28.02.2015 Views

Exchange Rate Economics: Theories and Evidence

Exchange Rate Economics: Theories and Evidence

Exchange Rate Economics: Theories and Evidence

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Purchasing power parity <strong>and</strong> the PPP puzzle 53<br />

of regression equation (2.17) α 0 <strong>and</strong> α 1 need not equal plus <strong>and</strong> minus unity,<br />

respectively.<br />

A further explanation for why α 0 <strong>and</strong> α 1 need not equal unity,in absolute<br />

terms,relates to the construction of the price series (see Patel 1990). Assume a<br />

representative agent, i,for whom PPP prevails. Therefore:<br />

s t = ζ p i t − ζ ∗ p ∗i<br />

t ,(2.25)<br />

where pt i <strong>and</strong> pt<br />

∗i are the ‘true’ prices facing the representative agent in the home<br />

<strong>and</strong> foreign country <strong>and</strong>,by definition,ζ = ζ ∗ = 1. In terms of measured price<br />

indices, p t <strong>and</strong> pt ∗ ,the weights used to construct them will differ from country to<br />

country:<br />

p t =<br />

p ∗ t =<br />

N∑<br />

ω jt p jt ,(2.26)<br />

j=1<br />

N∑<br />

ωjt ∗ p∗ jt ,(2.27)<br />

j=1<br />

where ω j denotes the weight given to commodity j in the construction of the price<br />

index <strong>and</strong> this will differ from country to country; that is, ω j ̸= ωj<br />

∗ (the potential<br />

time-variation of ω implies that the α terms are also potentially time-varying).<br />

Hence if the weights are not equal across countries,the only inference that may<br />

be tested on the α’s in (2.17) is that they are positive (because the weights are,by<br />

definition,positive).<br />

The basic message from cointegration-based tests of (2.17) is that the estimator<br />

used matters. For example,the application of the two-step Engle–Granger method,<br />

in which symmetry is generally imposed,produces little or no evidence of cointegration<br />

– see,for example,Baillie <strong>and</strong> Selover (1987),Enders (1988),Mark (1990)<br />

<strong>and</strong> Patel (1990) for evidence of a variety of bilateral currencies from the recent<br />

floating period. MacDonald (1995a) applies this test to a consistent data base for<br />

nine bilateral US dollar exchange rates for the recent floating experience <strong>and</strong> for<br />

illustrative purposes his results are presented in Table 2.1. This table contains a set<br />

of point estimates of the coefficients in (2.17),along with augmented Dickey–Fuller<br />

statistics. The sample period used to construct these numbers runs from March<br />

1973 to December 1992 <strong>and</strong> both wholesale <strong>and</strong> consumer prices (WPI <strong>and</strong> CPI)<br />

are used as the price measures.<br />

Using the WPI as the price measure we note that all of the α 0 <strong>and</strong> α 1 coefficients<br />

are correctly signed (positive <strong>and</strong> negative,respectively) while with the CPI,7 of the<br />

9 currencies produce correctly signed values of these coefficients <strong>and</strong> this seems to<br />

suggest that the WPI measure is the more appropriate. However,it is clear that most<br />

of the estimated coefficients are far from their hypothesised values of +1 <strong>and</strong> −1<br />

(although as we have seen this is not particularly damaging to the PPP hypothesis in<br />

the presence of transaction costs). Of more concern,however,is the fact that none

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!