An econometric estimation of the yen-dollar risk premium in a portfolio balance model with rational expectations

This dissertation has three objectives. First, to obtain maximum likelihood estimates of the yen-dollar risk premium by embedding a portfolio-balance model of the mean-variance type in a small open economy macromodel with rational expectations. Second, to test whether the cross-equation restrictions generated by the rational expectations solution of this model are satisfied. Third, to test the null hypothesis of constant risk premium against an alternative that the change in U.S. monetary policy from controlling interest rates to controlling aggregate reserves during 1979:4-1982:3, increased the exchange risk premium significantly during this period compared to the period prior to the change. We estimate our model for Japan using quarterly data for the period 1973:2-1987:4. We find evidence for risk aversion: a percentage change in the holdings of U.S. foreign assets by Japanese investors caused the yen-dollar risk premium to change by 0.96 of a percentage point over the fourteen year period 1973-87, or 0.069% on a per annum basis. Second, we find that our constrained rational expectations model cannot be rejected by the likelihood ratio test. Finally, we find that the risk premium is very large in magnitude (the risk parameter not statistically different than zero) if estimated over the period 1973:2-1982:3, in contrast to the period 1973:2-1979:3, when the old policy was in effect. Although our results support our hypothesis, they should be interpreted with caution because our estimation includes the overlapping sub-period 1973:2-1979:3, during which the old policy was in effect. The limited number of quarterly observations (12) for the sub-period 1979:4-1982:3, prevented us from estimating our model exclusively for this period. Using monthly observations would have required a more complex lag structure in our model, making the solution analytically intractable. In conclusion, our results are interesting compared to those of previous researchers, who failed to reject the null hypothesis of risk neutrality and zero risk premium. Despite the limitations of our study that stem from our assumptions about the structural model, this dissertation has shown that by incorporating rational expectations explicitly into the portfolio-balance framework, progress can be made towards obtaining satisfactory estimates of exchange risk premium in a small open economy macromodel.