key: cord-0927501-ab7vx51i
authors: Verick, Sher; Schmidt‐Klau, Dorothea; Lee, Sangheon
title: Is this time “really” different? How the impact of the COVID‐19 crisis on labour markets contrasts to the global financial crisis of 2008‐9
date: 2021-08-11
journal: Int Labour Rev
DOI: 10.1111/ilr.12230
sha: ee6c16b90e28acfe3d875c4c919d24b3e57d4b49
doc_id: 927501
cord_uid: ab7vx51i

The COVID‐19 pandemic has resulted in a more severe labour market crisis in 2020 than witnessed during the 2009 global financial crisis. Reflecting the effects of lockdown measures, which has been the main cause of damage to labour markets, the deepest impacts have been found in middle‐income economies, while certain sectors, such as accommodation and food services, and groups, especially young women, have been more negatively affected over 2020. Contrary to the adjustment process during previous shocks, the COVID‐19 crisis has resulted in a greater rise in inactivity than unemployment. Policy support needs to be maintained to avoid an unequal recovery.

Following the failure of Lehman Brothers in 2008 and the ensuing credit crunch, a financial crisis in the US became one of global dimensions, shifting to the real economy and sending shockwaves reverberating around the world. The global financial crisis (GFC) of 2009 marked the first year that the world economy contracted since World War II, a sharp correction on the boom years of [2002] [2003] [2004] [2005] [2006] [2007] . Global trade fell by 11 per cent in that year 1 , far more than the decline in output. Overall, the production, trade and consumption of goods, especially manufacturing goods, declined more during the global financial crisis than in the case of services. Due to the bursting of housing bubbles and the credit crunch, the construction sector was also badly hit, which resulted in a large number of job losses, especially among low-skilled and young workers.

The GFC was precipitated by not only the build-up of risk taking (i.e. sub-prime lending in the US and their subsequent repackaging in collateralized debt obligations masquerading as low-risk and high quality assets), but also lax financial regulation, loose monetary policy, and the collapse of real estate bubbles in a number of countries (e.g. Ireland, Spain, and the US), among other factors. 2 The years leading up to the GFC were also characterized by rising inequality in many countries, which was further exacerbated by gaps in governance and policy frameworks and made economies and labour markets vulnerable to such a crisis. 3 Though financial distress was evident already in the US in 2007, few economists predicted that this would result in a global downturn with most commentators expecting the golden years of the 2000s, which was a period of synchronized economic growth across a majority of countries, to continue. 4 Policymakers responded to the GFC by relying on a range of countercyclical macroeconomic tools, including cutting interest rates to encourage borrowing and investment, bailouts and injections of money into the financial system to address the credit crunch, and fiscal stimulus packages to support aggregate demand (Verick and Islam 2010) . The OECD estimates that the fiscal support announced during 2008-10 represented around 3.5 per cent of 2008 GDP in advanced economies with available data (OECD 2009) . The stimulus in the G20 grouping of countries amounted to around $692 billion in 2009 or approximately 1.4 percent of their combined GDP. 5 In terms of policy support in the wake of the GFC, debates on the size of fiscal multipliers and the relative effects of fiscal consolidation dominated the discourse over the years following the GFC, particularly in the context of the European Sovereign Debt Crisis. 6 The notion of "expansionary fiscal consolidation" was seen as a justification for austerity but has since been widely criticized. 7

Despite the moniker, "global", the GFC had different implications for economies and labour markets around the world. In fact, the crisis in 2009 had a greater impact on advanced economies, which experienced a decline in GDP of 3.3 per cent in that year compared with a positive 2.8 per cent growth for emerging market and developing countries (though these aggregate figures mask considerable variation within these groupings). 8, 9 Developing countries (separate from the emerging market economies) did suffer a growth slowdown in 2009, but still managed to register a relatively robust growth rate of 5.1 per cent on average 10 , reflecting that these economies were not as well integrated in the global economy and were, therefore, less affected by trade and credit shocks. Emerging economies, especially China and India, returned to strong economic growth in 2010, which generated strong demand for commodities around the world that, in turn, benefited exporters, especially developing countries, over the following years.

Although the global unemployment rate did decline slowly over the years following the GFC, it did not return to the pre-GFC low of 5.3 in the decade following the 2009 crisis. For youth, the situation was even more devastating: the global youth unemployment rate continued to rise after the GFC, reaching a peak of 13.8 per cent in 2016 (14.1 per cent for young men). Even in 2019, the youth unemployment rate still stood at 13.6 per cent, around 1.3 points higher than the rate witnessed in 2008, which reflects the long-lasting damage to labour markets, especially for young people. Though a lag in the labour market is consistent with previous financial crises (Reinhart and Rogoff 2009) , poor policy choices, namely the return to premature austerity in a number of advanced economies (e.g. in the European Union), prolonged the recovery. 11 Overall, output losses persisted following the GFC, while investment and total factor productivity remained below pre-crisis levels in a number of economies (Chen et al. 2019) .

