Small Business Credit Scoring

1
Small Business Credit Scoring

• An Empirical Analysis of the Viability of Pooled
  Data SME Scoring Models in Latin America
• Presented at the Conference on Small and Medium
  Enterprises, Washington, D.C., Oct. 15, 2004
• Margaret Miller

2
Project Objective

• To determine if pooled data SME credit scoring
  tools, which have proved very successful in the
  U.S. and in a few other select markets, could be
  developed for large emerging markets in Latin
  America

3
Relevant Literature on Credit Reporting

• Credit reporting can facilitate access to credit
• Positive relationship between private credit /
  GDP and credit reporting (Doing Business)
• Credit reporting reduces credit constraints on
  firms (Galindo Miller 2001, Love Mylenko
  2003)
• Credit reporting reduces the impact of bank
  concentration on access to finance (Beck,
  Demirguc-Kunt Maksimovic 2004)

4
Relationship Lending in the Small Business Credit
Market

• Old paradigm SME loan market focuses on
  relationship lending (Petersen Rajan 1994,
  Berger Udell 1995, Miller 1995)
• New paradigm SME loan market segmented with
  local lenders employing relationship lending
  technologies and national lenders using automated
  scoring (Petersen Rajan 2002, DellAriccia
  Marquez 2003, Hauswald Marquez 2002, Brevoort
  Hannan 2004)
• Newest paradigm? Many lenders using both
  technologies?

5
Literature on Small Business Credit Scoring (SBCS)

• Large banks in the U.S. were more likely to adopt
  SBCS first (Akhavein, Frame White 2001)
• U.S. banks that adopted SBCS increased their SME
  lending by 8.4 on average about 4 billion in
  increased lending per institution (Frame,
  Srinivasan Woosley 2001)
• Increased lending volumes from SBCS served to
  increase access to marginal or riskier borrowers
  (Berger, Frame Miller 2002)

6
Significant Increase in Number of SME Loans
Extended in U.S. Since Introduction of SBCS
7
Small Business Scoring Still Limited in
Developing Countries

• Only largest financial institutions have adopted
  SBCS
• Have funds to invest in expert or custom models
• View SBCS and their SME portfolio data as key
  elements of their competitive edge
• Lenders are reluctant to share data, especially
  on the SME market segment not used to working in
  common for a pooled model
• Difficult for external technology providers to
  create consortiums

8
Deficiencies in the small business lending

• Identify improvements in the banking and lending
  industry to allow small businesses growth and
  development
• Manage overall portfolio risk
• Streamline operations for increased cost
  efficiencies
• Increase number of profitable relationships
• Identify predictive data elements to analyze the
  credit risk

9
How scoring technology can help the lenders?

• Reduces cost by increasing efficiency and speed
• Make consistent ranking of risk and objective
  decisions
• Accurate risk prediction
• Majority of risk identified at origination!
• Competitive edge
• Faster response times
• Better risk assessment

10
Small business definition- United States

• Annual Sales of up to 5 million
• Credit of up to 250,000
• Types of Small Businesses
• Sole Proprietors
• Partnerships
• Corporations

11
Small Business Scoring Service (SBSS) in the
United States

• Fair Isaac partners with Robert Morris Associates
  (RMA)
• 17 banks each contribute 100 goods, 100 bads, 100
  declines
• Precoding sheets filled out by hand
• Paper credit bureau reports
• Fair Isaac data entry staff creates electronic
  database
• 1995 - Fair Isaac SBSS releases the first
  empirically derived commercial scorecards

12
Small Business Pooled Models Improvements

• Over 250,000 small businesses contributed by 25
  banks
• Data extracted electronically from application
  processing and master billing file systems 100
  availability of consumer bureau information
• Business data provides increase in predictive
  power
• Application, Business reports, Financial
  statements
• 2001 - Fair Isaac releases second generation of
  commercial scorecards
• Currently building the New Small Business Pooled
  Models

13
Small business scoring pooled models in Asia

• Pooled Models in Japan
• Over 3,500 small businesses contributed by 13
  banks
• Data transferred from application processing and
  master billing file systems and availability of
  consumer bureau information
• Business data provides increase in predictive
  power
• Application
• Financial statements
• Demographic information
• Currently building pooled models in Hong Kong

14
Definition of Small Business in Latin America
There is no consistent definition between lenders
in Latin America.

• MEXICO
• Number of Employees up to 100
• Annual Sales of up to 500,000
• Credit of up to 200,000
• Types of Small Businesses
• Sole Proprietors
• Partnerships
• Corporations

• COLOMBIA
• Number of Employees up to 200
• Annual Sales of up to 1 million
• Credit of up to 200,000
• Types of Small Businesses
• Sole Proprietors
• Partnerships
• Corporations

15
Distribution by business activity
COLOMBIA
MEXICO
16
Acceptance and booked rates
17
Distribution of delinquencies
18
Application processing time
DAYS
19
Pooled data development sampling

• A random representative portion of accounts with
  known payment behavior plus declined accounts
• Data provided by all institutions participating
  in the pool
• Predictive information and
  performance information

20
Data sources - Personal information about
principals

• Application
• Financial information
• Consumer credit reports

21
Value of data
High
Business
Personal
Value
Low
Size of Company
Small
Large
22
Report for participants

• Lending practices
• List of valuable variable that could be use for
  future model development or strategies
• Summary of individual portfolio performance
• Validation of questionnaire responses