Reserving in High Inflation. General Considerations by Alejandra Nolibos Specific example by Alejandro Ortega Presented at CLRS in Atlanta, September 11,2015 Argentina Auto Case Study

1. Reserving in High Inflation
CLRS
Atlanta, Georgia
September 11, 2015
Alejandra Nolibos, FCAS, MAAA
Alejandro Ortega, FCAS, CFA

2. Presenting Today
Alejandr@’s
Alejandra Nolibos, Towers Watson
Alejandra leads Towers Watson’s professional excellence team for North America financial reporting, with focus
on statutory issues . As a consultant, Alejandra leads and serves as advisor on complex engagements involving
national and global organizations.
Alejandro Ortega
Alejandro was most recently the Chief Actuary for AIG’s Latin America operations. He lead the reserving and
supported capital management in 15 countries. He is now focused on Actuarial Education.

3. Inflation and reserves – Why care?
Increases in inflation pose a serious risk to companies writing long tailed
business (especially excess) where inflation acts through the payment date
For self-insured corporations, changes in inflation would impact where
retentions are set (future) and breached (past)
Multinationals are interested in exposure to inflation changes around the
world and other related effects (e.g. FX, asset prices)
Extremely important in economies where inflation seems to be picking up
(e.g., Brazil)
Explicit recognition of inflation in reserve risk projections is a must for
economic capital modeling
…
towerswatson.com
2

4. U.S.: implications of the prolonged low inflation environment
for the current estimation of unpaid claims liabilities
 Unpaid claims liabilities have been running off in a persistent low
inflationary environment
 Typically, a return to longer term inflation levels would affect the runoff of
liabilities — even if gradual
 A.C.E. is an expected value over a range of reasonably possible
outcomes
 Should a return to longer term inflation levels be considered in that
range?
towerswatson.com
3

5. A related topic is the persistence of other benign trends
For many segments, claims development trends have been
consistently favorable for several years
 Here too a shift — even if gradual — would affect the runoff
 Very specific to each book
 Harder to articulate how likely
 Consider “social/residual inflation”
 Benefit level changes
 Judicial/legislative environment
 Favorable effects of claims cost control
 Utilization
 …
 May be very hard to quantify
towerswatson.com
4

6. Actuarial analyses should include thoughtful consideration of
the effects of past and future inflation
Is future inflation a driver of
the runoff of reserves?
How do A.C.E.s/loss
distributions reflect the impact
of inflation?
Have I considered this driver
in my assessment of reserving
risk? Am I communicating
clearly on the topic?
towerswatson.com
5
Key questions

7. Is future inflation a driver of
the runoff of reserves?
How do A.C.E.s/loss
distributions reflect the impact
of inflation?
Have I considered this driver
in my assessment of reserving
risk? Am I communicating
clearly on the topic?
towerswatson.com
6

8. Is future inflation a driver of the runoff of reserves?
Likely driver for long tailed classes
• Where payment amounts depend on environment at the time of payment
• Workers compensation medical
• Casualty
• Excess layers
Typically not a material issue where
• Claims resolve quickly
• Auto Physical Damage
• Claims costs depend mainly on occurrence date
• D&O
• Workers compensation indemnity (if no COLA)
Don’t forget
• Reinsurance
• Effects on losses in AADs, hitting corridors and retentions, exhaustion of contract limits
• ULAE – especially if estimated based on claim counts
towerswatson.com
7

9. How sensitive are reserves to changes in inflation?
Creating reserve estimates based on different inflation assumptions
 Loss development parameters using longer and shorter term and ex-
ante averages
 Explicit testing of inflated cashflows
 “+1% test”
 Applying methods that explicitly recognize inflation changes
towerswatson.com
8

10. Is future inflation a driver of
the runoff of reserves?
How do A.C.E.s/loss
distributions reflect the impact
of inflation?
Have I considered this driver
in my assessment of reserving
risk? Am I communicating
clearly on the topic?
towerswatson.com
9

11. How do A.C.E.s/loss distributions reflect the impact of
inflation?
 An A.C.E. represents an expected value over a range of reasonably
possible outcomes
 Should a return to longer term inflation levels be represented in that range?
 In practice
 Estimates based on the assumption of continuation of recent trends typically
would not reflect scenarios under higher inflation environments
 May be reflected in “high” estimates in a range
towerswatson.com
10

