Preparing ERTE for the Future

6. An in-depth evaluation of ERTE

Alexander Hijzen

Mateo Montenegro

Abstract

This chapter contributes to our understanding of the role of job retention schemes by performing an in-depth evaluation of the effects of ERTE on employment and productivity during the COVID‑19 crisis. It exploits the difference in co-financing requirements between small and large firms to identify the causal effects of ERTE on job retention, job reallocation and aggregate employment. Its main insight is that ERTE has been very effective in supporting employment during the COVID‑19 crisis.

In Brief

This chapter provides an in-depth evaluation of the causal effects of ERTE on employment and productivity. To this end, it exploits differences in co-financing requirements between firms with less than 50 employees and firms with 50 or more employees as measured on 29 February 2020 (before the outbreak of the COVID‑19 pandemic and the announcement of the rules for job retention support). To ensure that firms only differ in the generosity of job retention support to which they have access and not in the way they are treated by any other rules and regulations that make use of the same threshold, the analysis focuses on firms that cross the threshold just before are after of 29 February but have otherwise identical employment trajectories. Given the unprecedented use of job retention support, special emphasis is given to its macroeconomic effects by explicitly taking account of the effects of job retention support on the job-finding prospects of the unemployed. The analysis is implemented using unique administrative data covering the universe of workers and firms in the private sector.

The results of the evaluation show that ERTE has been a major success. It was not only widely used, but also proved highly effective in preserving jobs and preventing the labour market from becoming congested and depressing the job-finding prospects of the unemployed. Indeed, its effect on aggregate employment may have been larger than the number of jobs supported. This reflects in part the modest size of efficiency losses and in part the role of job retention support in limiting the increase in the expected duration of unemployment during the crisis by preventing congestion effects. As a result, the fiscal balance of job retention support was positive. The effects of ERTE on productivity are more uncertain but likely to be modest. More specifically, the following findings emerge:

• At the micro-level, ERTE was highly effective in preserving jobs. Lower co-financing requirements for firms with less than 50 employees increased the take‑up rate of ERTE by 4.5 percentage points on average during the period from March 2020 to August 2021, while it increased the job retention rate by 3.3 percentage points over the same period. Taken together, these estimates imply that efficiency losses – as a result of supporting jobs that would have been retained anyway or could not have been saved even with support – were modest at around 25%.

• At the macro-level, ERTE was even more effective since it prevented the labour market from becoming congested and depressing the job-finding prospects of the unemployed. It is shown using a difference‑in-difference design that the effective generosity of ERTE, as measured by the share of employment accounted for by small firms, supported the job-finding probability of the unemployed (or rather prevented it from falling further during the COVID‑19 crisis). Using a simple accounting framework, it is shown that congestion effects increase the effect of ERTE on employment by 50% or more. As a result, the effect of ERTE on aggregate employment exceeds the level of take‑up. For each worker supported between 1.1 and 2.2 jobs are saved. This implies in turn that the fiscal balance of job retention support was positive, i.e. the cost of supporting jobs was more than offset by lower expenditures on unemployment benefits and higher tax revenues from employed workers.

• The effects of ERTE on productivity are more uncertain but likely to be modest. While there is some indication that job retention support slowed the transition of workers to better firms in later periods, these effects were rather small. This is consistent with the fact that the use of job retention support was largely temporary and phased out in a timely manner as economic activity resumed.

6.1. Introduction

This chapter contributes to our understanding of the role of job retention schemes by performing an in-depth evaluation of the effects of ERTE on employment and productivity during the COVID‑19 crisis. It exploits the difference in co-financing requirements between small and large firms to identify the causal effects of ERTE on job retention, job mobility and employment. As such, it provides one of the first causal estimates of the effects of job retention schemes during the COVID‑19 crisis. Moreover, it explicitly takes account of the unprecedented scale at which job retention support was used by considering how ERTE may have affected the job-finding prospects of unemployed workers by preventing so-called congestion effects. The analysis relies on unique administrative data covering the universe of workers and firms in the private sector. A more detailed description of the analysis can be found in the corresponding background paper Montenegro and Hijzen (2024, forthcoming[1]).

6.2. The rationale and use of co-financing requirements

The evaluation of ERTE relies crucially on differences in co-financing requirements between firms of different size. This section therefore provides a detailed discussion of the rationale of co-financing requirements, their variation over time and across firms and their use in Spain.

