OECD Employment Outlook 2021

Nearly a year and a half into the economic crisis induced by the COVID‑19 pandemic, there is finally light at the end of the tunnel. But even as activity picks up across the OECD, labour markets face enormous challenges. As the crisis has evolved, so individuals most affected by its ravages have shifted. Certain groups however – including those in low-paid occupations, the low educated and the young – have persistently been in the eye of the storm. These groups not only suffered the most substantial reduction in hours worked but are more likely to have experienced this impact through joblessness.

The shape and speed of the labour market recovery is likely to be determined by: the extent to which the ultimate duration of the health emergency and economic crisis destroys those jobs currently “on ice”, triggering a new surge in job losses among those currently on temporary layoff or reduced hours; the ease with which those who have moved into inactivity can quickly be re‑engaged within the labour force; and finally the extent to which new job opportunities emerge to accommodate the growing number of those currently without jobs.

This chapter provides an examination of the unfolding labour market impact of the COVID‑19 crisis, as well as the challenges that are still emerging. The chapter shines a spotlight on those groups who have carried a heavy share of the burden of the crisis and points to areas where there may be a need for more profound and long-lasting support. The chapter is organised as follows. Section 1.1 briefly describes the ongoing development of the crisis and containment measures. The section charts how activity has responded to restrictions as OECD governments and populations have learnt more about the virus and how to live and work alongside it. Section 1.2 then turns to the labour market, examining the impact on unemployment and working hours in the various phases of the crisis and highlighting the labour market challenges that are still emerging. Section 1.3 is focused on those groups whose labour market outcomes have suffered the most during the crisis. The section examines the extent to which each group has been affected by loss of hours among the employed or loss of jobs, and the implications this may have for the speed of recovery. Finally, Section 1.4 reviews the available evidence about the acceleration of long-standing structural changes during the COVID‑19 crisis and their impact on the world of work, discussing the key role that upskilling and retraining policies to support vulnerable individuals will have in the recovery phase.

In early 2020, at the outbreak of the pandemic, most countries were unprepared for the speed of diffusion, the magnitude of the impact, and the duration of the struggle to contain the virus. Even with the recovery phase now underway, albeit with some stuttering because of further waves of contagion, we may yet be surprised by the lasting impact of the pandemic on OECD labour markets and livelihoods.

In March 2020, the speed and size of the shock precipitated by the pandemic plunged the global economy into a severe recession. Strict containment measures and behavioural guidelines, implemented to stymie contagion, had deep economic consequences, but were anticipated to be short-lived (Figure 1.1). The spread of the virus manifested first as an international supply shock – as workers were quarantined or sick, refrained from commuting or were subject to lockdowns, and as companies were forced to suspend operations or preferred to do so. It soon, however, spread to demand, as incomes plummeted and growing uncertainty reduced consumption and investment.

Initial hopes were for a rapid recovery. Indeed, over the course of the third quarter of 2020, many governments relaxed social distancing measures and began to plan for the roll back of support (Figure 1.2). This early optimism, combined with the reopening economies, prompted a strong rebound in GDP in Q3 (Figure 1.3, Panel B). By the end of 2020, however, this optimism had faded. As new variants were discovered across the world, and cases once more began to rise, many OECD countries – particularly in the Northern hemisphere – returned to stringent containment measures and even lockdown. This second wave, however, was far less uniform in its impact than that seen in Q1/Q2 2020. Indeed, as the recovery stalled across Europe, where strict containment measures were reintroduced, elsewhere, in countries such as Australia and Japan (where a substantial second wave never materialised – Figure 1.3, Panel A) and to a lesser extent Canada and the United States (where the second wave came later), the recovery continued throughout the second half of 2020 (Figure 1.3, Panel B).1

As the pandemic has progressed, and scientific understanding of transmission increased, non-pharmaceutical interventions have become increasingly targeted. The contractions in activity that accompanied the second and third waves of restrictions were smaller and less uniform than those seen in the second quarter of 2020 (Figure 1.2).2 They were also associated with more limited behavioural changes (e.g. use of public transport). Nevertheless, as the crisis has lengthened, the ground lost may prove harder to regain. Policymakers now recognise the dangers of rapid relaxation of restrictions while many individuals, after a year of severe precautions, remain uneasy about a quick return to economic activity. Meanwhile, though bankruptcies have been staved off through deep and widespread government support, the extended duration of the crisis has hit many businesses hard and further redundancies may yet materialise as support for business is rolled back.

Deployment of vaccines to combat the virus is now providing greater certainty, and most OECD countries are once again re-opening. Alongside this, increasingly targeted and effective measures to suppress the spread of the virus, and largescale additional fiscal support in many countries have, once more, renewed optimism that the end may soon be in sight. Reflecting this renewed optimism, OECD (2021[3]), forecasts that GDP growth will rise to 5.75% in 2021 and 4% in 2022 in the OECD area. Nevertheless, GDP per capita is unlikely to return to pre‑pandemic levels before 2022 in the majority of OECD countries and in a number of them the full recovery is further down the road.

The economic outlook remains uncertain. Vaccine rollout stalled in a number of countries, with shortages of doses, logistical delays, and scepticism among some populations delaying deployment in the first half of 2021. As a result, the strict containment measures that remained in place in a number of countries throughout the first and second quarters of 2021 may weigh on the recovery in the near term – particularly in the service sector. The evolution of new variants of the virus across the globe continues to temper the cautious optimism of many OECD countries as they plan for a gradual reopening of their economies. At the same time, widespread uncertainty remains about the extent of the financial distress facing employers – particularly small and medium-sized enterprises – see e.g. Hadjibeyli, Roulleau and Bauer (2021[4]). Furthermore, while this more positive outlook extends also to the labour market (see Section 1.2), even as unemployment rates fall it will be important to be aware that long-term scars are likely to remain.

The impact of the COVID‑19 crisis on labour markets across the OECD has been profound. In April 2020, following the onset of the crisis, the OECD unemployment rate saw an unprecedented 3 percentage point increase to reach 8.8% – the highest unemployment rate seen in a decade (Figure 1.4). In just one month, the entirety of the progress made since the financial crisis was erased. A large part of this surge in unemployment was driven by substantial increases in countries such as the United States, and Canada, where large numbers of temporary layoffs inflated unemployment figures as businesses closed and sent their workers home to shelter from the virus.3 In the United States alone, in just one month, the number of people in unemployment swelled by nearly 16 million, to reach over 23 million in April 2020 (see Figure 1.11 below). Numbers then tumbled, more than halving in the following six months, as economic activity recovered and businesses recalled their workers (Figure 1.5).

Elsewhere in the OECD (particularly in countries making heavy use of job retention schemes that support employers to reduce their labour costs by cutting the hours of retained employees – see Chapter 2), while unemployment has risen, the growth in the number of jobseekers has remained modest relative to the size of the shock: rising by around 1 percentage point in the majority of countries over the course of 2020 (see Box 1.1 for details regarding the comparability of unemployment data).4

The unemployment figures, while in some cases dramatic, do not capture the full extent of the impact of the COVID‑19 crisis on OECD labour markets; unemployment is just one form of labour market slack. This is because of the specific nature of the COVID‑19 shock and of the unprecedented policies introduced to support companies, jobs and people – see OECD (2020[5]) and Chapters 2 and 3. Alongside the unemployed, a large number of people both inside and outside the labour force would have liked more employment, either because they were working only few hours or because they were jobless but not available to work and searching for it – i.e. the conditions to be considered as unemployed.

The excess demand for employment is indeed made up of three components (i) the unemployed, those who are both seeking and available to work (ii) the marginally attached, people who are available for work but not searching for it and (iii) the underemployed, full-time workers working less than a full-week as well as part-time workers who want but cannot find full-time work.5 In the context of COVID‑19, and the labour market interventions that have accompanied the pandemic, these additional components of labour market slack have taken on increased importance.

At the height of the first wave of the coronavirus, widespread restrictions on mobility and social interactions, alongside fears of contracting the virus put a sharp break on job search activities as many of those who lost their job were not immediately able to search for a new one. In fact, while aggregate job search usually increases in times of recession, in many countries there is evidence of a reduction in job search during the COVID‑19 crisis (see Box 1.2). Indeed, acknowledging the difficulties, and dangers, of job search during the height of the pandemic, a number of countries temporarily lifted the job search requirements associated with benefit receipt (see Chapter 3).6

Nonetheless, to be considered unemployed, according to labour market statistics, an out-of-work individual must be actively looking for a job. Thus the limits on job search created by the pandemic pushed many of those who would, in normal times, have been classified as unemployed, into the inactive – or marginally attached – population. The contribution of these ‘marginally attached’ individuals to the swell of inactivity was particularly important in Chile, Mexico and Turkey, as well as a number of European countries such as Austria, Ireland, Finland, Portugal and Spain (Figure 1.6, Panel A).

Alongside individuals who remain marginally attached to the labour force, however, in a number of countries, a worrisome proportion of labour force withdrawals in the second quarter of 2020 were driven by increasing numbers who were no longer available for work. These withdrawals into inactivity were likely driven partially by school closures, and the increased demand for labour in the home, that left many, especially women, who may want to take up work, unavailable to do so.

During the third and fourth quarters of 2020, in the context of rolling back of mobility restrictions, the contribution of the marginally attached to the jobless fell back somewhat, reducing by over 2 percentage points in Mexico, Chile, Canada and Ireland and over 1.8 in Spain. Only in Iceland, Greece, Slovenia, Estonia and the Slovak Republic did marginal attachment continue to increase (Figure 1.6, Panel B). Nevertheless, marginal attachment remains – in all countries but Latvia, Luxembourg and Australia – above pre‑crisis levels.

While the contribution of other forms of inactivity fell alongside the numbers marginally attached to the labour market, it remained particularly elevated in Chile (3.6 percentage points), Iceland (2.4), the United States (1.3) and Italy (1.0), above pre‑pandemic levels. Apart from Iceland, these countries have been among those experiencing the longest school closures as a result of the pandemic (UNESCO, 2021[17]).

Just as business closures and the lifting of job search requirements have blurred the boundaries between traditional labour market categorisations of unemployment and inactivity by swelling the numbers putting job search on hold, so a heavy reliance on Job Retention (JR) schemes in many countries has blurred the boundary between employment and unemployment.

Across the OECD, the restrictions imposed by COVID‑19 containment measures were accompanied by support to help businesses in “non-essential” sectors to retain their workforce. Among these measures, JR schemes played a prominent role (see Chapter 2). JR schemes seek to minimise job losses by allowing firms, experiencing a temporary lull in business, to receive support for a significant share of the wages of employees working reduced hours. At the start of the pandemic, many countries, particularly in Europe, eased companies’ access to these schemes, or introduced new, temporary schemes. They increased coverage of sectors and firms (becoming, in most cases, universal); they increased their generosity, and lowered the associated conditionality, in efforts to minimise job losses and enable a quick resumption of economic activity when business closures came to an end. In response, use of the schemes rocketed, with take up in May 2020 being ten times as high as during peak of the global financial crisis. Alongside such wage support, to prevent the need for redundancies, many OECD countries provided largescale liquidity support to firms while a number of countries – such as Spain, France and Italy – directly imposed implicit or explicit bans on dismissals among companies making use of JR support.

To the extent that these schemes have enabled employers to avoid making largescale redundancies, they have prevented the impact of the crisis from translating into mass unemployment. And, given the unprecedented reliance on JR schemes in many countries, adjustments to the working time of workers who retained their employment are playing an unprecedented role. Figure 1.8, below, highlights the extent of the role played by underemployment in the absorbing the impact of the COVID‑19 pandemic on OECD labour markets. Underemployment saw a swift increase since the start of the pandemic, doubling from 5.4% to 11% of the labour force in just one‑quarter. This increase dominated the marked increase in the underutilisation of the OECD labour force in the second quarter of 2020.7 To put this dominance in perspective, in early 2010 – during the peak of the global financial crisis – unemployment accounted for close to two in every three individuals not working, or working less than they would normally, or would hope to. In the second quarter of 2020, the unemployed accounted for less than one in every two. Despite a much higher rate of labour underutilisation the unemployment rate remained below that seen during the financial crisis. This was true in all OECD countries except Australia which did not go into recession during the financial crisis, as well as Canada and the United States, the latter two countries being those in which temporary layoffs swelled employment numbers.

In the third and fourth quarters of 2020, labour underutilisation fell back sharply – by 4 percentage points – with falling underemployment driving the bulk of the decline. It is worth noting, however, that in many countries where underemployment contracted the most, unemployment increased.

