Exchange Rate Economics: Theories and Evidence

50 Purchasing power parity and the PPP puzzle
that the coefficients on relative prices (i.e. α 0 and α 1 ) to be equal and opposite and
insignificantly different from unity.
MacDonald (1988b) also used panel methods to estimate (2.17) and (2.18). The
key point in this paper was the recognition that since PPP was a long-run phenomenon
the most appropriate way to estimate it was by using low frequency
data,such as annual data. Such data was argued to be preferred to monthly or
quarterly data because it had a higher signal-to-noise ratio. Of course the problem
with using annual data to estimate PPP relationships for the recent float is
the relatively small number of annual observations available. MacDonald therefore
proposed using a panel estimator to increase the number of observations.
In particular,he used a time wise autoregressive/cross-sectionally heteroscedastic
estimator and found that for a panel of five countries over the period 1973–85 the
coefficients in both (2.17) and (2.18) were insignificantly different from unity,in
absolute terms. To our knowledge MacDonald (1988b) was the first to test PPP in
a panel context using annual data. As we shall see later,nearly all of the recent
panel estimates of PPP rely on annual data.
2.3 The recent empirical evidence on PPP and
the LOOP
Recent tests of PPP have focussed on using cointegration methods to test the
relationship between the nominal exchange rate and relative price relationship,
and unit root methods have been used to determine if real exchange rates are
mean-reverting processes and,in particular,estimate half-life adjustment speeds.
Before considering these type of tests we look at the sources of volatility in real
exchange rates in terms of the two key components q T and q NT ,T .
2.3.1 Real exchange rate volatility and systematic
movements of the real exchange rate: q T versus qNT ,T
As noted in Section 2.2,the early tests of the LOOP showed that there are significant
violations of the hypothesis. More recent tests of the LOOP have followed
on from the work of Engel (1993) who calculated the relative importance of the
two components in (2.12) by comparing the conditional variances of relative prices
within countries – V (p i −p j ) – and across countries – V (p i −s −pi ∗ ). Four dissagregated
indices of the CPI are used,namely,energy,food,services and shelter. These
indices are chosen to capture different degrees of tradability,with food taken to
be the traded good and the remaining components the non-traded elements. The
data are collected for the G7,over the period April 1973–September 1990 and
comparisons are made between the G6 and the US. The startling result to emerge
from Engel’s work is that out of a potential 2400 variance comparisons,2250 have
the variance of the relative price within the country smaller than the variance
across countries for the same type of good; that is, V (p i − p j )