The spread of the COVID-19 virus across countries and measures taken by governments to contain it – including shutdowns of many businesses and restrictions on travel and mobility – imply sharp contractions in GDP and associated employment losses that dwarf those experienced during the global crisis of 2008-09. The OECD projects the OECD-wide unemployment rate to increase by around 6 percentage points between the fourth quarter of 2019 and the second quarter of 2020 as compared to an increase of 2.2 percentage points between the third quarter of 2008 and the second quarter of 2009.

Labour market policies play a critical role in limiting social hardship and ensuring that employment rapidly rebounds once the shutdown of non-essential activities is eased.1 The focus is on the respective roles of policies aimed at preserving existing jobs (e.g. short-time work schemes, temporary layoff schemes and administrative measures to limit dismissals) and the unemployment insurance system. Economic downturns triggered by shocks that are both transitory and exogenous, such as natural disasters, typically require limited reallocation of resources. In this case, policies to preserve existing jobs may be the best course of action to both support workers and ensure businesses quickly resume activity once the initial shock fades. However, existing jobs may become unviable following shocks that require a sizeable reallocation of resources, such as financial and housing crises or persistent changes in commodity prices. In this case, partly relying on the unemployment insurance system allows for sufficient reallocation of resources rather than preserving existing jobs that may no longer be viable. The optimal mix of job preservation and unemployment benefit policies to support workers and ensure a rapid recovery thus depends on whether the exogenous COVID-19 shock turns out to be purely transitory or, as suggested by current OECD projections, more persistent.

Okun’s law quantifies the average response of the unemployment rate to changes in GDP growth. Previous studies generally find that Okun’s law is a strong empirical regularity in most countries (Ball et al., 2017). However, the size of the Okun coefficient – the effect of a 1% shock to GDP on the unemployment rate – is typically found to vary across countries and, to a lesser extent, within countries over time. Typical estimates of the Okun coefficient range from -0.1 to around -0.8, suggesting that a 1% decline in GDP may raise the unemployment rate by between 0.1 and 0.8 percentage points. These differences are typically interpreted as reflecting differences in labour market policies and institutions.

The country-specific Okun coefficients are estimated using quarterly unemployment and GDP data over the period 2000-2019 based on the following equation:



where Uq is the unemployment rate in quarter q, Yq is real GDP, β are the Okun coefficients, and ε is the error term; α denotes the intercept and can be interpreted as the change in the unemployment rate at zero  GDP growth. Consistent with previous studies, estimated Okun coefficients range from around -0.1 in some countries, including Japan, Korea and Norway, to -0.8 in Spain (Annex Figure A.1).

The Okun predictions conditional on GDP projections in a scenario with a single COVID-19 outbreak (“single-hit scenario”) suggest that the OECD unemployment rate could increase from around 5% in the fourth quarter of 2019 to 8% in the second quarter and 9% in the third quarter of 2020. In the single-hit scenario, GDP growth in the OECD would be around -2% in the first quarter of 2020, -13% in the second quarter and +6½ per cent in the third quarter. Predicted unemployment based on Okun’s law continues to increase in the third quarter despite positive GDP growth because the estimated Okun coefficients imply significant persistence in unemployment.

The positive average deviation of OECD projections from Okun’s law could reflect a number of factors. One factor could be the prevalence of COVID-19-related shutdowns in highly employment-intensive industries, which appears indeed to be the case for the OECD as a whole (Annex Figure A.2). Another reason that may be particularly relevant in the current context could be that the unemployment response is non-linear in the sense that large negative GDP shocks may have disproportionately large and rapid effects on the unemployment rate compared with more moderate shocks. In any case, the positive average deviation of projections from the Okun prediction hides significant differences across countries (Annex Figure A.3). In part, this may reflect the fact that OECD projections account for exceptional labour market policy measures taken in response to the COVID-19 crisis.

