Agriculture Reference
In-Depth Information
houses. The result indicates that incentives to invest in energy saving
heating technologies are higher for firms with small stocks of capital in
invested in structures. The results may also imply that it is optimal first to
depreciate the capital invested in structures before investing in new heating
systems.
Capital invested in machinery and installations has a positive and
significant effect on adoption of energy storage and a co-generator with
energy storage. Therefore, firms with a larger stock of capital invested in
machinery and installations are also more likely to make an additional
investment by adopting a co-generator with storage.
Vegetable firms and cut flower firms more likely adopt standard
heating with storage than pot plant firms do. However, vegetables firms
have smaller probability of adopting co-generators (significant at 8%) than
all other firm types. Firm size is increasing the probability of investing in
heating storage but is decreasing the probability of investing in co-
generators. Therefore, scale economies are important in getting benefits
from the heat storage but co-generators are technologies that are
sufficiently flexible to generate benefits in small firms.
Parameters estimated by the SML-method differ from the
corresponding parameters estimated by the multinomial Logit model (see
Tables 4.2 and 4.3). Firm size has a negative impact on investments in co-
generators in the multinomial logit model and a positive impact in the SML
model. Another difference is that vegetables firms have a lower probability
of adopting co-generators in the SML model, whereas the impact is
insignificant in the multinomial model. The joint significance of the
parameters in and and the differences found here suggest that firm
specific individual effects and serial correlation in the error terms play an
important role in investments and choices of heating technologies.
The mean log likelihood function had value -0.411 in the SML-
model and -0.430 in the Multinomial Logit model. In the restricted model,
including only the intercepts, the corresponding values were -0.713 and
-0.697. The goodness of fit of the SML and Multinomial Logit model is
assessed using McFadden's for the system of two choice equations
6
.
Values of 0.42 and 0.38 are found for the SML and Multinomial Logit
models, respectively. These values indicate that the goodness of fit of the
SML is rather good, in particular when taking into account the common
problem of modelling low frequency decisions (
e.g.
Dorfman 1996).
Elasticities of the choice probabilities to changes in the model
variables are found in Table 4.4. The elasticities indicate the relative
impact of changes in model variables on the choice probabilities of
different technologies;
i.e.
the impact is corrected for the measurement
scale of the variables. The parameter estimates in Table 4.2 are not
Search WWH ::
Custom Search