In contrast to the GFC, the COVID-19 crisis of 2020 is a very different shock in terms of its origin, transmission channels and impact. Although there were concerns of a global pandemic prior to 2020, which were influenced by the experiences of SARS and MERS, the COVID-19 pandemic was, nonetheless, unexpected and not explicitly linked to any other warning signs (economic or otherwise). Starting as a global health emergency in early 2020, the COVID-19 crisis evolved into a severe economic and labour market shock over the year, though this evolution has been complex and uncertain due to the changing nature of the pandemic. 6 See, for example, https://voxeu.org/article/fiscal-multipliers-during-european-sovereign-debt-crisis. 7 See Chowdhury and Islam (2012) for a critique. 8 In this paper, different terms are used to describe groupings of countries. Low-, middle-and high-income status are provided by the World Bank, while the IMF uses the groupings: advanced economies (i.e. highincome), emerging markets and middle-income countries, and low-income developing countries. According to the World Bank classifications, the cut-offs for countries by income levels are (GNI per capita in current USD, using the Atlas method exchange rates): low income: < 1,036; Lower-middle income: 1,036 -4,045; Uppermiddle income: 4,046 -12,535; High income: > 12,535. 9 Source: IMF World Economic Outlook database, April 2021, https://www.imf.org/en/Publications/WEO/weodatabase/2021/April 10 Source: https://databank.worldbank.org/source/world-development-indicators 11 See, for example, House et al. (2020) .

The economic impact of the crisis around the world has been driven mostly by the lockdown and other containment measures that have been necessary to control the spread of the pandemic. Since March 2020, the vast majority of workers have been affected by workplace closures though this has varied both across countries and time. At the end of August, the share of workers residing in countries with workplace closures stood at 94 per cent (ILO 2020a). In spite of some relief in the middle of the year, a number of economies have had to reinstate measures in the last quarter of 2020 as the number of cases increased again, especially in the northern hemisphere (with the onset of winter and the emergence of a second wave of infections). Even in January 2021, 93 per cent of workers were still affected by some form of workplace closures, though the measures have become more geographically targeted and sector specific (ILO 2021a). Border closures and restrictions on travel brought global (and, in many cases, domestic) tourism to a halt in 2020: according to the UN World Tourism Organization, international arrivals fell by 72 per cent in the first 10 months of the year 12 , which has severely affected tourism-dependent economies.

These lockdown and other containment measures have subsequently precipitated a sharp decline in both demand (consumption and investment) and the supply of both goods and services over 2020, especially in the second quarter. The supply shock has been driven by the lockdown and containment measures, which have subsequently prevented businesses from operating, particularly in sectors that require face-to-face interaction with customers (e.g. retail, accommodation and food services). This has led to a decline in demand for goods and services produced by other sectors (through forward and backward linkages). Disruptions to global supply chains have also affected supply in a number of sectors, though production and trade in goods managed to rebound in late 2020 and during the first half of 2021. On the demand-side of the economy, containment measures have significantly curtailed consumption of many goods and services, especially in restaurants, recreation places, and retail outlets, etc. The demand shock has been amplified by job and incomes losses, which further dampened consumption. Finally, the level of uncertainty has been very high, which would have had a further negative effect on consumption and investment decisions. 13 Thus, even in countries with a lower number of COVID-19 cases, an economic and jobs crisis ensued in 2020 because of the measures that closed borders, workplaces, places of retail and leisure, and other interventions that severely curtailed economic activity. As a result, the effect of the COVID-19 crisis on output has been severe and rapid in the vast majority of countries with the steepest decline in output and working hours evident in the second quarter of 2020. The latest IMF estimates indicate a global decline in economic output of -3.3 per cent in 2020 (though revised upwards from the estimates released earlier in 2020) with an estimated growth rate of -4.8 per cent in advanced economies and -2.2 per cent in emerging market and developing countries ( figure 1 (a) ). These estimates also suggest a sharp recovery in 2021, though the trajectory remains uncertain and fragile. This outcome stands in contrast to the global financial crisis, which did lead to a world recession (-0.1 GDP growth rate in 2009) but did not result in negative economic growth in most low-and middle-income countries. Overall, GDP growth was negative in 2009 in 92 countries or 47.9 per cent of the sample available for that year ( figure 1(b) ). In comparison, the 2020 GDP growth rate is estimated to be negative in 164 countries or 85.4 per cent of the sample. Consequently, only a handful of countries grew faster (or, rather, less negatively) in 2020 than they did in 2009 as illustrated by the number of observations in figure 1 (b) that are to the left of the 45-degree line. This illustrates that, for low-and middle-income countries, the decline in output in 2020 is far worse than that witnessed in 2009, which, in some cases, was even quite robust (a GDP growth rate of between 5 and 10 per cent). These economies are now experiencing their first decline in output for many decades, which will have negative implications for poverty, wiping some of the gains made in recent times. In 2020, only a handful of developing and emerging economies are estimated to have grown, including China, which registered a GDP growth rate of 1.9 per cent.

