Marc Cowling et al., Has previous loan rejection scarred firms from applying for loans during Covid-19? (2021)
Summary
A 2021 paper by Marc Cowling and co-authors asking whether previous loan rejection scarred firms from applying for loans during Covid-19. It is a 714,915-byte PDF whose original is on link.springer.com.
Summary drafted by a model from the text of the original file linked on this page 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
“Marc Cowling et al., Has previous loan rejection scarred firms from applying for loans during Covid-19? (2021).”
Original at link.springer.com. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/marc-cowling-et-al-has-previous-loan-rejection-scarred-firms-from-appl-004b2bc07d/
, SHA-256 7188b86ea47ca0dff4cc409f4b188f6fa1ad49b767a52591951e7661afe20326.Full text
Small Bus Econ (2022) 59:1327–1350
https://doi.org/10.1007/s11187-021-00586-2
Has previous loan rejection scarred firms from applying
for loans during Covid‑19?
Marc Cowling · Weixi Liu · Raffaella Calabrese
Accepted: 6 November 2021 / Published online: 18 December 2021
© The Author(s) 2021
Abstract The concept of the ‘discouraged’ bor- scarred firms faced such severe liquidity problems
rower is well documented. In this paper, we consider that they relaxed their distrust of banks during the
whether smaller firms in the UK who have been pre- Covid-19 crisis. However, their share of the govern-
viously rejected for bank loans have been scarred by ment-guaranteed loan portfolio was slightly lower
the experience so badly that even in the presence of suggesting that banks were treating each new loan
two exceptionally generous Covid-19 loan guarantee application on its merits.
schemes, they still refuse to make an application. Fur-
thermore, we also consider what happens when they Plain English Summary The Covid-19 crisis
do. As banks have either zero or minimal loss expo- hit smaller businesses so hard that even previously
sure, do they still maintain their normal strict lend- rejected borrowers were forced to apply for loans to
ing protocols or do they relax their standards to fulfil keep them afloat. Previous loan rejections have not
the governments’ objective of supporting struggling discouraged small businesses in the UK in applying
businesses through the crisis? Our findings show that for Covid-19 government-guaranteed loans. Banks
72% of previously rejected borrowers are reluctant to have used the loan guarantee schemes to continue to
request loans. We find some evidence that previously supply loans to small business during the pandemic.
Our paper analyses the important phenomenon of
borrower scarring and discouragement, when poten-
Supplementary Information The online version
contains supplementary material available at https://doi. tial debtors are self-excluded from the lending mar-
org/10.1007/s11187-021-00586-2. ket because they have previous rejections or expect
a negative bank response. We consider around
M. Cowling (*) 45,000 UK small businesses from 2018 to 2020. On
University of Derby, Kedlestone Road, Derby DE22 1GB,
UK the demand side, we find that the economic shock
e-mail: M.Cowling@derby.ac.uk for small businesses during the pandemic dissipates
the scarring effect. Specifically, we find that micro
W. Liu and small businesses had the highest loan demand in
University of Bath, The Avenue, Claverton Down,
Bath BA2 7AY, UK the first two quarters of the pandemic (from March
e-mail: w.liu@bath.ac.uk 2020). On the supply side, we show that scarred bor-
rowers were not routed onto Covid-19 government-
R. Calabrese guaranteed loan schemes. These findings show the
University of Edinburgh, Old College South Bridge,
Edinburgh EH8 9YL, UK importance of government-backed lending schemes
e-mail: Raffaella.Calabrese@ed.ac.uk for small businesses during crisis period.
Vol.:(0123456789)
13
1328 M. Cowling et al.
Keywords Scarred borrowers · Discouraged established firm behaviours during the crisis before we
borrowers · Bank loans · SMEs · Covid-19 can predict the potential longer run impacts.
The lens that we look through in this paper is
JEL Classification H81 · L26 · G21 · G01 · G4 the small firm debt market. We analyse 10 waves of
cross-sectional data on 45,000 UK SMEs from the
second quarter in 2018 to the third quarter in 2020.
Our primary research question is whether scarring
1 Introduction due to pre-Covid-19 loan application rejections has
prevented struggling small firms from putting forward
The concept of the discouraged borrower is well loan applications even in the presence of the two
documented (Kon & Storey, 2003) and has been most generous ever government-backed loan guaran-
identified as an important real-world phenomenon tee programmes (CBILS, the Coronavirus Business
(Cole & Sokolyk, 2016; Cowling et al., 2016; Fraser, Interruption Loan; and BBLS, the Bounce Back Loan
2009). For the main however, it is framed in a rather Scheme) which offer a guarantee coverage of 80% and
static way as certain groups of potential borrow- 100%, respectively, to the lending bank. In this sense,
ers self-exclude from the loan market on the basis we are questioning how bank loan rejections in the
of their perceptions of a negative bank response. In ‘good’ times pre-Covid-19 might have permanently
this paper, we extend this framing to include poten- scarred small businesses to the extent that even dur-
tial borrowers who have had previous loan applica- ing Covid-19, when trading activities were forcibly
tion rejections and are ‘scarred’ by this negative restricted, and many firms faced significant liquidity
experience so much that they self-exclude from the and cash flow pressures, they were more reluctant to
loan market after that point. In this sense, we argue put forward loan applications.
that the actual reality of being refused a loan adds We then consider the loan supply side of the small
more tangibility to the decision to self-exclude as it firms’ debt market and question whether banks have
is based on lived experience rather than perception. relaxed their lending standards in the presence of the
We term this group scarred borrowers. Furthermore, two generous loan guarantee schemes. Previous research
having identified the extent and nature of scarred highlights the fact that the incidence of financial delin-
borrowers, we overlay their current financing behav- quency (e.g. unauthorised overdraft limit transgressions,
iours with the Covid-19 crisis and question whether legal non-payment of debt obligations, late payment of
the sheer severity and scale of the crisis have encour- taxes, etc.) is a fundamental determinant of the banks’
aged scarred borrowers to change their behaviours. lending decision-making process (Cowling et al., 2016).
Scarring is a well-understood concept in Labour Thus, if previous loan application rejections informed
Economics where scarring from lengthy periods of their historical rejections, then it follows that banks
unemployment has been found to be an important char- viewed these firms as bad credit risks at that time. The
acteristic that also has very real and negative economic current state of the world and particularly the presence
effects (Arulampalam, 2001), particularly during ‘Great of a zero or 20% lending risk may have encouraged
Recessions’ and immediately afterward (Tumino, banks to lower their normal lending standards in this
2015). It has also been used as to describe the poten- respect which would result in an increased willingness
tial long-run legacy of the Covid-19 crisis. On this, to lend during the Covid-19 crisis.
Kozlowski et al., (2020: page 1) argue that the crisis The rest of the paper is organised as follows: in
has caused a scarring of beliefs, which they define as Section 2, we review the literatures on small firm debt
a ‘persistent change in the perceived probability of an markets and the loan contracting process. We draw in
extreme, negative shock in the future’. It is this aspect the related concepts of the discouraged borrower and
of this latter concept that is most relevant in this study consider the importance of an established pecking
given the current debates about what the post-Covid-19 order or financing hierarchy. We also present a theo-
recovery might look like and how long it might take retical model for scarred borrowers. In Section 3, we
firms to return to their steady-state patterns of sav- present our data, discuss the sample demographics and
ings, borrowing and investment. Our argument is that outline our analytical methodology for exploring our
we need to establish whether Covid-19 itself changed
13
Has previous loan rejection scarred firms from applying for loans during Covid‑19? 1329
fundamental research question. Section 4 presents our But the precise cause and severity of each crisis are
core modelling results, and we conclude in Section 5. different and hence any learning is non-parametric
as it is derived from the tails of the risk belief dis-
tribution (Kozlowski et al., 2020).
The most recent events that generate financial tur-
2 Literature review moil that had a significant impact on the credit avail-
ability for SMEs in the UK are the GFC, Brexit and
In this section, we draw from a broad literature relat- the Covid pandemic. The literature on the effects of
ing to small firm debt markets and loan contracting the GFC on access to finance for SMEs is extensive
with a particular focus on a financial pecking order (Cowling et al., 2012; Popov & Udell, 2012; Udell,
and the presence of borrower discouragement. Prior 2020 amongst others), but the number of empirical
to that, we consider the literature around negative studies on the impact of Brexit is still limited (Cala-
economic shocks and how they might influence brese et al., 2020; Brown et al., 2019). This num-
beliefs about the future potential for extreme events ber is even lower for the Covid crisis because it is a
to occur and the impact if they do. This has relevance recent event and, therefore, limited data is available
both to the concept of scarring and the perception (Cowling et al., 2021). One of the aims of this paper
that a negative future event will occur. is to fill this gap in the literature and provide some
empirical evidence on the Covid crisis.
Negative economic shocks could have differ-
ent effects on the demand for external finance. On
2.1 Negative economic shocks the one hand, it is likely that loan demand declines
when bank capital decreases, such as during the
The UK Brexit vote in June 2016 delivered an GFC, because SMEs are usually discouraged given
unexpected outcome (the vote to leave the Euro- the high number of unsuccessful loan applications
pean Union) and the intervening 4 years until the in such periods (Cowling et al., 2016). On the other
Covid-19 crisis hit was one of uncertainty and low hand, several empirical studies showed that the
borrowing and investment. In its’ own right, Brexit demand for external finance increases during eco-
could be seen as an extreme event which altered nomic downturn (Bank of England, 1993; Binks
behaviours. Ten months before the final exit from et al., 1992). Specifically, Cowling et al., (2012)
the European Union in January 2021, the Covid-19 and Popov and Udell (2012) obtained that as prof-
crisis unfolded another extreme, and unanticipated, its decrease and cash flow tightens, firms are more
event. And this ignores the global financial crisis likely to seek for external finance during the GFC.
