In the diversity of exchange rate regimes in sub-Saharan Africa (SSA) in the 1980s, there was a trend toward more flexible regimes and smaller parallels markets. How particular exchange rate arrangements affect such factors as output supply cannot be determined for SSA countries on the basis of experience in other developing countries, because SSA countries differ in the composition of their exports and imports, in level of industrialization, and in development of the financial sector. The authors supplement a survey of the literature with empirical testing, using pooled time-series and cross-section data for 22 countries in SSA for 1971-91. Among their findings: (a) When macroeconomic policies are inconsistent and there is a failure to adjust to adverse shocks, fixed regimes lead to overvaluation and the development of widespread parallel markets for foreign exchange. (b) SSA countries have attempted exchange-rate unification through occasional devaluations, a crawling peg, official dual markets, foreign exchange auctions, and a market pricing rule. Most such experiences have been gradual, and their outcomes mixed. Success in exchange-rate unification (as experience in Ghana and Uganda shows) depends on three crucial elements: supportive monetary and fiscal policy, external budgetary and balance-of-payments support, and official commitment to a credible reform process. (c) In the face of significant adverse real shocks (internal and external), built-in monetary and fiscal rules in fixed exchange rate regimes with currency convertibility (as in the CFA zone) may be inadequate to bring about the required short-term adjustment. As elsewhere in the developing world, the effect of real devaluation on output in SSA is mixed in the short run (contractionary when demand elasticities are low) and neutral in the long-run. Real depreciations have neutral effects on per capita growth of real output in the transition to steady state. (d) Farm producers in SSA respond to price incentives for a single agricultural crop as farmers do elsewhere in the developing world: they behave rationally. And econometric evidence confirms that growth in agricultural exports is not achieved at the expense of food production. (e) Applying the Granger-causality test to this data set reveals strong causality, running in both directions, between money growth rates and inflation. It also shows causality running from output to inflation, and from inflation to nominal devaluation.