Su et al provide new evidence on PPP

Su et al. (2014) provide new evidence on PPP hypothesis employed real effective exchange rate based on a new methodology that use the widely accepted KSS (Kapetanios et al., 2003) non-linear unit root tests which the authors wild bootstrapped. They applied this method to test PPP across 61 countries (for Latin America employed seven countries: Argentina, Brazil, Chile, Colombia, Peru, Mexico and Venezuela) over the period 1994–2012. After accounting for heteroskedasticity as well as non-linearity through the use of the wild-bootstrapped KSS test, the results provide strong evidence that most of the REERs are non-stationary. In particular, for Latin America the empirical results indicate that PPP is not hold for all countries. The results of Bahmani-Oskooee et al. (2008), Cheng et al. (2008), Divino et al. (2009), Su et al. (2011), Wu et al. (2011) and He et al. (2014) are not consistent with Drine and Rault (2008) and Su et al. (2014). Drine and Rault (2008) aimed to investigate whether the PPP concept could serve as a benchmark to determine the real exchange rate diacylglycerol kinase in a large sample of developed and developing countries (for Latin America employed 18 countries: Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, the Dominican Republic, Ecuador, El Salvador, Guatemala, Honduras, Mexico, Nicaragua, Panama, Peru, Paraguay, Uruguay, Venezuela). Considering panel unit root tests of strong version of PPP indicate that the null hypothesis of a unit root in the real exchange rate series cannot be rejected for Latin American, African, Asian, MENA and CEE countries). For the test of weak PPP, panel data cointegration tests indicate that weak PPP is rejected by data for Africa, Latin America, Asia and CEE countries (since the null hypothesis of no cointegration is not rejected for these countries).
Methodology
Empirical results Initially, we conducted the linearity test of Harvey et al. (2008), the power of which is not affected by the order of integration of the series, as described in the methodology. The goal of this first step is to identify which series of the real effective exchange rate is linear or non-linear. In the second step, this information is utilized to perform the linear or non-linear unit root test. According to Table 2, we reject the null hypothesis of linearity at the 5% level of significance for the real effective exchange rate of Mexico and Venezuela. The real effective exchange rate of these two countries is non-linear. Since the linearity of the series was identified, we analyzed the validity of the hypothesis of purchasing power parity through the stationarity test of the real effective exchange rate. In this part, for the linear series, we decided to use the Elliot et al. (1996) generalized least squares detrending tests (DF-GLS) and Ng–Perron (2001) linear unit root tests. The choice of these tests was made because of the problem of the size and power of the other traditional tests. Table 3 shows the results of these unit root tests. As shown in Table 3, we reject the null hypothesis of a unit root at the 5% level of significance only for the real effective exchange rate in Chile and Peru. In other words, the real effective exchange rate (henceforth called the REER) of these countries in the period analyzed is stationary and the PPP hypothesis is valid for these two countries. It is clear, too, that the linear unit root tests indicate that the REER series for Mexico and Venezuela would not be stationary. These results were expected since the series are considered non-linear by the Harvey et al. (2008) test. For the three countries with linear series that were not stationary in the DF-GLS and Ng–Perron tests (Argentina, Brazil and Colombia), there is the possibility of the presence of structural breaks in the REER. These structural breaks could interfere with the DF-GLS and Ng–Perron unit root tests. Thus, unit root tests with an endogenous structural break (Lee and Strazicich, 2003, 2004) were executed. The results are shown in Table 4.