(GFC) that occurred in late 2008 and persisted This evidence gives rise to our first hypothesis:
until 2011 in the UK during which banks severely H1: The general demand for loans will increase
rationed credit to smaller firms with loan applica- during Covid-19
tion acceptance reaching a nadir of 65% (Cowl- Negative economic shocks usually generate an
ing et al., 2012). In extremis, some small business increase in uncertainty that exacerbates information
owners, whose banks were happy to lend to prior asymmetries between borrowers and lenders (Mishkin,
to GFC who then faced rejection during GFC, may 2011; Sette & Gobbi, 2015; Jappelli and Pagano,
have downwardly revised their expectations of a 2002). Consequently, banks tend to cut the credit
positive loan outcome and become scarred at this supply during such periods toward small businesses
point, whilst some may have become scarred on an (Cowling et al., 2012). We therefore examine whether
individual basis through a simple loan rejection at the supply of bank loans to small businesses decreases
any time since. Taken overall, most small business during Covid-19 compared, so our second hypothesis
owners will know two things that might shape their is as follows:
view on the world: firstly, that banks do not treat H2: The willingness of banks to supply conven-
them well during a crisis (Udell, 2020), and sec- tional loans will decrease during Covid-19
ondly, that crises happen on a quite regular basis.
13
1330 M. Cowling et al.
2.2 Debt markets and SME loan contracting investment opportunities but have exhausted internal
cash which triggers higher borrowing.
Bank loans are the most prevalent and important The high general levels of non-borrowing amongst
form of external finance for smaller firms (Carpenter the small firm population suggest that in normal cir-
& Petersen, 2002). This in part relates to the relative cumstances, the majority of small firms would self-
unsophistication of the typical small business owner finance their operational and investment needs. In the
in respect of financial acumen and knowledge of Covid-19 crisis, recent research has identified a lack
more complex financial instruments, and also to the of precautionary saving and retained earnings amongst
ease of the application process and the speed of the micro and small businesses and this has directly
bank decision-making process (Howorth, 2001). The affected their liquidity position during the series of
key characteristic of the process of loan contracting business lockdowns that were imposed by government
between small firms and banks is the presence of (Brown & Cowling, 2021; Cowling, Brown, et al.,
asymmetric information—one agent in the contract 2020; Cowling, Liu, et al., 2020; Cowling, Nadeem,
having superior information about the true quality of et al., 2020). This lack of internal funds would imply
the lending proposal than the other (Stiglitz & Weiss, that more firms would move to their second preference
1981). This information advantage is usually assumed which is bank loans to shore up their cash position
to be in favour of the entrepreneur, although some even in the face of declining opportunities for profit-
researchers have argued that a large banking group able investment.
has millions of small business accounts and a his-
tory of repeated loan transactions and that this gives 2.4 Discouraged borrowers
it an informational advantage over the entrepreneur
(De Meza, 2002). Given the primacy of bank loans in The seminal work of Jappelli (1990) identified as
meeting the external financing needs of smaller busi- liquidity-constrained, borrowers who were either dis-
nesses and the offer of a government-guaranteed loan, couraged or turned down for credit, defined as denied
we hypothesise that loan demand and supply will firms. Cole and Sokolyk (2016) showed that discour-
increase during the Covid-19 crisis. aged and denied firms are significantly different in
terms of size, profitability and the number of sources
2.3 Pecking order hypothesis of financial services.
Given this difference, Kon and Storey (2003)
The pecking order hypothesis (Myers & Majluf, defined a theoretical model only for discouraged bor-
1984) states that firms will have a financial pecking rowers, defined as businesses who have a demand for
order within which internally generated funds will capital but do not translate that demand into a formal
be preferred to external debt, and external debt will application for bank loans. In the presence of imper-
be preferred to equity as that involves a legal claim fect information in small firm debt markets, banks
on the ownership of the firm and future stream of incur screening costs and importantly these costs are
profits, thus raising independence and control aver- passed on to the borrower. This raises the application
sion issues. This is a combined outcome of the per- costs to the small firm which makes some potential
sonal preferences of the small business owner and projects unviable as they are unable to make the nor-
also information-based capital constraints in external mal returns required to pay capital and interest plus
markets where insiders have a more perfect knowl- screening costs. Furthermore, banks make screening
edge about the true value of a project, and the firms’ errors, thus rejecting ‘good’ risk loan applications. It
prospects, than outsiders (Watson & Wilson, 2002). follows that the higher the screening costs imposed
It follows that any variation in debt financing will be by the banks and the greater the screening error, the
a direct result of a decline in internal funds (Shyam- higher the level of borrower discouragement.
Sunder & Myers, 1999). Problems with internal cash A subsequent body of empirical measurement,
flows are related with a firms’ investment opportunity quantification and testing of discouraged borrowers
set. A firm can be profitable but have limited invest- established that it is a genuine phenomenon amongst
ment opportunities leading to a big cash pile and low the small business population and quantifiably a non-
debt levels. Equally, a firm can have lots of profitable trivial subset of smaller business populations across
13
Has previous loan rejection scarred firms from applying for loans during Covid‑19? 1331
the world (Cowling et al., 2016; Han et al., 2009). H3: The deeper into the Covid-19 crisis we go,
Some studies (Cole & Sokolyk, 2016; Freel et al., the higher the demand for loans due to cash flow
2012) show that discouraged borrowers can outnum- pressure on firms and formerly scarred borrowers
ber actual turndowns. Other work has focused on (through choice) changing their behaviours
ethnicity (Fraser, 2009), gender (Freel et al., 2012), As the credit supply decreases during severe
Eurozone economies (Mac an Bhaird et al., 2016), recessions, such as the early thirties in the USA
the role of bank-firm trust (Tang et al. 2017) and the (Welsh, 1952) and during the Covid pandemic, the
potential for political connections to influence banks’ government can provide an increment to the sup-
decision-making (Qi & Nguyen, 2021). Importantly, ply of credit through loan guarantee schemes. Given
estimates from this body of empirical work also the unprecedented effects of the Covid-19 pandemic
suggest that between one third and one half of dis- on firms, the UK government supported the capi-
couraged borrowers would have been offered a loan talisation of businesses through the CBILS and the
had they applied. This suggests that self-rationing is BBLS. If a loan guarantee scheme is effective, most
often based on poor assumptions. of the firms obtaining bank loans through such a
In the context of the Covid-19 crisis, we would scheme were unlikely to obtain financing from tra-
expect that greater general uncertainty and higher ditional sources (Cowling & Mitchell, 2003; Rid-
levels of asymmetric information would increase ing et al., 2007). We expect that the more severe the
the costs of screening and discourage more firms economic downturn of the Covid crisis, the lower
from making loan applications. However, this are firm profits and weaker is their cash flow posi-
would potentially be mitigated by the presence of tion, so the higher is their demand for external
the government-backed guaranteed lending schemes finance in general (Cowling et al., 2012) and for the
as screening to reduce lending risk is less essential government-guaranteed loans, as our final hypoth-
when the government guarantee reduces lending risk esis states:
and loss given default significantly. H4: The deeper into the Covid-19 crisis we go,
the higher the use of government-guaranteed loans
to meet the higher demand of credit
2.5 Borrower scarring
In respect of a loan application outcome, the small 2.6 Theoretical model for scarred borrowers
business owner has a view on the potential distribu-
tion of outcomes which is informed by her previous Following Kon and Storey (2003), we propose in
history of loan applications. This initial view on the this section a theoretical model for scarred borrow-
outcome distribution might be revised in the light of ers that are discouraged by applying for external
a loan application event, particularly if it generates a finance given a previous loan rejection. We assume
less favourable outcome than expected. Some empiri- that there are two types of firms, good (G) and bad
cal studies showed that prior loan rejection often (B), and each firm needs one unit of funding for its
discourages a firm from applying again in the future investment that it can borrow from a bank or from
(Cavalluzzo et al., 2002; Xiang et al., 2015). There- an alternative source, called money lender. We also
fore, we define the scarred borrower, who, in the face consider the 10 assumptions reported in Kon and
of an unexpected loan rejection, is so scarred by the Storey (2003) and that a bank can correctly identify
experience that this fundamentally reshapes their view good and bad firms. Under these assumptions, the
on the distribution of future applications outcomes expected return from applying is
that they become reluctant to make future applica-
(1)
( )( )
tions. This means that scarred borrowers are the sub- 1 − pb XG − D − K + pb (w − K)
set of discouraged borrowers that have been previ-
where
ously turned down for credit.
In line with Cowling et al., (2012) and the previ-
• XG—gross return from the project funded by the
ous studies on prior loan rejections, we propose the
loan
following hypothesis:
13
1332 M. Cowling et al.
Fig. 1 The process of Scarred by previous
(pre-Covid-19) historical rejection
borrower scarring 77.50%
[0.91% POPN]
[6.51% borrower
Application rejected POPN]
8.40%
[1.18% POPN] Not scarred by
Previous loan previous rejection
application 22.50%
14.08% Application approved: [0.27% POPN]
Approved borrowers [1.89% borrower
SME Population POPN]
91.60%
No previous loan [12.90% POPN]
application event:
Non-borrowers
85.92%
• D—interest rate charged on the loan If the expected return for non-application is w, a
• K—application costs for the loan firm will apply for loan if the following condition is
• w—net return of the project funded through the satisfied
money lender after interest payment
(3)
( )( )
• pb—probability that the firm assesses itself as 1 − pb XG − D − K + pb (w − K) > w
bad
To incorporate the empirical evidence that prior
loan rejection can discourage a company from apply-
To compute the probability pb, we classify all the
ing in the future (Cavalluzzo et al., 2002; Xiang et al.,
borrowers as firms with no previous loan applica-
2015), we assume in our model that pb3 > pb2.
tions, borrowers whose previous application was
successful and scarred and unscarred rejected bor-
rowers, as displayed in Fig. 1. Based on the law of
3 Data, sample demographics and methodology
total probability, we can compute pb as follows:
(2) In this section, we discuss the source of the data
[ ]
pb = pb1 ∙ p(NPA) + pb2 ∙ p(APA) + pb3 ∙ 1 − p(APA) − p(NPA)
which we used for our empirical analysis of the poten-
where tial impact of previous loan rejection, scarring and
how these impacts on loan demand and supply during
• p(NPA)—the probability of not having a pre- the Covid-19 crisis. We then discuss our methodol-
vious history of loan application ogy for identifying these issues in the context of firms
• p(APA)—the probability of having a previous making a loan application, the banks’ decision and
loan application approved whether successful applications were routed through
• p b1—probability that the firm assesses itself the two UK government loan guarantee schemes.
