Bank culture and risky lending Linh Nguyen (University of St - - PowerPoint PPT Presentation

Bank culture and risky lending

Linh Nguyen (University of St Andrews) Louis Nguyen (University of St Andrews) Ben Sila (University of Edinburgh) Conference on Professional and Ethical Standards in Banking May 2016

Motivation

• Key economic features of the past decade:
• Reckless bank behavior
• Unprecedented wave of bailouts and failures around the world
• “Bank culture” – the source of all evils
• Research question: does bank culture affects financial stability?
• Via credit decisions
• A core business function of banks
• Require human discretion

What do we fi find?

• Bank culture affects financial stability via bank lending behavior
• Banks with compete-oriented culture are more likely to lend to sub-investment

borrowers, require fewer covenants and ask for a higher loan spread.

• They enjoy extraordinarily fast lending growth at the expense of more bad
• loans. As a result, they make greater contribution to systemic risk.
• Compete-banks behave like a slow-exploding bomb
• They make most contribution to systemic risk during boom time and only

realize losses (bad loans) during bust time.

How do we measure corporate culture?

Corporate cult lture used in in previous lit literature

1. Management scholars conduct case studies on a few selected companies

• Only captures some, but not all aspects of culture
• Temporary
• Generalization issue

2. Recent work in finance and economics use annual rankings of companies (“Best companies to work for”) or company surveys (“Glassdoor”)

• Limited to a subset of very large firms
• Firms pay to do surveys  self-selection issues

Our approach to capture corporate cult lture

• Textual analysis of company’s annual reports
• First used introduced to the finance literature by Fiordelisi and Ricci (2014)
• Assumption: Language mirrors values and culture
• Classifies into four major dimensions
• Compete: “compete hard, move fast, and play to win”
• Create: “create, innovate, and envision the future”
• Control: “better, cheaper, and surer”
• Collaborate: “development, empowerment, commitment”

Create and Compete  External, outward-looking Control and Collaborate  Internal, inward-looking

Culture type Bag of words Control capab*, collectiv*, commit*, competenc*, conflict*, consens*, control*, coordin*, cultur*, decentr*, employ*, empower*, engag*, expectat*, facilitator*, hir*, interpers*, involv*, life*, long-term*, loyal*, mentor*, monit*, mutual*, norm*, parent*, partic*, procedur*, productiv*, retain*, reten*, skill*, social*,tension*, value*
 Compete achiev*, acqui*, aggress*, agreem*, attack*, budget*, challeng*, charg*, client*, compet*, customer*, deliver*, direct*, driv*, excellen*, expand*, fast*, goal*, growth*, hard*, invest*, market*, mov*,

• utsourc*, performanc*, position*, pressur*, profit*, rapid*,

reputation, result*, revenue*, satisf*, scan*, succes* signal*, speed*, strong, superior, target*, win* Collaborate boss*, burocr*, cautio*, cohes*, certain*, chief*, collab*, conservat*, cooperat*, detail*, document*, efficien*, error*, fail*, help*, human*, inform*, logic*, method*, outcom*, partner*, people*, predictab*, relation*, qualit*, regular*, solv*, share*, standard*, team*, teamwork*, train*, uniform*, work group*
 Create adapt*, begin*, chang*, creat*, discontin*, dream*, elabor*, entrepre*, envis*, experim*, fantas*, freedom*, futur*, idea*, init*, innovat*, intellec*, learn*, new*, origin*, pioneer*, predict*, radic*, risk*, start*, thought*, trend*, unafra*, ventur*, vision*

Bag of f words

Dominating ti time-invariant corporate cult lture

In 2002, JPMorgan has 1145 (=3.11%) Compete-related, 451 (=0.75%) Create- related, 811 (=1.11%) Control-related and 433 (=0.25%) Collaborate-related words.

• These scores are meaningless
• Subjected to company-specific or industry-specific variations in 2002
• Or simply, what does a Compete=3.11% mean?
• We convert these scores into measures of “dominating culture” by comparing

with peer firms

• JPMorgan is said to have a compete-dominated culture if its Compete

score =3.11% is in the top 25% score of all banks in 2002.

• Although rare, a bank can have more than one or zero dominating culture.

