Julian Oliver Dörr et al., Small firms and the COVID-19 insolvency gap (2021)
Summary
Article by Julian Oliver Dörr, Georg Licht and Simona Murmann of the ZEW Leibniz Centre in Small Business Economics, published online 6 July 2021, on the insolvencies that COVID-19 aid in Germany prevented. Using data on about 1.5 million companies, they estimate an insolvency gap of around 25,000 firms, against about 16,300 actual insolvencies in 2020. The gap is concentrated among small firms that were already financially weak before the crisis. The authors attribute it to largely indiscriminate liquidity subsidies and the temporary suspension of the duty to file for insolvency, and they argue this blunted the cleansing effect that moves workers and capital to stronger firms.
Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used
Cite this document
“Julian Oliver Dörr et al., Small firms and the COVID-19 insolvency gap (2021).”
Original at link.springer.com. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/julian-oliver-d-rr-et-al-small-firms-and-the-covid-19-insolvency-gap-2-4ac92b00e6/
, SHA-256 58d75170c66642286ab5c58e27133650c645eac76e2038e747659c94e3857a42.Full text
Small Bus Econ (2022) 58:887–917
https://doi.org/10.1007/s11187-021-00514-4
Small firms and the COVID-19 insolvency gap
Julian Oliver Dörr · Georg Licht ·
Simona Murmann
Accepted: 12 May 2021 / Published online: 6 July 2021
© The Author(s) 2021
Abstract COVID-19 placed a special role on fiscal long-term implications on entrepreneurship and eco-
policy in rescuing companies short of liquidity from nomic recovery.
insolvency. In the first months of the crisis, SMEs as Plain English Summary This study analyzes the
the backbone of Germany’s economy benefited from extent to which the strong policy support to com-
large and mainly indiscriminate aid measures. Avoid- panies in the early phase of the COVID-19 crisis
ing business failures in a whatever-it-takes fashion has prevented a large wave of corporate insolvencies.
contrasts, however, with the cleansing mechanism of Using data of about 1.5 million German companies,
economic crises: a mechanism which forces unviable it is shown that it was mainly smaller firms that
firms out of the market, thereby reallocating resources experienced strong financial distress and would have
efficiently. By focusing on firms’ pre-crisis finan- gone bankrupt without policy assistance. In times of
cial standing, we estimate the extent to which the crises, insolvencies usually allow for a reallocation of
policy response induced an insolvency gap and ana- employees and capital to more efficient firms. How-
lyze whether the gap is characterized by firms which ever, the analysis reveals that this ‘cleansing effect’ is
were already struggling before the pandemic. With the hampered in the current crisis as the largely indiscrim-
policy measures being focused on smaller firms, we inate granting of liquidity subsidies and the temporary
also examine whether this insolvency gap differs with suspension of the duty to file for insolvency have
respect to firm size. Our results show that the COVID- caused an insolvency gap that is driven by firms which
19 policy response in Germany has triggered a back- were already in a weak financial position before the
log of insolvencies that is particularly pronounced crisis. Overall, the insolvency gap is estimated to
among financially weak, small firms, having potential affect around 25,000 companies, a substantial number
compared to the around 16,300 actual insolvencies in
2020. In the ongoing crisis, policy makers should pre-
fer instruments favoring entrepreneurs who respond
J. O. Dörr () · G. Licht · S. Murmann innovatively to the pandemic instead of prolonging the
Department of Economics of Innovation and Industrial survival of near-insolvent firms.
Dynamics, ZEW Leibniz Centre for European Economic
Research, L7 1, 68161 Mannheim, Germany
e-mail: julian.doerr@zew.de
Keywords COVID-19 · Liquidity support ·
Insolvency law · Corporate insolvency ·
J. O. Dörr
Department of Econometrics and Statistics, Cleansing effect · SMEs · Entrepreneurship
Justus Liebig University Giessen, Licher Straße 64,
35394 Gießen, Germany JEL Classifications G33 · H12 · H81 · L26
888 J.O. Do¨rr et al.
1 Introduction easing the burden of Small and Medium-sized Enter-
prises (SMEs) (OECD, 2020a). In the European Union
COVID-19 and its unprecedented economic impacts (EU), for instance, member states’ liquidity support
have ground economies worldwide to a halt. As a in form of public loan guarantees and tax deferrals
result of the early lockdown measures to contain the for distressed sectors has increased by an estimated 6
spread of the virus, many companies faced reduced percentage points (pp) of EU GDP compared to pre-
business activity and declining sales, which had an crisis levels (Council of the European Union, 2020).
immediate impact on their liquidity positions. Indeed, In most countries, policy measures have gone beyond
both the negative demand shock paired with a negative deferrals and loan guarantees, including instruments
supply shock in most industries have put numerous such as wage subsidies and adjustments in bankruptcy
companies under severe pressure to keep their oper- regimes. While there is no doubt that a strong policy
ations afloat. Previous crises have taught that small response was necessary to keep the struggling econ-
entrepreneurial firms are particularly prone to con- omy afloat, the need to respond quickly and the sheer
siderable liquidity constraints in deep recessions. For volume of firms seeking assistance left little time
example, literature on the financial crisis of 2007– for policymakers to assess the viability of firms that
2009 shows that especially small and entrepreneurial received early government support. Thus, many of the
enterprises were exposed to a severe liquidity early policy measures were not only unprecedented in
crunch due to the collapse of the interbank market scale but also largely granted indiscriminately with the
and its negative impact on corporate lending (e.g. primary focus to avoid corporate bankruptcies.1 Even
Cowling et al. 2012; Iyer et al. 2014; Lee et al. 2015; though some programs’ eligibility criteria are formally
McGuinness & Hogan 2016). In the COVID-19 cri- linked to pandemic-induced financial distress only,
sis, the early effects of the combined negative supply information asymmetries make drawing a line often
and demand shock are also characterized by a deep difficult in reality. We argue that these circumstances
liquidity shock in the real economy. The decline of have favored a substantial backlog of corporate insol-
trading activities and lack of business revenues made vencies as policy measures have also kept otherwise
many firms dependent on their cash reserves in order insolvent firms in the market. This phenomenon is
to meet their unchanged fixed cost obligations. As referred to as insolvency gap in the remaining of the
smaller and entrepreneurial companies are character- paper.
ized by strong dependence on internally generated The central purpose of this study is to analyze
funds to capitalize their business and provide the liq- whether the early policy response has indeed induced
uidity needed to finance day-to-day operations, both such an insolvency gap and, if so, by which firms
their cash reserves and collaterals for external financ- the gap is mainly driven. We do so by incorporating
ing are generally limited (Cowling et al., 2020). In the Schumpeterian cleansing effect usually observed
times of financial distress as in the current COVID-19 in economic crises into our analysis. In Schumpete-
crisis, this makes small ventures particularly vulnera- rian economics, crises are typically seen as cleansing
ble to financial insolvency (Fairlie, 2020; Bartik et al., mechanism forcing unviable firms out of the market
2020). Recent research suggests that severely affected thereby efficiently reallocating resources to more pro-
small entrepreneurial ventures even seek for alterna- ductive companies. Our hypothesis is that this cleans-
tive financing methods such as bootstrap financing ing mechanism is strongly compromised by the undif-
to keep their businesses alive (Block et al., 2021). ferentiated policy response which favors the survival
Trapped in a situation of thin capital reserves and of otherwise unviable firms. Since in times of crises
lack of collaterals for drawing new credit lines, small
firms face therefore a particularly high risk of business 1 In Germany, for instance, liquidity grants’ ‘application and
failure without the relief through policy intervention. payment process is to be swift and free from red tape’ accord-
Aware of the far-reaching consequences of a wave ing to the Ministry of Finance (Federal Ministry of Finance
of corporate insolvencies, governments in almost all 2020b, para. 2). Moreover, in context of public loan programs it
is stated that ‘the credit approval process does not involve addi-
countries have initiated a series of emergency mea- tional credit risk assessment by the bank’ and that ‘there are
sures to strengthen liquidity positions of their national no requirements for collateral security’ (Federal Ministry for
companies, some of which exclusively focusing on Economic Affairs and Energy 2020, para. 5).
Small firms and the COVID-19 insolvency gap 889
small firms tend to be particularly prone to liquidity Section 6 discusses the implications of our results and
shortages, we believe that the risk of unviable ‘sur- concludes.
vivors’ is especially high among smaller enterprises.
The strong policy focus on SMEs in many countries
(OECD, 2020a) reinforces this hypothesis. Finally, it 2 Related literature
is likely that the prolonged expansion prior to COVID-
19 along with the low interest rate environment have The fast growing literature on business failures in
already accumulated a substantial number of finan- response to the adverse economic impacts of COVID-
cially weak companies before the pandemic (Barrero 19 stresses that the early assistance packages may
et al., 2020). Normally, the COVID-19 crisis would bare high economic costs if they keep unviable firms
have been a ‘natural’ mechanism to force such ailing alive (Kalemli-Ozcan et al., 2020; Barrero et al., 2020;
firms out of the market. Given the interplay between Cowling et al., 2020; Juergensen et al., 2020; OECD,
prolonged expansion and sudden economic decline 2020b; Didier et al., 2021). Kalemli-Ozcan et al.
paired with a strong policy response, our hypothesis (2020), for example, find for a number of European
is that the insolvency gap is strongly driven by small countries that without appropriate targeting of policy
firms with weak financial conditions prior to the crisis. instruments, the fiscal costs of intervention and the
Our contribution to the fast growing literature on number of ‘ghost’ firms kept alive are substantially
the economic effects of the COVID-19 crisis is mani- higher compared to a scenario in which policies tar-
fold. First, we examine the heterogeneity with respect get only ‘viable’ firms. Besides the direct fiscal costs
to firm size in policy makers’ response to the risk associated with indiscriminate policy interventions,
of large-scale business failures. Second, we translate there is yet another source of economic costs asso-
Schumpeter’s theory of the cleansing effect in eco- ciated with keeping unviable firms alive. In Schum-
nomic crises into an empirical assessment by estimat- peterian economics, this may also impede the cleans-
ing the size of a policy-induced insolvency gap using ing effect of creative destruction (see, for example,
firm-specific credit rating data combined with infor- Legrand 2017 and in the COVID-19 context Barrero
mation on insolvency filings. Controlling for updates et al. 2020; Guerini et al. 2020). This effect describes
in a firm’s credit rating, we estimate the insolvency a process in which resources are reallocated from
gap induced by the COVID-19-related policy mea- less efficient and less creative firms to more efficient
sures using a potential outcome setting. Based on ones enhancing overall economic productivity and
pre-crisis observations of no policy intervention com- innovation (Schumpeter, 1942). Typically, this pro-
parable firms with closely matching changes in their cess of efficient resource reallocation is particularly
credit rating are used as control group for the esti- strong in times of economic crisis, allowing viable
mation of counterfactual insolvency rates. Finally, we and innovative firms to gain market share as unprof-
discuss the consequences for entrepreneurship if effi- itable firms exit the market (Caballero and Hammour,
cient resource reallocation and business liquidation 1994; Archibugi et al., 2013; Carreira & Teixeira,
are compromised. 2016). As such, without the intervention of fiscal pol-
The remainder of the paper proceeds as follows. icy, business failures of unviable firms are expected
Section 2 gives an overview of the relevant litera- to be substantial in economic recessions and, given
ture. In Section 3, we discuss the different COVID-19 the strong vulnerability of small and entrepreneurial
support instruments for firms in Germany and empha- firms, the effect is expected to be particularly pro-
size their different orientations depending on firm nounced among smaller businesses. In the current
size. Section 4 introduces the data sources and vari- crisis, however, there is growing public concern that
ables used to estimate the insolvency gap. Moreover, this process of creative destruction and ‘cleanse out’
the framework for the matching of counterfactual of unviable firms is seriously hampered by an increas-
survival states is introduced. Section 5 empirically ing policy-induced ‘zombification’ of the economy
examines the adverse impacts of the pandemic and (see, e.g., The Economist 2020a; The Washington Post
its heterogeneity across firms of different size and 2020). Analyzing only the short-term effects of pol-
sector affiliation. Moreover, it presents the empirical icy aid on firm survival, we do not want to go as far
results of the insolvency gap estimation. Ultimately, as speaking of a zombification which typically refers
890 J.O. Do¨rr et al.
to situations in which credit misallocation by banks (Brouwer, 2006). It appears, however, that the intro-
sustains the survival of de facto insolvent firms over duction of formal restructuring has barely been used
a longer period of time. Still, we hypothesize that the by small firms as the costs of reorganization proceed-
early policy measures with strong focus on SME relief ings are often too high for smaller, financially con-
induced an insolvency gap, defined as backlog of cor- strained companies (Cook et al., 2001; Dewaelheyns
porate insolvencies which are usually to be expected & Van Hulle, 2008). Thus, for small entrepreneurial
in a crisis like this. firms, insolvency declarations often offer no realis-
If efficient resource reallocation and business liqui- tic path towards reorganization but are more likely to
dation are compromised through policy interventions, end in liquidation. Since the prospects of reorganiza-
this has immediate consequences on entrepreneur- tion are low, insolvent small business owners have an
ship. Focusing on Germany, a country where liq- additional incentive not to file for bankruptcy, which
uidity support for SMEs has not only been particu- is why we argue that the temporary filing suspension
larly strong by international standards (Anderson & is disproportionately used by smaller firms.
et al. 2020; OECD, 2020a) but also been accompa- Besides Germany, further countries such as France,
nied by a temporary suspension of the obligation to Spain, Luxembourg, Poland, Portugal, Russia and the
file for insolvency (Federal Ministry of Justice and Czech Republic have temporarily released corporate
Consumer Protection, 2020), we identify the insol- directors and entrepreneurs from their insolvency fil-
vency law as an important institutional determinant for ing obligation in response to the pandemic. Other
entrepreneurship dynamics. Past literature has shown countries such as the USA temporarily raised the debt
that (changes in) the institutional environment have threshold for small businesses eligible to participate in
an important influence on entrepreneurial outcomes reorganization proceedings (Gurrea-Martı́nez, 2020).
