2011 International Conference on Alternative Energy in Developing Countries and Emerging Economies
- 189 -
After giving some factual regarding biodiesel fuel, the
main finding from the questionnaire showed that the
Bangkok survey residents, it was clear that the high price
of biodiesel fuel (4.22) was the greatest barrier to the
purchase of biodiesel fuel for the Bangkok residents. The
engine start factor (2.52) was the lowest ranked
obstruction in causing people to purchase biodiesel fuel
(Fig. 4). Furthermore, suppose the Bangkok residents
would have to buy biodiesel fuel at £0.5470 (40 Baht) per
litre as a possible biodiesel fuel price to completion with
diesel fuel at present, they would firstly be concerned
about the vehicle’s engine (3.55) such as high lubricates,
smoother operation, complete combustion and top speed.
On the other hand, economic reasons (1.75) such as
creating jobs and distributing income to farmers, was the
lowest ranked reason concerned by using the Friedman
test at a level of significance 0.05 (Fig. 5).
4.22
2.85
2.75
2.66
2.52
High price
Smooth engine
Operting cost
Start
engine
Switching
1
0
2
3
4
Fig. 4. The factors that affected biodiesel fuel concern.
3.55
3.48
3.30
2.91
1.75
Engine
Operating
Environment
Economy
Aesthetics
1
0
2
3
Fig. 5. Factors behind willingness to pay for biodiesel fuel at £ 0.5470
per litre.
Note: Figure from 3 to 5 on a scale ranging from 1= the lowest
concerned to 5 = the highest concerned. The numbers above
diagrams were the mean rankings by the Friedman’s test.
The conditional backward method of binary logistic
regression was the procedure utilizing a maximum
likelihood paradigm. The model is able to account for
most of the variance and therefore best suited to
predicting preferences for biodiesel fuel (H-L test Sig . =
0.000
1
, Pseudo R square = 0.115
2
).
The model
specification found to have the best fit of explanatory
variables with the most statistically significant
coefficients (Table II).
The overall fit of the model was acceptable by the
conventional standards used to describe probabilistic
choice models [9]. All of the fuel characteristic attributes
are significant factors in the choice of biodiesel fuel
characteristics scenario. All main attribute factors were
the coefficient signs with a priori expectations: FC, OC,
CN and PE. The sign of all attribute coefficients were
highly statistically significant at the 95% confidence level
(at the coefficient value each of the main attributes in
Table II, the values were 0.000). This indicated that most
attribute levels were negative on FC, OC and PE. For
example, as the price of biodiesel increases, the utility of
biodiesel decreases, and as pollution emission decreases,
the utility of biodiesel increases. On the other hand, the
cetane number was a positive coefficient estimate
implying that, with an increase in the cetane number, the
utility of biodiesel increases. In addition, it might be
implied that the respondents had the most concern with
fuel cost (FC), operating cost (OC), cetane number (CN)
and pollution emissions (PE), respectively (see the
coefficient value of each attribute in Table II).
To account for heterogeneity of preferences across
respondents, the effects of the social, economic and
attitudinal characteristics of the respondents on their
choice of fuel characteristic scenario must be
investigated. On account of possible multicollinearity
problems, it is not possible to include all the interactions
between the social, economic and attitudinal
characteristics of the respondents collected in the survey
(Table II) and the four fuel characteristic attributes when
estimating a basic conditional logit (CL) model with
interactions [10]. The CL model with interactions model
has a higher overall fit compared to the basic CL model
because of Pseudo R Square of 0.123 and log-likelihood
values of -7105.02. The coefficients of the main attributes
were all highly significant (the coefficient value of each
main attribute can be seen in Table II; the values were
0.000). Moreover, all main attributes using the interaction
model have larger coefficients than the basic CL model.
The coefficient signs of the CL model with interactions
were similar to the basic CL model. Of course, the socio-
economic variables and interaction of respondent-specific
characteristics with choice specific attributes were
influenced by the explanatory variable.
When considering the socio-economic variables, only
one variable was statistically significant. Gender was
positively related to choosing biodiesel at 95%
confidence level. Female residents preferred to choose
biodiesel fuel rather than male residents. Socio-economic
variables such as family size, family income, hours per
trip, number of cars in family and expenditure on fuel
were not found to have a significant effect at 95%
1
The model adequately fits the data. Hosmer-Lomeshow (H-L) test
is a test of goodness of fit.
2
The R
2
value in the logit model is similar to R
2
in conventional
analysis except that significance occurs at lower levels.