12. Sounds great, but how does one
 Select parameters that reflect the right levels of inflation?
 Estimate the past inflation embedded in the data to on-level with?
 Project future inflation?
Consideration
of inflation
How?
Implicit • Projections based on parameters that attempt to
conceptually reflect some long-term average level of
inflation
Explicit • Step 1: Experience on-leveled for historical inflation
• Step 2: On-level estimates/cashflows developed
• Step 3: On-level cashflows inflated
How to incorporate impact of inflation in A.C.E.s?
towerswatson.com
11

13. Methods for recognizing inflation vary in complexity and
effectiveness – some are more transparent than others, too
towerswatson.com
12
Method Steps Example
Traditional
methods
• Project payments based on raw
history
• Produce cashflows
• Paid loss development
Aggregate
inflation-
adjusted
methods
• Project on-level payments, or
payments per claim (which are
combined with projections of closed
claims)
• Generate on-level cashflows
• Inflate cashflows
• Inflation-adjusted ACPC
• Triangle GLM
Granular
methods
• Use individual claims information and
GLMs to predict on-level amounts
paid per claim by maturity
• Combine with claim counts
projections to generate cashflows
• Inflate cashflows
• Taylor & McGuire

14. Case study: Argentina
13
towerswatson.com

15. Mexico
Largest Insurance Markets
Colombia
Ecuador
Brazil
Chile
Argentina
Peru

16. Competition
Multinationals
Liberty
Ace
AIG
QBE
Zurich
Mapfre
Santander
RSA
Axa
Generali
Allianz
HDI
GNP (Mexico)
La Patronal (Argentina)
Itau (Brazil)
Local Insurers

17. Products
Consumer
 Personal Auto
Commercial Auto
(Fleets)
 Personal Property
 Extended Warranty
 Travel Insurance
 A&H Products (eg. Cancer,
AD&D)
Primary Casualty
Excess Casualty
Property
Energy (Oil production)
 Marine (inland/ocean)
Surety
Trade Credit
Management Liability
Professional Liability
Commercial

18. Distribution
 Alpha Brokers (Marsh, Aon, Willis)
 Second Tier Brokers
 Local / Small Brokers
 Agencies
 Banks
 Utilities
 Online
 Direct to Consumer

19. Size of Insurance Market in Latinamerica
2013 GPW $160 billion

20. Inflation
Low Inflation High Inflation
Short Tail US First Party Auto
Mexico Auto
Mexico Casualty
Venezuela – All
products
Argentina – Personal
Property
Long Tail US Casualty
Europe Casualty
Europe Financial
Lines
Argentina Auto
Argentina Casualty
Brazil Casualty

21. Venezuela Inflation
80%
60%
40%
20%
0%
TradingEconomics.com | Banco Central de Venezuela
2006 2008 2010 2012 2014
Claims that pay in 3‐4 months are not going to
have a large deviation due to inflation in that
short period

22. Inflation
Low Inflation High Inflation
Short Tail US First Party Auto
Mexico Auto
Mexico Casualty
Venezuela – All
products
Argentina – Personal
Property
Long Tail US Casualty
Europe Casualty
Europe Financial
Lines
Argentina Auto
Argentina Casualty
Brazil Casualty

23. Historical Inflation
Argentina
2001 Currency
Crisis
Government Reporting no longer reliable

24. High Inflation – Long Tail
• Auto third party bodily injury
• First party is short tailed
• Inflation makes the tail even fatter
12+

25. Assumptions of Chain ladder
Thomas Mack
1. Expected incremental losses are proportional
to losses reported (paid) to date
2. Losses in AY are independent of losses in other
accident years
3. Variance of incremental losses is proportional to
losses reported to date
• High and changing inflation produces CY effect
• Growth in litigiousness also a CY effect
• Assumptions 1 and 2 are violated

26. Assumptions of Chain ladder
•Chain ladder implicitly takes the inflation in the
triangle and forecasts from there
• When inflation is changing – this is not appropriate
• We will end up with a methodology that allows us to
forecast different levels of inflation

27. Estimating unpaid losses
• Adjust Paid Triangle for Inflation
• Adjust Incurred Triangle for Inflation
• Paid Only Triangle
• Average Severity to Date
• Future Closed Paid Claims x Future Severity
Closed Paid Claims Severity
X =
Unpaid Losses