In general, jobs and livelihoods are always vulnerable during such downturns, especially for youth, which is confirmed by both the GFC and the COVID-19 crisis, along with other crises (e.g. East Asian Financial Crisis). However, there have been marked differences in the COVID-19 crisis due to the specific nature of the shock as outlined above. A key distinguishing feature of the current downturn is that, due to the impact of lockdown and other containment measures, the service sector has suffered much more than during the global financial crisis, which was characterized by greater losses in manufacturing (and, to some extent, in construction). This sectoral aspect has meant that youth and women who are overrepresented in the service sector, have, in general, been more vulnerable to the employment impact of the COVID-19 crisis. This differs to the initial gender effects of the GFC on the labour market. In addition, the much more widespread economic shock to low-and middleincome countries has translated into considerable losses in their labour markets in 2020, which was not the case in 2009.

In response to the impact of the COVID-19 crisis, policymakers, including in emerging and developing economies, have responded with often unprecedented, monetary and fiscal policy interventions seeking to stop economies from collapsing. While the use of macroeconomic tools in response to the COVID-19 crisis has built on the experiences of the GFC, the approach to stimulus has been less constrained by concerns that were evident in 2009 and 2010, though the limits of such policies are being tested (e.g. the zero lower bound for interest rates in advanced economies, fiscal space in developing countries, etc.). This time around there has been less concern of debt levels, at least in advanced economies where borrowing is supported by historically low interest rates, which are expected to continue for some time to come. Significant resources have been allocated to job retention schemes in 2020, exceeding the allocations in 2009, which have kept millions of workers in jobs, including through new and innovative schemes outside advanced economies. Social protection schemes have been extended in duration and coverage, including, in some cases, the self-employed, platform workers and others not typically covered by existing measures.

Fiscal policy measures amounted to US$16 trillion (as at 17 March 2021), representing around 18 per cent of 2019 GDP, 14 which has been distributed across additional spending (and foregone revenue) (US$10 trillion) and government loans, guarantees, and capital injections ($6 trillion). However, this spending is not evenly distributed. The bulk of the spending has occurred in advanced economies, which account for more than $9.02 trillion of the additional spending and foregone revenue, along with almost all of the government loans, guarantees and capital injections. As a percentage of GDP, fiscal support in advanced economies represented 16.4 per cent of GDP, compared with just 4.2 per cent in emerging markets and 1.7 per cent in low-income developing countries (IMF 2021) . This disparity in policy support has not only impacted the ability of countries to respond to the immediate phase of the crisis (i.e. dealing with the health dimensions along with preventing the economy from collapsing) but also limits their capacity to steer the country towards recovery.

Against this backdrop, this paper focuses on the labour market impact of the COVID-19 crisis, based on available models and labour force survey data, with a particular focus on changes to employment, including differences by sectors, gender and age. Global (modelled estimates) and country-level data (from labour surveys) for 2020 are compared with estimates for 2009 where possible to illustrate both the similarities and differences between the GFC and the COVID-19 crisis.

The remainder of this paper is structured as follows: section 2 provides an overview of the aggregate labour market impact of the global financial crisis and COVID-19 crisis drawing on cross-country analysis, while section 3 turns to potential explanations for the variation in the decline in employment across countries. Section 4 highlights that most of the decline in employment during the COVID-19 crisis has translated into rising inactivity. Sections 5 and 6 investigates what has happened within the labour market by focusing on the sectoral impact along with the differences by age and gender, while section 7 concludes.

An economic shock will lead to an adjustment of labour demand through both internal and external channels, including by reducing hours worked, number of workers and wages/nonwage benefits. The adjustment in the labour market will, ultimately, depend on not only the nature of the shock, but also the effects of policies and institutions, which impact demand, including subsidies, such as job retention schemes that have played an important role in both crises in keeping workers in jobs, and employment protection legislation, which regulate how employers can adjust employment.

Due to the unique nature of the COVID-19 crisis and the suspension of economic activity during the lockdowns, working-hour losses has been an important summary indicator of the overall labour market impact of the pandemic, which captures both the internal adjustment that happens during a crisis (i.e. the reduction in working hours for those who remain employed) and the employment losses that result from the shock. The latest ILO estimates released on 25 January 2021 show that global working-hour losses in 2020 reached 8.8 per cent (compared with the fourth quarter of 2019), equivalent to 255 million full-time jobs (assuming a 48-hour working week) (ILO 2021a). Working-hour losses were highest in Latin America and the Caribbean, Southern Europe and Southern Asia. Overall, the reduction in working hours in 2020 was around four times greater than during the GFC.

