Discussion - Credit Boom and Lending Standard: Evidence...
Transcript of Discussion - Credit Boom and Lending Standard: Evidence...
INTRODUCTIONCOMMENTS
SUMMARY
DiscussionCredit Boom and Lending Standard: Evidence from Subprime
Mortgage Lending
Satyajit Chatterjee
Research DepartmentFederal Reserve Bank of Philadelphia
September 2008
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
OBJECTIVES AND FINDINGS
Consensus that the source of the current subprimemortgage crisis is increasingly lax lending standardsWhy did lending standards decline?
Rapid expansion of credit caused lower lending standardsIncrease in the number of lenders caused lower lendingstandards
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
OBJECTIVES AND FINDINGS
Consensus that the source of the current subprimemortgage crisis is increasingly lax lending standardsWhy did lending standards decline?
Rapid expansion of credit caused lower lending standardsIncrease in the number of lenders caused lower lendingstandards
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
OBJECTIVES AND FINDINGS
Consensus that the source of the current subprimemortgage crisis is increasingly lax lending standardsWhy did lending standards decline?
Rapid expansion of credit caused lower lending standardsIncrease in the number of lenders caused lower lendingstandards
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
OBJECTIVES AND FINDINGS
Consensus that the source of the current subprimemortgage crisis is increasingly lax lending standardsWhy did lending standards decline?
Rapid expansion of credit caused lower lending standardsIncrease in the number of lenders caused lower lendingstandards
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
APPROACHKey Idea
Use loan-level data (HMDA) for 2000 to estimate
Dj k = X ′j β + εj k
Identity of lender is proxy for prime or subprimeUse β̂ to predict denial rate of mortgage applications inyears 2001-2006 for different metropolitan areas (MSAs)Regress prediction error (actual - predicted) on log of thenumber of loan applicants in the MSA and other MSA-levelcontrolsLog of the number of applicants affects the prediction errornegatively – more applicants there are, the less choosylenders become
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
APPROACHKey Idea
Use loan-level data (HMDA) for 2000 to estimate
Dj k = X ′j β + εj k
Identity of lender is proxy for prime or subprimeUse β̂ to predict denial rate of mortgage applications inyears 2001-2006 for different metropolitan areas (MSAs)Regress prediction error (actual - predicted) on log of thenumber of loan applicants in the MSA and other MSA-levelcontrolsLog of the number of applicants affects the prediction errornegatively – more applicants there are, the less choosylenders become
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
APPROACHKey Idea
Use loan-level data (HMDA) for 2000 to estimate
Dj k = X ′j β + εj k
Identity of lender is proxy for prime or subprimeUse β̂ to predict denial rate of mortgage applications inyears 2001-2006 for different metropolitan areas (MSAs)Regress prediction error (actual - predicted) on log of thenumber of loan applicants in the MSA and other MSA-levelcontrolsLog of the number of applicants affects the prediction errornegatively – more applicants there are, the less choosylenders become
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
APPROACHKey Idea
Use loan-level data (HMDA) for 2000 to estimate
Dj k = X ′j β + εj k
Identity of lender is proxy for prime or subprimeUse β̂ to predict denial rate of mortgage applications inyears 2001-2006 for different metropolitan areas (MSAs)Regress prediction error (actual - predicted) on log of thenumber of loan applicants in the MSA and other MSA-levelcontrolsLog of the number of applicants affects the prediction errornegatively – more applicants there are, the less choosylenders become
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
APPROACHKey Idea
Use loan-level data (HMDA) for 2000 to estimate
Dj k = X ′j β + εj k
Identity of lender is proxy for prime or subprimeUse β̂ to predict denial rate of mortgage applications inyears 2001-2006 for different metropolitan areas (MSAs)Regress prediction error (actual - predicted) on log of thenumber of loan applicants in the MSA and other MSA-levelcontrolsLog of the number of applicants affects the prediction errornegatively – more applicants there are, the less choosylenders become
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
COMMENTSContribution
Use of data reported under the Home Mortgage DisclosureAct for assessing lending standards in local housingmarkets (MSAs)Exploit variations across local housing markets to shedlight on determinants of lending standards – great idea!!
