Riders of the Storm: Economic Shock and Bank Lending in a Natural Experiment Author: Matthieu Chavaz Comments by David Martinez-Miera (UC3M) BdE and World Bank 24 June 2014 David Martinez-Miera (UC3M) Discussion 24 June 2014 1 / 15
Introduction Empirical analysis of bank lending behaviour How do banks react to negative (natural) shocks? Framework: Panel data analysis (2002-2007) Counties a ected by 2005 Hurricanes If a ected there is a negative shock Banks exposed to those counties Branch location Show how banks more exposed lend more in a ected markets! In comparison to those less exposed David Martinez-Miera (UC3M) Discussion 24 June 2014 2 / 15
My personal view Nice framework for analysis Convincing empirical results Would like to learn more about underlying mechanism David Martinez-Miera (UC3M) Discussion 24 June 2014 3 / 15
Road Map My (brief) review of the results C1: Diversi cation vs Soft Information C1.1: Gambling for resurrection? C2: Competition setup David Martinez-Miera (UC3M) Discussion 24 June 2014 4 / 15
Data Yearly data 2002 to 2007 Mortgage data County level data for branches and hurricanes County is a ected if hit by hurricane (2005-2007) Bank is exposed if it lends in that county Compare counties a ected and not Compare banks more a ected Compare banks that are more exposed in counties that are a ected David Martinez-Miera (UC3M) Discussion 24 June 2014 5 / 15
A ected counties David Martinez-Miera (UC3M) Discussion 24 June 2014 6 / 15
Exposed measure Bank exposed measure Branches in a ected counties Total Branches Continuos variable [0, 1] If bank has higher fraction of branches in a ected markets higher exposure David Martinez-Miera (UC3M) Discussion 24 June 2014 7 / 15
Hypotheses Substitution Hypotheses Banks less a ected should lend more in a ected areas Because they do not su er looses Segmentation Hypotheses Banks more a ected should lend more in a ected areas Because they are less diversi ed Because they have better soft information David Martinez-Miera (UC3M) Discussion 24 June 2014 8 / 15
Main Result Banks that are more exposed increase their lending more Segmentation Hypotheses David Martinez-Miera (UC3M) Discussion 24 June 2014 9 / 15
Main Conclusion Banks that are more exposed increase their lending more This is because in downturns soft information is more valuable More exposed banks are less diversi ed and have higher soft information about the borrowers Soft information matters specially in downturns Really important for regulators (Macro prudential) Mi view- Not the author David Martinez-Miera (UC3M) Discussion 24 June 2014 10 / 15
Comment 1 Diversi cation vs Soft Information Segmentation hypotheses has di erent underlying hypotheses Highly diversi ed banks use lower soft information The result therefore is about value of soft information Probably right Highly diversi ed banks have better outside options Then they do not lend in a ected markets Can we tease them out? Soft information banks vs hard info banks - location? Do exposed banks open (not close) o ces? Do not exposed banks close o ces? Do soft info intensive banks enter? Diversi ed banks increase lending in others markets? Transfer of funds? David Martinez-Miera (UC3M) Discussion 24 June 2014 11 / 15
Comment 1 Diversi cation vs Soft Information Not so positive view of segmentation Gambling for resurrection by highly exposed banks? Possible non linearities in the result Average quality (Z-score, CDS spreads) of banks highly exposed Ex-post measures of mortgages (or other loans) originated by these banks Authors do control for "quality" using application and acceptance data However lots of these loans are sold Why care about reputation if bank is underwater? David Martinez-Miera (UC3M) Discussion 24 June 2014 12 / 15
Comment 2 Does the competitive structure matter? Does the local competitive structure pre merger matter for the results? If bank is highly exposed it might react di erent if local markets are more competitive Is there a restructuring of the branch network? Which type of banks does it favour? Probably another paper... David Martinez-Miera (UC3M) Discussion 24 June 2014 13 / 15
Comment 3- Exposed measure Bank exposed measure Branches in a ected counties Total Branches Continuos variable [0, 1] If you are big more probable to be exposed in a low manner If you are small you are either very exposed or none exposed Log of assets of exposed 17 vs 14 for non exposed Can you drop the banks that have 0 s? Intensity of the treatment but probably much more similar banks Can you create a bucket variable? Possible non linear e ects David Martinez-Miera (UC3M) Discussion 24 June 2014 14 / 15
Conclusions Clear results concerning banks response to natural shock Banks more exposed are the ones that lend more in a ected markets Can we tease out the underlying mechanism that drives these result Withdrawal of little exposed banks Value of soft information If soft information is the key driver then this opens very important issues Hard information oriented banking system might recover worse after crisis Basel III favours hard information! David Martinez-Miera (UC3M) Discussion 24 June 2014 15 / 15