1 the income elasticity of loan demand manthos d. delis, surrey business school iftekhar hasan,...

29
w w w .surrey.ac.uk/sbs w w w .surrey.ac.uk/sbs 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University of Piraeus Presentation prepared for the Academic Seminars Series 2014, Department of Banking and Financial Management, University of Piraeus

Upload: dontae-stafford

Post on 15-Dec-2015

214 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 1

The Income Elasticity of Loan Demand

Manthos D. Delis, Surrey Business School

Iftekhar Hasan, Fordham University and Bank of Finland

Chris Tsoumas, University of Piraeus

Presentation prepared for the Academic Seminars Series 2014, Department of Banking and Financial Management,

University of Piraeus

Page 2: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 2

Agenda

• Main questions• Theoretical and empirical background• Data• Empirical methodology• Results• Conclusions

Page 3: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 3

Main questions

• Q1: ◦ How does loan demand change with people’s incomes?

• Q2:◦ Is the income elasticity of loan demand stable over time?

• Answers have important implications for:◦ The origination of banking and financial crises◦ How policy should respond to fluctuations in loan demand

Page 4: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 4

Theoretical and empirical background• Primary reasons for the subprime crisis

(e.g., Mian and Sufi, 2009)◦ Credit expansion ◦ Accompanying decline in credit standards

Supply side forces

• Sources behind this developmentoEasy monetary policy increased incentives for banks to take on

higher risks in search of yield

(Jimenez, Ongena, Peydro, and Saurina, 2013) oGovernment policy toward subprime mortgage-credit expansion

(Rajan, 2010; Mian, Sufi, and Trebbi, 2014)

Equally supporting evidence:

(Keys, Mukherjee, Seru, and Vig, 2010; Demanyak and Van Hemert, 2011)

Page 5: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 5

Theoretical and empirical background

Rajan (2010): ◦ Provides anecdotal evidence for the role of household

income as a determinant of loan demand and a main source behind the subprime crisisRising inequality: the 90-50 hourly wage differential,

not the 50-10◦ Argues that the political response to this in the late 1990s and

early 2000s was to increase the availability of mortgage credit

“Bankers responded to implicit and explicit incentives that the system created”

Demand and supply side forces

Page 6: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 6

Theoretical and empirical background• Demand side forces may translate to:

1. Increased demand for credit,

2. Demand for riskier loans (ex ante riskier borrowers), and/or

3. Demand for larger credit amounts relative to income

Point 3 spans both points 1 and 2

Only Dell’Ariccia, Igan, and Laeven (2012) provide evidence for a demand side explanation of the crisis ◦ Exogenous increase in the demand for subprime credit

partially triggered the lower denial rates in specific areas of the U.S.

They touch upon points 1 and 2

We focus on point 3

Page 7: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 7

Anecdotal evidenceAverage income and loan demand (in nominal values and logs) for the top 10% of the income distribution, the 75%-90% income cohort, and the 25%-75% income group

3.5

4.0

4.5

5.0

5.5

6.0

1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

loan (90+) income (90+) loan(75-90)

income(75-90) loan(25-75) income(25-75)

The rich demand approximately the same amount of loans with their annual

income

The gap between loan demand and income for the 75%-90% income group is

considerable, and the gap for the 25%-75% group is even larger

There is almost a 1:1 increase in loan demand and income in 1992-2012,

quite similar for the 75%-90% and the 25%-75% income groups

Decreasing gap between income and the loan amounts demanded for the latter two

income groups for 2002 to 2006

Page 8: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 8

Theoretical and empirical background

• We are interested both in:

◦ The absolute differences in loan demand due to differences in income across income groups and how they evolve over time

◦ In the relative changes in loan demand due to differences in income between income groups Relates to the relative-income hypothesis

(Duesenberry, 1949) (“keep up with the Joneses” effect)

Page 9: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 9

Data

• Home Mortgage Disclosure Act (HMDA) database ◦ Applicant-level data on all mortgage applications

• 1992-2012• Characteristics for each application

◦ Loan information▪Requested loan amount▪Type of loan (i.e., conventional, insured by the FHA, guaranteed by the Department of Veterans Affairs or Farm Service Agency/Rural Housing Service)

▪Property type (one to four-family, manufactured housing, multifamily)

▪Purpose of the loan (i.e., home purchase, home improvement, or refinancing)

▪Owner’s occupancy status (e.g., house will be owner-occupied as a principal dwelling)

▪Loan status (originated, denied, withdrawn application, reason for denial etc.)