as Bad given that the firm did not previously
apply for a bank loan 3.1 The SME Finance Monitor
• p b2—probability that the firm assesses itself as
Bad given that the firm had a previous appli- The data we have available for analysis is the most
cation approved recent ten waves of the SME Finance Monitor (SFM)
• pb3—probability that the firm assesses itself as covering in total a cross-section of 45,018 individual
Bad given that the firm had a previous applica- firm-level observations and the period from 2018
tion rejected. quarter 2 and 2020 quarter 3. The last two waves are
the SFM special Covid-19 survey for the period April
Figure 1 shows that p(NPA) = 0.8592 and to September 2020. These specific survey waves are
p(APA) = 0.1290. unique within the context of the 36 SFM waves that
13
Has previous loan rejection scarred firms from applying for loans during Covid‑19? 1333
preceded them as they incorporated some specific and Q3 2020) and 0 if not; (ii) got loan offer, con-
questions relating to the UK government Covid- ditional on applying for a loan in the first instance,
19 lending schemes (CBILS and BBLS). It is this coded 1 if the firm received an offer from the bank and
additional information that we use to tackle our key 0 if refused; and (iii) government-backed loan offer,
research questions. conditional upon applying for a loan and receiving an
Because of their low natural representation in the offer, coded 1 if firms offer was a government-backed
SME population, larger SMEs are over-sampled in loan and 0 if not (i.e. a conventional bank loan).
the SFM survey using the classic B2B sample struc- Prior to our examination of the Covid-19 loan
ture in order to generate robust sub-samples of these demand and supply, and the use of the UK govern-
bigger SMEs. In particular, the SFM sample includes ment loan guarantee schemes, we need to go further
a pre-determined percentage of SMEs (quota) by back into survey time to the pre-Covid-19 period to
size: 20% zero-employee, 33% micro (1–9 employ- define scarring in the context of previous loan applica-
ees), 32% small (10–49 employees) and 32% medium tion rejections. The key survey questions that we use
(50–249 employees) firms. Additional quotas were for this preliminary classification of scarred and non-
set by sector and region, which were then allocated scarred loan applicants and approved applicants are
within employee size band to ensure that SMEs of all the following:
sizes were interviewed in each sector and region. To [Scarred Borrower Q1]: Thinking more broadly,
further qualify for the SFM interview, SMEs had to has the business EVER had either an application (for
meet three criteria in addition to the quotas by size, a loan) rejected or declined? (response, Yes/No)
sector and region: (i) not 50% + owned by another [Scarred Borrower Q2]: And has this previous
company; (ii) not run as a social enterprise or as a (loan) decline made you more reluctant to apply for
not-for-profit organisation; and (iii) turnover of less another loan? (response, Previous decline has made
than £25 m. Each quarter’s sample matched the previ- more reluctant to apply/*Previous decline has NOT
ous quarter’s results as closely as possible. made more reluctant)
Each of the 5,000 respondent firms interviewed per
wave (by CATI using the quota sampling method) is 3.2 Summary statistics
allocated a Dun & Bradstreet and Experian credit risk
rating and this is added to the data set on a case-by- Figure 1 in Section 2.6 graphically represents the step-
case basis. Sample weights were applied based on the by-step categorisation of our scarred and unscarred
above three strata—size, sector and region—and then rejected borrowers, successful borrowers and firms with
for firms trading for fewer than 2 years (start-ups). no history of making loan applications. We observe that
Weights were initially applied separately to each wave overall 85.9% of firms had no loan applications leaving
and both waves were then combined and grossed to the a potential lending pool of 14.1% of smaller businesses
total of 5,002,010 SMEs, based on UK SME popula- seeking loans. Of these loan applicants, the approval
tion data. This ensured that each individual wave is rate was 91.6% which is in line with other UK evidence
representative of all SMEs whilst the total interviews on banks’ willingness to supply loans to smaller busi-
conducted are weighted to the total of all SMEs. After nesses (Cowling et al., 2020a,b). The fact that the vast
eliminating missing values,1 we have 30,577 valid majority of the small business population do not make
observations across ten SFM waves. loan applications confirms the empirical validity of
The key dependent variables that link to our the pecking order hypothesis (Myers, 1984; Myers &
research questions are (i) sought loan, coded 1 if firm Majluf, 1984) which established those smaller busi-
sought a bank loan during the Covid pandemic (Q2 nesses will also have a stronger preference for financing
their operations and investment activities from internal
1
The missing values are a result of including additional con- funds whenever possible (Michaelas et al., 1999; Revest
trols, and primarily come from the sales data. A third of the & Sapio, 2012).
responses on sales are ‘don’t know’ or ‘refused’, and were Of the subset of the population that had been
subsequently treated as missing values. We confirm the robust- refused a loan prior to the Covid-19 crisis, 72.4%
ness of our results with respect to the missing data, by treat-
ing observations with missing sales data as a separate category reported that they were now reluctant to make fur-
(results available upon request). ther loan applications. We assign this group of
13
1334 M. Cowling et al.
Fig. 2 The demand and
supply of bank loans during
Covid-19 SME population
Thinking about seeking No plans to seek
Sought finance
finance finance
8.3% 17.4%
74.0%
Offered finance Refused finance
91.6% 8.4%
Government-backed
Conventional loan
finance
6.9%
93.1%
potential borrowers as scarred borrowers. Those guarantee schemes. This compares to a maximum of
27.6% of firms who were rejected but reported that 3% of total lending in normal pre-Covid-19 circum-
they remained willing to make further loan applica- stances (Cowling, 2010). Thus, banks are willing to
tions in the future are classified as non-scarred bor- commit loan funds during the crisis due to the size
rowers. Thus, scarred borrowers make up 6.1% of the of the guarantee which covers 80% of CBILS lending
total pool of potential lending pool, unscarred bor- and 100% of BBLS lending. The interesting subset of
rowers 2.3% and successful borrowers the remaining the 7.9% of firms who made a successful application
91.6%. In this sense, scarred borrowers are a small and were offered a conventional (non-government-
but potentially important segment of the small firm backed loan) suggests that the bank sought to lend to
potential borrowing population, although only 0.85% preferred customer firms outside of the scheme where
of the total small business population overall which is it had the ability to set different contractual terms
dominated by firms that do not seek external finance given both schemes had in-built repayment holidays
(85.9% of the total small business population). and a fixed rate of interest.
Figure 2 shows the loan demand and supply dur- Table 1 reports the definitions and summary statis-
ing the Covid-19 crisis when the two government tics of all dependent and independent variables. The
guarantee schemes (CBILS and BBLS) were in place. independent variables capture common business char-
We observe slightly higher loan application rates of acteristics, firm-level risk indicators and firm orien-
19.3% (up from 14.1% pre-Covid-19) and identi- tations. Firm characteristics include sales, employee
cal loan acceptance and refusal rates. An interesting number, legal status, sector, firm age and perfor-
feature was that a much smaller share of the small mance. Performance is measured by firstly a (made a
business population had no plans for borrowing than positive) profit dummy variable, secondly the banded
the permanent non-borrowers identified in the pre- percentage sales growth and thirdly a fast-growth var-
Covid-19 period (74.0% during Covid compared to iable. We use Experian risk classification to measure
85.9% pre-Covid). This suggests that sheer magnitude credit risk from minimum to above average. Firm ori-
of the Covid-19 crisis and its impact on smaller firms entation is proxied by two variables: the owner-man-
had weakened the resolve of many businesses that ager’s aims to grow the business (AIMGROW), and
have explicitly sought to avoid taking on debt. their perception of growth obstacle (FINPROBLEM).
The absolutely striking feature of the debt market We also include a control for firm-bank relationship.
landscape during the Covid-19 period was that 92% of As shown in Table 1, the firm size distribution
all issued loans were offered through the government suggests that zero-employee firms are the dominant
13
Table 1 Variable definition and descriptive statistics (weighted; N = 30,577*)
Variable Definition Full sample (1) SOUGHT = 1 (2) SOUGHT = 0 T test (3) GOT = 1 (4) GOT = 0 T test (5) (6) T test
GOT_ GOT_
GOV = 1 GOV = 0
Mean Std. Dev Mean Mean (1) = (2) Mean Mean (3) = (4) Mean Mean (5) = (6)
Dependent variables
SOUGHT (Q2&3 2020 0.193 0.394
only) = 1 if firm
applied for loan;
0 otherwise
GOT (Q2&3 2020 0.925 0.263
only) = 1 if
firm offered the
finance applied
for; 0 otherwise
GOT_GOVT (Q2&3 2020 0.921 0.269
only) = 1 if firm’s
offer was a gov’t-
backed loan; 0
otherwise
Historical rejec- = 1 if firm his- 0.037 0.190 0.094 0.033 *** 0.089 0.337 *** 0.080 0.237
tion torically rejected
loan; 0 otherwise
Scarred by rejec- = 1 if non-applying 0.724 0.447 0.793 0.711 0.733 0.964 *** 0.752 0.636
tion due to histori-
Has previous loan rejection scarred firms from applying for loans during Covid‑19?