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
COMMENTSContribution
Use of data reported under the Home Mortgage DisclosureAct for assessing lending standards in local housingmarkets (MSAs)Exploit variations across local housing markets to shedlight on determinants of lending standards – great idea!!
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
COMMENTSMain Concern
Why did the number of applicants go up so much across allMSAs?This question is not directly addressed in the paperbecause the authors focus on differences across MSAsBut thinking about why the number of applicants went upsuggests that the authors are ignoring causal influencesrunning the other way – from lower denial rates to moreapplicants
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
COMMENTSMain Concern
Why did the number of applicants go up so much across allMSAs?This question is not directly addressed in the paperbecause the authors focus on differences across MSAsBut thinking about why the number of applicants went upsuggests that the authors are ignoring causal influencesrunning the other way – from lower denial rates to moreapplicants
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
COMMENTSMain Concern
Why did the number of applicants go up so much across allMSAs?This question is not directly addressed in the paperbecause the authors focus on differences across MSAsBut thinking about why the number of applicants went upsuggests that the authors are ignoring causal influencesrunning the other way – from lower denial rates to moreapplicants
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
COMMENTSAn Alternative Story
The key event was the arrival of a cost-effective businessmodel for making higher-risk mortgages: “originate anddistribute”This meant lower denial rates on higher-risk loans – sothere were more buyers chasing homesThe resulting increase in house prices induced people toupdgrade or switch from renting to owning, whichgenerated more applicantions for mortgagesThis story explains how decline in denial rates could leadto an increase in applications and house pricesHowever, it is possile that the influx of these newapplicants may have led to further decline in lendingstandards thru channels suggested by the authors
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
COMMENTSAn Alternative Story
The key event was the arrival of a cost-effective businessmodel for making higher-risk mortgages: “originate anddistribute”This meant lower denial rates on higher-risk loans – sothere were more buyers chasing homesThe resulting increase in house prices induced people toupdgrade or switch from renting to owning, whichgenerated more applicantions for mortgagesThis story explains how decline in denial rates could leadto an increase in applications and house pricesHowever, it is possile that the influx of these newapplicants may have led to further decline in lendingstandards thru channels suggested by the authors
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
COMMENTSAn Alternative Story
The key event was the arrival of a cost-effective businessmodel for making higher-risk mortgages: “originate anddistribute”This meant lower denial rates on higher-risk loans – sothere were more buyers chasing homesThe resulting increase in house prices induced people toupdgrade or switch from renting to owning, whichgenerated more applicantions for mortgagesThis story explains how decline in denial rates could leadto an increase in applications and house pricesHowever, it is possile that the influx of these newapplicants may have led to further decline in lendingstandards thru channels suggested by the authors
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
COMMENTSAn Alternative Story
The key event was the arrival of a cost-effective businessmodel for making higher-risk mortgages: “originate anddistribute”This meant lower denial rates on higher-risk loans – sothere were more buyers chasing homesThe resulting increase in house prices induced people toupdgrade or switch from renting to owning, whichgenerated more applicantions for mortgagesThis story explains how decline in denial rates could leadto an increase in applications and house pricesHowever, it is possile that the influx of these newapplicants may have led to further decline in lendingstandards thru channels suggested by the authors
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
COMMENTSAn Alternative Story
The key event was the arrival of a cost-effective businessmodel for making higher-risk mortgages: “originate anddistribute”This meant lower denial rates on higher-risk loans – sothere were more buyers chasing homesThe resulting increase in house prices induced people toupdgrade or switch from renting to owning, whichgenerated more applicantions for mortgagesThis story explains how decline in denial rates could leadto an increase in applications and house pricesHowever, it is possile that the influx of these newapplicants may have led to further decline in lendingstandards thru channels suggested by the authors
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
COMMENTSEndogeneity and the IV Regression
Is the number of applicants for prime mortgages a goodinstrument for the number of applicants for subprimemortgages?With decline in the subprime denial rate, homes would sellat a faster rate and if some fraction of departing ownersapply for prime mortgages then applications for primemortgages will go up as wellSo, in theory, there could be a causal link from subprimedenial rates to the number of prime mortgage applicationsSuggestion: How about using number of prime mortgageapplications in “far away” MSAs as the instrument?