Policies that encourage the preservation of existing jobs, such as job retention schemes as well as administrative suspensions of dismissals, may lead to deviations of unemployment from the Okun benchmark. The Okun benchmark described above measures the average response of unemployment to changes in GDP growth, both during economic upturns and economic downturns. To the extent that a number of governments have put in place exceptional measures to dampen the unemployment increase in response to the COVID-19 crisis, one may expect the Okun benchmark to over-predict the increase in unemployment, or at least not to under-predict it as observed in Figure 1 for the OECD average. Although data on GDP growth and unemployment for the first half of 2020 are not yet available to formally test this hypothesis, a first assessment can be made by analysing OECD projections. OECD country experts integrate real-time information on GDP growth and unemployment from high-frequency indicators in their projections, as well as information on the extent of job-preserving measures, including past experience and real-time information on programme uptake. While by no means allowing a formal test of the hypothesis that job-preserving measures dampen increases in unemployment, the use of OECD projections allows summarising the currently available information in a synthetic way.

To compare and contrast outcomes, countries are split into those with large job retention schemes (“retention-based countries”) and those that have taken no exceptional measures in this area, continuing to rely mostly on unemployment insurance (“unemployment insurance-based countries”). Retention-based countries have either expanded existing job retention schemes or introduced large schemes during the crisis, with take-up suggesting that a significant share of businesses and workers are participating in them (Table A.1). Unemployment insurance-based countries do not have a job retention scheme in place, or take-up of existing schemes has been limited to a small fraction of businesses and workers.

Unemployment projections overshoot the Okun benchmark in unemployment insurance-based countries but are similar to the benchmark in job retention-based countries (Figure 2). The absence of significant deviation from the Okun benchmark in retention-based countries – despite the fact that the shutdown industries appear to be particularly labour-intensive, suggesting a particularly large response of unemployment – indicates that country desks project job retention schemes to significantly dampen increases in unemployment. In the second quarter, the deviation in retention-based countries is negligible while the overshoot in insurance-based countries is around 7½ percentage points. Taking the difference in the deviation from the Okun benchmark in retention-based and insurance-based countries at face value, on average country desks project job retention schemes to dampen increases in unemployment in the second quarter of 2020 by 7½ percentage points (0 deviation minus positive deviation of 7½ percentage points).2

A complementary way to assess the likely effectiveness of policies to preserve existing jobs is to compare real-time unemployment developments across countries with and without large job retention schemes. Such comparisons would ideally be based on deviations from the same Okun benchmark. However, GDP is available on a less timely and lower-frequency basis than unemployment data, which makes the computation of real-time deviations from Okun’s law impossible. The approach taken in Box 2 is to report both changes in registered unemployment and the workforce covered by applications to job retention schemes. Increases in unemployment have been systematically smaller in countries with larger coverage of applications, suggesting that these schemes have been effective in limiting increases in unemployment.

The above analysis based on OECD unemployment projections and real-time unemployment data suggests that measures to encourage the preservation of existing jobs are likely to be effective in limiting increases in unemployment in the short term. Previous research also suggests that such policies are effective in the sense that they do not primarily preserve jobs that would have been preserved even in the absence of job retention schemes (Hijzen and Venn, 2011; OECD, 2018). Searching for suitable jobs in terms of wage and non-wage attributes, such as location, working time or employer amenities, is costly for workers, as is the search of employers for suitable workers. Preserving existing jobs reduces such costs of matching employers to employees and may thereby promote a quicker labour market recovery as activity rebounds. To the extent that the COVID-19 shock is temporary and does not require a major reallocation of resources, freezing the existing allocation of resources by preserving existing jobs may also promote longer-term growth of employment and productivity by limiting the loss of firm-specific human capital.3 However, the preservation of existing jobs may not be efficient if the COVID-19 shock turns out to be more persistent than initially expected as some fraction of jobs preserved by short-time work schemes may not be viable in the long term. For instance, a number of non-essential activities (e.g. travel, hotels and restaurants, parts of the retail sector, recreational services) may suffer persistent rather than transitory declines as a result of new social distancing standards or changes in consumer preferences.4