Data and sample

Data

• All loans made by US lenders to US borrowers 1993-2007 from Dealscan
• Hand-clean lenders and match to Calls Report Data for bank characteristics
• Borrower data are collected from Compustat

Final sample:

• 30,000+ loan-level observations and 600+ bank-level observations

Empirical specification

Pr(risky lendingit) = α0 + α1compete-dominanti + α2create-dominanti + α3control-dominanti + α4collaborate-dominanti + controls + εitk

• Risky lending = Equals 1 if borrower is unrated or has a BB+ and below credit

ratings

• Controls: Bank, Borrower, Loan, Relationship lending, Same-culture indicator
• Fixed-effects: Year, Industry, State, Loan Purpose, Loan Type

Risky le lending

Dependent variable: Dummy equals 1 for unrated borrowers or those rated BB+ or worse (1) (2) (3) (4) (5) Compete-dominant 0.076*** 0.078*** (2.630) (2.638) Create-dominant

• 0.008

0.014 (-0.345) (0.626) Control-dominant

• 0.174**
• 0.171**

(-2.131) (-2.040) Collaborate-dominant 0.017 0.048 (0.454) (1.217)

[control variables omitted from tables]

Year dummies Yes Yes Yes Yes Yes Borrower sic-2 dummies Yes Yes Yes Yes Yes Borrower state dummies Yes Yes Yes Yes Yes Observations 38,875 38,875 38,875 38,875 38,875 Pseudo- R2 0.491 0.491 0.491 0.491 0.491

Risky le lending

Dependent variable: Dummy equals 1 for unrated borrowers or those rated BB+ or worse (1) (2) (3) (4) (5) Compete-dominant 0.076*** 0.078*** (2.630) (2.638) Create-dominant

• 0.008

0.014 (-0.345) (0.626) Control-dominant

• 0.174**
• 0.171**

(-2.131) (-2.040) Collaborate-dominant 0.017 0.048 (0.454) (1.217)

[control variables omitted from tables]

Year dummies Yes Yes Yes Yes Yes Borrower sic-2 dummies Yes Yes Yes Yes Yes Borrower state dummies Yes Yes Yes Yes Yes Observations 38,875 38,875 38,875 38,875 38,875 Pseudo- R2 0.491 0.491 0.491 0.491 0.491

Robustness

• Alternative cut-off points: results vary consistently with the notion of

“dominating culture”

• Excluding top 10 (“TBTF”) lenders
• Sample restricted to rated-only borrowers
• Remove years immediately after mergers
• Alternative standard-error clustering
• Additional location and loan controls

Endogenous le lender-borrower matching

• Borrowers do not randomly choose lenders
• We indeed find that borrower-lender has a corporate culture match, e.g.,

compete borrowers are more likely to obtain loans from compete banks

• This is a part of rather an alternative interpretation.
• That riskier borrowers approach riskier lenders suggest they are aware of

the “reckless reputation” of the banks

• There are other endogeneity concerns

Id Identifi fication: Russian Default Crisis

• An exogenous shock that produces a short-term negative sentiment to the

competitive culture of US banks and thus affects its lending behaviour.

• Plausibly exogenous to omitted factors such as credit demands of US

borrowers or borrower-lender matching.

• Event: The Russian Government defaulted their sovereign debt obligations on

17th August 1998, causing losses to exposed US banks, putting the entire US banking industry under distress.

• Chava and Purnanandam (2011) show that borrowers whose lenders affected

by the Russian crisis cut CAPEX and suffer valuation loss.

Id Identifi fication: Russian Default Crisis

(1) (2) (3) (4) 3-month 4-month 5-month Placebo event 1 y prior to August 1 Compete-dominant * Post Russian default

• 0.085**
• 0.087**
• 0.086**

0.004 (-2.204) (-2.091) (-2.124) (0.097) Create-dominant * Post Russian default

• 0.016
• 0.027
• 0.030
• 0.062

(-0.346) (-0.619) (-0.685) (-1.505) Control-dominant * Post Russian default

• 0.066
• 0.091
• 0.089

0.101 (-0.498) (-0.692) (-0.711) (0.635) Collaborate-dominant * Post Russian default 0.041 0.046 0.054 0.005 (0.699) (0.828) (1.050) (0.080)