(Baumol, 1990; Acs et al., 2008; Peng et al., 2010; Using insolvency data on French firms, Cros et al.
Levie et al., 2014; Chowdhury et al., 2015; Arcuri (2021) find that insolvency rates have substantially
& Levratto, 2020) determining both entrepreneurial fallen below pre-crisis rates. However, they argue that
exit but also firm entry (Melcarne & Ramello, 2020). the selection process to file for insolvency has not
Empirical results suggest that entrepreneur-friendly been distorted during the pandemic because firm char-
insolvency laws, characterized primarily by speed and acteristics that determine failure and survival have
efficiency in liquidation and reorganization processes, remained unchanged compared to pre-crisis times. For
have a positive impact on new firm entry (Chemin, the US economy, Wang et al. (2020) find a sharp
2009; Lee et al., 2011). Moreover, research shows that decline in insolvency filings among small firms, while
the design of bankruptcy laws can favor high-growth bankruptcy proceedings among large firms remain at
entrepreneurship (Estrin et al., 2017; Eberhart et al., normal levels. Despite the eased access to reorga-
2017). In fact, entrepreneurs seem to incorporate the nization for smaller firms, they suggest that small
efficiency of insolvency legislation into their founding businesses see insolvency proceedings only as a last
decision as regions with faster liquidation proceedings resort because successful reorganization is unlikely
appear to be associated with higher levels of busi- and often too costly. In general, official figures show
ness formations and firm growth (Garcı́a-Posada & that corporate insolvency numbers after the outbreak
Mora-Sanguinetti, 2015; Melcarne & Ramello, 2020). of the crisis have strongly decreased compared to
However, entrepreneur-friendly insolvency laws can 2019 levels especially in countries which implemented
also have adverse impacts on start-ups and SMEs as changes in their insolvency frameworks (see Fig. 7 in
refinancing costs may increase and access to credits is the Appendix). This underpins the idea that the large-
tightened by banks accordingly (Djankov et al., 2007; scale governmental support programs have, indeed,
Berger et al., 2011; Rodano et al., 2016). On the exit led to substantial distortions in business dynamics.
side, literature points out that changes in the design of Clearly, the suspension has allowed entrepreneurs
insolvency legislation strongly determine which type with viable business models to stay in the market and
of firms predominantly initiate insolvency proceed- use public liquidity subsidies to avert insolvency. But
ings. In the late 1990s, for instance, various European at the same time, if unprofitable firms do not exit the
countries have introduced formal restructuring pro- market because they are not required to do so, the
cedures to allow reorganization of distressed firms efficient reallocation of resources is impeded. Access
Small firms and the COVID-19 insolvency gap 891
to skilled human capital, physical resources such as revenues. In case of the ‘Soforthilfen’, for instance,
office space, and bank loans is limited for newly enter- only micro-firms with up to 10 employees were eligi-
ing firms when unviable firms congest the market. ble to receive injections between e9,000 and e15,000
This may prevent future entrepreneurs from starting for three months to cover their operational costs
up, but can also discourage existing entrepreneurs (Federal Ministry of Finance, 2020d). These immedi-
from initiating new, more promising ventures. Hence, ate subsidies have been accompanied by a large-scale
we assume that an insolvency gap along with a further stimulus package worth e25 billion covering a sub-
prolongation of aid measures in the ongoing crises is stantial part of SMEs’ fixed operating costs (Federal
likely to result in a decrease in entrepreneurial activity Ministry of Finance, 2020c). Generally, the subsi-
in the longer term. dies were granted in a non-bureaucratic fashion easily
accessible to all micro-businesses and SMEs which
assured that they were suffering financial distress
3 Policy response in Germany because of the COVID-19 pandemic (Federal Govern-
ment of Germany, 2020).
Official figures show that in Germany, the fiscal pol-
icy response to prevent corporate insolvencies due to 3.2 Liquidity loans under public guarantee schemes
crisis-related liquidity bottlenecks is particularly pro-
nounced by international comparison. According to a For SMEs with more than 10 employees the KfW
comparative study of the economic think tank Bruegel, Instant Loan Program has been launched. The pro-
nearly 40% of Germany’s 2019 GDP was spent on gram offers SMEs loans that are fully collateralized
COVID-19 measures to strengthen companies’ liquid- by the state. These loans amount up to 25% of a
ity positions (Anderson and et al. 2020). Compared firm’s 2019 revenues with a cap of e500k for small
with a number of selected OECD countries, this is companies and e800k for medium-sized companies,
the second strongest response in terms of a coun- respectively. No credit risk assessments are taking
try’s overall economic performance (see Fig. 6 in the place and no collaterals are required. The only eligi-
Appendix). In fact, the German Federal Government bility criterion is that the company was profitable in
itself describes the response as the ‘largest assistance 2019 or at least on average profitable between 2017
package in the history of the Federal Republic of and 2019 (Federal Ministry for Economic Affairs and
Germany’ (Federal Ministry of Finance 2020d, 3). Energy, 2020). This fairly broad criterion shows that
From a small business economics view, it is inter- the process is focused on speed and ease applied
esting to see that a number of intervention measures ‘without red tape’ (Federal Ministry of Finance
adopted by the German Federal Government have 2020b, 1) and not on elaborate screening mechanisms
been specifically designed to target SMEs (OECD, that could prevent providing liquidity to unviable
2020a). In the following, we describe the policy firms.
instruments to counter the economic impacts of the Furthermore, the COVID-19 support package
COVID-19 crisis in more detail, focusing on how the includes additional government guarantees on loans
instruments differ with respect to firm size (for a quick for both small and larger businesses, including lower
overview of the measures and possible effects on cor- interest rates for small firms compared to large firms.
porate insolvencies the reader is referred to Table 10 (Federal Ministry for Economic Affairs and Energy,
in the Appendix). 2020). Similar to the Instant Loan Program, the loans
are channeled through commercial banks and the
3.1 Direct liquidity subsidies state-owned bank KfW assumes risk coverage of 80%
for large enterprises and 90% for SMEs with a sim-
As an immediate response to the first lockdown, plified risk assessment (Federal Ministry of Finance,
the Federal Government granted liquidity subsi- 2020a). For commercial banks, this makes lending
dies through direct cash transfers (‘Sofort-’ and to SMEs particularly attractive and, given that they
‘Überbrückungshilfen’). The extent of liquidity sup- only bear 10% of the default risk, further disincen-
port is primarily determined by company size, mea- tivizes comprehensive risk assessments by the issuing
sured by the number of employees or previous bank.
892 J.O. Do¨rr et al.
3.3 Liquidity support through labor cost subsidies accompanied by a temporary amendment to the Ger-
man insolvency law. On March 27, 2020, the Federal
Another form of liquidity support to companies is the Government decided to temporarily suspend the insol-
use of short-time compensations (‘Kurzarbeitergeld’) vency filing obligation in order to avoid a massive
which are direct subsidies on firms’ labor costs. This increase in insolvencies as a result of COVID-19-
instrument has been available for quite some time; induced liquidity shortages. The obligation to file an
however, its eligibility criteria were relaxed in the pan- insolvency has been suspended until September 30,
demic. Now companies with only 10% of employees 2020, with an adjusted extension until the beginning of
working on short-time qualify for the wage subsidy 2021. Although the amended law stipulates that only
(instead of one third) (OECD, 2020b). In addition, those firms that are insolvent or over-indebted due
the subsidy has been increased compared to pre-crisis to the COVID-19 pandemic are temporarily exempt
levels, ranging now from 60 to 87% of the worker’s from insolvency proceedings, policy makers face the
dilemma that it is barely possible to assess whether
last net income. From a company perspective, short-
insolvent non-filers fulfill this eligibility criterion.
time compensations reduce labor costs, allow the
This is particularly true for smaller firms, whose lim-
company to retain specific human capital and avoid
ited disclosure requirements make such an assessment
the costs of new hires and training when the econ-
even harder. While there is no doubt that many viable
omy recovers again. Drawing on literature from the
companies facing illiquidity and over-indebtedness as
Great Recession, the usage of short-time work (STW)
a result of the economic shock will benefit from the
has a positive impact on firm survival (Cahuc et al.,
law change, it also creates loopholes for smaller, unvi-
2018; Kopp & Siegenthaler, 2021) but at the same
able companies to stay in the market and absorb public
time research results suggest that low productivity
liquidity aid.
firms have been taken up STW more often (Giupponi
The two cornerstones of the aid measures—public
& Landais, 2018). From a welfare perspective, this
liquidity support and the amendment of the insol-
may have adverse effects as it impedes the realloca-
vency law—have been implemented simultaneously
tion of workers from low- to high-productivity firms.
as a joint strategy to prevent widespread corporate
Since SMEs tend to be active in more labor-intense
insolvencies. Therefore, we cannot differentiate which
business activities than larger firms (Yang & Chen,
influence the individual measures have on the emer-
2009), it is reasonable to assume that SMEs as well as
gence of a possible insolvency gap. However, we
labor-intense sectors benefit disproportionately from
argue that the policy response must be understood
short-time compensations. Eligibility criteria for STW as a mix of interdependent policy actions that likely
are unrelated to firms’ pre-crisis performance, which would not have been effective in preventing business
allows unviable companies to benefit from the instru- failures had they been implemented separately. In par-
ment as well. ticular, the liquidity provision through state-supported
loans and the temporary suspension of the filing obli-
3.4 Intertemporal liquidity support gation have only had an insolvency-preventing effect
because they were implemented simultaneously and
To further improve the liquidity situation of companies, mutually. Without the filing suspension, companies
authorities have granted tax payment deferrals, allowed would have been discouraged from taking out gov-
lower tax prepayments and suspended enforcement ernment loans as this would have led many of them
measures for tax debts. The tax-related intertempo- into over-indebtedness, which in normal times would
ral liquidity assistance amounts to an estimated e250 have obliged firms to declare insolvency. Likewise,
billion and the policy measure applies equally to all without liquidity provision through easily accessible
company size classes (Anderson & et al. 2020). loans and other subsidies, the sole insolvency sus-
pension would have been ineffective since in light
3.5 Temporary change in insolvency law of strongly diminished turnovers the economic real-
ity of many liquidity constrained firms would have
Finally, the different elements of liquidity provision implied a de facto insolvency. Following this line
which have been granted to German businesses were of reasoning, the effect of liquidity support and
Small firms and the COVID-19 insolvency gap 893
temporary change in insolvency law on (non-) selec- with information concerning the firms’ insolvency sta-
tion into insolvency is best analyzed as a policy mix tus. In the following, we will introduce both data
used to combat the threat of mass insolvencies. While sources and the variables used in this study in more
the insolvency filing suspension allowed both small detail.
and large companies to avert insolvency and possi-
bly survive the crisis by taking advantage of liquidity 4.1.1 Survey data
injections by the state (Federal Ministry of Justice
and Consumer Protection, 2020), it has been shown We employ the survey to primarily assess which
that many of the liquidity support measures directly industries and company sizes are affected most by the cri-
target SMEs or provide indirect channels for espe- sis. Based on a representative random sample of German
cially smaller (and often entrepreneurial) businesses to companies, drawn from the MUP and stratified by
benefit disproportionately. With few screening mecha- firm size and industry affiliation the survey was con-
nisms in place, there is the risk that unviable firms will ducted three times spanning the period in which the
be kept alive, freezing up resources that could be used German insolvency regime was fully suspended.2 The
more productively elsewhere and possibly hampering survey includes questions on COVID-19-related eco-
entrepreneurial activity. nomic effects on various business dimensions. The
This section has highlighted the role of policy sup- collected data has then been supplemented with credit
port to counter the economic consequences of the rating scores from the MUP, which allows to control
pandemic in Germany—a country that has provided for the financial situation of the companies prior to the
substantial assistance to businesses to avoid a wave crisis. As shown in Fig. 1, we use the survey data to
of corporate bankruptcies. It has suggested that the investigate whether the adverse economic impacts of
joint implementation of widespread but undifferen- COVID-19 differ across sectors and firm size classes.
tiated liquidity support strongly focusing on SMEs These results together with the heterogeneity in public
together with the temporary amendment of the insol- aid programs with respect to firm size as outlined in
vency law, is likely to have favored a backlog of Section 3 motivates us to conduct our main empirical
corporate insolvencies particularly pronounced among analysis at the sector-size level.
small and possibly financially weak companies. In the Table 1 shows summary statistics of the relevant
next section, we introduce the data and methodology variables used to construct a COVID-19 Exposure
we use to estimate the existence and extent of such an Index, CEI , reflecting the extent to which firms expe-
insolvency gap. rienced negative impacts in relation to the pandemic.
Firms were asked on a Lickert scale of 0 to 4 in
which of the following areas they experienced nega-
4 Data, variables and methodology tive impacts as a result of the COVID-19 crisis: (1)
decrease in demand, (2) shutdown of production, (3)
4.1 Data and variables supply chain interruption, (4) staffing shortage, (5)
logistical difficulties, (6) liquidity shortfalls.3 From
The study uses two data sources which both originate these six questions we construct CEI as simple sum
from the Mannheim Enterprise Panel (MUP) covering of the response values. The average index is 6.31 out
the near universe of economically active firms in Ger- of a maximum possible value of 24. The most com-
many (Bersch et al., 2014). The first data source is mon and most severe impact relates to the decline in
a survey where the questioned companies have been demand, where respondents reported an average nega-
sampled from the MUP. The survey is used to exam- tive impact of 1.85. Shutdown of production facilities
ine how companies of different size and in different
sectors are affected by the adverse impacts of the 2 The surveys have been conducted in April 2020, in June
COVID-19 pandemic and motivates why we estimate 2020 and in September 2020 spanning the period of the full
the insolvency gap distinguishing between sector affil- suspension of the obligation to file for insolvency and is there-
fore particularly suitable for capturing the early policy-induced
iation and company size. For the estimation of the effects of the crisis.
insolvency gap we use a second data source: a large 3 0 indicates no negative effects, 4 signals strong negative
sample of firm-specific credit rating information along effects.
