MODELLING ENERGY SAVING TECHNOLOGY CHOICES IN DUTCH GLASSHOUSE HORTICULTURE - Economics of Sustainable Energy in Agriculture - page 39

modelled either in structural or reduced form using the optimal stopping
framework and dynamic programming ( e.g. Rust 1987). In this study the
model is first derived in structural form and then approximated by a
reduced form specification. Next, discrete choices in energy saving
technologies are estimated by explicitly allowing for a flexible error
structure. Individual, time constant (random) effects and first order serial
correlation in the choices are special cases of the general error structure in
the model. Controlling for random effects in estimating the firm operators'
technology choices is important because the choices may be affected by
unobservable factors related to the firm operator ( e.g. education and
managerial ability) and firm ( e.g. climate). It is also important to control
for serial correlation since the technology choices are persistent over time
due to the presence of adjustment costs.
The model is estimated using a Monte Carlo simulation technique,
known as the GHK simulator, developed in the early 90's by Geweke,
Hajivassiliou and McFadden, and Keane (Keane 1993; Hajivassiliou 1993).
Hajivassiliou et al. (1996) compares a number of probability simulators
and finds that the GHK simulator outperforms all other methods by
keeping a good balance between accuracy and computational costs. This
simulation approach is particularly tractable in simulating probabilities for
multiple choices that would otherwise require multidimensional integration
and intensive computation. Further, the GHK-method can easily be
extended to modelling choice alternatives recursively as sequences of
choices such that the choice sequences exhibit both time constant
individual effects and flexible time series characteristics. The method is
used for analysing factors determining the choice of energy saving
technologies by Dutch glass house firms using panel data over the period
1991-1995. The results obtained by the SML method are compared with
the results from a standard multinomial logit model.
The chapter proceeds with the presentation of the theoretical and
empirical models underlying the choice of an energy saving technology,
followed by a description of the data and a discussion of the results.

2. THEORETICAL MODEL

This section elaborates on a framework in which operators of glasshouse
firms decide among K possible technologies in each of N (finite) discrete
periods of time. Alternative technologies are indicated by a dummy

variable with if technology k is chosen at time t and

otherwise. The condition indicates that alternatives are