estimating the parameters of tourism demand functions.
On the other hand, the microeconomic approach is based on individual consumer expenditure theory and utilizes individual data rather than aggregated data. In order to gain efficiency in the estimation of tourism demand, Brown and Nawas (1973) suggested using individually-observed survey data. This data allows researchers to achieve more realistic results. Michael and Becker (1973) developed a new economic theory for estimating demand by incorporating consumer behaviour. Their theory was derived from traditional consumer theory, and the application of individual tastes and preferences. These variables are represented by demographic variables such as family size, family age-structure, education, housing tenure, occupation, race, socio-economic status, and other proxy variables (Pollak and Wales, 1981; Cai, 1996; Palakurthi and Parks, 2000; Heckman, 2001; Cannon and Ford, 2002; Alegre and Pou, 2006).
Count data model in tourism demand
Tourism studies have tried to examine tourism demand. This demand can be quantified in a variety of ways, some of which include tourism expenditure, the number of nights on a given trip, and the number of visitors to the destination. This kind of data has the unique statistical feature of non-negativity which hinders tourism researchers in achieving accurate demand estimations. To estimate accurate tourism demand, a number of alternative ways have been proposed, out of which count data econometric techniques (e.g. truncated OLS, Poisson regression, Negative binomial regression) were the most representative method to estimate demand function.
The count data econometric techniques were first applied to estimate recreational demand by Shaw (1988). Since then, the count data model has been applied with increasing frequency in estimating recreational and tourism demand (Englin & Shonkwiler, 1995). Because of the data distribution attribute, the count data model begins with a demand function which is specified with a semi-logarithmic functional form. Based on the survey data and sample selection process the authors of the present study assumed that the dependent variable, which is the number of nights away from home during the travel period, is a non-negative integer without zero value because the authors chose respondents who have travelled at least one night away from home and used any lodging facility (e.g. hotel, motel, family or friend's house, owned or rented cabin, camp site or RV). According to the work of Shaw (1988), only a small portion of the entire population is observed and shows non-zero demand for the tourism experience because only respondents who have recreational experience were selected and surveyed.
Duration models in tourism demand
Recently, duration modelling, a more sophisticated estimation method, has been used in estimating tourism demand when dependent variable is time-related. When time is crucial and constitutes main dependent variable of interest to the research, duration models can be used because the dependent variable has specific distribution functions which are not normally distributed and positively skewed (Cox and Oakes, 1984). In biostatistics and medical science, this type of model is referred to as a survival analysis model because it estimates and evaluates the number of days patients may survive given specific conditions. Moreover, in actuarial studies, and other insurance applications, the duration model is widely used to estimate the effect of explanatory variables, which are continuous or categorical, on a duration dependent variable (Martinez-Garcia & Raya, 2008).
In tourism demand studies, duration models have been used to estimate the effects of explanatory variables on the expected time that tourists stay at a destination. According to the work of Martinez-Garcia and Raya (2008), "The model also quantifies the likelihood that a tourist ends the trips conditioned on having travelled "t" day ... The estimated model allows evaluation of the effect of each variable or category of variables per se on the length of stay." Therefore, in the present study, the determinants of rural tourism demand for individual drive tourists could be empirically derived from the model.
The purpose of the current research
The current study recognizes the need for empirical research to provide practical marketing information and a theoretical foundation for estimating rural tourism demand. Therefore, the purpose of the research is twofold. The first goal is to establish an effective means of estimating rural tourism demand. The second goal is to determine the significant characteristics of rural drive tourists for market segmentation based on a major survey of household travel.
The current study will explore rural tourism demand with both a highly accurate statistical model and micro survey data, because it is necessary to examine drive tourism tourist's characteristics and travel patterns at the individual level. This research is beneficial for the academic field of tourism, as well as for rural stakeholders and decision makers.
The data used in the present study is long-distance trip from the National Household Travel Survey (NHTS, 2001). This survey has been conducted since 1969 by The Federal Highway Administration (FHWA) and has become the nation's representative travel survey to quantify traveller's trip-related behaviour. The survey has provided traveller's behaviour, and socio-demographic information relating to all modes of transportation, for all travel purposes, and for all travel distances. The NHTS 2001 was conducted from April 2001 through to May 2002 using Computer-Assisted Telephone Interviewing (CATI) technology, which collected travel data from a national sample. The data are the results of telephone interviews with 66,000 households.
A sub-sample of 2,063 respondents was selected from the complete dataset to include those respondents who live in the Midwest of the United States, and were away from home at least one night with a primary trip purpose of pleasure travel. While it is recognized that there are urban centres in the Midwest, this region has a large number of rural destinations and is therefore a useful region for this kind of exploratory examination.
