Testing for Purchasing Power Parity: Econometric Issues and an Application to Developing Countries

Abstract

There is now a vast literature on testing purchasing power parity (PPP). Any test is conditional on a particular econometric speci¢cation which embodies a set of auxiliary assumptions. This paper reviews the issues involved in econometric speci¢cation and estimation in the time series and panel models used to test PPP. We start from a general model and then systematically examine the implicit restrictions that are imposed to obtain the standard procedures and discuss the implications of these procedures for estimation and inference. The issues are illustrated on data for a panel of 31 developing countries, 1966^90. " IntroductionThere is now a vast literature on testing purchasing power parity (PPP) using national data on exchange rates and prices. Rogo¡ (1996) and Taylor (1995) provide recent reviews. Any test is conditional on a particular econometric speci¢cation, which embodies a set of auxiliary assumptions. In this paper we provide a systematic exposition of the alternative speci¢cations and illustrate the sensitivity of the results to speci¢cation on a sample of 31 developing countries over the period 1966^90. As with real exchange rates themselves, views about PPP have shown persistent swings. The orthodoxy of the early 1970s in favour of PPP was largely abandoned in the 1980s as new time series tests could not reject the hypothesis that there was a unit root in the real exchange rate. More recently failure to reject the unit root hypothesis has been interpreted as a product of the low power of the tests (e.g. Edison et al., 1997) and tests using either very long spans of historical data (e.g. Lothian and Taylor, 1996), panel data (e.g. Frankel and

"Aggregation Bias" DOES Explain the PPP Puzzle

“…13 See Boyd and Smith (1999), who conclude there is very little heterogeneity in real exchange rates dynamics between countries.…”

Real exchange rate effects on the balance of trade: cointegration and the Marshall–Lerner condition

“…Coakley et al (1996), the stationarity of the real exchange rate e t in the purchasing power parity literature, e.g. Boyd and Smith (1999), and the stationarity of deviations of domestic from foreign output, y t = y* t in the convergence literature, e.g. Lee et al (1997).…”

“…The econometric issues which arise when testing for PPP are analysed by Boyd and Smith (1999). They emphasize that, although PPP does appear to hold in the long run (in the developing countries, in their case) there are major difficulties in measuring the speed of convergence in cross-country panels, unless data for very long time periods are available and the structural parameters are constant, both across countries and over time.…”

Why Panel Tests of Purchasing Power Parity Should Allow for Heterogeneous Mean Reversion

“…The literature has identified several other biases that affect both methodologies, such as the time aggregation bias (see, e.g., Choi et al 2006) and the lagged dependent variable bias (see, e.g., Boyd and Smith, 1999). We do not address these biases here.…”