PUBLIC CREDIT GUARANTEES AND SME FINANCE

1
PUBLIC CREDIT GUARANTEES AND SME FINANCE

• Salvatore Zecchini
• University of Tor Vergata
• Rome
• and
• Marco Ventura
• Institute for Studies and Economic
  AnalysesISAE, Rome
• The World Bank Conference 13-14 March 2008
• Partial Credit Guarantee Schemes Experiences
  and Lessons

2
Is State intervention or a State-funded guarantee
scheme an effective instrument to promote lending
to small firms?

• Contrasting views in economic literature
• Against
• costly, financially unsustainable, no conclusive
  evidence of additional lending to SMEs, no
  substitute for correction of system failure.
• In favour
• new access to credit
• lower funding cost
• need of tight financial criteria.
• Our aim is to test the impact of a State-funded
  guarantee scheme on SME financing in terms of
  credit additionality, borrowing costs, financial
  sustainability.

3
The focus of our empirical tests is Italys Fund
for Guarantee to SME (SGS or Fund)

• We will present in turn
• A brief review of the economic literature on the
  subject of SME financial constraints
• Eligibility and selection criteria of the Fund
• Funds performance from the SME financing
  viewpoint
• Funds financial sustainability
• Our econometric approach as compared to that of
  other authors
• Econometric test of guarantee impact on SME
  borrowing cost
• Empirical evidence on credit additionality
• Conclusions this public guarantee instrument has
  had a positive effect on SME credit access and
  borrowing cost, albeit limited.

4
SME financial constraints in the economic
literature and in Italy

• Some empirical evidence shows financial
  constraints are inversely related to firm size
  (Bagella-Becchetti-Caggese 2001)
• Start-up firms, young enterprises, smaller ones,
  innovative ones, with fewer tangible assets and
  an uncertain track record are subject to much
  tighter financial constraints than other firms,
  especially, under the form of credit rationing by
  the banking system, Berger-Udell, 1998.
• Small firms are subject to impact of
  imperfections in bank credit market more than
  other firms, due to ex ante asymmetric
  information between bank lenders and borrowers,
  agency problems, relatively high evaluation and
  monitoring costs for the lender.
• The interest rate cannot often serve as a tool to
  distinguish good borrowers from bad ones, since
  info asymmetries can lead to adverse selection
  (Stiglitz-Weiss, 1981).

5
Under certain conditions, the provision of
collateral can lessen credit rationing and
borrowing costs (Coco,2000).

• Can the provision of outside guarantees be a tool
  to overcome market imperfections and lack of
  inside collateral, thereby giving SMEs broader
  access to bank financing?
• The guarantee value is a function of length and
  cost of legal procedures, as well as of the loan
  recovery rate.
• Given financial market imperfections and
  institutional weaknesses, Governments in general
  resort to various industrial policy tools to
  improve credit allocation to the advantage of
  SMEs.
• One of them is credit guarantees.

6
An outline of Italys guarantee system and the
role of mutually-based guarantee institutions and
the Fund

• Multi-pillar and multi-layer system, based on a
  mix of private and public funds.
• 3 pillars a) mutual guarantee institutions (MGI)
• b) banks and other financial institutions
• c) State- or Region Government-sponsored
  guarantee
• Funds.
• Multi-layer structure
• grassroots level MGIs and banks
• second level second-tier MGIs (credit
  reinsurance) and regional reinsurance
  institutions
• third level 3 State guarantee Funds for SME
  credit, namely SGS, Fund for crafts credit, and
  Fund for Farm credit.
• Italys MGI system is the largest one in Europe.
  About 600 MGI, 37 ot total outstanding
  guarantees in 2005 and 46 of beneficiary SMEs.

7
Italys State Fund for Guarantee to SMEs

• In 2005, equity base was 233.5 millions.
• In 6 years, 4.6 billion of guaranteed loans,
  equal to 3 of total lending to small firms that
  are eligible.
• Eligibility criteria

8
(No Transcript)
9
The economic performance of the Fund

• The Fund is run according to tight criteria
• Eligibility criteria lead to low rejection rate
  (83 of applications)
• Guarantee coverage rate was on average about 50
  of debt principal, with a 25-88 dispersion range
• Some preference was given to disadvantaged groups
  (women in business, micro firms)
• Manufacturing and construction industry received
  71 of guarantees
• Investment projects received 54 of total
  guarantee. But rising share of guarantees against
  lending for working capital
• Loan maturity structure concentrated on
  medium-term loans (48)
• Regional distribution industrial regions got
  60 South 26
• Emphasis on counter-guaranteeing MGIs (61 of
  total)
• Overall, significant degree of risk aversion,
  guarantees were driven to a significant extent by
  credit supply institutions, support to mutual
  credit guarantees .