The unemployment, underemployment and inactivity figures give an important indication of the large number of individuals affected by the COVID‑19 crisis. However, while stark, they each tell only part of the story. A complete picture of the depth of the impact of the pandemic on OECD labour markets must bring these multiple elements together. The change in hours worked since the start of the crisis provides just such a picture; capturing the impact both on the extensive margin (fewer employed workers) and the intensive margin (remaining workers working fewer hours). On average across the countries for which monthly data are available, total hours worked fell by close to 20% in just one month from March to April (Figure 1.9). The initial impact was felt most immediately among female workers, who saw their hours fall by over 21% compared to a fall of 19% among their male colleagues. As hours began to recover, over the course of the second quarter, however, women appear to have returned to work, and increased their hours at a faster rate than men. These averages, however, mask a degree of heterogeneity across countries, with a particularly stark initial decline in hours worked seen in Canada, Chile, Mexico, the United Kingdom, and the United States – where, by April 2020, hours worked had fallen by more than 20% with respect to the start of the year. In Sweden, where restrictions on activity were more limited, hours worked dropped by no more than 10%.

Beyond a fuller picture of the overall impact of the pandemic on OECD labour markets, a breakdown of the source of the reduction in hours offers a clearer picture of the channels through which the impact of the pandemic has been felt. Figure 1.10, below, divides the year-on-year fall in hours8 into the contribution of workers moving into joblessness, and that of the reduction in hours among workers that remained in employment.

At the onset of the COVID‑19 crisis, close to 4 in every 5 of the unworked hours were accounted for by some form of reduced working time. Even more impressive, the majority of unworked hours – more than 2 in every 3 – were accounted for by workers who, though employed, nonetheless reduced their working time to zero hours. This heavy reliance on the intensive margin to absorb the early labour market impact was driven, in large part, by European countries. Indeed, in countries such as Belgium, France, Greece, Hungary, Luxembourg, the Netherlands and the United Kingdom 9 in every 10 of the unworked hours were accounted for by reduced working time among the employed. In contrast, in the United States, the intensive margin accounted for just one‑quarter of unworked hours, with the majority of the adjustment channelled through joblessness (albeit temporary in many cases – see below).

Working time recovered markedly during the third quarter of 2020 in the majority of countries, as many shops and restaurants reopened, and workers returned to work. On average working time in the third quarter of 2020 was just 4.3% below the same quarter the previous year. This recovery appears to be largely driven by the reabsorption of the intensive margin. As a result, the composition of lost hours changed somewhat during this phase of the crisis, with joblessness taking on an increasingly important role in the adjustment – accounting for approximately 2 in every 3 unworked hours on average (Figure 1.10, Panel B).9 At the same time, in those countries which relied more heavily on temporary layoff in the second quarter of 2020 – notably Chile, the United States, Turkey and Canada, the recovery in hours worked in Q3 2020 was less pronounced.10

By the fourth quarter, as restrictions began to return in a number of countries (see Section 1.1.1), hours worked, once more began to decline – widening on average across the OECD from a 4.3% year on year fall in Q3, to a 5.6% year on year fall in Q4 (Figure 1.10). However, again, the OECD average disguises a degree of heterogeneity in these trends. Indeed, a number of countries – such as Austria, Belgium, the Czech Republic, Greece, Iceland, the Netherlands, Norway, Poland and the Slovak Republic – saw the year on year fall in hours worked more than triple between Q3 and Q4, while countries such as Australia, Chile, Canada, Denmark, Mexico, New Zealand, the United Kingdom and the United States, experienced an increasingly small year on year decline in hours worked or even returned to pre‑pandemic levels. Alongside the renewed fall in hours, the final quarter of 2020 saw a return to the use of the intensive margin to absorb much of this slack. By March 2021 hours worked were still 7% below their pre‑crisis level, on average across the ten countries for which up-to-date, monthly hours worked statistics are available (Figure 1.9), with Canada being the only country showing a full recovery. A gap of about 7% in Q1 2021 with respect to the level of Q4 2019 is also estimated for the OECD area as a whole, using figures from quarterly national accounts in countries for which labour force survey data for Q1 2021 are not available.

Given the external nature of the shock caused by the COVID‑19 pandemic, the resultant short-term liquidity problems experienced by many businesses provided little information regarding their long-run viability. To keep these businesses afloat until economic activity resumes, OECD governments have stepped in to provide unprecedented levels of support, including corporate bond purchases, direct lending, equity infusions, cash grants as well as direct support for labour costs. By allowing firms to reduce labour costs, such schemes were able to ease the immediate liquidity problems resulting from the pandemic and associated lockdowns.

As a result, the COVID‑19 crisis thus far has reversed the historical trend according to which bankruptcies track the business cycle. Indeed, according to the OECD bankruptcy index, in all quarters of 2020, the number of bankruptcies had fallen as compared to the previous year in almost all OECD countries for which data is available.11 However, while such schemes were designed to support firms and workers to weather the immediate impact of the pandemic, as the crisis lengthens, more firms will struggle to maintain solvency – see Demmou et al. (2021[18]) and Hadjibeyli, Roulleau and Bauer (2021[4]). This will necessarily entail permanent layoffs.

Beyond an increasing number of firms facing solvency issues, as the crisis lengthens, employers may increasingly find that labour hoarding encouraged by job retention schemes is a less attractive option – particularly as subsidies are rolled back and increasingly targeted. The benefits of labour hoarding (the tendency of firms to maintain more employees in response to a negative shock than would be needed to fulfil current optimal production) are particularly pronounced when shocks are temporary (Giupponi and Landais, 2018[19]). This is because, while the expected costs are time dependent, the savings, associated with the avoidance of firing and rehiring workers, are not. Expectations that the crisis would have been short-lived were largely set aside by the end of 2020, as uncertainty increased regarding the potential duration of the crisis and the structural changes it would imply. Indeed, employers dramatically reduced their claims for JR support as the economy reopened during Q3 2020, while the subsequent pickup in claims during the second lockdown did not reach the peak of April and May (see Chapter 2).

Similarly, in countries that have relied primarily on temporary layoffs, eventual recall may not be feasible for many of those who are still expecting to return to their previous employer.12 And, while the rapid employment movements between March and June were dominated by temporary layoffs and recalls, as time passes, workers who have not been recalled by their employers stand an increasing chance of becoming permanent (Cheng et al., 2020[20]). Indeed, the rapid fall in temporary layoffs over the course of Q3 2020 was accompanied by an increasing number of permanent job losses – both in Canada and in the United States (Figure 1.11). In the United States, for example, as the proportion of the labour force on temporary layoff fell – by over 10 percentage points between April 2020 and April 2021 – the proportion of the labour force unemployed but not on temporary layoff rose by 1.5 percentage points.

Recall rates among those on temporary layoff have, historically, been relatively high, with estimates suggesting that, in the United States, more than two in every three of those on temporary layoff were eventually recalled – the majority within the first 8 weeks (Katz and Meyer, 1990[21]).13 However, as the duration of the crisis has lengthened, beyond that initially expected at the time temporary layoff decisions were taken, and as businesses continue to grapple with how to adjust, for many the feasibility of recalling employees still on layoff has altered. As a result, historical patterns, observed during a more predictable labour market climate, may not provide an accurate guide. Indeed, by April 2021, of those reporting temporary layoff in the United States, more than 48% had been unemployed for 27 weeks and over,14 while a large number of workers are experiencing repeated unemployment spells during the pandemic. Indeed, using anonymised bank account data, Ganong et al. (2021[22]) have found that more than half of new unemployment insurance claims reflect workers who had already previously received unemployment insurance during the pandemic, suggesting that many of the workers who were recalled in the summer were subsequently laid off again.

With increasing numbers of unemployed, large numbers still working reduced hours or on temporary layoff, and elevated inactivity, the labour market remains vulnerable to a rapid build-up of longer-term unemployment. Many of those currently outside employment have put job search on hold, for a variety of pandemic-related reasons, (see Figure 1.7). As these individuals return to the labour force (alongside those currently working reduced hours and on temporary layoff who find their jobs no longer exist), current levels of labour market tightness – the ratio of job openings relative to individuals searching for employment – may well be expected to deteriorate. This could lead to lower job-finding rates and, potentially long-term unemployment. Over a year into the crisis provoked by the COVID‑19 pandemic, long-term unemployment is becoming an increasingly urgent concern.

Usually defined as the share of the unemployed who have been unemployed for 12 months or longer, the long-term unemployment rate usually begins to rise only one year after unemployment starts to increase. However, according to this definition, and given the lag in the availability of cross-country data, the unemployed made redundant at the start of the COVID‑19 crisis are not yet reflected in the long-term unemployed in the latest available data (Q4 2020). However, looming long-term unemployment can already be observed in the increasing numbers remaining unemployed between 6 months and 12 months. In the absence of a strong pickup in vacancies, these numbers provide a strong indication that the long-term unemployment rate will soon increase. As a result, the analysis below follows the Bureau of Labor Statistics in the United States in concentrating on those unemployed for greater than 6 months.15

By the fourth quarter of 2020, nine months after the start of the pandemic, on average across the OECD, the number of individuals unemployed between 6 and 12 months had more than doubled since the onset of the pandemic (Figure 1.12). This large increase reflects a climate of both limited vacancies and limited job search that led to relatively few of those made unemployed at the start of the pandemic returning to employment by the end of the year. In the United States and in Canada, where joblessness absorbed a large part of the early labour market shock, the proportion of the labour force experiencing an unemployment spell of 6‑12 months had risen by more than 540% and 370%, respectively.16 In Australia, and in countries across Europe, despite substantial job support, the proportion of the labour force with longer unemployment duration was already beginning to edge up, with countries such as Austria, Denmark, the Czech Republic, Spain, Lithuania and the Netherlands all seeing the numbers unemployed for 6‑12 months increase by more than two‑thirds while Iceland, Estonia, Slovenia, Ireland, Portugal, Norway and Australia saw those unemployed between 6‑12 months more than double by Q4 2020. In the United States, those unemployed for at least 6 months accounted for 43.4% of all the unemployed in March 2021, approaching the historical peak of 45.5% in April 2010, to fall slightly to reach 42.1% in June 2021 in the aftermath of the improvement of the US economy.17

The increase in the number of individuals unemployed for 12 months or more remains relatively limited in the majority of OECD countries. In the fourth quarter of 2020, less than a year had passed since the onset of the pandemic, hence long-term unemployed do not yet reflect its impact. Indeed, a number of countries (France, Greece, Ireland, Italy, Norway, Portugal, the Slovak Republic, Turkey) saw a declining proportion of the labour force with unemployment spells lasting longer than one year. This, however, likely results from individuals with longer unemployment spells becoming discouraged and abandoning job search in light of the additional hurdles created by the coronavirus pandemic alongside the suspension of mutual obligations in many countries (see Chapter 3).

The build-up of long-term unemployment also depends, in addition to the entries into unemployment, on ability to escape unemployment quickly. In the context of the continued uncertainty surrounding the spread of new variants of the virus, the date when social distancing will no longer limit economic activity, and the large degree of hidden slack, even a relatively moderate inflow may still be cause for concern.

While short periods of joblessness are of less concern, especially when unemployed persons are covered by unemployment insurance schemes or other forms of financial support, prolonged periods of unemployment are more problematic. In particular, as income insurance is exhausted and savings are depleted, long-term unemployment can lead to financial hardship. Indeed, recent figures collected via the OECD Risks that Matter Survey (OECD, 2020[23]) suggest that, on average across surveyed countries, close to one in three households affected by job loss since the start of the pandemic report being unable to pay a usual expense, while one in eight report having gone hungry (Figure 1.13). Financial hardship, while worrisome in and of itself, can also have potential long-run employment repercussions if it obliges jobseekers to accept lower quality job offers, potentially leading to skills mismatch.

Alongside financial hardship, and the mental and material stress that goes with it, long-term unemployment may lead to ‘scarring’ that can impede future job prospects; making future jobs harder to find, less lucrative, and more unstable. That jobseekers who have been unemployed for a relatively short period, find jobs at a faster rate than the long-term unemployed, is now relatively well established – see e.g. OECD (2018[24]). However, it is not yet clear what drives this association, nor the extent to which such scarring will occur in the context of COVID‑19 related unemployment. Where scarring is driven by human capital depreciation, and the deterioration of skills during time spent outside employment – see (Pissarides, 1992[25]), for example – then its effects may be of wider concern beyond the long-term unemployed. Individuals working reduced, or zero, hours for extended periods – even if they have not technically been unemployed – are also likely to experience human capital depreciation. This is particularly likely to be the case where those on working zero hours have not had access (or perceived need) to undertake labour market training (see Chapter 3).