In terms of the employment losses, ILO's estimates indicate that global employment declined by 114 million in 2020 or by 3.5 per cent (figure 2), which contrasts with the overall increase in global employment in 2009 (+0.4 per cent). The decline in employment in 2020 was, in particular, much more notable in low-and middle-income countries with a growth rate of -3.8 per cent compared with a positive employment growth rate in 2009 (+0.9 per cent). Employment in advanced economies experienced a decline in both the GFC and during the COVID-19 crisis but the drop in 2020 (-3.0 per cent) was significantly higher than in 2009 (-1.7 per cent). The falls in employment are largely in line with the declines in economic output (i.e. a similar employment elasticity), apart from the case of emerging and developing economies where the change in employment in 2020 is greater than for output (-3.8 per cent versus -2.2 per cent). Looking at a sample of countries with annual data for 2020, employment fell in all highincome countries, typically between 0 and -2 per cent, apart from Cyprus, Luxembourg, Macau, China, Malta, and New Zealand (figure 3 (a)). In advanced economies, the largest falls were witnessed in Chile, United States and Canada, which are characterized by more flexible labour markets (see also below for a discussion on the role of employment protection legislation in affecting labour market adjustment). In general, the decline in employment was much more substantial in middle-income countries, especially in Latin America, along with Montenegro and South Africa where the collapse in employment reached 8 per cent or more (figure 3 (b)). The decline in employment was more pronounced in middle-income economies characterized by higher levels of inequality. In fact, there is a strong negative relationship between employment growth (2020) and inequality, as measured by the Gini coefficient (figure 4). While this correlation masks complex interactions between economic, health and employment outcomes, it suggests that the COVID-19 crisis is likely to exacerbate pre-crisis inequalities. While up-to-date and representative labour force data is, in general, not available for lowincome countries, ad-hoc surveys and studies have shown that these vulnerable economies have also suffered from lockdown and containment measures. In particular, the impact of the crisis has been detrimental more broadly for incomes and livelihoods in developing countries (both low and middle-income) and not just their employment status. For example, based on a telephonic survey of selected countries carried out in April-May 2020, Bundervoet et al. (2021) find that 57.6 per cent of households in least-developed countries have experienced a decrease in total income. ILO's estimates indicate that an additional 108 million workers have become extremely or moderately poor (living on less than US$3.20 per day in PPP terms (ILO 2021b) . This has wiped out five years of progress in reducing working poverty rates (i.e. the rate has returned to the level witnessed in 2015).

In 2009, employment losses were driven by a demand shock and subsequent decline in output, which had been precipitated by a financial crisis and ensuing credit crunch. Under very different circumstances, the COVID-19 crisis has been caused by a parallel demand and supply-side shock, which hit economies and labour markets hard, especially as a result of the lockdown and other containment measures that severely curtailed economic activity. This section investigates how much the lockdown and other containment measures explain the decline in employment evident in the second quarter of 2020 (which constitutes the largest sample available to date), alongside other potential explanatory variables for which data is available.

In the context of typical crises, the standard approach to understanding how the labour market adjusts in response to an economic shock would be to look at adjustment over the Fitted line business cycle. As an input in the production of goods and services, labour demand will shift over the business cycle in line with economic conditions, declining during a recession, which has been precipitated, for example, by a financial crisis, and increasing in the recovery phase. Okun's law, which describes the negative correlation between GDP growth and unemployment, is the best-known empirical approach to analysing this relationship. 15 To the extent that policy affects output, such as fiscal stimulus packages, these interventions will also have implications for the adjustment of employment.

As noted above, labour demand is also affected by labour market institutions and policies that intermediate how output losses and gains affect the quantity and quality of employment. Labour laws, such as employment protection legislation (EPL) that determine how and when workers can be fired, are one such institution that can affect the speed of adjustment in response to an economic shock (Cazes et al. 2013) . For example, EPL is many countries regulates the conditions under which employers are able to fire workers. Moreover, in the case of mass layoffs, which occur during economic recessions, enterprises are often required to notify the local labour department/employment services. In addition to EPL, other labour market policies can impact decisions in an enterprises, most notably access to shorttime work and wage subsidies, which have been used extensively during the GFC and even more so in the COVID-19 crisis.

