PurposeThe primary purpose of this study is to examine whether the classification of industries into the tradable and nontradable matters for the Balassa–Samuelson (BS) effect.Design/methodology/approachThe study uses annual data for 38 countries from 1995 to 2014. To examine whether the classification of industries matter, the study proceeds with two approaches, that is, “traditional” and “benchmark”.FindingsFirst, by applying panel cointegration tests of Pedroni and Westerlund, the results validate the BS hypothesis. However, the coefficients of long-run elasticities show appreciation of real exchange rate (RER) due to increase in productivity in the case of “traditional approach”, whereas depreciation of RER in the case of “benchmark approach”. Second, by applying the Dumitrescu-Hurlin panel Granger causality test, the results reveal the bi-directional causality among RER and productivity for both the approaches. Further, to provide more insights, the study employs a fixed-effects panel threshold model. The results indicate that increase in productivity leads to both appreciation and depreciation of RER depending on threshold regimes.Practical implicationsThe study ascertains that the evidence of BS effect depends on the choice of approach considered. However, irrespective of the classification, there exists a BS effect beyond a threshold.Originality/valueAlthough the BS effect is well established in the literature; there is no study examining the importance of classification of industries at a disaggregated level. Furthermore, there is no consideration of threshold effects.
A Multidude of Econometric Tests: Forecasting the Dutsch Guilder