The sample was divided into two segments based on the following classification scheme. Respondents who report that that more than 50 percent of their trips requiring hotel stays are business related are considered to be business travelers. A total of 169 respondents were classified as business travelers. Respondents who report that more than 50 percent of their trips requiring hotel stays are leisure related are considered to be leisure travelers. A total of 691 respondents were classified as leisure travelers. Those respondents who were split equally, 50 percent business and 50 percent leisure were categorized as business travelers. This resulted in a total of 239 business travelers and 691 leisure travelers. Next, we explain the results of the innovative hotel choice experiment.
Innovative hotel choice modeling results
The primary analysis approach associated with DCA is the estimation of the MNL models based on a maximum likelihood estimation technique (Ben-Akiva and Lerman, 1991). Recall that each respondent had to evaluate eight choice sets, each containing two descriptions of hotels along with the option of not choosing either. Statistical details about MNL model estimation is described in extensive detail by Ben-Akiva and Lerman (1991) and Louviere et al. (2001). A more applied description of DCA and MNL mo del estimation is provided in Verma et al. (2002) and Verma and Plaschka (2003). Louviere etal. (2001) and Ben-Akiva and Lerman (1991) recommend that when estimating MNL models, experimental variables can be "effects-coded" to accurately estimate the relative impact on respondents' choices. The estimated MNL model for this study was statistically significant at the 5 percent level.
Table IV shows the relative impact of each experimental attribute on hotel choice decisions. Recall that we are only focusing on the innovative attributes that were included in the original broader dataset. Therefore, all results presented ignore the non-innovative attributes that were also included in the study. The estimated β weights for the innovative attributes are standardized to be between "zero" and "one" based on the highest and lowest part worth utility of an attribute. By transforming the data linearly, it is easy to compare and contrast the impacts of each attribute to one another. We estimated the relative main effect by subtracting the highest and lowest β weights for a given attribute (Louviere etaL, 2001). The main effects allow us to compare the overall impact of changing the levels of innovative attributes in hotel choice against each other.