Cross-country evidence on the allocation of COVID-19 government subsidies and consequences for productivity

Abstract

We study the consequences of the Covid-19 pandemic and related policy support on productivity. We employ an extensive micro-distributed exercise to access otherwise unavailable individual data on firm performance and government subsidies. Our cross-country evidence for five EU countries shows that the pandemic led to a significant short-term decline in aggregate productivity and the direct support to firms had only a limited positive effect on productivity developments. A thorough comparative analysis of the distribution of employment and overall direct subsidies, considering separately also relative firm-level size of support and the probability of being supported, reveals ambiguous cross-country results related to the firm-level productivity and points to the decisive role of other firm characteristics.

1. Introduction

The Coronavirus pandemic and related containment measures have led to the deepest disruption in global economic activity since the Second World War. The pandemic has evoked massive government support interventions, unprecedented in scope and scale. Although fiscal support to firms was well justified in order to limit bankruptcies, capital disruption and job losses, its longer-term effect on aggregate productivity is unclear. This study adds to the debate by providing comparable firm-level data-based evidence for EU countries along several dimensions. First, we cluster firms according to their pre-pandemic performance to evaluate the allocation of government subsidies to each cluster. We find that the pandemic state aid was distributed rather efficiently, as it supported mainly viable and productive firms in temporary need. Second, we document a large short-term decline in productivity due to the pandemic. Third, we quantify the impact of subsidies on productivity and find that it has only partially offset the large negative shock to productivity. Fourth, the aggregate outcome is largely driven by country-specific factors, including sectoral and size composition of the business sector, as the relationship between productivity and probability of being supported is inconclusive at the firm level.

While being extensive by now, the literature on the economic impact of the pandemic is still hampered by the significant lag in the availability of firm-level balance sheet data. At the same time, data on Covid-19 support often originate from different sources, making it usually difficult to combine with the balance sheet data. In addition, a cross-country analysis faces additional challenges as micro-level data are typically confidential and not harmonized across countries making cross-country comparisons difficult. We contribute to the literature by (i) combining unique and rich firm-level data on public Covid-19 support with firm-level data measuring firm performance and (ii) providing cross-country evidence for five EU countries (Croatia, Finland, Netherlands, Slovakia and Slovenia) by implementing a micro-distributed approach to the analysis. Our data preparation and analysis benefit from the established CompNet (The Competitiveness Research Network) infrastructure. We distribute a common code, independently executed by data providers on their respective national firm-level information. This method ensures high coverage and cross-country comparability while preserving confidentiality criteria.1

The subsidies we consider refer predominantly to the first and the most damaging wave of the Covid-19 pandemic. Although our calculations are based on subsidies provided over the entire year 2020, more granular data shows that subsidies assigned for the period between March and June 2020 represent between 45% to 70% of overall resources allocated in 2020. For Croatia, Slovakia and Slovenia we consider solely employment (wage) subsidies, i.e. support received by a firm related to employment contracts kept by the firm even if the work had been suspended. For Finland and Netherlands, we include a broader set of subsidies.

Our analysis adds cross-country empirical evidence to the literature on Covid-19 and productivity. It builds on the early considerations by di Mauro and Syverson (2020) which highlight the main channels through which the crisis may affect productivity growth. The Covid-19 shock affects production inputs and consequently overall productivity. With respect to labour inputs, we expect significant impacts on human capital. School closures, despite the huge progress made in distance learning, and increased difficulties in integrating young people into the labour market, may have a long-term negative effect on human capital (Martin et al. 2021). In addition to distance learning, we observe a historic increase in teleworking, that may also have consequences for productivity (see Bartik et al., 2020). The pandemic-related uncertainty and lack of financial resources influences capital inputs and investments. As suggested by Calligaris et al. (2021), lower investments may result in a long-term reduction in productivity.

Following Schumpeter (1939) we may assume that the pandemic recession accelerates the process of labour reallocation from low to high-productivity firms. However, substantial policy measures implemented to mitigate the Covid shock could potentially reduce productivity-enhancing reallocation. One of the first estimates published by Andrews et al. (2021a) suggests that job reallocation remained connected to firm productivity during the pandemic, i.e. high productivity firms were more likely to expand and low productivity firms were more likely to contract. The pandemic coincided with a temporary strengthening of the reallocation-productivity link in Australia – but a weakening in New Zealand – which appears related to the design of job retention schemes.

The Covid-19 pandemic brings two additional phenomena related to the extensive government support. First, as reported by Wang et al. (2020) for the United States, Freeman et al. (2021) for Netherlands or Müller (2021) for Germany, during the crisis, when one would expect higher probability of firm default, countries are experiencing lower numbers of bankruptcies compared to the pre-pandemic period. Second, despite fears and policy suggestions (e.g. Laeven et al., 2020), we do not observe an immediate increase in Zombie firms. On the public support allocation, Altomonte et al. (2021) find that it was allocated in line with firm productivity in Italy and Germany, while it was found productivity-neutral in France.2

Like Bénassy-Quéré et al. (2021), Demmou et al. (2021), or Lalinsky and Pal (2022), we benefit from micro data originating from balance sheets and income statements to compute productivity developments during the pandemic. Our findings challenge some results based on survey data. For example, Fernández-Cerezo et al. (2021) state that the Covid-19 shock had a stronger impact on small, young and less productive firms, which resorted relatively more to all available support schemes, including furloughs in Spain. Harasztosi et al. (2021) employ results of an EU-wide EIB Investment Survey and document that firms with low pre-Covid-19 productivity are significantly more likely to be supported than firms with high productivity and that being an exporter also matters, albeit to a lesser extent. Similarly, studies focusing on Japanese experience, amongst others Morikawa (2021), Hoshi et al. (2022) or Honda et al. (2023), find that less productive firms or firm with lower credit scores were more likely to apply for and receive subsidies.

The paper proceeds as follows. Section 2 describes the data and the methodology. In Section 3 we briefly characterize the Covid-19 support measures in EU countries and in the sample countries in particular. Section 4 analyses the allocative efficiency of the support measures. In Section 5 we study the consequences of the support measures for aggregate productivity. Section 6 concludes.

2. Data and methodology

Firm level data is of key importance in studying the Covid-19 crisis, as the pandemic has hit companies unevenly across, but even within, sectors. Macro or industry level data thus provides an incomplete picture of the heterogenous effects of the crisis across firms. We combine firm-level administrative data on firm-performance in 2019 with firm-level information on subsidies received by each firm during 2020. For each firm we observe characteristics such as revenues, value-added and input costs as well as other financial variables from balance sheet and income statements together with employment data. The data originate from national administrative sources and are representative of all non-financial firms.3 They are harmonised using the CompNet approach (CompNet, 2020)4 to comply with micro-level confidentiality restrictions and are combined with firm-level data on pandemic subsidies.

