Better pricing needs better underwriting
Transcript of Better pricing needs better underwriting
5/5/2010
1
© 2010 The Actuarial Profession www.actuaries.org.uk
Current issues in general insurance (CIGI)Julian Beardsworth Neil Chapman
Better pricing needs better underwriting
10 May 2010
Better pricing needs better underwriting
Our message:
Improvements in technical pricing are both good
and inevitable, however, with analytical power
comes underwriting responsibility.
1© 2010 The Actuarial Profession www.actuaries.org.uk
5/5/2010
2
Better pricing needs better underwriting
Agenda
• Introduction
• Distribution
• Bodily injury
• Other technical pricing advances
• Conclusion
2© 2010 The Actuarial Profession www.actuaries.org.uk
Introduction
Better pricing needs better underwriting
• Recent years have seen advancements in pricing techniques
and sophistication
• How has the role of underwriting changed and how does
underwriting fit alongside pricing?
3© 2010 The Actuarial Profession www.actuaries.org.uk
Getting the right balance
5/5/2010
3
Introduction
Better pricing needs better underwriting
• Pricing definition
– Setting prices to achieve target returns based on analysis of
historic experience incorporating trending, risk knowledge
and commercial expertise
• Underwriting definition
– Ensuring risks written are consistent with the assumptions
used in pricing– ….generic pricing assumptions include:
• utmost good faith;
• no fraudulent intent;
• the past can be used as a guide to the future;
• customer behaviour not changing overnight.
4© 2010 The Actuarial Profession www.actuaries.org.uk
Introduction
For presentational effect we’re working with mental caricatures:
5© 2010 The Actuarial Profession www.actuaries.org.uk
Underwriter Pricer
• Underwriting isn’t the preserve of Underwriters; it’s an activity that
complements technical pricing
• What we now go on to describe as underwriting is delivered by actuaries,
pricing teams and other insurance professionals as well as underwriters
(Noting that he’s
“playing blind”?)
5/5/2010
4
Introduction
Better pricing needs better underwriting
• Applicability
– Considerations based on commodity business supported by
statistical pricing (i.e. think personal motor)
– But the underlying principles apply much more widely
• Approach for today’s workshop
– We’ll consider:– Examples of better technical pricing
– Associated assumptions
– Underwriting response
6© 2010 The Actuarial Profession www.actuaries.org.uk
Better pricing needs better underwriting
Agenda
• Introduction
• Distribution
• Bodily injury
• Other technical pricing advances
• Conclusion
7© 2010 The Actuarial Profession www.actuaries.org.uk
5/5/2010
5
Distribution
Pricing
• Elasticity modelling and price optimisation have become
standard in recent years
• Price comparison sites have increased importance of a good
understanding of customer behaviour
• Winners on PCS will have a low cost base and the best risk
cost models
• Distribution changes the risk cost
• Pricing for the ultimate level of competition
8© 2010 The Actuarial Profession www.actuaries.org.uk
Distribution
Pricing
• Elasticity modelling considers how demand changes with price
9© 2010 The Actuarial Profession www.actuaries.org.uk
Low elasticity
Competitiveness
Retention Rate
High elasticity
0.5
0.5
2.0
1.0
1.0
Age 25 Age 50
Private Car Comprehensive Renewals
0.38
0.40
0.42
0.44
0.46
0.48
0.50
0.52
0.54
0.56
0.58
0.60
0.62
0.64
0.66
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
Policyholder Age
42%
62%
Renewal
Rate
Policyholder AgePrice
Retention
Rate
Policyholder Age
5/5/2010
6
0
2
4
6
8
10
12
14
Phone Web Aggregator Renewals
Private Motor Comprehensive
Elasticity
Distribution
Pricing
• Different channels have different sensitivities to any change in
price
• Elasticity = % Change in Demand / % Change in Price
10© 2010 The Actuarial Profession www.actuaries.org.uk
Distribution
Pricing
• Price optimisation considers both profit and volume
• Volume and profit can be traded to determine an optimal
strategy
11© 2010 The Actuarial Profession www.actuaries.org.uk
Premium for policy in question
400
410
420
430
440
450
460
470
480
490
500
510
520
530
540
550
560
570
580
590
600
Price
Optimal price?
