Dynamic Q-investment functions for Germany using panel balance sheet data and a new algorithm for the capital stock at replacement values

The paper explores the investment behaviour of German firms in the context of the Q-approach, which plays a dominant role in empirical investment research. The analysis is based on the Deutsche Bundesbank's corporate balance sheet statistics. The panel data set contains some 2,300 German firms' balance sheet data covering the years 1988-1998. While the Q-theory is mainly applied on the basis of stock market data, which facilitates the exploitation of market expectations and the calculation of average Q, the direct forecasting approach (Chirinko 1993) suggested by Abel and Blanchard (1986) and extended to panel data by Gilchrist and Himmelberg (1995, 1998) enables the Q-theory to be applied to non-quoted firms which are by far the majority in Germany. One of the key variables when using balance sheet data, which has attracted much detailed research, is firms' net capital stock at replacement costs. The challenge is to transform historical cost data, depreciated at non-economic, tax-oriented depreciation rates, into unreported and probably unknown economically meaningful data at actual replacement values. We suggest a complex procedure for calculating reliable replacement values of a firm's capital stock. To calculate Q we follow two different operationalisation strategies. First we estimate average Q based on balance sheet data by forecasting the present value of future profits using a VAR model. Second, we estimate marginal Q following the approach suggested by Gilchrist and Himmelberg. We compare the results from two different estimation techniques for dynamic investment models, GMM and direct bias correction. The results show that marginal as well as average Q influence investment significantly. When classifying the firms by size, we find that smaller firms react more strongly to Q and, to a lesser extent, to lagged investment.