The government support data refers to Covid-19 related employment subsidies allocated to firms in 2020 in Croatia, Slovakia and Slovenia. In the case of Netherlands, we investigate both employment subsidies and overall direct subsidies. The analysis of Covid-19 support in Finland builds on overall direct Covid-19 subsidies to firms in 2020.5

The size of support and share of supported firms, as well as the type of support, differ across countries. The support reached between 29% (Finland and Slovakia) and 59% (Croatia) of firms. The largest relative size of employment subsidies was recorded in Croatia (5% of firm revenue) and the largest relative size of overall support was observed in the Netherlands (11% of firm revenue).6

To study the distribution of the Covid-19 related support, we start with defining several clusters based on firm performance in the years before the pandemic and computing the share of subsidies allocated to each cluster to study the allocative efficiency of the support. We continue with estimating logit regressions to assess the relationships between firm characteristics and corresponding support. Finally, we perform OLS analysis to estimate to what extent the size of the support at the firm level correlates with firm's characteristics. These steps allow us not only to (i) quantify the share of subsidies allocated to “deserving” firms, but also to (ii) assess the impact of the Covid-19 pandemic on aggregate productivity.

The details of the econometric approaches and productivity decomposition used in the paper are described in the associated sections. Whereas the analysis of subsidy allocation connects pre-pandemic firm characteristics with pandemic subsidies, the analysis of productivity effects relies on projected firm developments in value added and employment during the pandemic.

Firm-level productivity is measured by labour productivity in firm i in sector s in year t using value added and the number of employees

$$\mathit{Prod}\mathit{ucti}\mathit{vit}{y}_{\mathit{ist}}=\frac{(\mathit{Sale}{s}_{\mathit{ist}}-\mathit{Cost}{s}_{\mathit{ist}})}{\mathit{Empl}\mathit{oyee}{s}_{\mathit{ist}}}$$ (1)

where value added is defined as the difference between sales and costs.7

The pandemic-time firm-level Sales_ist are derived from the pre-pandemic firm Sales_ist-1 and an annual index of sectoral turnover I_st recorded during the pandemic following the relationship

$$Sale{s}_{ist}=I_{st}.Sale{s}_{ist-1}$$ (2)

To quantify pandemic-time changes in costs associated with pandemic-time changes in sales we assume material costs Costs_ist to be proportional to total firm sales Sales_ist according to

$$\Delta Cost{s}_{ist}={\alpha }_s.\Delta Sale{s}_{ist}\text{with}0<{\alpha }_{\mathrm{s}}<1$$ (3)

Sectoral material cost elasticities α_s originate from Maurin and Pal (2020).8 Firm-level pandemic-time costs are calculated as

$$Cost{s}_{ist}=Cost{s}_{ist-1}+\Delta Cost{s}_{ist}$$ (4)

Analogously to material costs, the pandemic-time change in firm-level number of employees follows

$$\Delta Employee{s}_{ist}={\beta }_s.\Delta Sale{s}_{ist}\text{with}0<{\beta }_{\mathrm{s}}<1$$ (5)

where β_s stands for sectoral labour cost elasticities from Maurin and Pal (2020). Firm-level pandemic-time employment is then calculated as

$$Employee{s}_{ist}=Employee{s}_{ist-1}+\Delta Employee{s}_{ist}.$$ (6)

The firm-level pandemic-time productivity is then aggregated to the country level. For comparison between pandemic-time productivity with support and without support (see Section 5), we also aggregate firm-level productivity with support. The projection of the firm-level pandemic productivity with support follows Eqs. (1) to (4) but adds the value of the subsidy to the value added (numerator) and assumes fixed employment (to the value from year 2019).9

3. Covid-19 support in EU countries

EU countries have differed significantly in how the pandemic has hit the economies and in the respective policy responses. The cross-country heterogeneity in the spread of the virus including contagion, hospitalization and death rates has been large. The economic effect of the pandemic has also differed widely across countries, depending on, e.g., the relative size of the service sectors (in particular the industries with personal contacts such as the accommodation and food services industries).10 Policy measures have naturally played an important role for the economic consequences of the pandemic, but also self-imposed voluntary social distancing on the behalf of consumers has been important. Moreover, policy responses have varied both in size and type, depending on the above factors, but also importantly depending on the institutional features of the respective economies. The choice of Covid-19 support measures relied in many countries on pre-existing institutions, such as pre-pandemic distribution channels of public subsidies to firms or labour market adjustment channels, such as short time working schemes. Furthermore, the existing automatic stabilizers in the economy have influenced the need and choices of Covid-19 policy measures. Therefore, the Covid-19 support measures are not directly comparable. Reflecting these factors, the discretionary fiscal responses in EU countries range from under 5 percent to roughly 20 percent of annual GDP (Fig. 1 ).11

Fig. 1.

GDP growth in 2020 and discretionary fiscal response to the Covid-19 crisis in EU-economies.

Sources: IMF WEO Database and IMF Database of Country Fiscal Measures in Response to the COVID-19 Pandemic. Note: Estimates as of June 5, 2021. Numbers are based on July 2021 World Economic Outlook Update.

Fernández-Cerezo et al. (2021) use a Spanish sample to document that firms predominantly responded to the pandemic-time decline in sales by reducing their investments and by implementing working from home. Short-time working schemes (and employment support) were deemed as especially useful for medium-sized firms and firms facing significant (more than 15%) decline in sales. As mentioned by Anderton et al. (2020), job retention schemes reached unprecedented levels already in the first months after the onset of the Covid-19 pandemic. In April 2020, when lockdown measures to contain the spread of Covid-19 were in place in most euro area countries, 15% of all employees in Germany, 34% in France, 30% in Italy and 21% in Spain were on short-time work.12 Fig. 2

Fig. 2.

Share of allocated employment subsidies (% of year total).

Source: CompNet Data Providers.

Although the design of the employment support shares a similar structure of eligibility and replacement conditions across countries, the conditions themselves differ and their content has been changing during the first year of the pandemic. In response to the actual pandemic and economic developments, we could observe several adjustments in eligibility thresholds, wage replacement rates and wage caps in the analysed European countries. The largest cross-country convergence was recorded in case of the eligibility threshold for revenue decline. As shown in Table 2 , at the end of 2020, in all countries of our sample the official minimum eligibility threshold for employment support was a decline in revenue exceeding 20%. However, in practice firms with better performance were not necessarily excluded.13 Other conditions of employment support schemes differed across countries.

Table 1.

Summary statistics (mean value by country).

Variables	Croatia	Finland	Netherlands	Slovakia	Slovenia
Revenue (Thousand EUR)	1400	2542	8469	2765	2863
Employees	10.7	13.1	33.8	11.2	14.0
Supported firms (Share of total)	0.59	0.29	0.50	0.29	0.46
Size of support (Share on revenue)	0.05	0.07	0.11	0.02	0.03
Observations	85,424	113,454	120,211	93,520	36,339

Note: Relative support for supported firms only. Based on employment support (Croatia, Slovakia and Slovenia) and overall direct support (Finland and Netherlands). Employment support in the Netherlands reached 40% of firms (i.e. coefficient is 0.4) and its size was 0.5.