Profit per policy
Retention rate
Expected profit
Retention Rate
Total
Profit
Current prices
Maximise profit at
current retention
Maximise retention at
current profit
Increase profit and retention
5/5/2010
7
Distribution
Pricing
• Optimisation has
shown to deliver
material benefits
• Motor renewal
optimisation for direct
channel
• Pilot implementation
vs. “business as usual”
12© 2010 The Actuarial Profession www.actuaries.org.uk
Actual Outcomes
COR Improvement
Retention
Rate
Improvement
-2.5 points0 points
0 points
-1.5 points
+6.0 points
Actual
(BAU
Rates)
Actual
(Optimised)
Predicted
(Optimised)
Efficient
Frontier
Distribution
Pricing
• Distribution is causing pricing to cover more ground
– More competitive, smaller margins, different objectives
• Modelling makes use of more than just risk factors
• Has to use external data effectively
• Optimisation needs quality inputs (GIGO)
– risk premium models
– cost models
– elasticity and demand models
13© 2010 The Actuarial Profession www.actuaries.org.uk
5/5/2010
8
Distribution
Pricing
• Continued improvement of risk cost modelling embracing
– Structure
– Data
– Latest techniques
• Explicit risk and behaviour modelling of distribution
characteristics
• Second and third order elements of risk cost models need
increasingly to be used
14© 2010 The Actuarial Profession www.actuaries.org.uk
Distribution
Assumptions
• Web customer transaction is the same as telephone transaction
(70% by phone in 2006, up to 70% on web in 2010)
• Business models are the same as they were
• Pricing data mean the same thing today as pre-web
• The level of anti-selection hasn’t changed
• Claims inflation is claims inflation
15© 2010 The Actuarial Profession www.actuaries.org.uk
5/5/2010
9
The underwriting view:The nature of customer access has changed
Quote Driving Age Garaging Claims NCD Postcode Premium
% of
correct
pemium
Notes
1 IOD 24 Road 1 0 SE18 1PT £3,425 100% Correct details
2 IOD 24 Road 0 0 SE18 1PT £2,960 86% Accident was on someone else's policy
4 2 Named 24/51 Road 0 0 SE18 1PT £2,890 84% Add parent as driver
3 IOD 24 Road 0 2 SE18 1PT £2,168 63% Named driver NCD(?)
5 2 Named 51/24 Road 0 5 SE18 1PT £1,026 30% Fronting
6 2 Named 51/24 Garaged 0 5 SE18 1PT £926 27% Fronted address has garage
7 2 Named 51/24 Garaged 0 5 NR13 5NN £435 13% Insure at fronting address
• Very significant threat: estimated 5-10% more premium on internet transactions lost to
underwriting fraud compared to phone transactions
• Areas for motor underwriting fraud include: fronting, location, NCD, claims, mileage, garaging,
use, driving restriction, claims, convictions
• Household underwriting fraud tends to focus on previous claims, NCD and underinsurance
• Underwriting needs to find a way of making quotation engines tamper-resistant
A sequence of quotes on an internet quote engine progressing from correct risk to deeply
misrepresented and fronted………………….saving 87% of premium
16
The underwriting view:Business models have changed:
17© 2010 The Actuarial Profession www.actuaries.org.uk
InsurerTechnical premium
+ street pricing adjustment
Intermediary% Commission
Customer
InsurerNet technical premium
+ street pricing adjustment
+ optimisation for LTV
+ intermediary pricing
Intermediary+ optimisation for value
to intermediary
+ claims referral income
Aggregator+ paying flat rate fee
+ looking to increase
churn
Customer+ cashback from Quidco
• What does all this added complexity mean?