The dependent variable used for the research was the number of nights away from home in a travel period. In tourism demand studies, there are several commonly used proxy variables for tourism demand. These variables are: tourist expenditure, the number of visitors, and the number of nights away from home. The current study uses the number of nights away from home, which are important because the length of stay in a rural destination directly impacts on the revenue generated through hotel bed taxes, and is a good indirect indicator of tourist expenditure.
The conceptual model in the present research is based on previous tourism demand studies and consumer expenditure theory and assumes hypothetical relationships between dependent and independent variables, whilst the dependent variable is the number of nights at the rural destination. The independent variables are rural tourists' socio-demographic and trip-related variables, which include income, trip distance, age, gender, lodging preference, transportation mode, trip party size, weekend trip, and the attack of 9/11. The conceptual model is, thus, presented as follows:
Number of nights in rural destination = f (income, trip distance, age, gender, lodging preference, transportation mode, trip party size, weekend trip, 9/11 attack)
In order to estimate most accurate rural tourism demand model, four statistical models were used in this research. The statistical models are: Ordinary Least Square (OLS) model, Zero-truncated Poisson Model, Zero-truncated Negative Binomial Model, and a duration model estimated by Log-logistic regression. The Zero-truncated Poisson Model and the Zero-truncated Negative Binomial Model have been most commonly used in estimating tourism and recreation demand, but sometimes these models could be unreliable when key statistical assumptions are violated (e.g. over-dispersion issue) (Martinez-Espineira et al., 2008). Even though OLS results could be biased because of non-negative feature of tourism demand data (Perali & Chavas, 2000), the OLS model was utilized as reference model because one of the main objectives was to compare statistical models in terms of accuracy and effectiveness in estimation, and because it could also play a role as reference work. Moreover, the AIC criteria was utilized when authors determined which statistical model is the best model fit for estimating rural tourism demand.
The duration model focuses on times until the event of interest occurs. The event could be good or bad, or it could be favourable or unfavourable. It can be seen as "failure", the time until failure being the "survival time". In this study, the number of nights in at a rural destination is defined as survival time in that tourists' trip duration means their survival time during trip or tourism demand. The survival time can be estimated by the Log-logistic regression in duration model. The model form is the same as the Cox regression model, except for the assumption that the survival function follows the Log-logistic distribution. The number of nights (T) at a rural destination can be also regarded as a random variable with a probability distribution F(t) for survival time. The authors' research interest is the probability that tourists stay in a rural destination at least t days. The survival function is given by:
S(t) = P(T=t) = 1-F(t)
With respect to this research interest, there is a secondary function for survival data, the hazard function, which can be expressed by:
The hazard function is just the incidence rate, so it can be shown by:
S(t) = exp(-H(t))S(t) = exp(-H (t))
Where H(t) is the integrated hazard function also known as the cumulative hazard function. There are basically two kinds of estimation methods in survival analysis: nonparametric and parametric estimations. The Kaplan-Meier estimation and Cox regression are nonparametric methods. Based on the assumption for sample distributions, Exponential, Gompetz, Log normal, Log-logistic, Generalized gamma, and Weibull regression are all parametric methods. Among them, Cox regression, which employs a proportional hazard model, is a widely-used method, which provides robust test statistics, but it is inefficient in terms of information loss. It is, thus, defined by:
[h.sub.i](t) = [h.sub.0](t)exp([x'.sub.i][[beta])[h.sub.i](t) = [h.sub.0].(t) exp([x'.sub.i][beta])
Whereby [h.sub.0](t)[h.sub.0](t) is a baseline hazard function and a nonparametric part, "exp([x.sub.[down arrow]][i.sup.[up arrow]]' [[beta])" exp([x.sub.[down arrow]][i.sup.[up arrow]' [beta])" is a parametric part which contains covariates for analysis. The Cox regression method is able to estimate the baseline survivor function with covariates by relying on empirical evidence without theoretical distribution (Hamilton, 2008). Several alternative parametric approaches are widely used based on the assumption that survival, or the hazard function, follows a specific statistical distribution.
The following hypotheses established test the relationship between lodging demand, socio-demographic characteristics, and trip-related preferences. There are nine hypotheses in three categories of socio-demographic, economic, and trip-related variables. Among them, the last two hypotheses are in order to control and test the effect of both weekend trips and the 9/11 attacks in 2001. Detailed hypotheses are listed below.
H1: Total trip distance affects tourism demand.
H2: Income level affects tourism demand.
H3: Age affects tourism demand.
H4: Gender affects tourism demand.
H5. Lodging type affects tourism demand.
H6: Type of auto vehicle affects tourism demand.
H7: The number of household members affects tourism demand.
