Nowcasting GDP using machine learning methods
Published: 03 November 2022
This paper compares the ability of several econometric and machine learning methods to nowcast GDP in (pseudo) real-time. The analysis takes the example of Dutch GDP over the years 1992-2018 using a broad data set of monthly indicators. It discusses the forecast accuracy but also analyzes the use of information from the large data set of regressors. We find that the random forest forecast provides the most accurate nowcasts while using the different variables in a relative stable and equal manner.
Keywords: factor models; forecasting competition; machine learning methods; nowcasting
JEL codes C32; C53; E37
Working paper no. 754
Nowcasting GDP using machine learning methods
Research highlights
- A combination of a large set of monthly macro-economic indicators and machine-learning methods can help policy makers to get a more accurate forecasts of Dutch GDP growth in the nearby quarters.
- Until the financial crisis, the near-term GDP forecasts of the dynamic factor model were as accurate as those of the machine learning models.
- Since the financial crisis, however, machine learning methods, in particular the random forest, have been more accurate and this is true across various forecast horizons, possibly caused by its ability to catch potential nonlinearities in the data.
- These conclusions are based on a comparison of the forecasting accuracy of two popular econometric techniques used in central banks (dynamic factor models and MIDAS models) to regularization techniques, random subspace methods and the random forest for near-term forecast of Dutch GDP
Discover related articles
DNB uses cookies
We use cookies to optimise the user-friendliness of our website.
Read more about the cookies we use and the data they collect in our cookie notice.