I. Introduction and summary of main conclusions, 684. — II. A model with three tradable goods usable in production or consumption, 588. — III. Intermediate products, production technology, and the gains from trade, 601. — IV. The structure of production in protected industries, 610.

The choice of methods of production as related to factor costs is formulated differently from the ordinary production function by means of a particular activity set. Continuous factor substitutability is maintained in an ex ante sense. The resulting formulation is more amenable to existing data and is applied via a distributed lag model.

In the 1950's and 1960's a strong dollar standard existed; under fixed exchange rates the monetary policies of most nations were governed by the stable monetary policy of the United States. In the 1970's, the dollar standard was reduced to a weak form where industrial (but not less-developed) countries float their exchange rates to secure more domestic monetary independence, but the financial processes underlying international trade itself remain dollar based. What are the elements of the weak dollar standard necessary for world trade to remain largely monetized and multilateral? How can nations avoid a relapse to the bilateralism and barter that characterized the 1930's? I shall consider first, necessary restraints on foreign exchange intervention and reserve holdings by central banks of industrial economies other than the United States; and second, the proper American monetary policy for reconciling the weakened international role of the dollar with domestic monetary stability.

In Currency Substitution and Instability in the Dollar Standard (1982), Ronald McKinnon hypothesized that shifts in international portfolio preferences for dollar assets -including bonds -destabilized the effective demand for money in the United States. From the dollar depreciation of 1971-73 to the great dollar appreciation of 1981-83, these demand shifts were telegraphed by large changes in the dollar exchange rate against other hard currencies. They signaled sudden inflation in the United States when the dollar was unexpectedly weak, and deflation when the dollar became strong. In addition, McKinnon argued this ebb and flow in the demand for dollar assets provoked foreign central banks, but not the U.S. Federal Reserve System, to adjust their money growth rates to mitigate these exchange fluctuations. Consequently, since 1970, annual percentage growth in World MI-the sum of percentage growths in dollars, marks, yen, sterling and so on fluctuated more than annual money growth in the United States. The resulting international business cycle had a first-order impact on American income and prices. Other than discussing some suggestive money and price data for ten industrial countries, McKinnon provided no formal econometric testing of his theory. This note summarizes an exchange between McKinnon and Tan (M-T) and Radcliffe, Warga, and Willett (R-W-W) prompted by the latter's 1984 econometric test of McKinnon's hypothesis. Copies of the full exchange, including new econometric work, are available from the authors.' Consider the single-equation econometric technique of explaining U.S. prices or incomes by current or lagged changes in U.S. Ml. All authors agree that there was a significant deterioration in the fit of this basic monetary equation from the fixed exchange rate period of 1958-69 to the era of floating rates from 1972 to 1982. By itself, American money growth now gives a less satisfactory explanation of cyclical fluctuations in nominal income or prices. But how can changing international asset preferences be represented statistically in a mixed exchange rate regime of dirty floating? One proxy variable is money growth in the rest of the industrial world: MlROw. McKinnon (1982) originally hypothesized that world money inclusive of MlROW has a stronger impact on American prices than U.S. money by itself. But this conjecture turns out to be true only for American (and world) tradable goods prices-as approximated by the wholesale price index (WPI). (See McKinnon-Tan, 1983). Radcliffe, Warga, and Willett correctly pointed out that the influence of MlROw is not helpful in predicting changes in U.S. nominal GNP. Myles Wallace (1983) showed that domestic price indices, such as the American CPI or GNP deflator, which have large nontradable components are not well explained by world money. R-W-W's objection is important because monetary variables