894 J.O. Do¨rr et al.
Fig. 1 Data sources used in this study. Note: Observations of adverse effects of the pandemic on the sector-size level. Credit
the survey data (companies) and credit rating data (firm-specific rating data is used to estimate the existence of an insolvency gap
rating revisions) originate from the Mannheim Enterprise Panel on the sector-size level
(MUP) data base. Survey data allows to estimate exposure to
and liquidity bottlenecks are also frequently men- to examine the policy-induced heterogeneity of the
tioned consequences. COVID-19 related effects on business failures with
a special focus on possible size differences not only
4.1.2 Credit rating data among SMEs and large enterprises but also within the
group of SMEs. The latter estimation of the insol-
For the purpose of estimating whether the bankruptcy vency gap will be conducted on the sector-size level as
filing behavior has changed significantly as a result displayed in Table 2.
of the crisis-related aid measures and possibly cre- Assuming that the COVID-19 shock and its eco-
ated a backlog of insolvencies, we examine credit nomic consequences on liquidity and insolvency dis-
rating updates of close to all economically active firms tress of German businesses began by the end of March
listed in the MUP.4 The Mannheim Enterprise Panel 2020, we split our sample into a ‘pre-crisis’ period
is particularly suited for an analysis of the insolvency- and a ‘crisis’ period. This cutoff point also captures
related cleansing effect as it is constructed by process- COVID-19 policy dynamics as the German govern-
ing and structuring data collected by Creditreform, ment imposed the first countrywide lockdown that
the leading credit agency in Germany. Creditreform includes a shutdown of most customer service-related
regularly measures and updates the creditworthiness businesses on March 22 and suspended the obligation
of German companies. Overall our sample comprises to file for bankruptcy on March 27 (Federal Min-
2,373,782 credit rating updates of 1,500,764 distinct istry of Justice and Consumer Protection, 2020). Con-
German businesses whose ratings were updated at sequently, the pre-crisis period comprises all credit
least once during the last three years.5 Table 2 shows rating updates which took place between July 2017
that the sample of about 1.5 million companies is and December 2019. The crisis period includes all
very diverse in its industry and size composition. Most observations between April 2020 and end of July
important in the context of this study is the cover- 2020.6 In the later estimation of the insolvency gap,
age of SMEs, which not only is representative for rating updates from the pre-crisis period serve as pool
the German economy (see Table 2), but also allows of control observations. Closely matching credit rating
for a nuanced differentiation between medium-sized, updates from this pool are used to estimate counterfac-
small and micro-enterprises. Therefore, it suits well tual insolvency rates which will be compared against
4 In our analysis a company is defined as economically active if
it has received a credit rating update at least once over the last
three years spanning the period from July 2017 to July 2020. 6 Note that we exclude observations between January 2020 and
5 We observe one and the same company at most three times March 2020 which we see as transitional phase in which assign-
in our sample. Thus, credit rating updates normally do not take ment to either of the two periods is not straightforward. Also
place more often than once per year but may be conducted in a note that July 2020 is the latest month for which we observe
less regular cycle. credit rating information.
Small firms and the COVID-19 insolvency gap 895
Table 1 Descriptive statistics: survey data Table 2 Sample decomposition of credit rating data
Variables N Mean SD Min Max Sector affiliation Size of company Total
COVID-19 Exposure Index (CEI ) 2,344 6.31 5.40 0 24 Micro Small Medium Large (sample)
Questions used for the index calculation Business-related 89.4% 8.3% 1.9% 0.4% 28.6%
(1) Decrease in demand 2,344 1.85 1.56 0 4 services
(2) Lockdown of production 2,344 1.05 1.58 0 4 Manufacturing 84.9% 11.8% 2.7% 0.6% 22.5%
(3) Supply chain interrupted 2,344 0.88 1.24 0 4 Wholesale & retail 83.1% 13.4% 2.9% 0.6% 19.9%
(4) Staffing shortage 2,344 0.64 1.04 0 4 trade
(5) Logistical difficulties 2,344 0.81 1.28 0 4 Health & social 84.8% 10.6% 3.5% 1.1% 7.3%
(6) Liquidity shortfalls 2,344 1.08 1.41 0 4 services
Insurance & banking 93.6% 3.6% 1.8% 1.0% 4.5%
Note: The table shows descriptive statistics of the COVID-19
Accommodation & 88.5% 9.8% 1.6% 0.1% 4.1%
Exposure Index (CEI ). It also displays statistics of the survey
questions used to construct the index catering
Logistics & transport 80.5% 15.3% 3.5% 0.7% 4.1%
Others 82.7% 10.2% 4.6% 2.5% 3.9%
the actual insolvency rates observed after April 1,
Creative industry & 88.9% 8.8% 2.0% 0.3% 1.6%
2020.7
entertainment
For the estimation of insolvency rates, we enrich
our sample of firm-specific credit rating data with Mechanical 54.3% 27.5% 13.0% 5.2% 1.3%
information on the firm’s survival status after it has engineering
received an update on its rating. Information on firm- Food production 64.3% 23.0% 10.3% 2.4% 1.0%
specific survival states is obtained by the online reg- Chemicals & 49.1% 29.1% 16.5% 5.3% 0.7%
ister for bankruptcy filings of the German Ministry pharmaceuticals
of Justice. Besides information identifying the com- Manufacturing of data 58.9% 26.7% 10.9% 3.5% 0.5%
panies which have filed for insolvency, the register processing equipment
also contains the filing date, allowing us to match the Total (sample) 85.2% 11.1% 2.9% 0.8% 100%
most recent rating update that predates the filing date Total (population)a 81.8% 15.1% 2.5% 0.6% 100%
for that particular bankrupt firm. Our overall sample
comprises 15,634 credit rating updates that were fol- Note: The table shows the company size distribution within
sectors (rows) as well as the sector distribution (column ‘Total
lowed by an insolvency and 2,358,148 rating updates
(sample)’) in our credit rating sample. Size classification is
which did not result in an insolvency filing. With determined by number of employees, annual turnover and
this data, we are able to estimate two statistics. First, annual balance sheet total following the recommendation of
we use this information to estimate bankruptcy rates the EU Commission (European Commission, 2003) as outlined
in Table 8 in the Appendix. Sector groups are built to reflect
after the COVID-19 outbreak on the sector-size level
anecdotal heterogeneity in the context of COVID-19. Group-
based on firms for which we observe credit rating ing of sectors is based on EU’s NACE Revision 2 classification
updates during the pandemic. Second, using compara- scheme (European Union, 2006). In Table 9 in the Appendix
ble firms with closely matching credit rating updates an exact mapping of sector groups and NACE divisions can be
found. In all sectors the fraction of SMEs lies far above 90%
in non-crises times as control group, we are able to
which makes the data particularly useful to analyze the effects
estimate counterfactual insolvency rates. Comparing of COVID-19-related policy responses on smaller firms. Also
observed insolvency rates with counterfactual insol- note that the overall size composition of our sample compares
vency rates within each of the sector-size strata allows well against the official size distribution of the population of
German active companies as reported by the Federal Statistical
us to obtain sector-size-specific estimates of the insol-
Office (Destatis, 2020)
vency gap. In addition to firm size, industry affiliation, a Population size distribution according to official statistics of
and credit rating update, we consider an extensive set the Federal Statistical Office (Destatis, 2020)
7 Figure 2 provides an illustration of how closely matching of additional firm-specific variables when matching
observations from the pre-crisis period serve as controls for counterfactual survival states of pre-crisis observa-
rating changes of firms in the crisis period. tions with rating updates of firms observed in the
896 J.O. Do¨rr et al.
rating update non-insolvent insolvent
Example I: non-insolvent firm in crisis period
1
Control 1
Control 2
Control 3
Control 4
Firm identifier
Control 5
Example II: insolvent firm in crisis period
2
Control 6
Control 7
Control 8
Control 9
Control 10
01 7 01 7 18 01 8 01 8 01 8 19 01 9 01 9 01 9 pr .2
20 20
20 20 Ju l. 02 0
l. 2 ct.2 n. pr.2 l. 2 ct.2 n. pr.2 l. 2 ct.2 O
n. ct.2
20 20
Ju Ja Ju Ja Ju Ja Ja n. 02 0
O A O A O A 20 21
Date
Fig. 2 Matching: illustration. Note: The figure illustrates the the nearest neighbors also filed for insolvency and thus closely
nearest neighbor matching for two micro-enterprises in the reflect the actual survival status of firm 2. If we do not observe
accommodation and catering sector. In the top panel, we see that an insolvency filing four months after the rating update, we treat
firm 1 experienced a rating update in the crisis period which did the update as non-insolvent. Therefore, the time between rat-
not result in an insolvency filing. Furthermore, we see, how- ing update and the non-insolvent labelling in the visualization
ever, that two out of the k = 5 nearest neighbors from the always spans 4 months. The area shaded in gray highlights a
pre-crisis control period filed for insolvency after they received transitional phase which we intentionally exclude from our anal-
a very similar rating update. This signals that firm 1, given its ysis since assignment of observations falling in that phase to
financial information, faces a relatively high insolvency risk as either the pre-crisis or the crisis period is not straightforward.
almost half of its nearest neighbors indeed went bankrupt in The dashed vertical line at the end of July 2020 signals that we
times without policy intervention. The bottom panel shows the only have credit rating updates available up to this point. Note,
same approach but for firm 2 which filed for insolvency shortly however, that we observe insolvency filings beyond this point
after its rating update during the crisis period. We see that all of in time
COVID-19 period. In the following section, we intro- the credit update, it is likely that the update does not
duce all of these matching variables and provide some reflect the reasons why the company went bankrupt. A
descriptive statistics. more recent update of the firm’s rating (if that existed)
In our data used for the estimation of the insolvency would be necessary to capture the company’s finan-
gap firm survival status, ft+4 , serves as outcome vari- cial deterioration that contributed to the subsequent
able. It is equal to 1 if the company has filed for insolvency. Second, the COVID-19 period for which
insolvency no more than four months after its last rat- we have information on credit rating updates spans 4
ing update. If the firm has not gone bankrupt or it has months from April 2020 to the end of July 2020. Thus,
filed insolvency more than four months after its lat- for the latest in-crisis rating updates in July 2020, we
est rating update, it is 0. This means that we take four can observe the firm’s survival status at most 4 months
months as maximum time lag between a credit rating until November 2020 (the time of writing this paper).
update and the date at which the respective firm has Therefore, the maximum forecasting horizon for the
filed its bankruptcy to count the rating update as being rating updates observed in the crisis period is limited
predictive for the subsequent insolvency filing. We to 4 months.
choose this threshold for two reasons. First, we want The most important variable in finding counter-
to ensure that the rating update has a high information factual survival states in the matching procedure is
content in predicting a potential insolvency filing. If Creditreform’s credit rating index since it is the basis
the date of bankruptcy lies more than 4 months after for the calculation of the credit rating updates. The
Small firms and the COVID-19 insolvency gap 897
credit rating is calculated by Creditreform on the basis how severely the company’s financial standing has
of a rich information set relevant to assess a com- changed.10
pany’s creditworthiness. The metrics considered in Apart from the rating update, we also consider the
calculating the rating include, among other things, rating before the upgrade, rt−x , as a matching vari-
information on the firm’s payment discipline, its legal able when predicting counterfactual insolvency states.
form, credit evaluations of banks, credit line lim- This allows us to control for the location of the com-
its and risk indicators based on the firm’s financial pany in the rating distribution and consequently how
accounts (if applicable) (Creditreform, 2020b). Cred- high the default risk was before the down-/upgrade.
itreform attaches different weights to these metrics Moreover, we form two additional variables from the
according to their relevance on determining a firm’s firm’s credit rating information, both of which control
risk of credit default and calculates an overall credit for the medium-term path of the firm’s financial stand-
rating score which ranges from 100 to 500.8 The ing. First, we count the number of downgrades in the
higher the score, the worse the firm’s creditworthiness three years preceding the update at hand, dt . Second,
and thus the higher the risk of insolvency. In fact, we calculate the average credit rating in the three years
Creditreform’s rating index has a high forecasting prior to the current update under consideration, r̄t .11
quality to assess a firm’s credit default risk (Creditre- Finally, we consider firm age, at , as further matching
form, 2020b). Assuming that a high credit default risk variable acknowledging that younger firms tend to be
signals financial distress, which often results in insol- more prone to insolvency.
vency, we use Creditreform’s rating as the basis for Table 3 shows descriptive statistics of the variables
predicting corporate insolvency risk. The prediction of considered in the matching procedure. We see that
corporate bankruptcy via a scoring model goes back to an update which is followed by a bankruptcy filing
the seminal work of Altman (1968) and his develop- relates to a downgrade of close to 70 scoring points
ment of the Z-score model. Similar to Creditreform’s on average. This is a substantial deterioration in the
credit rating index, the Z-score model relies on sev- rating index compared to an update which is not fol-
eral accounting-based indicators which are weighted lowed by an insolvency filing. In fact, the difference
and summed to obtain an overall score. This score then in means between non-insolvency-related updates and
forms the basis for classifying companies as insolvent insolvency-related updates, as reported in column ‘
or non-insolvent (Altman, 2013). Today, this model Mean’, amounts to more than 65 index points and is
approach is still used by many practitioners to predict statistically significant. For all other matching vari-
firm insolvencies (Agarwal and Taffler, 2008). ables, we also find statistically meaningful differences
Based on the credit rating index, we construct the suggesting that firms which go bankrupt have a worse
following variables. Our main predictor variable is the credit rating both short-term and mid-term, have expe-
update in the rating index, rt , which is defined as the rienced more downgrades in the past and are younger
difference of the new rating assigned by Creditreform on average. The economically and statistically signif-
an the rating before the update (rt = rt − rt−x ).9 icant differences between non-insolvency-related and
Given the logic of the rating index, a positive sign insolvency-related credit updates across all variables
in the rating update reflects a downgrade in financial
solvency, a negative sign reflects an improvement in 10 Note that we define a rating update as a reassessment of
the rating, i.e. an upgrade of the company’s financial the company’s creditworthiness performed by Creditreform. We
standing. The amount of the down-/upgrade reflects have precise information on the date of reassessment, which
allows us to accurately assign the update to either the pre-crisis
8 The credit rating index suffers from a discontinuity as in case or the crisis period and also to accurately match the updates
of a ‘insufficient’ creditworthiness it takes on a value of 600 with insolvency dates. It should also be noted that a reassess-
(Creditreform, 2020a). We truncate credit ratings of 600 to a ment does not necessarily lead to a change in the rating index. If
value 500—the worst possible rating in our analysis. We do the creditworthiness of the company has not changed since the
so since our main predictor variable is the update in the rating last rating, the company gets assigned the same index as before,
index which can only be reasonably calculated if the index has resulting in a value of 0 for rt .
continuous support. 11 For example, for a credit rating observation in July 2017, we
9 Reassessments of the rating is conducted in an irregular fash- count how often the firm experienced a downgrade over the
ion such that the time between two updates, x, varies. On period June 2014 to June 2017 and also calculate the average
average, the time between two updates equals 20 months. rating over that period.