cal rejection; 0
otherwise
Independent variables
Sales Annual sales turno- 0.285 1.262 0.415 0.255 *** 0.414 0.370 *** 0.374 1.009
ver (£ ’Mil) of the
past year
Employment size Dummy vari-
ables: number
of employees
excluding owner
0 Zero-employee 0.764 0.425 0.660 0.783 *** 0.666 0.639 0.666 0.712
1–9 Micro 0.192 0.394 0.271 0.183 *** 0.267 0.313 0.269 0.204
10–49 Small 0.039 0.192 0.063 0.031 *** 0.062 0.042 0.060 0.080
50–99 Small-medium 0.004 0.062 0.005 0.003 *** 0.004 0.003 0.004 0.003
100–199 Medium 0.001 0.036 0.001 0.001 0.001 0.002 0.001 0.001
13
1335
Table 1 (continued)
1336
Variable Definition Full sample (1) SOUGHT = 1 (2) SOUGHT = 0 T test (3) GOT = 1 (4) GOT = 0 T test (5) (6) T test
GOT_ GOT_
13
GOV = 1 GOV = 0
Mean Std. Dev Mean Mean (1) = (2) Mean Mean (3) = (4) Mean Mean (5) = (6)
201–249 Large-medium 0.000 0.018 0.000 0.000
Firm age Dummy variable:
time when firm
was first estab-
lished
< 1 year Start-up 0.049 0.215 0.079 0.066 0.075 0.093 0.080 0.020 **
1–2 years New 0.156 0.363 0.165 0.124 * 0.171 0.076 ** 0.167 0.196
2–5 years Early stage 0.145 0.352 0.193 0.133 ** 0.195 0.209 0.211 0.010 ***
6–9 years Established 0.133 0.340 0.126 0.132 0.131 0.187 0.133 0.106
10–15 years Well established 0.156 0.363 0.145 0.182 * 0.139 0.172 0.137 0.122
> 15 years Old 0.361 0.480 0.292 0.363 *** 0.290 0.263 0.273 0.546 **
Legal Dummy variables:
legal form of the
firm
Sole proprietor 0.614 0.487 0.436 0.611 *** 0.405 0.552 0.414 0.245
Partnership 0.035 0.185 0.032 0.034 0.032 0.039 0.033 0.018
LLP 0.020 0.139 0.004 0.005 0.003 0.003
LTD Co 0.331 0.471 0.529 0.351 *** 0.560 0.408 0.550 0.738 *
Industry Dummy variables:
industry sector
Agriculture 0.030 0.170 0.018 0.034 * 0.018 0.019 0.019 0.004 ***
Manufacturing 0.055 0.228 0.039 0.059 ** 0.033 0.066 0.028 0.059
Construction 0.189 0.392 0.152 0.197 ** 0.146 0.199 0.157 0.019 ***
Wholesale/retail 0.100 0.300 0.133 0.094 ** 0.146 0.057 ** 0.150 0.124
Hotels and cater- 0.034 0.182 0.047 0.033 ** 0.051 0.035 0.050 0.066
ing
Transport and 0.125 0.330 0.171 0.111 ** 0.158 0.370 * 0.156 0.175
Comms
Real estate 0.265 0.441 0.275 0.264 0.281 0.199 0.277 0.373
Health 0.074 0.262 0.051 0.078 ** 0.055 0.009 *** 0.059 0.013 ***
Other services 0.128 0.334 0.113 0.130 0.111 0.046 ** 0.103 0.166
M. Cowling et al.
Table 1 (continued)
Variable Definition Full sample (1) SOUGHT = 1 (2) SOUGHT = 0 T test (3) GOT = 1 (4) GOT = 0 T test (5) (6) T test
GOT_ GOT_
GOV = 1 GOV = 0
Mean Std. Dev Mean Mean (1) = (2) Mean Mean (3) = (4) Mean Mean (5) = (6)
Region Dummy variable:
geographic region
Scotland 0.058 0.235 0.072 0.054 0.066 0.070 0.059 0.183 ***
North East 0.029 0.168 0.030 0.031 0.020 0.008 0.017 0.061
York and Humber 0.071 0.256 0.088 0.069 0.090 0.126 0.095 0.046
North West 0.098 0.298 0.106 0.097 0.102 0.116 0.106 0.041 *
West Midlands 0.071 0.257 0.091 0.068 0.096 0.112 0.095 0.084
East Midlands 0.070 0.255 0.083 0.069 0.094 0.065 0.102 0.002 ***
East England 0.099 0.299 0.067 0.106 ** 0.059 0.180 0.056 0.099
Wales 0.041 0.198 0.045 0.042 0.043 0.045 0.038 0.046
South West 0.100 0.300 0.085 0.109 0.086 0.025 *** 0.090 0.040
London 0.179 0.384 0.178 0.173 0.190 0.145 0.184 0.296
South East 0.163 0.369 0.134 0.162 0.139 0.064 ** 0.144 0.060 **
N. Ireland 0.020 0.141 0.020 0.020 0.017 0.044 0.015 0.042
Risk rating Dummy variables:
Experian risk
rating
Minimal 0.058 0.234 0.058 0.053 0.064 0.070 0.049 0.263
Has previous loan rejection scarred firms from applying for loans during Covid‑19?
Low 0.136 0.343 0.144 0.120 0.133 0.125 0.137 0.088
Average 0.278 0.448 0.233 0.279 * 0.241 0.222 0.245 0.205
Above average 0.398 0.489 0.454 0.427 0.455 0.435 0.455 0.444
Not known 0.130 0.336 0.111 0.121 0.106 0.149 0.114
PROFIT = 1 if firm made 0.742 0.438 0.702 0.717 0.706 0.655 0.724 0.403 **
a surplus over
the past year; 0
otherwise
FASTGROW = 1 sales 0.034 0.181 0.073 0.026 *** 0.072 0.023 0.077 0.005
growth ≥ 20% by
two consecutive
years; 0 otherwise
13
1337
Table 1 (continued)
1338
Variable Definition Full sample (1) SOUGHT = 1 (2) SOUGHT = 0 T test (3) GOT = 1 (4) GOT = 0 T test (5) (6) T test
GOT_ GOT_
13
GOV = 1 GOV = 0
Mean Std. Dev Mean Mean (1) = (2) Mean Mean (3) = (4) Mean Mean (5) = (6)
AIMGROW = 1 if firm aimed 0.464 0.499 0.420 0.293 *** 0.417 0.371 0.423 0.332
to grow the busi-
ness; 0 otherwise
FINPROBLEM = 1 if firm saw 0.143 0.350 0.248 0.138 *** 0.220 0.300 0.228 0.149
cash flow and/or
external finance
as main growth
obstacle; 0 oth-
erwise
Sales growth Growth of sales
over the past
12 months
> 40% 0.017 0.131 0.019 0.016 0.022 0.002 0.023 0.009
20 to 40% 0.070 0.256 0.089 0.055 * 0.092 0.039 ** 0.097 0.014 ***
0 to 20% 0.192 0.394 0.148 0.152 0.138 0.149 0.134 0.221
No growth 0.548 0.498 0.405 0.505 *** 0.400 0.311 0.389 0.487
< 0% 0.172 0.378 0.338 0.272 ** 0.349 0.499 0.357 0.269
Banking relation- Strength of rela-
ship tionship with
main bank
Strong 0.216 0.411 0.243 0.189 ** 0.260 0.118 * 0.261 0.210
Fair 0.654 0.476 0.572 0.676 *** 0.571 0.738 ** 0.572 0.567
None 0.130 0.336 0.186 0.135 ** 0.169 0.144 0.167 0.223
Except for SOUGHT (N = 5,052 and only observed in Q2 and Q3 2020 waves), GOT (N = 1,026 and only observed if SOUGHT = 1), and GOT_GOVT (N = 867 and only observed
if SOUGHT = 1)
M. Cowling et al.
Has previous loan rejection scarred firms from applying for loans during Covid‑19? 1339
size class of firm, and of employer firms, micro busi- 3.3 Methodology
nesses account for the vast majority. The median firm
would be 10–15 years old with nearly four in ten By definition, scarring only happens after previous
very old, but there are significant minorities of new loan rejection, and loan approval is only observed
and early-stage businesses. At the industry level, real for loan applicants. If we are to examine the effect
estate, renting and business services firms represent of scarring on loan application and application out-
the single largest group of businesses, although there come, the selection mechanism can be more than one
are also large shares in construction. On credit risk, layer. In particular, when sample selection is based
we find that very few smaller firms would be con- on historical rejection only, in the case of (a) scarring
sidered minimal risk, and quite small shares as low from previous rejection and (b) the effect of scarring
risk. In contrast, 4 in 10 smaller businesses would be on Covid-19 loan application, we apply the Heckman
classed as above average risk and a non-trivial share (1979) two-step procedure through the inclusion of
has no formal risk classification per se. This group an inverse Mills ratio (IMR) from the loan rejection
of firms would contain those who have never had a equation (step 1) in the outcome equation (step 2). In
borrowing event or any incidence of financial delin- the case of the effect of scarring on loan acceptance,
quency (for example, late payment of a tax bill), and sample selection arises first from previous rejection
also those that conduct their businesses through their and then current loan application. Similar is the offer
personal banking accounts or in cash. In terms of of Covid-19 government-guaranteed loans, where
performance, although nearly three quarters of SMEs there is one additional layer of selection mecha-
had made a profit in the previous year, the majority nism—the acceptance of Covid-19 loan application.
(73%) reported zero or negative sales growth and In the last two cases, we estimate the selection pro-
only a minimal proportion (3.4%) sustained high cess through a multi-variate probit model in a sequen-
growth. tial equation system, assuming correlated and jointly
We then conduct three sets of univariate mean- normal distributions of errors between each individ-
comparison tests, for finance applicants vs. non- ual selection equations. We explicitly account for the
applicants, approval vs. rejection and recipients of process of selection through the sequence of equa-
government- vs. non-government-backed loans, tions by calculating the IMR for each equation, which
respectively. It is shown that SMEs experienc- enters the next step, outcome equation (acceptance of
ing previous rejection were more likely to apply loan or government-guaranteed loan, respectively) as
for loans during the pandemic. Within-Covid loan explanatory variable. So that the model is correctly
rejection in general was higher for both historically specified, each selection equation should include a
rejected and scarred borrowers, but such difference vector of variables (exclusion restrictions) that is not
disappeared for government-backed loans. It is also included in the outcome equation, which will be dis-
found that non-applicants on average were smaller cussed in detail with individual specifications.