Yet the scarring associated with long-term unemployment is also likely to result partially from the stigma associated with protracted unemployment spells. Where employers view such spells as a negative signal of jobseeker quality, employer discrimination can cause longer unemployment spells to self-perpetuate – see for example Farber et al. (2018[26]). Indeed, experimental evidence, on the basis of CV-testing, suggests that, ceteris paribus, the likelihood a job applicant is called for interview significantly decreases with the length of their spell in unemployment (Kroft, Lange and Notowidigdo, 2013[27]) particularly for those with very long unemployment spells (Farber et al., 2018[26]). In the current economic climate, widespread joblessness is largely reflective of the exceptional limits on economic activity and the associated financial difficulties of many employers, rather than the quality of the work of the individuals who have been made redundant. As a result, the stigma associated with longer-term unemployment may be moderated in the context of the external shock of COVID‑19. Indeed Kroft, Lange and Notowidigdo (2013[27]) find the stigma effect to be weaker in less tight labour markets, suggesting that employers do recognise that the signal provided by unemployment duration is less informative when unemployment is high. Nevertheless, as the crisis lengthens, if newly unemployed continue to enter the pool of job-seekers, those with longer unemployment spells may increasingly find themselves at the back of the queue when vacancies pick up.

The extent and effects of long-term unemployment during the COVID‑19 induced crisis will have long-term implications for the widening vulnerabilities in the labour market. Already, early research suggests that groups that had the highest unemployment rates in April also tended to have the lowest reemployment rates (Cheng et al., 2020[20]). As the crisis continues, there is the risk that a gulf emerges, not just between those that have been able to work from home and those that have suffered job and income loss – see OECD (2020[5]) and Chapter 5 – but also between those that have weathered the crisis through recourse to reduced hours and short periods of temporary layoff, and those that have found themselves jobless, increasingly distant from the labour force and risking long term scars.

As vaccines are rolled out and economic activity is, once more, able to resume across the board, there is hope that excess savings have created a strong demand potential that may drive forward the eventual recovery. Indeed, OECD (2021[3]) forecasts further fall in unemployment in 2021 and 2022 to reach 5.7% in the last quarter of 2022. This improved outlook will, nevertheless, leave unemployment above pre‑crisis rates in most countries (Figure 1.14), with continued labour market slack throughout 2021‑22. This is largely due to the expectation that the absorption of the slack embodied in workers on JR support currently working reduced hours will precede largescale job creation.

Although the projected recovery is more optimistic than earlier forecasts, OECD (2021[3]) forecasts a significant degree of heterogeneity in the pace and pattern of the recovery across OECD countries. Indeed, a few countries have already recovered pre‑crisis employment rates and, by the end of 2022, many countries will see their employment rates at or approaching their pre‑pandemic levels. Yet, in a few others, employment is expected to take several years before returning to pre‑pandemic levels.

Only when the labour market is no longer at risk of being constrained by mandatory restrictions on activity, school closures, and individual concerns to avoid infection, will it be possible to gain a fuller grasp of the full extent of the required recovery. The shape and speed of the labour market recovery will indeed likely be determined by the extent to which the ultimate duration of the pandemic destroys those jobs currently ‘on ice’ – either on temporary layoff or reduced hours – triggering a new surge in job losses, the ease with which those who have moved into inactivity can be re‑engaged within the labour force and the extent to which new job opportunities emerge to accommodate the growing number of those currently without jobs.

As the rollout of vaccines brings renewed hope that the pandemic may be drawing to an end in a number of OECD countries, OECD labour markets still face enormous challenges. An unprecedented number of people on reduced hours, on temporary layoff, or out of work entirely, have done little or no work in over a year. The impact of such worklessness risks far outlasting the crisis itself. As stock can now be taken of the likely long-term implications of the past year, it is important to examine this experience, and what it tells us about who will bear the economic pain in the months, and years, to come.

During the first phase of the crises, at a time when many OECD countries were in lockdown, the severe reductions on mobility and social proximity triggered by the COVID‑19 pandemic put many sectors on hold. The initial shock of the pandemic was therefore shared across large swathes of the economy. As economies have slowly re‑opened, however, and as we have increasingly learned to live, and work, alongside the virus, the eye of the storm has become increasingly focused on sectors such as hospitality, tourism, arts and leisure.18

In accommodation and food service activities, the number of hours worked across the OECD more than halved in the second quarter of 2020. At this time, expectations that closures would be short-lived, prompted widespread use of job retention schemes as employers attempted to keep workers in their jobs in anticipation of a V-shaped recovery. As a result, close to two in every three of the lost hours in accommodation and food services were accounted for by individuals reducing their normal hours (see Figure 1.15). By the third quarter, as shops, restaurants and hotels were reopened, the fall in hours worked was a more modest 20%, as many of those on furlough and temporary layoff returned to work. However, the burden of adjustment moved to the extensive margin, with many on short hours returning to work while jobs destroyed were not recovered. As a result, job destruction accounted for over 80% of lost hours in the third quarter of 2020. A similar pattern was seen in the Arts sector, where hours worked fell by over 42% in the second quarter of 2020 before retrenching somewhat to a 14% year-on-year fall in the third quarter of 2020. A notable exception to this trend is seen in the United States, where reliance on temporary layoff has meant that the extent to which the extensive margin absorbed the reduction in hours worked – even in the second quarter of 2020 – was more pronounced with net job destruction accounting for approximately four in five of the reduced hours in sectors such as arts, as well as accommodation and food services, even in the second quarter of 2020.

In contrast to the modest reduction in hours lost due to net job destruction in the third quarter in sectors such as accommodation and food services and arts, the transportation and storage sector saw an increase in the hours lost because of job destruction in the third quarter of 2020. This may be due to limited openings of seasonal jobs and/or reflect the fact that some of those workers, initially on reduced or zero hours, found their jobs were destroyed by the third quarter – perhaps reflecting changing expectations regarding the duration of the crisis, in particular as regards future demand for travel.19

Also during the third quarter, retail, previously in the eye of the storm, began benefitting, alongside manufacturing, from shifts in spending away from services towards goods, dampening the reductions in hours worked seen during that period. Meanwhile, a number of sectors – including real estate, business services and construction – saw a strong rebound, as economies began to reopen during the third quarter and hours worked returned to levels seen the previous year, prior to the pandemic.

In contrast, hours worked in information and communication, as well as in financial and insurance activities, saw an increase in hours worked compared to the previous year, with these sectors making limited use of reduced hours while increasing labour along the extensive margin. Indeed, employment in these two industries continued to grow on average throughout the peak of the crisis. In the case of finance and insurance, or information and communication, this pattern likely results from the speed with which they were able to adapt by adopting changed work practices such as reduced travel and working from home – see Dingel and Neiman (2020[29]).

In sectors where the full throttle impact of the crisis was relatively short-lived, such as real estate, as well as the health and social work, the fall in unworked hours accounted for by workers on reduced or zero hours between the second and third quarters of 2020, was not accompanied by a rise in net job destruction in the third quarter. This pattern maybe suggestive of success, in these sectors, of the largescale use of job retention schemes that allowed employees working reduced hours to quickly return to work in in those sectors experiencing a rebound. Other sectors – such as agriculture – that were less reliant on physical proximity were also affected to a much smaller degree.

As case numbers again began to rise in the fourth quarter of 2020, the year on year fall in hours increased in almost all sectors relative to that seen in Q3. The exceptions – of information and communication, financial and insurance services and business services – being sectors in which workers have been relatively able to adapt to mobility restrictions through increased work from home. Despite the negative impact of the second wave on hours worked, the majority of sectors, were able to adjust largely through recourse to reduced working hours among the employed, with the extensive margin absorbing part of the impact only in the worst hit sectors – accommodation and food service activities, and the arts.

There remains a high degree of uncertainty regarding the duration and form of the ongoing (and evolving) restrictions on activity, as well as the permanence in the changes to habits and consumer preferences that have resulted from extended shutdowns. While some sectors, such as construction and real estate, may benefit, even before the pandemic has come to an end, from low interest rates, pent-up demand, and consumers’ desire to improve their living conditions, for others – such as accommodation and food services – the rebound may come too late for many businesses. If consumers make permanent shifts in how they work, shop, and spend their leisure, even with the acceleration of vaccination, a rebound in some sectors may never come. Recent OECD work estimates a large increase in firms that may become distressed as a result of the falling profits induced by the pandemic and associated restrictions (Demmou et al., 2021[18]). Indeed, while results differ across types of firms, in sectors experiencing the largest adverse impact, such as “Accommodation and food service activities” as well as “Arts and Entertainment”, up to 32% and 24% of otherwise viable firms are expected to become distressed, respectively, even in the context of current support measures. If a wave of bankruptcies is on the horizon, slack in certain sectors of the economy may still have some way to go.

The concentration of the impact of the COVID‑19 pandemic on the service sector is unusual. Indeed, in contrast to manufacturing and construction, which typically suffer more from cyclical downturns, services tend to be more resilient. The heavy impact of the current recession on the service sector may have implications both for the speed of recovery (Beraia and Wolf, 2021[30]), and – given the concentration of certain socio‑economic group in service sector occupations – the extent to which this crisis falls on the shoulders of the most vulnerable (Box 1.3).

The economic fallout of the coronavirus pandemic has had profoundly different impacts across socio‑economic groups, leaving some to shoulder the bulk of the burden, while others suffered little and recovered quickly. In the United States, for example, where transaction data from several private companies has been used to study the impact of the pandemic on individual’s employment and spending, employment is found to have largely recovered among the upper income quintiles while it remains subdued at the lower end of the distribution.20

Much ink has already been spent examining the impact of the pandemic on certain socio‑economic groups, including the low paid, low educated, youth and women – see for example OECD (2020[5]), Adams-Prassl et al. (2020[33]), Cheng et al. (2020[20]). Building on this existing work, the analysis below expands on how this impact has evolved over the course of the crisis, looking in particular at who has benefited from the shift of the labour market impact from the extensive to the intensive margin; from joblessness, to reduced hours among the employed. The form the impact of the pandemic has taken varies across socio-demographic groups and will likely have implications both for the speed of recovery, and for the longer-term challenges that may yet emerge.

The coronavirus pandemic has changed how we think of low-wage employees and highlighted the extent to which society depends upon essential workers. At the same time, praise for the heroic work of these workers, in conditions that are often dangerous and exhausting, has been widespread. In some quarters, this appreciation has been accompanied by concern that job quality in a number of essential sectors matches neither the importance of the work, nor the hazards involved. Indeed, recent research using data on 800 000 commercially insured individuals in Philadelphia, the United States, suggests that, during lockdown, essential workers were 55% more likely than others to get COVID‑19 (Song et al., 2021[34]). While the identification of work that is considered to be “essential” is not clear-cut and varies across – and even within – countries, the category ‘essential worker’ tends to include those working in: health and social care; education and childcare; food and other necessary goods; key public services; local and national government; utilities; public safety and national security, and transport.21 In Europe, such workers account for slightly more than one in every four employed individuals.22

Protecting all essential workers is indisputably important, yet only a subset must be physically present at their workplace. These, the most vulnerable to health risks, tend to be labelled “frontline” workers and, in many countries, have had priority access to childcare, protective equipment and vaccines. These ‘frontline workers’, not only tend to be more exposed to the virus, but are also likely to be less able to protect themselves from the financial consequences of the virus (see Box 1.4 on frontline workers in the long-term care sector). Indeed, building on the work of Dingel and Neiman (2020[29]) to identify those essential workers whose work requires their physical presence, recent work by Blau, Koebe and Meyerhofer (2020[35]) attempts to identify ‘frontline workers’ as distinct from the larger pool of essential workers. The authors find that, in the United States, while the broader group of essential workers tends to mirror the demographic characteristics of the labour force, frontline workers are less educated, tend to earn lower wages and encompass a relatively high proportion of immigrants.

Alongside the increased vulnerability to infection among low-wage frontline workers, however, workers in low-wage occupations more widely have been disproportionately vulnerable to loss of income, job loss and loss of hours as a result of the pandemic. Furthermore, JR support appears to have been less effective at protecting the labour market attachment of those in low-paid occupations, who have seen the fall in their hours of work manifest largely along the extensive margin – see Chapter 2 for more analysis of JR schemes. This is likely a result of the smaller proportion of low-paid occupations on stable and protected employment contracts, as well as the more limited costs associated with hiring/firing low-paid employees.