6.2.1. The rationale of co-financing

The intention of job retention schemes is to support jobs that are temporarily under pressure but remain viable in the medium term. However, ensuring that subsidies go to the right job is not obvious for governments since they typically do not observe whether jobs are really at risk or will be able to resume relatively soon. Firms and workers typically have a better sense of the viability of jobs than governments, particularly in a context where social distancing restrictions remain important and market signals are weak. Governments therefore tend to make use of financial incentives for firms and workers to enhance the targeting of support to jobs that are temporarily at risk but remain viable. Co-financing requirements for firms for hours not worked have been widely used for modulating the degree of support, including in Spain.

The main idea of co-financing requirements for firms for hours not worked is to reduce the risk of supporting jobs that have weak prospects of resuming in the medium term and, in doing so, risk slowing down the reallocation of jobs from low productivity to high productivity firms (Burdett and Wright, 1989[2]). Locking workers in permanently unviable jobs results in so-called “displacement effects” as it makes it more difficult for firms wishing to expand to fill their vacancies. Requiring firms to share in the cost of reduced working hours reduces the attractiveness of short-time work for all firms, but particularly among firms with jobs that have weak long-term prospects (see Box 6.1 for details). Firms for which the long-term value of a job falls short of the cost of labour hoarding will instead lay the worker off. Whether displacement effects are empirically important depends on the extent to which job retention support is concentrated in low productivity firms, the duration of support relative to the evolution of labour demand and the speed of job reallocation in the absence of job retention schemes.1

6.2.2. Variation in co-financing

Most countries vary co-financing requirements over time, by making them stricter in good times and weaker in bad times (pro-cyclical). There are two main arguments for this. First, in bad times, a larger share of firms is likely to face liquidity constraints irrespective of their longer-term prospects. Consequently, the risk of supporting jobs that do not need support or have become permanently unviable is relatively less important. Second, in bad times, job reallocation tends to slow as good firms typically put recruitments on hold during a recession. Consequently, there is little risk of locking works in dead-end jobs that would otherwise have able been to find another job in a better firm. Workers who are laid off simply would have become unemployed. For these reasons, most governments set the cost of job retention support for firms to zero at the start of the COVID‑19 crisis and gradually re‑introduced co-financing requirements as economic activity was allowed to resume (OECD, 2022[3]).

Some countries also vary co-financing requirements across firms of different size. For example, Japan, Korea and Spain provided more favourable conditions to small firms during the COVID‑19 crisis, whereas the United States operated a specific job retention programme for small and medium-sized firms (e.g. the Paycheck Protection Programme in the United States). The usual justification for providing more generous support to small firms is that such firms are more likely to experience liquidity constraints due to their more limited access to external finance (Sharpe and A, 1994[4]; Doniger and Kay, 2023[5]; Chodorow-Reich et al., 2020[6]). In the specific context of the COVID‑19 crisis, an additional argument for targeting small firms may have been that small firms were more affected, either because they were concentrated in sectors most impacted or because the scope for teleworking was more limited.

6.2.3. Co-financing in Spain

Spain imposed higher co-financing requirements for firms with 50 or more employees than smaller firms during most of the COVID‑19 crisis. The co-financing difference was particularly important during the initial period of the COVID‑19 crisis when small firms were fully exempted from paying employer social security contributions over hours not worked, whereas larger firms received a partial exemption of 75% (Figure 6.1, Panel A). The difference corresponds to about 6% of labour costs (Figure 6.1, Panel B). As the crisis evolved, the difference was phased out in three steps: in July 2020, the difference in the exemption rate for employer contributions was reduced to 20 percentage points; in October 2020, it was further reduced to 10 percentage points; and in November 2021, the difference in co-financing was removed entirely. While these patterns in co-financing are quite complex, the empirical analysis only exploits the fact that there was a sizeable difference in co-financing since the start of the crisis until November 2021.

Figure 6.1. The costs of hours not worked for employers by firm size in Spain

New applications for COVID‑19 related job retention support in the case of full suspensions by firm size and time (March 2020 – October 2021)

Source: OECD calculations based on OECD Tax-Benefit models.

Box 6.1. The role of co-financing for the effectiveness of short-time work

This box develops a simple theoretical framework to illustrate how the generosity of job retention support for firms affects the ability to preserve jobs temporarily at risk due to liquidity constraints and its potential costs by supporting jobs that either do not need support or cannot be saved. The framework is illustrated visually in Figure 6.2.

