The views expressed in the papers are the views of the author(s) and do not necessarily reflect the views of the Centre or its sponsors. [...] The study neither included specification tests for the consis- tency of the pooled OLS and GLS estimators (e.g., whether the constant coefficients α and β are appropriate for the purpose of using pooled OLS estimator) nor formally reported forecast performance of the models. [...] One way is to directly exponentiate the logged estimated mean of expenditure y for the unlogged mean of forecast ey; and the logged estimated standard error of the fore- √ cast σ for the unlogged standard error of the forecast eσ2. [...] The alternative way is to exponentiate both the logged estimated mean and the logged variance for the √ 2 unlogged mean of forecast ey+(σ )/2, and for the unlogged standard error of forecast e2y+σ2(eσ2 − 1) (Lin, 2012). [...] So the N and T differ for the panels of THE and PHE, and the expenditure amounts cannot be compared between the two panels.