Large shocks test an economy’s ability to adapt, adjust, and continue – a capability called ‘resilience’ – in response to the unexpected (Brunnermeier 2021). Critical infrastructure may be damaged or unavailable, leaving other systems stretched beyond capacity. Whether a natural disaster, terrorism or cyber attacks, or a global pandemic, the severity of a crisis is determined not only by the size of the shock, but also by the resilience of the response. The COVID-19 economic and health crisis greatly stressed the major economies (Altig et al. 2020, Aneyi et al. 2021) and led to labour market adjustments (Barrero et al. 2020), but in some cases also elicited unplanned resilience. In a recent paper (Eberly et al. 2021), we argue that fungibility of factors of production at
Eberly, Haskel, Mizen considers the following as important:
This could be interesting, too:
Ehrmann, Holton, Kedan, Phelan writes Views on monetary policy communication by former ECB policymakers
Agresti, Calvino, Criscuolo, Manaresi, Verlhac writes Tracking business dynamism during the COVID-19 pandemic
Itskhoki, Mukhin writes The Mussa puzzle and the optimal exchange rate policy
Gaynor, Sacarny, Sadun, Syverson, Venkatesh writes Opening the black box of hospital mergers
Large shocks test an economy’s ability to adapt, adjust, and continue – a capability called ‘resilience’ – in response to the unexpected (Brunnermeier 2021). Critical infrastructure may be damaged or unavailable, leaving other systems stretched beyond capacity. Whether a natural disaster, terrorism or cyber attacks, or a global pandemic, the severity of a crisis is determined not only by the size of the shock, but also by the resilience of the response. The COVID-19 economic and health crisis greatly stressed the major economies (Altig et al. 2020, Aneyi et al. 2021) and led to labour market adjustments (Barrero et al. 2020), but in some cases also elicited unplanned resilience.
In a recent paper (Eberly et al. 2021), we argue that fungibility of factors of production at different locations contributed substantially to economic resilience during the COVID-19 pandemic. This required not only the much-discussed ability of some (fortunate) workers to work from home (Bloom et al. 2015), but also the capital they needed to deploy from these remote-from-work locations, including the capital to connect them to each other and the workplace. We call this ‘potential capital’. Conceptually, it is the equipment – home offices, laptop computers, and internet connections – that can be combined with remote-from-workplace labour to produce output.
We identify two puzzles and two results due to the resilience provided by both potential capital and remote-from-workplace labour.
Why did output fall by so little in the pandemic?
Consider the Covid pandemic – one of the largest economic shocks in living memory and the largest in 300 years for some countries.1 The dark bars in Figure 1 show the peak-to-trough (2020Q1 to Q2) actual decline in GDP for the seven countries we study in our paper. The decline is dramatic: it shows quarterly declines of between 9 and 20 log points.2 Yet, catastrophic as the collapse was, it could have been worse. Many workplaces were closed and people were advised to isolate and to work from home (WFH). As we shall document, since much of the workforce was working from home, there was a fall in hours at the workplace of nearly 30 log points from 2020Q1 to 2020Q2 (an unweighted average in our seven countries). Based on an output elasticity of two-thirds, output should have fallen by 20 log points – far more than the actual (average) drop of 14 log points. Thus, first puzzle: why did output fall by so little in response to these large changes in hours at the workplace? Could the decline in output have been buffered by economic resilience arising from remote labour?
Figure 1 Actual output and workplace output
Notes: Actual Output is the decline in GDP from Q1 to Q2 of 2020 in National Accounts data. Workplace Output is calculated based on factor use at the workplace, from Eurostat, BEA, National Accounts and own calculations (see data appendix in Eberly et al. 20201). Work from home from ONS and Google mobility data is described in the appendix. Average is an unweighted average of all countries.
A second puzzle emerges if we consider capital in addition to labour. As Mokyr (2001) documents, the history of industrialisation shows the point of going to the workplace is that workers have capital with which to work. It seems hard to believe that the economy substituted a 30-log-point fall in hours with a rise in workplace capital in a matter of weeks. Rather, with fewer workers at the workplace, it seems very likely that workplace capital utilisation also fell. Based on industrial electricity consumption, we estimate a fall in capital use at the workplace of just over 20 log points, which, with an output elasticity of one-third, should have reduced output by a further 7 log points. Thus in Figure 1, the light bars, which are the implied workplace output changes, lie below actual bars illustrating our puzzle: why did output fall by so little in response to these large changes at the workplace? Expanding our question above, could the decline in output have been buffered by economic resilience arising from capital deployed from home?