H8: Weekend trips affect tourism demand.
H9: The attacks of September 11, 2001 affect tourism demand.
The typical profile of the self-drive tourist is presented in Table 1. A total of 2,063 respondents were selected and they were those who travelled to the Midwest of the United States. Among the respondents, 46% were male and 53% were female. Age groups were nearly equally distributed in Fig. 1, but the dominant age group was from 35-54 years.
[FIGURE 1 OMITTED]
The median household income level was $50,000 and household income was converted into income per household member. Interestingly, one of the key variables was household size. The largest group was the two-member household, which accounted for 44.64% of the total samples. Education level was also examined; thus, half of the respondents were high school graduates or less educated (54.77%), followed by college graduates (29.76%). According to Figure 2, it can be found that most of the self-drive tourists own at least two cars, whilst drive tourists could select transportation mode as tourists want. Therefore, selection of cars for travel can be viewed as a trip-related variable.
Results of OLS estimation for rural tourism demand are presented in Table 2. To ensure the key OLS assumptions, the Variance Inflation Factor (VIF) was checked. According to the results, there is no multicollinearity problem between the independent variables. Moreover, Breusch-Pagan was performed to test heteroskedasticity, showing that heteroscedasticity exists. To resolve that problem, the heteroscedasticity robust error was used in testing coefficients of the demand model. Results showed that the gender variable is insignificant and dummy variables for transportation mode, which are the SUV and van users, are partially insignificant. Trip distance, income level, age, lodging selection preference, type of auto vehicle, trip party size, weekend trip, and 9/11's effect are however significant, as these variables affected the rural tourism demand of the self-drive tourism market.
Table 3 shows the results of Zero-truncated Poisson regression. The Pseudo R2 is 0.3998 and Log likelihood is -4097.10. Compared to previous OLS results, there is some difference between them. In the case of the present study, most coefficients are significant except for one dummy variable in the lodging type (i.e., rented cabin or condo). In terms of the estimate coefficients, it represents how this variable influences drive tourist's lodging and tourism related behaviour. The overall sign of the Poisson model is the same as for the OLS estimation results. There are a few differences between OLS estimation and Poisson estimation. However, in this case, one of the different coefficient sign variables is statistically insignificant. One can conclude that the OLS model and Poisson models explain the effect of the independent variable in approximately the same way.
Table 4 shows the results of employing the Zero-truncated Negative Binomial Regression Model. The sign of the parameter estimates implies the Negative Binomial Regression Model is the same as the OLS and Poisson models, suggesting, thus, similar meaning. However, the goodness of fit value for the zero-truncated NBRM is quite different. The Pseudo R2 is calculated by using the information from the -2 Log likelihood (-2LL) for the full model and for the intercept only. The result is a measure for the improvement in fit of the model due to the introduction of independent variables into the model. In the case of Zero-truncated Negative Binomial Regression Model, its Pseudo R2 is 0.1601 and Log likelihood is -3298.31. These results were compared to the Poisson model, and as superior.
The results of duration model estimation are presented in Table 5. The duration model, using Log-logistic regression, was determined as most suitable for estimating the rural tourism demand in the self-drive market, as it best represents the travel behaviour of the drive tourist in terms of parametric distribution. Also, the duration model better explains the tourism demand for drive tourists by using a generated statistic called a time ratio. This indicator represents the probability of a drive tourist's willingness to continue to stay at a rural tourism destination. The results show that the socio-demographic and trip-related variable are significant except for individual income and gender, indirectly affecting the drive tourists' lodging demand.
In order to compare each model's fitness and explanation power, four models based on Akaike Information Criterion (AIC) have been evaluated. Results showed that each of the four regression analyses demonstrated its unique characteristics in fitting the random sample data. However, due to the feature of limited dependent data, the OLS model was excluded. Thus, the three remaining regression models were compared in terms of Akaike Information Criterion (AIC) with each model's AIC value being shown in Table 6.
Because the AIC indicated that the Log-logistic regression model is the most suitable model among alternatives (Table 4-6), the present study selected it as the primary research model, whilst parameters from Log-logistic regression estimation have been used in testing the hypotheses.
H1: Total trip distance affects tourism demand. The estimated coefficient is 0.0009, and p-value is less than 0.001. The hypothesis is supported. Total trip distance can be crucial a factor to affect tourism demand.
H2: Income level affects tourism demand. The estimated coefficient is -0.0054, and p-value is with 0.227, the hypothesis is not supported. Household income does not affect the tourism demand.
H3: Age affects tourism demand. The hypothesis is also supported. The estimated coefficient is 0.0223, and p-value is less than 0.01. Senior drive tourists may spend more nights at rural destinations than young age group tourists.
