How would our model behave if the demand for money were made to depend on income, in addition to depending on interest rates and wealth? Although we have argued that this specification is inconsistent logically with other features of our model, especially its setting in continuous time, the question is worth answering. Many other models dealing with our subject–the functioning of pegged and flexible exchange rates and implications of asset-market integration–adopt this specification. It is thus useful to compromise with logical consistency in order to achieve comparability. More generally, we need to know how heavily our own results depend on the specification adopted in Part II and employed again in Parts IV and V.

When income affects the demand for money, asset markets cannot be partitioned from goods markets. It is still possible, of course, to classify disturbances according to the markets in which they originate. It is no longer possible, however, to claim that the exchange rate is determined in the short run by money and bond markets. The demands for assets respond immediately to any disturbance or policy change affecting the level of income. Does this mean, in turn, that a flexible exchange rate can confer continuous insulation? Does it mean that we must modify our conclusions about monetary and fiscal policies? Will it still be true, for example, that the impact effect of a tax cut is unaffected by the exchange-rate regime or by the degree of asset-market integration?