Set up

Job matches between pairs of firms and workers are assumed to be heterogenous in their present values. The present value of job match j in period t is denoted by V_jt. All job matches are assumed to be hit by a common negative productivity shock in period t equal to -φ_t. The shock contains a temporary component of -(1‑p)φ_t, which is reversed in the next period and hence does not affect the long-term value of job matches, as well as a permanent component which affects the long-term value of job matches and is equal to -pφ_t. Thus, the present value of the job match at time t can be expressed as V_jt=-pφ_t + V_jt+1.

Job matches have limited liquidity. For the present purposes, liquidity is assumed to be linearly proportional to the present value of job matches, A=aV_jt, with a<1. The presence of liquidity constraints justifies the use of STW schemes to preserve jobs that are temporarily at risk but remain viable in the longer term. However, when doing so, STW may also end up supporting jobs that have become permanently unviable without the subsidy (“displacement effects”) or jobs that are not actually at risk of being destroyed (“deadweight effects”).

In the absence of public support, low productivity job matches between 0 and V₁ will be destroyed because they have become permanently unviable as a result of the permanent productivity shock, V_jt+1 < pφ_t. Moreover, some higher productivity job matches between V₁ and V₂ will be destroyed because they are temporarily unviable as a result of the temporary productivity shock. This is the case when liquid assets are insufficient to offset the temporary shortfall in revenue relative to labour costs, aV_jt+1 <(1‑p)φ_t.

Job retention subsidies

Now assume there is a job retention scheme which fully protects job matches against the shock in productivity by providing a subsidy that is equal to the size of the shock. Since this allows job matches to reduce operating costs one‑to‑one with the decline in revenue, all job losses will be prevented. It allows liquidity-constrained job matches at risk of being destroyed to be preserved as intended (V₁ to V₂). However, it will also yield some “displacement effects” by allowing job matches that would have become permanently unviable in the absence of the subsidy to be preserved (V₀ to V₁). Since all job matches are assumed to be subject to the same productivity shock, they are all eligible for support, giving rise to “deadweight effects” as high productivity job matches with a long-term value above V₂ would have been retained even in the absence of support.

Job retention loans

Assume instead that the government runs a job retention scheme which insures job matches against the temporary shock in productivity by providing liquidity loans rather than subsidies. In this case, job matches with a positive long-term value that are at risk of being destroyed due to liquidity constraints – those between V₁ and V₂ – are preserved since the availability of public loans entirely removes liquidity constraints. Moreover, there are no efficiency losses as in the case of job retention subsidies. Job matches with negative long-term values below V₁ will be destroyed (no displacement effects) whereas job matches that do not use support will take up any loans (no deadweight effects). Loans therefore provide a fully efficient way to promote job retention during a temporary crisis.

Intermediate cases

In practice, job retention schemes tend to be based on subsidies rather than loans.1 Instead, to increase the efficiency of job retention schemes, governments tend to rely on partial subsidies for hours not worked and a sharing of their costs with firms (and workers). The requirement of firms to cover part of the costs of hours not worked is referred to as direct co-financing whereas the requirement to payback part of the costs later may be referred to as experience‑rating. Co-financing reduces the effectiveness of preserving liquidity-constrained jobs, but also reduces the risk of inefficiency losses. Since experience‑rating is more akin to providing a loan, it tends to be more effective in preserving liquidity- constrained job matches and limiting inefficiency losses.

Figure 6.2. The effects of job retention support

No public support versus full public insurance against productivity shocks

Source: Montenegro and Hijzen (2024, forthcoming[1]), “Job Retention at Scale”. OECD background paper.

1 There are several of reasons for this: i) uncertainty about the nature of the shock may reduce the effectiveness of loans; ii) the aggregate productivity shock is not seen as an entrepreneurial risk; iii) outlays on short-time work subsidies are expected to reduce expenditures on unemployment benefits.

6.3. Methodology and data

This section briefly discusses the empirical approach and data used.

6.3.1. Empirical approach

A natural way of exploiting the firm-size threshold for co-financing would be to make use of a regression discontinuity design (RDD). This strategy consists of comparing the outcomes of firms just below the 50‑employee threshold to those just above it. The idea is that firms just above and below the threshold are identical in all respects, except in the cost of JR support. Unfortunately, this strategy is not viable in the present case. The reason for this is that Spanish law makes use of the 50‑employee threshold to regulate many different policies, including regulations about tax, firing procedures and labour representation (Arregui and Shi, 2023[7]). As a result, any comparison between firms above and below the threshold also incorporates the effect of these other regulations, and hence cannot be used to isolate the effect of co-financing.

