Credit Growth and the Financial Crisis: A New Narrative

Credit Growth and the Financial Crisis: A New Narrative Stefania Albanesi, University of Pittsburgh Giacomo De Giorgi, University of Geneva Jaromir Nosal, Boston College Fifth Conference on Household Finance and Consumption Banque du France, Paris December 14-15, 2017

Introduction - Prevailing narrative about the financial crisis: credit growth during boom concentrated in subprime segment defaults during financial crisis also concentrated in this segment expansion of subprime credit leading cause for the crisis - Mechanism: mortgage defaults drop in house prices contraction in credit for high MPC households drop in consumption and employment (Lorenzoni & Guerreri 2015, Midrigan & Philippon 2016, Justiniano & al. 2016, Berger & al. 2015, Kaplan, Mittman &Violante 2017, Hedlund & Garriga 2016, etc.)

Our Contribution - Study household debt and delinquency in 1999-2013: based on large administrative panel of credit report data

Our Contribution - Study household debt and delinquency in 1999-2013: based on large administrative panel of credit report data Findings:

Our Contribution - Study household debt and delinquency in 1999-2013: based on large administrative panel of credit report data Findings: I. Credit growth during boom primarily for mid-high credit score borrowers (consistent with Adelino, Shoar & Severino 2015, Ferreira & Guyourko 2015 and Foote, Loewenstein & Willen 2016 ) II. Larger rise in defaults for mid-high credit score borrowers during crisis III. High credit score defaults driven by real estate investors

Data - FRBNY Consumer Credit Panel/Equifax Data 1% of all individuals with an Equifax credit report (2.5 mil borrowers per quarter) - Information quarterly, 1999:Q1-2013:Q4 all consumer debt except pay day loans delinquent behavior public record items credit score, age, ZIP code matched to payroll data for 2009

Prevailing Narrative - Initial credit score used to assess borrower quality (Mian&Sufi 2009 and 2017)

Prevailing Narrative - Initial credit score used to assess borrower quality (Mian&Sufi 2009 and 2017) Individuals by Initial Credit Score 3 2.5 2 1.5 1 0.5 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Real per capita real mortgage balances, ratio to 2001Q3. (FRBNY CCP/Equifax Data.)

Prevailing Narrative - Initial credit score used to assess borrower quality (Mian&Sufi 2009 and 2017) Individuals by Initial Credit Score 3 Zip Codes by Initial Subprime Share 2.5 2.5 2 2 1.5 1.5 1 0.5 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 1 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Real per capita real mortgage balances, ratio to 2001Q3. (FRBNY CCP/Equifax Data.)

Prevailing Narrative - Initial credit score used to assess borrower quality (Mian&Sufi 2009 and 2017) Stronger mortgage debt growth for subprime borrowers Individuals by Initial Credit Score 3 Zip Codes by Initial Subprime Share 2.5 2.5 2 2 1.5 1.5 1 0.5 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 1 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Real per capita real mortgage balances, ratio to 2001Q3. (FRBNY CCP/Equifax Data.)

Problems with Initial Credit Score Ranking - Low credit score borrowers disproportionately young Median Age Quartile 1: 39 Quartile 2: 44 Quartile 3: 48 Quartile 4: 58 0.035 0.03 0.025 0.02 0.015 0.01 0.005 0 20 25 30 35 40 45 50 55 60 65 70 75 80 85 Age distribution by credit score quartile, 2004-2012 average. (Experian Data.)

Problems with Initial Credit Score Ranking - Low credit score borrowers disproportionately young - Young experience life cycle debt and credit score growth Credit Score Debt 160 $40,000 140 120 100 $30,000 $20,000 $10,000 Total Debt Balances Mortgage Balances 80 60 40 20 0 21 26 31 36 41 46 51 56 61 66 71 76 81-20 $0 21 26 31 36 41 46 51 56 61 66 71 76 81 -$10,000 -$20,000 -$30,000 -$40,000 -$50,000 Estimated age effects. (FRBNY CCP/Equifax Data.)