[control variables omitted from tables]

Year dummies Yes Yes Yes Yes Borrower sic-2 dummies Yes Yes Yes Yes Borrower state dummies Yes Yes Yes Yes Observations 1,198 1,582 2,003 54,195 Pseudo- R2 0.703 0.641 0.604 0.491

Lo Loan terms

Dependent Variable = Covenants Dependent Variable = Ln(Spread) Full sample Sub- investment grade Investment Full sample Sub-investment grade Investment (1) (2) (3) (4) (5) (6) Compete-dominant

• 0.041**
• 0.055***

0.021 0.045*** 0.029***

• 0.002

(-2.513) (-2.777) (1.097) (3.991) (2.916) (-0.117) Create-dominant

• 0.096***
• 0.106***
• 0.001

0.047*** 0.053***

• 0.020

(-6.215) (-5.514) (-0.085) (4.617) (5.430) (-1.617) Control-dominant 0.208*** 0.224*** 0.094* 0.011 0.104**

• 0.096*

(3.319) (2.719) (1.782) (0.248) (2.369) (-1.695) Collaborate-dominant 0.060*** 0.066*** 0.009 0.012

• 0.004

0.029 (3.032) (2.919) (0.328) (0.985) (-0.394) (1.574)

[control variables omitted from tables]

Year dummies Yes Yes Yes Yes Yes Yes Borrower sic-2 dummies Yes Yes Yes Yes Yes Yes Borrower state dummies Yes Yes Yes Yes Yes Yes Observations 30,877 23,080 7,797 30,877 23,080 7,797 Pseudo- R2 0.376 0.266 0.431 0.491 0.529 0.556

Lo Loan terms

Dependent Variable = Covenants Dependent Variable = Ln(Spread) Full sample Sub- investment grade Investment Full sample Sub-investment grade Investment (1) (2) (3) (4) (5) (6) Compete-dominant

• 0.041**
• 0.055***

0.021 0.045*** 0.029***

• 0.002

(-2.513) (-2.777) (1.097) (3.991) (2.916) (-0.117) Create-dominant

• 0.096***
• 0.106***
• 0.001

0.047*** 0.053***

• 0.020

(-6.215) (-5.514) (-0.085) (4.617) (5.430) (-1.617) Control-dominant 0.208*** 0.224*** 0.094* 0.011 0.104**

• 0.096*

(3.319) (2.719) (1.782) (0.248) (2.369) (-1.695) Collaborate-dominant 0.060*** 0.066*** 0.009 0.012

• 0.004

0.029 (3.032) (2.919) (0.328) (0.985) (-0.394) (1.574)

[control variables omitted from tables]

Year dummies Yes Yes Yes Yes Yes Yes Borrower sic-2 dummies Yes Yes Yes Yes Yes Yes Borrower state dummies Yes Yes Yes Yes Yes Yes Observations 30,877 23,080 7,797 30,877 23,080 7,797 Pseudo- R2 0.376 0.266 0.431 0.491 0.529 0.556

Lo Loan terms

Dependent Variable = Covenants Dependent Variable = Ln(Spread) Full sample Sub- investment grade Investment Full sample Sub-investment grade Investment (1) (2) (3) (4) (5) (6) Compete-dominant

• 0.041**
• 0.055***

0.021 0.045*** 0.029***

• 0.002

(-2.513) (-2.777) (1.097) (3.991) (2.916) (-0.117) Create-dominant

• 0.096***
• 0.106***
• 0.001

0.047*** 0.053***

• 0.020

(-6.215) (-5.514) (-0.085) (4.617) (5.430) (-1.617) Control-dominant 0.208*** 0.224*** 0.094* 0.011 0.104**

• 0.096*

(3.319) (2.719) (1.782) (0.248) (2.369) (-1.695) Collaborate-dominant 0.060*** 0.066*** 0.009 0.012

• 0.004

0.029 (3.032) (2.919) (0.328) (0.985) (-0.394) (1.574)

[control variables omitted from tables]

Year dummies Yes Yes Yes Yes Yes Yes Borrower sic-2 dummies Yes Yes Yes Yes Yes Yes Borrower state dummies Yes Yes Yes Yes Yes Yes Observations 30,877 23,080 7,797 30,877 23,080 7,797 Pseudo- R2 0.376 0.266 0.431 0.491 0.529 0.556