To overcome this issue, the analysis exploits the fact that the regulation about ERTE co-financing is based on firm size on a specific date, 29 February 2020, 17 days before the rules on ERTE co-financing were announced (17 March 2020). Since these regulations were announced after the measurement date of firm size, firms could not alter their size to get beneficial co-financing rates. This feature of the scheme allows exploiting transitions of firms across the threshold in the period just before or just after the measurement date (and well before the announcement date). Conditional on the employment trajectory of firms, i.e. whether they move above or below the threshold, the timing of these transitions and hence their co-financing requirements can be considered as good as random (see Box 6.2 for details on the methodology).

Importantly, this evaluation strategy allows providing one of the first causal estimates of the use of job retention support during the COVID‑19 crisis and the first to specifically focus on the role of co-financing requirements for JR support. This is important since co-financing has been the main parameter used by governments to modulate the level of support over the course of the COVID‑19 crisis, with potentially important implications for the cost-effectiveness of support.

6.3.2. Data and implementation

The analysis makes use of a unique dataset based on the combination of two administrative sources. The first is the Altas y Bajas de la Seguridad Social, an employer-employee dataset of the universe of workers and firms during the period 2019‑22, with information on job spells, type of contract, firm size, province and industry. The second is a worker-level dataset recording when workers were put on ERTE (March 2020 to February 2022).

To avoid errors in the measurement of firm size on 29 February 2020, the analysis is restricted to single-establishment firms, without atypical forms of contracts in the private sector, excluding agriculture.2 Firms with only fixed-term workers are also excluded.

To analyse firm dynamics, monthly firm variables are constructed first (e.g. employment or suspended workers) and then averaged over five consecutive six‑month periods between March 2020 and October 2022. The first three periods correspond to those in which there were differences in co-financing requirements by firm size and the last two to those in which there were no longer any differences in co-financing.

For the purposes of this report, the results are presented visually by displaying the average outcome for firms just above the 50‑employee threshold, the average outcome of firms just below and the statistical significance level of the estimated difference in average outcomes between firms just below and above threshold.

Box 6.2. Empirical strategy to identify the causal effects of co-financing

Intuition

The present analysis relies on a novel identification strategy based on the measurement date of firm size used for determining the required level of co-financing by firms. More specifically, the rules for co-financing are based on the measurement of firm size on 29 February 2020, 17 days before the rules were announced on 17 March. Since these regulations were announced after the measurement date of firm size, firms could not alter their size to get more beneficial co-financing rates. This feature of the scheme allows exploiting firm transitions around the threshold in the period just before and after the measurement date. Conditional on the employment trajectory of firms, i.e. whether they move above or below the threshold, the timing of these transitions and hence the applicable co-financing regime, can be considered as good as random.

The intuition behind the identification strategy is conveyed by the comparison of firms A and B which exhibit identical employment trajectories except for their timing as displayed in Figure 6.3. Both firms have 49 employees one week before 29 February 2020 and 50 employees one week after. However, firm A crosses the 50‑employee threshold before 29 February 2020, while firm B does so after this date. According to ERTE regulations, this means that firm B will have access to more favourable co-financing rates than firm A. Since ERTE regulations were established after 29 February 2020, firms could not alter their firm size in response to them. Thus, firms A and B should be ex-ante identical in all respects but their access to different co-financing. Comparing their outcomes will thus isolate the causal effect of co-financing.

Figure 6.3. Illustration of the identification strategy

Source: Montenegro and Hijzen (2024, forthcoming[1]), “Job Retention at Scale”, OECD background paper.

Model

One can generalise the example discussed above to firms on different trajectories using the following econometric specification:

$y_i=\beta I\left(E_i^{Feb29}<50\right)+{\theta }_i(E_i^{-1w},E_i^{+1w}, I(T_i>\mu (E_i^{-1w},E_i^{+1w}))+\gamma X_i+{\epsilon }_i$ (1).

where $Y_i$ is an outcome variable for firm i; $\left(E_i^{Feb29}<50\right)$ is an indicator function for whether firm size ($E_i^{Feb29})$ is below 50;${\theta }_i(E_i^{-1w},E_i^{+1w}, I(T_i>\mu (E_i^{-1w},E_i^{+1w}))$ are fixed effects for firms which have identical employment size one week before and after 29 February 2020, denoted as $E_i^{-1w}$and $E_i^{+1w}$ respectively, and have a similar level of volatility in their hires and separations throughout this time window, as measured by whether their excess turnover in this period, denoted by $T_i$, is above the median of the group of firms experiencing the same change in employment in the time window, denoted by $\mu (E_i^{-1w},E_i^{+1w})$. Including a measure of volatility is important since firms with higher turnover are more likely to cross the 50-employee threshold but are also likely to be different in terms of their behaviour to less volatile firms.