898 J.O. Do¨rr et al.
Note: Non-insolvent observations comprise credit rating updates which have not resulted in an insolvency filing in the first four months after the update. Insolvent observations
−65.6793*** suggest that they serve well as matching variables in a
−148.0728*** counterfactual estimation of insolvency rates.
−0.0254***
Mean
We also report univariate descriptive statistics of
−101.8475***
our credit rating sample differentiating between the
8.9405***
pre-crisis and the crisis period in Table 4. We see
that before the COVID-19 outbreak 0.71% of rating
Max
359
500
3
500
400 updates were followed by a bankruptcy filing. This
69.6892 translates into an insolvency filing rate of 1.05% on the
414.6607 firm level (note that firms can receive more than one
Mean
0.5051
367.6691 credit rating update in that period). In the crisis period,
13.3199
however, despite the worsened economic conditions,
−226
141 it turns out that only 0.33% of rating updates were fol-
Min
0
138
1 lowed by a bankruptcy filing. This fraction also equals
the firm-level insolvency filing rate as in the 4-month
N firms
15,634
15,634 crisis period each firm is only observed once. ‘
15,634
15,634 Mean’ reporting the difference between the variable
15,166
means of the pre-crisis and the crisis period suggests
Insolvent
15,634
15,634 that the difference of 0.38 pp in the average survival
include observations which have been followed by an insolvency filing in the first four months after filing. N refers to the number of rating updates, Nf irm to the number of unique
15,634
15,634 status is statistically significant. The lower average
N 15,166
insolvency rate in the crisis period contrasts with the
351 finding that the financial rating of firms observed in
Max
500
3
500
1,017 the crisis period has deteriorated on average. In fact,
firms experience, on average, a significantly higher
4.0099
266.5879 downgrade of more than three index points during
the crisis period.12 This decline of insolvencies in
Mean
0.4797
265.8216
22.2604 the COVID-19 period is consistent with official fig-
Table 3 Descriptive statistics: non-insolvent observations and insolvent observations
−356 ures (The Economist, 2020b) and is a first indication
100
Min that there is indeed an insolvency gap in the German
0
100
1
economy. The strong political reaction to strengthen
1,489,376 firms’ liquidity and to prevent German companies
N firms
1,489,376
1,489,376 from going bankrupt is likely to be a driving force
1,489,376
1,479,383 behind the low insolvency rate in the crisis period.
Non-insolvent
It remains to be analyzed if there are specific sector-
2,358,148
2,358,148 size combinations for which the number of insolven-
2,358,148 cies is significantly below the counterfactual num-
2,358,148
N 2,346,686 ber that one would expect given the observed rating
firms which experienced at least one rating update. Significance levels: *p < 0.10, **p < 0.05, ***p < 0.01
updates and information from pre-crisis insolvency
Predictor variables
paths. Also we aim to tackle the question whether the
gap is mainly driven by firms which already before the
crisis were characterized by a weak financial stand-
Credit rating update: rt
ing. In the next section, we introduce a matching
approach that allows us to predict counterfactual insol-
Credit rating (prior to update): rt−x vency filings based on pre-crisis observations where
no policy intervention saved struggling firms from
Number of downgrades (3-year horizon): dt insolvency. With this approach, we are able to derive
counterfactual insolvency rates at the sector-size level
Average credit rating (3-year horizon): r̄t 12 See also Fig. 8 in the Appendix for a comparison of the distri-
Variable
bution of the credit rating updates in the pre-crisis and the crisis
Company age: at period.
Small firms and the COVID-19 insolvency gap 899
Note: Pre-crisis period comprises all credit rating observations from July 2017 to December 2019. Crisis period includes all observations starting from April 2020 to July 2020.
−3.2336*** and provide an estimate regarding the existence of an
0.0038***
−0.2017**
insolvency gap by comparing them with the actual
Mean
0.0098***
−0.9384*** filings observed during the crisis period.
−0.7408***
349 4.2 Methodology
500
3
Max 1
500
1,016.00 4.2.1 Nearest neighbor matching
7.2161
267.7361 This paper focuses on the extent to which govern-
Mean 0.0033
0.4714
267.2973 ment policy in the COVID-19 crisis may have induced
22.8378 ailing firms to stay in the market. To answer this
−293 question, we compare the survival status of closely
100
Min matching firms observed before the COVID-19 out-
0
100
0 1
Although the mean differences in the predictor variables are statistically significant, their magnitude seems to be rather negligible (except credit rating update), particularly when
break with the survival status of firms observed during
337,679 the pandemic. Besides general company characteris-
N firms 337,679
337,679
337,679
337,679 tics such as company size, industry affiliation and
337,679 company age, our matching approach takes particular
337,679 account of firm-specific solvency information as pre-
337,679
337,679
sented in the previous section. The core idea of the
Crisis
337,679
337,679
N 337,679 matching procedure is to find comparable firms which
359 have experienced very similar rating updates and have
followed an almost identical path in their financial sol-
Max
500
3
1 500
1,017 vency but in times prior to COVID-19 and the related
3.9825 policy interventions that keep struggling firms afloat.
267.5344
We conduct a nearest neighbor matching approach
Mean 0.0071
0.4812
266.3589
22.0970 in order to find for each of the in-crisis observations
a number of matches from the pre-crisis period. Near-
comparing with the differences between non-insolvent and insolvent observations (Table 3). This suggests that the crisis sample is not biased in the sense that it primarily includes
0 −356
100 est neighbor matching in observational studies goes
Min
0
100
back to the work of Donald Rubin (1973) and aims
Table 4 Descriptive statistics: pre-crisis observations and crisis observations
1
at reducing bias in the estimation of the sector-size-
1,377,671
1,377,671
specific insolvency gap. A simple comparison of the
N firms
1,377,671
1,377,671
1,377,671 mean values of the survival status of observations
1,367,244 before the crisis and during the crisis (as in Table 4)
is likely to give a highly biased picture of the insol-
2,036,103
2,036,103
Pre-crisis
2,036,103 vency gap. First, policy measures to rescue firms from
2,036,103
2,036,103 failing have been highly heterogeneous with respect
N 2,024,173
to firm size as highlighted in Section 3. Therefore,
comparing the survival status of firms of different size
Outcome variable
bears high risk of firm size acting as confounding
Survival status: ft+4 variable in the estimation of a policy-induced back-
credit updates of firms with poor financial records. Significance levels: *p < 0.10, **p < 0.05, ***p < 0.01
log of insolvencies. For this reason, we only search for
Predictor variables
matches within the same company size group. Next,
Credit rating update: rt the evaluation of our survey suggests that there is great
heterogeneity in the COVID-19 exposure across sec-
Credit rating (prior to update): rt−x
tors (as becomes apparent in Section 5.1). For this
Number of downgrades (3-year horizon): dt reason, we only match firms that are in the same sector
class. Ultimately, the previous section has shown that
Average credit rating (3-year horizon): r̄t
Variable
in the crisis period the distribution of rating updates
Company age: at has systematically shifted to the right implying that the
900 J.O. Do¨rr et al.
in-crisis observations have, on average, experienced analysis, this requires us to consider weights which
larger downgrades in their ratings. For an unbiased reflect whether a pre-crisis unit falls in the matched
estimation of the insolvency gap, this shift needs to be sample more than once. In Section 5.2 where we esti-
controlled for. Our nearest neighbor matching aligns mate the insolvency gap on the sector-size level, we
the in-crisis distribution of updates with the distribu- need to consider these weights for inference (Stuart,
tion of matched observations as we put a strict caliper 2010). In this way, we can not only predict the crisis
on the credit rating variable when searching for match- observations’ probability to file for bankruptcy if there
ing observations. In fact, comparing the distribution was no policy intervention but also make a statement
of the predictor variables between pre-crisis and crisis whether the differences between the observed insol-
period before and after matching indicates that control vency rates and the predicted counterfactual insol-
observations and crisis observations are much more vency rates on the sector-size level are statistically
balanced after matching (see Table 12 in the Appendix significant.
for an assessment of covariate balance). Before presenting the results of the counterfactual
The details of our matching algorithm look as fol- insolvency rate prediction and insolvency gap estima-
lows. Acknowledging the heterogeneity with respect tion, we use our survey results in the next section
to firm size and sector affiliation, we estimate the to show how the pandemic affected sectors to vary-
insolvency gap within each of the 52 sector-size ing degrees. The observed heterogeneity in sector
combinations. Therefore, we only consider pre-crisis exposure motivates our further empirical analysis.
observations that share the same sector-size stratum
as the crisis observation of interest. In that sense
we perform exact matching on both sector affiliation 5 Empirical results
and company size group. Next, within each sector-
size stratum the algorithm selects for each in-crisis 5.1 COVID-19 exposure and firm characteristics
observation i the k nearest neighbors from the pre-
crisis period which have the smallest distance from Anecdotal evidence suggests that industries are asym-
i. The maximum number of nearest neighbors, k, metrically affected by the COVID-19 recession
reflects the ratio of pre-crisis and crisis observations because lockdown measures as well as supply and
within each sector-size stratum. Distance is measured demand effects differed between sectors. To verify
by the Mahalanobis distance metric (Rubin, 1980), this observation, we empirically investigate to what
MD, which is computed on all predictor variables extent the economic effects of the COVID-19 crisis
X = (rt rt−x dt r̄t at ) . For the key predictor vari- have asymmetrically hit sectors by making use of our
able, rt , we additionally impose a caliper, c, of 0.25 survey data. In addition, we analyze whether firm size
standard deviations. Thus, a pre-crisis observation, j , and the pre-crisis credit rating is correlated with the
only falls under the k nearest neighbors if it does not perceived shock by the COVID-19 recession at the
exceed the caliper on rt . firm level.
The regression results of the analyses are shown
(Xi−Xj )−1 (X i−Xj ) if |rt,i −rt,j | ≤ c
MDij = in Table 5. Model (1) reveals that the COVID-19
∞ if |rt,i −rt,j | > c
Exposure Index indeed significantly differs between
with as the variance covariance matrix of X in the sectors. We choose chemicals and pharmaceuticals as
pooled sample of in-crisis and all pre-crisis observa- reference category since this sector is least negatively
tions. The strict caliper implies that the number of affected. The sectors accommodation and catering as
matches on each crisis observation can be smaller well as creative industry and entertainment experience
than k or, in case that there is no control observa- very strong and significant negative shocks in com-
tion fulfilling the caliper condition, there may even parison to the baseline sector. This is in line with the
be no match. If this the case, the crisis observation strong restrictions experienced in these sectors. Since
for which no match could be found is disregarded the business activities in these sectors often require
from further analysis. Moreover, we conduct matching direct human interactions, corresponding companies
with replacement allowing pre-crisis units to match have been severely affected by lockdown measures.
to more than one crisis observation. In the outcome Interestingly, firm size categories show no statistically
Small firms and the COVID-19 insolvency gap 901
Table 5 Regression: COVID-19 Exposure Index on firm characteristics
(1) (2) (3) (4)
CEI CEI CEI CEI
Business-related services 0.637 0.646 0.473
(0.611) (0.609) (0.615)
Manufacturing −0.004 −0.023 −0.073
(0.605) (0.603) (0.604)
Wholesale & retail 1.479** 1.476** 1.427**
(0.647) (0.644) (0.646)
Health & social services 1.087* 1.085* 0.855
(0.660) (0.657) (0.661)
Insurance & banking 0.643 0.618 0.653
(0.689) (0.686) (0.682)
Acc. & catering 6.024*** 6.046*** 5.835***
(0.711) (0.710) (0.712)
Logistics & transport 1.454** 1.464** 1.396**
(0.650) (0.647) (0.646)
Creative i. & entertainment 5.444*** 5.445*** 5.224***
(0.832) (0.831) (0.831)
Mechanical engineering 2.464*** 2.477*** 2.433***
(0.665) (0.659) (0.658)
Food production 2.564*** 2.559*** 2.394***
(0.701) (0.699) (0.696)
Manufac. of data proc. equip. 0.147 0.156 0.208
(0.653) (0.652) (0.650)
Micro-enterprise 0.311 −0.048 −0.509
(0.423) (0.418) (0.455)
Small enterprise 0.269 −0.248 −0.538
(0.447) (0.433) (0.448)
Medium-sized enterprise −0.0216 −0.128 −0.209
(0.457) (0.440) (0.444)
Credit rating (pre-crisis) 0.008***
(0.003)
N 2,344 2,344 2,344 2,344
Note: Chemicals and pharmaceuticals serve as baseline sector among the sector dummies, large enterprises serve as baseline size
group. Dummy coefficient estimates need to be read relative to the baseline group(s). Standard errors are reported in parentheses.
