Economic expected losses and downside risks due to the Spanish flu
Quantifying the expected output loss due to the spread of Covid-19 and assessing the macroeconomic risks associated with the pandemic are both challenges for the economic profession. In a recent paper (De Santis and Van der Veken 2020), we try to learn from the Great Influenza Pandemic (known also as the Spanish flu), which began in 1918 and persisted until 1920, causing the death of about 40-100 million people across the world (Hatchett et al. 2007, Barro et al. 2020). Pandemic outbreaks are rare events generating non-normal economic effects across countries. Therefore, we employ a non-linear model, covering the vast majority of countries in the world. The model allows us to construct the conditional distribution of real per-capita GDP growth as a function of the Spanish flu mortality rate, which is then used to quantify the impact of the Spanish flu.
Impact of the Spanish flu
Figure 1 shows the unconditional distribution of real per capita GDP growth using 42 countries on an annual basis. Real per-capita GDP growth is on the horizontal axis and the relative frequency is on the vertical axis. The distributions depicted by the bold and dashed lines use growth rates over the sample period 1901-1929 and the Spanish flu (1918-1920) period, respectively. The distribution over the Spanish flu period is characterised by a relatively higher number of negative growth rates and by a curve, which is more skewed, highlighting the advantage of the non-linear approach.
Figure 1 Real per capita GDP growth (%)
Source: Figure 1 in De Santis and Van der Veken (2020).
By exploiting the death rates across countries due to the Spanish flu and controlling for the death rates due to WWI, Barro et al. (2020) show that the loss in terms of real per-capita GDP for a typical country due to the Spanish flu amounts to 6.0% cumulated over the 1918-1920 period. Using the non-linear model and the same dataset, we corroborate the cumulated drop in economic activity of 7% for the typical country due to the pandemic disease (see expected value in Panel A of Table 1). About 70% of the estimated economic damage occurs in the first year.
Table 1 Estimates of the impact of the Spanish flu on real per-capita GDP (%)
Source: Tables 1-2 in De Santis and Van der Veken (2020).
We also study the downside risks, which ought to be interpreted as output losses in addition to the expected losses. Two measures are computed, associated to a realisation of (1) a ‘bad’ scenario (a draw from the conditional distribution of the GDP growth rate that results in a negative GDP growth rate), and (2) a ‘worse’ scenario (a draw from the 10% worst possible outcomes of the conditional distribution). In the bad scenario, we expect economic activity to drop by 5.5% in 1918 relative to the median, while in the worse scenario, we expect economic activity to drop by an astonishing 19.6% relative to the median (Table 1, Panel A). This is visible in Figure 2, which shows that the left tail of the conditional distribution (i.e. more negative values) of real per-capita GDP growth increases sharply due to the Spanish flu.
Figure 2 Estimates of the impact of the Spanish flu on real per capita GDP growth
Source: Figure 3 in De Santis and Van der Veken (2020).
Impact of the Spanish flu on higher-income and lower-income countries
The pandemic flu can increase income inequality across countries, because it has a disproportionate burden on low-skilled workers and because infectious diseases and their associated death toll can spread more rapidly in lower-income countries, given the high healthcare costs and public spending needed to contain the virus. Economists are concerned that pandemic diseases raise inequality. We corroborate this concern. We estimate that the expected real income loss due to the pandemic is twice as large for lower-income countries (9.8%) compared to higher-income countries (4.7%) cumulated between 1918 and 1920 (see expected value in Table 1, Panel B). Furthermore, macroeconomic risks associated with the pandemic, while large for higher-income countries, are immense for lower-income countries (see scenarios in Panel B of Table 1).
As for the US, our results (see Panel C of Table 1) are in line with the observation in Burns and Mitchell (1946) and Velde (2020), who argued that the US recession in 1918-1919 was brief and of moderate amplitude. However, the estimates of the macroeconomic risks due to the Spanish flu for the US are not negligible in the case of the worse scenario. Another useful case is Spain, which was one of only a few major European countries to remain neutral during WWI. The media in this country could report freely and in depth on the developments of the pandemic. This explains why the pandemic became commonly known as the ‘Spanish flu’. Our results show that the output loss in Spain was rather large, together with the associated downside risks in economic activity.
Death rates used as regressors
These results rely on the hypothesis that the death rates during the Spanish flu do not depend on government policies affecting economic activity. To the extent that policies aimed at reducing the spread of the influenza virus caused a drop in output, the impact of the Spanish flu on economic activity would be underestimated. In this case, the estimates would not only reflect the negative impact of death rates on economic activity, but they would also capture the positive relation between both variables induced by the policy measures (e.g. lockdown measures).
We believe that this problem during the Spanish flu is negligible for three main reasons related to: (1) the structure of the global economy, which was more dependent on agriculture and manufacturing, and the lack of generalised lockdown measures; (2) communication, including the wartime censorship; and (3) the influence of WWI on economic activity.