The baseline specification also includes a set of control variables $X_i$(province fixed effects, 3‑digit industry fixed effects, the proportion of workers on fixed-term contracts in January 2020, employment growth between January 2019 and 2020, employees in January 2020 that were still employed or in the same firm as in January 2019, an indicator for firms in protected 4‑digit industries according to ERTE regulations, excess turnover in 2018‑19 and an index of firm poaching measured in 2018‑19). Estimates without these control variables are qualitatively similar, but less precisely estimated. Finally, ${\epsilon }_i$ is a random error term. See Hijzen and Montenegro (2024, forthcoming[1]) for a discussion of the validity of the identification strategy.

Source: Montenegro and Hijzen (2024, forthcoming[1]), “Job Retention at Scale”, OECD background paper.

6.4. Results

This section first presents the microeconomic estimates of lower job retention support for take‑up and job retention using the identification strategy laid out above. It then extends the analysis of direct effects (partial equilibrium) to assess the aggregate effects of JR support on employment by taking account of its indirect effects on the job-finding prospects of the unemployed through so-called “congestion” effects. It concludes with an analysis of its effects of productivity growth through its impact on efficiency-enhancing job reallocation.

6.4.1. Micro effects on take‑up and job retention

Lower co-financing increases take‑up

On average during the period from March 2020 to August 2021, firms with less than 50 employees, and thus with lower average co-financing rates, exhibited higher levels of take‑up than larger firms (

Figure 6.4, Panel A). More specifically, the estimates suggest that co-financing increased the take‑up rate of ERTE by an additional 4.5 percentage points on average during the period. This is a large effect since it represents a 38% increase with respect to the average in larger firms.

Since the difference in co-financing rates across the 50‑employee threshold tended to vary between 10 and 25 percentage points (which represents respectively 65 and 162 euros for the average worker),3 this suggests that a 10 percentage-point decrease in co-financing increases take‑up by between 1.8 and 4.5 percentage points (which corresponds to 15% and 38% compared to mean take‑up).

Lower co-financing also increases job retention

Higher take‑up among firms with lower co-financing requirements translates into higher job retention (Panel B). While larger firms retained 83% of their employees during the first 18 months of the pandemic, the job retention rate in smaller firms with lower co-financing requirements was 86%, 3.3 percentage points higher (which corresponds to a 4% increase).

A natural way to measure the importance of deadweight or displacement effects (i.e. workers that would have been retained in the absence of JR support or that were supported but could not be retained) is to compare the effects of lower co-financing on take‑up and job retention. If the ratio of the effect on job retention to the effect on take‑up is smaller than one, this indicates that more workers are suspended than retained, pointing to the presence of efficiency losses. Based on the estimates, the ratio is about 0.73 (3.3/4.5). This suggests that on average during the first 18 months since the start of the COVID‑19 crisis, efficiency losses were modest, amounting to 27%. Note that these effects do not take account of broader employment effects that arise due to the implications of ERTE for the job-finding opportunities of the unemployed (labour market congestion effects). This will be discussed further below.

The effects of co-financing are concentrated among workers on open-ended contracts

The effects of co-financing are concentrated among workers on open-ended contracts. The same reduction in co-financing induced a 5.2 percentage point increase in take‑up for workers on open-ended contracts but only a 3.4 percentage point increase for workers on fixed-term contracts. Similarly, it increased job retention by 4.2 percentage points for workers on open-ended contracts, while the effect on workers with fixed-term contracts is statistically insignificant. This suggests that co-financing reinforced labour market duality by limiting the use of ERTE to workers with open-ended contracts. It is not entirely clear to what extent this also increased cost-effectiveness. Since the estimates for fixed-term contracts are statistically insignificant, there is considerable uncertainty about the cost-effectiveness of ERTE for workers on such contracts.