Using the sample of countries with annual data for 2020, it is possible to explore correlations between a number of explanatory variables and the employment growth rate (table 1). The first regression on the real GDP growth rate shows, as expected, a positive and statistically significant relationship with the employment growth rate in 2020 (column (1), table 1). The coefficient estimates indicates that a 10-percentage point decline in the real GDP in 2020 is associated, on average, with a 2.8 per cent decline in employment. However, this positive correlation becomes statistically insignificant once the average Oxford Stringency Index (over 2020) is included as an explanatory variable (columns (2) and (3), table 1). The Oxford Stringency Index is a composite measure based on nine response indicators including school closures, workplace closures, and travel bans, rescaled to a value from 0 to 100 (100 = strictest), and, therefore, captures the overall impact of lockdown and other containment measures. The weakening of the relationship between employment and real GDP growth is not surprising since the strictness of the lockdown measures is a key determinant of the GDP growth rate in 2020 in the first place. Overall, there is a strong and negative relationship between the employment decline in 2020 and the average stringency of lockdowns (see also figure 5). Note: Sample consists of 54 high-and middle-income countries. Oxford COVID-19 Government Response Stringency Index, 2020 is a composite measure based on nine response indicators including school closures, workplace closures, and travel bans, rescaled to a value from 0 to 100 (100 = strictest). If policies vary at the subnational level, the index is shown as the response level of the strictest sub-region. The average Index value for 2020 is used. Employment growth rate is the percentage change in the employment level from 2019 to 2020.

In terms of other explanatory variables, there is no evidence in this sample of a statistical relationship between employment growth in 2020 and fiscal support (above-the-line items -additional spending and foregone revenue) (column (4), table 1). This contrasts with an analysis presented in the 6 th ILO Monitor, which indicated a positive and statistically significant relationship between working-hour losses and fiscal stimulus (ILO 2020a). However, care is needed in interpreting such relationships since the effectiveness of such policy support will need to be evaluated not only on the basis of data for 2020 but also 2021 and beyond. For this reason, more research on this topic is needed, also at the country level to identify the impact of such policies.

At the same time, there is a positive correlation (though only statistically significant at the 10 per cent level) between the employment growth rate and the strictness of the employment protection legislation as measured by the OECD index of employment protection covering individual and collective dismissals (regular contracts). In line with previous crises, the latter implies that employment adjustment is more rapid (and negative) in countries with weaker employment protection legislation, i.e. where it is easier to fire workers. EPL smooths adjustment over the business cycle -that is, these protections lead to not only smaller reductions in employment following a negative shock but also lower employment growth during the recovery. Moreover, as discussed in Lee et al. (2020) and Cazes et al. (2013) , there is empirical evidence that adjustment is asymmetric insofar that, due to labour market institutions (i.e. EPL), reductions in employment during downturns are smaller than employment growth during recoveries.

Building on the policy lessons from the GFC, many governments relied heavily on job retention schemes in 2020 to keep workers in jobs -these measures include both short-time work arrangements and wage subsidies. At the peak of the lockdowns in May last year, these schemes were supporting 50 million workers in OECD countries, which is 10 times greater than the case of the GFC (OECD 2020). The most well-known programmes is the German Kurzarbeit, which successfully helped the manufacturing sector in Germany cope with the GFC. This scheme was simplified and extended (to agency workers and in terms of duration, which is now currently at 24 months) during the COVID-19 crisis. Consequently, the number of workers covered by Kurzarbeit reached around 5.7 million in April 2020, up from just under 45,000 in the year before (and a peak of 1.4 million in 2009 during the GFC). The latest estimates suggest that 2.6 million German workers are still supported by Kurzarbeit in March 2021. 16 Did the surge in these subsidies make a difference? ILO estimates for the European Union (27 countries) shows that working-hour losses reached 8.3 per cent in 2020 though employment fell by a much lower rate (-2 per cent), reflecting that labour market adjustment in these countries was largely achieved through a reduction in working hours, which was supported by job retention schemes. Turning to a sample of 20 OECD countries with data on these measures there is evidence of a positive and significant relationship between approved applications to job retention schemes as a share of dependent employees and the employment growth rate in 2020 (figure 6). The positive relationship is evident for countries, such as New Zealand, which didn't have a scheme prior to the crisis. The US is reflected at the other end of the spectrum. However, in the case of the US, workers who are temporarily laid off (i.e. furloughed) are counted as unemployed (on temporary layoff). In May 2020, there were 15.3 million American workers in this category, which has since fallen to 1.8 million in May of this year. 17

Source: Approved applications to job retention schemes as a share of dependent employees -OECD (2020). Employment growth rate -authors' calculations, labour force surveys, ILOSTAT, accessed 17 June 2021.

A decline in employment during a recession, such as that highlighted in figure 3, would normally translate into an increase in unemployment as evident following the GFC when the global youth unemployment rate went up from 12.4 per cent in 2008 to 13.4 per cent in 2009. Either as a new labour market entrant or as a person who has recently lost their job or livelihood, jobseekers will undertake job search -the criteria to be considered as unemployed -unless they believe the benefits of search are less than the costs (or if job search is constrained for other reasons).