By searching for detailed information on the US Library of Congress tool ‘Chronicling America’ as well as other public means, we concluded that governments across the globe did not impose the closure of non-essential businesses during the Spanish flu period, but mostly adopted social distancing measures, which affected business activities such as theatres, picture houses, and music halls that accounted for a marginal fraction of real GDP globally. At that time, most of the communication was channelled through newspapers. In several countries, in order to maintain morale, wartime censors refused newspapers to report on the disease and its severity. At the time, WWI was at its end. It was unconceivable to shut down non-essential businesses. Governments could not afford it. There may have been considerable social as well as economic pressures to stay on the job (IMF 2006). In the US, the private sector even asked employees to work on Sundays to fulfil the war-related demand despite the scarcity of labour, as workers became ill.
Lessons for the COVID-19 pandemic
Studies assessing the economic implications of the COVID-19 pandemic are growing fast. We tried to gain insights from the Spanish flu, an episode for which we estimate a large output loss together with high downside risks. These potential extreme economic losses, which a country could face, can explain the request for swift policy actions by lead economists and policymakers around the world and the prompt for large fiscal and monetary policy interventions globally, after the spread of COVID-19 in 2020. Although the mortality rate due to the COVID-19 pandemic will be much lower, due to the health infrastructure, the stringent lockdown measures, and the general knowledge about the disease, the insights from the Spanish flu should not be dismissed. The risk of a sharp fall in economic activity close to two digits is real.
If one could isolate the impact of government macroeconomic policies introduced to counteract the Covid-19 pandemic on the economy, one might even expect a larger output loss, given the adverse impact of the pandemic on international supply chains, due to the restrictions on international travel, and the sizeable role of services in the global economy, which are more affected by social distancing measures.
Finally, our results support the view of international economic organisations, such as the World Bank and the United Nation Development Programme (UNDP), which argue that the COVID-19 pandemic will leave deep scars in lower-income countries, unless urgent policy action is taken with support from the international community.
Barro, R J, J F Ursúa and J Weng (2020), “The coronavirus and the Great Influenza Pandemic. Lessons from the “Spanish Flu” for the Coronavirus’s potential effects on mortality and economic activity”, NBER Working Paper, No. 26866.
Hatchett, R J, C E Mecher and M Lipsitch (2007), “Public health interventions and epidemic intensity during the 1918 influenza pandemic”, Proceedings of the National Academy of Sciences of the United States of America 104 (18): 7582-7587.
Lilley, A, M Lilley and G Rinaldi (2020), “Public health interventions and economic growth: Revisiting the Spanish Flu evidence”, Harvard University Working Paper.
Ludvigson, S C, S Ma and S Ng (2020), “COVID-19 and the Macroeconomic Effects of Costly Disasters”, NBER Working Paper No. 26987.
McKibbin, W J and R Fernando (2020), “The Global Macroeconomic Impacts of COVID-19: Seven Scenarios”, CAMA Working Paper 19/2020.
Velde, F (2020), “What happened to the US economy during the 1918 Influenza Pandemic? A view through high-frequency data”, FED Bank of Chicago Working Paper Series, No. 2020-11.
World Bank (2020), Global Economic Prospects, Washington DC.
2 The extent to which the expected value of that distribution drops in response to the presence of Spanish flu deaths indicates how strongly one can expect GDP to fall due to the Spanish flu in each particular country and year. Similarly, the extent to which the left tail of this conditional distribution becomes more important as flu deaths arise, allows us to quantify how strongly macroeconomic risk responds to the Spanish flu. For the quantification of macroeconomic risk based on the left tail, we either use the part of the distribution that covers negative growth rates (i.e. ‘bad scenario’, cf. infra) or the part of the distribution that covers the 10% most adverse outcomes (i.e. ‘worse scenario’, cf. infra).
3 The macroeconomic risk is computed by subtracting the median to control for a potential shift of the central tendency. To estimate the impact of the Spanish flu, we subtract from the macroeconomic risk a similar measure conditioning on the observed aggregate war death rates but setting at zero the flu death rates. For technical details see De Santis and Van der Veken (2020).
7 Only one out of 17 US cities declared general business closures (Hatchett 2007). Only prophylactic measures were taken to reduce the spread of the Spanish flu (World Bank 2020, pg. 135).
8 In support of our assumption, Velde (2020) finds that the epidemic and public health interventions had small negative to zero effects on US retail sales, employment, pig-iron production, and bank loans.
9 As far models focusing on output, Ludvigson et al. (2020) quantify the macroeconomic impact of Covid-19 in the US, Chudik et al. (2020) develop a multi-country econometric model, and McKibbin and Fernando (2020) explore the global macroeconomic effects using scenarios analysis.
10 See the collection of articles in Baldwin and di Mauro (2020).