Does purchasing power parity hold following the launch of the euro? Evidence from the panel unit root test

Abstract

Using a richer panel data set with more representative price indexes, this article adopts Pesaran's (2007) panel unit root test to study the validity of Purchasing Power Parity (PPP) across the eurozone countries both before and after the launch of the euro in 1999. We find that PPP holds before the introduction of the single currency, whereas it fails to hold after 1999. Our result is consistent with the finding in Engel and Rogers (2004) and Rogers (2007) that there is no longer a tendency to support the law of one price following the birth of the euro.

Notes

¹See Haan et al. (2005, p. 1).

²See Taylor and Taylor (2004) for a survey about the source of nominal-exchange-rate fluctuations in the short run. They also propose that such fluctuations lead to violations of the LOP.

³In the literature, there are several studies on the European prices related to the euro. For example, Drine and Rault (2006) show that PPP holds in the east-European countries during the period 1995 to 2000, which covers the launch of the euro in 1999. Arghyrou (2007) finds that the Greek consumer prices do not converge to the eurozone price level following this country's adoption of the euro in 2001. Koukouritakis (2009) studies the validity of PPP among the 12 new countries that joined the European Union in 2004; however, there are no consensus agreements regarding whether PPP holds in this article. Nevertheless, those studies have not yet considered the change in persistency across the eurozone countries because of the introduction of the euro in 1999.

⁴These two approaches measure the magnitude of the greater or lesser amount that a consumer in a particular city has to pay compared with the average price across cities at a specific time.

⁵The panel unit root tests are widely adopted to investigate the mean-reverting property of real exchange rates. See, for example, Coakley and Fuertes (1997), Coakley et al. (2005) and Wickremasinghe (2009).

⁶The previous studies collect data on the actual prices of 139 narrowly defined items from the eurozone cities. The number of covered items, as a result, is only a small proportion relative to the total number of items within a country.

⁷More specifically, the log of the real exchange rate between countries h and f at time t is constructed according to , where , , , and denotes the log of the nominal exchange rate converting two countries' prices into US dollars. M − which denotes the number of countries available in our data set – equals 11 for the CPIs, while it equals 10 for the PPIs. i is the possible combination of h and f. Therefore, N^k equals 55 for the real exchange rates based on the CPIs, and 45 for these in terms of the PPI data.

⁸The conventional tests are also known as the first-generation panel unit root tests, while those tests accounting for CD are so-called second-generation panel unit root tests. See Pesaran (2007).

⁹Following Elliot and Pesavento (2006), we do not include a deterministic time trend here.

¹⁰Monte Carlo simulations in Pesaran (2007) reveal that this test has satisfactory size and power even for relatively small values of N^k and (e.g. even in the case of ).

¹¹The following conventional unit root tests can be implemented by a popular econometric software, such as Eviews.

¹² We obtain the same result of rejecting the unit root hypothesis when p is chosen to be 0 (see Table 3).