The European Central Bank has published a working paper finding that local linear forests outperform random walks, ridge regressions and random forests in forecasting negotiated wage growth in France. The results indicate that French wage dynamics contain nonlinear patterns and that incorporating economic indicators from Germany and Italy improves forecast accuracy beyond models using French data alone. Using quarterly data from 1985 through the first quarter of 2025, the authors evaluated forecasts one to five quarters ahead based on indicators including unemployment, productivity, inflation expectations and past wage growth. Local linear forests performed best overall when using the full French, German and Italian data set, particularly at longer horizons. Their ability to track smooth signals also produced better results than random forests when wage growth moved to historically unusual levels, including the low growth period after the global financial crisis.
European Central Bank working paper finds local linear forests improve French wage forecasts
A European Central Bank working paper finds that local linear forests improve forecasts of negotiated wage growth in France, particularly at longer horizons. Forecast accuracy rises further when German and Italian economic indicators are added, pointing to nonlinear wage dynamics and substantial cross-country information effects.