28. Fisher Lange
• Closed claims are easy to estimate
• Allows different assumptions for future inflation (and
interest)
•Generates squared triangle of cashflows
• Sensitivity testing vs. case reserves
Closed Paid Claims Severity
X =
Unpaid Losses

29. Closed Claims
Forecast the Following
• Newly Reported Claims at each age
• % of Claims Closed Without Payment (CWP)
•% of Claims Closed With Amount (e.g., paid)
Closed Paid Claims

30. Severity
Underlying Components of Severity:
• % Disability awarded by the Court (similar to WC)
• Cost of a point of disability in each jurisdiction
•(2,500 – 4,000 pesos)
• The final cost of the claim is proportional the product of
these two
• Four General Categories of a Claim:
• Indemnity
• Treatment Expenses
• Court and Attorney Fees
• Interest and Inflation
Severity

31. Severity
Severity
Interest Costs
•In addition to the base cost of the claim, the insurer must pay interest
from the date of the accident
• A claim occurring in 2010, and closing in 2015, we would pay 5 years
of interest
Inflation (Calendar Year Effect)
• The base cost of this claim is based on the Cost per Point in 2015 –
not, 2010
By waiting one more year to close it in 2015
• We pay an additional year of interest (~12%)
• Cost of a Point is also increased (~9%)
• Total cost of claim goes up about 22%

32. Severity
Severity
Forecasting Severity
• Forecast Severity on the Diagonal
• Forecast Down the Triangle using Inflation (CY Trend)
• Reasonability Check – going Across the Triangle for
Interest, and Development Year Trend

33. Severity
Severity
AY 12 24 36 48 60 72 84 96 108 120 132 144
2002 35 47 124 28 44 55 38 110 265
2003 57 144 51 24 37 127 55 107 241 292
2004 29 50 64 75 95 140 89 221 217 265 321
2005 10 18 55 68 103 74 70 164 193 238 291 353
2006 9 19 65 74 101 117 162 175 213 262 320 388
2007 11 17 43 70 95 182 155 193 234 288 352 427
2008 9 19 41 69 147 144 174 212 257 317 388 470
2009 7 20 49 73 128 158 191 233 283 349 426 517
2010 11 20 58 86 141 174 210 257 311 384 469 569
2011 10 24 65 95 155 192 231 282 342 422 516 626
2012 11 28 71 104 171 211 254 311 377 465 568 688
2013 13 30 78 114 188 232 280 342 414 511 624 757
Historical Severity (On-level)
Selected Diagonal Severity
Forecast Severity (Inflated)
All scaled by a factor
Pesos (000)

34. Unpaid Losses
Reasonability Checks are Performed
• Compare Ultimate Losses to Prior Analysis
• Look at Loss per Exposure across accident years
• Compare Unpaid Losses to Case Reserves
• This method does not calculate IBNR, but rather Unpaid
Losses
Unpaid Losses

35. January 2014
•The Peso was losing about 7% per year against the
USD from 2009-2011; or about 0.5% per month
• 2012: lost 1% per month
•2013: lost 2% per month
•2014 – January: 19%
•Short Term Bonds went from 18% to 25%
•This caused us to revise the future inflation and
interest assumptions
•The model allowed us to immediately have an
estimate of Unpaid Losses

36. High Inflation Environment
• Argentina has additional complications due to changing
legal environment
• High inflation is typically associated with a weak
currency, and changing interest rates
• Sometimes it is associated with social trends (e.g.,
higher litigiousness)
• Understanding the underlying drivers of Claim Costs is
key
• Fisher-Lange allows you to forecast different levels of
inflation and interest
• Great tool for sensitivity testing

37. Beyond Argentina
36
towerswatson.com

38. The challenge of “predicting” past inflation
Traditional methods

What is the level of inflation implied by selected development factors?

Fundamental in assessing the future inflation implied by the projection

Consider history of relevant inflation indices

Tricky when using benchmarks – must understand source and process
Methods that use inflation explicitly

What is the appropriate inflation index (or indices) to use in on-
leveling?

Parameterizing based on published inflation indices for specific cost drivers
of a line of business (e.g. medical cost, construction costs, wage, etc.)

