Equipment Expenditures by Input-Output Industries

Abstract

N the past two decades our understanding of the investment process has expanded greatly. Central to this expansion is the notion that the desired or stock of capital depends, not only upon the level of output, but also upon such factors as prices, interest rates, technological change, and tax policy. Equally central is the closely related proposition that the actual stock of capital adjusts to its optimal value only after a considerable lag, possibly several years. In contrast, dynamic input-output forecasting models typically treat investment in much the same fashion as early accelerator theories: for each industry net investment is given by the change in output multiplied by a fixed capital coefficient. Thus lags and endogenously determined changes in the capital coefficients cannot be accounted for without considerable difficulty. However valuable this traditional treatment may have been in the past (and there is no doubt that it was), a wholly different approach is now required if input-output models are to benefit fully from accumulated knowledge about the investment process. In particular, fitted investment equations, rather than matrices of capital coefficients, should be the basis for forecasts. A primary purpose of this paper is to estimate such a set of equations for 68 equipment purchasing sectors as defined in the 1958 inputoutput study [10]. The regression model, first developed by the author in [19] and subsequently used by Robert Coen [3], is a generalization of work done by Robert Hall and Dale Jorgenson [11]. Their model, in addition to handling the conventional determinants of investment, is able to account for changes in a wide variety of tax policy variables: depreciation methods, tax lives, emergency amortization provisions, and investment tax credits. These variables affect investment by altering the user cost of capital and, consequently, the optimal stock of capital. Unfortunately, the extent to which changes in the user cost affect the optimal stock of capital is not estimated by Hall and Jorgenson. Rather, the effect is constrained by assuming a Cobb-Douglas production function. To remedy this deficiency, the generalized approach presented here actually estimates the impact of the user cost on optimal stocks with the aid of a constant elasticity of substitution (CES) production function [1]. Since the Cobb-Douglas is a special case of the CES, estimates of the generalized model provide a test of the Hall-Jorgenson hypothesis. The plan of the paper is as follows. Sections II and III outline the theory of investment underlying the regression model and explain the various assumptions and sources relied upon in the construction of the data. Then, section IV presents the regression results and summarizes the most important findings. Finally, in order to illustrate the usefulness of the fitted equations, section V estimates the impact on investment of a hypothetical repeal of the tax credit for 1962-1967.