VoxEU Column COVID-19

Decomposing demand and supply shocks during COVID-19

Recent academic discussions have sought to understand whether the economic impact of the COVID-19 crisis and associated lockdown should be ascribed to demand or supply shocks. This debate is of some importance since the underlying shock can have significant implications for stabilisation policy. This column tries to answer these questions by using data on hours worked and wages to estimate labour demand and supply shocks for the aggregate economy and for different sectors through an econometric model. It finds that while labour supply shocks accounted for a larger share of the fall in hours, both shocks were important.

In response to the COVID-19 outbreak, governments and public health authorities around the world implemented confinement and mitigation measures such as social distancing. These measures effectively led to the controlled shutdown of entire sectors of the economy, especially those that supply economic activities and services involving high physical contact with other people, such as restaurants, hairdressers, airlines, etc. On one hand, authorities forced many such establishments to close and send their workers home (the so-called lockdowns). On the other hand, consumers themselves also reduced their consumption of these services, regardless of public health policy recommendations.1 Furthermore, as workers in some of these services lose their jobs and income, they also reduce their purchases of other goods and services. This, combined with uncertainty about the evolution of the pandemic, leads to a reduction in demand for goods and services across the board, affecting not just these locked down sectors (Gourinchas 2020).

For this reason, most economists would agree that the economic effects of the outbreak and mitigation measures combine aspects of so-called ‘supply’ and ‘demand’ shocks (Baldwin and Weder di Mauro 2020). A supply shock is something that reduces the economy’s ability to produce goods and services at given prices. Public health authorities and employers preventing service workers from doing their jobs can be thought of as a supply shock. A demand shock, on the other hand, is something that reduces consumers’ ability or willingness to purchases goods and services at given prices. People staying at home and not going to restaurants or movie theaters for fear of contagion is an example of a demand shock. Additionally, as service workers lose their jobs they may stop purchasing other goods such as cars or appliances, which can also be thought of as a demand shock in those specific sectors.

Conventional monetary and fiscal policy can be used to offset aggregate demand shocks, but other policies may be more appropriate to stabilise the economy after a supply shock. Understanding whether a shock is caused by supply or demand is therefore very important for the design and implementation of stabilisation policies. Due to the nature of this specific shock, it is not clear that the government wants to stimulate/stabilise activity in certain service sectors, as that could run counter to the objectives of public health policy. The government could, however, target policies (such as fiscal or credit policy) to sectors that are not part of the lockdown but are subject to aggregate shocks. This implies that it is not just important to understand whether this shock in the aggregate has to do with supply and demand, but it is also crucial to understand this at the sector level (Guerrieri et al. 2020).

In a recent paper (Brinca et al. 2020), we try to answer these questions by using data on hours worked and wages to estimate labour demand and supply shocks for the aggregate economy and for different sectors, using an econometric model.2 The basic assumptions we use for identifying supply and demand shocks are very simple: if we observe hours and wages (prices and quantities) moving in the same direction, we assign more probability to those movements being caused by a demand shock. On the other hand, if we observe hours and wages moving in opposite directions, we assign that to a supply shock.

Figure 1 Shock decomposition for March 2020

Figure 1 plots the shock decomposition we estimate for March 2020, when the lockdown began, for the growth rate of hours worked. The sum of the red and blue bars is the percentage point change in the growth rate of hours worked relative to its historical average; the size of the red bar relative to the blue bar shows how important supply shocks were relative to demand shocks in that sector. The first ‘sector’ is total private employment, and our results show that two-thirds of the decline in hours were attributable to supply shocks. By far the most affected sector was leisure and hospitality, where the growth rate of hours worked fell by almost ten percentage points. Again, supply played a slightly larger role than demand. While most sectors experienced negative supply shocks, some sectors experienced positive demand shocks; for example, retail trade likely benefitted as people stopped going to restaurants and started buying more groceries and cooking at home. The information sector also benefitted, likely due to increased interest of firms by telework software and arrangements.

Figure 2 Shock decomposition for April 2020

Figure 2 repeats the exercise for April 2020, the first full month of lockdown. The total effect on hours worked during this month was much larger across sectors, with total private employment falling by almost 17 percentage points. Again, for most sectors, two-thirds of the decrease seemed to be associated with supply. Also, during this month, the positive demand shocks in sectors such as retail and information vanished, or even reversed.

There are some potential caveats with our approach, ranging from potential nonlinearities caused by such large shocks (our model is linear), and composition effects driving the joint dynamics of hours and earnings. To address these, we take our estimated measures of shocks in April 2020 and compare them to a sectoral measure of how many jobs can be done at home (Dingel and Neiman 2020), something that should affect labour supply more than labour demand. We show that there is a stronger (positive) correlation between the April 2020 supply shocks and the fraction of jobs that can be done at home.3 We believe that this helps validate our methodology and decomposition.

All in all, our results seem to suggest that labour supply shocks accounted for a larger share of the fall in hours caused by the pandemic shock in the months of March and April, but that both shocks were important. In particular, there were significant demand shocks in sectors that should not be directly affected by the lockdown, such as manufacturing. This suggests that targeted stabilisation policy could help assuage some of the effects of the current crisis.


Andersen, A L, E T Hansen, N Johannesen and A Sherdian (2020), “Pandemic, Shutdown and Consumer Spending: Lessons from Scandinavian Policy Responses to COVID-19”, COVID-19 e-print available at arXiv:2005.04630.

Baldwin, R and B Weder di Mauro (2020), Economics in the time of COVID-19: A new eBook,, 6 March.

Baumeister, C and J Hamilton (2015), “Sign restrictions, structural vector autoregressions, and useful prior information”, Econometrica 83(5): 1963–1999.

Brinca, P, J B Duarte and M Faria-e-Castro (2020), “Measuring Sectoral Supply and Demand Shocks during COVID-19”, Covid Economics, Issue 20, London: CEPR Press.

Dingel, J I and B Neiman (2020), “How many jobs can be done at home?”, Covid Economics, Issue 1, London: CEPR Press.

Gourinchas, P O (2020), “Flattening the pandemic and recession curves”,, 3 June.

Guerrieri, V, G Lorenzoni, L Straub and I Werning (2020), “Viral recessions: Lack of demand during the coronavirus crisis”,, 6 May.


1 For example, a recent paper (Andersen et al. 2020) compares personal expenditure between Denmark, a country that imposed a lockdown, and Sweden, a country that did, and finds that consumption expenditures fell in both countries by similar amounts.

2 We apply the methodology of Baumeister and Hamilton (2015), who estimate a Bayesian structural vector-autoregression of the labour market. The SVAR models the joint dynamics of the growth rate of hours and real wages.

3 Furthermore, when we remove leisure and hospitality, which is a clear outlier, the correlation between supply and this measure of telework remains while the correlation between demand shocks and this measure vanishes.

32,126 Reads