Lo Loan terms

Dependent Variable = Covenants Dependent Variable = Ln(Spread) Full sample Sub- investment grade Investment Full sample Sub-investment grade Investment (1) (2) (3) (4) (5) (6) Compete-dominant

• 0.041**
• 0.055***

0.021 0.045*** 0.029***

• 0.002

(-2.513) (-2.777) (1.097) (3.991) (2.916) (-0.117) Create-dominant

• 0.096***
• 0.106***
• 0.001

0.047*** 0.053***

• 0.020

(-6.215) (-5.514) (-0.085) (4.617) (5.430) (-1.617) Control-dominant 0.208*** 0.224*** 0.094* 0.011 0.104**

• 0.096*

(3.319) (2.719) (1.782) (0.248) (2.369) (-1.695) Collaborate-dominant 0.060*** 0.066*** 0.009 0.012

• 0.004

0.029 (3.032) (2.919) (0.328) (0.985) (-0.394) (1.574)

[control variables omitted from tables]

Year dummies Yes Yes Yes Yes Yes Yes Borrower sic-2 dummies Yes Yes Yes Yes Yes Yes Borrower state dummies Yes Yes Yes Yes Yes Yes Observations 30,877 23,080 7,797 30,877 23,080 7,797 Pseudo- R2 0.376 0.266 0.431 0.491 0.529 0.556

Bank-level outcomes

• Banks with compete-oriented culture makes risky lending. So what?
• Through the lending channel, bank culture influences:
• Performance and risk of individual banks
• Bank’s contribution to systemic risk

Financial leaders must take values as seriously as valuation, culture as seriously as

• capital. This makes abundant sense to me—culture and capital each promote

financial stability Christine Lagarde

Lo Loan performance

Dependent variable: Loan growth (1) (2) (3) (4) (5) Compete-dominant 0.025*** 0.027*** (2.675) (2.661) Create-dominant 0.005 0.011 (0.336) (0.701) Control-dominant

• 0.015
• 0.004

(-1.134) (-0.311) Collaborate-dominant 0.003 0.008 (0.433) (0.908) [control variables omitted from tables] Year dummies Yes Yes Yes Yes Yes Observations 317 317 317 317 317

Bad lo loans

Dependent variable: Non-performing loans/Total assets (1) (2) (3) (4) (5) Compete-dominant 0.927** 0.892** (2.344) (2.058) Create-dominant 1.436 1.775* (1.457) (1.785) Control-dominant

• 1.560***
• 1.234**

(-3.244) (-2.264) Collaborate-dominant 0.329 0.523 (0.712) (1.118) [control variables omitted from tables] Year dummies Yes Yes Yes Yes Yes Observations 317 317 317 317 317

Bad lo loans

Dependent variable: Non-performing loans/Total assets (1) (2) (3) (4) (5) Compete-dominant 0.927** 0.892** (2.344) (2.058) Create-dominant 1.436 1.775* (1.457) (1.785) Control-dominant

• 1.560***
• 1.234**

(-3.244) (-2.264) Collaborate-dominant 0.329 0.523 (0.712) (1.118) [control variables omitted from tables] Year dummies Yes Yes Yes Yes Yes Observations 317 317 317 317 317

Systemic ri risk

Dependent variable = CoVar (x100) (1) (2) (3) (4) (5) Compete-dominant

• 0.182***
• 0.164***

(-3.287) (-2.830) Create-dominant 0.025

• 0.003

(0.345) (-0.033) Control-dominant 0.202*** 0.160** (3.220) (2.347) Collaborate-dominant 0.003

• 0.002

(0.075) (-0.050) [control variables omitted from tables] Year dummies Yes Yes Yes Yes Yes Observations 317 317 317 317 317

Systemic ri risk

Dependent variable = CoVar (x100) (1) (2) (3) (4) (5) Compete-dominant

• 0.182***
• 0.164***

(-3.287) (-2.830) Create-dominant 0.025

• 0.003

(0.345) (-0.033) Control-dominant 0.202*** 0.160** (3.220) (2.347) Collaborate-dominant 0.003

