Travel Reference
In-Depth Information
The dependent variable here is the fi nal choice of tourists, which can be formulated either in
terms of whether or not a certain destination is chosen (dichotomous variable) or, which
destination among a few options is chosen (categorical variable). In this situation, it is possible to
use logistic regression, also known as a logit model, to fi nd out the possibility of each outcome
based on the independent variables (the predictors). There are two types of logistic regressions
that are used frequently in tourism destination choice studies, multinomial logit and conditional
logit. Basically, multinomial logit is used to identify the infl uence of individual characteristics
(e.g. Morley 1994) such as demographics or attitudes of tourists in decision-making while the
conditional logit is used for testing the importance of destination characteristics on fi nal choice
(e.g. Seddighi and Theocharous 2002).
One thing to be noted is that importance of the attributes measured by general regression
analysis is represented by the coeffi cient value, which describes the changing ratio between
dependent and independent variables. It is not a measure of the absolute importance value
assigned to each attribute by tourists but rather a value that indicates the elasticity of each
attribute. For example, if we fi nd the coeffi cient values for price and local temperature on
destination choice are 0.4 and 0.2, it does not mean that for tourists price is twice as important
as local temperature or tourists would consider price fi rst then local temperature. It only means
that every unit of change of price level would generate twice the effects on overall preference
than every unit of change in local temperature. Furthermore, regression analysis simplifi es the
complex mental decision-making process into an input-output relationship between independent
variables and dependent variables. The simplifi cation enables statistical calculations for such a
complex problem but it does not allow explanatory insights concerning the true process of
tourists' decision-making.
In recent years, a more sophisticated method named Analytic Hierarchy Process (AHP) has
been widely used in a variety of multi-criteria decision-making fi elds including government,
industry, healthcare and education. The AHP was initially introduced by Saaty (1997) for
operations management studies. It is a methodology that provides a systematic problem-
solving framework. Specifi cally, it enables the researcher to estimate the relative priority
of elements within the hieratical structure by conducting a series of paired comparisons.
Compared with traditional multi-criteria decision-making analysis methods such as the
regressions mentioned above, the respondents found the AHP method required less diffi cult
mental processing since it is quite straightforward and due to the systematic guide provided by
AHP during the comparisons. The respondents perceived the fi ndings about the importance of
each attribute more trustworthy (Schoemaker and Waid 1982). A brief summary of how this
method works is presented below.
Firstly, the AHP decomposes the decision-making problem into a hierarchy. A simple
hierarchical structure of decision-making from the top to bottom is comprised as follows: choice
objective; criteria; sub-criteria; and alternatives (see
Figure 23.1
).
Actually, the criteria can be
further divided into many layers of sub-criteria. Secondly, decision makers pair-wise compare the
criteria (N = 3 in
Figure 23.1
)
at level 2 by expressing their preference between every
2 criteria. For the example listed in
Figure 23.1
,
criterion 1 is two times more important than
criteria 3 but equally important as criterion 2, criterion 2 is two times more important than
criteria 3. These paired comparisons can be formed into a (N ∗ N) preference matrix and then by
using the eigenvector solution it is possible to convert the preference matrix into the numerical
priority values of each criterion. Thus the sum of priority values at each level equals 1.
As can be seen in
Figure 23.1
,
the calculated priority values for the three criteria at level 2
are 0.4, 0.4 and 0.2. Following the same paired comparison and calculation process, the local
priority values for the sub-criteria within each criterion at level 3 can be calculated. In order to