Table 2.

Summary of main eligibility criteria and coverage of employment support.

	Croatia	Netherlands	Slovakia	Slovenia
Eligibility threshold	At least 20% decline in sales (revenue)	At least 20% decline in sales (revenue)	At least 20% decline in sales (revenue)	At least 20% decline in sales (revenue)
Wage replacement rate	Not set	Up to 85%	Up to 80%	Up to 100% (on average 80%)
Wage cap	4000 HRK (approx. 532 EUR)	Not set	1100 EUR	1754 EUR (country average wage)

Note: The conditions may differ depending on the specific type of support and the time of application.

Source: Eurofound and data providers’ information.

In addition to employment support analysis, data for the Netherlands and Finland allow us to investigate the distribution and effects of overall subsidies to companies. The list of considered subsidies by countries is summarized in Table 3 .

Table 4.

Allocation of subsidies to selected firm clusters.

	Wage subsidies				Overall subsidies
Cluster	Croatia	Netherlands	Slovakia	Slovenia	Finland	Netherlands
High productive	33.9	32.2	32.2	30.2	24.7	29.6
Low Productive	7.0	21.6	5.8	9.7	13.1	18.5
Young and high productive	2.8	1.4	1.1	1.3	N/A	2.0
Zombie	3.5	3.7	4.6	2.4	2.2	3.5
High-tech, knowledge intensive	15.6	41.1	32.7	22.6	30.8	38.4
Low-tech, not knowledge intensive	76.2	55.4	61.9	70.9	59.9	57.1
Growing	16.3	12.6	10.9	12.4	19.7	13.8
Declining	2.5	0.9	3.0	3.2	1.5	0.8

Note: High (low) productive firms are firms belonging to the highest (lowest) quintile of the labour productivity distribution. Young and high productive firms are firms from the highest quintile of productivity distribution and active for less than three years. Zombie firms are firms recording negative profits for three consecutive years and are not high labour growth prior the pandemic. High-tech (low-tech) firms are defined following the sectoral classification shown in Table A7 in Appendix. Growing firms are defined as firms in the highest quartile of the rate of change of labour productivity distribution and in the highest quartile of the rate of change of size distribution. Declining firms are defined as firms in the lowest quartile of the rate of change of labour productivity distribution and in the lowest quartile of the rate of change of size distribution.

Source: Authors’ calculations based on micro-data from Croatia, Finland, Netherlands, Slovakia and Slovenia.

Table 3.

List of subsidies included in the analysis.

Croatia	Finland	Netherlands	Slovakia	Slovenia
Employment subsidies		Employment subsidies	Employment subsidies	Employment subsidies
	Direct costs subsidies	Fixed costs subsidies
		Tax deferrals
		Pandemic loans

Source: Data providers’ information.

The evolution of allocated subsidies is highly correlated across countries, and it follows the economic impact of the pandemic. Governments paid most of the subsidies to counteract the huge drop in sales recorded in April 2020 and they allocated 45% to 70% of all subsidies between March and May 2020.

4. Allocative efficiency of the support

To understand the distribution of the Covid-19 related support, we conduct three different types of analysis. First, we start with an unconditional analysis: we cluster firms according to their performance during the pre-pandemic years and then we quantify the amount of subsidies allocated to each cluster. Second, we estimate logit regressions to assess the relationships between firm characteristics and probability of receiving support. And third, we estimate OLS regressions to analyse to what extent the size of the support at the firm level correlates with firm characteristics.

The presence of eligibility thresholds suggests the use of more advanced methods. Following the literature, we might consider the discontinuity in the probability of being supported around the sales growth threshold. For example, Kawaguchi et al. (2022) use regression discontinuity design (RDD) to evaluate the impacts of two Covid pandemic related subsidy programs in Japan, where they observed strictly implemented eligibility thresholds. However, in the countries of our interest, the value of the eligibility criterion has evolved in time and it did not represent a necessary condition for receiving support in all cases. In addition, eligibility thresholds were based on higher (monthly) frequency data and the balance sheet data at our disposal are in annual frequency and they finish in 2019 for most of the countries.

For the robustness analysis, we utilize Slovak annual balance sheet data available up to 2020 and construct sub-samples of firms, so that we set an upper bound to the annual growth in sales (in 2020) in each subgroup (from −20% to 18%). We then compute alternative margin estimates for the probability of receiving support for 39 sub-samples of firms to understand how the coefficient for labour productivity varies with respect to the annual growth in firm revenue (sales). The results show that the alternative margin estimates are relatively stable and do not differ much with respect to the threshold.14

4.1. Unconditional cluster analysis

One of the main concerns related to the massive Covid-19 stimulus packages implemented around the world is whether governments allocated the support to viable and productive firms in a temporary (financial) need, instead of insolvent unviable firms. In other words, to what extent are governments supporting firms that would (or should) quit the market even without the pandemic shock?

To form an aggregate view, we define several clusters and assign firms to the clusters based on their performance in the years preceding the pandemic. We find that around one third of the wage subsidies in Croatia, Netherlands, Slovakia and Slovenia was allocated to productive firms, defined as firms that belonged to the highest quintile of the labour productivity distribution in 2019. And a significantly lower share of subsidies was distributed to low productivity firms. Only a very small share, less than 3%, went to young productive start-ups, defined as firms in the highest quintile of productivity distribution and active for less than three years.15

To consider the sustainability of long term economic growth, we investigate the extent to which the Covid-19 aid has supported innovative or technologically advanced firms. Our calculations show that high-tech and knowledge intensive firms have received a relatively low share of subsides. Low-tech and low-knowledge intensive firms received the majority of the subsidies (between 55% and 76% depending on the country).16

Another important concern is the misallocation of Covid-19 related support to zombie firms. We find that in all countries under review, only a small share of subsidies went to firms recording negative profits and at most low/moderate employment growth for three consecutive years prior the pandemic (i.e. to zombie firms).17

The relatively high and low shares of subsidies allocated, respectively, to growing and declining firms confirm this evidence. As shown in Fig. 3 , the firms that in 2019 were amongst the firms which experienced the largest growth in labour productivity and largest growth in number of employees (first quartile of the two growth distributions) received between up to 20% of subsidies. Contrarily, firms from the lowest quartile of the two growth distributions, called declining firms, received a negligible share of the subsidies.18

Fig. 3.

Share of productivity deciles on overall wage subsidies, revenue, employment and firm population.

Source: Authors’ calculations based on micro-data from Croatia, Netherlands, Slovakia and Slovenia.