– Distance between insurers and their customers has increased
– Would expect reduced loyalty and increased willingness by customers to play games
– Price paid by customer is only tenuously based on risk premium + distribution cost
– Underwriting needs to reappraise wordings, acceptance criteria and underwriting rules
5/5/2010
10
The underwriting view:Data doesn’t mean the same as it did
Distribution changes have to some degree compromised:
• Mileage
• NCD and past claims
• Occupation (Money Saving Expert advice on cheapest way to describe yourself)
• Driving restrictions
• Agent
• Underwriting needs to call the future trend on these and find
improved means of validation where necessary
18© 2010 The Actuarial Profession www.actuaries.org.uk
Motor: Margin Benefit by Panel Size
0%
5%
10%
15%
20%
25%
30%
35%
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
The underwriting view:The winner’s curse has changed profitability
• The more competitors the lower the average price
• When an insurer is the cheapest of many, are they more likely to have a pricing error than
when they’re average?
• Underwriting can help to manage the extremes of pricing error by developing tighter
underwriting and contributing to the pricing model on areas of unintentional overfitting19
5/5/2010
11
The underwriting view:When is claims inflation not claims inflation?
• Any unidentified increase in risk mix results in an increase in claims cost that
will be misattributed to claims inflation
• Underwriting must be wise to this and ensure that the Claims Director isn’t
sacked for an element of claims inflation he has no control over
• Acid test is whether market claims costs increase in line with the high claims
inflation rates identified by insurers 20
© 2010 The Actuarial Profession www.actuaries.org.uk
Adverse selection example where customers get better at spotting value for money:
Time t-1
Risk type
A
Risk type
B
Risk type
C Time t
Risk type
A
Risk type
B
Risk type
C
Premium £100 £100 £100 Premium £100 £100 £100
Risk cost £85 £70 £55 Risk cost £85 £70 £55
Volume 5 3 2 Volume 7 2 1
Insurer's risk cost models and pricing doesn't differentiate risks A, B and C
With increasing adverse selection risk type A sells better while risk types B and C sell worse
Results for period t-1 Results for period t
Total premium £1,000 Total premium £1,000
Total risk cost £745 Total risk cost £790
Loss ratio 74.5% Loss ratio 79.0%
Actual claims inflation of 0% is measured at 6%
Better pricing needs better underwriting
Agenda
• Introduction
• Distribution
• Bodily injury
• Other technical pricing advances
• Conclusion
21© 2010 The Actuarial Profession www.actuaries.org.uk
5/5/2010
12
0%
20%
40%
60%
80%
100%
120%
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Proportion of TPBI and TPPD Buring Cost over time
TPBI TPPD Other
Bodily Injury
Pricing
• Proportion of bodily injury has increased significantly
• Have modelling approaches adapted sufficiently?
22© 2010 The Actuarial Profession www.actuaries.org.uk
-10%
-5%
0%
5%
10%
15%
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
TPBI Inflation
BI Frequency BI Severity BI Burning Cost
Bodily Injury
Pricing
• Enhance risk models
– Consider different model structures – Split by type of event: Rear shunt, Head-on, …
– Tiered capping levels: 0-10k, 10k – 25k, …
– Split by injury type: Whiplash vs other
– Propensity for BI given TPPD
– Inferences from AD & TPPD severity for BI frequency and severity
– Reflect geographic and socio-demographic influences– Best use of external data
– Up to date geographical classification
23© 2010 The Actuarial Profession www.actuaries.org.uk
0.900
1.000
1.100
1.200
1.300
1.400
1.500
1.600
Change in Frequency Since 2005
Aggregate Direct Aggregate Intermediated
5/5/2010
13
Bodily Injury
Pricing
• Enhance risk models
– Understand range of uncertainty in risk estimates
• Utilise available fraud data
– Fraud models and fraud scoring– Policy underwriting and claims underwriting
– Impact of any changes in fraud approach– How does increasing fraud screening impact claims cost?
24© 2010 The Actuarial Profession www.actuaries.org.uk
Bodily Injury
Assumptions
• Assumptions
– Fraud doesn’t need to be considered because detected fraud
has £0 cost and so excluded from cost models
– BI fraud is an unpredictable, random occurrence perpetrated
by unknown third parties, so fraud propensity in claim
unrelated to first party characteristics (other than geography)
– Number of injuries per accident is constant over time
25© 2010 The Actuarial Profession www.actuaries.org.uk
5/5/2010
14
The underwriting view:Bodily Injury
• The expertise of Claims need to be brought
into future pricing and underwriting
• Are underwriting rules, acceptance criteria,
policy terms and conditions aligned with the
preponderance of third party property
damage and bodily injury claims? Many will
target AD and theft risk.