During COVID-19 crisis, the shift from employment to non-employment has been very different due to the unique nature of the downturn. In particular, the lockdown and other containment measures have meant that the jobless have been restricted in their ability to undertake job search due to workplace closures and social distancing requirements. Moreover, the same measures have prevented people from being available to take up work. Other explanations, which deserve further investigation, including the effects of the unequal childcare burden on women, which increased their inactivity (and, hence, prevented them from being available for work and undertaking job search) and the impact of voluntary social distancing. Overall, global employment losses (114 million) have translated into a rise in inactivity of 81 million versus an increase in unemployment of 33 million (ILO 2021a).

A simple decomposition of aggregate labour market outcomes 18 shows how the decline in employment in 2020 at the country level has been accompanied by a larger increase in 18 Separating changes between unemployment and inactivity is based on a simple decomposition, Working-age population = Employment + Unemployment + Inactivity, which can be transformed into, −∆ � � = inactivity than in unemployment (the latter requires jobless to be actively searching and available for work, as noted above) in 31 high-income countries or 57.4 per cent of the sample with annual data at the time of analysis (table 2). The rise in unemployment in these advanced economies was significantly higher in such labour markets as Canada, Latvia, Lithuania and the United States. In the sample of middle-income countries, the tendency for a greater increase in inactivity in 2020 has been much stronger with 15 or 83.3 per cent of cases showing this trend. In some countries, such as Brazil, Montenegro, South Africa and Turkey, the decrease in employment has resulted in a much stronger rise in inactivity than unemployment (where the latter has even declined). In comparison to the COVID-19 crisis, the decrease in employment during the GFC, which was most pronounced in high-income countries, translated into a greater increase in unemployment than inactivity (figure 7). ILO modelled estimates show that, in the case of highincome economies, the decline in the employment-to-population ratio in 2009 (over the previous year) is estimated at 1.5 percentage points, which resulted in an increase in the unemployment-to-population ratio 19 by 1.3 points, while the inactivity rate rose by just 0.2 points 20 . The global estimates for 2020 reflect the trends highlighted in table 2: the fall in the employment-to-population ratio has resulted in a much greater increase in the inactivity rate in developing countries. In the case of middle-income countries, the employment-to-population ratio fell by 2.9 percentage points in 2020, leading to an increase in the unemployment ratio by 0.4 points and in the inactivity rate by 2.5 points. 

Based on the available data at the end of March 2020, the ILO identified four sectors as being at high risk of a substantial impact on the labour market: accommodation and food service activities; real estate; business and administrative activities; manufacturing; and wholesale and retail trade (ILO 2020b). At the same time, it was expected that education, human health and social work, public administration, and utilities would be less susceptible to the initial impact due to the nature of the shock and the sector (e.g. more protected jobs, etc.). 19 The unemployment ratio is equal to the number of young people unemployed divided by the youth population (aged 15 to 24), while the unemployment rate is the number of young people unemployed divided by the youth labour force (=employment + unemployment). 20 The rise in inactivity in 2009 reflects also longer-term trends -the increase in 2009 was, in fact, similar to the rise witnessed in 2008. At the same time, the increase in 2020 was far greater than had been seen in the years prior to the COVID-19 crisis (i.e. an increase of 2.2 percentage points in 2020 over 2019 versus no change in the inactivity rate from 2018 to 2019). Drawing on ILO modelled estimates for 2020 made available at the end of last year, the decline in employment at the sectoral level (ISIC classification, revision 4) reveals both strong consistencies with the predictions made in April 2020 and some variations (table 3) . In particular, the largest declines in employment are evident in sectors classified as at high-risk, namely accommodation and food services (-9.4 per cent) and manufacturing (-7.9 per cent). Employment declines were also significant in the cases of construction (-7.6 per cent) and other services (-6.2 per cent), which includes arts, entertainment and recreation and activities of households as employers (including domestic work). In contrast, employment in the public sector and financial and insurance activities have both experienced positive growth in 2020, the only two sectors to sdo so.

These sectoral employment growth rates diverge from the global estimates for 2009 (column 4, table 3). Firstly, global employment in all sectors grew in 2009 apart from manufacturing (decline of 1.3 per cent) and agriculture (-0.9 per cent). However, the decline in agriculture is part of a longer-term trend linked to structural transformation (i.e. the shift of workers out of agriculture) rather than the impact per se of the GFC. In fact, employment in the agricultural sector declined by a similar amount in 2008 (and continued to do so over period following the GFC). Thus, as noted above, manufacturing was, in general, the hardest hit sector in 2009 in terms of global employment losses. In contrast to the COVID-19 crisis, employment in accommodation and food services grew strongly by 3.6 per cent in 2009, as did other segments of the service sector.

