Description:
We contribute to the burgeoning literature on macroeconomic forecasting by exploring the benefits of machine learning methods. The performance of machine learning algorithms, factor models and MIDAS in forecasting China’s real GDP growth rate is evaluated in an extensive pseudo out-of-sample simulation. The findings can be summarized in three points. First, ridge, LASSO and elastic-net, random forest and factor-augmented autoregressive perform relatively well compared to other models. The ML method that deserves more attention is ridge regression, which dominates all other models in terms of minimum rRMSE. Second, we find that ML methods, especially ridge regression, are effective to recognise the impacts of Global Financial Crisis and Covid-19 shocks. Third, we find predictors from monetary policy, industrial and consumption groups have the most influential powers. These findings may shed light on future research aimed at uncovering the dynamic of real variables on aggregate economic activity