A Test of Separability and Random Effects in Production Function with Decomposed IT Capital

Abstract

In productivity analysis, many studies have used real value-added function for estimating productivity. These studies have made explicit or implicit assumption that real value-added function exists. As real value-added is the residual of real gross output from real intermediate input through the double deflation method, the existence of real value-added function is not guaranteed automatically. In order to test this, we have used an additively strong separability test. We could not accept the existence of real valueadded function from the data of 32 industries during the period of 1981~2002 in Korea. This means that it is more appropriate to use gross output based productivity rather than value-added based one. In addition, in order to identify the contribution of IT investments, we have decomposed capital stock into IT capital stock and non-IT capital stock. We have failed to find the evidence that IT capital has increased productivity in the entire economy which supports the Solow(1987) paradox. However, when we decomposed the industries by the IT capital intensity, there is a significant contribution of IT capital to gross output in the highly IT capital intensive industries. This phenomenon is related to the substitution elasticities between IT capital and non-IT capital.