Upload
structuralpolicyanalysis
View
178
Download
3
Embed Size (px)
Citation preview
Distressed Banks, Distorted Decisions
Gareth Anderson, University of OxfordRebecca Riley, NIESRGarry Young, NIESR
January 2018
Motivation
Figure: 5 year CDS premia
Four banking groups account for around 80% of loans to UK firms.The crisis had a heterogeneous impact on these banking groups.
Contribution
Exploit an exogenous source of credit constraints faced by UK firms,induced by banking relationships they maintained prior to crisis.Use natural experiment approach to assess whether the exit margin offirms was distorted by credit constraints following the crisis.
Main Messages
Maintaining relationships with Distressed Banks increased theprobability of firms exiting.Increased probability of exiting was not concentrated in the mostunproductive firms.
I Tight credit may have limited the cleansing impact of the recession.I While micro impact is significant, macro impact is limited.
Related Literature: Credit Constraints and Firm Activity
Global financial crisis provides a natural experiment for studying theimpact of tight credit conditions on firm activity.Offers a solution to disentangling credit demand and credit supply.Tighter credit conditions following the crisis adversely affected firminvestment and firm employment (e.g. Duchin et al. (2010) andBentolila et al. (2013)).
Related Literature: Credit Constraints and Firm Dynamics
Contrasting views on how recessions and crises affect the process offirm entry and firm exit.“Cleansing” view: recession accelerates the exit of inefficient firmswhich are insulated during boom times. (e.g. Schumpeter (1934);Caballero and Hammour (1994)).“Sullying” / “Scarring” view: frictions associated with recessions mayadversely affect productivity (Barlevy (2002, 2003); Ouyang (2009);Osotimehin and Pappadà (2016)).
Productivity Distortions from Banking Crises
Firms that need external credit to survive may be more productive(e.g. Barlevy (2003)).Less productive firms which are less reliant on bank finance facereduced competition.Forbearance/ ”zombie lending” (e.g. Peek and Rosengren (2005);Arrowsmith et al. (2013); Roland (2016)).
Model: Credit frictions and productivity cutoffs
Figure: Impact of credit market frictions on productivity cutoffs
Model: Productivity distribution
0 0.5 1 1.5 2 2.5 3 3.5 40
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Figure: Impact of credit market frictions on the productivity distribution
Related Literature: Credit Constraints and Firm Dynamics
Recent empirical studies have found some support for the view thatcredit constraints can weaken the “cleansing” effect of recessions.Evidence that the negative relationship between firm productivity andthe probability of exit weakened following the Great Recession (e.g.Foster et al. (2016); Harris and Moffat (2016)).Some highly productive, credit constrained firms are forced to exit(e.g. Eslava et al. (2010); Hallward-Driemeier and Rijkers (2013)).
Distressed and Non Distressed Banks
Follow the approach of Bentolila et al. (2013):I Define Distressed Banks as those which obtained state funding between
2008 and 2009 or required a takeover.I Define NonDistressed Banks as those which did not receive state
funding and did not require a takeover.I Divide firms into Treatment and Control groups based on which banks
they had relationships with in 2008.
Identification: Distressed and Non Distressed Banks
For identification, require thatI i) Credit supply conditions tightened by more for firms which had
pre-crisis relationships with Distressed Banks.I ii) Firms subject to a tightening of credit conditions cannot easily
switch bank.
Identification: Distressed and Non Distressed Banks
i) Credit supply conditions tightened by more for firms which hadpre-crisis relationships with Distressed Banks.
I CDS premiums of banks similar prior to crisis, but different followingcrisis. Increase in CDS spreads particularly pronounced for RBS andLBG.
I Little evidence that lending commitments made by Distressed Banks inreturn for public support influenced their lending behaviour.
I Overall, evidence suggests the contraction in credit supply byDistressed Banks was greater than that of NonDistressed Banks.
Identification: Distressed and Non Distressed Banks
ii) Firms cannot easily switch bank.I Asymmetric information makes switching difficult.I Bank switching by UK firms is low, with no pick-up following crisis.
Figure: Evolution of Switching Rates Over Time
Data: Corporate Balance Sheet
Consider key measures of corporate activity (2002-2012 BvD/FAME).Control for firm age, relationship length, foreign owned firms,exporters, credit score band, court judgements, firm size, account type.