Problems with Initial Credit Score Ranking - Low credit score borrowers disproportionately young - Young experience life cycle debt and credit score growth Initial credit score lower than at time of borrowing

Life Cycle Credit Scores, Debt and Income - Credit score and debt growth for young in 1999 rise with 2009 income 25-34 year olds in 1999 by income quintile in 2009 Credit Score Mortgage Balances Difference from 2001 0 15 30 45 60 75 Quintile 1 (Lowest) Quintile 5 (Highest) 0 15 30 45 60 75 Ratio to 2001 1 1.5 2 2.5 3 Quintile 1 (Lowest) Quintile 5 (Highest) 1 1.5 2 2.5 3 2001 2003 2005 2007 2009 2001 2003 2005 2007 2009 Difference with 2001 (credit score) and ratio to 2001 (mortgage balances). (FRBNY CCP/Equifax Data.)

Life Cycle and Borrowing by Initial Credit Score I. Removing differences in age distribution Individuals by Initial Credit Score 3 Age Distribution Set to Quartile 4 3 2.5 2.5 2 2 1.5 1.5 1 1 0.5 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 0.5 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Real per capita mortgage balances by 1999 Equifax Risk Score, ratio to 2001. (FRBNY CCP/Equifax Data.)

Life Cycle and Borrowing by Initial Credit Score I. Removing differences in age distribution Differences in debt growth across initial credit scores attenuated Per Capita 2001Q3-2007Q4 Real Mortgage Balance Growth Difference with Quartile 4 Explained by Age Distribution Quartile 1 Quartile 2 Quartile 3 25% 20% 14% Borrowers ranked by 1999 Equifax Risk Score. (FRBNY CCP/Equifax Data.)

Life Cycle and Borrowing by Initial Credit Score II. Removing life cycle effects Individuals by Initial Credit Score 3 3 Life Cycle Effects Removed 2.5 2.5 2 2 1.5 1.5 1 1 0.5 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 0.5 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Real per capita mortgage balances by 1999 Equifax Risk Score, ratio to 2001. Life cycle effects removed by assigning to each 1999 age bin balances of borrowers in that age bin in current quarter. (FRBNY CCP/Equifax Data.)

Life Cycle and Borrowing by Initial Credit Score II. Removing life cycle effects Differences in debt growth by initial credit score mostly eliminated Individuals by Initial Credit Score 3 3 Life Cycle Effects Removed 2.5 2.5 2 2 1.5 1.5 1 1 0.5 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 0.5 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Real per capita mortgage balances by 1999 Equifax Risk Score, ratio to 2001. Life cycle effects removed by assigning to each 1999 age bin balances of borrowers in that age bin in current quarter. (FRBNY CCP/Equifax Data.)

Credit Scores, Debt and Defaults - Alternative to initial credit score? recent credit score

Credit Scores, Debt and Defaults - Alternative to initial credit score? recent credit score Strongly positively related to income, given age 850 800 750 700 650 600 550 5,000 20,000 35,000 50,000 65,000 80,000 95,000 110,000 125,000 140,000 155,000 170,000 185,000 200,000 215,000 25 30 35 40 45 50 55 60 65 Predicted relation between credit score and total labor income by age in 2009. (FRBNY CCP/Equifax Data.) 230,000 245,000

Debt and Defaults by Recent Credit Score - Analysis from lender s perspective

Debt and Defaults by Recent Credit Score - Analysis from lender s perspective Regression Specification Dependent variable: future change in balances (4-12 quarter ahead)

Debt and Defaults by Recent Credit Score - Analysis from lender s perspective Regression Specification Dependent variable: future change in balances (4-12 quarter ahead) Explanatory variables: 1 quarter lagged credit score quartile lagged change in credit score (4-8 quarter change) time effects, age effects time and age effects interacted with 1 quarter lagged credit score

Debt by Recent Credit Score: Mortgage Balances - Growth strongest for quartiles 2-3 during boom Predicted 8 quarter ahead change in mortgage balances 20,000 15,000 10,000 5,000 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011-5,000-10,000 Age adjusted, by 1Q lagged Equifax Risk Score quartile, USD. (FRBNY CCP/Equifax Data.)

