Bayesian Estimation of Capital Stock and Depreciation in the Production Function Framework

Abstract

We propose a Bayesian approach to estimating productive capital stocks
and depreciation rates within the production function framework, using annual
data on output, employment and investment only. Productive capital stock
is a concept related to the input of capital services to production, in contrast
to the more common net capital stock estimates, representing market value of
fixed assets. We formulate a full Bayesian model and employ it in a series of
illustrative empirical examples. We find that parameters of our model, from
which the time-path of capital is derived, are weakly identified with the data
at hand. Nevertheless, estimation is feasible with the use of prior information
on the production function parameters and the characteristics of productivity
growth. We show how precision of the estimates can be improved by augmenting
the model with an equation for the rate of return.