Business fixed investment and nonlinear time series analysis

This dissertation explores two topics. Chapters I and II investigate the representative firm's decision-making about business fixed investment (non-inventory) spending. Chapter III introduces a new class of time-trend specification in econometrics.;For a firm, investment is a frictional process of adjusting its capital stock from an inherited to a desired level. Thus, factors which effect the investment decision are related to which sources of friction are included in the analysis. From this point of view, Chapter I surveys the related literature.;Chapter II analyzes the empirical importance of firm-specific external borrowing cost on the level of investment spending. Though the effect of external borrowing cost (in a form of credit rationing) on the firm-level investment behavior is widely acknowledged, previous research emphasizes the validity of standard investment models for a non-credit rationed firm. Thus the core of earlier analyses is to find sample-splitting criteria which distinguish non-rationed firms from rationed firms.;In contrast, I modify the standard neoclassical model by incorporating the costly external borrowing premium, and derive an investment function where the debt-capital ratio affects investment spending. The model is estimated by a modified GMM using panel data of 80 U.S. manufacturing companies. The standard firm's borrowing cost is found to be approximately nine percent of the usual capital installation cost.;Many growing macroeconomic time series appear to contain persistent smooth components called trends. It has been common to model such trend components by a random walk with a positive constant drift. Chapter III extends this idea, and introduces a generalized random walk process, which I call non-linear stochastic trend. Theoretical basis and a various specifications of growth processes are presented. Empirical examples include univariate models and a bivariate error correction model with a nonlinear stochastic common trend. This chapter introduces an estimation procedure of the bivariate error correction system, and re-examines the cointegrating relationship between U.S. real consumption and disposable income.