By George Kapetanios & Fotis Papailias
Apart from the damage to public health and the great toll on human lives, the COVID-19 pandemic also shocked the economic structure of almost every country. The impact was greater than that of the 2007/2008 economic crisis. However, the effects on financial conditions have not been as protracted, due to the reaction of policy makers. Nevertheless, academic researchers and industry practitioners are currently debating how economic forecasting could have been done differently during the pandemic and what lessons may be learned from this exogenous economic shock.
One strand of the literature is currently looking into various models and methodologies to improve forecasting ability. However, is it only a matter of methodologies and modelling? Should we invest all our research efforts in the investigation of methodologies which should (or could) have been used during the pandemic?
In this study, we attempt to shed light on the data availability side and to highlight the need for novel, high-frequency, or big data-based datasets to enable detection of early patterns in the changing behaviour of economic variables. We do not aim to present a horserace for the competing models provided by various methodologies. Instead, we focus on a brand-new dataset of weekly indicators for the United Kingdom (UK) and evaluate whether it improves the timeliness of estimates of economic activity. Most of these indicators were quickly made available by the Office of National Statistics (ONS) soon after the outbreak of the pandemic in an effort to facilitate real-time economic research. The indicators include many weekly variables such as VAT indices, the use of debit and credit cards, transport use, business activities, online job advertisements, traffic cams data, COVID-19 surveys, online retail prices and traffic near ports.
In our paper, we demonstrate two uses of this dataset: (i) the construction of a real-time coincident indicator, and (ii) the evaluation of this index in an out-of-sample nowcasting exercise. The real-time coincident indicator is for the UK in real-time from 2020M01 to 2021M01, which covers the whole COVID-19 pandemic period. It seems to accurately track the gross value added over the course of the pandemic. We evaluated the predictive ability of the indicator in an out-of-sample nowcasting exercise for the UK, using the monthly Gross Value Added as our target variable.
As the figure shows, the “faster indicators” dataset provides an accurate real-time estimate of economic activity and improves timeliness. Our empirical results provide ample evidence that timely indicators are essential ingredients in real-time estimation and greatly help to track economic activity even in turbulent times, such as during the pandemic. Given that the faster indicators dataset we worked with had more predictive information than the traditional standard data variables, it is our recommendation that these should be considered in forecasting and nowcasting exercises from now on. Further research could focus on undertaking more detailed studies on each individual indicator.
Read the full ESCoE Discussion Paper here.
George Kapetanios is Professor of Finance and Econometrics at King’s Business School, King’s College London
Fotis Papailias is a Lecturer in Banking & Finance at King’s Business School, King’s College London
ESCoE blogs are published to further debate. Any views expressed are solely those of the author(s) and so cannot be taken to represent those of the ESCoE, its partner institutions or the Office for National Statistics.
