Project CEMAPRE internal
|Title||Bayesian Validation of Behavioural Parameters in Economic Models|
|Participants||Rui Paulo (Principal Investigator), João F.D. Rodrigues|
|Summary||Models for assessing the impact of economic policies often require the specification of two types of|
parameters: structural parameters, which can be calibrated using data from a baseline year, and
behavioural parameters, which require econometric estimation and/or assumptions which are not
amenable to direct testing. In this paper we address the problem of validating behavioural
parameters given a set of empirical observations that can be compared with the model predictions.
Our starting point is Bayesian model selection, which allows for assessing the change in a
practitioner's belief in different models given observations. We depart from the classical formulae
because a behavioural parameter is defined within a model and the practitioner has different models
at his/her disposal. We therefore introduce an explicit probability that an empirical observation
falls within the scope of a particular model. The Bayesian validation method developed here is
applied to the selection of technology assumptions in supply-use models, which disambiguate how an
industry uses a given input in the production recipe of multiple outputs, using the WIOD database as
a case study.