One policy option to preserve existing jobs are job retention schemes. These schemes typically operate on the principle that businesses are subsidised to preserve existing job matches while workers experience no or limited wage losses (Box 3). In practice, businesses continue to pay employees a significant part of their monthly wages even though they are working only part-time or not at all. In return, they can claim a wage subsidy that covers part of the excess wage cost.5 Short-time work schemes allow for work sharing in the sense that working time for all workers is typically reduced by a fixed proportion, whereas temporary layoff schemes allow businesses to put all or a proportion of workers on “furlough” (i.e. zero hours). In practice this difference is less significant, as most temporary layoff schemes set up during the COVID-19 crisis allow for some degree of work sharing. A number of countries, including Australia and New Zealand, have introduced broad wage subsidy schemes that are not conditional on working time reductions but may be used as short-time work or temporary layoff schemes.6 Businesses resorting to job retention schemes are typically required not to dismiss workers while using the schemes, although this is not the case in all countries, e.g. in the German short-time work scheme (Box 3).7

Job retention schemes that are used for work sharing preserve human capital of workers particularly effectively, as workers continue to work part-time but receive a subsidy for being partially unemployed. The adjustment in hours worked allows firms to adjust working time rather than employment, thereby preserving the job match while allowing workers to maintain their human capital and avoiding the trauma of job loss. In order for job retention schemes to promote work sharing among all workers, an important consideration is the coverage of non-standard workers, such as temporary or dependent self-employed workers. Broad coverage can ensure that the burden of employment adjustment does not disproportionately fall on non-standard workers (OECD, 2020c).

Another labour market policy option to preserve existing jobs and freeze the existing allocation of resources is to suspend the dismissal of workers for economic reasons. A number of countries, including Italy and Spain, have introduced such suspensions to varying degrees (Box 2). In contrast to short-time work schemes – whose cost is typically shared between workers, firms and the government – the cost of administrative suspensions of dismissals is fully borne by firms if no compensating subsidies are in place. This may put firms that may otherwise be viable at risk of failure. In Italy and Spain, for instance, this risk is mitigated by providing subsidies through liquidity support measures (OECD, 2020a) or by combining suspensions of dismissals with short-time work schemes. However, a significant drawback of such suspensions is that they do not cover non-standard workers, such as temporary workers with imminent contract expiration dates or dependent self-employed workers who are not covered by dismissal regulations. Limited coverage of non-standard workers by short-time work schemes could further re-inforce such uneven employment adjustment across different groups of workers (OECD, 2020c).

A number of countries, including many in Central and Eastern Europe and the United States, have taken very limited labour market measures to support the preservation of existing jobs.8 Firms in these countries have greater incentives to lay off workers in response to the COVID-19 shock. For instance, US data on registered unemployed shows that about 13% of the labour force have been laid off in the United States between mid-March and end-April 2020.9 This partly reflects the ease of layoffs in the United States and the absence of significant job retention schemes at the federal level.10

This approach allows for the possibility that the COVID-19 shock may have more persistent economic implications than initially expected and may therefore require a significant reallocation of resources in the future. Laid-off workers are more likely to engage in job search than workers on short-time work schemes. At the same time, an exceptionally high share of layoffs during March and April in the United States appears to be temporary, with around 90% of all laid-off workers in the April labour force survey reporting to be on temporary layoff.11 Temporarily laid-off workers have explicitly been provided with a recall date by their employers or expect to be recalled in the future, suggesting that the employer-employee relation has not been fully severed and a degree of attachment of the employee to the previous employer remains intact (Groshen, 2020). The recall rate is particularly high for temporary layoffs – around 85% according to Fujita and Moscarini (2017) – suggesting that a significant share of these workers may be recalled by their previous employers if and when economic conditions normalise.12 By contrast, employers for which the COVID-19 shock adversely affects longer-term growth prospects are likely to permanently sever the employment relationship.

The main drawback of relying on unemployment insurance rather than preserving existing jobs is the risk of excess dismissals and social hardship. Businesses do not immediately bear the cost of laying off workers while they partly bear the cost of short-time work schemes. Even if firms expect job matches to be viable in the long term they may choose to lay off workers to reduce costs, thereby creating a negative externality on the unemployment insurance system (Cahuc and Zylberberg, 2008).13 Such risk of excess dismissals is particularly pronounced in countries with weak employment protection. At the same time, relying on unemployment insurance rather than preserving existing jobs may lead to social hardship, especially where earnings replacement rates are low or a significant part of the workforce may not be eligible for unemployment benefits (e.g. the dependent self-employed) or may only be entitled to low benefits (e.g. temporary workers with patchy employment histories (OECD, 2020c). Even with extended coverage and enhanced generosity of unemployment benefits, this approach may nonetheless lead to social hardship in countries where health and/or pension insurance are provided by employers or linked to peoples’ employment status.