A closer look at the overall allocation of employment subsidies across firm productivity deciles (instead of quintiles presented above) reveals other common, but also some country specific findings. As shown in Fig. 3, the allocation of employment subsidies across productivity deciles closely follows the contribution of the deciles to aggregate employment in all countries.19

At the same time, firms from the upper (above median) parts of the (within-country) productivity distribution frequently received shares of employment subsidies exceeding their shares of firm population, but frequently slightly exceeding also their (high) shares of employment. This does not hold for the Netherlands, where, on one hand, only the highest decile was disproportionately subsidized and, on the other hand, a large part of subsidies was allocated to the lowest productivity deciles that, in contrast to the other three countries, highly contribute to the overall employment.

A more granular decomposition across macro-sectors shows that the allocation of subsidies to low productivity firms in the Netherlands is largely sector specific. Low-productive firms from administrative and support service activities, accommodation and food services and wholesale and retail trade received a relatively large part of the subsidies.20 To at least some extent this allocation is the result of the more uneven presence of low productive firms in these sectors and the relatively importance of these sectors in the dutch economy (also compared to other countries considered in the analysis).

The cluster analysis provides a useful snapshot of how governments allocated subsidies across groups of firms with similar characteristics. However, it has some limitations: first it suffers from some confounding factors (e.g. firm size or sector) and second, it does not provide information on the relationship between firm characteristics and the probability of getting the support or the size of it. We focus on these aspects in the following sections.

4.2. Firm probability to receive the support

We perform conditional analysis of the firm probability to receive employment subsidies and overall direct subsidies. We start with the analysis of subsidy allocation by firm productivity and continue with other firm characteristics. We regress the dependant variable - binary dummy variable equal to 1 for a supported firm and 0 otherwise – on different explanatory variables of interest and a set of covariates.

$$\text{Pr}({\mathrm{Y}}_t=1|X_{t-n})=\frac{1}{1+\text{exp}\left(-X_{t-n}\beta \right)}$$ (7)

where Pr(Y_t=1|X_t-n) denotes the probability of receiving support for a firm in period t given X_t-n, which is a row vector of explanatory variables and β is the corresponding column vector of regression coefficients.21

We find several common patterns. In line with the purpose of the support and nature of the shock, firms supplying accommodation and food services had the highest chance of being supported. And even after controlling for sectoral characteristics, larger firms were supported with higher probability than the smaller ones. However, our estimates related to firm productivity show an increased level of heterogeneity and non-linearity, pointing to a decisive role of sectoral and size characteristics. A positive relationship between firm probability to be supported and firm productivity was confirmed only in Slovakia and Croatia, i.e. countries with a higher contribution of large firms to overall productivity and relatively higher share of productive firms in the affected service sectors.

Allocation probability by productivity deciles

The conditional probability of receiving support with respect to firm productivity differs across countries. As shown in Table 5 , we find that the chance of being supported increases with firm productivity in Croatia and Slovakia, while the relationship is negative in Slovenia, Finland and Netherlands. In particular our cross-country results do not confirm universal validity of previous evidence on higher probability of employment support allocated to high productive firms concluded by Andrews et al. (2021b). And results for Slovenia, Finland and Netherlands are more in line with Morikawa (2021), who investigated the relationship between government relief programs take-up and firm productivity in Japan.

Table 6.

Probability of receiving support – by macro-sector.

	Wage subsidies				Overall subsidies
Variables	Croatia	Netherlands	Slovakia	Slovenia	Finland	Netherlands
Construction	−0.1249***	−0.2056***	−0.0724***	−0.1229***	−0.1807***	−0.1919***
	(0.0070)	(0.0063)	(0.0066)	(0.0102)	(0.0054)	(0.0065)
Wholesale and retail trade	−0.0138**	0.0020	0.0344***	0.0784***	−0.0115**	0.0262***
	(0.0061)	(0.0054)	(0.0057)	(0.0091)	(0.0058)	(0.0053)
Transportation and storage	−0.0733***	−0.0604***	−0.0179**	−0.0550***	−0.0309***	−0.0425***
	(0.0089)	(0.0082)	(0.0081)	(0.0122)	(0.0066)	(0.0083)
Accommodation and food service activities	0.1447***	0.2968***	0.2647***	0.3059***	0.4362***	0.3215***
	(0.0066)	(0.0080)	(0.0092)	(0.0121)	(0.0081)	(0.0072)
Information and communication	−0.2225***	−0.1176***	−0.1217***	−0.1322***	0.0338***	−0.0846***
	(0.0093)	(0.0070)	(0.0076)	(0.0134)	(0.0081)	(0.0071)
Real estate activities	−0.1246***	0.0476	−0.0885***	−0.0545***	-	0.0326
	(0.0134)	(0.2435)	(0.0085)	(0.0195)	-	(0.2406)
Professional, scientific and technical activities	−0.0865***	−0.1393***	−0.0613***	−0.0824***	−0.0551***	−0.1248***
	(0.0066)	(0.0057)	(0.0059)	(0.0096)	(0.0061)	(0.0057)
Administrative and support service activities	0.0322***	−0.0522***	−0.0367***	0.0866***	−0.0737***	−0.0117
	(0.0091)	(0.0072)	(0.0068)	(0.0164)	(0.0071)	(0.0073)
Control variables:
Productivity	Yes	Yes	Yes	Yes	Yes	Yes
Size class	Yes	Yes	Yes	Yes	Yes	Yes
Observations	71,180	99,925	76,005	30,701	90,855	99,925

Note: The table shows average marginal effects from the logit regression for binary dummy representing receipt of COVID-19 government support in 2020. Lagged explanatory variables from year 2019. Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

Table 7.

Probability of receiving support – by labour intensity.

	Wage subsidies				Overall subsidies
Variables	Croatia	Netherlands	Slovakia	Slovenia	Finland	Netherlands
Wage share	0.0312***	−0.0024	0.0612***	−0.0174***	0.0184***	−0.0039*
	(0.0028)	(0.0023)	(0.0019)	(0.0044)	(0.0023)	(0.0023)
Control variables:
Productivity	Yes	Yes	Yes	Yes	Yes	Yes
Sector	Yes	Yes	Yes	Yes	Yes	Yes
Size class	Yes	Yes	Yes	Yes	Yes	Yes
Observations	71,180	99,925	76,005	30,701	90,855	99,925

Note: The table shows average marginal effects from the logit regression for binary dummy representing receipt of COVID-19 government support in 2020. Lagged explanatory variables from year 2019. Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

Table 8.

Relative size of support – by productivity.

	Wage subsidies						Overall subsidies
Variables	Croatia	Netherlands		Slovakia		Slovenia	Finland	Netherlands
Labour productivity	−0.0484***		−0.0226***		−0.0078***	−0.0283***	−0.0239***	−0.0507***
	(0.0005)	(0.0004)		(0.0002)		(0.0007)	(0.0008)	(0.0010)
Constant	−0.2065***	−0.1397***		0.0452***		0.1256***	0.1656***	0.3242***
	(0.0021)	(0.0020)		(0.0010)		(0.0025)	(0.0033)	(0.0049)
Control variables:
Sector	Yes	Yes		Yes		Yes	Yes	Yes
Size class	Yes	Yes		Yes		Yes	Yes	Yes
Observations	44,523	43,193		23,986		14,838	28,481	51,785
R-squared	0.1034	0.2119		0.0764		0.1660	0.1643	0.1309

Note: The table shows coefficients of OLS regressions for supported firms with the share of firm subsidies on revenue as dependant variable. Continuous variables in logarithm. Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

Table 5.