H4: Gender affects tourism demand. The hypothesis is rejected because its p-value is 0.133. There is no difference in tourism demand between different gender groups.
H5. Lodging type affects tourism demand. This hypothesis has been tested with four dummy variables by lodging type. Results showed that all dummy variables are significant, and the hypothesis is supported.
H6: Type of vehicle affects tourism demand. The hypothesis is partially supported because only pick-up truck and RV users show statistical difference in tourism demand.
H7: Size of travel party affects tourism demand. The size of the travel party also has a significant effect on the tourism demand. The estimated coefficient is -0.0254 with p-value of 0.022. The hypothesis is supported.
H8: Weekend trips affect tourism demand. The estimated coefficient is 0.1557 with p-value of less than 0.01. The hypothesis is supported, which shows weekend trip is a crucial factor in tourism demand.
H9: Events of September 11, 2001 affects tourism demand. The estimated coefficient is -0.082 with p-value of less than 0.001. The hypothesis is supported. The overall tourism demand for the self-drive market decreased after the attacks of September 11, 2001.
Discussion and conclusions
The two fold purpose of the present study was to establish an effective means to estimate tourism demand, and to determine the significant characteristics of rural drive travellers for market segmentation based on the National Household Travel Survey (NHTS). In addressing the first goal, the study estimated the rural tourism demand function for the self-drive travel market by using four statistical models: the Ordinary Least Square regression (OLS) model, the Zero-truncated Poisson regression model, the Zero-truncated Negative Binomial regression model, and the duration model with Log-logistic regression. Among them, the duration model is the most adequate statistical model for estimating rural tourism demand. The duration model provides key insights about tourism demands in the self-drive travel market when the dependent variable is length of stay. The model can fully utilize the feature of event data. Addressing the second goal of the research, according to the results of the analysis, most socio-demographic variables and trip-related variables are significant. Results showed that these variables can be used as market segmentation variables because they show the effects of self-drive tourists' lodging selection, type of transportation model, trip party size, weekend trip, trip distance, and the socio-demographic variable of length of stay, the most important tourism behaviour. What drive tourists want on rural trips is revealed by examining a combination of trip-related variables like lodging preference, trip distance, and type of auto vehicle. When drive tourists take long distance trip, rural tourism demand increases. Camper and Recreational Vehicle (RV) travellers usually stay longer than those who stay in hotel, while, travellers driving fuel inefficient cars stay shorter than travellers in more fuel efficient vehicles. The impact of the events of the September 11, 2001 attacks, while significant in the current study, are considered to be unique to the data collection period of the study. The data was collected immediately following the attacks at a time of great disruption.
The current study utilized relatively new statistical methods to estimate duration of specific events for the rural tourism demand study. The duration method has been widely used in a clinical studies and the insurance industry. The current study found that the method is relevant to the tourism industry, as well as for the study of rural tourists' behaviour. Additionally, this study has explored the unique rural tourism market, the self-drive travel market, providing insightful information about the profile of self-drive tourists, and influential variables on tourists' behaviour. The study can be considered as an exploratory attempt to quantitatively examine the demand for rural tourism.
Theoretically, the study examined the relevance of socio-demographic variables in both market segmentation and tourism demand estimation. A current research trend has emphasized the importance of psychological aspects of tourists' behaviours. It tends to neglect the usefulness of socio-demographic variables in marketing research and the study of tourist behaviour. However, socio-demographic variables are cost-effective variables in marketing research; they are useful in examining tourists' behaviour when relevant variables are obtained. The study re-confirmed the availability of socio-demographic variables, which is also supported by consumer demand and expenditure theory.
Methodologically, the present study has considered the unique feature of a dependent variable (i.e. number of night at destinations). Contrary to conventional demand studies, the feature caused serious methodological problems in estimating demand function. The study explained why the duration model is appropriate, and showed that it provides rigorous and accurate estimating results. Additionally, the study demonstrated how much tourists' micro data can contribute to quantify tourists' preferences and socio-demographic aspects with various statistical models.
The present study has some managerial implications for the rural tourism industry. Rural tourism is subject to geographical limitations, but propagation of private transportation mode enables rural tourists to access tourism destinations. Therefore, since self-drive tourists can be seen as the main target market for rural tourism, rural tourism marketers should focus on that target market to facilitate the influx of rural tourists into such destinations. Research findings also provide an interesting concept for a successful marketing program. Lodging selection, trip distance, and type of vehicle are all significant variables, showing that tourists' travel behaviour is influenced by economic factors and tourism preferences. Quantified effects of these variables can help rural tourism practitioners to understand their customers and generate effective marketing strategies.