Figure 1.17 builds upon the occupational categories defined in Goos, Manning and Salomons (2014[36]), to aggregate occupations into those which are highly paid, middle paid and low paying.23 Low-paying occupations took a strong hit in the initial months of the crisis. Indeed, the average reduction in hours in these occupations, across the OECD, at 28% exceeded that experienced among high-paying occupations by over 18 percentage points. In countries such as Portugal and Spain these low-paying occupations saw hours fall by over 40% when compared to the previous year (see Annex 1.B). These patterns stand in contrast to the trends in vacancies following the initial onset of the crisis, when falling vacancies were comparable in both high and low paying occupations (OECD, 2020[5]) with middle paying occupations experiencing a slightly stronger negative impact.24

Furthermore, the apparently homogenous impact on vacancies also appears to have disguised a strongly differential impact on the extensive and intensive margins as revealed in the hours of existing employees. In the second quarter of 2020, across the OECD, over 34% of the hours reduction in low-paying occupations were the result of net job destruction. Among highly paid occupations, meanwhile, even in the context of reduced total hours, hours worked on the extensive margin remained incrementally positive. This suggests that the low paid have faced a double disadvantage. Alongside the magnitude of the fall in hours among low-paying occupations, the extent to which this reduction translated through joblessness suggests that the low paid may have disproportionately suffered from instability, income loss, and longer-term career damage.

By the third quarter of 2020, while total hours worked by those in highly paid occupations were largely back to their pre‑pandemic levels, among those in low-paid occupations total hours worked remained 10% below those seen in 2019, with the preponderance of the adjustment – over 80% – now accounted for by job destruction. It is also notable that, in Q3 2020, as hours begin to recover from the peak of the crisis, some convergence is observable in the pattern of recovery among low and middle paid occupations – while many of those working reduced hours returned to work, little recovery was observed on the extensive margin.

In the fourth quarter of 2020 the year on year fall in hours once more widened. However where, once again, the entirety of the contained hours losses among high-paying occupations has been, on aggregate, absorbed by employees working reduced hours (indeed an increase in job creation has mitigated the hours lost to reduced hours), among low-paying occupations hours lost to job destruction have not recovered. In fact, representing over half of the reduced hours, job destruction among low-paying occupations accounted for an even larger share in the second wave than it did in Q2, during the first. In contrast to the middle and high-paying occupations, the lion’s share of the remaining lost hours were accounted for by individuals moving down to zero (as opposed to reduced) hours.

The heavy impact of the pandemic on low-paying occupations is partially a reflection of the concentration of many low-paying occupations in those sectors most affected by closures and reduced demand – particularly in retail and trade (Box 1.3). It may also partially reflect the skill composition of those in low-paying sectors and the concomitant implications for the incentives of firms to retain (or not) these skills through extended reliance on JR schemes.

In the early phases of the COVID‑19 pandemic, a strong and widespread fall in hours worked was seen among workers at all levels of education. The initial labour market impact was felt most strongly, however, among those with a more limited education. Across the OECD, average hours worked fell by 8.5% among the high skilled, 20% among those with a medium level of education, and 24% among those holding just a lower secondary education diploma or less (Figure 1.18). This disparate reduction was most marked in countries such as Ireland, the Slovak Republic, Slovenia, the United States and Finland, where the reduction in hours worked among the low skilled was 25 percentage points larger than that among those with a high level of education (some tertiary). The trend was less notable in Mexico, Greece, the Czech Republic, Austria, Denmark and Switzerland, where the difference remained under 10 percentage points. In Latvia and Lithuania, highly educated workers experienced a larger impact on their labour market.

Alongside having experienced a larger contraction in demand for their labour, the contraction in hours worked among the low educated was also more frequently experienced on the extensive margin. Indeed, among workers holding a low level of education, the increase in net joblessness accounted for about half of the hours lost in the year to the second quarter of 2020. This is likely to be partly a reflection of the fact that temporary jobs are more widespread among the low-educated (see Box 1.5) and were less covered by JR schemes. Conversely, consistent with the finding that net job destruction was limited in high-paying occupations, net joblessness did not increase for the highly educated – rather the entirety of the reduction in hours was channelled through the intensive margin.25

Perhaps more worrisome still, from the perspective of the growing labour market inequalities provoked by the COVID‑19 pandemic, is the evolution of the composition of unworked hours from the second to the third quarter of 2020. In contrast to the patterns observed among low/medium/highly paid occupations, the third quarter of 2020 saw educational disparities consolidated, as many of the medium and highly educated returned to work from reduced or zero hours while, among the low-skilled, joblessness persisted – even increasing in some countries such as Latvia, Lithuania, Slovenia, the Slovak Republic, Belgium, the Czech Republic, Sweden, Chile, Norway, Greece and Poland.

In the fourth quarter of 2020, the year on year change in hours worked again deteriorated with the increasing spread of the virus and concomitant restrictions. This was largely felt through increased reliance on reduced hours at all levels of education. However, among the highly educated the reliance on the intensive margin to absorb the shock was accompanied by net job creation, while among those with a mid or low level of education the fall in hours was seen along both the extensive and intensive margins.26 Overall, among those holding only a low level of education, at the end of 2020 almost 10% fewer people had a job than one year prior, while employment of those with at least a college diploma grew by 3% during the same period.

It is important to note that an examination of the impact of employment and hours alone cannot capture the full extent of the impact on earnings. This is because it is not possible to distinguish in the hours worked data between those who reduced their employment through JR schemes, and those who were on unstable contracts and whose reduced hours were not compensated through JR schemes. The proportion of hours reduction that was compensated is likely to be even lower among the low paid. Indeed, the Low Pay Commission in the United Kingdom found that the proportion working reduced hours on full pay was increasing in income (Low Pay Commission, 2020[39]).

From the perspective of employers, much of the benefit of retaining employees in the face of temporary reductions in demand through relying on subsidised unworked hours, accrues through savings made by avoiding the costs associated with firing and rehiring workers (see Chapter 2). These costs are likely to be higher for highly educated workers, who tend not only to hold more stable positions benefiting from employment protection, but are, in many cases, harder to replace. Furthermore, in addition to these firing/rehiring costs, refilling certain positions also implies substantial firm-specific, or positions-specific, reskilling costs. To the extent that such positions are also more likely to be occupied by more educated workers, it is to be expected that joblessness accounts for a smaller proportion of reduced hours among highly educated workers (Pfann and Palm, 1993[40]). While intuitive, the finding that subsidised labour hoarding benefits disproportionately the highly educated has important implications for the complementarity of job support and out-of-work benefits and suggests the need for a strong unemployment insurance system and other forms of out-of-work income replacement.

Young people have shouldered a heavy part of the burden of the COVID‑19 pandemic and associated restrictions on activity. At the best of times, the youth labour market is highly sensitive to economic cycles; having been hired relatively recently, they tend to have had less chance to accrue firm-specific skills and experience. And, as the last in, young workers are often the first out. The crisis prompted by COVID‑19 has also been particularly damaging to the youth labour market because young people tend to be more likely to work in those sectors most affected by lockdown and social distancing measures, notably in hospitality and non-food retail (Box 1.3).

At the onset of the pandemic, unemployment among 15‑24 year‑olds in the OECD surged, from historical lows of just 11.5% in February 2020, to 19% in just two months. This was more than two times the percentage point increase seen in the unemployment rates of those aged 25 and over. As with headline unemployment figures, these dramatic fluctuations are driven, in large part, by vast swings in those countries that have relied heavily on temporary layoffs. Both in the United States and in Canada youth unemployment rates increased by 17 percentage points in just two months, reaching over 27% in April 2020. In the European Union, however, youth unemployment has, thus far, remained substantially below levels seen during – and for some time after – the global financial crisis (Figure 1.20). Nevertheless, even in Europe, the 3 percentage point increase in youth unemployment rates since the start of the year, significantly outpaced that seen among their older peers (by 1 percentage point). And in many countries rates continue to rise (Figure 1.20). This strong increase in youth unemployment is likely driven, both by flows from employment to unemployment, but also by increased numbers of those who join the labour market but, in the context of limited hiring, are not able to access an initial foothold in employment. With further restrictions on economic activity introduced during Q4 2020 and Q1 2021, alongside large numbers of young people leaving education into a labour market with limited vacancies, youth unemployment rates will likely remain elevated for some time to come.

Given the concentration of young workers in the sectors affected by lockdown and social distancing (Box 1.3), and in less stable contracts, it is unsurprising that, alongside unemployment, young workers have seen heavy reductions in their hours of work. In the second quarter of 2020, across the OECD, reductions in working time contributed to close to half of the 24% fall in hours worked by this age group. And, in contrast to the pattern among prime aged and older workers, only in a minority of countries – including Austria, Iceland, Turkey, and to a lesser extent Greece, the Slovak Republic and Italy – did the adjustment occur through a partial reduction in working hours – see Figure 1.21 and Figure 1.22. Indeed, the vast majority of reduced hours among youth was accounted for by young workers moving down to zero-hours employment. It is likely that, given the concentration of young workers in non-standard contracts, a proportion of this zero hours employment may have been outside job retention support (for example through paid or unpaid leave), however, the available data do not allow examining the extent of this.

During the third quarter of 2020, as lockdown and social distancing measures began to ease across the OECD, many of those young people working zero hours began to return to work. However, given the relatively large role played by joblessness in the reduced hours among youth, in contrast to prime‑aged and older workers, the fall in their unworked hours has not seen a substantial retrenchment.27 In the fourth quarter of 2020, as workers both young and old, returned to shorter hours, the year on year fall in hours resulting from joblessness among young workers remained prominent.

The increase in joblessness among youth prompted by the pandemic arises, largely, through two channels. In the first place, there are those young people – often working in hard hit sectors and on precarious contracts – who lose their jobs. In addition, however, there are large numbers leaving the education system – either dropping out, or reaching the end of their educational career – who struggle to find employment in the context of limited vacancies. While timely cross-country data on youth hires are limited, some tentative indication of this entry margin can be gained by looking at the hiring rate, defined here as the proportion of employed youth who started their jobs in the past three months (Figure 1.23, Panel A). The second quarter of 2020 saw the hiring rate among youth fall substantially – falling, year on year, by more than 5 percentage points in Spain, Greece, Ireland, Portugal, Italy, France and Estonia (Figure 1.23, Panel B).28 This fall in the hiring rate, sustained, albeit to a lesser degree, in the third and fourth quarters of 2020, suggests that new entrants to the labour market have accounted for a relatively large degree in the increase in joblessness among youth.

Many young people remain in full-time education, and have struggled through the crisis learning at distance. This has taken a heavy toll on the mental health of many young people, and the true costs in terms of lost learning, particularly among the most vulnerable, may not be fully realised for many years. Others, those young people who are now leaving the education system, or had only a very tentative foot in the labour market when the pandemic struck, are particularly exposed – to unemployment in the short-term, and to greater risk of scarring if they go through a long spell of unemployment and inactivity. Finally, there are those who may have planned to work alongside their studies in order to finance their education. For these young people, the paucity of employment prospects may compromise, not only their labour market activity, but also their educational careers and, as a result, their long-term career prospects.

Work-based learning opportunities and apprenticeships have also been hard hit as employers have often be forced to cut such schemes, or conduct them at distance due to social distancing measures and business closures. In Germany, for example, less than half a million people agreed on new apprenticeship contracts in 2020, down 9.4% from the previous year. In the United Kingdom, only around 61 000 apprenticeship began between March and July 2020 – a year-on-year fall of over 45% (OECD, 2021[44]).

In 2019, the share of young adults not in employment, education or training (NEET) was one of the lowest since the turn of the century. At the end of 2019, prior to the onset of the pandemic, on average across OECD countries, just over 1 in every 10 young people aged between 15 and 29 were NEET. In the early phases of the COVID‑19 crisis, however, the NEET rate swelled across the OECD (Figure 1.24, Panel A), reversing the trend of the past decade – rising more than 4 percentage points in countries such as Canada, the United States,29 Ireland, Turkey, Spain and Portugal. Particularly worrisome in the second quarter of 2020 was the extent to which the increasing number of NEET were concentrated in inactivity.

By the third quarter of 2020, as mobility restrictions diminished, economies began to re‑open, and opportunities returned (albeit in limited numbers). Many young NEET returned to active job search and the year on year growth in inactive NEET fell substantially back. Nevertheless, by the fourth quarter of 2020, inactive NEET numbers remained elevated in the majority of countries – including Turkey, the United States, Greece, Italy, Iceland, Ireland, Hungary, Estonia and France (Figure 1.24, Panel B).