Probability of receiving support – by productivity.

	Wage subsidies				Overall subsidies
Variables	Croatia	Netherlands	Slovakia	Slovenia	Finland	Netherlands
Labour productivity	0.0202***	−0.0690***	0.0213***	−0.0673***	−0.0280***	−0.0918***
	(0.0020)	(0.0019)	(0.0013)	(0.0045)	(0.0020)	(0.0019)
Control variables:
Sector	Yes	Yes	Yes	Yes	Yes	Yes
Size class	Yes	Yes	Yes	Yes	Yes	Yes
Observations	71,180	99,925	76,005	30,701	90,855	99,925

Note: The table shows average marginal effects from the logit regression for binary dummy representing receipt of COVID-19 government support in 2020. Lagged explanatory variables from year 2019. Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

As shown in Figs. 4 and Fig. 5 , the relationship between the probability of receiving support and firm productivity is non-linear. The positive relationship increases up to the 5th or 7th deciles and then it declines in Croatia and Slovakia. In Slovenia and Finland, the overall negative relationship between productivity and probability of being supported is driven by the upper part of the productivity distribution, when only the firms belonging to the highest probability deciles have significantly lower chance to be supported. In the Netherlands, only firms in the 2nd and 3rd productivity deciles tend to be supported with higher probability and then the probability steeply declines.22

Fig. 4.

Firm probability of receiving employment support – by productivity deciles.

Note: Within country – firms assigned to deciles of the country-level distribution of labour productivity. Source: Authors’ calculations based on micro-data from Croatia, Finland, Netherlands, Slovakia and Slovenia.

Fig. 5.

Firm probability of receiving overall support – by productivity deciles.

Note: Within country – firms assigned to deciles of the country-level distribution of labour productivity. Source: Authors’ calculations based on micro-data from Croatia, Finland, Netherlands, Slovakia and Slovenia.

Allocation probability by industry and firm-size

In line with the nature of the corona crisis that hit mostly in-person services, our analysis confirms that firms supplying accommodation and food services have the highest chance to be supported. This finding holds for all considered countries and both types of subsidies. Conditional on firm size and productivity, these firms have up to 0.44 percentage points higher probability to be supported than manufacturing firms.

The allocation of subsidies by firm size is more homogenous across countries. We find that larger firms receive support with higher probability in all countries and the relationship is statistically significant for both employment and overall subsidies.23

Given the nature of the Covid-19 pandemic and various containment measures, including lockdowns, labour intensive firms suffered more. Our firm-level estimates controlling for firm size and sector show that firms with higher share of wages on revenue had higher probability to receive subsidies in Croatia, Slovakia and Finland.

4.3. Firm-level size of the support

The conditional firm probability of being supported tells only a part of the story. It needs to be accompanied by a conditional analysis of the size of the subsidy at the firm-level. Following the structure of the previous section, we start with the analysis of relative subsidy distribution by firm productivity and continue with the analysis by other main firm characteristics.24

We estimate Eq. (8)

$${\mathrm{Y}}_t=\alpha +\beta X_{t-n}$$ (8)

where Y_t denotes the relative size of the firm-level subsidy with respect to revenue and X_t-n is a row vector of explanatory variables (including sector and size controls) and β is the corresponding column vector of regression coefficients.25

We find several common patterns across countries. More productive firms received smaller relative subsidies. The support decreases also with firm size. More labour-intensive firms and firms from the most severely hit industries received higher support.

Role of firm productivity

Our conditional OLS estimates for supported firms show that the relative size of the support decreases with firm productivity. The relationship between the size of the subsidies and firm productivity is more linear than the relationship between productivity and the probability of being supported. This holds especially for the employment subsidies (see Fig. A1 and A2 in the Appendix for details).

Fig. A1.

Relative size of the employment support – by productivity deciles.

Note: Within country – firms assigned to deciles of the country-level distribution of labour productivity. Conditional OLS estimates for supported firms. Source: Authors’ calculations based on micro-data from Croatia, Netherlands, Slovakia and Slovenia.

Fig. A2.

Relative size of the overall direct support – by productivity deciles.

Note: Within country – firms assigned to deciles of the country-level distribution of labour productivity. Conditional OLS estimates for supported firms. Source: Authors’ calculations based on micro-data from Finland and Netherlands.

Role of firm industry and size

Our conditional estimates controlling for firm size, region and productivity (Table 9 ) confirm the larger relative size of support allocated to firms supplying accommodation and food services in most of the countries. At the same time, the macro sectors receiving the highest relative support differ across countries and types of support. Finland and Netherlands allocated higher relative subsidies to ICT sector.26 In Croatia and Slovenia, the largest relative amount reached Real estate activities and in Slovakia it was Professional, scientific and technical activities.

Table 10.

Relative size of support – by firm wage share.

	Wage subsidies				Overall subsidies
Variables	Croatia	Netherlands	Slovakia	Slovenia	Finland	Netherlands
Wage share	0.0144***	0.0250***	0.0081***	0.0217***	0.0101***	0.0443***
	(0.0008)	(0.0004)	(0.0003)	(0.0006)	(0.0009)	(0.0011)
Constant	0.1972***	0.1570***	0.0510***	0.1170***	0.1729***		0.3336***
	(0.0021)	(0.0019)	(0.0010)	(0.0024)	(0.0034)		(0.0047)
Control variables:
Productivity	Yes	Yes	Yes	Yes	Yes		Yes
Sector	Yes	Yes	Yes	Yes	Yes		Yes
Size class	Yes	Yes	Yes	Yes	Yes		Yes
Observations	44,523	43,193	23,961	14,838	28,476		51,785
R-squared	0.1916	0.2649	0.0983	0.2318	0.1675		0.1486

Note: The table shows coefficients of OLS regressions for supported firms with the share of firm subsidies on revenue as dependant variable. Ownership information not available for Slovenia. Continuous variables in logarithm. Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

Table 9.

Relative size of support – by macro sectors.