Debt by Recent Credit Score: Mortgage Balances - Sizable estimated age effects only for quartiles 2-4 Age effects for 8 quarter ahead change in mortgage balances 25,000 20,000 15,000 10,000 5,000 0-5,000 20 25 30 35 40 45 50 55 60 65 70 75 80 By 1Q lagged Equifax Risk Score quartile, USD. (FRBNY CCP/Equifax Data.)

Credit Growth by Credit Score: More Evidence - No growth in new originations for quartile 1 Fraction with New Originations 0.3 0.2 0.1 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 By 8Q lagged Equifax Risk Score quartile. Quartile cutoffs: 615, 720, 791, 840. (FRBNY CCP/Equifax Data.)

Credit Growth by Credit Score: More Evidence - No growth in new originations for quartile 1 - No growth in fraction with first mortgages for quartile 1 Fraction with First Mortgages 1 0.8 0.6 Quartile 2 Quartile 3 Quartile 4 Quartile 1 1 0.8 0.6 0.4 0.4 0.2 0.2 0 2001 2003 2005 2007 2009 2011 2013 0 By 8Q lagged Equifax Risk Score quartile. Quartile cutoffs: 615, 720, 791, 840. (FRBNY CCP/Equifax Data.)

Defaults by Recent Credit Score: Balances - Delinquent mortgage balances grow most for quartiles 2-4 during crisis Predicted 8 quarter ahead change in delinquent mortgage balances 5,000 3,000 1,000-1,000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011-3,000-5,000-7,000 Age adjusted, 90+ day delinquent, by 1Q lagged Equifax Risk Score quartile, USD. (FRBNY CCP/Equifax Data.)

Defaults by Recent Credit Score - Quartile 1 share of foreclosures drops during crisis Fraction with Share of Fraction (3QMA) 0.002.004.006 Quartile 1 (Lowest) Quartile 2 Quartile 3 Quartile 4 (Highest) 0.002.004.006 Share (3QMA) 0.2.4.6.8 Quartile 1 (Lowest) Quartile 2 Quartile 3 Quartile 4 (Highest) 0.2.4.6.8 2001 2003 2005 2007 2009 2001 2003 2005 2007 2009 2011 2013 Foreclosures in the last 4 quarters by 8 quarter lagged Equifax Risk Score quartile. (FRBNY CCP/Equifax Data)

Explaining High Credit Score Defaults - Why did borrowers with good credit default during crisis? Rise in investors borrowers with 2 or more first mortgages

Explaining High Credit Score Defaults - Why did borrowers with good credit default during crisis? Rise in investors borrowers with 2 or more first mortgages Fraction of Investors 2001Q3-2004Q3 mean 0.063 0.103 0.110 0.107 Investor Share of Mortgage Balances 2001Q3-2004Q3 mean 0.123 0.196 0.212 0.226 By 8 quarter lagged Equifax Risk Score. (FRBNY CCP/Equifax Data.)

Explaining High Credit Score Defaults - Why did borrowers with good credit default during crisis? Rise in investors borrowers with 2 or more first mortgages Fraction of Investors 2001Q3-2004Q3 mean 0.063 0.103 0.110 0.107 2007Q4 peak 0.082 0.156 0.162 0.142 Investor Share of Mortgage Balances 2001Q3-2004Q3 mean 0.123 0.196 0.212 0.226 2007Q4 peak 0.183 0.333 0.350 0.317 By 8 quarter lagged Equifax Risk Score. (FRBNY CCP/Equifax Data.)

High Credit Score Defaults: Role of Investors - Rise in foreclosure rate more pronounced for investors Investors (2+) Non Investors (1) 0.045 0.04 0.035 0.03 0.025 0.02 0.015 0.01 0.005 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 0.045 0.04 0.035 0.03 0.025 0.02 0.015 0.01 0.005 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Foreclosure rate by 8 quarter lagged Equifax Risk Score, 3QMA. (FRBNY CCP/Equifax Data.)