Given the high degree of uncertainty on the consequences of the COVID-19 shock for the reallocation of resources, the challenge for policy makers is to find the right balance between measures to promote the preservation of jobs that are viable in the long term and the reallocation of workers in unviable jobs. The prudent course of action is to combine policies to preserve existing jobs with temporary expansions of unemployment benefits to limit the income loss for laid-off workers. One policy option is to adjust the relative cost for firms of choosing short-time work over lay-offs (OECD, 2018). If more reallocation is deemed to be required, for instance because activity in high-contact sectors does not fully recover in the medium term, lay-offs could be made relatively more attractive by reducing the government subsidy to short-time work schemes while possibly protecting workers’ income by expanding unemployment insurance. This may become increasingly relevant as shutdowns are eased in countries with particularly generous government subsidies to job retention schemes, such as Denmark, France and the United Kingdom. In countries with a risk of excess layoffs, such as most Central and Eastern European countries and the United States, there may be room to promote the use of existing short-time work schemes and/or making access to various aid programmes set up in response to the COVID-19 shock conditional on preserving employment.14 In US states with short-time work schemes, for instance, firms could be encouraged to reduce working time rather than lay off workers, which would give workers on reduced hours access to the full weekly USD 600 lump-sum payment in the COVID-19 rescue package on top of pro-rated unemployment benefits (Von Watcher, 2020).

Other options to adjust the balance between the preservation of existing jobs and reallocation is to promote reallocation in job retention schemes and promote recalls where firms rely predominantly on lay-offs to adjust total hours. Reallocation in job retention schemes could be promoted by lifting restrictions on taking new jobs without workers losing their benefits. Workers in job retention schemes could also be provided with training subsidies, for instance in the area of digital skills, which may allow them to look for and perform jobs online. In countries where firms predominantly rely on layoffs, preservation of existing jobs that are viable in the long term could be promoted by subsidising recalls of previously dismissed workers. Firms do not account for the positive externality of recalls on workers’ wages as the gains from recalls only partly accrue to firms through higher productivity while part of them go to workers through higher wages. In Israel, for instance, the government introduced a recall subsidy of around USD 2100 at the end of May. One option of subsidising recalls in the context of the COVID-19 crisis is to partly convert liquidity support in the form of interest-free loans or tax deferrals into subsidies conditional on firms recalling their previously dismissed workers (Fujita et al., 2020).

As the COVID-19 crisis evolves, finding the right balance between job preservation and reallocation of resources may require adjusting the parameters of existing labour market measures. This will involve some degree of experimentation regarding the sharing of costs related to job retention schemes between employers, employees and the government, as well as a focus on restoring viable job matches in countries that have experienced large numbers of layoffs. As activity in a number of industries resumes, a renewed focus on active labour market policies, including training and public employment, on top of appropriate income support may limit the costs of reallocation for workers. In order to provide timely and granular labour market policy advice, the OECD is monitoring ongoing reallocation across firms, industries and regions using real-time data on online job advertisements. The results of this work will be reported in forthcoming OECD policy briefs.


Abraham, K. and S. Houseman (2020), Shared-Work Programs Can Ease the Coronavirus’s Economic Impact,

Adams-Prassl, A. et al. (2020), “Inequality in the Impact of the Coronavirus Shock: Evidence from Real Time Surveys”, IZA Discussion Papers, No. 13183, Institute of Labor Economics (IZA), Bonn.

Ball, L., D. Leigh and P. Loungani (2017), “Okun’s Law: Fit at 50?”, Journal of Money, Credit and Banking, 49/7, 1413-1441,

Barrero, J., N. Bloom and S. Davis (2020), COVID-19 Is Also a Reallocation Shock, National Bureau of Economic Research, Cambridge, MA,

Borup, D. and E. Montes Schütte (2019), In Search of a Job: Forecasting Employment Growth Using Google Trends.