	Wage subsidies				Overall subsidies
Variables	Croatia	Netherlands	Slovakia	Slovenia	Finland	Netherlands
Construction	0.0029*	−0.0085***	−0.0055***	−0.0027*	−0.0197***	−0.0127***
	(0.0017)	(0.0014)	(0.0010)	(0.0015)	(0.0023)	(0.0034)
Wholesale and retail trade	−0.0082***	−0.0191***	−0.0018**	−0.0028**	−0.0221***	−0.0359***
	(0.0015)	(0.0009)	(0.0008)	(0.0012)	(0.0020)	(0.0024)
Transportation and storage	0.0142***	−0.0002	−0.0054***	−0.0004	−0.0225***	0.0030
	(0.0022)	(0.0015)	(0.0012)	(0.0017)	(0.0023)	(0.0038)
Accommodation and food service activities	0.0117***	0.0264***	0.0040***	0.0042***	−0.0331***	0.0459***
	(0.0017)	(0.0013)	(0.0010)	(0.0015)	(0.0021)	(0.0033)
Information and communication	0.0000	0.0337***	0.0031**	0.0101***	0.0708***	0.0794***
	(0.0026)	(0.0014)	(0.0014)	(0.0021)	(0.0026)	(0.0034)
Real estate activities	0.0324***	−0.0160	0.0012	0.0360***		−0.0311
	(0.0035)	(0.0420)	(0.0015)	(0.0028)	N/A	(0.1138)
Professional, scientific and technical activities	0.0081***	0.0250***	0.0041***	0.0123***	0.0398***	0.0453***
	(0.0016)	(0.0011)	(0.0009)	(0.0014)	(0.0022)	(0.0027)
Administrative and support service activities	0.0195***	0.0245***	0.0029***	0.0032	0.0052*	0.0285***
	(0.0023)	(0.0013)	(0.0010)	(0.0020)	(0.0027)	(0.0032)
Constant	−5.5331***	−5.5134***	0.0440***	0.1248***	0.1605***	−5.3014***
	(0.3031)	(0.3032)	(0.0011)	(0.0025)	(0.0035)	(0.6310)
Control variables:
Productivity	Yes	Yes	Yes	Yes	Yes	Yes
Size class	Yes	Yes	Yes	Yes	Yes	Yes
Region	Yes	No	Yes	Yes	Yes	No
Observations	44,523	43,193	23,986	14,838	28,469	51,785
R-squared	0.1034	0.1101	0.0768	0.1664	0.1673	0.0492

Note: The table shows coefficients of OLS regressions for supported firms with the share of firm subsidies on revenue as dependant variable. Manufacturing represents a base value for macro sectors. Firm location not available for the Netherlands. Continuous variables in logarithm. Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

As documented in Table A5 and Table A6 in Appendix, the relative size of support decreases with firm size and firm age. And these relationships are statistically significant for all countries and both types of subsidies.

Role of labour intensity

In line with the nature of the employment support, we find that more labour-intensive firms receive not only higher employment subsidies, but also higher overall subsidies.

The analysis of the allocation of the subsidies reveals several interesting findings. The group of viable and highly productive firms received a much higher share of subsidies than the groups of low productive, declining or zombie firms. These aggregate observations originating from our cluster analysis are partially supported by our analysis at the firm-level. The relationship between firm probability to be subsidised and its level of productivity is concave – non-linear and heterogeneous across countries in its average effect. The relationship between firm probability and the size of firm-level support adjusted by revenue is negative across all countries. As confirmed by the accompanying analysis of other firm characteristics, the allocation of the Covid-19 government support to individual firms was driven mostly by their sectoral and size characteristics. Although, the general distribution rules, based mostly revenue decline thresholds, resulted in an efficient distribution of the Covid-19 government support to the firms from the most severely hit industries, the support may have also reached less productive firms with higher probability or size.

5. Consequences for productivity

Covid-19 related lockdowns and temporary supply chain disruptions resulted in significant annual declines in sales in most developed countries. As documented by Dhyne and Duprez (2021) the health crisis had a greater adverse impact on a larger percentage of firms than the 2008–2009 financial crisis. Although, firms tried to compensate the gaps in revenue by adjusting their costs, most of the industries experienced declines in value added. With generally lower elasticities of labour costs or employment than elasticities of material costs to sales, the majority of firms recorded lower labour productivity in 2020.

Based on Eurostat estimates, aggregate EU labour productivity declined by 4.6% in 2020.27 All countries analysed in this study recorded smaller declines, only Croatia faced a somewhat steeper drop in overall labour productivity.28

The contraction of economic activity during the pandemic varies dramatically across sectors and the pattern of the sectoral impact is very different from normal recessions (EC 2021). It is well documented that sectors that require physical proximity, such as the cultural and creative industries, or accommodation and food services, have been the hardest hit by the crisis. While domestic containment measures and the voluntary behavioural response of individuals had a severe impact on e.g. recreational services, the mostly short-term retrenchment in manufacturing was largely driven by external factors.29

Available data on sectoral developments suggests that the most significant reduction in revenue, value added and consequently also in labour productivity took place in the sectors with initially low productivity levels. As documented in Fig. 6 , showing EU productivity by a main industry breakdown,30 the steepest drop in labour productivity was recorded in the least productive sector of Arts, entertainment and recreation services.31 At the same time, this sector creates only between 2% to 4% gross value added depending on the country. Fig. 7

Fig. 6.

EU labour productivity by sectoral breakdown.

Note: The left-hand figure presents EU-27 labour productivity from 2019 in logarithm versus annual difference in labour productivity logarithms from year 2020. The right-hand figure presents sectoral EU-27 gross value- added shares in 2020. Sectors displayed: (A) Agriculture, forestry and fishing, (C) Manufacturing, (F) Construction, (G-I) Wholesale and retail trade, transport, accommodation and food service activities, (J) Information and communication, (K) Financial and insurance activities, (L) Real estate activities, (M-N) Professional, scientific and technical, activities; administrative and support service activities, (O-Q) Public administration, defence, education, human health and social work activities, (R-U) Arts, entertainment and recreation; other service activities. Source: Eurostat, authors’ calculations.

Fig. 7.

Annual growth in micro-aggregated productivity with and without support in the pandemic year (in%) – all firms.

Note: Projected values based on firm-level value added from 2019, industry level changes in sales in 2020 and cost elasticities to sales from Maurin and Pal (2020). Only sustaining firms, entries and exits not considered. Source: Authors’ calculations.

Micro-data, compiled with a significant time lag, remains still unavailable. To shed more light on the granular consequences of the pandemic on productivity, we follow Lalinsky and Pal (202) and utilize cost elasticities to sales estimated by Maurin and Pal (2020). Our projections employing pre-pandemic firm-level value-added figures, together with industry level sales developments in 2020 and industry level cost elasticities to sales, suggest that significant pandemic-time drops in sales and value added were accompanied by reductions in the number of employees.32

Employment subsidies substituted part of the lost income and sustained value added, at the cost of smaller adjustments in number of employees, however. The aggregate changes in value added and employment with and without government support suggest that the overall impact of Covid-19 subsidies on productivity was relatively mild in all countries, except the Netherlands.33 The significantly stronger effect of subsidies in the Netherlands is to a large extent driven by much larger scale of employment support that was at the same time targeting relatively high share of firms.34 In the vein of the labour productivity definition, subsidies supplementing a larger part of the lost income, helped to sustain value added and mitigate the decline in productivity.

