Forecasting models for the Chinese macroeconomy in a data‐rich environment: Evidence from large dimensional approximate factor models with mixed‐frequency data

Our paper contributes to the nascent literature on forecasting the Chinese macroeconomy in a data‐rich environment. We perform a horse race among a large set of traditional models and two classes of factor models with 251 monthly and 34 quarterly macroeconomic variables over the period from January 2002 to June 2018. We find evidence that mixed‐frequency factor models provide superior forecasts of the CPI, RPI, investment, and consumption when compared to the simple benchmark. During the Global Financial Crisis period, we find little evidence of superiority of factor models over the simple benchmark AR(p) model.