• 0.002

(0.075) (-0.050) [control variables omitted from tables] Year dummies Yes Yes Yes Yes Yes Observations 317 317 317 317 317

Co Consequences of f risk risky cult lture are not t im immedia iately ly reali lized

Lending growth Bad loans Risky lending Distress=0 Distress=1 Distress= 0 Distress=1 Distress=0 Distress=1 (1) (2) (3) (4) (5) (6) Compete-dominant 0.024** 0.028 0.104 1.207* 0.110*** 0.054 (1.988) (1.655) (0.173) (1.749) (2.839) (1.541) Create-dominant

• 0.002

0.033 6.155*** 0.130 0.037 0.000 (-0.100) (1.399) (4.965) (0.130) (1.238) (0.011) Control-dominant

• 0.025

0.021

• 1.666**
• 2.217***
• 0.081
• 0.282***

(-1.316) (0.983) (-2.438) (-2.790) (-0.658) (-2.664) Collaborate-dominant

• 0.000

0.019 0.460 1.037 0.009 0.070 (-0.014) (1.447) (0.785) (1.336) (0.218) (1.253) Other controls Yes Yes Yes Yes Yes Yes Year dummies Yes Yes Yes Yes Yes Yes Borrower sic-2 dummies Yes Yes Yes Yes Yes Yes Borrower state dummies Yes Yes Yes Yes Yes Yes Observations 178 101 178 101 25,857 28,158 Pseudo- R2 0.266 0.431 0.529 0.556 0.497 0.500

Co Consequences of f risk risky cult lture are not t im immedia iately ly reali lized

Lending growth Bad loans Risky lending Distress=0 Distress=1 Distress= 0 Distress=1 Distress=0 Distress=1 (1) (2) (3) (4) (5) (6) Compete-dominant 0.024** 0.028 0.104 1.207* 0.110*** 0.054 (1.988) (1.655) (0.173) (1.749) (2.839) (1.541) Create-dominant

• 0.002

0.033 6.155*** 0.130 0.037 0.000 (-0.100) (1.399) (4.965) (0.130) (1.238) (0.011) Control-dominant

• 0.025

0.021

• 1.666**
• 2.217***
• 0.081
• 0.282***

(-1.316) (0.983) (-2.438) (-2.790) (-0.658) (-2.664) Collaborate-dominant

• 0.000

0.019 0.460 1.037 0.009 0.070 (-0.014) (1.447) (0.785) (1.336) (0.218) (1.253) Other controls Yes Yes Yes Yes Yes Yes Year dummies Yes Yes Yes Yes Yes Yes Borrower sic-2 dummies Yes Yes Yes Yes Yes Yes Borrower state dummies Yes Yes Yes Yes Yes Yes Observations 178 101 178 101 25,857 28,158 Pseudo- R2 0.266 0.431 0.529 0.556 0.497 0.500

Bu But, t, compete banks make sile silent contr trib ibutio ion to system ris risk

CoVar (x100) Distress=0 Distress=1 (1) (2) Compete-dominant

• 0.247***
• 0.114

(-3.411) (-1.298) Create-dominant 0.046 0.023 (0.452) (0.186) Control-dominant 0.009 0.407*** (0.106) (4.364) Collaborate-dominant

• 0.003
• 0.004

(-0.058) (-0.046) Other controls Yes Yes Year dummies Yes Yes Borrower sic-2 dummies Yes Yes Borrower state dummies Yes Yes Observations 178 101 Pseudo- R2 0.266 0.431

• Compete-banks make more contributions to systemic risk during normal

times, and not during distress.

• Slow-exploding bombs!

Corporate cult lture or business models?

• Corporate culture
• Soft: Dressing, ethos, body language, unwritten agreements
• Business model: Loans/Assets, Deposits/Assets, Interest/Non-interest

Incomes, soft-compete=residuals of the regressions of three basic business model variables: Loans/Assets, Deposits/Assets, Interest/Non-interest Incomes on compete.

• Results are robust

What underlies the ‘soft culture’?

• Appointment-Compensation-Retention Model
• Three mechanisms
• Compete-oriented banks recruit more reckless executives
• Compete-oriented banks pay executives with more high-incentivised

compensation packages

• Compete-oriented banks more like to fire execs following poor

performance

Thank you.