This increase in inactive NEET is unsurprising given the restrictions on activity imposed by the virus, and more limited availability of employment and childcare services. Nevertheless, this trend stands in contrast to that seen during the global financial crisis (Carcillo et al., 2015[45]). And, given that periods of inactivity have been shown to be particularly damaging for the career prospects of young people, high levels of inactivity among young NEET risk enhancing scarring. As a result, it will be important to identify and contact those who have dropped out of the education system, and to ensure that those youth who are NEET are not left to drift further and further from the labour market. Often, the most vulnerable young people do not get in contact with the Public Employment Services (PES) or youth services, because they are not entitled to income support, because they lack trust in public authorities, or simply because they are not aware of the support they can receive. Rapid and proactive outreach will be particularly important in the current crisis.

A young person’s level of education has typically been an important determinant of NEET status. Across OECD countries, 25‑29 year‑olds with below upper secondary education are four times more likely to be NEET than those with a tertiary education (OECD, 2020[46]). And, in nearly all OECD countries, youth holding only a low level of education (lower-secondary at most) are strongly overrepresented among NEETs. However, in recent years, young people with a medium to higher level of education have accounted for a growing proportion of the NEET (Carcillo et al., 2015[45]) and in a number of countries this share increased further during the COVID‑19 crisis (Figure 1.25). Across much of the OECD, the share of NEETs with at least some tertiary education has risen since the onset of the crisis, and a number of countries saw large increases in the early phases of the pandemic with Denmark, France, Hungary, Latvia, Portugal and Sweden, all seeing year-on-year increases of the share of NEET with a high or medium level of education of over 7 percentage points in Q2 2020. And, while year-on-year growth in the share of NEET with a medium/high level of education fell back somewhat in the third quarter of 2020 in the majority of countries, it increased once more in the final quarter of the year – even outstripping the second quarter year-on-year growth in a number of countries.

The careers of young people can be significantly disrupted by poor labour market conditions at the time they leave school. The long-lasting labour market consequences, directly related to the impact of economic crises can play particularly heavily on the career trajectories of new labour market entrants. When it comes to youth, scarring tends to work through two channels. For those who are unable to find employment upon labour market entry, spells in unemployment and, particularly, in inactivity can weigh upon their future employment and earnings prospects (Dorsett and Lucchino, 2018[47]). Those who successfully find a job can, however, also face lasting disadvantage from scarring if they are forced to accept lower level starting positions, if their mobility is compromised by more limited vacancies, or if they are able to access more limited training and promotion opportunities.30 Indeed research has found that a large recession at time of graduation, not only reduces earnings upon graduation, but the effect persists in subsequent years. Looking at the United States, Altonji, Kahn and Speer (2016[48]) identify an earnings reduction of roughly 10% for the average graduate joining the labour market in a typical recession, with a yearly earnings loss of approximately 1.8% over the first 10 years following graduation. Results of a similar magnitude are found by Oreopoulos, von Wachter and Heisz (2012[49]) who use employer-employee matched data in Canada. There is also evidence that these scarring effects extend to health and well-being (Garrouste and Godard, 2016[50]).

Certain groups of youth may be particularly vulnerable to the effects of scarring. For example, research has found that lower skilled youth, as well as those graduating from fields characterised by relatively low pay, tend to be particularly vulnerable to scarring effects – see Kroft, Lange and Notowidigdo (2013[27]) and Altonji, Kahn and Speer (2016[48]). Furthermore, the effects of the distributional burden tend to differ by recession (Altonji, Kahn and Speer, 2016[48]). And, there may be reason to believe that the social distancing and home working brought about by COVID‑19 will have a negative impact upon the career prospects of even those young people who have managed to secure a job – particularly in fields where post-schooling human capital accumulation is important.

Despite substantial progress in recent decades, women still tend to be less firmly attached to the labour market than their male colleagues. Women in employment tend to work fewer paid hours than men, earn less, and have shorter job tenure – see OECD (2018[24]; 2020[51]; 2020[52]; 2020[53]). This can leave them more vulnerable than men and easier to lay off. On top of this vulnerability, and in contrast to previous crises, which are often concentrated mainly in male‑dominated sectors – see for example Bredemeier, Juessen and Winkler (2017[54]) – retail, catering, and hospitality – a sector characterised by high female employment – suffered particularly heavily at the start of the COVID‑19 induced crisis. The rise of the service economy in recent decades has been found to account for an important share of the observed trends in the number of hour’s women work, as well as their relative wages – see Ngai and Petrongolo (2017[55]) and Blau and Kahn (2017[56]). And the negative impact on employment in these sectors, prompted by the COVID‑19 pandemic, led to concern that greater job and income loss among women may undermine recent progress made toward closing the gender gap (Adams-Prassl et al., 2020[33]). Indeed, in the initial phases of the pandemic, the majority of OECD countries saw the gender employment rate gap increase – by more than 1 percentage point in a number of countries, including Slovenia, Canada, Sweden and Finland (Figure 1.26). However, the following quarter saw a reversal of this widening gender employment gap in the majority of countries, albeit with a number of exceptions including Slovenia, Lithuania, Canada, Sweden, Finland, Hungary, Colombia and Belgium.

Nevertheless, as well as being overrepresented in the lockdown sectors, women also make up a disproportionate share of workers in many of the sectors, defined as essential, that have often been required to work additional hours to cope with heavy demand (see Box 1.3). And, beyond jobs either shut down or defined as essential, the impact of the pandemic on employment in the early phases of the crisis was, to a large extent, dependent on the ability to work from home – which tended to be possible in many female dominated sectors such as education.

As understanding of COVID‑19 advances, and as new variants-of-concern emerge, restrictions are constantly evolving. In later phases of the crisis, many of those working in male‑dominated sectors (such as construction, repairs, and large parts of manufacturing) that do not require close interaction with colleagues or clients have been able to return to work. At the same time, as restaurants have once again closed, and as new more contagious strains of the virus have led to school closures in a number of countries (see Figure 1.2), many of both the push and pull factors that kept women from work at the start of the outbreak have returned. These trends may still be reflected in later data releases.

Indeed, increased caregiving typically affects gender disparities in labour market outcomes slowly, over time, as women – in particular – move down to part-time work, leave the labour market entirely, or merely search for jobs with more flexibility or a shorter commute. These choices often translate into slower wage growth – through limiting the pool of jobs, weaker bargaining power and scarcer opportunities for promotion once in situ – see e.g. OECD (2018[24]). In this respect, the pernicious repercussions of the pandemic may yet be felt for many years.

In recent years, the earnings penalty associated with motherhood has remained stubbornly stable (Ngai and Petrongolo, 2017[55]). And, indeed, beyond what happens to their job, the labour market attachment of women has been tested by the closure of school and childcare facilities that has accompanied efforts to contain the virus (see Figure 1.2). The increased care burden that accompanied widespread school closures fell largely upon the shoulders of women – see Hupkau and Petrongolo (2020[57]) in the United Kingdom, Farré et al. (2020[58]) in Spain, and del Boca et al. (2020[59]) in Italy – prompting many to withdraw from the labour market entirely – even in cases where their jobs remain active.

In a number of OECD countries that have introduced JR schemes, or specific care leaves, women have been able to request to move to reduced hours to avoid being pulled from the labour market by home schooling and care responsibilities.31 As a result, the impact on the employment rate gap of the first wave of the virus was not clear cut, but rather varied across countries (Figure 1.26).32

When it comes to hours, the magnitude of impact on hours lost following the initial impact of the crisis tended to be larger among women than among men – both along the intensive and extensive margins (Figure 1.27, Panels A and B). – However, this disparate impact appears to have been short lived in many countries or even reversed. Indeed, hours lost in the third quarter of 2020 were comparable, with male hours falling marginally further than those of women – 4.3 percentage points among women on average across the OECD, and 4.5 percentage points among men. These trends tend to suggest that, for women, policies that channelled the impact of the crisis through the intensive margin appear to have been relatively successful in cushioning the impact of the crisis and enabling a rapid return of women to work. Indeed, in the United States, where the bulk of the impact was felt through the extensive margin, the fall in hours among the female workforce in the third quarter remained larger than that among men.33

The return of the virus in many countries during the fourth quarter of 2020 once again saw hours worked by women fall marginally more, on average across the OECD, than those worked by their male counterparts – 6.2% among women compared to 5.7% among men (Figure 1.27, Panels C and D). And, while the intensive margin again absorbed the majority of this increase, joblessness carried a larger share of the fall in hours during this second wave – particularly among women. However, particularly during this second wave of restrictions, the more substantial fall in hours among women was driven by large disparities in a handful of countries – such as Chile, Slovenia, Turkey and Lithuania. In many countries, particularly across Europe, the impact on hours worked among women during Q4 was comparable or, in many cases, more limited, than among men.

Despite the substantial impact of the pandemic on employment and earnings, governments across the OECD were able to protect household income through deep and wide use of government support – see OECD (2020[42]; 2020[5]). Indeed, between Q4 2019 and Q2 2020, despite falling GDP per capita by 12.4% across the OECD area, real household gross disposable income grew by 3.9% on the back of largescale COVID‑19 government support measures. Such growth was particularly marked in Canada and in the United States where large but temporary support led to growth of 12.6% and 11.1%, respectively (Figure 1.28, Panel A). In these two countries, however, this growth has since retrenched somewhat and, by the fourth quarter of 2020, growth in household disposable income had fallen to a more modest 6.6% and 4% since pre‑pandemic levels. This is reflective of the temporary nature of the increase in net transfers to households (Figure 1.28, Panel B).

Smaller increases in disposable income were also observed between Q4 2019 and Q2 2020 in Ireland (4.5%), Australia (4.1%) and Poland (3.5%). In Ireland and Australia this growth continued through Q4 2020. Despite the increase in net cash transfers to households in almost all countries (Figure 1.28, Panel B), many still experienced reductions in disposable income per capita in Q2 2020. However, Chile, Austria, Sweden, Hungary and Slovenia, as well as Italy, Mexico and the Netherlands made strong progress in reversing the negative shock to household disposable per capita income experienced between Q4 2019 and Q2 2020.

These figures provide some insight into the extent to which support measures have been effective in maintaining livelihoods in the face of the COVID‑19 pandemic. However, they tell us little about how the impact of the pandemic on disposable income was distributed across the income distribution, and among certain socio‑economic groups; they tell us little about the success of government support in protecting the most vulnerable segments of the population.

Unfortunately cross-country micro data on the impact of government transfers are not yet available. Reliable indicators of economic inequality at a high frequency are lacking, with most official statistics on income inequality available only on an annual basis, and often with a long delay. Fortunately, however, recent research – see for example Aspachs et al. (2020[60]), Bick and Blandin (2021[61]), Chetty et al. (2020[62]), Cox et al. (2020[63]) and Ganong and Noel (2019,[64]) – has been able to harness big data, from private sources, to understand the rapid changes and to inform policy making in a timely manner. Such work provides tentative evidence that, while rapidly designed and implemented measures have done a remarkable job in protecting the economic well-being of households on average, the labour market characteristics of certain groups may have left them vulnerable and disproportionately exposed (see Box 1.6).

Nearly a year and a half into the COVID‑19 induced crisis, there is, once again, hope that there is light at the end of the tunnel. But even now, as economic activity resumes, labour markets across the OECD face enormous challenges. As the crisis has evolved, so those most affected by its ravages have shifted. Certain groups however – including those in low-paid occupations, the low-educated and the young – have consistently been in the eye of the storm. These groups not only saw the most substantial impact on their hours of work, but have been more likely to experience this impact through joblessness. This finding has important implications: it sheds light on the de facto targeting of employment support policies and the potential impact this has on inequalities, as well as it informs us about the likely long-term implications of the crisis on the careers of those affected.

Despite efforts to reduce the hardship of the crisis, the economic and employment effects of the COVID‑19 crisis are likely to extend well beyond the short term, into the medium and long term.

This section concludes the chapter by reviewing the available evidence about the acceleration of long-standing structural changes and their impact on the world of work. In particular, many companies facing severe containment guidelines and uncertainty may have sped up their plans to digitalise and automate production processes. This may, in turn, lead to a ‘double impact’ on vulnerable workers who have lost their jobs during the pandemic and may not be able to recover them in its aftermath due to accelerated automation and technology adoption.

Before the COVID‑19 pandemic hit countries around the globe, technological change, automation, digitalisation as well as the advent of artificial intelligence and the use of big data were already among the megatrends reshaping societies and the world of work – see e.g. OECD (2019[67]).