Received January 22, 2011. Resubmitted March 25, 2011
ALEGRE, J. & POU, L. (2006), The length of stay in the demand for tourism. Tourism Management, 27, 1343-1355.
BROWN, W. G. & NAWAS, F. (1973). Impact of aggregation on the estimation of outdoor recreation demand functions. American Journal of Agricultural Economics, 55, 246-249.
BUREAU OF TRANSPORTATION STATISTICS (2003), Highlights of the 2001 National Household Travel Survey. Washington, DC: U.S. Department of Transportation, Bureau of Transportation Statistics.
CAI, L. A. (1996), An analysis of household expenditure behavior for hospitality and tourism goods and services: Implications for market segmentation. Thesis (PhD), Purdue University, 1996.
CANNON, T. F. & FORD, J. (2002), Relationship of demographic and trip characteristics to visitor spending: an analysis of sports travel visitors across time. Tourism Economics, 8, 263-271.
CARSON, D. B. & TAYLOR, A. J. (2008), Sustaining four wheel drive tourism in desert Australia: exploring the evidence from a demand perspective. The Rangeland Journal, 30, 77-83.
CLAWSON, M. & KNETSCH, J. L. (1966), Economics of outdoor recreation, Baltimore, Published for Resources for the Future by Johns Hopkins Press.
CONNELL, J. & PAGE, S. J. (2008), Exploring the spatial patterns of car-based tourist travel in Loch Lomond and Trossachs National Park, Scotland. Tourism Management, 29, 561-580.
COX, D. R. & OAKES, D. (1984), Analysis of survival data, Chapman & Hall.
DICKINSON, J. E. & ROBBINS, D. (2008), Representations of tourism transport problems in a rural destination. Tourism Management, 29, 1110-1121.
DOWNWARD, P. & LUMSDON, L. (2004), Tourism transport and visitor spending: A study in the north york moors national park, UK. Journal of Travel Research, 42, 415.
ENGLIN, J. & SHONKWILER, J. S. (1995), Estimating social welfare using count data models: an application to long-run recreation demand under conditions of endogenous stratification and truncation. The Review of Economics and Statistics, 77, 104-112.
ENGLISH, D. B. K., MARCOUILLER, D. W. & CORDELL, H. K. (2000), Tourism dependence in rural America: Estimates and effects. Society & Natural Resources, 13, 185 202.
FLEISCHER, A. & TCHETCHIK, A. (2005), Does rural tourism benefit from agriculture? Tourism Management, 26, 493-501.
GARTNER, W. C. (2004), Rural tourism development in the USA. International Journal of Tourism Research, 6, 151-164.
HAMILTON, L. C. (2008), Statistics with Stata: Updated for version 10, Cengage Learning.
HECKMAN, J. J. (2001), Micro data, heterogeneity, and the evaluation of public policy: nobel lecture. Journal of Political Economy, 109, 673-748.
HEGARTY, C. & PRZEZBORSKA, L. (2005), Rural and agri-tourism as a tool for reorganising rural areas in old and new member states--a comparison study of Ireland and Poland. International Journal of Tourism Research, 7, 63-77.
HOWAT, G., BROWN, G. & MARCH, H. (2006), Self-drive tourism: Travellers to South Australia's limestone coast region. Technical Report.
HU, C. (2002), Advanced tourism demand forecasting: Artificial neural network and Box-Jenkins modeling. Thesis (PhD), Purdue University, 2002.
LANE, B. (1994), What is Rural Tourism? Journal of Sustainable Tourism, 2.
LANE, R. & WAITT, G. (2007), Inalienable places: Self-drive tourists in northwest Australia. Annals of Tourism Research, 34, 105-121.
LOUREIRO, M. L. & JERVELL, A. M. (2005), Farmers participation decisions regarding agro-tourism activities in Norway. Tourism Economics, 11, 453-469.
MACKAY, K. J., ANDERECK, K. L. & VOGT, C. A. (2002), Understanding vacationing motorist niche markets. Journal of Travel Research, 40, 356.
MALLETT, W. J. & MCGUCKIN, N. (2000), Driving to distractions: Recreational trips in private vehicles. Transportation Research Record, 1719, 267-272.
MARTINEZ-ESPINEIRA, R., LOOMIS, J. B., AMOAKO-TUFFOUR, J. & HILBE, J. M. (2008), Comparing recreation benefits from on-site versus household surveys in count data travel cost demand models with overdispersion. Tourism Economics, 14, 567-576.
MARTINEZ-GARCIA, E. & RAYA, J. M. (2008), Length of stay for low-cost tourism. Tourism Management, 29, 1064-1075.
MICHAEL, R. T. & BECKER, G. S. (1973), On the new theory of consumer behavior. Swedish Journal of Economics, 75, 378-396.
