LEARNING ABOUT OTHERS ACTIONS AND THE INVESTMENT ACCELERATOR

The paper considers a simple general equilibrium model of investment determination. In this type of model, the resulting sequence of investment levels and capital stocks are in general independent of the past history of output unless some gestation lags are introduced. It shows that when there is imperfect observability of other agents' actions and, consequently, of some pay-off relevant information, aggregate output as well as other variables would be used to update expectations. This would naturally lead to output Granger causing investment. Therefore the investment accelerator is the outcome of "learning' in the economy. The existence of externalities and learning add to the persistence of the business cycle fluctuations. The evidence from US and UK data is mixed but it seems to be the case that output has some predictive power for investment and employment. To discriminate between the explanation based on learning and other explanations, it tested whether the predictive power of output falls as its variance increases. This is so in the UK where the conditional variance of output changes over time. The evidence thus offers some support for our theory. -from Author