Despite initial fears of potential massive technological unemployment, recent evidence on the impact of automation on labour markets suggests that employment levels have been trending upwards, with the exception of the period of global financial crisis (GFC). While there is no clear-cut evidence of a negative effect of automation on employment at the aggregate level, important concerns remain as to the negative effect that technological change (including digitalisation and automation) can have on specific groups of individuals such as the low skilled or those with poor digital skills. Further evidence (Georgieff and Milanez, 2021[68]) indicates, in fact, that occupations that were at higher risk of automation in 2012 experienced lower employment growth than average, or even modest declines in employment levels in the subsequent period, up to 2019. The risk that technological change could create more inequalities is exacerbated by the fact that many of the workers employed in occupations at high-risk of automation are generally low-skilled or older workers who are less likely to engage in lifelong learning and retraining (OECD, 2019[69]; 2021[70]).

Even in the aftermath of the COVID‑19 crisis, the pace of technology adoption is expected to remain unabated or even accelerate (World Economic Forum, 2020[71])). Similarly, other megatrends such as population ageing and climate change are still expected to play a key role in shaping employment trends, boosting the demand for workers in health care or in sectors related to the green economy, in turn, likely spurring further adoption of new technologies. In order to anticipate these changes, some countries have produced employment projections (see Box 1.7 and Annex 1.D) that account for the short to long-run effects that megatrends are expected to have on jobs.

Most of the available employment projections were elaborated prior to the COVID‑19 pandemic and, therefore, account for structural factors driving employment growth and decline but not for the expected rebound (or further decline) that employment in different occupations is likely to experience in the aftermath of the COVID‑19 crisis, as economic activity is gradually restarting and vaccines are rolled out to increasingly larger shares of the population.

These employment projections, however, allow to investigate the key question as to whether jobs that before the COVID‑19 crisis were already facing a high risk of displacement due to megatrends, have also been hit particularly hard during the pandemic downturn, leading to a double negative effect on already vulnerable workers.

Recent evidence shows, for instance, that the adoption of digital technologies during the coronavirus pandemic has helped protect the jobs of millions of workers who were able to carry out their activities remotely and working from home – see OECD (2020[5]; 2021[73]) and Chapter 5. But, while many employers and employees have used digital technologies to weather the COVID‑19 crisis, others have instead been unable to do so due, among other reasons, to the lack of adequate skills or the necessary technological infrastructures in their workplace.

Mounting evidence suggests that vulnerable workers, especially those facing higher risks from automation and digitalisation, may have been particularly exposed during the COVID‑19 crisis, experiencing a more pronounced employment hit than other workers. Going forward, these same workers could experience a far weaker recovery as many of the jobs and tasks they held before the pandemic may end up being automated in the meantime. According to the UK Commission on Workers and Technology,34 for instance, approximately 61% of jobs furloughed in the first half of 2020 in the United Kingdom were in sectors where automation is most likely to lead to job losses. Similarly, a recent report from the Royal Academy of Science (Wallace-Stephens and Morgante, 2020[74]) shows a positive and significant correlation between the probability of automation and the JR scheme take up in the United Kingdom, which can be considered a rough measure of the contraction of activity in different occupations (see also Section 1.3 and Chapter 2).

To investigate the association between the impact of the COVID‑19 crisis on jobs and workers’ vulnerability to megatrends, Figures 1.32, 1.33, 1.34 and 1.35 combine information on the evolution of job postings published online during the crisis and country-specific employment projections that account for the impact that megatrends are expected to have on future labour markets.

Results suggest that, on average across the countries for which information is available, many of the jobs that were projected to decline the most due to structural changes already ongoing prior to the pandemic have also bore the brunt of the current crisis, experiencing among the strongest declines in the number of new job openings published during the pandemic. In Australia, for instance, online job postings for Secretaries and Personal assistants dropped in 2020 by 28% and 37% relative to 2019. Prior to the pandemic, employment in those occupations was already projected to decline by 30% and 9% by 2024 due to structural trends, among which technology adoption.

Similarly, in Canada, online job postings for Data entry clerks or Banking insurance and other financial clerks decreased by 39% and 31%, respectively, relative to the same period in 2019. Projections elaborated prior to the pandemic were already expecting those jobs to decline by 20% and 14% by 2028 (relative to 2019). Results are qualitatively similar for the United States, where travel agents and word processors and typists, jobs projected to decline significantly in the next decade, also experienced among the strongest drops in the number of new job postings published online during the pandemic. Data for the United Kingdom show a particularly strong and statistically significant correlation between the projected decline in employment and the drop in new job postings during the pandemic.35 In particular, occupational groups such as Secretarial and related occupations or Process, plant and machine operatives were already expected to decline by more than 15% and 7%, respectively, by 2024 but they have also been occupational groups hit particularly hard by the pandemic, experiencing a decrease in online job postings of 32% and 14%, respectively, in 2020 relative to the previous year.

Results above also show that those jobs that have performed relatively well during the COVID‑19 crisis are also among those that, prior to the pandemic, were projected to grow the most in coming years. Not surprisingly, among the occupations with the strongest increase in online job postings during the pandemic, several of those jobs are in the health care sector: Aged and disabled carers (Australia, +35%), Licensed practical nurses (Canada, +39%), Community health workers (United States, +91%) or Health professionals (United Kingdom, +25%).

While the extreme pressure on countries’ health care systems is surely at the core of the sudden increase in demand for these professionals, it is also worth noticing that those are occupations that were already projected to grow well above 10% prior to the COVID‑19 shock. In addition to health care jobs, other occupations whose employment is projected to increase significantly and that also experienced a strong demand during the crisis are in the green-economy sector, such as Solar photovoltaic installers (United States, +91%).

These results hint to the existence of an association between the vulnerability of certain workers to megatrends and the intensity with which the crisis has hit them (and their jobs) in the 2020.

Despite being suggestive, these results also show a wide heterogeneity across occupations which can be hard to disentangle with the available data. Regression analyses (see Annex 1.C), however, indicate that the association between the impact of the COVID‑19 on jobs and the projected employment trends due to megatrends remains significant also when controlling for time‑invariant occupational group effects.

While caution should be used when interpreting the results in this section – as it may be still too early to say whether the COVID‑19 crisis has indeed accelerated ongoing megatrends – there are several reasons to believe that labour markets may not go back to ‘business as usual’ after the crisis and that, instead, the effect of the COVID‑19 pandemic will extend to a more profound reorganisation of work, accelerating, among others, the adoption of new technologies, the importance of certain occupations in the labour market (in particular in the health care sector) and increasing the risk of displacement for those who were already particularly vulnerable prior to the crisis.

The uncertainty as to when human ‘manual’ labour will be fully available again, for instance, is likely to accelerate firms’ plans to adopt automation technologies earlier than expected. A recent survey of large employers run by the World Economic Forum (World Economic Forum, 2020[71]) indicates that, in addition to cloud computing, big data and e‑commerce, employers increased their interest for encryption, nonhumanoid robots and artificial intelligence – signalling that more investment is likely to go into digitalisation of processes and the deployment of automation technologies.

Similarly, new work by the Bureau of Labor Statistics in the United States (Ice, Rieley and Rinde, 2021[75]) suggests the possibility of widespread, permanent changes to consumer and firms’ behaviours. Increasing adoption of telework, even after the end of the health crisis, are expected to have both direct and spillover effects on individuals, firms and the economy through changes in the need for office space, individual choices about non-residential construction, the demand for food and accommodation and the location of retail stores, and that for information technology (IT) and computer-related occupations, particularly those involved in IT security. Public demand for better prevention, containment, and treatment of infectious diseases is also expected to lead to increased scientific and medical research funding and to a further boost to the health care sector on top of what already projected.

If confirmed, these trends would imply that workers in occupations that have been hit hard during the pandemic may struggle more than others to return to their previous job (assuming they lost it during the crisis), not only because of the layoffs in their sector (which are likely to take time to fully recover) but also because firms may use the crisis period to accelerate pre‑existing trends (automation, digitalisation as well as the boost in the demand for professionals in the health care and green sectors), restructuring profoundly the way they produce and combine human labour with new technologies.

Predicting what will happen in the near future can prove to be extremely difficult, especially as many intertwined factors are likely to play a role in the path towards recovery. With the ongoing (or even accelerated) speed of technology adoption, it is reasonable to assume that many of the workers who will be able to go back to the jobs they held before the pandemic, will still experience significant changes in the tasks they will be expected to perform in their jobs. More vulnerable workers, instead, may not even be able to re‑enter the labour market in their previous roles and will need to consider career changes as some of their jobs are expected to disappear.

In both cases, however, retraining and upskilling will be key for all workers going forward as some will need to update their skills for new tasks and others will retrain and look for new employment opportunities through career changes. The identification of targeted and responsive retraining pathways will, therefore, be key for individuals to navigate such uncertain and challenging landscape and to reduce the risk of persistent skill mismatch and under-qualification among the most vulnerable workers.

During the pandemic several countries developed short training programmes to meet the pressing demand for frontline and health care workers. In many cases, training programmes have been targeted at health and medical professionals who needed to acquire specific knowledge related to the pandemic response (OECD, 2020[76]). In other cases, the training recipients have been workers displaced by the pandemic whose skills were deemed to be relevant to fulfil roles in high-demand essential services (see Box 1.8).

Going forward, despite the uncertainty surrounding the shape of the recovery and its timing, countries should put more efforts in anticipating the potential impact that the pandemic could have in the medium to long run on a wide range of jobs and to provide support to workers who may be displaced and struggle to go back to their original occupations (see Box 1.9).

Concrete examples of these efforts are those implemented by Public Employment Services (PES) in some countries. In Ireland, for instance, the Department of Employment Affairs and Social Protection created a site to connect displaced workers from recent business closures with jobs in health care, retail, life sciences, infrastructure and IT, customer support and other sectors facing short-term staffing requirements. In France, Pôle Emploi launched an online platform to facilitate recruitment by those sectors currently in need of more labour, including agriculture, agrifood, health, transport and telecommunications. Lithuania’s PES also partnered with the massive open online course (MOOC) provider, Coursera, to provide free courses for unemployed adults during the summer and autumn of 2020. The initiative involved already thousands of unemployed adults who participated in online learning. The PES in Brussels has also developed an active campaign on its website, using its newsletter to advertise training offers (in particular basic digital skills and language training) and encourage adults to use them. Italy’s government also set up a website gathering various short courses that can help managers and employees develop the skills and competencies to telework more effectively.

Several governments also partnered with education institutions to make quick progress in delivering online learning during the pandemic. France, for instance, has launched online VET courses free of charge for a period of three months, including the core curriculum of vocational schools and main training courses for professional qualifications. Korea started a virtual training platform enabling learning providers to upload their course content while in the Netherlands, in-person VET in small groups was organised for students who do not have sufficient digital resources (OECD, 2020[77]). Going forward, similar efforts should be boosted also in other countries in order to provide displaced workers the necessary skills to remain active in high-quality jobs in the recovery phase and in the longer run.

After a significant increase at the start of the pandemic, unemployment is now retracting in many countries, although it is projected to remain above pre‑crisis rates in most of them. Yet in the context of this pandemic – and the labour market policy that has accompanied it – unemployment offers only a partial picture. In the early phases of the crisis, large numbers withdrew from the labour market because of job search restrictions and increased burden of household duties. At the same time, many of those who remained in employment saw their hours reduced, often supported by job retention schemes. In the second quarter of 2020 working time reductions among workers still in employment accounted for around 80% of the decline in hours worked.

While many of those on temporary layoffs or working reduced, or even zero, hours, have been relatively quick to return to work over the course of the past year, others are struggling to bounce back. Many young people have lost their jobs during the crisis or failed to find one upon entry in the labour market. The fall in hours worked by 15‑24 year‑olds was almost twice as large as that of prime‑aged and older workers, and hours lost through increased joblessness are likely to be more durably gone. Other vulnerable groups – such as those with a limited education and those in low-paying occupations – have also been affected more deeply. As the recovery phase progresses in the coming months and years, this unequal impact – if left unaddressed – risks translating into a more lasting increase in inequality.

As we now look towards a recovery, promoting a return to work will require supportive policy, implemented with careful timing. While loosening restrictions, countries must begin to promote a return to active job search, ensuring the re‑engagement of those who have become increasingly distant from the labour market over the past year. As job search picks up, measures to support job creation may be needed, such as carefully timed and targeted hiring subsidies, while jobseekers may need support and intelligently designed training pathways, to build on their existing skills and guide them towards emerging opportunities.

