As shown in the preceding chapters, early neo-classical and accelerator models dealt with uncertainty in an unsatisfactory way, super-imposing distributed lag structures on top of essentially static models in order to capture expectations of the future. Quite apart from the estimation problems that accompany the use of distributed lag structures to capture the different causes of investment lags, Jorgenson’s critics argued that these refinements lacked the theoretical basis that an explicit behavioural assumption about rationality provides. In response, q models of dynamic optimization were formulated by incorporating uncertainty directly: expectations appear explicitly in the firm’s optimization problem and can be linked directly to underlying assumptions about technology and expectations. A key problem for the empirical estimation of these models is that expectations of the future are unobservable but some solutions to this problem are outlined in this chapter, the most influential of which is probably Brainard and Tobin’s q approach, in which expectations are captured using Stock Market data. Other solutions discussed include the ‘transformation’ and ‘direct forecasting’ approaches identified by Chirinko (1993). The outcome of empirical work on adjustment cost models, whilst more encouraging than results from the estimation of Jorgenson’s theory, were still empirically disappointing, but did pave the way for real options models of investment, outlined in Chapter 9.
KeywordsCapital Stock Rational Expectation Adjustment Cost Investor Sentiment Efficient Market Hypothesis
Unable to display preview. Download preview PDF.