(both in sales and employment), grew more slowly
and had higher credit risk, which lend support to the
discouraged borrower literature that lower expected 4 Results
(or perceived) approval probability would deter
firms from submitting loan applications. Further- Here, we present the multi-variate regression results
more, SMEs with higher growth orientations were from our models relating to (a) previous incidences of
more likely to seek finance. There is some, but lim- loan rejection, (b) scarring effects of previous rejec-
ited, evidence that government-backed finance was tion, (c) current loan demand, (d) current loan supply
targeting at most deprived businesses (firms less and (e) whether loan offers used the UK government
than 5 years old) and regions (such as North West Covid-19 guarantee schemes. The models we use are
and Midlands). We also report the summary statis- pooled probit models allowing for the potential pres-
tics by region and industry in the online Appendix ence of sample selection effects as discussed in the
(Table A1) to highlight potential location- and sec- previous section. All the regressions are population
tor-driven difference in financing activities.
13
1340 M. Cowling et al.
weighted and Table 22 reports the coefficient esti- of historical rejection is increased by 5.6 percentage
mates as well as the marginal effects to show their points, or more than doubled (mean rejection = 3.7%
economic significance. from Table 1), for firms with a weak banking relation-
ship. This suggests that building a strong relationship
4.1 Previous loan rejection with a main bank creates the conditions for a mutu-
ally beneficial relationship where softer information
We initially found that only a very small subset of can be shared.
small firms, around 1.18% of the total business pop-
ulation and 8.4% of the borrower population, had a 4.2 Scarring from previous loan rejection
previous negative loan application outcome. But of
the rejected borrower population, 72.4% were scarred Here, we distinguish (Model 2) between firms that
and had become reluctant to make future loan appli- were rejected for loans in the pre-Covid-19 period
cations. Given that the general loan refusal rate in sta- and then became scarred in terms of reporting a
ble economic times is around 10% (Krasniqi, 2010; reluctance to make any future loan applications
Scott, 2006) and in this sample 8.4%, this implies which we term Scarred Borrowers. We use Model
that a substantial share of the rejected set of firms 1 as the selection equation to calculate the IMR.
had received more than one loan rejection in the past. Two measures of historical performance, sales
This strengthens the potential effect of borrower scar- growth and FASTGROW, are used as the exclusion
ring, which is central to our study. restrictions. Here, we contend that the intention of
The results (Model 1) show that micro businesses finance seeking is more associated with the firm’s
(1–9 employees) have a higher probability of hav- subjective perception of credit supply, but lend-
ing a historical loan rejection, by 10%, compared to ers would use more objective measures of perfor-
all other size classes of firm. This is consistent with mance and riskiness to make lending decisions.
information-based problems for the smallest of firms. The first point of note is that there is no evi-
In contrast, real estate businesses have a lower rejec- dence of a selection effect, since the IMR is sta-
tion probability than firms in all other industry sec- tistically insignificant. From the results, it is
tors. Risk rating plays a key role in determining loan very clear that the probability of scarring dimin-
rejection. Compared to firms with minimum risk, all ishes increasingly in firm size class. Thus, once
other firms have a significantly lower probability of a micro business has received a loan application
having their loan application accepted, by magnitudes rejection, it is much more likely never to make an
between 12 and 20%. Rejection probability is highest application again in the future. Scarring is also
amongst firms with above average risk. This is con- higher amongst the oldest age class of businesses
sistent with previous empirical work on loan rejec- (> 15 years since start) which suggests that start-
tion (Sannajust, 2014) and with standard bank lend- up and early-stage firms are less prone to scarring
ing processes which place credit risk at the forefront through loan rejection.
of their decision-making process’s (Berg, 2018). We There is evidence that the construction sector had
also note that firms with low or no sales growth were a higher probability of scarring although the reasons
less likely to experience historical rejection, implying for this industry effect are unclear. However, there
that banks care more about the volatility of growth, was tremendous variation in the probability of scar-
as a proxy for business risk, than the performance of ring across the regions of the UK with each single
a single year when making lending decisions. Lastly, region having a significant and different coefficient.
a good relationship with the main bank significantly On this, firms located in the South East of England
reduces the odds of loan rejection. Compared to (the most prosperous region of the UK) had the single
those with strong banking relationship, the chance highest probability of being scarred by a loan rejec-
tion. This may relate to expectations as firms in poor
regions might expect rejection and cope with it bet-
2
To save space, Table 2 only reports the coefficient estimates ter. In general, there is no clear pattern across UK
of the outcome equations when sample selection is considered.
The full regression results including the selection equations are regions based on relative economic prosperity which
reported in the Online Appendix Table A2. reinforces our general argument that behaviours and
13
Table 2 Regression results (population weighted)
Variables (1) Historical loan rejection (2) Scarred by rejection (3) Covid-19 loan application (4) Covid-19 loan acceptance (5) Covid-19 government
(probit) (probit with sample selection (probit with sample selection (GOT) (probit with sample guarantee (GOT_GOVT)
correction) correction) selection correction) (probit with sample
selection correction)
Coeff dy/dx Coeff dy/dx Coeff dy/dx Coeff dy/dx Coeff dy/dx
Scarred by rejection 0.667*** 0.159 − 1.317*** − 0.143 − 0.522 − 0.048
(0.189) (0.249) (0.454)
Survey wave (base: Q2 2018)
Q3 2018 0.023 0.002 0.177 0.054
(0.104) (0.310)
Q4 2018 0.088 0.008 0.274 0.081
(0.111) (0.303)
Q1 2019 − 0.160 − 0.012 0.425 0.121
(0.115) (0.333)
Q2 2019 − 0.035 − 0.003 0.110 0.034
(0.106) (0.293)
Q3 2019 − 0.059 − 0.005 − 0.122 − 0.039
(0.113) (0.324)
Q4 2019 − 0.169 − 0.012 0.626* 0.168
(0.109) (0.331)
Q1 2020 − 0.086 − 0.007 0.293 0.086
Has previous loan rejection scarred firms from applying for loans during Covid‑19?
(0.110) (0.294)
Q2 2020 0.052 0.005 0.467 0.131
(0.107) (0.292)
Q3 2020 − 0.014 − 0.001 0.181 0.055 0.257*** 0.060 0.357 0.044 0.293 0.028
(0.108) (0.309) (0.080) (0.231) (0.303)
ln(sales) 0.010 0.001 0.019 0.005 0.254*** 0.060 0.409*** 0.044 − 0.353 − 0.032
(0.022) (0.076) (0.037) (0.124) (0.276)
Employment size (base: 1 employee)
1–9 0.117** 0.010 − 0.470*** − 0.139 − 0.148 − 0.034 − 0.538** − 0.065 0.793* 0.072
(0.057) (0.163) (0.097) (0.249) (0.469)
10–49 − 0.000 0.000 − 0.638** − 0.194 − 0.217 − 0.049 − 1.145*** − 0.187 1.049 0.085
(0.087) (0.249) (0.144) (0.335) (0.737)
13 50–99 − 0.182 − 0.012 − 1.425*** − 0.453 − 0.352* − 0.074 − 1.824*** − 0.375 1.956* 0.110
(0.123) (0.392) (0.198) (0.494) (1.058) 1341
Table 2 (continued)
1342
Variables (1) Historical loan rejection (2) Scarred by rejection (3) Covid-19 loan application (4) Covid-19 loan acceptance (5) Covid-19 government
(probit) (probit with sample selection (probit with sample selection (GOT) (probit with sample guarantee (GOT_GOVT)
13
correction) correction) selection correction) (probit with sample
selection correction)
Coeff dy/dx Coeff dy/dx Coeff dy/dx Coeff dy/dx Coeff dy/dx
100–199 − 0.213 − 0.014 − 1.069** − 0.340 − 0.826*** − 0.141 − 3.400*** − 0.795 3.125 0.116
(0.156) (0.490) (0.263) (0.769) (1.925)
200–249 − 0.217 − 0.014 − 1.522 − 0.482 − 1.110*** − 0.165
(0.342) (1.010) (0.402)
Firm age (base: < 1 year)
1–2 years − 0.108 − 0.009 0.484 0.154 0.173 0.048 0.535 0.055 − 1.409** − 0.081
(0.139) (0.453) (0.173) (0.459) (0.692)
2–5 years − 0.027 − 0.003 0.558 0.175 − 0.062 − 0.016 0.423 0.046 − 0.137 − 0.003
(0.141) (0.438) (0.174) (0.524) (0.735)
6–9 years − 0.244* − 0.019 0.688 0.210 − 0.093 − 0.024 − 0.183 − 0.028 − 0.888 − 0.034
(0.144) (0.472) (0.181) (0.424) (0.658)
10–15 years − 0.110 − 0.010 0.335 0.109 − 0.243 − 0.058 0.037 0.005 − 1.516** − 0.094
(0.138) (0.428) (0.173) (0.466) (0.686)
> 15 years − 0.100 − 0.009 0.660* 0.203 − 0.316* − 0.074 0.260 0.031 − 1.667** − 0.115
(0.131) (0.400) (0.162) (0.447) (0.649)
Legal status (base: sole proprietorship)
Partnership − 0.042 − 0.003 − 0.212 − 0.064 0.122 0.029 − 0.114 − 0.017 0.507 0.025
(0.077) (0.233) (0.170) (0.535) (0.443)
LLP − 0.259** − 0.017 0.123 0.034 − 0.002 0.000
(0.106) (0.414) (0.283)
LTD Co − 0.021 − 0.002 0.116 0.032 0.125 0.030 0.441* 0.049 − 0.619* − 0.057
(0.066) (0.187) (0.098) (0.236) (0.363)
Industry (base: agriculture)
Manufacturing − 0.112 − 0.010 0.301 0.090 0.046 0.008 − 0.489 − 0.073 − 0.949* − 0.079
(0.108) (0.312) (0.183) (0.460) (0.576)
Construction − 0.138 − 0.013 0.527* 0.149 0.336** 0.069 − 0.031 − 0.004 0.305 0.012
(0.099) (0.316) (0.164) (0.449) (0.505)
Wholesale/retail − 0.044 − 0.004 0.014 0.004 0.203 0.039 0.482 0.041 − 1.171** − 0.110