Cahuc, P. (2019), “Short-time work compensations and employment”, IZA World of Labor,

Cahuc, P. and A. Zylberberg (2008), “Optimum income taxation and layoff taxes”, Journal of Public Economics, 92/10-11, 2003-2019,

DARES (2020), Situation sur le marché du travail au 12 mai 2020, (accessed on 14 May 2020).

Fujita, S. and G. Moscarini (2017), “Recall and Unemployment”, American Economic Review, 107/12, 3875-3916,

Fujita, S., G. Moscarini and F. Postel-Vinay (2020), The labour market policy response to COVID-19 must save aggregate matching capital, VoxEU CEPR Policy Portal,

Goldsmith-Pinkham, P. and A. Soujourner (2020), Predicting Initial Unemployment Insurance Claims Using Google Trends, (accessed on 26 March 2020).

Groshen, E. (2020), It Matters that Most COVID Layoffs in March were Furloughs,

Hijzen, A. and D. Venn (2011), The Role of Short-Time Work Schemes during the 2008-09 Recession, OECD Publishing.

OECD (2018), Good Jobs for All in a Changing World of Work: The OECD Jobs Strategy, OECD Publishing, Paris,

OECD (2020a), Corporate vulnerabilities during the Covid-19 outbreak: assessment and policy responses.

OECD (2020b), Evaluating the Initial Impact of COVID Containment Measures on Activity.

OECD (2020c), Identifying workers most vulnerable to COVID-19-driven economic activity containment: Stylised facts and policy considerations.

OECD (2020d), OECD Unemployment Rates News Release: March 2020, (accessed on 13 May 2020).

OECD (2020e), Supporting people and companies to deal with the Covid-19 virus: Options for an immediate employment and social-policy response.

OECD (2020f), Supporting livelihoods during the COVID-19 crisis: Closing the gaps in safety nets.

OECD (2020g), Combatting COVID-19's effect on children.

OECD (2020h), Women at the core of the fight of COVID-19 crisis.

OECD (2020i), Public employment services in the frontline for employees, jobseekers and employers.

OECD (2020j), VET in a time of crisis: Building foundations for resilient vocational education and training systems.

Von Watcher, T. (2020), A Proposal for Scaling Enrollments in Work Sharing (Short-Time Compensation) Programs During the Covid-19 Crisis: The Case of California,

Data from private businesses can provide an additional indication of unemployment developments in real time. Google Trends (GT) aggregates data on searches on Google, and assigns them to topics based on proprietary algorithms. The topic “unemployment benefits” aggregates search interest in this topic independent of the original language and exact terms the user typed into Google. Unlike verbatim search queries that are used in the forecasting literature on US labour markets (Borup and Montes Schütte, 2019; Goldsmith-Pinkham and Soujourner, 2020) this facilitates the use of Google Trends for cross-country studies. GT captures online search interest in unemployment benefits, which can provide an indication of the inflows into unemployment. The data are provided as an index relative to the overall search volume in the same period and location.

Figure B.1 shows that countries with a larger share of the workforce covered by applications to job retention schemes typically experienced smaller increases in Google searches for unemployment benefits than countries where such schemes either covered only a small fraction of the labour force, or countries without such schemes. In some countries, average search volume between March and May 2020 was more than 15 times higher than during a typical month in 2019. This pattern mirrors the evidence from registered unemployment (Figure 3). A caveat is that interest in unemployment benefits might not always be associated with an inflow into unemployment as measured by labour force surveys or in unemployment registers. In some countries, such as Israel, Norway or Spain, workers placed on temporary lay-offs receive unemployment benefits, but are not counted as unemployed in the labour force surveys if the lay-off lasts three months or less, in accordance with ILO statistical guidelines (OECD, 2020d). In some countries, such as Spain, such workers are not counted in the unemployment register whereas in other countries, such as Israel or Norway, they are included in the register. An additional limitation is that some countries that introduced new job retention schemes, such as New Zealand, experienced strong search interest in unemployment benefits especially in March, even though actual inflows into registered unemployment were low. Other countries that substantially extended unemployment benefits are Canada, Chile, Israel, Ireland, Norway and the United States. A final limitation of Google Trends data lies in the possibility that a given increase in search interest might not translate into the same increase in unemployment across countries, due to differences in internet penetration, demographic structure, administrative procedures, or search behaviour across countries.