High Credit Score Defaults: Role of Investors - Rise in foreclosure rate more pronounced for investors Rise in investor share of defaults for high credit score borrowers Investor Share of Foreclosures 0.6 0.5 0.4 0.3 0.2 0.1 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 By quartile of the 8 quarter lagged Equifax Risk Score, 3QMA. (FRBNY CCP/Equifax Data.)

Macroeconomic Implications - Aggregate consequences of growth in subprime lending Mortgage defaults drop in house prices contraction in credit for high MPC households drop in consumption and employment - Causal link identified from geographical variation (zip code, MSA, county, state) (Mian & Sufi 2014, Mian, Rao & Sufi 2013, Kehoe, Midrigan & Pastorino 2014, Mian, Sufi & Trebbi 2014, Midrigan & Philippon 2016, Justiniano, Primiceri & Tambalotti 2016, Guren, Nakamura, Steinsson 2017 etc )

Growth in Mortgage Balances By Zip Code - Strongest growth for prime borrowers in all zip codes Quartile 1 Quartile 2 2.5 2.5 2.25 2.25 2 2 1.75 1.75 1.5 1.5 1.25 1.25 1 1 0.75 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Above 660 Below 660 2011 2012 2013 0.75 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Above 660 Below 660 2011 2012 2013 Quartile 3 Quartile 4 2.5 2.5 2.25 2.25 2 2 1.75 1.75 1.5 1.5 1.25 1.25 1 1 0.75 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Above 660 Below 660 2011 2012 2013 0.75 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Above 660 Below 660 2011 2012 2013 Real per capita mortgage balance growth by fraction of subprime borrowers in 2001. Ratio to 2001. (FRBNY CCP/Equifax Data.)

Zip Code Variation: Role of Age Distribution - Highest debt growth in high subprime zip codes for all borrowers

Zip Code Variation: Role of Age Distribution - Highest debt growth in high subprime zip codes for all borrowers - More young borrowers in high subprime zip codes 2001 subprime share 19% 32% 44% 60% median age 50 49 48 46 Fraction in each age bin 20-34 0.22 0.25 0.28 0.30 35-54 0.42 0.41 0.41 0.41 55-85 0.38 0.34 0.32 0.30 By fraction of subprime in 2001. 2001Q1-2013Q4 averages. (FRBNY CCP/Equifax Data.)

Zip Code Variation: Role of Age Distribution - Highest debt growth in high subprime zip codes for all borrowers - More young borrowers in high subprime zip codes Quartile 4-Quartile 1 difference mostly explained by age distribution 2001Q1-2007Q4 Real Per Capita Mortgage Balance Growth Difference relative to Quartile 1 explained by age distribution Quartile 2 Quartile 3 Quartile 4 44% 43% 84% By fraction of subprime in 2001. (FRBNY/CCP Equifax Data.)

Defaults By Zip Code - Level differences in foreclosure rates, similar rise during crisis 0.02 Foreclosure Rate 0.015 0.01 0.005 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 By fraction of subprime in 2001. (FRBNY CCP/Equifax Data.)

Defaults By Zip Code - Level differences in foreclosure rates, similar rise during crisis - Large rise in prime share of defaults in all zip codes during crisis Prime Share of Foreclosures 0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 By fraction of subprime in 2001. (FRBNY CCP/Equifax Data.)

Defaults By Zip Code - Level differences in foreclosure rates, similar rise during crisis - Large rise in prime share of defaults in all zip codes during crisis Higher default rates for prime borrowers in high subprime zip codes Prime Share of Foreclosures 0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 By fraction of subprime in 2001. (FRBNY CCP/Equifax Data.)

Defaults By Zip Code: Role of Investors - Larger rise in investors for prime borrowers, similar across zip codes - More subprime investors in low subprime zip codes Prime Borrowers 0.2 0.15 0.1 0.05 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Subprime Borrowers 0.2 0.18 0.16 0.14 0.12 0.1 0.08 0.06 0.04 0.02 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Fraction with 2+ first mortgages by fraction of subprime borrowers in 2001. Prime status based on 8Q lagged credit score. (FRBNY CCP/Equifax Data.)

