Docstoc

steffen

Document Sample
steffen Powered By Docstoc
					The Sources of Value of Relationships:
Screening, Monitoring and the
Likelihood of Consumer Default
Manju Puri (Duke University)
Joerg Rocholl (ESMT)
Sascha Steffen (University of Mannheim)

Carefin Conference “Matching Stability and Performance: The Impact of New
   Regulations on Financial Intermediary Management”

Milan, September 2010
                                                The Sources of Value of Relationships   1
What are the sources of value of relationships?


     “The modern literature on financial intermediaries has primarily
    focused on the role of banks as relationship lenders… (However)
    existing empirical work is virtually silent on identifying the precise
               sources of value in relationship banking.”
                                                           (Boot (2000))

•    Do banks obtain valuable information from relationships that
     improves the quality of the loans made?

•    Should discretion in the loan granting process be allowed or
     should the loan granting process be codified?



                                                   The Sources of Value of Relationships   |2
Relationship specific information can obtained through
simple transaction accounts


 •   Importance of relationships has been documented particularly
     for banks lending to corporate customers

 •   Literature has largely defined relationships in the context of
     repeat lending but simple transaction accounts can be an
     economically meaningful source of relationship specific
     information

 •   Also, relationships can matter in a variety of ways: relationship
     type, length, breath and depth




                                                The Sources of Value of Relationships   |3
Research questions


 •   Do relationships matter in granting loans and do they matter in
     terms of ex-post performance?

 •   Do relationship loans have lower default rates?

 •   If so, what form of relationships is particularly valuable?

 •   And what are the potential mechanisms through which
     relationships allow for lower default rates?




                                                 The Sources of Value of Relationships   |4
Relationships prior to loan granting is important both to
screen applications and monitor borrowers


 •   Vast literature on the general benefits of relationships (e.g.
     James and Wier, 1990; Petersen and Rajan, 1994)
      – This research examines the benefits of relationships between
        banks and corporate customers both for banks and borrowers

 •   A few recent studies identify that banks monitor by bundling
     loans and checking accounts
      – Mester et al. (2007) find that transaction accounts provide financial
        intermediaries with valuable information for monitoring of small
        business borrowers
      – Norden and Weber (2009) find that checking account information is
        valuable and affects subsequent loan contract terms
      – We focus on retail lending and more importantly on information
        from bank-depositor relationships prior to granting the loan


                                                   The Sources of Value of Relationships   |5
Final accept / reject decision is made by loan officer


 •   Data from 2004-2008 provided by the German Savings Banks
     Association
     – Applications with accept / reject decision
     – 1 million loans made by 296 banks
     – Performance data for all loans including default events based on
       Basel II definitions
     – Relationship characteristics
         • Existence and length of relationship
         • Scope of relationships: credit and/or debit cards; existence and
           utilization of credit lines; savings accounts
         • Prior consumer loans and performance of these loans
     – Borrower characteristics and credit scores


 •   12 month default rate: 0.6%
                                                      The Sources of Value of Relationships   |6
Different forms of relationships


 •   Peculiarities of relationships between depositors and savings
     banks:
      – Only 2.5% of applicants have no prior relationships
      – 47.6% of the applicants have a relationship with their savings bank
        of more than15 years, 80% have been customer of more then 5
        years
      – 46.5% of the applicants have both credit and debit cards
      – 28.9% of the applicants have a checking account but neither a
        credit nor a debit card
      – 94.5% of the applicants already have a credit line but these are not
        used in 30.1% of the cases, 25% have utilized their line by more
        than 80%
      – 76.8% also have savings accounts (includes investments in stocks)
      – 19.2% already had a loan in the past (within our observation period)
                                                  The Sources of Value of Relationships   |7
Multivariate analysis


 Estimate:

             Yi,b = Ab + δ*Xi,b,t + β1*RELATIONSHIPS + εi,b,t

 •   Yi,b =1 if customer i at bank b defaulted within 12 months after
     loan origination; 0 otherwise.
 •   A are time fixed effects
 •   Xi,b,t is a vector of individual controls that capture in particular
     each borrower’s risk at loan origination date t
 •   RELATIONSHIPS are various relationship proxies

 •   Probit with standard errors clustered at bank level, do
     robustness check with LPM and fixed effects to show we get
     similar results with either model.
                                                  The Sources of Value of Relationships   |8
         Relationships from transaction accounts reduce default
           rates consistent with enhanced monitoring ability

