Expectation Revision and Optimal Learning in a Foreign Investment Model

Abstract

BETTER DECISIONS CAN BE MADE with improved information. The anticipation that more information will be available tomorrow generally influences today's decisions which themselves may generate information as a by-product. Investigations of this two-way relationship may explain market phenomena which might otherwise seem irrational. The effects of and learning can be quite diverse, depending on the exact nature of the and of learning. This diversity does not seem to have received emphasis in recent contributions to the theory of adaptive optimization; see, for example, [3], [6] and [7]. Furthermore, existing models assume that only one side of the market learns. If both sides learn, today's decisions must take account of the fact that each side's later decisions depend on the new state of information. In order to explore some of the diverse implications of learning and a model of foreign investment is studied below under alternative specifications about learning. Basically, we consider the case of learning about an unknown parameter. Two polar types of are postulated: type one uncertainty, where the value of the parameter becomes known after one observation, and type two uncertainty, where the subjective beliefs about the value of the parameter are modified after each observation, but complete learning is not possible. Under type one uncertainty we show that the truly optimal level of foreign investment may differ from the myopic level: it may be higher, to encourage foreign learning, or lower, to prevent such learning, depending on the initial states of information (beliefs) and the nature of the foreign supply curves of capital. In particular, when the foreign investment is in the exhaustible resource industry, we show that the irreversibility of resource depletion dictates a cautious approach, with a bias in favor of less initial extraction, even if all economic agents are risk neutral. Essentially, this is because the expected benefits of an irreversible decision should be adjusted to allow for the resulting loss of options. This result is in agreement with that of Arrow and Fisher [2], where only the government learns. In addition, we show that the more is the and hence the scope of learning, the more pronounced is the bias in favor