10
(No Transcript)
11

12
Financial sustainability of the Fund

• The degree of financial sustainability is
  assessed through the following equation
• 1 L A I F O S
• where
• L loan losses
• A administration expenses
• I public debt service cost (cost of use of
  borrowed capital)
• F guarantee fees
• O other income, such as the return from the
  investment of reserves
• S the amount of public subsidy to cover any
  losses.
• The subsidy element is the balancing item that
  allows to avoid the exhaustion of the capital
  base.

13
Financial performance
14

• Funds financial performance is better than
  similar schemes in other European countries
  (losses 0.25 vs. 2-10 of guaranteed loan
  portion)
• The default ratio is lower than that of Italys
  banking system (1.83 vs. 5.89)
• Small firms and micro firms generated less losses
  for the Fund than medium-sized firms (49 of
  losses in medium-sized firms)
• Loan default rates appear as being an increasing
  function of the loan size and guarantee size
  (following figure).
• Highest default rate is in the smallest loan
  class (up to 10,000), but a relatively low loss
  rate, because of low guarantee coverage.
• 69 of defaults are among loans for working
  capital needs
• 59 of defaults happens after the first 2 years
  of the loan
• Regional distribution of defaults is heavily
  dependent on sectoral distribution.

15
(No Transcript)
16
(No Transcript)
17
Subsidy rate

• S/G (( L A I - F - O )/G ) 100
• namely,
• 0.25 0.39 0.47 ( 0.66) - 0.35 - 0.012
  0.75 ( 0.94)
• The subsidy rate was much higher for those
  enterprises that were charged no fee it is
  estimated to go on average up to 1.29 of the
  guarantee.
• Fees did not cover either losses or operating
  expenses. There was in fact a current-account
  deficit averaging 0.28 per guarantee, that
  prevented the scheme from breaking even.
• Such a deficit (0.14 per euro of guarantee1)
  looks, however, very low compared to the other
  State-funded subsidy schemes for enterprises,
  1 This is the ratio of the deficit to the
  amount of guaranteed loans, and is equal to the
  product of the deficit ratio by the guarantee
  coverage ratio (tab. 3).

18
Approach to credit additionality and interest
cost reduction

• To carry out quantitative tests of guarantee
  impact, reference to empirical literature on the
  presence of tighter financial constraints for
  some groupings of firms and on their underlying
  factors.
• Applications to models of investment demand
  together with notion of capital mkt imperfections
  and disparities in firms access to credit mkt
  (Hubbard 1998).
• Cash flow sensitivity of investment demand
  (Fazzari et al. 1988, Kaplan Zingales 1997).
• Credit rationing some firms inability to get
  credit, even if they would pay higher interest
  rate (Stliglits Weiss 1981)
• Attempts to estimate impact of Gvt credit
  programmes (Gale, Boocock Shariff, Riding Haines)
• Novelty of our approach.

19

• Is the Fund effective?
• 2. What happens once the guarantee expires?

• Definition of effectiveness of the Fund
• Guaranteed firms are charged lower financial
  costs once guaranteed
• and/or
• Guaranteed firms receive a greater amount of bank
  loans. Additionality or incrementality effect.

20

• Answer to question 1
• Yes it is, the Fund is effective
• Answer to question 2
• Puzzling evidence formerly guaranteed firms are
  charged lower financial costs, but they are not
  granted higher quantity of bank loans.
• Further investigation

21
HOW TO REACH THESE ANSWERS

• Before giving any sensible answer to qn. 1 2
  one must rule out the
• ANTICIPATION EFFECT
• The Fund systematically guarantees better
  performers, therefore we want to make sure that

22

• Guaranteed firms are not charged lower financial
  costs BEFORE receiving the guarantee
• and/or
• Guaranteed firms do not receive higher quantity
  of loans BEFORE receiving the guarantee

23
where rt Nx1 vector of (log of) financial costs
in 1999 x1t Nx1 vector of (log of) number of
employees in 1999 x2t Nx1 vector of (log of)
sales in 1999 x3t Nx1 vector of (log of) bank
debt in 1999 dtn dummy variables, that takes on
value of 1 if the firm is guaranteed at time tn
(where t1999) and to 0 otherwise ut error term.
24
Table 4. Estimates of the ? parameter using data
prior to 1999 for firms receiving the guarantee
in the following years (cost effect)
Standard errors in parenthesis. , and
indicate 1, 5 and 10 significance levels,
respectively. Different regressions are reported
in each column by changing the dummy in order to
account for the firms guaranteed in different
years. For instance, in column 3 we report the
estimated ? coefficient related to the 1999
financing cost for firms that received a
guarantee in 2001.
25

