Files
Abstract
In econometrics there is a long history of using continuous functions to force
distributed lag coefficients to behave in an economically accepted way. For
example, geometrically declining lags have often been used to model coefficients
that we believe should be declining. Polynomial lags have been used to model lag
coefficients expected to increase and then decrease. In this paper a more flexible
way of imposing such prior information is investigated. Inequality constraints are
used to impose knowledge about the relative magnitudes of coefficients without
forcing them to lie on a smooth continuous curve. A Metropolis algorithm is used
to get posterior density functions for the lag coefficients and functions of those
coefficients for the Nerlove orange data and the Almon capital expenditures data.