New panel unit root and cointegration tests of purchasing power parity

Cerrato, Mario (2003) New panel unit root and cointegration tests of purchasing power parity. Doctoral thesis, London Metropolitan University.


Purchasing Power parity (PPP) is one of the most investigated topics in international finance. The empirical analysis on PPP in the 1980s and 1990s relied on univariate tests such as Dickey Fuller and Augmented Dickey Fuller and cointegration tests. The empirical evidence from these methodologies seemed to produce very little empirical evidence favouring PPP. However, these methodologies have been shown to have low power and be inadequate when used with highly persistent stochastic processes (for unit root tests of the real exchange rate).

One solution followed in the literature to overcome the low power problem was pooling data on two dimensions (i.e. time and cross section dimensions) instead of only one dimension (i.e. time series dimension). Panel unit root tests of PPP have re-affirmed the existence of this parity condition in some studies. However, the empirical evidence favouring PPP in most of the studies using panel unit root tests might be overvalued due to cross section dependence (O'Connell, 1998).

One of the objectives of this thesis is to consider cross section dependence by extending the bootstrap panel unit root proposed by Maddala and Wu (1999) and apply the latter to a panel of twenty OECD real exchange rates. We also use Monte Carlo simulations to examine the size properties of the proposed bootstrap panel test.

If on one hand the literature on testing for PPP by using panel unit root tests is wide, on the other hand very little has been done on testing for cointegration between nominal exchange rate and domestic-foreign prices in a panel context. We shall also address this issue by using some new, heterogeneous, and more powerful panel cointegration tests. Furthermore, we also test for the joint symmetry and proportionality restriction by using likelihood ratio tests extended to a panel context.

As we have pointed out above there exists a large literature on PPP, but most of the analysis conducted has been undertaken on OECD coun¢es. Studies of PPP using data for developing countries are limited. In addition, very few of the studies have used black market exchange rates. One of the main objectives of this thesis is to investigate the validity of PPP in developing countries using black market exchange rates. We construct and use a unique data set consisting of twenty emerging market economies and black market nominal exchange rates, spanning over 1973M1-1993M12. As far as we know such a big data set has never been used in the studies of PPP using black market nominal exchange rates. Furthermore, we use new developed panel unit root and cointegration tests.

Another unresolved puzzle in the PPP literature is the low degree of mean reversion of the real exchange rate towards PPP. In fact, if deviation from PPP were due to monetary factors, one would expect a much faster degree of mean reversion of the real exchange rate towards PPP than what reported in the literature (i.e. 3-5 years). Rogoff called this "the purchasing power parity puzzle". We investigate the PPP puzzle in emerging markets using black market real exchange rates and econometric techniques, such as median unbiased estimation and impulse response function, that have been shown to be more appropriate to measure persistence of the real exchange rate (Murray and Papell, 2002). We also employ non-parametric bootstrap to construct confidence intervals for half-lives.

426599.pdf - Published Version

Download (7MB) | Preview


Downloads per month over past year

Downloads each year

View Item View Item