M. Cowling et al.
Table 2 (continued)
Variables (1) Historical loan rejection (2) Scarred by rejection (3) Covid-19 loan application (4) Covid-19 loan acceptance (5) Covid-19 government
(probit) (probit with sample selection (probit with sample selection (GOT) (probit with sample guarantee (GOT_GOVT)
correction) correction) selection correction) (probit with sample
selection correction)
Coeff dy/dx Coeff dy/dx Coeff dy/dx Coeff dy/dx Coeff dy/dx
(0.103) (0.309) (0.159) (0.492) (0.503)
Hotel and catering − 0.093 − 0.009 0.409 0.119 0.340** 0.070 0.693 0.052 − 1.662*** − 0.201
(0.101) (0.315) (0.163) (0.469) (0.629)
Trans and Comms − 0.187* − 0.016 0.331 0.098 0.672*** 0.158 − 0.527 − 0.081 − 0.559 − 0.037
(0.111) (0.339) (0.177) (0.517) (0.535)
Business service − 0.222** − 0.019 0.221 0.067 0.425*** 0.091 0.050 0.006 − 0.440 − 0.027
(0.096) (0.319) (0.161) (0.470) (0.473)
Health − 0.326** − 0.025 0.605 0.167 0.439** 0.094 0.724 0.053 − 0.435 − 0.027
(0.127) (0.477) (0.216) (0.603) (0.683)
Other services − 0.028 − 0.003 0.139 0.043 0.144 0.027 0.498 0.042 − 0.872* − 0.069
(0.106) (0.310) (0.170) (0.436) (0.512)
Region (base: Scotland)
North East − 0.031 − 0.003 0.870** 0.276 − 0.094 − 0.024 0.805 0.049 − 0.787 − 0.195
(0.134) (0.381) (0.215) (0.527) (0.691)
York and Humber − 0.024 − 0.002 0.543 0.182 − 0.072 − 0.019 − 0.180 − 0.020 1.460** 0.186
(0.114) (0.340) (0.174) (0.471) (0.597)
Has previous loan rejection scarred firms from applying for loans during Covid‑19?
North West − 0.077 − 0.006 0.773** 0.250 − 0.052 − 0.014 − 0.305 − 0.037 1.494*** 0.188
(0.115) (0.346) (0.184) (0.557) (0.498)
West Midlands − 0.069 − 0.006 1.030*** 0.316 0.032 0.009 − 0.434 − 0.056 1.317*** 0.177
(0.113) (0.375) (0.177) (0.475) (0.485)
East Midlands − 0.039 − 0.003 0.799** 0.257 − 0.093 − 0.024 0.468 0.035 2.418*** 0.216
(0.116) (0.374) (0.181) (0.444) (0.605)
East England − 0.112 − 0.009 0.710** 0.232 − 0.312* − 0.074 − 0.844* − 0.136 1.059 0.157
(0.115) (0.334) (0.185) (0.487) (0.696)
Wales − 0.040 − 0.003 0.486 0.164 − 0.085 − 0.022 − 0.211 − 0.024 0.957* 0.147
(0.117) (0.364) (0.187) (0.522) (0.529)
South West 0.047 0.004 0.706** 0.231 − 0.345* − 0.081 0.766 0.048 0.510 0.092
(0.110) (0.322) (0.188) (0.491) (0.487)
13
1343
Table 2 (continued)
1344
Variables (1) Historical loan rejection (2) Scarred by rejection (3) Covid-19 loan application (4) Covid-19 loan acceptance (5) Covid-19 government
(probit) (probit with sample selection (probit with sample selection (GOT) (probit with sample guarantee (GOT_GOVT)
13
correction) correction) selection correction) (probit with sample
selection correction)
Coeff dy/dx Coeff dy/dx Coeff dy/dx Coeff dy/dx Coeff dy/dx
London − 0.085 − 0.007 0.631* 0.209 − 0.178 − 0.044 0.002 0.000 1.107*** 0.161
(0.107) (0.324) (0.165) (0.477) (0.390)
South East − 0.031 − 0.003 1.083*** 0.329 − 0.203 − 0.050 0.127 0.012 0.772* 0.127
(0.105) (0.299) (0.169) (0.432) (0.425)
N. Ireland − 0.154 − 0.012 0.411 0.140 − 0.160 − 0.040 − 0.841* − 0.135 0.659 0.113
(0.142) (0.468) (0.203) (0.505) (0.653)
PROFIT − 0.089 − 0.007 − 0.134 − 0.038 0.054 0.013 − 0.263 − 0.029 1.143*** 0.104
(0.056) (0.156) (0.091) (0.224) (0.257)
Risk rating (base: minimum)
Low 0.189* 0.012 − 0.056 − 0.016 0.382 0.059 0.680* 0.054
(0.097) (0.312) (0.369) (0.372)
Average 0.217** 0.014 0.046 0.013 0.502 0.074 0.282 0.027
(0.093) (0.333) (0.383) (0.443)
Above average 0.293*** 0.020 − 0.085 − 0.024 0.741* 0.097 − 0.115 − 0.013
(0.096) (0.350) (0.443) (0.504)
Not known 0.226** 0.015 0.076 0.021 0.172 0.029
(0.108) (0.358) (0.482)
Sales growth (base: > 40%)
20 to 40% − 0.191 − 0.022
(0.161)
0 to 20% − 0.277* − 0.030
(0.156)
No growth − 0.426*** − 0.041
(0.155)
< 0% − 0.175 − 0.020
(0.157)
FASTGROW 0.078 0.006
(0.132)
M. Cowling et al.
Table 2 (continued)
Variables (1) Historical loan rejection (2) Scarred by rejection (3) Covid-19 loan application (4) Covid-19 loan acceptance (5) Covid-19 government
(probit) (probit with sample selection (probit with sample selection (GOT) (probit with sample guarantee (GOT_GOVT)
correction) correction) selection correction) (probit with sample
selection correction)
Coeff dy/dx Coeff dy/dx Coeff dy/dx Coeff dy/dx Coeff dy/dx
Banking relationship (base: strong)
Fair 0.158** 0.010 0.436** 0.148
(0.065) (0.205)
Weak 0.595*** 0.056 1.111*** 0.319
(0.077) (0.224)
AIMGROW 0.170** 0.041
(0.081)
FINPROBLEM 0.415*** 0.099
(0.098)
Inverse Mills ratios
Historical rejection − 0.065 − 0.018 − 1.244*** − 0.296 1.007 0.109 − 1.607 − 0.147
(0.695) (0.368) (0.944) (1.035)
SOUGHT 0.998* 0.109 − 0.291 − 0.026
(0.536) (0.723)
GOT − 1.458 − 0.133
(1.148)
Has previous loan rejection scarred firms from applying for loans during Covid‑19?
Constant − 1.371*** − 1.337 0.239 − 4.326* 7.945**
(0.261) (1.703) (0.758) (2.300) (3.556)
N Obs 30,577 1077 30,577 30,577 30,577
Censored N 5,052 1026 867
Wald χ2 137.18*** 158.86*** 260.90*** 125.54*** 114.59***
Pseudo R2 0.026 0.147 0.120 0.250 0.364
Log likelihood − 555.75 − 67.16 − 253.54 − 18.45 − 12.27
Models 1 and 2 estimate historical rejection and scarring prior to Covid-19. Models 3 to 5 estimate loan demand, loan supply and use of government-guaranteed loan schemes
during Covid-19. Full regression results including the selection equations are reported in the Online Appendix. * p < 0.10; ** p < 0.05; *** p < 0.01. Asymptotic robust standard
errors reported. Weights applied
13
1345
1346 M. Cowling et al.
changing behaviours are a combination of many fac- precautionary savings gap amongst these size classes
tors including a spatial one. of business which meant that they were in the most
precarious liquidity position entering the Covid-19
4.3 Current Covid‑19 loan demand crisis (Cowling, Brown and Rocha, 2002; Brown
& Cowling, 2021). The evidence suggests that this
Model 3 estimates the probability that a firm weakened cash flow position led to an increase in the
demanded a loan during the first two quarters of the demand for loans during the Covid-19 crisis. There
Covid-19 crisis (2020 q2 and q3) and whether such is also a clear sector effect on within-pandemic loan
demand differs for SMEs scarred by previous rejec- demand. Businesses experiencing more severe dis-
tion. We use the same selection adjustment as in ruptions because of the lockdown, such as construc-
Model 2, and the significantly negative coefficient tion, hotel and catering and transportation and com-
estimate of the IMR justifies the treatment of sample munications, exhibited higher demand caused by the
selection. This is also consistent with the previous significant loss of revenue during the pandemic. The
empirical evidence that historical rejection would dis- higher application rate for the health industry is likely
courage loan application. to be a result of inflated demand for health-related
The key finding is that scarring due to historical products and services in the Covid-19 crisis.
rejection is positively associated with loan applica-
tion during the Covid-19 crisis. Scarred borrow- 4.4 Loan supply and use of government‑guaranteed
ers were 16% more likely to apply, an 80% increase schemes
compared to the average loan application probabil-
ity of 19%. In this sense, the severity of the crisis in Model 4 reports the coefficient estimates of the loan
terms of its impact on firms through reduced trading acceptance equation. As discussed previously, the
and the associated income losses and cash flow pres- selection process includes two equations, histori-
sures have obviously changed the behaviours of firms cal rejection which is the same as before and loan
that were previously scarred by rejection. Financing application. For the second selection equation,
constraints have become more severe for scarred bor- entrepreneurial growth objectives (AIMGROW) and
rowers, and cash flow pressure accumulated through the perception of cash flow problems as the main
historical non-borrowing means they can no longer growth obstacle (FINPROBLEM) are used as exclu-
rely on internal finance to fund the business. With sion restrictions. Growth objectives are commonly
liquidity problems exacerbated as businesses faced used as the exclusion restriction in previous credit
restricted trading conditions and lower revenues in rationing studies (c.f. Cowling et al., 2016), as they
the formal lockdown further during the pandemic, are obviously unobservable by lenders but found
the financing shortfall soon extended to the whole to be significant in explaining the finance-seeking
small business sector: the demand for bank loans behaviour by individual businesses (Michaelas et al.,
increased by 6% during Q3 2020 and thus H1 is sup- 1999; Psillaki and Daskalakis, 2009). The use of
ported. If this key behaviour was sustained into the the second exclusion restriction is justified by the
post-Covid-19 period, then we would expect to see relationship between cash constraints and external
an upward shift in the general demand for bank loans finance need according to the pecking order theory
which had reached a historical low in the period from (Myers & Majluf, 1984).