Global employment growth rate (%) During crises induced by economic/financial shocks, a downturn typically results in the shedding of jobs in the formal sector due to their exposure to a fall in demand, including through trade channels, and the effects of a credit crunch. The informal sector often acts as an absorber of workers who have lost jobs in the formal sector due to the low entry costs (e.g. being a street trader). 21 In the context of household labour supply, job losses can result in an increase in the supply of women's labour to the informal economy in response to a layoff affecting the spouse, which is termed an "added-worker effect". For example, during the East Asian Financial Crisis, informal sector employment increased in Indonesia as a household coping mechanism. 22

In contrast, during the COVID-19 crisis, informal employment has been negatively impacted due to the lockdown and other containment measures, which prevented informal enterprises and workers from engaging in economic activity. The ILO has identified that one of the key reasons behind the upward revisions in working-hour losses made during 2020 and the greater losses in developing countries is the higher share of informal employment in these economies (ILO 2020a) . Based on the analysis of data for 11 countries, informal wage workers were, in fact, three times more likely, on average, than those in formal employment to lose their jobs during the COVID-19 crisis (ILO 2021b).

As evident in 2009 during the global financial crisis 23 and previous downturns, youth are especially vulnerable to the impact of an economic shock on the labour market. 24 Young people are more sensitive to a decline in GDP because they represent a large share of new job seekers, while they are more likely to work in less protected jobs and tend to be less expensive to fire (both due to provisions of employment protection and other forms of labour legislation and the inherent costs for employers of losing firm-specific capital, which is greater for more experienced (and older) workers). Young people also have weaker networks (with those who can assist them find a job) and face greater barriers to undertaking job search (Verick 2009 ).

Previous recessions have led to not only greater falls in employment and increases in unemployment for young people over the short-term, but also to a rise in long-term unemployment and overall detachment of youth from the labour market as captured by the not in employment, education or training (NEET) rate. Consequently, the situation for young people in the labour market typically takes longer to recover with damage persisting for many years after the initial economic decline, as witnessed in the years following the global financial crisis. As highlighted above, the global youth unemployment rate has not yet returned to the 21 For a discussion on informality, see, for example, Jutting and de Laiglesia (2009). 22 See, for example, Islam and Chowdhury (2009) and Matsumoto and Verick (2010) . 23 See, for example, Verick (2009) . 24 At the same time, there is some early evidence that, in contrast to previous crises, COVID-19 has adversely affected the labour market outcomes of older workers (see, for example, Bui et al. (2020) ). Moreover, older people have suffered much more in terms of the health effects of pre-GFC levels after 2009 (and before the COVID-19 crisis hit). The scarring effects of crises for young people have been well documented in the empirical literature. 25

The COVID-19 crisis has been no different and arguably worse for youth than previous crises because they have suffered across three dimensions (ILO (2020c); ILO (2021c)): (1) disruptions to education, training and work-based learning; (2) increased difficulties for young jobseekers and new labour market entrants; and (3) job and income losses, along with deteriorating quality of employment.

The greater sensitivity of youth employment to a decline in economic output during the COVID-19 crisis is reflected in the data for 2020 (figure 8), which shows that, for a given decline in economic output, there is a larger reduction in employment for young people than for adults. For young people, on average, a 10-percentage point decline in GDP translates into a 6.7 percentage-point decline in youth employment compared with 2.6 points for adults aged 25 and above. Due to the specific nature of the COVID-19 crisis, particularly in terms of the greater impact on the service sector, women, especially young women, have suffered more in the labour market. In addition, job losses for women have been exacerbated by the unequal distribution of the increased care responsibilities during the crisis, which have affected women disproportionately and constrained women's ability to be available for work, including remote work, and to undertake job search (ILO 2020d). 25 See, for example, Kahn (2010) and Schwandt and von Wachter (2019) . Using the sample of (middle-and high-income) countries with available data for 2020, the differences in employment losses by gender and age are apparent, along with differences between high-and middle-income countries. These figures show that employment declined in high-income countries, on average, by 15.6 per cent for young women and 11.4 per cent for young men compared with growth rates of -0.8 and -0.7 per cent for adult women and men aged 25 and above, respectively (figure 9). As noted above, employment fell by a greater amount in middle-income economies: in 2020, the decrease in employment reached 10 per cent for young women and 8.1 per cent for young men in these countries, while adult women and men experienced a decline of 4.7 and 2.8 per cent, respectively. Compared with the COVID-19 crisis, global employment declined in 2009 during the GFC for youth at a lower rate than in 2020, at -2.4 per cent for young women and -2.5 per cent for young men 26 . At the same time, employment continued to grow for adults around the world in 2009 (around 1 per cent). In high-income countries, employment fell across all age groups and gender breakdowns in 2009 but was more pronounced among youth, especially young men (-9.3 per cent for young men and -5.8 per cent for young women).