Figure 6.4. Effect of co-financing on take‑up and job retention

Take‑up and job retention on average during the period March 2020‑August 2021 as a share of workers employed in February 2020

Notes:

**, *** statistically significant at the 5 and 1 percent level respectively. For further details, please refer to Box 6.2.

The figure displays the mean take‑up of ERTE during the first 18 months of the pandemic from March 2020 to August 2021 in firms with 45‑49 employees (small firms) and 50‑54 employees (large firms) on 29 February 2020, as estimated using equation (1). Take‑up in Panel A refers to the share of workers employed on February 2020 that were suspended on average each month during the period from March 2020 to August 2021. Job retention in Panel B refers to the share of workers employed in small (large) firms in February 2020 that were still employed in the same firm on average in each month during the period from March 2020 to August 2021.

Source: Ministry of Inclusion, Social Security and Migration (Ministerio de Inclusión, Seguridad Social y Migraciones).

6.4.2. Aggregate employment effects

The estimates of ERTE presented so far are direct effects or partial equilibrium effects in the sense that they do not take account of general equilibrium feedback or spillover effects. However, given the widespread use of ERTE, at least during the initial phases of the COVID‑19 pandemic, such effects are potentially important. General equilibrium effects may operate through different channels. Here we concentrate on the effects of ERTE through labour market tightness, and its implications for the job-finding probability of the unemployed (labour market congestion effects). The effects of ERTE on aggregate demand through higher consumption are not considered.

To analyse congestion effects, we start from the observation that ERTE may affect unemployment both by reducing inflows to unemployment as analysed in the previous subsection and by increasing outflows from unemployment, relative to what would have happened in the absence of ERTE, by limiting the competition for job vacancies among unemployed jobseekers. To assess the aggregate unemployment effects of ERTE, we complement the estimates of the direct effects of ERTE on job retention (unemployment inflows) with estimates of the effect of ERTE on unemployment outflows. This is done by using a simple difference‑in-differences design that exploits differences in the effective generosity of ERTE across local labour markets that result from differences in firm-size structure (see Box 6.3 for details). The results indicate that ERTE significantly increased the unemployment outflow probability (relative to what would have happened without support). Combining the estimates with those on the unemployment inflow probability using the framework laid out in Box 6.3, the number of jobs saved per job supported varies between 1.1 and 2.2. In other words, for each job supported between 1.1 and 2.2 jobs were saved. This means that between one‑third and two‑thirds of the effect of ERTE on employment is driven by its direct effect on job retention and the remainder by its indirect effect on job-finding among the unemployed through positive congestion externalities.

With the estimates of the aggregate employment effects in hand it is relatively straightforward to calculate the fiscal costs of using ERTE. The fiscal costs of using ERTE are determined by the costs of ERTE, its effects on unemployment benefit expenditures for the unemployed and its effects on tax revenues (personal incomes taxes and social security contributions) from the employed. Since aggregate employment increased and the costs of short-time work and unemployment benefits are the same in Spain, it is straightforward to see that ERTE yielded a net benefit. Note that without incorporating the congestion externality, it would have been almost fiscally neutral. The latter is similar to what Kopp and Siegenthaler (2019[8]) find for Switzerland for the use of short-time work during the global financial crisis.

Box 6.3. Analysing congestion effects

The aggregate effect of ERTE on employment can be written as the sum of its effect on the employment outflow probability times employment and its effect on the employment inflow probability times unemployment. In steady-state, employment is constant, and employment outflows equal employment inflows, such that $E^{*}=\frac{f}{1+f-r}.$ In steady-state, the aggregate effect of ERTE on employment can be written as:

1) $\frac{\partial E^{*}}{\partial JRS}=\frac{\partial \left[\frac{f}{1+f-r}\right]}{\partial JRS}=\frac{f^{\text{'}}\left(1+f-r\right)-f\left(f^{\text{'}}-r^{\text{'}}\right)}{{(1+f-r)}^2}$

where $E^{*}$ is the steady-state employment, r is the monthly job retention rate and f is the monthly employment inflow rate. From the partial equilibrium analysis, the impact of ERTE on the job retention rate ($r\text{'})$was estimated to be 0.7 (the estimate on job retention over the estimate on take‑up). The average monthly job retention rate before the COVID‑19 crisis (r) is equal to 0.8 and the average monthly employment inflow rate is equal to 0.08. The effect of ERTE on the employment inflow rate (f’) needs to be estimated.