Page 10: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 10

Data

◦ Applicant information▪ income, sex, gender and race; co-applicant's sex, gender and race

◦ Property location information▪MSA, state, county, and census-tract codes of the property

◦ Financial institution information▪ Identity; Supervisory agency

◦ Census info for property location▪Population; Minority population; Median family income; Tract-to-MSA median family income percentage; Number of owner occupied units; Number of 1- to 4-family units

• Supplemented with data for◦ Personal income and employment summary (MSA level) (Bureau

of Economic Analysis)◦ Median income estimates and percent of county population below

the poverty line (county level) (SAIPE)◦ All-transactions house-price indices (MSA level) (FHFA)◦ Annual list of subprime lenders from HUD (Dell’ Ariccia et al.,

2012)

Page 11: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 11

Data

• Use of Both accepted and denied loans

▪First step for distinguishing between supply and demand Only conventional loans for home purchase

▪To avoid distortions from loan guarantees and/or the presence of past mortgages

• Data cleaned from missing info for variables of interest

• Creation of randomized samples for◦ 1992-2012 period (“stacked” sample) (pseudo panel)

▪657,292 obs. (the 1% of each year’s applications)

◦ Each year (“cross-sectional samples”)▪~ 300,000 obs. each

Page 12: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 12

0

5,000,000

10,000,000

15,000,000

20,000,000

25,000,000

30,000,000

35,000,000

40,000,000

45,000,000

1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

All mortgage applications

Conventional type loans for home purchase

Conventional type for home purchase - No missing data

Page 13: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 13

Empirical analysis• Estimation of a loan-demand equation

: (log) quantity of loan demanded for application i

: offered lending rate

: (log) income of the borrower

: k-dimension vector of observed applicant’s characteristics

(i.e., sex, race, occupancy status) The coefficient of is the income elasticity of loan demand• Estimation for six different income brackets

(90+ (top 10%); 70%-90%; 50%-70%; 30%-50%; 10%-30%;

10- (bottom 10%))• Estimation using “stacked” and annual “cross-sectional”

samples

Page 14: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 14

1st Identification challenge

Offered lending rate is unobserved until the loan is originated

Available only for a small fraction of originated loans and only after 2004

Solution proposed: Assume that applicants observe the same lending rates

within a particular census tract, given that we estimate equation (1) separately for each income group.

Thus, include census tract dummies to control, inter alia, for

The error term structure becomes For cross-sectional samples estimation:

For stacked sample estimation:

Page 15: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 15

Results with census tract fixed effects – Elasticities

0.0

0.2

0.4

0.6

0.8

1.0

1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

90+ 70-90 50-70 30-50 10-30

The middle class demands larger amounts of loans relative to its

income

Loan demand gradually shrinks for middle-income and especially lower-middle-income people

relative to their income during the 2000sConvergence towards the rich until 2006

Findings in line with Mian and Sufi (2009)who show that increases in subprime borrowers

incomes do not justify larger loan originations

Page 16: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 16

2nd identification challenge

Distinguishing loan supply from loan demandBanks may influence borrowers’ decisions through their

marketing strategies and/or their loan termsThus, supply side (bank) unobserved effects

Solution proposed

1. Use of all applications – not only originated loans, and

2. Add bank fixed effects

Page 17: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 17

Results with bank and census tract fixed effects - Elasticities

0.0

0.2

0.4

0.6

0.8

1.0

1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

90+ 70-90 50-70 30-50 10-30

Same pictureElasticities about 10% smaller

Thus, it seems that the system became riskier not because of increased loan demand

of individual applicant, but…

Page 18: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 18

3rd identification challenge Potential endogeneity of due to unobservable

characteristics of individuals affecting loan demand Wealth, age, education, expected income in future years,

other

Solution proposed:

1. Grouping individuals based on a large number of their observed characteristics (income bracket, sex, race, co-applicant’s sex and race, occupancy, loan status, reason for denial (if present), bank, census tract) Large number of singleton groups (about 65% to 85% each

year)

Page 19: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 19

3rd identification challenge

2. Keep only groups with two “identical” individuals

3. Calculate differences on (log) loan demand and (log) income Aim to difference out any unobservable characteristics

4. Estimate equation:

Page 20: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 20

Results when individuals are grouped based on observed characteristics

0.0

0.2

0.4

0.6

0.8

1.0

1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

90+ 70-90 50-70 30-50 10-30

Same picture once moreUpward trend after 2007,

more intense for the 70%-90% and 30%-50% income brackets

Page 21: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 21

Results when individuals are grouped based on observed characteristics – Obs.