The Covid-19 shock did not hit firms homogenously. Besides the standard accounting framework, one might find it important to learn to what extent the pandemic-time decline in productivity is driven by individual firm behaviour and to what extent it can be explained by reallocation of resources and productivity between firms. In line with the literature (e.g. Bloom et al., 2020), we may expect the drop in productivity to be predominantly driven by a temporary deterioration in within-firm productivity. This part of the analysis will remain a topic of further research until the actual firm-level data for the entire pandemic period becomes available.

6. Conclusion

Supporting companies with public funds was a key part of the economic policy response to the Covid-19 crisis. Without this strong response, the recession could have been much deeper and longer than what it turned out to be. The considerable size of the funds spent on business support, their distribution on a fast schedule and the possible harmful side effects of support raise legitimate questions about the targeting and effectiveness of support.

The Covid-19 crisis has hit companies unevenly across and even within sectors. Consequently, macro or industry level data only partially uncover the effects of the pandemic and related support measures on the economy. Therefore, the targeting and effectiveness of the Covid-19 subsidies should be examined on the basis of firm-level data.

By employing information on actual business developments and early available firm-level information on the distribution of the state aid, we contribute to the discussion on the impact of the Covid-19 pandemic and government support on productivity. We analyse the experience of a group of five EU member states representing both less and more advanced economies, as well as economies experiencing mild and severe economic consequences, or weak and strong policy responses to the pandemic.

We bring the first in-depth cross-country firm-level evidence on the allocation of government employment support and other direct subsidies implemented in 2020, as we decompose the overall allocation of government subsidies to the firm-level probability of being supported, but also the relative firm-level size of the support.

We find that, although, the conditions and scale of support differ across countries, a relatively large share of subsidies has been allocated to “deserving” productive and growing firms in temporary need of support, while only a small share of support has accrued to zombies or declining firms.

Our analysis points to the decisive role of the sectoral (and size) characteristics, when firms from the most severely hit industries (firms supplying accommodation and food services) had the highest chance to be supported and received higher relative size of subsidies across countries. Otherwise, the decomposition reveals different findings between the two “margins”.

In terms of the relative size of the support, after controlling for firm size and sector, only more labour-intensive firms showed consistently higher support. More productive or growing firms received smaller relative subsidies. The support adjusted by revenue decreased also with firm size or firm age.

The analysis of the firm probability to be subsidized shows fewer common patterns across countries and leads to the conclusion that larger or older firms had a higher chance to be supported. However, our cross-country analysis based on harmonized hard data from five EU countries confirms that the relationship between firm productivity and probability of receiving support differs across countries, as suggested by earlier studies focusing mostly on individual countries and drawing on survey data.

Relatively timely and efficient state aid has probably reduced not only the long-run scarring effect of the pandemic on the labour market, but also on output and productivity. Importantly, our further findings suggest that Covid-19 state aid has positively affected productivity and the magnitude of the impact appears to largely depend on the country-specific characteristics, including sectoral composition. However, the implemented stimulus packages only partially offset the large negative shock of the pandemic on productivity.

These results underscore important policy consequences. In the acute crisis, productivity issues and the renewal of the economy were hardly the main concern of policy makers, as the main objective of the subsidies was, with good reason, to prevent an economic collapse. Under normal circumstances, public support to firms should be more tightly linked to the long-run development and productivity of the economy, addressing possible market failures and favouring companies with potentially positive spillovers to the macroeconomy.

Appendix

Fig. A3.

Overall allocation of wage subsidies by sectors and firm productivity.

Note: Colours represent deciles of firm productivity. Vertical axis shows value of subsidies in thousands of euro. Horizontal axis represents macro sectors: (C) Manufacturing; (F) Construction; (G) Wholesale and retail trade; (H) Transportation and storage; (I) Accommodation and food service activities; (J) Information and communication; (L) Real estate activities; (M) Professional, scientific and technical activities; (N) Administrative and support service activities. Source: Authors’ calculations based on micro-data from Croatia, Netherlands, Slovakia and Slovenia.

Table A1.

Description of explanatory variables.

Variable	Description
Labour productivity	Value added divided by number of employees
Wage share	Labour costs divided by revenue
Firm growth class	High – firm with annual growth of employment more than 20% over period of the last 3 years Normal – firm with positive annual growth of employment more less than 20% over period of the last 3 years Negative – firm with negative or zero annual growth of employment over period of the last 3 years
Control (dummy) variables
Firm age category	Start-up – less than 3 years Young – from 3 to 4 years Mature – from 5 to 24 years Old- 25 and more years
Size class	0–9 employees 10–19 employees 20–49 employees 50–249 employees 250 and more employees
Region	Croatia – Jadranska Hrvatska, Kontinentalna Hrvatska (NUTS 2016) Finland – Länsi-Suomi, Helsinki-Uusimaa, Etelä-Suomi, Pohjois- ja Itä-Suomi, Åland Slovakia – Bratislavský kraj, Západné Slovensko, Stredné Slovensko, Východné Slovensko Slovenia – Vzhodna Slovenija, Zahodna Slovenija
Sector	Manufacturing; Construction; Wholesale and retail trade; Transportation and storage; Accommodation and food service activities; Information and communication; Real estate activities; Professional, scientific and technical activities; Administrative and support service activities

Table A2.

Cost elasticities to sales.

NACE Rev.2	ISIC Code	Industry	Labour cost elasticity to sales (βs)	Material cost elasticity to sales (αs)
1–3	A	Agriculture	0.298	0.668
5–9	B	Mining	0.353	0.824
10–12	CA	Food manufacturing	0.437	0.861
13–15	CB	Textiles	0.466	0.943
16–18	CC	Wood	0.415	0.911
19	CD	Coke and petroleum	0.324	1.461
20	CE	Chemicals	0.370	0.889
21	CF	Pharmaceuticals	0.266	0.776
22–23	CG	Rubber and plastic	0.431	0.966
24–25	CH	Basic metal	0.468	0.987
26	CI	Manuf. Of computer electronics	0.381	0.881
27	CJ	Manuf. Of electrical equipment	0.398	1.088
28	CK	Machinery	0.420	0.992
29–30	CL	Transport equipment	0.420	0.851
31–33	CM	Other manufacturing	0.478	0.957
35	D	Electricity and gas	0.276	0.725
36–39	E	Water	0.381	0.855
41–43	F	Construction	0.423	0.804
45–47	G	Trade	0.386	0.767
49–53	H	Transportation	0.479	0.848
55–56	I	Accommodation and food services	0.569	0.786
58–60	JA	Publishing	0.357	0.636
61	JB	Telecommunication	0.353	0.698
62–63	JC	IT	0.436	0.700
68	L	Real estate	0.293	0.637
69–71	MA	Legal and accounting	0.352	0.645
72	MB	R&D	0.309	0.640
73–75	MS	Other professional services	0.382	0.701
86–88	Q	Health	0.512	0.779
90–93	R	Art and recreation	0.393	0.638

Source: Maurin and Pal (2020).

Table A3.

Probability of receiving support – by firm size.