Defaults By Zip Code: Role of Investors - Stronger rise in balances and foreclosures for prime investors in high subprime zip codes Prime Borrowers 2001Q3-2007Q4 net mortgage balance growth no. first mortgages 2 86% 85% 97% 104% 3 94% 104% 117% 118% 4+ 102% 122% 133% 125% 2005Q4-2007Q4 change in foreclosure rate no. first mortgages 2 0.023 0.027 0.045 0.053 3 0.040 0.063 0.087 0.115 4+ 0.076 0.096 0.123 0.151 Zip code level investor activity for prime borrowers by fraction of subprime in 2001. (FRBNY CCP/Equifax Data.)

Zip Code Variation: Role of Demographics - Why did high subprime zip codes experience more severe recession?

Zip Code Variation: Role of Demographics - Why did high subprime zip codes experience more severe recession? Young, low education, high minority share Zip Code Level Indicators Associate+ degree (2012) 45% 31% 23% 17% Percent white 93% 90% 83% 63% Percent black 1.7% 3.6% 7.6% 24.6% By fraction of subprime in 2001. PDI in 2012 USD. (FRBNY CCP/Equifax Data, IPUMS, IRS, ACS.)

Zip Code Variation: Role of Demographics - Why did high subprime zip codes experience more severe recession? Young, low education, high minority share High unemployment, low income, high inequality Zip Code Level Indicators Average UR 2001-2007 4.94% 5.19% 5.38% 5.72% Average PDI 2001-2007 $41k $30k $26k $21k PDI Growth 2001-2007 25% 16% 10% 4% Mean Income $200K Mean Income (2006-11) 6.4 7.9 9.4 11.8 By fraction of subprime in 2001. PDI in 2012 USD. (FRBNY CCP/Equifax Data, IPUMS, IRS, ACS.)

Zip Code Variation: Role of Demographics - Why did high subprime zip codes experience more severe recession? Young, low education, high minority share High unemployment, low income, high inequality Higher population density, more pronounced housing cycle Zip Code Level Indicators Pop per sq mile 1,214 1,380 1,386 2,322 HPI Growth 2001-2007 29% 37% 42% 47% HPI Growth 2007-2010 -21% -30% -27% -36% By fraction of subprime in 2001. PDI in 2012 USD. (FRBNY CCP/Equifax Data, IPUMS, IRS, ACS.)

Zip Code Variation: Role of Demographics - Why did high subprime zip codes experience more severe recession? Young, low education, high minority share High unemployment, low income, high inequality Higher population density, more pronounced housing cycle Prevalence of business cycle sensitive, high MPC populations = stronger impact of recession on employment and consumption

Zip Code Variation: Role of Demographics - Why did high subprime zip codes experience more severe recession? Young, low education, high minority share High unemployment, low income, high inequality Higher population density, more pronounced housing cycle Prevalence of business cycle sensitive, high MPC populations = stronger impact of recession on employment and consumption Prevalence of urban areas = accentuated house price cycle gentrification (Guerrieri et al. 2013) international capital inflows

Conclusions I. Reassessment of role of subprime credit II. Important role of real estate investors for foreclosure crisis

Conclusions I. Reassessment of role of subprime credit II. Important role of real estate investors for foreclosure crisis - drivers of investor activity? - alternative default risk indicators?

Conclusions I. Reassessment of role of subprime credit II. Important role of real estate investors for foreclosure crisis - drivers of investor activity? - alternative default risk indicators? III. Geographical variation - larger rise in debt and defaults for prime borrowers everywhere - more severe recession in high subprime areas linked to demographics Why stronger housing cycle and investor activity in high subprime areas? - preference for urban locations - labor market factors rise in initial local income (Ferreira and Gyourko 2012) concentration of growing industries (Liebersohn 2017)