Significance levels: *p < 0.10, **p < 0.05, ***p < 0.01
significant heterogeneity in their correlation with the heterogeneity in the negative exposure to the eco-
COVID-19 Exposure Index as Model (2) reveals. The nomic consequences of the pandemic with respect to
effects with respect to sectors and firm size also hold sector affiliation hold when controlling for the firms’
when both measures are incorporated simultaneously pre-crisis credit rating in Model (4).
as in Model (3). Controlling further on the firms’ The heterogeneous COVID-19 exposure at the sec-
pre-crisis credit rating and thus on the financial sit- tor level shows that differences in insolvency dynam-
uation prior to the outbreak shows that the rating is ics with respect to industry affiliation may play an
significantly correlated with the perceived COVID-19 important role. Taking further into consideration that
impact. Although the effect is low in magnitude, the many of the policy measures in Germany have been
marginal effect suggests that a higher (worse) credit specifically tailored to SMEs, the subsequent esti-
rating is associated with a stronger exposure to the mation of the insolvency gap is conducted at the
negative impact of the crisis. Ultimately, the strong sector-size level.
902 J.O. Do¨rr et al.
5.2 The COVID-19 insolvency gap micro-enterprises (except for some outliers in the
large enterprise size class). In the group of micro-
5.2.1 Results on the sector-size level enterprises, we see that actual insolvency rates are
highest in the sectors which according to our sur-
Estimating the insolvency gap requires us to derive vey results are also severely affected by the negative
two statistics. First, we calculate actual insolvency impacts of the crisis. In the accommodation and cater-
rates, I Rsactual , observed after the COVID-19 out- ing sector, for example, the actual insolvency rate
break for each sector-size stratum s.13 The calculation amounts to 1.11%, in the logistics and transport sector
is based on firms for which we observe credit rating which includes the strongly affected aviation indus-
updates after April 1, 2020. try we observe an insolvency rate of 0.94% and in the
creative industry and entertainment sector the rate is
Nsinsolvent
I Rsactual = 0.76%. These results appear intuitive and are in line
Ns
with the survey results. At the same time, we find that
Second, taking the matched sample of observations in all sectors within the group of micro-enterprises the
from the pre-crisis period which includes for each firm expected insolvency rates exceed the actual rates and
observed in the crisis period at most k nearest neigh- in most sectors this gap is statistically significant. The
bors, we are able to estimate counterfactual insolvency average insolvency gap across all sectors in the group
counterf actual
rates, I Rs , as follows of micro-enterprises amounts to 0.80 pp which is sub-
stantial when being compared to the overall pre-crisis
insolvency rate of 1.05%.
In the group of small enterprises, we see similar
patterns although at a lower magnitude both in terms
Ñs
with Ñs = j =1 wj,s as the number of matched
of actual insolvency rates and counterfactual rates. In
observations from the pre-crisis period for stratum s. fact, Table 6 suggests that the rates expected in most
wj,s is the weight assigned to pre-crisis observation sectors exceed actual rates for small enterprises; how-
j reflecting how often j is selected as control obser- ever, this gap is in no sector statistically significant.
vation in the matching process and On average, the insolvency gap in the group of small
equals 1 if control observation j filed for insolvency businesses amounts to 0.03 pp.
at most four months after its last rating update and 0 Moving on to the group of medium-sized enter-
otherwise. prises, the patterns observed in the smaller size classes
Comparing actual insolvency rates with counterfac- start to vanish. While in two of the most severely hit
tual insolvency rates for each of the sector-size strata sectors accommodation and catering as well as logis-
allows us to obtain sector-size-specific estimates of tics and transport expected insolvency rates are higher
the insolvency gap, I Gs , defined as than the ones observed, the difference (i.e. the insol-
counterf actual vency gap) is statistically not significant. For the other
I G s = I Rs − I Rsactual . sectors, the picture is even more mixed. In two sec-
In other words, the insolvency gap measures the tors (food production and mechanical engineering),
extent to which observed insolvencies during the pan- some insolvencies took place yet almost none were
demic deviate from the counterfactual insolvencies predicted in the counterfactual scenario. For all other
that would be expected in a pre-crisis setting without sectors, actual and counterfactual rates are very simi-
policy intervention. Figure 3 contrasts actual insol- lar. Table 6 shows that none of the differences (except
vency rates against counterfactual insolvency rates for the sector mechanical engineering) are statistically
and Table 6 displays the sector-size specific insol- significant.
vency gap estimates along with their statistical signif- Ultimately, the patterns break down completely
icance. Several insights can be gained from there. for the group of large enterprises. Barely any insol-
First of all, it becomes obvious that actual insol- vency filing can be observed in either the crisis period
vency rates are in almost all sectors highest among or the counterfactual setting. In general, insolven-
cies among large corporations are rather rare events
13 s ∈ [1, 52]. which is reflected by our results. Two sectors stand
Small firms and the COVID-19 insolvency gap 903
Counterfactual Actual
Small Medium-sized Large
Micro-enterprise
enterprise enterprise enterprise
Accommodation & catering
Logistics & transport
Wholesale & retail trade
Manufacturing
Business-related services
Creative industry & entertainment
Sector Food production
Health & social services
Insurance & banking
Others
Manufacturing of data processing equipment
Chemicals & pharmaceuticals
Mechanical engineering
01 02 01 02 0.01 02 01 02
0. 0. 0. 0. 0. 0. 0.
Insolvency rate
Fig. 3 Actual and counterfactual insolvency rates. Note: The figure displays actual insolvency rates with estimated counterfactual
insolvency rates for each of the S = 52 sector-size strata
Table 6 Outcome analysis: insolvency gap estimation results
Sector affiliation Size of company
Micro Small Medium Large
I Gs I Gs I Gs I Gs
Accommodation & catering +0.0115*** +0.0005 +0.0028 0.0000
Logistics & transport +0.0070*** +0.0002 +0.0030 0.0000
Wholesale & retail trade +0.0107*** +0.0004 +0.0001 −0.0006
Manufacturing +0.0103*** +0.0002 −0.0004 −0.0035
Business-related services +0.0070*** −0.0001 −0.0005 0.0000
Creative industry & entertainment +0.0012 +0.0017 0.0000 0.0000
Food production +0.0027 +0.0024 −0.0019 −0.0105**
Health & social services +0.0037*** +0.0005 −0.0011 +0.0004
Insurance & banking +0.0037*** 0.0000 0.0000 0.0000
Others +0.0037*** −0.0002 0.0000 0.0000
Manufacturing of data processing equipment +0.0044* −0.0009 0.0000 −0.0122*
Chemicals & pharmaceuticals +0.0033* +0.0003 0.0000 0.0000
Mechanical engineering +0.0003 +0.0018 −0.0025*** 0.0000
Note: Significance levels: *p < 0.10, **p < 0.05, ***p < 0.01. Statistical significance is based on the χ 2 -test for equality in the
insolvency proportions in the actual and counterfactual samples using Rao-Scott corrections to the χ 2 statistic (Rao & Scott, 1981) to
account for the matching weights
904 J.O. Do¨rr et al.
out with high actual insolvency rates: food produc- firms. This becomes particularly problematic if the
tion and manufacturing of data processing equipment. non-filing firm does not fulfill the criteria to be eligi-
Both cases are somewhat special as they are driven by ble for the suspension as it enables these companies to
only one insolvency for which no insolvent pre-crisis further absorb state subsidies. Similarly, the early on
control observation with comparable financial charac- provision of direct and indirect liquidity without red
teristics exists. Thus, one needs to be cautious when tape has targeted smaller firms in particular and thus
interpreting the results of the large size class. enabled them to bridge plummeting revenues in a sit-
The finding that counterfactual insolvency rates uation in which they usually would have been forced
persistently, and in most sectors also significantly, out of the market due to illiquidity.
exceed actual rates among micro-enterprises strongly In order to better understand the magnitude of the
suggests that there is a substantial backlog of insol- insolvency gap, it is possible to aggregate and convert
vencies in this size class. As company size increases, the insolvency gap estimates on the distinct sector-
the backlog of insolvencies gradually vanishes which size levels into an absolute number describing the
is in line with our hypothesis that Germany’s fiscal overall backlog of insolvencies (see Table 11 in the
policy response in the COVID-19 crisis disproportion- Appendix). Based on the total number of economi-
ately favored the survival of smaller companies. Both cally active companies in Germany, we estimate that
the temporary change in Germany’s insolvency regime the insolvency gap makes up around 25,000 com-
and the high provision of liquidity subsidies allowed panies as shown in Fig. 4. The figure reveals two
especially micro-enterprises to stay in the market. We further aspects. Firstly, the time series shows that dur-
argue that the temporary suspension of the obligation ing the last economic shock, the Great Recession of
to file for insolvencies has made it particularly easy 2008–2009, the number of insolvencies noticeably
for smaller firms to use the amendment as a loop- increased, which in light of the Schumpeterian cleans-
hole to avert insolvency proceedings. Since disclosure ing mechanism is an expected response in business
requirements are more limited the smaller a company dynamics. Secondly, in contrast to the Great Reces-
is, it becomes particularly difficult for policy makers sion, it can be seen that in the current crisis the actual
to enforce insolvency filings among non-filing small number of corporate insolvencies has declined. The
40,000
25,000 firms)
Number of insolvency filings
Insolvency gap
30,000
(
Great Recession
20,000
COVID-19 Crisis
Actual insolvencies
16,300 firms)
10,000
(
0
20 20 20 20 20 20 20 20 3 20 20 20 20 20 20 0 1
05 06 07 08 09 10 11 12 20 1 14 15 16 17 18 19 20 2 20 2
Year
Fig. 4 Evolution of corporate insolvencies in Germany and the 2020, both actual number of insolvencies as well as the evolu-
COVID-19 insolvency gap. Note: The figure displays yearly tion of insolvencies in the counterfactual scenario are displayed.
absolute number of corporate insolvencies in Germany accord- Shaded areas reflect months of economic downturn
ing to official statistics of the Federal Statistical Office. In
Small firms and the COVID-19 insolvency gap 905
observation that bankruptcy filings are lower in an firms’ pre-crisis financial standing in the estimation of
economic crisis than in non-crisis times underpins that the insolvency gap.
the large-scale governmental support programs have
led to substantial distortions in business dynamics. 5.2.2 The insolvency gap and firm viability
Indeed, policy measures in Germany have prevented
a significant number of companies from insolvency. In order to examine whether the insolvency gap is
The crucial question is, which firms were saved from driven by companies that are characterized by a poor
insolvency proceedings. The following section fur- financial standing before the crisis and had faced a rel-
ther narrows down this question by incorporating the atively high risk of market exit when the pandemic hit,
Table 7 Outcome analysis: insolvency gap estimation results incorporating firms’ pre-crisis financial condition
Viability Sector affiliation Size of company
Micro Small Medium Large
I Gs I Gs I Gs I Gs
Strong financial standing (pre-crisis)
Accommodation & catering −0.0029*** −0.0013 0.0000 0.0000
Logistics & transport −0.0008 −0.0001 +0.0003 0.0000
Wholesale & retail trade +0.0003 −0.0001 +0.0005 −0.0007
Manufacturing −0.0001 −0.0001 −0.0008 −0.0038
Business-related services −0.0002 −0.0006 −0.0010 0.0000
Creative industry & entertainment −0.0025** 0.0000 0.0000 0.0000
Food production −0.0007 +0.0020 −0.0027* −0.0112*
Health & social services −0.0001 −0.0004 −0.0003 0.0000
Insurance & banking +0.0007 0.0000 0.0000 0.0000
Others −0.0002 +0.0002 0.0000 0.0000
Manufacturing of data processing equipment +0.0007 −0.0015 0.0000 −0.0127*
Chemicals & pharmaceuticals +0.0017 +0.0020 0.0000 0.0000
Mechanical engineering −0.0003 −0.0017 −0.0015* 0.0000
Weak financial standing (pre-crisis)
Accommodation & catering +0.0171*** +0.0018 +0.0030 0.0000
Logistics & transport +0.0128*** +0.0004 +0.0049 0.0000
Wholesale & retail trade +0.0196*** +0.0020 −0.0035 0.0000
Manufacturing +0.0184*** +0.0015 +0.0060 −0.0060
Business-related services +0.0122*** +0.0013 +0.0033 0.0000
Creative industry & entertainment +0.0029 +0.0032 0.0000 –
Food production +0.0051 +0.0030 +0.0025 0.0000
Health & social services +0.0059*** +0.0022 +0.0010 +0.0060
Insurance & banking +0.0055*** 0.0000 0.0000 0.0000
Others +0.0073*** −0.0012 0.0000 0.0000
Manufacturing of data processing equipment +0.0065 0.0000 0.0000 0.0000
Chemicals & pharmaceuticals +0.0050 −0.0050 0.0000 0.0000
Mechanical engineering +0.0014 +0.0126* −0.0056 0.0000
Note: The upper panel displays insolvency gap estimates for firms with ‘strong financial standing’ comprising all firms whose three
year average credit index prior to the crisis is better than the median rating index. The lower panel shows results for companies with
a ‘weak financial standing’ including those with a rating worse than the median rating. For large firms in creative and entertainment
sector with weak financial standing the insolvency gap could not have been calculated as no firm in this strata has been observed
during the crisis period. Significance levels: *p < 0.10, **p < 0.05, ***p < 0.01. Statistical significance is based on the χ 2 -test for
equality in the insolvency proportions in the actual and counterfactual samples using Rao-Scott corrections to the χ 2 statistic (Rao &
