Forecasting Gross Domestic Product Growth with large Unbalanced Data Sets: The Mixed Frequency Three-pass Regression Filter

Abstract

Summary Gross domestic product (GDP) is a key summary of macroeconomic conditions and it is closely monitored both by policy makers and by decision makers in the private sector. However, it is only available on a quarterly frequency, and in many countries it is released with a substantial delay. There are, however, many higher frequency and more timely economic and financial indicators that could be used for nowcasting and short‐term forecasting GDP. Against this backdrop, we propose a modification of the three‐pass regression filter to make it applicable to large mixed frequency data sets with ragged edges in a forecasting context. The resulting method, labelled MF‐3PRF, is very simple but compares well with alternative mixed frequency factor estimation procedures in terms of theoretical properties and actual GDP nowcasting and forecasting for the USA and a variety of other countries.

Forecasting GDP with many predictors in a small open economy: forecast or information pooling?

“…In this study, we consider mixed frequency three-pass regression filter recently proposed in Hepenstrick and Marcellino ( 2019 ), who extend the 3PRF method in Kelly and Pruitt ( 2015 ) to a mixed frequency context. MF-3PRF enables us to obtain targeted factors for forecasting a specific variable of interest, rather than only summarising the information in a large number of predictors.…”

“…Moreover, it has several asymptotic optimality properties and is quite straightforward to implement, being primarily based on ordinary least squares. Based on a dataset drawn from many different countries, Hepenstrick and Marcellino ( 2019 ) and Marcellino and Sivec ( 2021 ) provide evidence that MF-3PRF is a promising tool for nowcasting GDP growth.…”

“…To nowcast quarterly GDP growth using the 3PRF methodology, Hepenstrick and Marcellino ( 2019 ) consider the case in which the target variable (or the proxies ) are of lower frequency than predictors . Their MF-3PRF method can be summarised in three steps as follows.…”

“…Marcellino and Schumacher ( 2010 ) provide more details on these approaches in a MIDAS context. More recently, Hepenstrick and Marcellino ( 2019 ) proposed a mixed frequency three-pass regression filter (MF-3PRF), which is built upon the work of Kelly and Pruitt ( 2015 ). This method enables us to obtain targeted factors in an intuitive and less computationally intensive manner, thereby overcoming the drawback of principal components-based methods that ignore the comovement between indicators and the variable of interest when computing factors.…”

Markov-Switching Three-Pass Regression Filter

“…16 Third, the forecasts are constructed as 15 This result still holds when doing PCA at a monthly frequency and then aggregating the factor at a quarterly frequency. 16 As a side note, the first pass of the 3PRF filter could possibly accommodate mixed-frequency data using the techniques outlined in Foroni et al (2015); whereas, in the third pass of the filter, unrestricted mixed data sampling (MIDAS) polynomials could be used as in Hepenstrick and Marcellino (2016), and regime-switching parameters in the mixed-frequency predictive equation could be modelled as in Guérin and Marcellino (2013).…”