the Brexit vote in June 2016 through to the onset of Both exclusion restrictions are positively and
Covid-19 in March 2020. Therefore, we find strong significantly associated with loan application
support for H3. (Table A2), but the coefficient estimates of the two
The notable feature of these results was an IMRs show little evidence of sample selection bias.
inverted ‘U’-shaped pattern in loan demand as firm The marginally significant and positive IMR for
size increases. Medium- (100–199 employees) and the loan application equation suggests that finance
large–medium (200–249 employees) firms are on demand was generally based on an unbiased estima-
average 15% less likely to seek external loans than tion of the true loan approval probability. Our head-
their smaller counterparts. This is consistent with line finding is that applicants scarred by previous
earlier UK empirical evidence which identified a rejection were significantly less likely to be approved,
13
Has previous loan rejection scarred firms from applying for loans during Covid‑19? 1347
by a margin of 14%. This translates to 7,300 more (1–9 employees) and start-up (< 1 year) businesses
rejections3 out of 6 million SME population com- had a higher, whilst limited liability companies had
pared to businesses not scarred by previous rejection. a lower likelihood to receive government-guaran-
This suggests that whilst scarred SMEs were forced teed loans. There was also evidence of businesses
to seek external finance during the pandemic due to located in more deprived regions (York and Hum-
more severe financial constraints, they appeared to be berside, North West, West and East Midlands) hav-
more disadvantaged as lenders treated their previous ing a higher probability of receiving a government-
rejection as a negative signal of borrower quality. guaranteed loan offer. Surprisingly, sectors worst hit
There is no evidence of changes in conventional by the shutdown of businesses during the pandemic,
loan supply during the pandemic, and thus H2 is namely wholesale/retail and hotel/catering, were the
not supported. We have noted that the overwhelm- least likely to receive government support. This may
ing majority of business lending offers in the Covid- relate to the wage protection schemes available which
19 crisis were backed by the government guaran- covered 80% of pay for workers in sectors forced to
tee schemes. In this sense, despite general model lock down and the relatively low capital requirements
significance, it was difficult to distinguish between in service sectors. Regarding credit risk, we find that
firms who got a loan offer and those who got refused the probability of receiving a government-guaranteed
by section, region or credit risk. Banks are shown to loan offer was higher for low-risk firms and then
have relied on a few common indicators, primarily decreasing for average- and above average-risk firms.
sales and business size, when processing loan appli- This implies that the government only had a moderate
cations. Firms with higher sales were more likely to risk tolerance, and the recoverability of public funds
be approved, implying that application decisions were was a key criterion for government supporting initia-
generally based on an unbiased evaluation of the true tives. It also suggests that banks still conducted due
approval probability. On the other hand, loan rejec- diligence on loan applications.
tion increased monotonically with the number of Given the low rejection rate on loan application
employees. This suggests that during the pandemic, and the low percentage of commercial loans approved
banks have imposed stronger restrictions in lending (Fig. 2), the legitimacy of conventional binary (pro-
to larger firms which normally have higher capital bit or logit) regressions may be compromised by the
requirements. ‘rare-event’ problem first noted by King and Zeng
Model 5 examines government-guaranteed lend- (2001). To test the robustness of our results against
ing. One additional mechanism enters the selection potential small sample bias, we re-estimate Models 4
process because loan type is only observed if a loan and 5 in Table 2 using the penalised maximum likeli-
application is approved. We use firm-bank relation- hood estimation proposed by Firth (1993), as statisti-
ship to identify the selection equation, drawn from cal simulations have found this method of bias cor-
the extant relationship lending literature on commer- rection to be more effective with large sample size
cial bank lending (Berger & Udell, 1995 and 2002). (Leitgöb, 2013). As reported in Table A3, our head-
Similarly, although the exclusion restrictions seem to line findings regarding the effect of borrower scarring
be significant (Table A2), all three IMRs are insignifi- on general and government-backed loan approval and
cant, showing no evidence of selection bias. the distribution of government lending with respect to
We find that the offer of government guarantee various business characteristics still hold.
was not discriminating against scarred borrowers, and
there is no sign of more intense government support
the deeper into the crisis; thus, H4 is also not sup- 5 Conclusion
ported. However, there was a clearer differentiation
with respect to the targets of public support. Micro We set out to establish how important a previous
loan rejection and any scarring that might arise from
a rejection was in informing the small firms’ deci-
sion on whether to make a new loan application dur-
3
Fourteen percent (Table 2) of 0.85% (Fig. 1) SME population. ing the Covid-19 crisis. We then traced out the pro-
cess from application for a loan to the banks’ decision
13
1348 M. Cowling et al.
and potential offer of a government-backed Covid- In terms of what these findings might mean for a
19-guaranteed lending scheme. We find that over the future post-Covid-19 world, we tentatively suggest
last 10 years, the loan rejection rate was fairly constant that the small, but non-trivial, group of scarred bor-
even though there was variation in loan demand. But rowers may have revised their expectations about
for those businesses that had a rejection, there was banks and lending upward. If this was the case, then
evidence of scarring in the sense that they changed we would see a structural shift in the potential bor-
their expectations about the probability that future rower pool with a greater share of the business pop-
loan applications would be accepted to such an extent ulation willing to make bank loan applications. It
that they became reluctant to even apply. This subset remains the case that when the government guaran-
of the business population was estimated to be 6.51% tee schemes cease to exist, if banks raise their lend-
of the potential borrower pool and 77.5% of rejected ing standards and begin to choke off lending, then
borrowers. They were found to be highly concentrated we could potentially be back to square one. In paral-
amongst micro and small businesses and businesses lel with this de-scarring, we also observed a decline
that had gone beyond their formative stages. This sug- in the share of what have been termed permanent
gests that they were shocked by a rejection given their non-borrowers. Again, if this group of businesses
relatively advanced stage in their life cycle. At this sustained their new found willingness to apply for
stage, their expectations based on track record would loans, then the pool of borrowers would expand sig-
have been high regarding a positive outcome. nificantly. However, given the strong evidence of a
We then explored their borrowing behaviours financial pecking order where internal funds are pre-
during the Covid-19 crisis when the vast majority ferred over all external debt (and equity particularly),
of firms faced reduced trading activity, significant any expansion in the economy which puts more cash
income reductions and problems with cash `flow. into the small business sector would see them return
Prior research found that micro and small businesses to their self-financing status quo.
were disproportionately impacted on all three meas- In terms of our contribution, we think we have
ures and we established that it was precisely these size added a new and important dimension to the wider
classes of firm that had the highest loan demand in the debates and research around credit rationing by
first two quarters of the Covid-19 crisis which began broadening the focus of the issue to consider how pre-
in March 2020 in the UK. Importantly, we found no vious credit rationing then shapes future behaviours
evidence that prior rejection OR scarring from prior and effectively increases pool of small firms operat-
rejection impacted on the desire to make a Covid-19 ing on the first step of the pecking order ladder which
period loan application. In this sense, the scarring is financing all operations and investment from inter-
effect which influenced borrowing behaviours dissi- nal resources. We think that previous literature has
pated when firms were faced with the sheer magnitude identified borrower discouragement as an important
of the economic shock wrought by Covid-19. phenomenon, but the concept of the scarred borrower
On the supply side of the loan market, banks were adds nuance and additional insights. Indeed, some
more than willing to make loan offers to firms that previously rejected loan applicants become scarred
had experienced historical loan rejections suggest- and some do not which is interesting in its own right.
ing that they treat each funding proposal, even from Furthermore, credit rationing will always be more
the same firm, on its merits. However, the presence strongly associated with smaller (and younger) firms
of the two most generous government-backed guar- due to their information opacity.
anteed loan schemes (CBILS and BBLS) may have The simple UK Covid-19 loan guarantee take-up
given banks the leeway to drop their lending stand- figures show that 971,302 small firms accessed loans
ards when in a normal crisis they would have raised on the Bounce Back Scheme compared to 76,704
them and choked off lending. In terms of routing firms for the mid-range scheme (CBILS), and 569
loan applications through normal lending channels or large firms on the big firm scheme (CLBILS) rein-
government-guaranteed loan schemes, again, we find force this point. In the wider context of firm size,
no evidence that formerly rejected or scarred bor- small firms when seeking external finance are heavily
rowers were routed onto government-backed lending reliant on debt. Medium and large firms increasingly
schemes. have much greater access to the full range of debt and
13
Has previous loan rejection scarred firms from applying for loans during Covid‑19? 1349
equity sources which means they are less depend- Bank of England. (1993). Bank lending to smaller busi-
ent upon a single source. They also have far higher nesses. London: Bank of England.
Berg, T. (2018). Got rejected? Real effects of not getting a loan.
approval rates for all sources of finance. So, although The Review of Financial Studies, 31(12), 4912–4957.
the UK does have a well-developed loan market, it Berger, A. N., & Udell, G. F. (1995). Relationship lending and
has always been the case that its transactional lending lines of credit in small firm finance. Journal of Business,
and banking structure means that the ability of a firm 68(3), 351–381.
