| Abstract: | A common problem with micro‐level analysis is that capital stock data is missing. Typically, a feasible measure of capital is calculated by accumulating investment flows from an initial value of the capital stock. As the time dimension of most disaggregated data is rather short, the choice of this initial value can have significant effects on the resulting capital estimates. Most empirical studies impute the initial value using a single arbitrary proxy. In this paper, we propose a panel data framework that assigns weighting coefficients to multiple proxy variables. We conduct a series of Monte Carlo experiments to test the performance of the proposed method and apply the method to a U.S. manufacturing dataset. The results suggest that our method improves the approximation of the capital stock and thus in turn reduces the bias in the production function estimation. |