                                         (1)                       (2)                        (3)                        (4)                        (5)                        (6)

Relationship Characteristics
Relationship Yes/No
Relationship (Yes)           -0.006***         (.001)
Relationship Length
Relationship <2years                                    0.014***         (.001)                                                                                    0.011***          (.001)

Relationship >=2, <5years                               0.008***         <0.001                                                                                    0.006***          (.001)

Relationship >=5, <9years                               0.004***         <0.001                                                                                    0.003***          <0.001

Relationship >=9, <15years                              0.002***         <0.001                                                                                    0.001***          <0.001
Scope of Relationships: Cards & Checking Account Information
Debit and Credit Card                                                             -0.003***         <0.001                                                         -0.002***         <0.001
Debit Card                                                                        -0.003***         <0.001                                                         -0.002***         <0.001
Credit Card                                                                       -0.001**          <0.001                                                         -0.001**          <0.001
Scope of Relationships: Credit Line
Credit Line (Yes)                                                                                            -0.008***         (.001)
Used > 150%                                                                                                                             0.003***          (.001)   0.003***          (.001)
Used > 120%, <=150%                                                                                                                      -0.001           <0.001   -0.00002          <0.001
Used > 100%, <=120%                                                                                                                     -0.001***         <0.001   -0.001*           <0.001
Used > 80%, <=100%                                                                                                                      -0.003***         <0.001   -0.002***         <0.001
Used > 20%, <=80%                                                                                                                       -0.005***         <0.001   -0.004***         <0.001
Used > 0%, <=20%                                                                                                                        -0.004***         <0.001   -0.003***         <0.001
Positive account balance                                                                                                                -0.006***         <0.001   -0.005***         <0.001

                                                                                                                    The Sources of Value of Relationships                            |9
Analyzing relationships measures related to savings
accounts and previous lending gives similar results


 •   Loans where borrowers have previous transaction account
     based relationships have lower default rates

 •   Next, we examine the impact of relationships through savings
     accounts on default rates using two measures
     – Existance of savings accounts
     – Assets held in accounts


 •   We assess the importance of repeat lending relationships
     – 1 year and 2 year look-back window


 •   Transaction account based measures are relatively more
     important compared to other measures
                                             The Sources of Value of Relationships   | 10
Relationships provide loan officers with information in
addition to information from internal credit ratings


 •   Does relationship specific information provide information beyond
     codified hard information?

 •   Internal ratings are used by loan officers in their credit decision.

 •   Results are consistent with previous findings:
      – Higher rated customers are less likely to default
      – Relationship customers are less likely to default (e.g. having a
        checking account reduces the likelihood of default by 0.3%)


 •   There is more information available to loan officers through private
     information than is captured in ratings (Puri et al. (2010b))

                                                   The Sources of Value of Relationships   | 11
Relationship specific information is valuable and cannot
be obtained from external credit bureaus


 •   Is this relationships specific information or can that information
     be obtained from third parties, i.e. credit bureaus?
      – And therefore is available to „outside lenders“ as well?


 •   We have information about external credit scores for 86,628
     loans in our sample provided by a German credit bureau.

 •   The results are again consistent
      – Borrowers with better external credit rating have a lower likelihood
        to default
      – Relationship borrowers are less likely to default (coefficient is
        similar to excluding external ratings)


                                                   The Sources of Value of Relationships   | 12
Endogeneity & identification: Employ bivariate probit model
and use a proxy for availability of financial intermediaries as
instrument

 •   What determines a relationship itself?
      – Do banks build relationships with high quality customers?


 •   Identification requires exogenous variation along the
     relationship/non-relationship margin that is uncorrelated with
     unobservable borrower quality.

 •   Instrument
      – Proxy for the availability of financial intermediaries in a region
      – Use number of savings bank branches in region -> Log(HHI)


 •   Binary probit model to account for binary nature of variables

                                                     The Sources of Value of Relationships   | 13
Availability of financial intermediaries increases the
likelihood of having a relationship, unobservable
borrower characteristics don‘t affect our results

                                                                  Coefficents                        Marginal Effects
                                                 Relationships                    Default                Default

      Relationship (Yes)                                N/A               -0.706***         (.238)        -0.023