• Are there policyholder characteristics that
make bodily injury claims outcomes more
likely – what should be done about them?
• The significance of Distribution on BI claim
frequency – and what to do about it?
• Can lessons be taken from motor markets
such as Texas where the injury rates are
even higher than the North West?
26© 2010 The Actuarial Profession www.actuaries.org.uk
Proportion of collisions giving rise to injuries -
data now 10 years old
Maryland
LiverpoolOhio
Manchester
Preston
Watford
SE London
Edinburgh
Aberdeen
Los Angeles
Texas
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
The underwriting view:Bodily Injury and Fraud
• Critical for underwriting to understand changing fraud risk
• Claims expertise is a vital input
• Establish acceptable means for potential fraud to be identified
• Establish data requirements associated with fraud in order to better
underwrite and mitigate it
• Underwriting must assist the BI claims function
• Establish the links between underwriting fraud and claims fraud
• Learn from the past – if fraud wasn’t a major issue ten years ago what
lessons can be learnt?
• Using fraud propensity scores in underwriting and claims handling
27© 2010 The Actuarial Profession www.actuaries.org.uk
5/5/2010
15
Better pricing needs better underwriting
Agenda
• Introduction
• Distribution
• Bodily injury
• Other technical pricing advances
• Conclusion
28© 2010 The Actuarial Profession www.actuaries.org.uk
Other technical pricing advances
Geographic classification using spatial smoothing
• Pricing– Spatial smoothing is common practice for geographical analysis
– Best results enhance effect of external factors
– Geographical effects change over time and need to be maintained
29© 2010 The Actuarial Profession www.actuaries.org.uk
5/5/2010
16
Geographic classification using spatial smoothing
– Adjacency based
• Postcodes are similar to
neighbouring areas
• Assumptions– Distance based
• Influence of a postcode inversely
proportional to distance away
Other technical pricing advances
30© 2010 The Actuarial Profession www.actuaries.org.uk
Distance Adjacency
Other technical pricing advances
Vehicle or Trade Smoothing
• Spatial smoothing can be extended to vehicle or trade codes
31© 2010 The Actuarial Profession www.actuaries.org.uk
5/5/2010
17
Length
Pow
er
Other technical pricing advances
Example adjacencies
32© 2010 The Actuarial Profession www.actuaries.org.uk
Length
Pow
er
Example adjacencies
Other technical pricing advances
33© 2010 The Actuarial Profession www.actuaries.org.uk
5/5/2010
18
Other technical pricing advances
Vehicle Smoothing
• Delivers material benefit
34© 2010 The Actuarial Profession www.actuaries.org.uk
0
20,000
40,000
60,000
80,000
100,000
120,000
140,000
160,000
180,000
200,000
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
Less than -
50%
-50%
--45%
-45%
--40%
-40%
--35%
-35%
--30%
-30%
--25%
-25%
--20%
-20%
--15%
-15%
--10%
-10%
--5%
-5%
-0%
0%
-5%
5%
-10%
10%
-15%
15%
-20%
20%
-25%
25%
-30%
30%
-35%
35%
-40%
40%
-45%
45%
-50%
Mo
re than 5
0%
Veh
icle
Year
Rescale
d V
alu
e
Percentage change between proposed and current groop
Proposed Vehicle Group vs ABI 50 Group
Exposure Additional Ef fect above ABI Group
1.5x
Percentage change between modelled and current
Other technical pricing advances
Saddles
• Systematic review of interactions can be simplified into “saddle
spotting”
• This can be designed to work equally well at 2, 3, 4, 5, …
dimensions without losing process efficiency
• Can rapidly build list of possible interactions for inclusion in
models
35© 2010 The Actuarial Profession www.actuaries.org.uk
5/5/2010
19
Other technical pricing advances
Saddles
36© 2010 The Actuarial Profession www.actuaries.org.uk
Other technical pricing advances
Saddles
37© 2010 The Actuarial Profession www.actuaries.org.uk
5/5/2010
20
Other technical pricing advances
Saddles
38© 2010 The Actuarial Profession www.actuaries.org.uk
Saddles
Other technical pricing advances
39© 2010 The Actuarial Profession www.actuaries.org.uk
5/5/2010
21
Saddles
Other technical pricing advances
40© 2010 The Actuarial Profession www.actuaries.org.uk
Saddles
Other technical pricing advances
41© 2010 The Actuarial Profession www.actuaries.org.uk
5/5/2010
22
Saddles
Other technical pricing advances
42© 2010 The Actuarial Profession www.actuaries.org.uk
Saddles
Other technical pricing advances
43© 2010 The Actuarial Profession www.actuaries.org.uk
5/5/2010
23
The underwriting view:Geographic Spatial Smoothing
• Which geographic elements should over-ride spatial
smoothing?