The effect on the employment outflow rate is estimated using a difference‑in-differences design, which specifies the monthly employment inflow rate in local labour market i at time t ($f_{it})$as a function of the effective generosity of ERTE as measured by the share of employment in small firms ($e^{<50}/e^{all})$during the COVID‑19 period (Post) as follows:

2) $f_{it}=\frac{\beta e^{<50}}{e^{all}}*Post+ {\delta }_i+{\delta }_t+X_{it}^{\text{'}}\gamma +{\epsilon }_{it}$

where ${\delta }_i$and ${\delta }_t$ respectively denote local labour market (region) and time (month*year) fixed effects, and $X_{it}^{\text{'}}$ is a vector of time‑varying covariates. The results indicate that the effective generosity of ERTE increased the employment inflow rate by between 0.01 and 0.034 or between 13 to 42% compared to the mean at baseline.

Having estimated the effects of ERTE on the employment inflow and outflow rates, one can now plug all values in equation (1). Using f’=0.01 or f’=0.034 yields $\frac{\partial E^{*}}{\partial JRS}=1.1 \mathrm{o}\mathrm{r}\mathrm{ }\frac{\partial E^{*}}{\partial JRS}=2.2.$This implies that for each worker on JR support between 1.1 and 2.2 jobs were saved.

Source: Montenegro and Hijzen (2024, forthcoming[1]), “Job Retention at Scale”, OECD background paper.

6.4.3. Effects on efficiency-enhancing reallocation

Apart from employment, ERTE may also have implications for productivity. It may enhance productivity by preserving match-specific human capital in jobs that have become temporarily unviable but remain viable in the longer term, while it may undermine productivity by propping up jobs in firms with structural difficulties and slowing the process of efficiency-enhancing job reallocation across firms. This sub-section concentrates on the possible effects of ERTE on slowing down efficiency-enhancing job reallocation.

The few previous studies that have looked into this have adopted a local labour market approach (Aiyar and Dao, 2021[9]) or done so indirectly by examining the characteristics of the firms using JR support (Giupponi and Landais, 2023[10]). Here a different approach is taken by exploiting the worker-level information in the data to analyse the impact of co-financing on the probability of transitioning to a better firm using the same micro‑econometric framework as above for the analysis of take‑up and job retention.4 Since information on productivity is not available in the data, different measures of firm quality are used instead. These are the poaching index, i.e. the extent to which firms recruit workers from other firms rather than from non-employment (Dustmann et al., 2021[11]), and excess turnover, i.e. the level of hiring and separations in excess of those needed to grow at a given rate (Davis, Haltiwanger and Schuh, 1998[12]).

The results indicate that JR support only had very limited effects on efficiency-enhancing job reallocation (Montenegro and Hijzen, 2024, forthcoming[1]). Only during later periods is there some evidence that workers in firms with access to more generous support are less likely to transition to better firms, but the effects are not statistically significant. This suggests that in later periods some workers were kept in jobs that had better opportunities elsewhere. This is driven by the effect of JR support on slowing job transitions in general. It did not render job mobility less allocative since conditional on moving JR support increased rather than reduced the probability of moving to a better firm. This is likely to reflect the possibility that ERTE prevented workers from being knocked off the job ladder and pushed into taking lower quality jobs. The conclusion that the reallocation of JR schemes are modest echoes previous findings by Calligaris et al. (2023[13]) who study this question in a cross-country context.

6.5. Conclusion

This chapter provides an in-depth evaluation of the causal effects of ERTE on employment and productivity. To this end, it exploits differences in co-financing requirements between firms with less than 50 employees and firms with 50 or more employees as measured on 29 February 2020 (before the outbreak of the COVID‑19 pandemic and the announcement of the rules for job retention support). Given the unprecedented use of job retention support special emphasis is given to its macroeconomic effects by explicitly taking account of the effects of job retention support on the job-finding prospects of the unemployed. The analysis is implemented using unique administrative data covering the universe of workers and firms in the private sector. The results of the evaluation show that ERTE has been a major success. It was not only widely used, but also proved highly effective in preserving jobs and preventing the labour market from becoming congested.

An important avenue for future work would be to focus more in-depth on the role of programme design and, more specifically, the optimal modulation of co-financing requirements over the course of a crisis. However, such an analysis requires information on productivity which, for the moment, is not available. Such information would allow 1) documenting the trajectory of financial variables of firms that have benefited from ERTEs, 2) estimating the impact of access to ERTEs on the financial performance of firms and their employees, and 3) understanding how the effectiveness of ERTEs is mediated by the financial strength of firms. In principle, information can be obtained by merging the present dataset with Central de Balances de Empresas of the Bank of Spain.

