Developing consistent marginal effect estimates in a simulataneous equation model with limited dependent variables

Type Report
Title Developing consistent marginal effect estimates in a simulataneous equation model with limited dependent variables
Author(s)
Publication (Day/Month/Year) 2013
URL https://www.researchgate.net/profile/Joseph_Atwood/publication/267254817_DEVELOPING_CONSISTENT_MARGI​NAL_EFFECT_ESTIMATES_IN_A_SIMULATANEOUS_EQUATION_MODEL_WITH_LIMITED_DEPENDENT_VARIABLES/links/558331​4f08aefa35fe30b7b6.pdf
Abstract
Economic studies often utilize data that is the result of simultaneously determined choices and/or
simultaneously occurring economic forces. Examples of simultaneous determination include
market prices and quantities, work/leisure decisions, and technology adoption. A common and
complicating feature of many studies is that one or more of the observed endogenous variables
may be observed only as a limited dependent variable. Later in the paper we discuss the results
of a study with simultaneously determined limited dependent variables occurring as discrete
indicators of whether given technologies was adopted or not. In development research the
researcher is commonly interested in examining or identifying the marginal effect of one or more
personal or demographic characteristics and/or governmental policies upon the likelihood of
adopting a particular set of technologies. In our study we are interested in the factors influencing
the likelihood of a developing country?s producers choosing to utilize improved seed varieties
and chemical fertilizer. In this situation it is probable that the decisions of whether to use
improved seed and/or chemical fertilizer are made concurrently and both are likely to be
dependent upon unobserved latent variables. For our study, we use the classical assumption of an
unobserved random utility function in which the decision to adopt is yes if the expected utility
exceeds some threshold level
?
. While we utilize a pair of classical probit models below, the
complications discussed in this paper are relevant over a broader range of models including
multistage decision models and systems with combinations of continuous, truncated, and/or
binary endogenous variables. Indeed this paper?s example is a small part of a larger study in
which we examine a number of indicators of technology use including combinations of
continuous, truncated, and binary endogenous variables. Given the combinations of continuous,
truncated, and binary endogenous dependent variables in our broader study, an obvious choice
was to consider the two stage process presented by Nelson and Olson (1978).

Related studies

»