Young people will need particular attention. The ravages of COVID‑19 itself fell with particular force on the older population. And, in response, OECD countries implemented deep and wide‑ranging measures to control the virus and protect the vulnerable. The impact of these measures, however, has fallen disproportionately on the livelihoods and labour market prospects of the young. OECD countries must now, with equal purpose, develop a programme of measures to protect these young workers, with the aim of providing them with an initial foothold in the labour market, preventing scarring of young careers, and preparing them for future opportunities.

Much remains to be learned regarding who has shouldered the burden of the COVID‑19 crisis, and it is not yet clear what form the recovery will take. This chapter has made a first attempt to survey the impact on a few broad groups. However, emerging evidence suggests that, in addition to those groups, the impact has also varied depending on race, ethnicity and migrant status. There is also considerable scope to examine the impact on inequality and on livelihoods across the income distribution. The limited availability of timely micro data in many OECD countries has meant that a thorough investigation of these disparities is not yet possible. Building on the micro data as they become available, future analysis will be able to provide important insights into how these groups have fared through the crisis and how they benefited from the protective policies introduced to cushion its impact on OECD labour markets.

The wide‑ranging labour market policies introduced over the course of the past year have had a profound effect. They have protected the livelihoods of many and prevented widespread hardship. But these policies were introduced at speed, as a rapid response to the emergent structural weaknesses revealed by the crisis, and not all have benefited to an equal degree. As OECD countries now turn to navigating the recovery, addressing these structural weaknesses in a manner that is both holistic and durable will be an important priority.

Going forward, many of the workers hit hardest during the pandemic may struggle to return to their previous occupations, due to a lack of skills and as firms profoundly restructure the way they produce and combine human labour with new technologies. Targeted support in the form of upskilling and retraining should be provided to the most vulnerable to ensure that the recovery is inclusive and does not leave anyone behind.

[33] Adams-Prassl, A. et al. (2020), “Inequality in the impact of the coronavirus shock: Evidence from real time surveys,”, Journal of Public Economics, Vol. 189, https://doi.org/10.1016/j.jpubeco.2020.104245.

[82] Alon, T. et al. (2020), “This Time It’s Difference: The Role of Women’s Employment in a Pandemic Recession”, CEPR Discussion Paper 15149.

[48] Altonji, J., L. Kahn and J. Speer (2016), “Cashier or Consultant? Entry Labor Market Conditions, Field of Study, and Career Success”, Journal of Labor Economics, Vol. 34/S1, pp. S361-S401, http://dx.doi.org/10.1086/682938.

[60] Aspachs, O. et al. (2020), “Real-Time Inequality and the Welfare State in Motion: Evidence from COVID-19 in Spain”, CEPR Discussion Paper, No. 15118.

[66] Aspachs, O. et al. (2020), Tracking inequality in real-time: impact of the actvity rebound, https://inequality-tracker.caixabankresearch.com/en/.

[15] Baker, S. and A. Fradkin (2017), “The Impact of Unemployment Insuarnce on Job Search: Evidence from Google Search Data”, Review of Economics and Statistics, Vol. 99 (5), pp. 756-768, https://doi.org/10.1162/REST_a_00674.

[14] Balgova, M. et al. (2021), “Job Search during a Pandemic Recession: Survey Evidence from the Netherlands”, IZA Discussion Papers, Vol. IZA DP No. 14180, https://www.iza.org/publications/dp/14180/job-search-during-a-pandemic-recession-survey-evidence-from-the-netherlands.

[78] Bell, B., M. Codreanu and S. Machin (2020), “What can previous recessions tell us about”, A CEP Covid-19 analysis, Vol. Paper No 007, https://cep.lse.ac.uk/pubs/download/cepcovid-19-007.pdf?platform=hootsuite.

[30] Beraia, M. and C. Wolf (2021), Demand Composition and the Strength of Recoveries, https://economics.mit.edu/files/20297.

[61] Bick, A. and B. A (2021), Real Time Labor Market Estimates During the 2020 Coronavirus Outbreak, http://dx.doi.org/10.2139/ssrn.3692425.

[56] Blau, F. and L. Kahn (2017), “The Gender Wage Gap: Extent, Trends, and Sources”, Journal of Economic Literature, Vol. 55/3, pp. 789-865, https://doi.org/10.1257/jel.20160995.

[35] Blau, F., J. Koebe and P. Meyerhofer (2020), Who are the Essential and Frontline Workers?, https://ssrn.com/abstract=3687947.

[41] Blundell, J., S. Machin and M. Ventura (2020), Covid-19 and the Self-Employed: Six Months into the Crisis, https://cep.lse.ac.uk/pubs/download/cepcovid-19-012.pdf.

[54] Bredemeier, C., F. Juessen and R. Winkler (2017), “Man-cessions, fiscal policy, and the gender composition of employment”, Economics Letters, Vol. 158, pp. 73-76, http://dx.doi.org/10.1016/j.econlet.2017.06.022.

[65] CaixaBank Research (2020), CaixaBank Inequality Tracker, https://inequality-tracker.caixabankresearch.com/.

[45] Carcillo, S. et al. (2015), “NEET Youth in the Aftermath of the Crisis: Challenges and Policies”, OECD Social, Employment and Migration Working Papers, No. 164,, https://doi.org/10.1787/5js6363503f6-en.

[20] Cheng, W. et al. (2020), Back to Business and (Re)employing Workers? Labor Market Activity During State COVID-19 Reopenings, National Bureau of Economic Research, Cambridge, MA, http://dx.doi.org/10.3386/w27419.

[62] Chetty, R. et al. (2020), The Economic Impacts of COVID-19: Evidence from a New Public Database Built Using Private Sector Data, National Bureau of Economic Research, Cambridge, MA, http://dx.doi.org/10.3386/w27431.

[63] Cox, N. et al. (2020), “Initial Impacts of the Pandemic on Consumer Behavior: Evidence from Linked Income, Spending, and Savings Data”, SSRN Electronic Journal, http://dx.doi.org/10.2139/ssrn.3633008.

[59] del Boca, D. et al. (2020), “Women’s Work, Housework and Childcare, before and during COVID-19”, Review of Economics of the Household, Vol. 18, pp. 1001-1017, http://dx.doi.org/10.1007/s11150-020-09502-1.

[18] Demmou, L. et al. (2021), “Liquidity shortfalls during the Covid-19 outbreak: assessment and policy responses”, OECD Economics Department Working Papers, No. 1647, OECD Publishing, Paris, https://doi.org/10.1787/581dba7f-en.

[29] Dingel, J. and B. Neiman (2020), “How Many Jobs Can be Done at Home?”, Journal of Public Economics, Vol. 189/104235, https://doi.org/10.1016/j.jpubeco.2020.104235.

[47] Dorsett, R. and P. Lucchino (2018), “Young people’s labour market transitions: The role of early experiences,”, Labour Economics,, Vol. 54, pp. 29-46, https://doi.org/10.1016/j.labeco.2018.06.002.

[8] Eurofund (2021), European Restructuring Monitor: Restructuring Support Instruments, https://www.eurofound.europa.eu/observatories/emcc/erm/support-instrument?field_english_name_erm_si_value=temporary+layoffs&field_native_name_erm_si_value=&field_ef_involved_actors_tax_tid=All&field_phase_erm_si_value_selective=.

[79] European Commission (2020), Labour Market and Wage Developments in Europe 2020, European Commission, https://ec.europa.eu/social/main.jsp?langId=en&catId=89&furtherNews=yes&newsId=9873.

[43] European Commission JRC (2020), New Evidence on Platform Workers in Europe: Evidence from the Second COLLEEM Survey, EU Publications, https://op.europa.eu/en/publication-detail/-/publication/f7d9af38-5224-11ea-aece-01aa75ed71a1.

[7] Eurostat (2016), EU Labour Force Survey Explanatory notes,, https://ec.europa.eu/eurostat/documents/1978984/6037342/EU-LFS-explanatory-notes-from-2016-onwards.pdf.

[26] Farber, H. et al. (2018), Whom Do Employers Want? The Role of Recent Employment and Unemployment Status and Age, National Bureau of Economic Research, Cambridge, MA, http://dx.doi.org/10.3386/w24605.

[58] Farré, L. et al. (2020), “How the COVID19 Lockdown Affected Gender Inequality in Paid and Unpaid Work in Spain”, IZA Discussion Paper 13434.

[12] Forsythe, E. et al. (2020), “Searching, Recalls, and Tightness: An Interim Report on the COVID Labor Market”, National Bureau of Economic Research Working Paper 28083, http://dx.doi.org/10.3386/w28083.

[81] Fujita, S. and G. Moscarini (2017), “Unemployment and Recall”, American Economic Review, Vol. 102/12, pp. 3875-3916, http://dx.doi.org/10.1257/aer.20131496.

[22] Ganong, P. et al. (2021), “Spending and Job Search Impacts of Expanded Unemployment Benefits: Evidence from Administrative Micro Data”, BFI Working Paper, No. 2021-19, Becker Friedman Institute for Economics at the University of Chicago.

[64] Ganong, P. and P. Noel (2019,), “Consumer spending during unemployment: Positive and normative implications”, American Economic Review, Vol. 109/7, pp. 2383–2424, http://dx.doi.org/0.1257/aer.20170537.

[50] Garrouste, C. and M. Godard (2016), “The Lasting Health Impact of Leaving School in a Bad Economy: Britons in the 1970s Recession”, Health Economics, Vol. 25, pp. 70-92, http://dx.doi.org/10.1002/hec.3391.

[68] Georgieff, A. and A. Milanez (2021), “What happened to jobs at high risk of automation?”, OECD Social, Employment and Migration Working Papers, No. 255, OECD Publishing, Paris, https://dx.doi.org/10.1787/10bc97f4-en.

[19] Giupponi, G. and C. Landais (2018), “Subsidizing Labor Hoarding in Recessions: The Employment & Welfare Effects of Short Time Work”, CEP Discussion Papers, No. 1585, Centre for Economic Performance, London School of Economics.

[36] Goos, M., A. Manning and A. Salomons (2014), “Explaining Job Polarization: Routine-Biased Technological Change and Offshoring”, American Economic Review, Vol. 104(8), pp. 2509-26, http://dx.doi.org/10.1257/aer.104.8.2509.

[4] Hadjibeyli, B., G. Roulleau and A. Bauer (2021), “Live and (don’t) let die: The impact of Covid-19 and public support on French firms”, French Treasury working paper, No. 2021-2, https://www.tresor.economie.gouv.fr/Articles/2021/04/20/live-and-don-t-let-die-the-impact-of-covid-19-and-public-support-on-french-firms.

[2] Hale, T. et al. (2020), Oxford COVID-19 Government Response Tracker, http://dx.doi.org/www.bsg.ox.ac.uk/covidtracker.

[1] Hale, T. et al. (2020), Variation in Government Responses to COVID-19, http://dx.doi.org/www.bsg.ox.ac.uk/covidtracker.

[13] Hensvik, L., T. Le Barbanchon and R. Rathelot (2021), “Job search during the COVID-19 crisis”, Journal of Public Economics, Vol. 194, http://dx.doi.org/10.1016/j.jpubeco.2020.104349.

[57] Hupkau, C. and B. Petrongolo (2020), “Work, Care and Gender during the COVID‐19 Crisis”, Fiscal Studies, Vol. 41, pp. 623-651, https://doi.org/10.1111/1475-5890.12245.

[75] Ice, L., M. Rieley and S. Rinde (2021), “Employment projections in a pandemic environment”, Monthly Labour Review, Bureau of Labor and Statistics, U.S. Bureau of Labor Statistics.

[9] Istituto Nazionale di Statistica (2021), Monthly historical series, https://www.istat.it/it/files//2021/02/202012_serie-storiche.xls.

[31] Joyce, R. and X. Yu (2020), “Sector shutdowns during the coronavirus crisis: which workers are most exposed?”, IFS Briefing Note, No. BN 278, Institute of Fiscal Studies, London, https://ifs.org.uk/uploads/BN278-Sector-Shutdowns.pdf.

[21] Katz, L. and B. Meyer (1990), “Unemployment Insurance, Recall Expectations, and Unemployment Outcomes”, The Quarterly Journal of Economics, Vol. 105(4), pp. 973-1002, https://doi.org/10.2307/2937881.

[32] Kong, E. and D. Prinz (2020), “The Impact of State Shutdown Policies on Unemployment During a Pandemic”, Journal of Public Economics, Vol. 189, pp. 104-257, https://doi.org/10.1016/j.jpubeco.2020.104257.