Scott, 1981) to account for the matching weights
906 J.O. Do¨rr et al.
we split the sector-size strata further according to the accommodation and catering sector and the creative
observations’ pre-crisis financial standing. More pre- and entertainment sector observed insolvencies signif-
cisely, we split each sector-size strata into two further icantly exceed expected insolvencies. We know from
sub-strata. The first sub-strata contains all observa- our survey that these two sectors are by far the most
tions whose three year average credit rating prior affected industries. Given the severe impairments in
to the crisis is better than the overall median rating these industries, it seems plausible that companies
index. We refer to these as observations with ‘strong which had been rated relatively well before the pan-
financial standing’, viable to survive the crisis based demic nevertheless file for insolvency more frequently
on their pre-crisis conditions. The other sub-strata than the counterfactual estimation would suggest. This
comprises all observations worse than the overall means that, despite the cushioning effect provided by
pre-crisis median rating. Firms falling in such sub- fiscal policy, micro-firms with a strong pre-crisis rat-
strata are referred to as having a ‘weak financial stand- ing filed for bankruptcy more frequently than would
ing’. Given their pre-crisis financial circumstances, we have been possible to learn from the financial paths
expect them to be more vulnerable to default in the of similarly strong firms in pre-crisis times. Most
current crisis or even if the pandemic had not hit the important, however, is the finding that among micro-
economy. enterprises the insolvency gap as backlog of expected
Table 7 shows the insolvency gap estimates analo- insolvencies is not driven by firms with a strong finan-
gous to Table 6 with the additional distinction between cial standing prior to the crisis. In contrast, it is driven
strata comprising financially strong companies and by less viable companies with a rating worse than the
financially weak ones. Several aspects become appar- median rating. This results from the insolvency gap
ent from these results. First, we observe that among estimates among micro-enterprises with weak finan-
micro-enterprises with above median credit rating (top cial standing (bottom panel). It becomes apparent that
panel) there is in almost no sector a significant devi- throughout all sectors the counterfactual insolvency
ation between actual and counterfactual insolvency rates exceed the actual rates indicating a backlog of
rate. There are, however, two exceptions. Both in the insolvencies which in the majority of sectors is not
Strong financial standing Weak financial standing
1.5
Average insolvency gap (in pp)
1
0.5
0
-0.5
ro al l m rge
ic Sm iu La
M Med
Size of company
Fig. 5 Average insolvency rates by size class and pre-crisis insolvencies is strongly driven by micro-enterprises with weak
viability. Note: The figure displays average insolvency gaps credit rating prior to the crisis and also that with increasing firm
(weighted by the number of observations falling into the size the gap becomes less pronounced among financially weak
matched strata) distinguishing by company size and pre-crisis companies
financial conditions. It becomes apparent that the backlog of
Small firms and the COVID-19 insolvency gap 907
only statistically significant but in some also substan- in the short term, liquidity subsidies and loan guar-
tial in magnitude. In fact, the gap amounts to more antees have been necessary to save companies under
than 1.70 pp in accommodation and catering, whole- severe liquidity pressure from insolvency. In Germany,
sale and retail as well as manufacturing which is a a country where fiscal policy played a crucial role in
substantial backlog when taking into consideration mitigating the crisis’ impact, liquidity subsidies were
that the overall pre-crisis insolvency rate lies at 1.05%. accompanied by a temporary suspension of the insol-
In the small business group, we see that in the strata vency regime. While both measures are different in
with strong financial performance, the size and sign design, they target the same objective: preventing an
of the insolvency gap estimates are comparable to the unprecedented wave of corporate insolvencies. Study-
results in the micro firm group (except for accom- ing Germany’s policy response, it becomes also appar-
modation/catering and creative/entertainment sector). ent that a number of aid schemes were either explic-
These results indicate that there is no significant gap in itly designed to save smaller companies or at least
insolvency filings among small businesses with above implicitly favored the survival of particularly small
median credit rating. In the strata of small and finan- entrepreneurial firms. This policy environment is the
cially weak businesses, in turn, we observe for most basis for our hypothesis that a substantial backlog
sectors a positive sign in the insolvency gap estimation of insolvencies has accumulated particularly among
albeit only statistically significant in the mechanical SMEs as a result of the COVID-19 policy response.
engineering sector. Again, this suggests that also in If, however, support schemes postpone or even prevent
the group of small firms the backlog of insolvencies is the exit of financially weak SMEs, there is the danger
driven by companies with weak pre-crisis conditions of negative long-term effects on the entire economy. In
even if magnitude and significance is less pronounced fact, in the ongoing crisis it is likely that early liquidity
in comparison to the micro size group. issues increasingly translate into an erosion of firms’
Similar to the results in Table 6 and in line with our equity. Suspending bankruptcy proceedings of such
hypothesis that the fiscal policy response in Germany over-indebted firms over a longer period of time not
disproportionately favored survival of smaller compa- only is ‘to deny reality’ (The Economist 2020a, 3) but
nies, the observed patterns for small and especially also hampers the efficient reallocation of resources.
micro-sized firms gradually vanish with increasing firm In this vein, economic crises also serve as cleansing
size as shown in Fig. 5. Consequently, the insolvency mechanism to release resources from inefficient and
gap estimates for medium-sized and large enterprises non-innovative firms which typically find more pro-
do not reveal clear patterns in the sign of the estimates ductive use elsewhere. The early policy response of
nor significant deviations between observed and the German government not only has been targeted
predicted rates (apart for some aforementioned disproportionately at smaller firms but also did so with
exceptions). little screening mechanisms in place (see, for example,
Our results show that the COVID-19-induced pol- Federal Ministry for Economic Affairs and Energy
icy response has created a non-negligible insolvency 2020) rescuing companies from insolvency in a fairly
gap that is strongly driven by micro-enterprises, which indiscriminate manner.
were already in a comparatively weak financial situa- Making use of both survey data and a unique and
tion before the crisis. large dataset of firm-specific credit rating data along
This suggests that the early policy answer to with information on firm insolvency filings, we inves-
dampen the economic impacts of the COVID-19 cri- tigate whether the German policy response has indeed
sis has indeed hampered the natural cleansing effect caused distortions in the natural cleansing mechanism
typically observed during economic crises. typically encountered in liquidity crises. While the
policy response to the economic impact of COVID-19
in Germany suggests notable differences in firm size,
6 Discussion and conclusion our survey results reveal strong heterogeneity across
economic sectors in their exposure to the adverse
The ongoing COVID-19 crisis has placed a special effects in the current crisis. With these findings, we
role on policy in order to soften the adverse economic estimate the extent of an insolvency gap, defined as
impacts faced by many firms. There is little doubt that, the deviation of observed insolvency rates during the
908 J.O. Do¨rr et al.
COVID-19 pandemic and expected insolvency rates creating barriers to entrepreneurship. It is likely that
based on a counterfactual pre-crisis setting with no once the policy instruments will cease, i.e. liquid-
policy intervention, for 52 distinct sector-size strata. ity support will terminate and the German insolvency
In line with our hypothesis, our results show that regime returns back to the filing obligation, a number
the insolvency gap is particularly significant in the of small business insolvencies will follow. Without an
group of micro-enterprises (at most 10 employees) ‘evergreening’ of policy support it is, however, doubt-
and that the gap gradually vanishes with increasing ful if they can be prevented at all. In the ongoing crisis,
firm size. Furthermore, we distinguish between finan- it will therefore become increasingly important to
cially strong and financially weak firms in our analysis think about policy measures that remove entry barriers
with the latter being defined as companies with below for young and innovative businesses and create growth
median credit rating prior to the crisis. Thus, we opportunities for firms which respond innovatively to
refer to financially weak firms as companies being the pandemic instead of prolonging the survival of
relatively more vulnerable to default in the current cri- near-insolvent firms. For example, policy makers are
sis based on their pre-crisis financial standing. Our well advised to consider law reforms that lower the
findings suggest that the backlog of insolvencies is barriers to corporate restructuring for viable smaller
mainly driven by firms with a relatively poor credit firms while streamlining and encouraging liquidation
rating prior to the crisis. This indicates that particu- procedures for unviable companies. Past experience
larly financially weak, small firms may take advantage suggests that this would stimulate the reallocation of
of the less stringent screening processes associated capital to more productive entrepreneurial endeavors
with many of the COVID-19-related policy instru- (Adalet McGowan et al., 2018).
ments or absorb the liquidity injections as windfall Understanding the effects of the interplay between
gains, especially during the first months of the crisis liquidity support on the one hand and temporary adjust-
when eligibility criteria were low. ments to insolvency regimes on the other hand will be
From a welfare perspective, this comes at the an important lesson from the COVID-19 crisis. Does the
burden of high fiscal costs that are associated with interplay of these two instruments impair entrepreneur-
granting financial aid to unviable firms. Favoring ship and economic recovery as it primarily discour-
the survival of financially weak firms as our find- ages struggling firms from exiting the market or does
ings indicate, however, also imposes indirect costs it, if well dosed, even serve as a useful policy mix in liq-
in the longer term as such firms tie up resources uidity crises? Our results which only look at the early
whose efficient redistribution would have facilitated policy effects in the pandemic suggest the former. It
entrepreneurship. Past experience shows that keeping is left to future research to investigate the long-term5
distressed firms alive may severely obstruct business effects on productivity, innovation and entrepreneur-
dynamism and structural change. Literature on Japan ship induced by the policy responses to COVID-19.
(Caballero et al., 2008), but also on other OECD
economies (Adalet McGowan et al., 2018), suggests Acknowledgements We are indebted to our anonymous
that granting life-sustaining credit to near-insolvent reviewers and Tobias Weih for comments and criticism that
greatly improved the manuscript. Any remaining errors are our
firms has not only lowered aggregate productivity own.
but also deterred market entry of new entrepreneurs.
Although in these cases the survival of insolvent firms Funding Open Access funding enabled and organized by Pro-
is mostly attributed to questionable bank lending prac- jekt DEAL. The study extends upon a project analyzing the
tices and not to a crisis-related policy response, some economic effects on SMEs in the COVID-19 crisis which was
funded by the German Federal Ministry of Economic Affairs
lessons can still be learned from these experiences: (BMWi) under the grant agreement no. 15/20. In addition, the
keeping unviable firms alive causes severe market study received research funding from the European Union Hori-
congestion which creates barriers to market entry and zon 2020 Research and Innovation Action for the ‘GrowInPro’
limits the growth of young companies. The persis- project under grant agreement no. 822781.
tence of crisis-induced SME support along with a
Availability of data and materials The data that support the
further prolongation of the (at least partial) mora- findings of this study are deposited in the ZEW Research Data
torium of Germany’s insolvency regime increasingly Center (ZEW-FDZ) and will be made available upon reasonable
favors such a market congestion with the risk of request from the corresponding author (J.O. Dörr).
Small firms and the COVID-19 insolvency gap 909
Appendix
Table 8 Mapping firm
characteristics to size group Size of company
Micro Small Medium Large
Note: The table shows
Number of employees ≤10 11–49 50–249 ≥250
translation of firm
characteristics into company Annual turnover (in M e) ≤2 2–10 10–50 >50
size classes used in this study as Annual balance sheet ≤2 2–10 10–43 >43
defined by European total (in M e)
Commission (2003)
Table 9 Mapping EU
NACE Revision 2 divisions Sectors Divisions
to sector groups
Business-related services 58–63, 68, 69–82
Manufacturing 5–9, 12–19, 23–25, 27,
31–33, 35–39, 41–43
Wholesale & retail trade 45–47
Health & social services 86–88, 94–96
Insurance & banking 64–66
Accommodation & catering 55, 56
Logistics & transport 49–53
Creative industry & entertainment 90–93
Mechanical engineering 28–30
Food production 10, 11
Chemicals & pharmaceuticals 20–22
Manufacturing of data processing 26
Note: The table shows
translation of EU’s NACE equipment
Revision 2 divisions (European Others Any division not
Union, 2006) into sector listed above
groupings used in this study
910 J.O. Do¨rr et al.
Guarantees Immediate fiscal impulse Deferral
Italy
Germany
Belgium
France
UK
Country
Portugal
Spain
Denmark
Netherlands
United States
Hungary
Greece
0 10 20 30 40 50
% of 2019 GDP
Fig. 6 COVID-19 liquidity support through fiscal policy mea- amount (as share of 2019 GDP) of fiscal policy measures to
sures by international comparison. Note: Calculations are address adverse COVID-19 impacts on companies for selected
retrieved from Anderson and et al. (2020). Numbers reflect the OECD countries. Numbers are as of November 18, 2020
Jan-Aug 2020, % decrease on same period a year earlier
Aug 2020, % decrease on a year earlier
Japan
United States
Country UK
France
Germany
-4 0 -3 0 -2 0 -1 0
0
Change (in %)
Fig. 7 Decline in corporate insolvencies during the COVID-19 crisis. Note: The figure shows the percentage change of insolvencies
in the crisis year 2020 compared to 2019 for a number of selected countries. Bar chart is adapted from The Economist (2020b)
Small firms and the COVID-19 insolvency gap 911
Table 10 COVID-19 first-round policy measures in Germany: Overview
Instrument Description Scope Target group Suggested effect on
(by firm size) insolvency filings
Liquidity support
Direct liquidity ‘Soforthilfen’: fully Up to e15k per month Weakly lowering short-term
subsidy subsidized payments e50bn overall insolvency risk, weakly
over 3 months favoring insolvency gap
‘Überbrückungshilfen’: Up to 80% of fixed costs Lowering short-term insolvency
fully subsidized payments (max. e50k) per month risk, moderately favoring
over 3 months e25bn overall insolvency gap
Liquidity ‘KfW-Schnellkredite’: Up to 3 monthly turnovers Lowering mid-term insolvency
loan under low-interest loans hedged (max. e800k) in total risk, favoring insolvency gap
public guarantee by a 100% guarantee from
scheme the Federal Government
Labor cost ‘Kurzarbeitergeld’: public Up to 87% of last net Lowering mid-term insolvency
subsidies wage compensations for income for up to 21 months risk, favoring insolvency gap
employees’ reduced (including social security
working hours if at least charges) per employee
10% of workforce in
short-time
Intertemporal Various tax-related e250bn overall Weakly lowering mid-term
liquidity support deferrals (estimated) insolvency risk, weakly
favoring insolvency gap
Change in insolvency regime
Temporary ‘German COVID-19 Full suspension until No effect on actual insolvency
suspension Insolvency Law Amendment’: September 30, 2020 risk eventually giving the firm
of the obligation Temporarily releases from Suspension until January time to take up liquidity support
to file for the legal obligation to 31, 2021 in case of and to make arrangements for
insolvency disclose insolvency in case over-indebtedness their financing and restructuring
of (1) iliquidity, (2) with its creditors, strongly
imminent iliquidity or favoring insolvency gap
(3) over-indebtedness
Note: The table provides an overview of the early policy measures to support companies depending on the size of the company. Only
the most important first-round policy instruments which are likely to have an impact on corporate insolvencies are presented
Size classes: micro-enterprise, small enterprise, medium-sized enterprise, large enterprise