Cyrille SCHWELLNUS  (✉

Michael KOELLE (✉

Balazs STADLER (✉

This paper is also available as Issue Note 5 OECD (2020), OECD Economic Outlook, Volume 2020 Issue 1: Preliminary version, OECD Publishing, Paris,


← 1. Limiting social hardship and ensuring a rapid rebound of employment also requires closing gaps in social safety nets (OECD, 2020f); supporting people particularly exposed to social hardship, including children and women (OECD, 2020g; OECD, 2020h); supporting public employment services (OECD, 2020i); as well as adjusting public training schemes (OECD, 2020j). The central role of employment and social policies in supporting well-being and limiting social hardship will be further explored in forthcoming OECD work.

← 2. Note that employment intensity in the shutdown sectors is not systematically higher in insurance-based countries than in retention-based countries, suggesting that differences in the deviation from the Okun benchmark do not simply reflect differences in employment intensity.

← 3. Evidence for the United States suggests that the probability of recalling a worker that has previously been laid off is positively correlated with pre-unemployment tenure and wage declines among laid off workers are smaller for those who are eventually recalled (Fujita and Moscarini, 2017), which suggests that firm-specific human capital is important.

← 4. Based on a business survey, Barrero et al. (2020) estimate that in the United States the COVID-19 shock has thus far caused three new hires in the near term for every 10 layoffs.

← 5. Job retention schemes can be viewed as involving a short-term loan from businesses to the government if the subsidy payment is made to businesses and there is a delay between the payment of wages and the reception of government subsidies, e.g. in the Danish and German schemes (Box 3).

← 6. These schemes typically require the wage paid by businesses to equal or exceed the government subsidy.

← 7. In countries with strict employment protection, short-term work and/or job retention schemes can be viewed as a liquidity support measure for firms since employment is not adjustable in the short term. In fact, short-term work schemes tend to be more prevalent in countries with strict employment protection (Cahuc, 2019).

← 8. These countries have generally taken significant non-labour market measures to preserve existing businesses, including through liquidity support (OECD, 2020a), and shelter workers from income losses through expanded unemployment insurance (OECD, 2020e).

← 9. The cumulated increase in initial unemployment claims over the same period is even higher (around 21%). Initial claims can overstate the inflow into registered unemployment if applications are rejected due to ineligibility, withdrawn, or if duplicate claims are filed for the same individual.

← 10. In the United States, short-term work schemes exist in 26 states but take up has been extremely limited (Von Watcher, 2020).

← 11. By way of comparison, the overwhelming part of layoffs during the economic crisis of 2008-09 reflected permanent layoffs, with the share of temporary layoffs in total layoffs declining from around 30% in early 2008 to around 20% in late 2009.

← 12. Recalls are particularly likely if the COVID-19 shock is short-lived, as the likelihood of recall is high for low unemployment durations but declines over time (Fujita and Moscarini, 2017).

← 13. Employers may bear some of the cost in the medium term if social security contributions are experience rated, in the sense that they depend on firms’ past dismissal behaviour as is the case in the United States.

← 14. The Paycheck Protection Programme enacted as part of the COVID-19 rescue package (CARES Act) allows for conversion of loans into grants if funds are used for payroll, with allowances for rent and utilities. However, caps on the size of loans per firm and high rents in large cities appear to have limited the extent of job retention.


This paper is published under the responsibility of the Secretary-General of the OECD. The opinions expressed and the arguments employed herein do not necessarily reflect the official views of OECD member countries.

This document, as well as any data and map included herein, are without prejudice to the status of or sovereignty over any territory, to the delimitation of international frontiers and boundaries and to the name of any territory, city or area.

© OECD 2020

The use of this work, whether digital or print, is governed by the Terms and Conditions to be found at