Transcript Deloitte Consulting, 2005

© Deloitte Consulting, 2005
Predictive Modeling – Panacea or
Placebo?
Cheng-Sheng Peter Wu, FCAS, ASA, MAAA
CAS 2005 Spring Meeting
Scottsdale, AZ
May 16-19, 2005
© Deloitte Consulting, 2005
Agenda
What is Predictive Modeling
A Case Study of Successful Predictive
Modeling - Credit Scoring Revolution
From Credit Scoring to Predictive Modeling
What Does Predictive Modeling Mean for
Actuaries?
2
© Deloitte Consulting, 2005
What is Predictive Modeling
What is predictive modeling?
– Predictive modeling is an application of
mathematical and statistical techniques and
algorithms to produce a mathematical model
that can effectively predict and segment future
events
3
© Deloitte Consulting, 2005
Why is PM a Hot Topic?
Is it just a new actuarial fashion?
Is it just the “flavor of the month” that we all
like to talk about at conferences but nobody
really does?
No to both!
– It is a new addition to the actuary’s toolkit.
– It is here to stay.
4
© Deloitte Consulting, 2005
Why is PM a Hot Topic?
PM is a natural extension of what actuaries
have done all along.
– Use data to make predictions and forecasts.
It allows us to add statistical rigor and
additional info to traditional areas of
actuarial practice.
– Large scale data mining and multivariate
analysis
– GLM-based ratemaking
– Stochastic Loss reserving
5
© Deloitte Consulting, 2005
Why is PM a Hot Topic?
PM also allows us to broaden actuarial
practice.
– Underwriting models
– Credit scoring
– Retention and cross-sell modeling
– Target marketing models
– Agency monitoring tools
– Other industries
6
© Deloitte Consulting, 2005
What is New About “Today’s”
Predictive Modeling?
Rapid advancement of cheap computing power
Moore’s Law
Year
1980 1985 1990
1995
2000
2004
Storage Cost per
Megabyte
$190
$ 70
$ 10
$0.90
$0.05
$0.001
5-8
16
33
75
200
400
Microprocessor Speed,
MHz
7
© Deloitte Consulting, 2005
What is New About “Today’s”
Predictive Modeling?
Availability of wide range of data from
internal and external sources.
“Data Mining”:
“Data mining is a process that utilizes predictive
modeling techniques to analyze large
quantities of internal and external data, in
order to unlock previously unknown and
meaningful business relationships”
8
© Deloitte Consulting, 2005
What is New About “Today’s”
Predictive Modeling?
Development of new and powerful modeling and
data exploration techniques
– Examples: regression, GLM, neural networks,
decision trees, clustering analysis, MARS, ...
– Explore complicated patterns in data such as nonnormality, non-linearity, interactions, etc.
Statistical analysis is no longer restricted to what
you can do with pencil and paper...
…or spreadsheets.
9
© Deloitte Consulting, 2005
What is New About “Today’s”
Predictive Modeling?
Multivariate analysis with large amount of
data and many variables
– Analyze multiple variables “simultaneously”
instead of one or two at a time.
– Use large amounts of data
No need to use summarized data for actuarial
analyses.
– Create and analyze novel predictive variables.
10
© Deloitte Consulting, 2005
A Case Study of Successful PM
Which Company is This?
12
Premium, in Billions
10
8
6
4
2
0
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
Year
11
© Deloitte Consulting, 2005
Credit Score Revolution
0.35
115%
0.3
110%
0.25
105%
0.2
100%
0.15
95%
0.1
90%
0.05
85%
0
Combined Ratio
Growth Rate
Progressive vs Industry
Industry Growth Rate
Progressive Growth Rate
Industry Combined Ratio
Progressive Combined Ratio
80%
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
Year
12
© Deloitte Consulting, 2005
Why?
Multiple Choice
Progressive provided foosball tables and
free snacks to their trendy, 20-something
workforce
Progressive built a compound GammaPoisson GLM model to design their class
plan
Progressive pioneered the use of credit in
pricing/underwriting
13
© Deloitte Consulting, 2005
Credit Score Revolution
Personal line rating history:
– Few rating factors before World War II
– Explosion of class plan factors after the War
– Auto class plans:
Territory, driver, vehicle, coverage, loss and violation, others,
tiers/company…
– Homeowners class plans:
Territory, construction class, protection class, coverage, prior
loss, others, tiers/company...
