Improving forecasting accuracy of the Phillips curve in OECD countries: The role of commodity prices

We investigate the power of commodity prices to improve inflation forecast performance in 21 Organization for Economic Cooperation and Development (OECD) countries, within the framework of commodity prices‐augmented Phillips curve model. Using monthly data spanning over 57 years, we use single and multi‐factor predictor models, and Westerlund and Narayan (2012, 2015) estimator to address inherent issues of heteroscedasticity, persistence and endogeneity, which present empirical challenges to erstwhile attempts to forecast inflation. Contrary to Stock and Watson (1999) findings, our results overwhelmingly show that commodity prices significantly improve the power of Phillips curve‐based inflation forecasts in OECD countries. These findings hold for both core and headline measures of inflation, and within extensive in‐sample and various out‐of‐sample forecast horizons. In addition, we find differential degrees of commodity prices‐inflation pass‐through, with agricultural and energy commodity prices exerting the highest pass‐through to inflation. We further evaluate and compare the forecast performance of our augmented inflation forecast models with that of conventional random walk model as a benchmark. Our results overwhelmingly confirm that commodity prices‐augmented versions of inflation forecast models outperformed popular random walk models in all forecast horizons.