PALAKURTHI, R. R. & PARKS, S. J. (2000), The effect of selected socio-demographic factors on lodging demand in the USA. Management, 135, 142.
PENNINGTON-GRAY, L. (2003), Understanding the domestic VFR drive market in Florida. Journal of Vacation Marketing, 9, 354-367.
PENNINGTON-GRAY, L. & VOGT, C. (2003), Examining welcome center visitors' travel and information behaviors: Does location of centers or residency matter? Journal of Travel Research, 41, 272-280.
PERALI, F. & CHAVAS, J. (2000), Estimation of censored demand equations from large cross-section data. American journal of agricultural economics, 82, 1022.
POLLAK, R. A. & WALES, T. J. (1981), Demographic variables in demand analysis. Econometrica, 49, 1533-1551.
PRIDEAUX, B. (2000), The role of the transport system in destination development. Tourism Management, 21, 53-63.
PRIDEAUX, B. & CARSON, D. (2003), A framework for increasing understanding of self-drive tourism markets. Journal of Vacation Marketing, 9, 307-313.
PRIDEAUX, B., MCCLYMONT, H. & CASSIDY, F. (2005), Promoting to the drive tourists: an exploratory Queensland study.
PRIDEAUX, B., WEI, S. & RUYS, H. (2001), The senior drive tour market in Australia. Journal of Vacation Marketing, 7, 209-219.
SCOTT, N. (2002), Product market perspective of self-drive tourism. Carson et al., ref, 1, 81?0.
SHAW, D. (1988), On-site samples' regression: Problems of non-negative Integers, truncation, and endogenous stratification. Journal of Econometrics, 37, 211-23.
SHIH, H. Y. (2006), Network characteristics of drive tourism destinations: An application of network analysis in tourism. Tourism Management, 27, 1029-1039.
SKURAS, D. D. (2006), Demand for Rural Tourism: The Effects of Quality and Information. Agricultural economics, 35, 183.
SMERAL, E. (1988), Tourism demand, economic theory and econometrics: An integrated approach. Journal of Travel Research, 26, 38-43.
STONE, R. (1954), Linear expenditure systems and demand analysis: An application to the pattern of british demand. The Economic Journal, 64, 511-527.
TAYLOR, A. & CARSON, D. (2010), Four wheel drive tourism and economic development opportunities for remote areas. Tourismos, 5, 69-85.
TRENT, R. B. & POLLARD, C. (1983), Gasoline prices and discretionary auto travel: Evidence from a panel survey. Journal of Travel Research, 21, 12-16.
WILSON, S., FESENMAIER, D. R., FESENMAIER, J. & VANES, J. C. (2001), Factors for success in rural tourism development. Journal of Travel Research, 40, 132.
Sangchoul YI *, Jonathon DAY *, Liping A. CAI *
* Purdue Tourism & Hospitality Research Center and School of Hospitality and Tourism Management, Purdue University, 700 W. State Street, West Lafayette, U.S.A.; email@example.com
Table 1 Profile of Rural Tourist Freq. Pct. Age 18-24 177 8.58% 25-34 361 17.50% 35-44 444 21.52% 45-54 418 20.26% 55-64 353 17.11% 65~ 310 15.03% Total 2,063 100% Gender Male 959 46.49% Female 1,104 53.51% Total 2,063 100% Race Not Ascertained 7 0.34% Refused 4 0.19% White 1,932 93.65% African American 35 1.70% Asian Only 3 0.15% American Indian, Alaskan Native 11 0.53% Hispanic or Mexican Only 22 1.07% White & American Indian 17 0.82% White & Hispanic 31 1.50% Other Combination 2 Races 1 0.05* Total 2,063 100% Household size 1 201 9.74% 2 921 44.64% 3 348 16.87% 4 360 17.45% 5 150 7.27% 6 67 3.25% 7 14 0.68% 8 1 0.05% 9 1 0.05% Total 2,063 100% Household income << $30,000 401 19.44% $30,000-$49,999 539 26.13% $50,000-$69,999 416 20.16% $70,000-$100,000 419 20.31% $100,000 or more 288 13.96% Total 2,063 100% Number of vehicles 0 23 1.11% 1 297 14.40% 2 948 45.95% 3 460 