THE PERSONAL SIDE OF RELATIONSHIP BANKING A RANDOMIZED EVALUATION START-TO-FINISH. M&E and Sustainability Reporting Training Dhaka, April 2012

Transcription

1 THE PERSONAL SIDE OF RELATIONSHIP BANKING A RANDOMIZED EVALUATION START-TO-FINISH M&E and Sustainability Reporting Training Dhaka, April 2012

2 Course Overview Importance of evidence based decision making Theory of change in program evalua:on Evalua:on methods A randomized evalua:on: start to finish Innova:ve technologies for data collec:on Sustainability repor:ng

3 The Need Small and medium sized enterprises (SMEs) are integral to economic development SMEs biggest growth challenge is access to capital OFen, however, we lack hard informa:on on their credit worthiness How do we evaluate credit worthiness? How do we minimize risk?

4 Mo1va1ons Policy Is rela:onship lending an effec:ve way to decrease default rates on SME loans? Background Does a personal rela:onship between the lender and borrower lead to: (1) more sof informa:on for the lender, (2) less willingness to default for the borrower (or (3) both)? Previous literature focused on (1)

5 Program Theory What about giving people one on one :me with a rela:onship manager at the bank? A behavioral interpreta:on: With a personal rela:onship, it becomes personally costly due to fairness norms for a borrower to default on their loan officer Feelings! A non behavioral interpreta:on: A personal rela:onship leads to embedded sof informa:on, which is costly to acquire with a new bank Clients s:ck with the bank, not out of feelings but due to pure, cold ra:onality!

6 ICICI: India s largest commercial bank The Se9ng

7 Example of Relationship Lending for SMEs: ICICI Small Business Loans At the :me, was only uncollateralized lending facility to SMEs in the formal banking system in India Loan size between $10k $50k 7

8 Small Business Loan Facility of ICICI Problem: Credit bureau informa:on are very important in se`ng up reliable credit scores, but India does not have a well func:oning credit bureau. SBL had ini:ally high default rates, above 7% Challenge: Design/test monitoring and incen:ve schemes to improve repayment rates and increase client loyalty Can the incen:ves that bank provides subs:tute for the lack of financial infrastructure in the market?

9 Original Setup Credit assessment is based on a score card approach Loans are sourced by sales agents in branches: not involved in credit assessment Centralized risk team makes credit decisions based on observable informa:on, e.g. tax filings, bank statements Goals: Efficient screening and reduced transac:on costs Loans are structured as one year overdraf facili:es Borrowers can pay online or in the branch No dedicated loan officer assigned to borrowers Start collec:on calls in case borrowers are 30+ days late Goal: Reduce transac:on costs 9

10 Design Challenge: Focus on Relationship Assign rela:onship manager Call every 2 weeks not only when borrower pays late Borrower is less willing to default on a specific loan officer than an anonymous bank Reduce cost by using phone and SMS rather than inperson mee:ng Tele callers not loan officers 10

11 Experimental Design Group A: Personal touch treatment Assign individual rela:onship manager to create ongoing rela:onship with client. Group B: Medium touch treatment Random rela:onship manages follow up with clients regularly to solve problems, understand the nature of the business. Parallel to treatment A Group C: Reminder treatment Send SMS with interest and principal due every month. Only follow up with phone call if clients have outstanding balances. Group D: No monitoring treatment Control Group During the project: the bank decided to implement reminder calls for all borrowers with delayed payments thus blurring C and D 11

12 Implementation of the Experiment Loan applicants were randomized into treatment and control groups Dura:on: July 2008 to April 2009 Sample: 1319 SBA loans across all regions of India were assigned to treatment and control groups Rela:onship managers have preset scripts to reach out to borrowers Check in every two weeks independent of loan status and payment behavior. Solve problems with accounts, remind customers of delays in payment if necessary No cross selling (!) would destroy personal trust 12

13 Data Sources Administra1ve bank data: Daily and monthly repayment reports for all clients in the study (+ info on business characteris:cs, loans, loca:on) Monitoring data on clients: Complaints and queries Exit survey of all clients: Sa:sfac:on based on the type of loan treatment and monitoring they received

14 Outcome Indicators Late payment (binary) Frequency of delinquency (intensive margin sta:s:cs) Based on the first year of the account (months 1 11) Complaints

15 Variable (1) Number of Accounts 1319 Geographic Region East 10% North 50% South 19% West 21% Constitution Type Overall Sample Proprietor 66% Partnership 19% Private Ltd 11% Other, incl Missing 4% Distribution Month % % % % % % % % Initial Loan Limit (USD $ 20,723 Product Group SBL Power 29% SBL 71% City Size CPA City 82% Non CPA City 18% Start: Baseline Descrip1ve Stats Average ini:al loan limit: US$20,723 Geographic loca:on of loans: 50% North India 20% Western India 19% South India Sample size: 1,319 individual accounts

16 Delinquency Results The Higher Touch Treatment led to a 1.91% reduc:on in number of accounts ever 30+ Days past due. The Higher Touch treatment also nega:vely impacted the number of mul:ple delinquencies. Coefficient of Treatment Dummy from OLS Regressions on Delinquency Measures Group A Group B Group C Account U:liza:on 16

17 Complaint and Resolution Results The dedicated rela:onship officer assigned to Group A seems to influence the level of service expected from the bank The dedicated Results rela:onship of OLS officer Regressions assigned on Complaint to Group Data A seems to influence Dependent Variable Coefficient of Treatment Group A Dummy Variable Ever Complained 0.082*** Group A is less likely to complain Number of Complaints if Ever Complained 0.661** but those that do complain more % of Complaints & Disputes Unresolved % of Complaints, Disputes and Issues Unresolved and the bank is less likely to be able to resolve those complaints 0.057*** *** denotes coefficient significant at 99% confidence interval; ** denotes 95%, and * denotes 90%. 17

18 Take away Delinquency rate and onset of delinquency fall significantly with personalized treatment Responsibility for rela:onship manager Borrowers feel reciprocity when well treated by the lender: protect rela:onship Importance of building loyalty Technology enhanced outreach allows for cost effec:ve approach to rela:onship building Delinquency rate and onset of delinquency fall significantly with personalized treatment Responsibility for rela:onship manager 18