Berger, A. N., & Udell, G. F. (2002). Small business credit
to provide collateral against loans is central to the availability and relationship lending: The importance of
banks’ decision. Therefore, we have had a loan guar- bank organisational structure. The Economic Journal,
antee scheme for small firms since 1981 and current 112(477), F32–F53.
discussions with UK policymakers have stressed the Binks, M. R., Ennew, C. T., & Reed, G. V. (1992). Information
asymmetries and the provision of finance to small firms.
importance of maintaining a public loan guarantee International Small Business Journal, 11(1), 35–46.
scheme in the post-Covid-19 period and enhancing its Brown, R., Liñares-Zegarra, J., & Wilson, J. O. S. (2019). The
capacity in a crisis period. (potential) impact of Brexit on UK SMEs: Regional evi-
There are also clear implications for small busi- dence and public policy implications. Regional Studies,
53(5), 761–770.
nesses themselves. First, developing and maintain- Brown, R., & Cowling, M. (2021). The geographical impact
ing a strong relationship with ones’ main bank is of the Covid-19 crisis on precautionary savings, firm
important for reducing information asymmetries and survival and jobs: Evidence from the United Kingdom’s
increasing the probability of future loan application 100 largest towns and cities. International Small Business
Journal, 0266242621989326
success. Furthermore, banks treat each loan applica- Calabrese, R., Degl’Innocenti, M., & Zhou, S. (2020). Expecta-
tion in isolation at a specific point in time and their tions of access to debt finance for SMEs in times of uncer-
decision reflects the prevailing economic condi- tainty. Journal of Small Business Management. https://doi.
tions at that time, although a poor application will be org/10.1080/00472778.2020.1756309
Carpenter, R. E., & Petersen, B. C. (2002). Is the growth of
treated as such at any time. So, some previous loan small firms constrained by internal finance? Review of
rejections may simply mean that economic conditions Economics and Statistics, 84(2), 298–309.
were poor or that the funding proposal was deemed Cavalluzzo, K. S., Cavalluzzo, L. C., & Wolken, J. D. (2002). Com-
inadequate. Returning later with a new funding pro- petition, small business financing, and discrimination: Evidence
from a new survey. The Journal of Business, 75, 641–679.
posal will be treated on its relative merits at that time; Cole, R., & Sokolyk, T. (2016). Who needs credit and who
thus, we encourage entrepreneurs not to simply drop gets credit? Evidence from the surveys of small business
out of the loan market as that may damage their firm finances. Journal of Financial Stability, 24, 40–60.
by reducing potentially profitable new investment. Cowling, M. (2010). The role of loan guarantee schemes in
alleviating credit rationing in the UK. Journal of Finan-
cial Stability, 6(1), 36–44.
Open Access This article is licensed under a Creative Com- Cowling, M., & Mitchell, P. (2003). Is the small firms loan
mons Attribution 4.0 International License, which permits guarantee scheme hazardous for banks or helpful to small
use, sharing, adaptation, distribution and reproduction in any business? Small Business Economics, 21(1), 63–71.
medium or format, as long as you give appropriate credit to the Cowling, M., Liu, W., & Ledger, A. (2012). Small business financ-
original author(s) and the source, provide a link to the Crea- ing in the UK before and during the current financial crisis.
tive Commons licence, and indicate if changes were made. The International Small Business Journal, 30(7), 778–800.
images or other third party material in this article are included Cowling, M., Liu, W., Minniti, M., & Zhang, N. (2016). UK
in the article’s Creative Commons licence, unless indicated credit and discouragement during the GFC. Small Busi-
otherwise in a credit line to the material. If material is not ness Economics, 47(4), 1049–1074.
included in the article’s Creative Commons licence and your Cowling, M., Liu, W., & Zhang, N. (2018). Did firm age, expe-
intended use is not permitted by statutory regulation or exceeds rience, and access to finance count? SME performance
the permitted use, you will need to obtain permission directly after the global financial crisis. Journal of Evolutionary
from the copyright holder. To view a copy of this licence, visit Economics, 28(1), 77–100.
http://creativecommons.org/licenses/by/4.0/. Cowling, M., Brown, R., & Rocha, A. (2020). ? covid19? Did you
save some cash for a rainy COVID-19 day? The crisis and
SMEs. International Small Business Journal, 38(7), 593–604.
Cowling, M., Liu, W., & Zhang, N. (2020). In the post-crisis
References world, did debt and equity markets respond differently to
high-tech industries and innovative firms? International
Arulampalam, W. (2001). Is unemployment really scarring? Small Business Journal, 0266242620947281.
Effects of unemployment experiences on wages. The Eco- Cowling, M., Nadeem, S. P., Foster, C., & Baranova, P. (2020).
nomic Journal, 111(475), F585–F606. Can local finance add value to local small business?
13
1350 M. Cowling et al.
Evidence from a UK Local Loan and Grant Fund. Interna- Myers, S. C., & Majluf, N. S. (1984). Corporate financing and
tional Review of Entrepreneurship, 18(1). investment decisions when firms have information that
Cowling, M., Liu, W., & Conway, E. (2021). Ethnicity and investors do not have. Journal of Financial Economics,
bank lending before and during COVID-19. Interna- 13(2), 187–221.
tional Journal of Entrepreneurial Behavior & Research, Popov, A., & Udell, G. F. (2012). Cross-border banking, credit
Forthcoming. access, and the financial crisis. Journal of International
De Meza, D. (2002). Overlending? The Economic Journal, Economics, 87(1), 147–161.
112(477), F17–F31. Psillaki, M., & Daskalakis, N. (2009). Are the determinants of
Firth, D. (1993). Bias reduction of maximum likelihood esti- capital structure country or firm specific? Small Business
mates. Biometrika, 80(1), 27–38. Economics, 33(3), 319–333.
Fraser, S. (2009). Is there ethnic discrimination in the UK mar- Qi, S., & Nguyen, D. D. (2021). Government connections and
ket for small business credit? International Small Business credit access around the world: Evidence from discour-
Journal, 27(5), 583–607. aged borrowers. Journal of International Business Stud-
Freel, M., Carter, S., Tagg, S., & Mason, C. (2012). The latent ies, 52(2), 321–333.
demand for bank debt: Characterizing “discouraged bor- Revest, V., & Sapio, A. (2012). Financing technology-based
rowers.” Small Business Economics, 38(4), 399–418. small firms in Europe: What do we know? Small Business
Han, L., Fraser, S., & Storey, D. J. (2009). Are good or bad Economics, 39(1), 179–205.
borrowers discouraged from applying for loans? Evidence Riding, A., Madill, J., & Haines, G. (2007). Incrementality of
from US small business credit markets. Journal of Bank- SME loan guarantees. Small Business Economics, 29(1–
ing & Finance, 33(2), 415–424. 2), 47–61.
Heckman, J. J. (1979). Sample selection bias as a specification Sannajust, A. (2014). Impact of the World Financial Crisis to
error. Econometrica: Journal of the Econometric Society SMEs: The determinants of bank loan rejection in Europe
47(1):153–161 and USA. IPAG Business School, 327, 1–29.
Howorth, C. A. (2001). Small firms’ demand for finance: A Scott, J. A. (2006). Loan officer turnover and credit availability
research note. International Small Business Journal, for small firms. Journal of Small Business Management,
19(4), 78–86. 44(4), 544–562.
Jappelli, T. (1990). Who is credit constrained in the US Sette, E., & Gobbi, G. (2015). Relationship lending during a
economy? The Quarterly Journal of Economics, 105(1), financial crisis. Journal of the European Economic Asso-
219–234. ciation, 13(3), 453–481.
Jappelli, T., & Pagano, M. (2002). Information sharing, lending Shyam-Sunder, L., & Myers, S. C. (1999). Testing static trade-
and defaults: Cross-country evidence. Journal of Banking off against pecking order models of capital structure.
& Finance, 26(10), 2017–2045. Journal of Financial Economics, 51(2), 219–244.
King, G., & Zeng, L. (2001). Logistic regression in rare events Stiglitz, J. E., & Weiss, A. (1981). Credit rationing in markets
data. Political Analysis, 9(2), 137–163. with imperfect information. The American Economic
Kon, Y., & Storey, D. J. (2003). A theory of discouraged bor- Review, 71(3), 393–410.
rowers. Small Business Economics, 21(1), 37–49. Tang, Y., Deng, C., & Moro, A. (2017). Firm-bank trusting
Kozlowski, J., Veldkamp, L., & Venkateswaran, V. relationship and discouraged borrowers. Review of Mana-
(2020). Scarring body and mind: the long-term belief- gerial Science, 11(3), 519–541
scarring effects of Covid-19 (No. w27439). National Tumino, A. (2015). The scarring effect of unemployment from
Bureau of Economic Research. the early ‘90s to the Great Recession (No. 2015–05).
Krasniqi, B. A. (2010). Are small firms really credit constrained? ISER Working Paper Series.
Empirical evidence from Kosova. International Entrepre- Udell, G. F. (2020). SME access to finance and the global
neurship and Management Journal, 6(4), 459–479. financial crisis. Journal of Financial Management, Mar-
Leitgöb, H. (2013).The problem of modeling rare events in kets and Institutions, 8(01), 2040003.
ML-based logistic regression: Assessing potential rem- Watson, R., & Wilson, N. (2002). Small and medium size
edies via MC simulations. Paper presented at the 5th con- enterprise financing: A note on some of the empiri-
ference of European survey research association, Lisbon, cal implications of a pecking order. Journal of Business
Portugal. Finance & Accounting, 29(3–4), 557–578.
Mac an Bhaird, C., Vidal, J. S., & Lucey, B. (2016). Discour- Welsh, E. C. (1952). Government aid to business expansion.
aged borrowers: Evidence for Eurozone SMEs. Journal of The American Economic Review, 42(2), 418–427.
International Financial Markets, Institutions and Money, Xiang, D., Worthington, A., & Higgs, H. (2015). Discour-
44, 46–55. aged finance seekers: An analysis of Australian small and
Michaelas, N., Chittenden, F., & Poutziouris, P. (1999). Finan- medium-sized enterprises. International Small Business
cial policy and capital structure choice in UK SME: Journal, 33(7), 689–707.
Empirical evidence from company panel data. Small Busi-
ness Economics, 12(2), 113–130. Publisher’s note Springer Nature remains neutral with regard
Mishkin, F. S. (2011). Over the cliff: From the subprime to the to jurisdictional claims in published maps and institutional
global financial crisis. Journal of Economic Perspectives, affiliations.
11(1), 49–70.
Myers, S. C. (1984). The capital structure puzzle. The Journal
of Finance, 39(3), 574–592.
13
File and source
- File
- cowling-2021-has-previous-loan-rejection-scarred-firms-from-applying-for-loans-during-covi.pdf
- Size
- 714,915 bytes
- SHA-256
- 7188b86ea47ca0dff4cc409f4b188f6fa1ad49b767a52591951e7661afe20326
- Our copy
- cowling-2021-has-previous-loan-rejection-scarred-firms-from-applying-for-loans-during-covi.pdf
- Original
- link.springer.com