22.30% 4 215 10.42% 5 93 4.51% 6 16 0.78% 7 3 0.15% 8 5 0.24% 9 3 0.15% Total 2,063 100% Education High or less 1,130 54.77% Some college 614 29.76% Post Graduate 312 15.12% No answer 7 0.34% Total 2,063 100% Life cycle one adult, no children 145 7.03% 2+ adults, no children 560 27.14% one adult, youngest child 0-5 14 0.68% 2+ adults, youngest child 0-5 361 17.50% one adult, youngest child 6-15 23 1.11% 2+ adults, youngest child 6-15 337 16.34% one adult, youngest child 16-21 8 0.39% 2+ adults, youngest child 16-21 110 5.33% one adult, retired, no children 55 2.67% 2+ adults, retired, no children 450 21.81% Total 2,063 100% Table 2 Lodging Demand Function of Drive Tourism by Using OLS with Heteroskedasticity Robust Error Variable Parameter Estimate (heteroskedasticity robust error) Constant -2.388(0.120) * Total trip distance [Distance] 0.524(0.019) * Income level [Income] -0.058(0.019) * Age 0.015(0.008) ** Gender [Male] (Female) (a) -0.028(0.024) Lodging type [VFR] 0.128(0.028) * [Rented cabin or condo] 0.314(0.072) * [Own cabin or condo] 0.477(0.049) * [Camper or RV] (Hotel) (a) 0.456(0.064) * Type of auto vehicle [Van, Suv] 0.008(0.027) [Pick-up truck, Truck, RV] (Car) (a) -0.081(0.034) * Trip party size -0.060(0.024) * Weekend trip [Weekend] (Weekday) (a) 0.320(0.030) * 9/11 attacks [After] (Before) (a) -0.082(0.025) * F-value 72.79 [R.sup.2] 0.44 Variable t p-value VIF Constant -19.86 <0.001 Total trip distance [Distance] 27.12 <0.001 1.11 Income level [Income] -3.07 <0.002 1.03 Age 1.93 0.054 1.10 Gender [Male] (Female) (a) -1.18 0.24 1.02 Lodging type [VFR] 4.64 <0.001 1.46 [Rented cabin or condo] 4.37 <0.001 1.13 [Own cabin or condo] 9.58 <0.001 1.27 [Camper or RV] (Hotel) (a) 7.19 <0.001 1.34 Type of auto vehicle [Van, Suv] 0.33 0.740 1.13 [Pick-up truck, Truck, RV] (Car) (a) -2.38 0.017 1.21 Trip party size -2.51 0.012 1.08 Weekend trip [Weekend] (Weekday) (a) 10.55 <0.001 1.05 9/11 attacks [After] (Before) (a) -3.35 <0.001 1.07 F-value <0.001 [R.sup.2] * Significant to within 5%.; ** Significant to within 10%. (a) Parenthesis refers to variable used as a reference group Table 3 Lodging Demand Function of Drive Tourism by Using Zero- truncated Poisson Regression Variable Parameter Estimate (Std. error) Constant -4.047(0.112) * Total trip distance [Distance] 0.773(0.016) * Income level [Income] -0.098(0.025) * Age 0.068(0.009) * Gender [Male] (Female) (a) -0.037(0.028) Lodging type [VFR] 0.046(0.036) [Rented cabin or condo] -0.070(0.066) [Own cabin or condo] 1.033(0.046) * [Camper or RV] (Hotel) (a) 0.658(0.056) * Type of auto vehicle [Van, Suv] -0.198(0.032) * [Pick-up truck, Truck, RV] (Car) (a) -0.302(0.047) * Trip party size -0.301(0.038) * Weekend trip [Weekend] (Weekday) (a) 0.793(0.032) * 9/11 attacks [After] (Before) (a) -0.128(0.029) * Log-likelihood -4097.10 Pseudo [R.sup.2] 0.3998 Variable t p-value Constant -36.14 <0.001 Total trip distance [Distance] 47.23 <0.001 Income level [Income] -3.86 <0.012 Age 7.15 <0.001 Gender [Male] (Female) (a) -1.31 0.189 Lodging type [VFR] 1.30 0.195 [Rented cabin or condo] -1.06 0.287 [Own cabin or condo] 22.64 <0.001 [Camper or RV] (Hotel) (a) 11.76 <0.001 Type of auto vehicle [Van, Suv] -6.19 <0.001 [Pick-up truck, Truck, RV] (Car) (a) -6.38 <0.001 Trip party size -8.02 <0.001 Weekend trip [Weekend] (Weekday) (a) 24.99 <0.001 9/11 attacks [After] (Before) (a) -4.40 <0.001 Log-likelihood Pseudo [R.sup.2] * Significant to within 5 %.; ** Significant to within 10%. (a) Parenthesis refers to variable used as a reference group Table 4 Lodging Demand Function of Drive Tourism by Using