Data: Corporate Banking Relationships
Lenders commonly require companies to provide security against aloan, in the form of a mortgage, a floating charge or a fixed charge.In the UK, registered companies are required to report charges andmortgages to Companies House within 21 days of their creation date.We identify firm-bank relationships using a textual algorithm.
Empirical Specification
Baseline specification- Linear probability model.How does having a relationship with a Distressed Bank at time t affectthe likelihood of exit in the subsequent period for firm i in industry j :
Yi ,t = gj ,+Xi ,tk +b1⇥Distressed Banki ,t +b2⇥Post Crisist (1)+b3⇥Distressed Banki ,t ⇥Post Crisist + ei ,t
where
Yi ,t is an indicator variable equal to 1 if firm i subsequentlyexits in the specified time frame and 0 otherwise
Summary Statistics, by Banking Relationship
2004 2008
ND D No Bank ND D No Bank
Exit in 2 years 11% 10% 13% 11% 11% 14%
Exit in 4 years 21% 20% 26% 19% 20% 25%
Start-Up 14% 13% 33% 6% 8% 24%
Young 34% 33% 58% 26% 28% 56%
Foreign Owned 2% 3% 3% 2% 3% 2%
Exporter 1% 2% 1% 2% 1% 1%
Observations 66334 70695 403407 70400 81237 546880
Results: Baseline specification
Table: Baseline model
2 Year Exit 3 Year Exit 4 Year Exit
Distressed -0.001 0.001 0.001(0.001) (0.002) (0.002)
Post-Crisis -0.001 0.012⇤⇤⇤ 0.026⇤⇤⇤
(0.002) (0.005) (0.006)Distressed * Post-Crisis 0.006⇤⇤⇤ 0.006⇤⇤⇤ 0.008⇤⇤⇤
(0.001) (0.002) (0.002)
Industry Fixed Effects Yes Yes YesFirm Controls Yes Yes YesR-Squared 0.143 0.112 0.117Observations 303953 300244 288648
Robust standard errors. ***, **, * shows significance at the 1%, 5% and 10% levels.
Results: Baseline specification
Change in probability of exit significantly higher for firms withDistressed Banks following the crisis than firms attached toNonDistressed Banks.Change in the probability of exit within 2 years was around 0.6percentage points higher for firms which had a relationship withDistressed Banks relative to firms with NonDistressed Banks.
Exit Dynamics and Productivity
Consider a smaller sample of firms for which we are able to calculate aproxy for their gross value added productivity:
Productivityi ,t =GVAi ,t
Employees
where
GVAi ,t is a proxy of gross value added in real terms given bythe sum of a firm’s reported Operating Profits and theCost Of Employees, deflated by industry deflators.
Exit Dynamics and Productivity
Split the observations into productivity quintiles, based on proxy forgross value added productivity.
Figure: Firms which exited andbanked with “Non-Distressed” banks
Figure: Firms which exited andbanked with “Distressed” banks
Exit Dynamics and Productivity
Interact variables of interest in our Baseline specification(Distressed Bank , Post Crisis and Post Crisis⇥Distressed Bank) withindicator variables for the productivity quintile a firm is in:
Yi ,j ,t, = gj +Xi ,tk +5
Âk=1
b1,k(Distressed Banki ⇥Prodi ,k,t) (2)
+5
Âk=1
b2,k(Post Crisist ⇥Prodi ,k,t)
+5
Âk=1
b3,k(Distressed Banki ⇥Post Crisist ⇥Prodi ,k,t)+ ei ,j ,t
Table: Firm Exit, by Productivity Quintile
2 Year Exit 3 Year Exit 4 Year Exit
Lowest Productivity Quintile
Distressed * Post-Crisis -0.036⇤ -0.033⇤ -0.020
(0.018) (0.018) (0.032)
Productivity Quintile 2
Distressed * Post-Crisis 0.011 0.022 0.022
(0.016) (0.020) (0.021)
Productivity Quintile 3
Distressed * Post-Crisis 0.031⇤⇤⇤ 0.017 0.025
(0.011) (0.016) (0.017)
Productivity Quintile 4
Distressed * Post-Crisis 0.012 0.003 0.018
(0.010) (0.012) (0.016)
Highest Productivity Quintile
Distressed * Post-Crisis 0.019⇤ 0.011 0.014
(0.010) (0.010) (0.014)
Industry Fixed Effects Yes Yes Yes
Firm Controls Yes Yes Yes
R-squared 0.227 0.168 0.204
Observations 18284 18638 19016
Robust standard errors. ***, **, * shows significance at the 1%,5% and 10% significance levels.
