| This dissertation explores the dynamic adjustments and structure of real exchange rates in 32 Sub-Saharan African countries over four decades (1960–2000). It consists of three essays. The first paper investigates the Purchasing Power Parity (PPP) hypothesis using panel unit root tests based on institutional integration and identical country characteristics. Using French franc (FF) and German mark (GM) as numeraires, we find some support for PPP in a development group, SADC and in an economic community, COMESA. No evidence is found in CFA-franc zone, a monetary union. We interpret this result as the real exchange rates for developing nations in economic and development unions tend to support PPP; whereas, those in a monetary union weaken it.;The second paper examines the linear dynamic adjustments of real exchange rates employing sequential univariate and panel unit root tests with a variant of structural changes. We find British pound-based real exchange rates are characterized as stationary around one to two broken trends. Next, we conduct structural change tests in a univariate and multivariate context and find evidence of one to two breaks in most series. Most common break dates, occurring in late 1970s, early to mid-1980s, and early 1990s, reflect high inflation induced by the second oil price shocks, debt crisis factored by drastic deterioration of the terms of trade, currency realignments, and currency devaluation policies instituted by the institutions of Bretton Woods through the structural adjustment programs.;The third paper uses the Exponential Smooth Transition Autoregressive (ESTAR) model to examine the nonlinear dynamics of the real exchange rates in the post Bretton Woods. It intends to assess whether the failure to provide enough evidence for PPP and the existence of underlying unresolved puzzles are due to linear model misspecifications. We find nonlinear mean reverting processes adequately characterize all series, in which case PPP holds true. The GM-based real exchange rates are slightly more persistent than the GM-based real exchange rates. Finally, the half-lives computed using Monte Carlo integration range from 1.5 to 3 years and are similar to those reported for major real exchange rates by Taylor et al. (2001). |