We propose that the answer to these puzzles is that capital equipment and structures at home were brought into use for production, providing the capacity to respond to a large unanticipated shock. During Covid, potential capital alongside labour working from home helped to offset the expected decline in output due to low workplace labour and capital utilisation rates, which explains why output did not fall as much as it might have done. Production was buffered by use of capital and labour at home. Productive labour and capital at home gave firms flexibility to respond to a large shock like Covid. Fungibility between capital and labour at home and the workplace is therefore a source of economic resilience because capital at home has both productive capacity (e.g. it can run software) and connectivity with other workers, making WFH both possible and productive.
Some growth accounting
We will assume the standard structure for growth accounting and TFP estimation, that is, constant returns to scale, perfect competition, and optimising behaviour.3 So as to isolate the effects of interest, the only deviation is to add home versus workplace production. Changes in output, Y, are therefore changes in the inputs times their output elasticities. We suppose that labour services, L*, are hours and capital services, K*, are capital (later we shall write capital services as a product of capital stocks and capital utilisation.) Trends in online sales follow a similar (but more modest) pattern, with some increase in the share of sales being made online during the pandemic, some of which is expected to persist over the medium term (Figure 3). Just over a third of firms expect to sell more online than they did pre-pandemic. This is a significant share, but it affects only around half as many firms as the increase in homeworking. Our new distinction is to differentiate factors at the workplace and those at home. Them the growth in total nominal output (YW and YH being output produced at the workplace and home, respectively) is the growth in labour hours at work and at home (HW and HH) times the relevant elasticities, and similarly for capital services at work and at home (KW and KH), plus the contributions of total factor productivity at work and at home. To get back to the concept of resilience, it is important to note that Y falls if L and K at work fall, which is conventional, but is offset if L or K migrates to home.
To measure the change in log GDP we set the L and K shares as two-thirds and one-third, respectively. By assuming that labour is paid the same at home or at work, then the payments weights are the share of hours at home and at work. But in the labour market we must account for the possibility of furloughs. In the UK, a furlough policy allowed firms to temporarily suspend workers rather than laying them off, with the government paying 80% of their wages. We remove these workers from the calculations. Other countries did not have the same scheme, but we can make sector-by-sector adjustments to remove those who were effectively furloughed.
For the capital shares, using national accounts conventions we can measure share of KH as the dwelling payments as a share of total capital payments. This estimate needs to be adjusted in at least three ways. First, we multiply dwellings capital by labour force participation, assuming that dwellings capital can be potentially brought into production in proportion to the fraction of the population who may potentially be WFH. Second, KH is not just dwellings capital but, for example, domestic computers and the internet, which are not counted as investment in national accounts but as consumption. We also need to adjust for utilisation, which we proxy using final commercial and domestic energy use, corrected for seasonality and temperature, and excluding the output of the energy generation sector itself and (the very volatile) transport sector.
The results below show a growth accounting exercise for seven advanced economies. Data are drawn from the OECD, Eurostat, ONS and Bank of England, with details in Eberly et al. (2021). Figure 2 shows the time series response of four key variables: an index of output, labour hours and energy consumption (100 = 2019Q4), and WFH share of the total workforce. The first three indices show the decline in output, labour hours, and energy use demonstrating the impact of the shock. The fourth shows the response in the workforce, as the share of working from home increased peaking in 2020Q1 in Japan and 2020Q2 in the remaining six countries.
Figure 2 Output, labour, energy, and working from home
Actual output and workplace output
Our results in Figure 3 show log point changes in output and workplace output, as in Figure 1, but for all quarters in 2020. In all countries, output fell in 2020Q1 and precipitously in 2020Q2. After lockdowns were eased in the summer output rebounded across the board. As the lighter line shows, workplace output fell by more in 2020Q1 and 2020Q2, but this was offset to a degree by the use of potential capital and labour working remotely, demonstrating the resilience effect of working from home. In later quarters, workplace output recovers – in some countries by more than actual output (Japan, Germany, Italy) – as it replaces some remote work, which declines. We estimate economic resilience in the pandemic accounted for 8–14% of GDP in the trough of the COVID recession.