Table: Firm Exit, by Leverage and Productivity
Lowest Leverage Tercile Middle Leverage Tercile Highest Leverage Tercile
2 Year Exit 4 Year Exit 2 Year Exit 4 Year Exit 2 Year Exit 4 Year Exit
Lowest Productivity Quintile
Distressed * Post-Crisis 0.008 -0.003 -0.032 0.015 -0.081⇤⇤ -0.049
(0.029) (0.041) (0.034) (0.047) (0.034) (0.058)
Productivity Quintile 2
Distressed * Post-Crisis 0.026 0.024 -0.014 0.012 0.019 0.018
(0.024) (0.025) (0.023) (0.031) (0.040) (0.040)
Productivity Quintile 3
Distressed * Post-Crisis 0.005 -0.007 0.020 0.044⇤ 0.069⇤⇤ 0.010
(0.016) (0.033) (0.017) (0.025) (0.030) (0.035)
Productivity Quintile 4
Distressed * Post-Crisis 0.008 -0.008 0.026⇤ 0.030 0.006 0.012
(0.016) (0.017) (0.015) (0.025) (0.023) (0.036)
Highest Productivity Quintile
Distressed * Post-Crisis 0.011 -0.010 -0.006 0.005 0.054 0.044
(0.015) (0.022) (0.022) (0.024) (0.035) (0.041)
Industry Fixed Effects Yes Yes Yes Yes Yes Yes
Firm Controls Yes Yes Yes Yes Yes Yes
R-Squared 0.191 0.162 0.145 0.145 0.295 0.278
Observations 6032 6274 6034 6275 6218 6467
Robust standard errors. ***, **, * shows significance at the 1%,5% and 10% significance levels.
Robustness: Placebo Crises
DiD specification relies on the assumption of parallel pre-crisis trends.Undertake placebo tests considering alternative placebo “crises”.Consider the impact of having relationships with Distressed Banks onthe two year exit probability for two placebo “crises”: 2004 and 2006.
Robustness: Placebo Crises
Table: Placebo Crises, 2 Year Exit Rate
Placebo “Crisis”=2004 Placebo “Crisis”=2006 True Crisis=20082 year exit 2 year exit 2 year exit
Distressed * “Crisis” -0.003 -0.003 0.006⇤⇤⇤
(0.003) (0.002) (0.001)
Industry Fixed Effects Yes Yes YesFirm Controls Yes Yes YesR-Squared 0.088 0.113 0.143Observations 267902 289347 303953
Robust standard errors. ***, **, * shows significance at the 1%, 5% and 10% levels.
Robustness: Weighted Regression
Productivity sample limited to firms which report Operating Profits,Employees and Cost Of Employees.Under-represents smaller firms.Weight observations in productivity sample to match the number offirms in each industry-size-bank group-year cell in baseline sample.
Table: Firm Exit, Weighted Regression
2 year exit 3 year exit 4 year exit
Lowest Productivity Quintile
Distressed * 2008 -0.003 0.028 0.024
(0.031) (0.043) (0.054)
Productivity Quintile 2
Distressed * 2008 0.016 -0.020 0.038
(0.041) (0.056) (0.042)
Productivity Quintile 3
Distressed * 2008 0.058⇤⇤ 0.031 0.001
(0.027) (0.031) (0.044)
Productivity Quintile 4
Distressed * 2008 0.033 -0.025 0.026
(0.022) (0.044) (0.050)
Highest Productivity Quintile
Distressed * 2008 0.039 0.056⇤ 0.035
(0.028) (0.031) (0.030)
Industry Fixed Effects Yes Yes Yes
Firm Controls Yes Yes Yes
R-Squared 0.276 0.252 0.298
Observations 18284 18638 19016
Robust standard errors. ***, **, * shows significance at the 1%, 5% and 10% levels.