912 J.O. Do¨rr et al.
pre-crisis crisis
0.05
0.04
0.03
Density
0.02
0.01
0.00
-3 5
-3 0
0
-2 0
5
-2 0
0
-1 0
5
-1 0
00 -3 5
-3 0
0
-2 0
5
-2 0
0
-1 0
5
-1 0
00
-5
10
15
20500
0
0
0
0 -5
10
15
20 0
500
0
0
0
25
30
35 0
0
0 25
30
35 0
0
0
Credit rating update ( rt )
Fig. 8 Distribution of credit rating update in pre-crisis and last rating, the company gets assigned the same index as before,
crisis period. We define a rating update as a reassessment of resulting in a value of 0 in rt . Form the figure it becomes
the company’s creditworthiness performed by Creditreform. We apparent that there is a rightward shift in the distribution of rat-
have precise information on the date of reassessment, which ing updates during the crisis period, indicating that there were
allows us to accurately assign the update to either the pre-crisis more credit rating downgrades as compared to the pre-pandemic
or the crisis period and also to accurately match the updates period. This reflects that the financial situation deteriorated for
with insolvency dates. It should also be noted that a reassess- a larger share of companies in the crisis period than in the three
ment does not necessarily lead to a change in the rating index. If years preceding the crisis
the creditworthiness of the company has not changed since the
Table 11 Calculation of the insolvency gap in absolute terms
Sector Size of company
Micro Small Medium
Ns I Gs (in %) Ns I Gs (in %) Ns I Gs (in %)
Accommodation & catering 37,633 0.0115 4,852 0.0005 810 0.0028
Creative industry & entertainment 16,057 0.0012 1,910 0.0017 476 0.0000
Food production 8,191 0.0027 3,674 0.0024 1,962 −0.0019
Health & social services 69,029 0.0037 12,331 0.0005 4,269 −0.0011
Insurance & banking 46,670 0.0037 2,583 0.0000 1,290 0.0000
Logistics & transport 43,899 0.0070 10,756 0.0002 2,773 0.0030
Chemicals & pharmaceuticals 5,170 0.0033 3,980 0.0003 2,342 0.0000
Small firms and the COVID-19 insolvency gap 913
Table 11 (continued)
Sector Size of company
Micro Small Medium
Ns I Gs (in %) Ns I Gs (in %) Ns I Gs (in %)
Manufacturing of data proc. eq. 4,270 0.0044 2,449 −0.0009 1,057 0.0000
Mechanical engineering 10,567 0.0003 6,828 0.0018 3,386 −0.0025
Business-related services 287,115 0.0070 40,448 −0.0001 9,871 −0.0005
Manufacturing 251,027 0.0103 50,447 0.0002 12,399 −0.0004
Others 37,695 0.0037 5,381 −0.0002 2,398 0.0000
Wholesale & retail trade 201,838 0.0107 46,342 0.0004 10,549 0.0001
Weighted insolvency gap (in %) 0.0080 0.0003 −0.0003
Number of active firms (official statistics) 3,109,261 293,610 63,928 3,466,799
Insolvency gap (absolute) 24,933 90 −19 25,004
Note: Weighted insolvency gap of each size class is calculated as average of the sector specific insolvency gap estimates weighted by
the number of observations of the overall sample in the respective stratum. Number of active firms in Germany reflect the latest official
statistics of the Federal Statistical Office. Insolvency gap in absolute terms is calculated as product between the weighted insolvency
gap and the total number of active German firms within the respective size class. Due to the small number of large firm insolvencies,
we refrain from converting the estimates into absolute numbers in this size class
Table 12 Improvement in balance through matching
Sector Size % Improvement in eCDF mean Variance ratio
rt rt−x r̄t dt at rt rt−x r̄t dt at
Accommodation & catering Micro 97 89 90 96 71 1.00 1.01 1.01 1.01 1.38
Small 96 47 45 67 −15 1.00 1.10 1.11 1.04 1.66
Medium 96 6 6 −16 −119 1.00 1.25 1.26 1.10 2.26
Large 91 22 3 47 −71 1.13 0.71 0.67 1.31 6.59
Business-related services Micro 95 91 92 99 89 1.01 1.01 1.01 1.00 1.02
Small 88 57 72 95 81 1.02 1.05 1.06 1.01 1.09
Medium 83 78 84 88 65 1.02 1.06 1.09 1.01 1.14
Large 74 22 −2 47 25 1.04 1.09 1.12 1.05 1.21
Chemicals & pharmaceuticals Micro 81 40 41 19 40 1.01 1.09 1.12 1.04 1.09
Small 74 47 64 83 47 1.02 1.07 1.10 1.06 1.16
Medium 81 14 63 66 −13 1.03 1.14 1.13 1.09 1.22
Large 74 −30 1 3 11 1.01 1.23 1.23 1.18 1.22
Creative industry & entertainment Micro 95 71 76 59 36 1.01 1.03 1.03 1.02 1.09
Small 95 53 26 65 23 1.01 0.99 0.97 1.08 1.26
914 J.O. Do¨rr et al.
Table 12 (continued)
Sector Size % Improvement in eCDF mean Variance ratio
rt rt−x r̄t dt at rt rt−x r̄t dt at
Medium 94 −71 −64 −3 −79 1.03 1.29 1.33 1.11 2.15
Large 87 1 19 87 −19 1.07 0.81 0.72 1.05 0.60
Food production Micro 85 77 79 67 32 1.01 1.04 1.03 1.04 1.16
Small 83 36 42 89 −28 1.00 1.03 1.03 1.02 1.18
Medium 71 −50 −19 12 −58 1.03 1.28 1.29 1.07 1.14
Large 62 55 47 33 34 1.00 1.48 1.48 1.22 1.02
Health & social services Micro 95 89 91 92 80 1.01 1.01 1.01 1.01 1.18
Small 89 47 50 25 35 1.02 1.05 1.05 1.03 1.17
Medium 84 42 24 78 30 1.01 1.12 1.12 1.08 1.11
Large 71 54 39 79 −39 1.03 1.24 1.27 1.10 1.44
Insurance & banking Micro 90 86 88 89 80 1.01 1.02 1.02 1.00 1.05
Small 73 49 67 82 73 1.03 1.04 1.04 1.07 1.07
Medium 52 61 50 92 63 1.04 1.05 1.07 1.00 1.06
Large 79 59 59 96 71 1.08 1.17 1.16 1.02 0.98
Logistics & transport Micro 93 87 87 93 62 1.01 1.02 1.02 1.01 1.05
Small 89 −27 49 0 53 1.02 1.12 1.13 1.04 1.09
Medium 84 −9 35 52 11 1.02 1.14 1.16 1.07 1.19
Large 85 −78 −107 77 −9 1.05 1.30 1.26 1.09 1.26
Manufacturing Micro 93 92 93 97 82 1.01 1.01 1.01 1.00 1.03
Small 87 72 80 99 68 1.02 1.04 1.05 1.00 1.04
Medium 84 −2 54 67 37 1.02 1.09 1.11 1.03 1.07
Large 85 −27 19 73 −23 1.03 1.23 1.20 1.12 1.25
Manufacturing of data processing equipment Micro 74 −20 0 89 44 1.01 1.06 1.05 1.05 1.19
Small 71 19 43 81 32 1.01 1.20 1.16 1.07 1.28
Medium 73 −31 23 −35 45 1.05 1.55 1.58 1.11 1.27
Large 79 32 56 65 3 1.04 1.11 1.04 1.17 1.60
Mechanical engineering Micro 83 27 31 92 45 1.01 1.07 1.06 1.01 1.11
Small 77 −9 37 89 25 1.02 1.10 1.13 1.02 1.12
Medium 85 −22 37 78 −7 1.01 1.22 1.18 1.05 1.19
Large 87 3 18 54 28 1.02 1.30 1.39 1.08 1.16
Others Micro 93 88 90 91 64 1.01 1.01 1.02 1.01 1.12
Small 86 32 52 93 35 1.02 1.02 1.04 1.01 1.26
Medium 84 45 63 72 42 1.03 1.08 1.08 1.02 1.05
Large 74 25 48 80 −47 1.01 1.08 1.06 1.12 1.11
Wholesale & retail trade Micro 96 91 92 98 82 1.01 1.01 1.01 1.00 1.03
Small 92 11 49 93 67 1.01 1.04 1.04 1.01 1.04
Medium 87 33 66 82 42 1.02 1.11 1.11 1.02 1.10
Large 83 30 45 75 20 1.03 1.18 1.17 1.08 1.15
Note: The table shows balance assessment statistics for all matching variables. % improvement in empirical cumulative density func-
tion (eCDF) mean shows by how much percent the deviation in the eCDF mean between pre-crisis and crisis observations has improved
through nearest neighbor matching. It becomes apparent that for most covariates in all sector-size strata a substantial improvement in
balance has been achieved through the matching process. Variance ratio statistics refer to the ratio of the variance among the matched
control observations and the variance among the crisis observations for the respective variable. Values closer to zero indicate better
balance in variance
Small firms and the COVID-19 insolvency gap 915
Open Access This article is licensed under a Creative Com- the National Academy of Sciences, 117(30), 17656–17666.
mons Attribution 4.0 International License, which permits https://doi.org/10.1073/pnas.2006991117.
use, sharing, adaptation, distribution and reproduction in any Baumol, W. J. (1990). Entrepreneurship: productive, Unproduc-
medium or format, as long as you give appropriate credit to tive, and Destructive. Journal of Political Economy, 98(5,
the original author(s) and the source, provide a link to the Cre- Part 1), 893–921. https://doi.org/10.1086/261712.
ative Commons licence, and indicate if changes were made. The Berger, A. N., Cerqueiro, G., & Penas, M. F. (2011). Does
images or other third party material in this article are included debtor protection really protect debtors? Evidence from
in the article’s Creative Commons licence, unless indicated oth- the small business credit market. Journal of Banking
erwise in a credit line to the material. If material is not included and Finance, 35(7), 1843–1857. https://doi.org/10.1016/
in the article’s Creative Commons licence and your intended use 12.010j.jbankfin.2010.
is not permitted by statutory regulation or exceeds the permit- Bersch, J., Gottschalk, S., Mueller, B., & Niefert, M.
ted use, you will need to obtain permission directly from the (2014). The Mannheim Enterprise Panel (MUP) and Firm
copyright holder. To view a copy of this licence, visit http:// Statistics for Germany. ZEW Discussion Paper, 14–104.
creativecommons.org/licenses/by/4.0/. https://doi.org/10.2139/ssrn.2548385.
Block, J. H., Fisch, C., & Hirschmann, M. (2021). The
determinants of bootstrap financing in crises: evi-
dence from entrepreneurial ventures in the COVID-19
pandemic. Small Business Economics, (forthcoming).
References https://doi.org/10.1007/s11187-020-00445-6.
Brouwer, M. (2006). Reorganization in US and European
Acs, Z. J., Desai, S., & Hessels, J. (2008). Entrepreneur- bankruptcy law. European Journal of Law and Economics,
ship, economic development and institutions. Small 22(1), 5–20. https://doi.org/10.1007/s10657-006-8978-2.
Business Economics, 31(3), 219–234. https://doi.org/10. Caballero, R. J., & Hammour, M. L. (1994). The cleansing
1007/s11187-008-9135-9. effect of recessions. American Economic Review, 84(5),
Adalet McGowan, M., Andrews, D., & Millot, V. (2018). The 1350–1368. http://www.jstor.org/stable/2117776.
walking dead? Zombie firms and productivity performance Caballero, R. J., Hoshi, T., & Kashyap, A. K. (2008). Zom-
in OECD countries. Economic Policy, 33(96), 685–736. bie lending and depressed restructuring in Japan. Amer-
https://doi.org/10.1093/epolic/eiy012. ican Economic Review, 98(5), 1943–1977. https://doi.
Agarwal, V., & Taffler, R. (2008). Comparing the performance org/10.1257/aer.98.5.1943.
of market-based and accounting-based bankruptcy predic- Cahuc, P., Kramarz, F., & Nevoux, S. (2018). When short-
tion models. Journal of Banking & Finance, 32(8), 1541– time work works. Banque de France Working Paper, 692.
1551. https://doi.org/10.1016/j.jbankfin.2007.07.014. https://doi.org/10.2139/ssrn.3247486.
Altman, E. I. (1968). Financial ratios, discriminant analysis Carreira, C., & Teixeira, P. (2016). Entry and exit in severe
and the prediction of corporate bankruptcy. The Journal recessions: lessons from the 2008–2013 Portuguese eco-
of Finance, 23(4), 589–609. https://doi.org/10.1111/j.1540- nomic crisis. Small Business Economics, 46(4), 591–617.
6261.1968.tb00843.x. https://doi.org/10.1007/s11187-016-9703-3.
Altman, E. I. (2013). Predicting financial distress of compa- Chemin, M. (2009). The impact of the judiciary on
nies: revisiting the Z-Score and ZETA® models. In A.R. entrepreneurship: evaluation of Pakistan’s “Access to Jus-
Bell, C. Brooks, & M. Prokopczuk (Eds.) Handbook of tice Programme”. Journal of Public Economics, 93(1–2),
research methods and applications in empirical finance 114–125. https://doi.org/10.1016/j.jpubeco.2008.05.005.
