LlewellynConsulting | Macro series | John Llewellyn and Soyon Park 12 August 2020
Our reference G7 multi-year GDP outlook
On historical experience, G7 GDP will likely re-attain its 2019 Q4 level only in 2024
Forecasters make their biggest mistakes when an economy is hit by a shock that is both large (taken to be 2% of GDP or more) and novel.1 This time, the COVID-19 lockdown shock is unquestionably large, and indubitably novel, having originated outside the economic system, and provoking an unusual response: neither on the occasion of the 1957 influenza pandemic, which killed an estimated 1 million people worldwide, nor after the 1968 pandemic (which killed an estimated 1 to 4 million) did policymakers lock down their economies.2
Clearly, even if it is an overworked truism, forecasting GDP in present circumstances is more than usually uncertain. Nevertheless it is appropriate to have a view, not least – a prime reason for forecasting – to serve as a reference against which to assess the data as they come in, such that unexpected developments stand our clearly and are recognised early.
Current debate is focussing largely on the likely shape of the initial recovery – whether it will be V-shaped, a U, an L or whatever. But that is scarcely the most pertinent issue: more important is what stands to happen longer term. Here is what history suggests:
▪ In previous post World War II recessions it has typically taken 3 to 6 years for GDP to reattain its pre-recession level.3 As is happens, that was almost how long it took for the G7 aggregate after the 2008 Global Financial Crisis – from 2007 Q4 to 2011 Q2 (see the figure). It is hard to see why the evolution should be much more favourable this time around. On such a reckoning it will be between 2023 and 2026 – mid-point 2024 – before G7 GDP4 is back to its 2019 level.
▪ On that basis the G7 economies will by 2024 have foregone over 7½% of their combined 2019 GDP as a result of the COVID-19 recession – the pale-pink panel in the figure.
▪ Arguably however the GDP loss will be greater than that. Typically, when an economy has recovered from a recession, real GDP remains below where it would have been had it stayed on its pre-recession path5 – so-called ‘persistent’ GDP loss.6 Had there not been the COVID-19 recession, G7 GDP would, presumably, have grown at least somewhat, even if only at its somewhat lacklustre 1.2% 2007-19 rate. On this basis, the additional GDP loss stands to be about 3½% – the light-grey panel – making for a total loss of 11%.
▪ How fast the G7 economy will grow, particularly after it reaches more nearly normal levels of employment, will depend importantly on the extent to which the resources in the recovery phase are devoted to capital formation.7 Typically in a recession it is private sector investment that is hardest-hit.8
▪ Indeed, the extent to which policy can promote investment – not least in infrastructure – stands to be one of the defining determinants of the outcome longer term.
1 Following the 1973/74 oil price shock, for example, forecasters over-predicted countries’ year-ahead GDP on average by fully 4 percentage points, yet following the similar-sized, but no longer novel and hence better understood, 1978/79 oil price shock the mean error was only 0.9 of a percentage point1. For more on this, see Llewellyn, J., Potter, S., and Samuelson, L, 1985. Economic forecasting and policy – the international dimension. Chapter 6, p. 101. Routledge and Kegan Paul.
2 Source: Honingsbaum, M., 2020. The art of medicine: Revisiting the 1957 and 1968 influenza pandemics. The Lancet, May 25. Available at [Accessed 3 August 2020]
3 Typically, the time that it takes from start to finish – i.e. for GDP to reach its trough and then to re-attain its pre-crisis level – is to be measured in years, not quarters. The right-hand side of the figure below shows, for a selection of the largest depressions and recessions, that over the post-WWII period their duration is typically of the order of 3 to 6 years. In one case – Italy – GDP never got back to where it was in 2007 before it was hit again, by the COVID-19 pandemic.
Depth and duration of major recession episodes
For more, see Llewellyn, J., Jones, R., and Park, S., 2020. Recessions last years, not quarters. Llewellyn Consulting, 7 May.
4 The G7 group of countries has been chosen as the aggregate here because it represents a substantial proportion (around 60%) of global GDP, and is made up of an intrinsically similar set of (OECD) economies.
5 For more on this, see Llewellyn. J. and Park. S. 2020. The economic cost of recessions. 1 June. Llewellyn Consulting; and Llewellyn, J., and Park, S., 2020. The economic cost of recessions II. July 2020. Both available upon request.
6 The OECD put the median loss in aggregate potential output at about 3½ per cent for its 34 member countries, taken all together, over the period following the 2008 Global Financial Crisis; and at about 5½% for the subset of 19 OECD countries that directly experienced the banking crisis. See Ollivaud, P., and Turner, D., 2015. The effect of the global financial crisis on OECD potential output, especially p. 42 and Table 1. Available at [Accessed 12 July 2020]
The IMF, examining several decades of data for a wider group of 190 countries, both developed and developing, found the persistent loss in output to have been on average about 5 percent of GDP in the case of recessions caused by balance of payments (BOP) crises, 10 percent for recessions caused by banking crises, and 15 percent following ‘twin’ crises. See Cerra, V. and Saxena, S., 2017. Booms, Crises, and Recoveries: A New Paradigm of the Business Cycle and its Policy Implications. IMF Working Papers, 17(250), p.6. Available at . [Accessed 12 July 2020]
7 The consequences can be so severe that sometimes recessions result not only in a persistent loss of output relative to the extrapolated trend level but also a reduction in the (post-crisis) trend rate of growth thereafter. The
IMF found that the average growth rate after a recession was typically about a ½ percentage point lower in the first four years of recovery than the average growth rate for all expansion years. See Cerra, V, and Saxena, S., (2017), op. cit.. p.7. As the authors observe: “Output does not cycle around a long-term upward trend. Instead, shocks result in a complete shift in the trend line itself. In short, the “business cycle” is not a cycle.”
8 Moreover, this time changes in the structure of demand – reductions for example in the demand for public transport, airlines, conference centres, cruise ships, even office space – may well result in premature scrapping of some capital stock. I am grateful to Saul Eslake for reminding me of the importance of this phenomenon, which I first saw following the 1973/74 oil crisis. The quadrupling of the price of oil led to hundreds of still technically viable, but now economically unviable, piston-engined propeller-driven aircraft being mothballed in the Mojave Desert – an extraordinary sight – prior to their being broken up.
©Copyright Llewellyn Consulting LLP 2020. All rights reserved. This report is for exclusive use by the addressee only. The content of this report, either in whole or in part, may not be reproduced, or transmitted in any form or by any means, electronic, photocopying, digitalisation or otherwise without prior specific written permission from Llewellyn Consulting LLP.
The information, tools and material presented herein are provided for informational purposes only and are not to be used or considered as an offer or a solicitation to sell or an offer or solicitation to buy or subscribe for securities, investment products or other financial instruments. All express or implied warranties or representations are excluded to the fullest extent permissible by law.
Nothing in this report shall be deemed to constitute financial or other professional advice in any way, and under no circumstances shall we be liable for any direct or indirect losses, costs or expenses nor for any loss of profit that results from the content of this report or any material in it or website links or references embedded within it. This report is produced by us in the United Kingdom and we make no representation that any material contained in this report is appropriate for any other jurisdiction. These terms are governed by the laws of England and Wales and you agree that the English courts shall have exclusive jurisdiction in any dispute.