ABSTRACT
This paper considers the real interest rate parity (RIRP) in OECD countries applying a sequential panel selection (SPS) method on alternative panel unit-root tests. Our approach exploits the enhanced power of panels to uncover evidence of stationarity, but also identifies the exact countries for which the RIRP holds in a panel. Moreover, we construct real interest rate measures using alternative approaches, including a Markov regime-switching procedure, which is consistent with the forward-looking nature of inflation expectations formation. Considering US as the benchmark economy, we produce strong evidence of stationarity in real interest rate differentials, which resuscitates RIRP, especially given the inconclusive results in the related literature. Our results are robust to different panel unit-root tests, measures of inflation expectations, and interest rate maturities. The RIRP appears quite resilient in the face of the global financial crisis and the low real interest rate environment after the great recession. The SPS allows to calculate half-lives, which avoid the pitfalls of over/underestimating the speed of adjustment and are lower as compared to the typical estimates in the literature.
Notes
1 See Perron (Citation1989).
2 For the IPS, we use 5% as the level of significance, while for the PES and CS, we use the tables provided by Pesaran (Citation2007) and Chang and Song (Citation2009), respectively.
3 There are three sets of these probabilities; the inference, the smoothed, and the forecast ones. For more details, see Hamilton (Citation1990) and for a concise review see Franses and van Dijk (Citation2000).
4 We do not include Finland, Spain, and Mexico due to lack of data.
5 We do not report these results, but they are available upon request.
6 These are Belgium, Canada, Denmark, Finland, France, Germany, Italy, Japan, Korea, Mexico, South Africa, Spain, Switzerland, and UK.
7 In the case of short-run ex ante rates, the IPS univariate and panel tests find the same stationary series. Such a finding cannot be excluded as an outcome of the testing procedure. The most probable reason is IPS test inability to take into account any potential cross-sectional dependencies, as we mentioned in Section 3.