Zero- truncated Negative Binomial Regression Variable Parameter Estimate (Std. error) Constant -4.299(0.192) * Total trip distance [Distance] 0.805(0.029) * Income level [Income] -0.123(0.040) * Age 0.048(0.015) * Gender [Male] (Female)a -0.065(0.047) Lodging type 0.117(0.056) ** [VFR] 0.350(0.111) * [Rented cabin or condo] 0.836(0.084) * [Own cabin or condo] [Camper or RV] (Hotel) (a) 0.748(0.098) * Type of auto vehicle -0.074(0.052) [Van, Suv] [Pick-up truck, Truck, RV] (Car) (a) -0.261(0.075) * Trip party size -0.219(0.055) * Weekend trip [Weekend] (Weekday) (a) 0.722(0.049) * 9/11 attacks [After] (Before) (a) -0.158(0.048) * log-likelihood -3298.31 Pseudo [R.sup.2] 0.1601 Variable t p-value Constant -22.38 <0.001 Total trip distance [Distance] 27.51 <0.001 Income level [Income] -3.04 0.002 Age 3.09 0.002 Gender [Male] (Female)a -1.37 0.170 Lodging type 2.06 0.040 [VFR] 3.15 0.002 [Rented cabin or condo] 9.93 <0.001 [Own cabin or condo] [Camper or RV] (Hotel) (a) 7.64 <0.001 Type of auto vehicle -1.44 0.150 [Van, Suv] [Pick-up truck, Truck, RV] (Car) (a) -3.51 <0.001 Trip party size -4.00 <0.001 Weekend trip [Weekend] (Weekday) (a) 14.65 <0.001 9/11 attacks [After] (Before) (a) -3.29 <0.001 log-likelihood Pseudo [R.sup.2] * Significant to within 5 %.; ** Significant to within 10%. (a) Parenthesis refers to variable used as a reference group Table 5 Result of Survival Analysis with Log-logistic Regression Variable Coefficient (Std. error) Constant 0.188(0.050) Total trip distance [Distance] 0.0009 (0.0003) * Income level [Income] -0.0054(0.0044) Age 0.0223(0.0078) * Gender [Male] (Female) (a) -0.035(0.023) Lodging type [VFR] 0.1407(0.0280) * [Rented cabin or condo] 0.4354(0.0597) * [Own cabin or condo] 0.4143(0.0444) * [Camper or RV] (Hotel) (a) 0.3763(0.0519) * Type of auto vehicle [Van, Suv] 0.0023(0.0262) [Pick-up truck, Truck, RV] (Car) (a) -0.0684(0.0346) * Trip party size -0.0254(0.0111) * Weekend trip [Weekend] (Weekday) (a) 0.1557(0.0277) * 9/11 attacks [After] (Before) (a) -0.0820(0.0241) * Log-likelihood -1641.80 Variable Time Ratio t p-value Constant 3.75 <0.001 Total trip distance [Distance] 1.000934 30.84 <0.001 Income level [Income] .9946068 -1.21 0.227 Age 1.02259 2.83 0.005 Gender [Male] (Female) (a) .9654269 -1.50 0.133 Lodging type [VFR] 1.151153 5.03 <0.001 [Rented cabin or condo] 1.545683 7.30 <0.001 [Own cabin or condo] 1.513406 9.33 <0.001 [Camper or RV] (Hotel) (a) 1.457018 7.24 <0.001 Type of auto vehicle [Van, Suv] 1.002337 0.09 0.929 [Pick-up truck, Truck, RV] (Car) (a) .9338493 -1.98 0.048 Trip party size .97499 -2.29 0.022 Weekend trip [Weekend] (Weekday) (a) 1.168586 5.62 <0.001 9/11 attacks [After] (Before) (a) 0.9205766 -3.44 <0.001 Log-likelihood * Significant to within 5 %.; ** Significant to within 10%. (a) Parenthesis refers to variable used as a reference group Table 6 AIC and Log Likelihood Values in Three Regression Models Poisson Neg. Binomial Survival regression regression analysis Log-likelihood -4097.10 -3298.31 -1641.80 AIC 8220.20 6622.62 3309.60 * AIC = -2log(L) + 2K Fig. 2 The number of vehicles that self-drive tourists own 9 0.1% 8 0.2% 7 0.1% 6 0.8% 5 4.5% 4 10.4% 3 22.3% 2 46.0% 1 14.4% 0 1.1% Note: Table made from bar graph.
|Printer friendly Cite/link Email Feedback|
|Author:||Yi, Sangchoul; Day, Jonathon; Cai, Liping A.|
|Publication:||Journal of Tourism Challenges and Trends|
|Date:||Jun 1, 2011|
|Previous Article:||Rural tourism to the rescue of the countryside? Oltenia as a case study.|
|Next Article:||Rural tourism in Lijiang, China, and its impact upon the local minority communities.|