The emergence of a vaccine may mean that there is light at the end of the pandemic tunnel, but as Rishi Sunak pointed out in his spending review, the economic fallout has only just begun.
The Chancellor revealed the coronavirus pandemic would raise Government borrowing in 2020 to a peacetime record of £394 billion, while 2.6 million people or 7.5 per cent, and could be out of work by the second quarter of 2021.
In these unprecedented times, figuring out when the UK economy will be back to its pre-crisis levels is grotesquely difficult. The Office for Budget Responsibility, whose data and forecasts the Chancellor relies on, made the unprecedented step of producing three different scenarios of the country’s economy; from optimistic that predicts the UK’s GDP will return to its pre-pandemic level in the first quarter of 2021; to a central forecast of midway through 2022; and a pessimistic take of late 2024.
Macroeconomic forecasting typically relies on stable relationships between economic indicators. But thanks to the pandemic and subsequent lockdowns recently published macroeconomic data has been quite unusual.
UK GDP quarterly growth rates usually average at around 0.5 per cent. The Office for National Statistics has estimated a fall of 19.8 per cent in the second quarter of 2020 and an increase of 15.5 per cent in the third quarter. As indicated in the graph below, these numbers imply that in the third quarter of 2020 the UK economy was 10 per cent smaller than it was a year earlier, but that still represents a recovery if compared with the 21.5 per cent decline in the second quarter.
In the graph below, I consider a range of possible alternatives for the UK economy’s likely path, including a very fast recovery in blue and an even deeper recession in red. These paths suggest that forecasting uncertainty is also unusually large, as the differences across possible paths are beyond the usual +1.5 or -1.5 per cent range.
Popular economic forecasting models are unable to pick up the massive swings in growth observed in the last two quarters, as they are extraordinary when compared to past data. The best practice in macroeconomic forecasting is to compute forecasts for the future, say for 2021, using the most recent data.
But are these recent swings useful to forecast the path of the economy for 2021? Are we likely to observe again these sizeable rates or are these just once-in-a-century events? Would 2021 be more like 2019, and consequently, should we discard 2020 data when computing growth forecasts?
Some solutions to this problem were proposed by researchers in the area of macroeconomic forecasting in a recent online workshop I organised with the National Institute of Economic and Social Research. Ivan Petrella, Associate Professor at WBS, proposed a method to downweigh some of the large movements in macroeconomic data by using fat-tailed shock distributions.
A fat tail is a statistical distribution that indicates a high probability of rare and extreme outcomes. A normal distribution sees 99 per cent of the outcomes generally fall within three deviations of the mean. Under a fat tail distribution, on the other hand, the percentage of outcomes that fall more than three standard deviations from the mean is much higher.
Professor Petrella believes a forecasting model should recognise that outliers are likely to occur, even if with low probability.
Massimiliano Marcelino, of Bocconi University, instead proposed to remove the impact of the recent data when computing the most likely future path for the economy. He argued that the most recent data would have a partial effect on raising the measurement of the uncertainty around the most likely path.
Even if we can find statistical methods capable of fitting the recent growth swings, these methods may not be adequate to forecast economic growth in 2021. The main reason is that they may lead to an underestimation of the uncertainty around 2021 forecasts. Especially as one of the few things, that researchers and policymakers agree on is that one cannot be sure on the timing of the full economic recovery.
The timing of the recovery requires not only estimates of the effect of past policies on the economy, but also the likelihood of future constraints on economic activity as the Government attempts to mitigate the health effects of the COVID-19 pandemic.
One may argue that as popular statistical models are struggling with the recent economic data, a better choice is to rely instead on expert evaluation when computing macroeconomic forecasting.
Indeed, my current research with WBS Professors James Mitchell and Anthony Garratt suggests that expert judgement may improve the quality of GDP growth forecasts during a time of heightened uncertainty. A disadvantage is that subjective measures of uncertainty attributed to these forecasts are not as accurate as measures obtained with statistical models.
My suggestion then is to rely on the recent advances in statistical modelling using fat-tailed distributions or Professor Marcelino’s idea. These new methods are more likely to deliver accurate estimates of the uncertainty of macroeconomic forecasts as they take the view that recent extraordinary growth fluctuations indicate similar events are likely in the future, even with a small probability.
This article was originally published on the Warwick Business School (WBS) website, and written by Ana Galvao, who teaches on the Executive MBA at WBS London, located at The Shard, and the online Global Central Banking & Financial Regulation qualification.