References

[9] Aiyar, M. and M. Dao (2021), “The Effectiveness of Job-Retention Schemes: COVID-19 Evidence From the German States”, IMF Working Papers, https://ideas.repec.org/p/imf/imfwpa/2021-242.html (accessed on 12 February 2024).

[7] Arregui, N. and Y. Shi (2023), “Labor Productivity Dynamics in Spain: A Firm-Level Perspective”, IMF Selected Issues Paper, Vol. SIP/2023/002.

[2] Burdett, K. and R. Wright (1989), “Unemployment Insurance and Short-Time Compensation: The Effects on Layoffs, Hours per Worker, and Wages”, Journal of Political Economy, Vol. 97/6, pp. 1479-1496, https://doi.org/10.1086/261664.

[14] Cahuc, P. and S. Nevoux (2017), “Inefficient short-time work”, IZA DP, No. 11010, http://www.iza.org (accessed on 15 February 2019).

[13] Calligaris, S. et al. (2023), “Employment dynamics across firms during COVID-19: The role of job retention schemes”, OECD Economics Department Working Papers, No. 1788, OECD Publishing, Paris, https://doi.org/10.1787/33388537-en.

[6] Chodorow-Reich, G. et al. (2020), Bank Liquidity Provision Across the Firm Size Distribution, National Bureau of Economic Research, Cambridge, MA, https://doi.org/10.3386/w27945.

[12] Davis, S., J. Haltiwanger and S. Schuh (1998), “Job Creation and Destruction”, MIT Press Books, Vol. 1, https://ideas.repec.org/b/mtp/titles/0262540932.html (accessed on 8 December 2020).

[5] Doniger, C. and B. Kay (2023), “Long-lived employment effects of delays in emergency financing for small businesses ✩”, https://doi.org/10.1016/j.jmoneco.2023.08.002.

[11] Dustmann, C. et al. (2021), “Reallocation Effects of the Minimum Wage”, The Quarterly Journal of Economics, Vol. 137/1, pp. 267-328, https://doi.org/10.1093/QJE/QJAB028.

[10] Giupponi, G. and C. Landais (2023), “Subsidizing Labour Hoarding in Recessions: The Employment and Welfare Effects of Short-time Work”, The Review of Economic Studies, Vol. 90/4, pp. 1963-2005, https://doi.org/10.1093/RESTUD/RDAC069.

[8] Kopp, D. and M. Siegenthaler (2019), “Short-Time Work and Unemployment in and after the Great Recession”, KOF Working Papers, Vol. 462, https://doi.org/10.3929/ethz-b-000359533.

[1] Montenegro, M. and A. Hijzen (2024, forthcoming), “Job Retention at Scale”, OECD background paper.

[3] OECD (2022), Riding the waves: Adjusting job retention schemes through the COVID 19 crisis, https://read.oecd-ilibrary.org/view/?ref=1130_1130248-vu3d9r6eog&title=Riding-the-waves-Adjusting-job-retention-schemes-through-the-COVID-19-crisis (accessed on 27 September 2023).

[4] Sharpe and S. A (1994), “Financial Market Imperfections, Firm Leverage, and the Cyclicality of Employment”, American Economic Review, Vol. 84/4, pp. 1060-1074, https://ideas.repec.org/a/aea/aecrev/v84y1994i4p1060-74.html (accessed on 7 October 2019).

Notes

← 1.  Displacement effects also can arise when there is a recurrent use of short-time work as this introduces a distortion that shifts employment to seasonal activities where workers are only employed for part of the year (Cahuc and Nevoux, 2017[14]).

← 2.  Namely, we exclude firms with indeterminate types of contracts and those with “fijo-discontinuo” contracts, which are seasonal contracts, for which the dataset used does not determine the period in which the contract is active. We define the public sector as those activities belonging to sections O,P and Q of the NACE Rev 2.0 classification. Private firms are those that do not belong to these activities.

← 3.  This is assuming the average gross monthly salary in 2020 of EUR 2 160 for full-time worker, and 30% of employer social security contributions.

← 4.  These are likely to be underestimates to the extent that they do not take into account the impact of ERTE on job transitions in firms that do not use ERTE (same argument as for congestion).