– Motorways
– Rivers
– Boundaries
• Predictable characteristics:
– Theft near ports
– Windscreen claims in broader East Anglia
• What to do without exposure
• What competitors are doing geographically and why
44© 2010 The Actuarial Profession www.actuaries.org.uk
The underwriting view:Vehicle or Trade smoothing
• Virtual spatial smoothing is broadly equivalent to a traditional
underwriting approach
• Risk is assessed in a broader, more holistic way with
experience being developed more generally than on particular
trades or vehicles
• The approach leans toward partial implementation in terms of
underwriting rules
• Underwriting input will include:
– What to do without exposure
– What competitors are doing and why
45© 2010 The Actuarial Profession www.actuaries.org.uk
5/5/2010
24
The underwriting view:Saddles
• There should be an underwriting rationale for each interaction
adopted into risk cost models
• Underwriting will propose deep interactions that appeal:
– Teachers living in rural areas with more than one vehicle in
the household doing a low mileage?
• Underwriting will manage the balance between pricing
complexity and its ease of interpretation and maintenance
46© 2010 The Actuarial Profession www.actuaries.org.uk
The underwriting view:Saddles Deep interactions
47© 2010 The Actuarial Profession www.actuaries.org.uk
Subject
matter of
insurance
Occupation
& Age
Postcode
and socio-
demographic
• Have you ever wondered what “Occupation” was
doing in motor and home rating?
• Deeper interactions have always been part of the
underwriter’s armoury
• Consistency is the key
• Increased sophistication hasn’t delivered the
ability to spot the real risk from the fraud
• The case underwriter scrutiny is badly missed and
needs to be reinvented
Consistency?
5/5/2010
25
Better pricing needs better underwriting
Agenda
• Introduction
• Distribution
• Bodily Injury
• Technical pricing activities
• Conclusion
48© 2010 The Actuarial Profession www.actuaries.org.uk
Conclusion
We started off defining underwriting as ensuring risks are
consistent with pricing assumptions
• Technical pricing development is vital to general insurance
• Technical pricing developments are rarely assumption free and generally add
to the cumulative assumptions already in play
• We’ve shown lots of instances where pricing assumptions need to be
supported by underwriting
• Underwriting provides a context that allows technical pricing development to
be highly technical
• Insurers don’t necessarily need Underwriters but they do need Underwriting
• What would our new underwriting world look like?
49© 2010 The Actuarial Profession www.actuaries.org.uk
5/5/2010
26
Conclusion
What does this new underwriting world look like?
• Commercial motivation
• Micro and macroeconomic insight
• Risk expertise
• Insurance expertise
• Numerate
• Legal skills and knowledge
• Lateral thinking
• Project and relationship management
• Worldly
• Pragmatic
50© 2010 The Actuarial Profession www.actuaries.org.uk
An analogy from popular culture?
Questions or comments?
Expressions of individual views by
members of The Actuarial Profession
and its staff are encouraged.
The views expressed in this presentation
are those of the presenter.
51© 2010 The Actuarial Profession www.actuaries.org.uk