(pp. 428–456). Edward Elgar Publishing. https://doi.org/10. Chowdhury, F., Terjesen, S., & Audretsch, D. (2015). Vari-
4337/9780857936097.00027. eties of entrepreneurship: institutional drivers across
Anderson, J., et al. (2020). The fiscal response to the entrepreneurial activity and country. European Journal of
economic fallout from the coronavirus. Accessed: 28 Law and Economics, 40(1), 121–148. https://doi.org/10.10
Dec 2020. https://www.bruegel.org/publications/datasets/ 07/s10657-014-9464-x.
covid-national-dataset/. Cook, G. A., Pandit, N. R., & Milman, D. (2001). Formal
Archibugi, D., Filippetti, A., & Frenz, M. (2013). Economic cri- rehabilitation procedures and insolvent firms: empirical
sis and innovation: is destruction prevailing over accumu- evidence on the British Company voluntary arrangement
lation? Research Policy, 42(2), 303–314. https://doi.org/10. procedure. Small Business Economics, 17(4), 255–271.
1016/j.respol.2012.07.002. https://doi.org/10.1023/A:1012293605945.
Arcuri, G., & Levratto, N. (2020). Early stage SME bankruptcy: Council of the European Union (2020). Report on the com-
does the local banking market matter? Small Business Eco- prehensive economic policy response to the COVID-19
nomics, 54(2), 421–436. https://doi.org/10.1007/s11187- pandemic. Accessed: 12 Dec 2020. https://www.consilium.
018-0042-4. europa.eu/en/press/press-releases/2020/04/09/report-on-the
Barrero, J. M., Bloom, N., & Davis, S. (2020). COVID-19 Is -comprehensive-economic-policy-response-to-the-covid-
Also a Reallocation Shock. NBER Working Paper 27137. 19-pandemic/.
https://doi.org/10.3386/w27137. Cowling, M., Liu, W., & Ledger, A. (2012). Small business
Bartik, A. W., Bertrand, M., Cullen, Z., Glaeser, E. L., Luca, financing in the UK before and during the current financial
M., & Stanton, C. (2020). The impact of COVID-19 on crisis. International Small Business Journal, 30(7), 778–
small business outcomes and expectations. Proceedings of 800. https://doi.org/10.1177/0266242611435516.
916 J.O. Do¨rr et al.
Cowling, M., Brown, R., & Rocha, A. (2020). Did you save Federal Ministry for Economic Affairs and Energy (2020).
some cash for a rainy COVID-19 day? The crisis and KfW Instant Loan for small and medium- sized enter-
SMEs. International Small Business Journal, 38(7), 593– prises to be launched tomorrow. Accessed: 29 Dec 2020.
604. https://doi.org/10.1177/0266242620945102. https://www.bmwi.de/Redaktion/EN/Pressemitteilungen/20
Creditreform (2020a). Creditreform Commercial Report Inter- 20/20200414-kfw-instant-loan-for-small-and-medium-siz
national. Accessed: 1 Dec 2020. https://www.creditreform. ed-enterprises-to-be-launched-tomorrow.html.
com/fileadmin/user upload/CR-International/Bilder/PB Federal Ministry of Finance (2020a). Additional KfW
International-Commercial-Report web.pdf. Special Programme 2020 for the economy to be
Creditreform (2020b). Creditreform Solvency Index. Accessed: launched today. Accessed: 5 Jan 2021. https://www.
1 Dec 2020. https://en.creditreform.de/fileadmin/user bundesfinanzministerium.de/Content/EN/Pressemitteilungen/
upload/crefo/download eng/commercial information/ 2020/2020-03-23-KfW-special-programme.html.
Flyer Solvency Index.pdf. Federal Ministry of Finance (2020b). Corona virus: immedi-
Cros, M., Epaulard, A., & Martin, P. (2021). Will Schumpeter ate federal economic assistance now available. Accessed:
Catch Covid-19? CEPR Discussion Paper, 15834. https:// 5 Jan 2021. https://www.bundesfinanzministerium.
ssrn.com/abstract=3795217. de/Content/EN/Pressemitteilungen/2020/
Destatis (2020). Shares of small and medium-sized enter- 2020-04-01-corona-federal-economic-assistance.html.
prises in selected variables, 2018. Accessed: 17 Dec 2020. Federal Ministry of Finance (2020c). German government
https://www.destatis.de/EN/Themes/Economic-Sectors-En launches temporary aid scheme as part of coronavirus stim-
terprises/Enterprises/Small-Sized-Enterprises-Medium-Size ulus. Accessed: 28 July 2020. https://www.bundesfinanz
d-Enterprises/Tables/total-cik.html. ministerium.de/Content/EN/Standardartikel/Topics/Priority-
Dewaelheyns, N., & Van Hulle, C. (2008). Legal reform Issues/Artvicles/2-020-07-08-temporary-aid-scheme-launch
and aggregate small and micro business bankruptcy ed.html.
rates: evidence from the 1997 Belgian bankruptcy Federal Ministry of Finance (2020d). German Stability Pro-
code. Small Business Economics, 31(4), 409–424. gramme 2020. Accessed: 28 Dec 2020. https://www.
https://doi.org/10.1007/s11187-007-9060-3. bundesfinanzministerium.de/Content/EN/Standardartikel/
Didier, T., Huneeus, F., Larrain, M., & Schmukler, S. L. Press Room/Publications/Brochures/2020-04-17-german-
(2021). Financing firms in hibernation during the COVID- stability-programme-2020.pdf? blob=publicationFile&v=9.
19 pandemic. Journal of Financial Stability, 53, 100837. Federal Ministry of Justice and Consumer Protection (2020).
https://doi.org/10.1016/j.jfs.2020.100837. Act to Mitigate the Consequences of the COVID-19
Djankov, S., McLiesh, C., & Shleifer, A. (2007). Private credit
Pandemic under Civil, Insolvency and Criminal Proce-
in 129 countries. Journal of Financial Economics, 84(2),
dure Law. Accessed: 28 July 2020. https://www.bmjv.de/
299–329. https://doi.org/10.1016/j.jfineco.2006.03.004.
SharedDocs/Gesetzgebungsverfahren/Dokumente/Bgbl
Eberhart, R. N., Eesley, C. E., & Eisenhardt, K. M. (2017). Fail-
Corona-Pandemiev EN.pdf? blob=publicationFile&v=2.
ure is an option: institutional change, entre-preneurial risk,
Garcı́a-Posada, M., & Mora-Sanguinetti, J. S. (2015). Does
and new firm growth. Organization Science, 28(1), 93–112.
(average) size matter? Court enforcement, business demog-
https://doi.org/10.1287/orsc.2017.1110.
Estrin, S., Mickiewicz, T., & Rebmann, A. (2017). Prospect the- raphy and firm growth. Small Business Economics, 44(3),
ory and the effects of bankruptcy laws on entrepreneurial 639–669. https://doi.org/10.1007/s11187-014-9615-z.
aspirations. Small Business Economics, 48(4), 977–997. Giupponi, G., & Landais, C. (2018). Subsidizing labor hoard-
https://doi.org/10.1007/s11187-016-9810-1. ing in recessions: the employment & welfare effects of
European Commission (2003). Commission recommendation short time work. CEP Discussion Papers, 13310. https://
concerning the definition of micro, small and medium- ssrn.com/abstract=3287057.
sized enterprises. Official Journal of the European Guerini, M., Nesta, L., Ragot, X., & Schiavo, S. (2020). Firm
Union, L124, 36–41. https://eur-lex.europa.eu/LexUriServ/ liquidity and solvency under the Covid-19 lockdown in
LexUriServ.do?uri=OJ:L:2003:124:0036:0041:EN:PDF. France. OFCE Policy Brief, 76.
European Union (2006). Regulation (EC) No 1893/2006 of the Gurrea-Martı́nez, A. (2020). Insolvency Law in Times of
European Parliament and of the Council of 20 December COVID-19. SSRN Electronic Journal. https://doi.org/10.
2006 establishing the statistical classification of economic 2139/ssrn.3562685.
activities NACE Revision 2 and amending Council Reg- Iyer, R., Peydró, J. L., Da-Rocha-Lopes, S., & Schoar,
ulation (EEC) No 3037/90 as well as certain EC Regula. A. (2014). Interbank liquidity crunch and the firm
Official Journal of the European Union, L393, 1–39. http:// credit crunch: evidence from the 2007-2009 cri-
data.europa.eu/eli/reg/2006/1893/oj. sis. Review of Financial Studies, 27(1), 347–372.
Fairlie, R. (2020). The impact of COVID-19 on small https://doi.org/10.1093/rfs/hht056.
business owners: evidence from the first three months Juergensen, J., Guimón, J., & Narula, R. (2020). European
after widespread social-distancing restrictions. Journal of SMEs amidst the COVID-19 crisis: assessing impact and
Economics and Management Strategy, 29(4), 727–740. policy responses. Journal of Industrial and Business Eco-
https://doi.org/10.1111/jems.12400. nomics, 47(3), 499–510. https://doi.org/10.1007/s40812-
Federal Government of Germany (2020). Soforthilfen für Selb- 020-00169-4.
stständige und kleine Unternehmen. Accessed: 29 Nov Kalemli-Ozcan, S., Gourinchas, O. P., Penciakova, V., &
2020. https://www.bundesregierung.de/breg-de/aktuelles/ Sander, N. (2020). COVID-19 and SME failures. NBER
corona-soforthilfen-1737444. Working Paper, 27877. https://doi.org/10.3386/w27877.
Small firms and the COVID-19 insolvency gap 917
Kopp, D., & Siegenthaler, M. (2021). Short-time work in and data from complex sample surveys: Chi-squared tests for
after the Great Recession. Journal of the European Eco- goodness of fit and independence in two-way tables. Jour-
nomic Association (forthcoming). https://doi.org/10.1093/ nal of the American Statistical Association, 76(374), 221–
jeea/jvab003. 230. https://doi.org/10.1080/01621459.1981.10477633.
Lee, N., Sameen, H., & Cowling, M. (2015). Access to finance Rodano, G., Serrano-Velarde, N., & Tarantino, E. (2016).
for innovative SMEs since the financial crisis. Research Bankruptcy law and bank financing. Journal of Finan-
Policy, 44(2), 370–380. https://doi.org/10.1016/j.respol. cial Economics, 120(2), 363–382. https://doi.org/10.1016/j.
2014.09.008. jfineco.2016.01.016.
Lee, S. H., Yamakawa, Y., Peng, M. W., & Barney, J. B. Rubin, D. B. (1973). Matching to remove bias in observa-
(2011). How do bankruptcy laws affect entrepreneurship tional studies. Biometrics, 29(1), 159–183. https://doi.org/
development around the world? Journal of Business Ven- 10.2307/2529684.
turing, 26(5), 505–520. https://doi.org/10.1016/j.jbusvent. Rubin, D. B. (1980). Bias reduction using Mahalanobis-Metric
2010.05.001. matching. Biometrics, 36(2), 293–298. https://doi.org/10.
Legrand, M. D. P. (2017). Retrospectives: do produc- 2307/2529981.
tive recessions show the recuperative powers of capi- Schumpeter, J. A. (1942). Capitalism, socialism and democ-
talism? Schumpeter’s analysis of the cleansing effect. racy. New York: Harper.
Journal of Economic Perspectives, 31(1), 245–256. Stuart, E. A. (2010). Matching methods for causal inference: a
https://doi.org/10.1257/jep.31.1.245. review and a look forward. Statistical Science, 25(1), 1–21.
Levie, J., Autio, E., Acs, Z., & Hart, M. (2014). Global https://doi.org/10.1214/09-STS313.
entrepreneurship and institutions: an introduction. Small The Economist (2020a). The corporate undead: what
Business Economics, 42(3), 437–444. https://doi.org/10. to do about zombie firms. Accessed: 5 Jan 2021.
1007/s11187-013-9516-6. https://www.economist.com/leaders/2020/09/24/
McGuinness, G., & Hogan, T. (2016). Bank credit and trade what-to-do-about-zombie-firms.
credit: evidence from SMEs over the financial crisis. The Economist (2020b). The corporate undead: why covid-19
International Small Business Journal, 34(4), 412–445. will make killing zombie firms of harder. Accessed: 5 Jan
https://doi.org/10.1177/0266242614558314. 2021. https://www.economist.com/finance-and-economics/
2020/09/26/why-covid-19-will-make-killing-zombie-firms-
Melcarne, A., & Ramello, G. B. (2020). Bankruptcy delay and
off-harder.
firms’ dynamics. Small Business Economics, 54(2), 405–
The Washington Post (2020). Here’s one more economic
419. https://doi.org/10.1007/s11187-018-0041-5.
problem the government’s response to the virus has
OECD (2020a). Coronavirus (COVID-19): SME pol- unleashed: Zombie firms. Accessed: 11 Nov 2020.
icy responses. Accessed: 17 Dec 2020. https://read. https://www.washingtonpost.com/business/2020/06/23/
oecd-ilibrary.org/view/?ref=119 119680-di6h3qgi4x& economy-debt-coronavirus-zombie-firms/.
title=Covid-19 SME Policy Responses. Wang, J., Yang, J., Iverson, B. C., & Kluender, R. (2020).
OECD (2020b). Corporate sector vulnerabilities during the Bankruptcy and the COVID-19 Crisis. SSRN Electronic
Covid-19 outbreak: assessment and policy responses. Journal. https://doi.org/10.2139/ssrn.3690398.
Accessed: 17 Dec 2020. https://doi.org/10.1787/ Yang, C. H., & Chen, K. H. (2009). Are small firms less
6434b1e4-en. efficient? Small Business Economics, 32(4), 375–395.
Peng, M. W., Yamakawa, Y., & Lee, S. H. (2010). Bankruptcy https://doi.org/10.1007/s11187-007-9082-x.
laws and entrepreneur- friendliness. Entrepreneurship:
Theory and Practice, 34(3), 517–530. https://doi.org/10. Publisher’s note Springer Nature remains neutral with regard
1111/j.1540-6520.2009.00350.x. to jurisdictional claims in published maps and institutional
Rao, J. N. K., & Scott, A. J. (1981). The analysis of categorical affiliations.
File and source
- File
- dorr-2021-small-firms-and-the-covid-19-insolvency-gap.pdf
- Size
- 1,402,209 bytes
- SHA-256
- 58d75170c66642286ab5c58e27133650c645eac76e2038e747659c94e3857a42
- Original
- link.springer.com