Irreversible investment, uncertainty and hysteresis: a New Zealand investigation

Abstract

This thesis investigates the propensity of aggregate investment to display hysteresis, whereby a temporary exogenous shock leaves behind permanent effects, even once the shock has departed from the system. Chapter 1 reviews traditional investment models and critiques their treatment of uncertainty. The recently developed theory of irreversible investment under uncertainty is set out in Chapter 2. Rather than invest immediately upon a positive present value of profits, it may be optimal for a firm to wait and see how uncertainty resolves itself. To invest/abandon requires a profit/loss larger than the option value of waiting. Recessionary hysteresis results when a large negative shock forces the representative firm to abandon; next period, the shock disappears, once more returning the firm to inertia, but leaving the economy with a permanently lower capital stock. Chapter 3 considers several complications. Firstly, the contradictory nature of existing theories' attempts to combine irreversibility with perfect competition is set out, and given irreversibility, it is shown that hysteresis can still result under 'competition'. Secondly, it is shown how credit rationing makes recessionary hysteresis more likely. Finally, aggregation problems mean that a minor change in uncertainty might have minor effects, rather than no effects at all. Chapter 4 considers specific channels for uncertainty such as output prices, input costs, interest rates, exchange rates, and the general environment. It is concluded that theoretically, an increase in uncertainty depresses investment. The econometric investigation tests two hypotheses arising from the foregoing: (i) does certainty of an up/downturn in the general business situation have a positive/negative effect on business investment ? (ii) does the inclusion of uncertainty remove structural breaks from an otherwise well-specified investment equation ? A novel entropy-related measure of uncertainty is constructed and estimation of data primarily gathered from the R.B.N.Z. Model XII proceeds using the Engle-Granger two step error correction technique and Hendry's 'general to specific' modelling procedure. Notwithstanding aggregation problems, strong support is found for the first hypothesis when lagged uncertainty is considered, while the second hypothesis sees functional misspecification and/or structural breaks removed from an otherwise well-specified equation.