What is the inflation embedded in the data?

Reviewing triangles of average paid amounts

GLMs using individual claims data (normalizes for other effects)

GLMs using aggregate development data
towerswatson.com
37

39. -2.0%
0.0%
2.0%
4.0%
6.0%
8.0%
10.0%
U.S. Inflation Rates
CPI Medical CPI Average CPI Average Medical CPI
Calibrating traditional methods:
What experience periods would typically reflect a low inflation
environment?
towerswatson.com
38
High LowModerate
Source: bls.gov

40. Detecting past inflation: the Argentina motor case study
Background
 We know that macro inflation is an important factor in Argentina
 We have been performing client work using various inflation-adjusted
techniques
 Including granular modeling such as in the case study
The case study
 TW has access to granular claims data reported to Argentina motor
bureau
 Large volume of data (~290k closed claim counts)
 Using detailed claim level data allowed for GLMs that normalize for
changes in mix of other key variables
 We also tested aggregate triangle GLMs
 Part of the output was a historical calendar period trend index
towerswatson.com
39

41. Detecting past inflation: the Argentina motor case study
Identified predictive variables
 Payment period
 Age of driver
 Gender
 State
 Age of vehicle
 Duration
 Company
 Seriousness
 Location
 Litigation
 Measure of inflation
 Other cost drivers
towerswatson.com
40
In many cases, “operational time”, i.e., the order in which a claim is closed
relative to other claims in the same “class” is a better measure of the
relative severity than duration (T&M)

42. Detecting past inflation: the Argentina motor case study
 In addition to using GLMs applied to individual claims information, we used
aggregate triangle GLMs
 Indications produced by individual claims GLM were more in line with
expectations and average payments per claim observed in triangles
towerswatson.com
41
0
50
100
150
200
250
300
350
6/1/2008 12/1/2008 6/1/2009 12/1/2009 6/1/2010 12/1/2010 6/1/2011 12/1/2011 6/1/2012 12/1/2012 6/1/2013 12/1/201
Implied inflation index
Emblem ResQ AY + CY Unofficial Official

43. Another challenge: reflecting future inflation
Selecting future inflation assumptions
 Project future inflation as a function of wage, price and medical
inflation available from proprietary ESGs and other public sources
 Consider adjustments for social/residual inflation if applicable and not
captured elsewhere
 A simpler approach or reasonability check may be to consider the spread
between expected returns and inflation
 For traditional methods, assuming a spread over implied inflation
assumption
 Important to consider inflation assumptions underlying other
calculations
 E.g., capital modeling, pricing, financial projections
towerswatson.com
42

44. Is future inflation a driver of
the runoff of reserves?
How do A.C.E.s/loss
distributions reflect the impact
of inflation?
Have I considered this driver
in my assessment of reserving
risk? Am I communicating
clearly on the topic?
towerswatson.com
43

45. Assessing and communicating inflation-related reserving risk
Qualitative
assessment
• Type of business
• Limits/retentions
• Limitations of
analysis
• …
Quantification
• Sensitivity
• Stress-testing
• Benchmarking
• Stochastic
modelling
• …
towerswatson.com
44

46. Inflation effects can make communication tricky
Background: why you are talking about inflation
 Recent years’ favorable/unfavorable claim cost environment and likely drivers
 Relevant inflation rates (CPI, tort, etc.)
 How sensitive reserves are to inflation, which classes are most affected and why
Mechanics: what you did and why
 What the low inflation environment means to the methods and assumptions you used
 How you have dealt with those effects
 Whether and how your projections (implicitly or explicitly) reflect outcomes where
inflation is higher and why
Findings: what drives the uncertainty
 Uncertainty around whether or when a shift in inflation may occur
 Whether you have explicitly quantified the potential impact of rising inflation rates
 Quantification of risk
Transparency
 Sources of assumptions and why you think they are relevant
 Ties to other calculations (capital modeling, pricing, etc.)
towerswatson.com
45

47. The bottom line…
In many cases, changes in inflation (including a
return to longer term inflation levels) present a
risk to reserve run-off
A number of actuarial methods allow the actuary
to assess this risk and consider in his/her
projections, but their application may be tricky
Assessment and communication to the audience
of these risks and uncertainties are key
towerswatson.com
46