[27] Kroft, K., F. Lange and M. Notowidigdo (2013), “Duration Dependence and Labor Market Conditions: Evidence from a Field Experiment*”, The Quarterly Journal of Economics, Vol. 128/3, pp. 1123-1167, http://dx.doi.org/10.1093/qje/qjt015.

[39] Low Pay Commission (2020), Low Pay Commission Report 2020, Crown copyright, https://www.gov.uk/government/publications/low-pay-commission-report-2020.

[16] Marinescu, I., D. Skandalis and D. Zhao (2020), “Job Search, Job Posting and Unemployment Insurance During the COVID-19 Crisis”, SSRN Electronic Journal, http://dx.doi.org/10.2139/ssrn.3664265.

[55] Ngai, L. and B. Petrongolo (2017), “Gender Gaps and the Rise of the Service Economy”, American Economic Journal: Macroeconomics, Vol. 9/4, pp. 1-44, http://dx.doi.org/10.1257/mac.20150253.

[70] OECD (2021), “Adult Learning and COVID-19: How much informal and non-formal learning are workers missing?”, OECD Policy Responses to Coronavirus (COVID-19), OECD Publishing, Paris, https://dx.doi.org/10.1787/56a96569-en.

[3] OECD (2021), “OECD Economic Outlook No 109 (Edition 2021/1)”, OECD Economic Outlook: Statistics and Projections (database), https://dx.doi.org/10.1787/4229901e-en (accessed on 28 May 2021).

[73] OECD (2021), OECD Skills Outlook 2021: Learning for Life, OECD Publishing, Paris, https://dx.doi.org/10.1787/0ae365b4-en.

[28] OECD (2021), “Risks that matter 2020: The long reach of COVID-19”, OECD Policy Responses to Coronavirus (COVID-19), OECD Publishing, Paris, https://dx.doi.org/10.1787/44932654-en.

[44] OECD (2021), “What have countries done to support young people in the COVID-19 crisis?”, OECD Policy Responses to Coronavirus (COVID-19), OECD Publishing, Paris, https://doi.org/10.1787/ac9f056c-en.

[46] OECD (2020), Education at a Glance 2020: OECD Indicators, https://doi.org/10.1787/69096873-en.

[11] OECD (2020), OECD employment and unemployment statistics during the COVID-19 crisis, OECD, Paris, https://www.oecd.org/sdd/labour-stats/OECD-employment-and-unemployment-statistics-during-the-COVID-19-crisis.pdf.

[52] OECD (2020), OECD Employment database, http://www.oecd.org/employment/emp/onlineoecdemploymentdatabase.htm.

[5] OECD (2020), OECD Employment Outlook 2020: Worker Security and the COVID-19 Crisis, OECD Publishing, Paris, https://doi.org/10.1787/1686c758-en.

[53] OECD (2020), OECD Family Database, http://www.oecd.org/els/family/database.htm.

[23] OECD (2020), OECD Risks That Matter Survey 2020, https://www.oecd.org/social/risks-that-matter.htm (accessed on  March 2021).

[10] OECD (2020), OECD unemployment rate up sharply in March 2020 with data for April pointing to an unprecedented rise, OECD, Paris, http://www.oecd.org/sdd/labour-stats/unemployment-rates-oecd-05-2020.pdf.

[76] OECD (2020), “Skill measures to mobilise the workforce during the COVID-19 crisis”, OECD Policy Responses to Coronavirus (COVID-19), OECD Publishing, Paris, https://dx.doi.org/10.1787/afd33a65-en.

[42] OECD (2020), “Supporting livelihoods during the COVID-19 crisis: Closing the gaps in safety nets”, OECD Policy Responses to Coronavirus (COVID-19), OECD Publishing, Paris, https://doi.org/10.1787/17cbb92d-en.

[77] OECD (2020), “VET in a time of crisis: Building foundations for resilient vocational education and training systems”, OECD Policy Responses to Coronavirus (COVID-19), OECD Publishing, Paris, https://dx.doi.org/10.1787/efff194c-en.

[38] OECD (2020), Who Cares? Attracting and Retaining Care Workers for the Elderly, OECD Health Policy Studies, OECD Publishing, Paris, https://doi.org/10.1787/92c0ef68-en.

[51] OECD (2020), “Women at the core of the fight against COVID-19 crisis”, OECD Policy Responses to Coronavirus (COVID-19), OECD Publishing, Paris, https://doi.org/10.1787/553a8269-en.

[37] OECD (2020), “Workforce and safety in long-term care during the COVID-19 pandemic”, OECD Policy Responses to Coronavirus (COVID-19), OECD Publishing, Paris, https://doi.org/10.1787/43fc5d50-en.

[67] OECD (2019), OECD Employment Outlook 2019: The Future of Work, OECD Publishing, Paris, https://dx.doi.org/10.1787/9ee00155-en.

[69] OECD (2019), OECD Skills Outlook 2019 : Thriving in a Digital World, OECD Publishing, Paris, https://dx.doi.org/10.1787/df80bc12-en.

[24] OECD (2018), OECD Employment Outlook 2018, OECD Publishing, Paris, https://dx.doi.org/10.1787/empl_outlook-2018-en.

[72] OECD (2017), Getting Skills Right: Skills for Jobs Indicators, Getting Skills Right, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264277878-en.

[49] Oreopoulos, P., T. von Wachter and A. Heisz (2012), “The Short- and Long-Term Career Effects of Graduating in a Recession”, American Economic Journal: Applied Economics, Vol. 4/1, pp. 1-29, http://dx.doi.org/10.1257/app.4.1.1.

[40] Pfann, G. and F. Palm (1993), “Asymmetric Adjustment Costs in Non-linear Labour Demand Models for the Netherlands and U.K. Manufacturing Sectors”, The Review of Economic Studies, Vol. 60(2), pp. 397–412, https://doi.org/10.2307/2298064.

[25] Pissarides, C. (1992), “Loss of Skill During Unemployment and the Persistence of Employment Shocks”, The Quarterly Journal of Economics, Vol. 107/4, pp. 1371-1391, http://dx.doi.org/10.2307/2118392.

[34] Song, H. et al. (2021), The Impact of the Non-essential Business Closure Policy on Covid-19 Infection Rates, National Bureau of Economic Research, Cambridge, MA, http://dx.doi.org/10.3386/w28374.

[17] UNESCO (2021), Global Monitoring of School Closures caused by COVID-19, https://en.unesco.org/covid19/educationresponse#schoolclosures (accessed on 10 May  2021).

[6] United Kingdom Office of National Statistics (2020), Coronavirus and its impact on the Labour Force Survey, https://www.ons.gov.uk/employmentandlabourmarket/peopleinwork/employmentandemployeetypes/articles/coronavirusanditsimpactonthelabourforcesurvey/2020-10-13.

[80] van den Berge, W. (2018), “Bad start, bad match? The early career effects of graduating in a recession for vocational and academic graduates”, Labour Economics, Vol. 53, pp. 75-96, http://dx.doi.org/10.1016/j.labeco.2018.05.011.

[74] Wallace-Stephens, F. and E. Morgante (2020), Who is at risk? Work and automation in the time of COVID-19, The royal society for arts, manufactures and commerce, https://www.thersa.org/reports/work-automation-covid.

[71] World Economic Forum (2020), The Future of Jobs Report 2020, https://www.weforum.org/reports/the-future-of-jobs-report-2020.

Let t denote time, e at-work employed workers, j jobless workers (inactive plus unemployed), o 0‑hour employees, H total hours worked, N number of people and h(=H/N) hours per at-work worker.

The change in hours worked between t and t+1 can be decomposed into the contribution of hours per at-work employed worker (intensive margin) and number of at-work employed workers (extensive margin) as follows:

${∆H}_{t+1}=H_{e,t+1}-H_{e,t}={h_{e,t+1}N}_{e,t+1}-{h_{e,t}N}_{e,t}=$

$=\left({h_{e,t+1}N}_{e,t+1}-{h_{e,t}N}_{e,t+1}\right)+\left({h_{e,t}N}_{e,t+1}-{h_{e,t}N}_{e,t}\right)=∆{h_{e,t+1}N}_{e,t+1}+{h_{e,t}∆N}_{e,t+1}$

Taking into account that $∆N_{e,t+1}=-(∆N_{j,t+1}+∆N_{o,t+1}-{∆N}_{p,t+1})$, where $p$ denotes the relevant population, the above expression can be further decomposed as:

$∆H_{t+1}=\underset{\begin{array}{c}Contr. of hours per \\ at-work employee\end{array}}{\underbrace{∆{h_{e,t+1}N}_{e,t+1}}}-\underset{\begin{array}{c}Contr. of 0 hours\\ employment\end{array}}{\underbrace{{h_{e,t}∆N}_{o,t+1}}}-\underset{\begin{array}{c}Contr. of \\ joblessness (net of\\ population changes)\end{array}}{\underbrace{h_{e,t}({∆N}_{j,t+1}-{∆N}_{p,t+1})}}$

That is, the change in hours can be decomposed in the contribution in the change in the average hours worked for at-work employees, the net change in the level of 0‑hour employees and the net change in the level of jobless individuals (inactive and unemployed), net of population changes.

Correlations presented in Figures 1.32, 1.33, 1.34 and 1.35 hint to the existence of an association between the strength by which the COVID‑19 crisis hit jobs in 2020 and the employment projections (due to megatrends) for those occupations going forward. Results are suggestive that jobs that have been hit particularly hard during the pandemic were also already projected to decline substantially in the future. Results based on the correlation, however, still highlight a substantial deal of occupation-heterogeneity. The reasons behind this heterogeneity can be multiple. During the pandemic, for instance, certain jobs (and sectors) may have been disproportionally hit due to imposed lockdowns while others may have been thriving due to a sudden increase in demand, that little have to do with the impact of megatrends on employment. In order to account in part for such heterogeneity at the occupation level, regression analysis in Annex Table 1.C.1 estimates the relationship between the growth in online job postings and the growth in employment projections controlling for occupational dummies (at 2 digit level) in separate OLS regressions. In the case of the United States, additional controls at the occupation level are also available so that results also account for occupational skill and educational heterogeneity, experience and on-the‑job training differences across occupations (see note to Annex Table 1.C.1). Results of the OLS regression broadly confirm that the association between the growth/decline in job postings and future employment trends remains statistically significant even after accounting for occupational heterogeneity and other controls but small sample size (especially in the case of Canada) calls for caution when interpreting and generalising these results.

This chapter makes use of available country-specific employment projections. Data sources and a selection of results are provided below.

Australia: The employment projections presented in Section 1.4 are based on detailed data from the Australian Bureau of Statistics Labour Force Survey. The projections have been derived from time series models that summarise the information that is in a time series and convert it into a forecast. The projections are made by combining forecasts from autoregressive integrated moving average (ARIMA) and exponential smoothing with damped trend (ESWDT) models, with some adjustments made to take account of research undertaken by the National Skills Commission and known future industry developments.

Canada: The projections presented in Section 1.4 draw from the current Canadian Occupational Projection System (COPS) analysis completed in 2019, before the 2020 COVID‑19 outbreak. Employment projections by occupation are first calculated at the industrial level, by multiplying total employment projected in a given industry times the projected employment share of the occupation in the industry. The result can then be summed up across all industries to produce the total employment projection for each occupation. Employment projections by industry are derived from the macroeconomic and industrial outlook (including GDP and productivity projections), while the projected shares of occupational employment by industry are derived from historical trends and other assumptions (including output gap). The projections were developed for 42 industrial groupings that cover the entire economy (based on the North American Industry Classification System – NAICS) and 293 occupational groupings that cover the entire workforce (based on the National Occupational Classification – NOC).

United Kingdom: The projections presented in Section 1.4 draw from the work of the UK Commission for Employment and Skills (UKCES) and the Warwick Institute for Employment Research / Cambridge Econometrics. Projections are calculated from a number of different data sources, using a variety of econometric and statistical techniques. For further details, see the Working Futures Technical Report (available at https://www.gov.uk/government/publications/uk-labour-market-projections-2014-to-2024).

United States: The projections presented in this section draw from the US National Employment Matrix database produced by the Bureau of Labor Statistics (BLS). The matrix displays data on base‑ and projected-year employment and employment change. BLS produces occupational employment projections by analysing current and projected future staffing patterns (the distribution of occupations within an industry) in an industry – occupation matrix. Changes in the staffing pattern for each industry are projected and applied to the final industry projections, yielding detailed occupational projections by industry. This projected employment matrix includes estimates for 790 occupations across 295 industries. The Occupational Projections Data database displays data on employment, employment change, occupational openings, education, training, and wages for each detailed National Employment Matrix occupation.