– Credit scoring introduced in late 80s and early 90s
14
© Deloitte Consulting, 2005
Credit Score Revolution
About credit score:
– First important factor identified over the past 2
decades
– Composite multivariate score vs. raw credit
information
– Introduced in late 80s and early 90s
– Viewed at first as a “secret weapon”
– Quiet, confidential, controversial, black box, …etc
“Early believers and users have gained
significant competitive advantage!”
15
© Deloitte Consulting, 2005
Credit Score Revolution
Current environment of credit score:
– Now everyone is using it:
Marketing and direct solicitation
New business and renewal business pricing and
underwriting
– Regulatory constraints:
Many states have conducted studies on the true
correlation with loss ratio and potential discrimination
issues - WA study, TX study, MO study
Many states have/are considering restricting the use of
credit scores or certain type of information, MI.
More states want the “black box” filed and opened
16
© Deloitte Consulting, 2005
Credit Score Revolutions
What is “credit score”?
– A composite score that usually contains 10 to
40 pieces of credit information
Payment pattern information, account history,
bankruptcies/liens, collections, inquiries, bad
debt/defaults…
Formula scoring or rule-based scoring
Industry scores and proprietary scores
17
© Deloitte Consulting, 2005
Credit Score Revolutions
Why “credit score” is so successful?
– “Large scale” “multivariate” scoring using
“external data source”
– Loss ratio lift is significant, a powerful class plan
factor or rate tiering factor
– “Brilliant” marketing approach for credit score:
Benefits/ROI are measurable and lift curve can be
translated into bottom-line benefit
Blind test and independent validation can be done to
verify the benefit
18
© Deloitte Consulting, 2005
Loss Ratio Lift Curve
120
90
82
78
Loss Ratio
74
66
70
62
58
50
Credit Score Decile
19
© Deloitte Consulting, 2005
Credit Score Revolution
1997 NAIC/Tillinghast Study of 9 Companies' Data
Loss Ratio Relativity of the Best and Worst 20% of Credit Score
Co1
Co2
Co3
Co4
Co5
Co6
Co7
Co8
Co9
Avg
Best 20%
-38% -29% -19% -15% -14% -34% -22% -22% -36% -25%
Worst 20%
48%
20%
32%
30%
46%
59%
20%
22%
95%
41%
20
© Deloitte Consulting, 2005
From Credit Scores to
Predictive Modeling
A credit score is just “one example” of an
insurance predictive model
The same methods used to build credit scores
are used in data mining to build insurance
predictive models:
– Fully utilize all sources of internal and external data
sources
– Fully utilize all available data
Not just credit
– Other lines of business?
21
© Deloitte Consulting, 2005
What Does PM Mean for
Actuaries?
© Deloitte Consulting, 2005
What Does PM Mean for
Actuaries?
New ways of analyzing data
– New data sources
– New technologies
– New analytical tools
– True Multivariate analysis
No longer one or two variables at a time
– Analysis of risk-, policy-, or HH-level data,
rather than aggregated data.
23
© Deloitte Consulting, 2005
What Does PM Mean for
Actuaries?
New emphasis on the “business” side of the
analytical work and out-of-box thinking
Who thought of credit a decade ago?
How to stay competitive if everyone is using credit and
GLM?
What is the “next” big thing out there?
Are you using the same “lift curve” and ROI concept in
your analytical work?
How do you tie in your model/analytical work to
business benefit?
Can you demonstrate the business benefits of your
analytical work through a blind test?
…etc
24
© Deloitte Consulting, 2005
What Does PM Mean for
Actuaries?
New challenges to “actuarial” methodologies
and principles
– Actuarial Ratemaking Principle #1: “A rate is an
estimate of the expected value of future costs”
– Actuarial Ratemaking Principle #4: ” A rate is
reasonable, not excessive, not inadequate, and not
unfairly discriminatory
– But is that really the way profit-seeking companies
price their products? Are rates ultimately based on
costs or on what the market will bear?
25
© Deloitte Consulting, 2005
What Does PM Mean for
Actuaries?
New challenges to actuarial principles and
methodology:
– “Unfairly discriminatory”:
If we develop a powerful new segmentation model,
is it discriminatory to certain risks?
If we don’t introduce it, is it discriminatory to other
risks?
How do we know if we don’t do the analysis?
Actuaries’ “Static/Equalibrium” Principles vs.
Business’ “Ever Changing/Dynamic” Principles
26
© Deloitte Consulting, 2005
Placebo or Panacea?
So which is it?
Not a placebo
– PM is here to stay
– A permanent addition to the actuary’s toolkit
– Has the power to both deepen and expand
actuarial practice.
Not a panacea
– PM complements, doesn’t replace
fundamental actuarial principles
– PM does nothing without sound business
strategy and implementation.
27