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
COMMENTSEndogeneity and the IV Regression
Is the number of applicants for prime mortgages a goodinstrument for the number of applicants for subprimemortgages?With decline in the subprime denial rate, homes would sellat a faster rate and if some fraction of departing ownersapply for prime mortgages then applications for primemortgages will go up as wellSo, in theory, there could be a causal link from subprimedenial rates to the number of prime mortgage applicationsSuggestion: How about using number of prime mortgageapplications in “far away” MSAs as the instrument?
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
COMMENTSEndogeneity and the IV Regression
Is the number of applicants for prime mortgages a goodinstrument for the number of applicants for subprimemortgages?With decline in the subprime denial rate, homes would sellat a faster rate and if some fraction of departing ownersapply for prime mortgages then applications for primemortgages will go up as wellSo, in theory, there could be a causal link from subprimedenial rates to the number of prime mortgage applicationsSuggestion: How about using number of prime mortgageapplications in “far away” MSAs as the instrument?
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
COMMENTSEndogeneity and the IV Regression
Is the number of applicants for prime mortgages a goodinstrument for the number of applicants for subprimemortgages?With decline in the subprime denial rate, homes would sellat a faster rate and if some fraction of departing ownersapply for prime mortgages then applications for primemortgages will go up as wellSo, in theory, there could be a causal link from subprimedenial rates to the number of prime mortgage applicationsSuggestion: How about using number of prime mortgageapplications in “far away” MSAs as the instrument?
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
COMMENTSCompetition and Lending Standards
In the “originate and distribute model”, incumbentmortgage lenders earn money in origination feesAll else remaining the same, lower denial rates mean moreoriginations per lenderThe additional return could have attracted new entrantsSo, again, in theory there could be a link between a drop indenial rates and entry of new lenders
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
COMMENTSCompetition and Lending Standards
In the “originate and distribute model”, incumbentmortgage lenders earn money in origination feesAll else remaining the same, lower denial rates mean moreoriginations per lenderThe additional return could have attracted new entrantsSo, again, in theory there could be a link between a drop indenial rates and entry of new lenders
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
COMMENTSCompetition and Lending Standards
In the “originate and distribute model”, incumbentmortgage lenders earn money in origination feesAll else remaining the same, lower denial rates mean moreoriginations per lenderThe additional return could have attracted new entrantsSo, again, in theory there could be a link between a drop indenial rates and entry of new lenders
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
COMMENTSCompetition and Lending Standards
In the “originate and distribute model”, incumbentmortgage lenders earn money in origination feesAll else remaining the same, lower denial rates mean moreoriginations per lenderThe additional return could have attracted new entrantsSo, again, in theory there could be a link between a drop indenial rates and entry of new lenders
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
SUMMARY
The idea that expansion of credit leads to lax standards isplausibleRich data set, but the regressions suffer from potentialendogeneity issuesSimultaneous equation approach?Be more specific about the exact channel via whichexpansion affects standards?
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
SUMMARY
The idea that expansion of credit leads to lax standards isplausibleRich data set, but the regressions suffer from potentialendogeneity issuesSimultaneous equation approach?Be more specific about the exact channel via whichexpansion affects standards?
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
SUMMARY
The idea that expansion of credit leads to lax standards isplausibleRich data set, but the regressions suffer from potentialendogeneity issuesSimultaneous equation approach?Be more specific about the exact channel via whichexpansion affects standards?
Satyajit Chatterjee Discussion
INTRODUCTIONCOMMENTS
SUMMARY
SUMMARY
The idea that expansion of credit leads to lax standards isplausibleRich data set, but the regressions suffer from potentialendogeneity issuesSimultaneous equation approach?Be more specific about the exact channel via whichexpansion affects standards?
Satyajit Chatterjee Discussion