0

50,000

100,000

150,000

200,000

250,000

300,000

350,000

1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

90+ 70-90 50-70 30-50 10-30

Fast accumulating burden on the systembecause of a very large number of loans

Page 22: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 22

A "Keeping up with the Joneses" Effect?

• Two possible forms of the effect (non-mutually exclusive)

1. Individuals demand bigger loans relative to their incomes so as to keep up with their richer neighbors

“first order” effect

2. Individuals apply for loans to keep up with their neighbors whereas they would not apply at all otherwise

“second order” effect

• We examine the “first order” effect

Page 23: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 23

A "Keeping up with the Joneses" Effect?

Solution proposed:1. Take averages at year t-1 by income bracket and census tract

for the applicant’s loan amount and income2. For each application at year t, calculate the difference

between the loan amount and income for applicant i minus the average loan amount and income, respectively, in the income bracket just above that of i in the same census tract at year t-1.

Aim to identify mimicking behavior among loan applicants stemming from the observable loan decisions of their richer neighbors one year earlier.

Higher value on the differenced coefficient of income supports an enhanced “keeping up with the Joneses” effect.o Bank and census tract dummies included

Page 24: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 24

Results – "Keeping up with the Joneses" Effect

0.0

0.2

0.4

0.6

0.8

1.0

1993 1995 1997 1999 2001 2003 2005 2007 2009 2011

70-90 50-70 30-50 10-30

In general, not definitive support for “first-order “ “keeping up with the Joneses” effect

• Roughly similar behavior of loan-demand elasticities for the middle-class income brackets

with that in previous specifications• Exception for the 70%-90% bracket

(quite lower, more unstable) • Short-term upward spikes for the 50%-70% and

30%-50% income groups

Page 25: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 25

Sensitivity analysesA. Results driven by subprime lending applications?Solution proposed:

1. Following Dell’Ariccia et al. (2012), use the annual list of subprime lenders to identify financial institutions that specialize in subprime mortgage lending for 1993-2005

2. Repeat the analysis with bank and census tract dummies

0.0

0.2

0.4

0.6

0.8

1.0

1993 1995 1997 1999 2001 2003 2005

70-90 50-70 30-50 10-30

Page 26: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 26

Sensitivity analyses

B. Results driven by local inequality conditions? Solution proposed:

1. Only the counties at the top or bottom 20% centile of the poverty line measure, i.e., the percent of county population that lie below the poverty line, estimated by the Census Bureau, are included in each year.

0.0

0.2

0.4

0.6

0.8

1.0

1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

90+ 70-90 50-70 30-50 10-30

Higher poverty rate counties

0.0

0.2

0.4

0.6

0.8

1.0

1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

90+ 70-90 50-70 30-50 10-30

Lower poverty rate counties

Page 27: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 27

Sensitivity analyses

• C. Are the elasticities different or the originated loans? Solution proposed:◦ Replicate the analysis with grouped individuals based on

observed characteristics, but only for loans that were actually originated

0.0

0.2

0.4

0.6

0.8

1.0

1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

90+ 70-90 50-70 30-50 10-30

Page 28: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 28

Conclusions• Income elasticity of mortgage demand for all middle-class

income groups is:◦ Higher than that of the rich◦ Falling and converging with that of the rich before the

eruption of the subprime crisis• Thus, proponents of income inequality idea are right, but:

◦ Middle-class income levels affect the loan amounts less and less before the crisis

◦ Impact through the number of applications Behavioral decisions of households can be systemically risky

at the aggregate level even if they could be prudent from a microeconomic perspective

• Implications◦ “Systemic view” of loan policies from financial institutions◦ Foreclosure policies

Page 29: 1 The Income Elasticity of Loan Demand Manthos D. Delis, Surrey Business School Iftekhar Hasan, Fordham University and Bank of Finland Chris Tsoumas, University

www.surrey.ac.uk/sbs 29

Possible caveats and extensions• But…

◦ What if income data are not quite reliable for the period before the crisis due to the advent of ‘low doc/no doc’ loans?

Solution proposed (to be conducted):▪Use aggregate data on income (census data) and loans at the census tract level

▪Other?

• Does the economic structure of counties/MSAs affects the results through its impact on expected income?

• Other?