	Wage subsidies				Overall subsidies
Variables	Croatia	Netherlands	Slovakia	Slovenia	Finland	Netherlands
Firm size	0.0630***	0.0589***	0.0927***	0.0526***	0.0495***	0.0438***
	(0.0016)	(0.0010)	(0.0012)	(0.0023)	(0.0009)	(0.0011)
Control variables:
Productivity	Yes	Yes	Yes	Yes	Yes	Yes
Sector	Yes	Yes	Yes	Yes	Yes	Yes
Observations	71,180	99,925	76,005	30,701	90,855	99,925

Note: The table shows average marginal effects from the logit regression for binary dummy representing receipt of COVID-19 government support in 2020. Lagged explanatory variables (logarithm of number of employees) from year 2019. Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

Table A4.

Probability of receiving support – by firm age.

	Wage subsidies				Overall subsidies
VARIABLES	Croatia	Netherlands	Slovakia	Slovenia	Finland	Netherlands
Firm age	0.0164***	0.0096***	0.0279***	0.0220***	N/A	−0.0089***
	(0.0020)	(0.0016)	(0.0021)	(0.0032)		(0.0016)
Control variables:
Productivity	Yes	Yes	Yes	Yes		Yes
Sector	Yes	Yes	Yes	Yes		Yes
Size class	Yes	Yes	Yes	Yes		Yes
Observations	59,497	98,147	74,687	30,197		98,147

Note: The table shows average marginal effects from the logit regression for binary dummy representing receipt of COVID-19 government support in 2020. Lagged explanatory variables (logarithm of age in years) from year 2019. Age of Finnish firms not available. Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

Table A5.

Relative size of support – by firm size.

	Wage subsidies						Overall subsidies
Variables	Croatia	Netherlands		Slovakia		Slovenia	Finland	Netherlands
Firm size	−0.0484***		−0.0226***		−0.0078***	−0.0283***	−0.0239***	−0.0507***
	(0.0005)	(0.0004)		(0.0002)		(0.0007)	(0.0008)	(0.0010)
Constant	−0.2065***	−0.1397***		0.0452***		0.1256***	0.1656***	0.3242***
	(0.0021)	(0.0020)		(0.0010)		(0.0025)	(0.0033)	(0.0049)
Control variables:
Productivity	Yes	Yes		Yes		Yes	Yes	Yes
Sector	Yes	Yes		Yes		Yes	Yes	Yes
Observations	44,523	43,193		23,986		14,838	28,481	51,785
R-squared	0.1034	0.2119		0.0764		0.1660	0.1643	0.1309

Note: The table shows coefficients of OLS regressions for supported firms with the share of firm subsidies on revenue as dependant variable. Lagged number of employees in logarithm. Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

Table A6.

Relative size of support – by firm age.

	Wage subsidies				Overall subsidies
VARIABLES	Croatia	Netherlands	Slovakia	Slovenia	Finland	Netherlands
Firm age	−0.0119***	−0.0290***	−0.0089	−0.0014*	N/A	−0.1000***
	(0.0008)	(0.0016)	(0.0065)	(0.0008)		(0.0045)
Constant	0.2058***	0.3156***	0.1971***	0.1293***		0.8308***
	(0.0028)	(0.0070)	(0.0207)	(0.0030)		(0.0200)
Control variables:
Productivity	Yes	Yes	Yes	Yes		Yes
Sector	Yes	Yes	Yes	Yes		Yes
Size class	Yes	Yes	Yes	Yes		Yes
Observations	37,100	42,511	23,646	14,621		50,877
R-squared	0.0972	0.0909	0.0038	0.1609		0.0377

Note: The table shows coefficients of OLS regressions for supported firms with the share of firm subsidies on revenue as dependant variable. Lagged firm age (years) in logarithm. Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1.

Table A7.

Classification of sectors according to technology and knowledge intensity.

Industry classification	Nace 2-digit industry	Description
High-medium technology and knowledge intensive services	20–21	Manufacture of basic pharmaceutical products and pharmaceutical preparations; Manufacture of chemicals and chemical products
	26 - 30	Manufacture of computer, electronic and optical products; Manufacture of electrical equipment; Manufacture of machinery and equipment n.e.c.; Manufacture of motor vehicles, trailers and semi-trailers; Manufacture of other transport equipment
	50–51	Water transport; Air transport;
	58–63	Publishing activities: Motion picture, video and television program production, sound recording and music publish activities; Programming and broadcasting activities; Telecommunications; computer programming, consultancy and related activities; Information service activities
	64–66	Financial and insurance activities
	69–75	Legal and accounting activities; Activities of head offices, management consultancy activities; Architectural and engineering activities, technical testing and analysis; Scientific research and development; Advertising and market research; Other professional, scientific and technical activities; Veterinary activities
	78,80,84–93	Employment activities; Security and investigation activities; Public administration and defence, compulsory social security; Education, Human health and social work activities; Arts, entertainment and recreation.
Medium-low technology and less knowledge intensive services	19	Manufacture of coke and refined petroleum products
	22–25	Manufacture of rubber and plastic products; Manufacture of other non-metallic mineral products; Manufacture of basic metals; Manufacture of fabricated metals products, excepts machinery and equipment
	33	Repair and installation of machinery and equipment
	10–18	Manufacture of food products, beverages, tobacco products, textile, wearing apparel, leather and related products, wood and of products of wood, paper and paper products, printing and reproduction of recorded media
	31–32	Manufacture of furniture; Other manufacturing
	45–47,49,52–53,55–56	Wholesale and retail trade; Repair of motor vehicles and motorcycles (section G); Land transport and transport via pipelines; Warehousing and support activities for transportation; Postal and courier activities; Accommodation and food service activities (section I)
	68,77,79,81,82	Real estate activities; Rental and leasing activities; Travel agency, tour operator reservation service and related activities; Services to buildings and landscape activities; Office administrative, office support and other business support activities;
	94–99	Activities of membership organization; Repair of computers and personal and household goods; Other personal service activities; Activities of households as employers of domestic personnel; Undifferentiated goods- and services-producing activities of private households for own use; Activities of extraterritorial organizations and bodies

Notes: The shows the classification of NACE Rev.2 2-digit sectors according to technology and knowledge intensity.

Source: Eurostat.

Table A8.

Micro-aggregated labour productivity growth and its components in the pandemic year (in%).

Country	Value added	Employment	Labour productivity
Croatia	−12.58	−4.98	−8.00
Netherlands	−9.25	−2.53	−6.89
Slovakia	−5.30	−3.09	−2.30
Slovenia	−10.60	−3.54	−7.32
Finland	−4.25	−1.83	−2.46
Netherlands	−9.25	−2.53	−6.89

Note: Based on projected firm-level values calculated using cost elasticities to sales and sectoral turnover index.

Source: Authors’ calculations based on micro-data from Croatia, Finland, Slovakia and Slovenia.

Data Availability

• The data that has been used is confidential.