Does inflation targeting matter for PPP? An empirical investigation

Abstract

This article examines whether Inflation Targeting (IT) matters for long-run Purchasing Power Parity (PPP). For this purpose, we formally assess the evidence on PPP for a panel of 19 countries using two price indices and two panel unit root tests with cross-sectional dependence. The empirical results show that IT plays an important role in providing favourable evidence for long-run PPP.

Keywords:

• purchasing power parity
• real exchange rate
• inflation targeting
• cross-sectional dependence
• panel unit root

JEL Classification:

• F30
• F31

Notes

¹ See Lothian (1998), Papell (2002), Taylor (2003) and Taylor and Taylor (2004) for details.

² See Taylor and Sarno (1998) for details.

³ See Svensson (2000) and Mishkin and Schmidt-Hebbel (2007) for details.

⁴ See Moon and Perron (2007) for details.

⁵ Taylor and Sarno (1998) proposed a test based on the Johansen's (1991) method, but it is suitable for small panels. Instead, we regroup countries according to their monetary policy rule (i.e. whether they follow IT). See Taylor and Sarno (1998) for details.

⁶ For the European and Monetary Union (EMU) countries, we have data from 1974:1 to 1998:2.

⁷ They are Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, the Netherlands, New Zealand, Norway, Spain, Sweden, Switzerland, the United Kingdom and the United States. For PPI, we have 15 countries namely Australia, Austria, Canada, Finland, Germany, Ireland, Italy, Japan, the Netherlands, New Zealand, Norway, Spain, Switzerland, the United Kingdom and the United States. This is determined by the availability of data.

⁸ For details, see Bernanke and Mishkin (1997).

⁹ BIC is used for p in Equation 4(4).

¹⁰ The Data-generating Process (DGP) underlying the bootstrap is given by for i = 1, … , N, where the contemporaneous error covariance matrix is , This asserts the null hypothesis that q_it is a driftless unit root process and is fitted by iterated Seemingly Unrelated Regression (SUR). With the SUR estimates in hand, the nonparametric bootstrap distribution applied to both tests is built as follows: First, we generate innovation by random resampling with replacement of the fitted residuals. Second, the initial values are obtained by block resampling, as described by Berkowitz and Kilian (2000). Third, after dropping the first 100 pseudo-observations to avoid start-up effects, we run the ADF and CADF regressions ( Equation 4(4)) on the pseudo-data and obtain the values. We do this 10 000 times, and the collection of 10 000 values forms the nonparametric bootstrap distributions of the test statistics from which p-values are computed for the IPS and CIPS tests, respectively.

¹¹ They argue that PPI produce more support for PPP since this index contains more tradable goods.