      Borrower Characteristics
      Log (Income)                          -0.07***          (.007)      -0.132***         (.009)        -0.002
      Repayment (% of Income)
                                   < 5%     -0.064***          (.01)      -0.227***         (.022)        -0.002
                           >= 5%, < 11%     0.040***          (.007)      -0.213***         (.012)        -0.003
                           >=11%, < 13%       0.005           (.011)      -0.143***          (.02)        -0.002
                           >=13%, < 20%      -0.018**         (.008)      -0.082***         (.014)        -0.001
      Age
                           18 to 23 years   -0.256***         (.019)       0.499***         (.034)        0.012
                           23 to 25 years   -0.255***         (.019)       0.448***         (.034)        0.010
                           25 to 30 years   -0.300***         (.017)       0.392***         (.032)        0.008
                           30 to 45 years   -0.215***         (.015)       0.380***          (.03)        0.006
                           45 to 50 years   -0.140***         (.017)       0.282***         (.032)        0.005
                           50 to 60 years   -0.082***         (.016)       0.259***          (.03)        0.004
      Job Stability
                              <= 2 years    -0.163***         (.007)       0.270***         (.011)        0.004

      Log(HHI)                              0.118***          (.009)                 N/A                   N/A
                                                                 Wald Test: ρ=0.15573 (p=0.1172)
      Number of observations                                                 1,036,901
                                                                         Χ²(76)=6,921.75
                                                                         Prob > Χ²=0.000


                                                                                      The Sources of Value of Relationships   | 14
Use selection models to separate screening (loan
application stage) from monitoring


 •   What are the sources of value of relationships?
      – Do relationships help banks to better screen customers and thus to
        reduce default rates?
      – Or does relationship specific information improve monitoring?


 •   Potential selection bias (Heckman (1979))
      – If the banks’ screening process is not based on credit scores but
        also includes subjective elements or discretion
      – If the error terms of loan granting model and default model are
        correlated
      – We use a Log(HHI) as instrument for identification.
          • More competition has a positive effect on credit supply (Jayaratne and
            Strahan (1996), Black and Strahan (2002), Zarutskie (2006)).


                                                      The Sources of Value of Relationships   | 15
Discretion at the loan application stage is associated with
lower frequencies of default, relationships still significant

                                                                     Coefficents                                 Marginal Effects
                                                     Accepted                                 Default                Default

       Relationship (Yes)                    0.547***           (.012)            -0.521***             (.013)        -0.036


       Borrower Characteristics
       Log (Income)                          0.151***           (.007)            -0.168***             (.006)        -0.007

       Repayment (% of Income)
                                    < 5%      0.381***          (.016)            -0.335***             (.015)        -0.010
                            >= 5%, < 11%      0.252***          (.008)            -0.253***             (.008)        -0.009
                            >=11%, < 13%      0.035***          (.012)            -0.079***             (.012)        -0.003
                            >=13%, < 20%     -0.034***          (.008)             -0.004               (.009)        0.000
       Age
                            18 to 23 years   -0.894***          (.022)            0.853***              (.022)        0.078
                            23 to 25 years   -0.547***          (.023)            0.573***              (.022)        0.041
                            25 to 30 years   -0.484***          (.021)            0.502***              (.021)        0.031
                            30 to 45 years   -0.353***           (.02)            0.408***               (.02)        0.019
                            45 to 50 years   -0.241***          (.022)            0.292***              (.021)        0.015
                            50 to 60 years   -0.128***          (.021)            0.212***               (.02)        0.010
       Job Stability
                               <= 2 years    -0.394***          (.007)            0.394***              (.007)        0.022

       Log(HHI)                              -0.177***       (.009)                     N/A                            N/A
                                                         Wald Test: ρ= -0.97571 (p<0.001)

       # of censored observations                                        23,060

       # of uncensored observations                                   1,036,901
                                                                   Χ²(38)=17740.59
                                                                   Prob > Χ²=0.000

                                                                                                The Sources of Value of Relationships   | 16
Conclusion


 •   We take advantage of a unique dataset to study the sources of
     value of relationships.
     – Relationships matter in different forms (transactions accounts,
       savings accounts, prior loans) and scope (credit and debit cards,
       credit line) and depth (relationship length, utilization of credit line,
       money invested in savings accounts)
     – Importantly, though, even the simplest forms of relationships such
       as transaction accounts (e.g., savings or checking accounts) are
       economically meaningful in reducing defaults.
     – We find that banks’ screening process is not deterministic but
       includes elements of subjective assessment or discretion which is
       associated with lower frequencies of defaults.
     – However, even after taking screening into account, prior
       relationships still have a first order impact in reducing borrower
       defaults.
                                                     The Sources of Value of Relationships   | 17

				
DOCUMENT INFO