26 Source: Authors' calculations, ILO modelled estimates, ILOSTAT. 

While it is too early to fully conclude on the impact of the COVID-19 crisis, the analysis presented above reveals a number of trends that often diverge from the experience of the GFC in 2009. Overall, the impact of the COVID-19 crisis has been far more widespread than the GFC in terms of both breadth and depth of its impact on economies and labour markets around the world, mainly as a result of lockdown and other containment measures that severely curtailed. In contrast to the GFC, few countries and groups within national boundaries have escaped some of the negative effect of the COVID-19 crisis. However, certain sectors, such as accommodation and food services, and population groups, including young people and women, have been more negatively affected. The unique nature of the crisis has also resulted in unexpected labour market adjustment patterns, such as the greater rise in inactivity.

Based on a sample of middle-and high-income countries, employment losses in 2020 are best explained by the strictness of lockdown and other containment measures, which curtailed economic activity by impacting both the supply and demand sides of the economy. Building on the empirical analysis presented in this paper, the benefits of fiscal stimulus packages and other policy measures, including job retention schemes, need to be further evaluated once more data becomes available on the implementation of measures (though attribution of single policies will remain difficult).

Give the likely persistence of the crisis-induced jobs deficits, policymakers around the world will need to continue supporting the economy and labour market for some time to come. Over the coming years, these policy responses will need to take into account five key risks.

Firstly, the COVID-19 crisis is likely to lead to longer-term structural changes to the economy and labour market (i.e. effects on the structure of the economy and shifts between and within sectors). Some of these changes might not be in the direction of fairer and more inclusive economies. Many sectors, such as those linked to tourism are unlikely to return to their pre-pandemic situation, at least in the near future. This implies that policies will need to adapt to these changes by both providing continuing protection to some sectors, while also facilitating the shift of resources to others, including new sectors that have the potential to create decent and productive employment.

Secondly, given the shift of jobless into inactivity and the subsequent rise in unemployment further into the crisis, employment policies and active labour market programmes, matched by the necessary capacity and resources, will need to support transitions of people, particularly in terms of getting people back into employment.

Thirdly, young people around the world will need specific support to get into the labour market. As learned during the global financial crisis, measures targeting youth need to be comprehensive -e.g. the EU Youth Guarantee. 27 In addition, given the more severe impact on women, an integrated approach is needed to ensure that some of the losses can be overcome, while addressing gender disparities. This will involve looking at how investments and job creation can benefit women, including in such sectors as the digital and care economies. Constraints to women's participation, which have been further exacerbated by the COVID-19 crisis, need to be tackled with renewed efforts and resources (e.g. care services, leave policies and flexible work arrangements to help families juggle better caring responsibilities).

Fourthly, policy response will need to overcome worsening inequality that appeared during the crisis, and which may suppress aggregate demand and economic recovery. The biggest concern is the severe impact on developing and emerging economies, which will result in not only a worsening of employment outcomes for millions of people around the world but also an increase in poverty. The global convergence of previous decades has come undone. To bring these economies back to pre-pandemic growth trajectories will require strong international solidarity, including efforts to bring the vaccine as fast as possible to these countries and greater support on development financing (i.e. debt renegotiations and relief).

Finally, given the highly differentiated impacts especially between different groups, there is a high risk of K-recovery. If policymakers only focus on "averages" during the recovery without its distributional aspects, policies may withdraw prematurely, which will have a negative impact on the recovery itself through its effects on aggregate demand but also on exacerbating inequalities in the labour market, which will have long-term consequences.

Economic uncertainty in the wake of the COVID-19 pandemic

COVID-induced economic uncertainty and its consequences

Early Evidence on the Impact of COVID-19 and the Recession on Older Workers

The Short-Term Impacts of COVID-19 on Households in Developing Countries

Why did unemployment respond so differently to the global financial crisis across countries? Insights from Okun's Law

The global economic recovery 10 years after the 2008 financial crisis

The Debate on Expansionary Fiscal Consolidation: How Robust Is the Evidence?

From Crisis to Crisis: Capitalism, Chaos and Constant Unpredictability

The impact of financial crises on labor markets, household incomes, and poverty: a review of evidence

A global panel database of pandemic policies (Oxford COVID-19 Government Response Tracker

IMF. International Labour Organization (ILO). 2020a. ILO Monitor: COVID-19 and the world of work

Growth, Employment and Poverty Reduction in Indonesia

The long-term labor market consequences of graduating from college in a bad economy

2009 Is Informal Normal? Towards More and Better Jobs in Developing Countries

Does economic growth deliver jobs? Revisiting Okun's Law

Is the Indonesian labour market really so stable? Reviewing employment trends from over a decade of reforms, crises and recovery

The Aftermath of Financial Crises

Unlucky Cohorts: Estimating the Long-Term Effects of Entering the Labor Market in a Recession in Large Cross-Sectional Data Sets

Who is hit hardest during a financial crisis? The vulnerability of young men and women to unemployment in an economic downturn

The Great Recession of 2008-2009: causes, consequences and policy responses