Figure 3 Actual output and workplace output
Exploring the contributions that were made ‘at work’ and ‘at home’, we see in Figure 4 the use of inputs at the workplace fell in 2020Q1 and 2020Q2 (light bars), rose in 2020Q3, and was somewhat flatter in 2020Q4. The contributions from home inputs (dark bars) were the opposite: rising in 2020Q1 and Q2 and then falling. Notice that the contributions were considerable in 2020Q1 in all countries, especially in countries where the virus struck relatively early, such as Japan. The data are more mixed in 2020Q2, reflecting the later impact of the virus in the US and UK, where there was a substantial offset from the contribution of home output.
Figure 4 Capital and labour contributions to home and workplace output
Total factor productivity
Figure 5 plots the effect on TFP. Workplace TFP takes actual output and subtracts off the contribution of workplace input. This generates an apparent rise in TFP in the UK and US initially, since these countries had a large cushion from working from home. That is, in those countries the ‘hidden’ input from working from home causes an apparent rise in TFP. There is then a large drop in workplace-based TFP, as resources switch back to the workplace.
Figure 5 Total factor productivity
Ignoring home capital, the UK and US would have been misconstrued as seeing a productivity boom during the pandemic. Similarly, in continental European countries, productivity would have been higher, although not quite the boom that would have been observed in the US and the UK. The exception is Germany, where manufacturing for export continued to produce output without much interruption, but labour hours across all sectors was lower in 2020Q2 than it had been before COVID; after making adjustments for labour and capital at home the boom is still observed
Home capital and home working proved to be a source of economic resilience that we estimate accounted for between 8% and 14% of GDP in the trough of the COVID recession. Following one of the largest economic shocks in living memory, our results emphasise that the quantitative effects were not as large as they might have been due to the large-scale restructuring of production at pace. If WFH were to be ignored along with potential capital, output would appear to have fallen and productivity would have been exceptionally strong in 2020Q1 and 2020Q2, due to mismeasurement of labour and capital inputs to production of goods and services. The pandemic has revealed under-utilised capital across the economy and across the globe. The gig economy previously uncovered and deployed some of this capacity, such as the part-time driver who uses their domestic vehicle for commercial rides, yet none of these explorations envisioned the deployment of home capital at the scale and speed, with the potential consequences, observed in the context of the COVID-19 crisis. The pandemic revealed unused capacity as a macroeconomic phenomenon.
Altig, D, S Baker, J M Barrero, N Bloom, P Bunn, S Chen, S Davis, J Leather, B Meyer, E Mihaylov, P Mizen, N Parker, T Renault, P Smietanka and G Thwaites (2020), “https://voxeu.org/article/economic-uncertainty-wake-covid-19-pandemicEconomic uncertainty in the wake of the COVID-19 pandemic”, VoxEU.org, 24 July.
Anayi, L, J M Barrero, N Bloom, P Bunn, S Davis, J Leather, B Meyer, M Oikonomou, E Mihaylov, P Mizen and G Thwaites (2021), “Labour market reallocation in the wake of Covid-19”, VoxEU.org, 13 August.
Barrero, J M, N Bloom, and S J Davis (2020), “COVID-19 and Labour Reallocation: Evidence from the US”, VoxEU.org, 14 July.
Bloom, N, J Liang, J Roberts and Z J Ying (2015), “Does Working From Home Work? Evidence From A Chinese Experiment”, The Quarterly Journal of Economics 130: 165–218.
Brunnermeier, M K (2021), The Resilient Society, Endeavor Literary Press.
Dingel, J and B Neiman (2020), "How many jobs can be done at home?", VoxEU.org, 7 April.
Eberly, J C, J Haskel and P Mizen (2021), “’Potential Capital’, Working from Home, and Economic Resilience”, NBER Working Paper 29431.
Mokyr, J (2001) “The Rise and Fall of the Factory System: Technology, Firms, and Households since the Industrial Revolution”, Carnegie-Rochester Conference Series on Public Policy 55(1): 1–45.
1 “U.K. Economy Suffers Biggest Slump in 300 Years Amid Covid-19 Lockdowns”, Wall Street Journal, 12 February 2021.
2 We use natural log changes (times 100) throughout to be consistent with implementation of our growth accounting framework in section 2. For a change to y from x, the log point change = 100*ln(y/x).
3 While these assumptions are standard, they are not necessarily innocuous. Crouzet and Eberly (2018) demonstrate the impact for TFP measurement of relaxing perfect competition and allowing for errors in capital measurement, which can be substantial.