• The dummy coefficient is always positive and
  significant
• guaranteed firms were not better performers in
  terms of cost, they were charged higher financial
  costs in 1999, wrt other SMEs

26
where yt Nx1 vector of (log of) bank debt in
1999 x1t Nx1 vector of (log of) number of
employees in 1999 x2t Nx1 vector of (log of)
sales in 1999 x3t Nx1 vector of (log of) total
assets in 1999 dtn dummy variables, that takes
on value of 1 if the firm is guaranteed at time
tn (where t1999) and to 0 otherwise ut error
term.
27
Table 5. Estimates of the ? parameter using data
prior to 1999 for firms receiving the guarantee
in the following years (additionality)
Standard errors in parenthesis. , and
indicate 1, 5 and 10 significance levels,
respectively. Different regressions are reported
in each column by changing the dummy in order to
account for the firms guaranteed in different
years. For instance, in column 3 we report the
estimated ? coefficient related to the 1999 bank
debt for firms that received a guarantee in 2001.
28
EVIDENCE

• The dummy coefficient, d, is negative and
  significant for 2000-2004 Guaranteed firms in
  those years received a lower quantity of bank
  loans, i.e. they were not better performers,
• d is not significantly different from zero for
  2005. Guaranteed firms in that year did not
  receive a higher quantity of bank loans, i.e.
  they were not better performers,
• d is positive and significant for 2006 Guaranteed
  firms in 2006 were better performers, we rule
  them out from the sample otherwise we cannot
  single out the causal effect of the Fund on SMEs.

29
ESTIMATION STRATEGY

• To isolate the causal effect of the Fund we use
  the DID estimator
• once defined an appropriate outcome variable the
  DID compares the average time difference of the
  treated to the average time difference of the
  control group.
• We exploit a twofold counterfactual
• 1. guaranteed firms before receiving the
  guarantee, 1999
• 2. non guaranteed firms

30

• When the anticipation effect did not occur the
  underlying hypothesis of the DID are satisfied.
• We can isolate the causal effect of the Fund on
  the treated unit ATT
• The proves of answers to questions 1 and 2

31
Table 6 - DID estimate of the causal effect of
the guarantee on the (log of) financial costs
1999-2005 (cost effect)
Robust standard errors in parenthesis. ,
and indicate 1, 5 and 10 significance
levels, respectively. S.E. are computed through
the SUR (PCSE) coefficient covariance matrix to
account for both cross-section heteroskedasticity
and correlation.
32

• The dummy coefficient is not significantly
  different from zero, therefore financial costs
  charged to guaranteed SMEs are in line with non
  guaranteed.
• Recall before the Fund started operating
  guaranteed firms were charged higher financial
  costs, other things being equal. See anticipation
  effect.
• The effect persists once the guarantee expires

33
Table 6 - DID estimate of the causal effect of
the guarantee on the (log of) bank debt 1999-2005
(additionality effect)
Robust standard errors in parenthesis. ,
and indicate 1, 5 and 10 significance
levels, respectively. S.E. are computed through
the SUR (PCSE) coefficient covariance matrix to
account for both cross-section heteroskedasticity
and correlation.
34

• The dummy coefficient is positive and
  significant, guaranteed firms are granted a
  higher quantity of bank loans, 13.77 computed as
  exp(d)-1100. In the same direction Bocock
  Sharif (2005), Gale (1991), Riding Madill Haines
  (2006).
• Recall before the Fund started operating
  guaranteed firms were granted a lower quantity of
  bank loans (see estimates of anticipation effect
  for additionality).
• Once the guarantee expires no traces are left,
  guaranteed SMEs go back to the previous situation
  of lower bank loans.

35
SUMMING UP

• THE FUND HAS BEEN PROVED TO BE EFFECTIVE IN
• DECREASING FINANCIAL COSTS FOR GUARANTEED FIRMS
• INCREASING THE AMOUNT OF BANK LOANS

36

• ONCE THE GUARANTEE EXPIRES
• THERE ARE TRACES OF PERSISTENCE IN THE COST
  EFFECT, WHILE
• 2. THERE IS NO EVIDENCE OF PERSISTENCE IN THE
  ADDITIONALITY EFFECT.