Deviations from PPP and UIP in a financially open economy: the Turkish evidence

Abstract

This paper investigates the interrelations between purchasing power parity (PPP) and uncovered interest parity (UIP) in Turkey using Johansen cointegration analysis for a system containing Turkish and US inflation rates, interest rates, and exchange rate. The results of a structural model obtained by data-acceptable over-identifying restrictions over the cointegration space suggest the existence of two cointegration vectors representing UIP and PPP with proportionality and symmetry conditions, respectively. Consistent with the capital enhanced equilibrium exchange rates (CHEERs) approach, each of the international parity hypotheses is rejected when formulated independently. This is a theiry-consistent result for a financially open economy for which equilibrium conditions of asset and commodity markets may not be independent of each other.

Notes

See Obstfeld and Rogoff (2000), Taylor (2001) and Sarno and Taylor (2001) for recent reviews.

Defining equilibrium exchange rates appropriately may be crucially important also for the design of an exchange rate based stabilization programme. If the evolution of exchange rates is not independent of interest rate differentials, then a PPP-based exchange rate targeting policy may not be sustainable. This is basically because, the adjustment of exchange rates to capital flows due to interest differentials may lead to targeted exchange rates substantially diverging from the equilibrium rates for a financially open economy.

The recent related studies for the Turkish data include Metin (1994), Telatar and Kazdaglı (1998), Sarno (2000) and Erlat (2001). Metin (1994) tests UIP and PPP jointly and finds that neither of the parities is supported by the Turkish annual data pertaining fixed and flexible exchange rate periods. Telatar and Kazdaglı (1998) provides evidence against PPP on the basis of conventional unit root tests. Sarno (2000) employs a non-linear modelling approach and provides a strong support for the validity of PPP. The results by Erlat (2001) suggest that the PPP hypothesis cannot be rejected when fractional integration and endogenous break point estimation methods are employed.

See Sarno and Taylor (2001) for a recent comprehensive survey.

McCallum (1994) and Flood and Rose (2001) provide the recent accounts. Note that, under a maintained hypothesis that the Fisher parity holds for each of the countries, Equation 8 gives a relationship between real exchange rate and real interest rate differential. See Edison and Melick (1999) for a recent application.

The data are from the IMF's International Financial Statistics CD-Rom (December 2000) and the Central Bank of the Republic of Turkey database. The data can be obtained by request from the authors at their internet addresses.

Note that the possible stationarity of Δe _t does not necessarily preclude it from being contained in a cointegration space. This is because cointegration tests may still be used even if some series are stationary as Dickey and Rossana (1994) note.

The sequential likelihood ratio (LR) test of system reduction from VAR(5) to VAR(4) yielded χ²(25) = 53.96 (p = 0.001). Thus the reduction appears not to be data-acceptable. This choice is supported also by the Akaike Information Criterion (AIC).

Note that these t-statistics has a meaning only if the corresponding cointegration vector defines a plausible economic relationship (Hansen and Juselius, 1995).