Robustness: Alternative Measure of Productivity
Productivity sample limited to firms which report Operating Profits,Employees and Cost Of Employees.Use an alternative measure of productivity, given by the ratio of GVAto the cost of employees, following Giordano et al. (2015).Productivity_ai ,t =
GVAi ,t
Cost Of Employees
Table: Firm Exit, Alternative Measure of Productivity
2 Year Exit 3 Year Exit 4 Year Exit
Lowest Productivity QuintileDistressed * 2008 -0.027⇤ -0.056⇤⇤ -0.018
(0.015) (0.024) (0.021)Productivity Quintile 2Distressed * 2008 0.036⇤⇤ 0.028 0.032
(0.017) (0.025) (0.020)Productivity Quintile 3Distressed * 2008 -0.003 0.012 0.018
(0.009) (0.013) (0.015)Productivity Quintile 4Distressed * 2008 0.007 0.017 0.025
(0.010) (0.014) (0.020)Highest Productivity QuintileDistressed * 2008 0.017 0.004 0.006
(0.012) (0.017) (0.016)
Industry Fixed Effects Yes Yes YesFirm Controls Yes Yes YesR-Squared 0.247 0.234 0.272Observations 23635 24282 24716
Robust standard errors. ***, **, * shows significance at the 1%, 5% and 10% levels.
Conclusion
Exploit pre-crisis banking relationships as an exogenous source ofcredit constraints faced by UK firms.Credit constraints faced by UK firms following the financial crisis hada detrimental impact on their probability of survival.Credit constraints distorted productivity distribution of exiters.Some credit constrained firms may have been forced to exit theirindustry, despite being more productive than surviving competitors.While micro impact is significant, macro impact is limited.
Model: Overview
Closed economy Melitz (2003) model with credit constraints, a’laManova (2013).Firms must finance a fraction of their fixed costs upfront, usingfinancial intermediaries.Fraction of fixed cost to be financed varies across firms.Credit market frictions raise the productivity threshold for firms.Firms more dependent on external finance are more adversely affectedby credit market frictions.
Model: Consumers
Constant elasticity of substitution (CES) preferences over a continuumof goods indexed by w over ⌦:
U =⇥R
w2⌦ q(w)rdw⇤ 1
r
where s = 11�r > 1
With optimal consumption and expenditure for different varieties givenby:
q(w) = Qhp(w)P
i�s
r(w) = p(w)q(w) = Rhp(w)P
i1�s
Model: Producers
Continuum of firms, each of which produces a variety w .Entrants are required to pay an entry cost, fe , draw productivity level,j , from g(j)Labour consists of a fixed cost f and a variable cost which depends ona firm’s productivity, j :
l = qj + f
Model: Producers
Firms have to pay a fraction, di of fixed cost, f , upfront.Fraction low, dL , with probability c or high, dH , with probability1�c .Upfront fixed cost financed by a financial intermediary.Financial contract: at start of period, firms offer repayment F to bepaid at the end of the periodIntermediaries obtain agreed repayment F with probability l 1.With probability 1�l , firm defaults and the intermediary does notreceive F , but seizes collateral tfe . from the firm.In the case of default, the firm replaces its collateral, tfe .
Model: Producers
Upon entry, the problem of the firm is to choose its price, quantityand repayment to maximise profits subject to three constraints:
max
p(j),q(j),F (j,di )p(j)q(j)�
hq(j)
j +(1�di )f +lF (j,di )+(1�l )tfei
subject to
(1) q(j) = Qhp(j)P
i�s
(2) F (j,di ) p(j)q(j)� q(j)j � (1�di )f
(3) di f lF (j,di )+(1�l )tfe
Assume constraint (3) binds with equality.Define a productivity threshold, j⇤
di, below which producers choose
not to produce.
Model: Comparative Statics
Solve the model to find the two productivity thresholds, j⇤dL
and j⇤dH
.To illustrate comparative statics, calibrate the model, closely followingMelitz and Redding (2013).Consider how contract imperfections, given by l , affect the cutoffproductivities and the productivity distribution.
Model: Credit frictions and productivity cutoffs
Figure: Impact of credit market frictions on productivity cutoffs
Model: Productivity distribution
0 0.5 1 1.5 2 2.5 3 3.5 40
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Figure: Impact of credit market frictions on the productivity distribution