+ All Categories
Home > Documents > SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

Date post: 30-Dec-2015
Category:
Upload: eileen-roberts
View: 215 times
Download: 0 times
Share this document with a friend
Popular Tags:
27
SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013
Transcript
Page 1: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

SAM NLUR Working Group

Industry workshopQIS 2 to QIS 3

Johannesburg Country Club 13 June 2013

Page 2: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

2

Introduction

NLUR Workbook

Premium risk calibration

Reserve risk calibration

Natural catastrophe risk calibration

Man-made catastrophe risk calibrationo Hannes van Rensburgo Lisa Pines

(QIS 2 feedback on issues throughout)

Further issueso Premium definitiono Cash-back bonuseso Variable commission structureso Segmentation

Agenda

Page 3: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

3

Pillar 1 sub-committee request

Purposeo NLUR WG Update on changes to underwriting risk calculationo Facilitate STI industry discussion on underwriting risk issues o Strict time keepingo Methodology rather than numbers

NLUR WGo 31 memberso 14 qualified actuarieso 19 organisationso Many hourso Various sub-working groups:

FSB calibration team Workbook Premium risk Reserve risk Natural catastrophe risk Man-made catastrophe risk

Introduction

FSB Primary Insurers RI'ers & RI Brokers ConsultantsABSA Hannover Re DeloitteCGIC Guy Carpenter E&YHollard Willis KPMGM&F AON PWCNatsure DynamicModelOffi ceStandard Bank IACZurich GrailAfrica

Page 4: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

4

Separate NLUR workbooko Feedback from QIS 1o Completeness of reporting to FSBo Standard formula vs internal model

Issues from QIS 2 feedbacko Risk mitigation

o Complexo Not enough flexibilityo Allow calculations outside workbooko Etc

o Errorso Limited number of counterpartieso Lack of guidance

NLUR WorkbookQIS 2

Page 5: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

5

Identified errors fixed

Reinsurance structureso Allowing more counterpartieso Multiple NLUR workbooks to allow for different portfolios covered by

different structureso Allow for features such as aggregate limits, deductibleso Guidance on NP adjustmento Allowance on man-made and nat cats for XL then prop then XL then prop

NLUR WorkbookProposed changes

Page 6: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

6

Guidanceo Guidance manualo Case studies with accompanied workbook exampleso NP adjustmento Two workbooks – one not protected; button to transfer to protected with

error outputo Unprotected version circulated beforehando Scaling of nat cat charge when splitting exposures

Automated summary sheet to aid analysis of results; capital allocation

NLUR WorkbookProposed changes

Page 7: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

7

How much freedom do we allow in a standard formula?

Data availability

Complexity

Other?

NLUR WorkbookOutstanding issues

Page 8: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

8

Structure

Volume measures by segment

Proportional RI %’s per counterparty per segment

NP volatility adjustment factor – optional!

Premium risk calibrationQIS 2 recap

Page 9: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

9

Compulsory data request – end of 2012o By segment (more detailed than QIS 2)o By accident year/underwriting yearo Earned premiumso Paid and reported claims triangleso Gross and net of reinsurance

Data cleaning and quality checko FSB; anonymityo Error correction; follow-up with individual companieso Re-run calibrationso Import key parameters into calibration toolo Exclusion of some outliers (e.g. very low loss ratios)o Graph all loss ratioso Confirm net of cat

Premium risk calibrationData collection

Page 10: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

10

Premium risk calibrationAnalyses by segment – 4 methods

Method 1 Method 2 Method 3 Method 4 Assumed distribution of Loss ratios and fitting method

Normal (Least squares)

Lognormal (Maximum Likelihood Estimation)

Lognormal (Maximum Likelihood Estimation)

Lognormal (Maximum Likelihood Estimation)

µ Fit LR Mean by undertaking

Fit LR Mean by undertaking

Fit LR mean across all undertakings

Fit LR Mean by undertaking

σ Fit LR sd by undertaking

Fit LR sd across all undertakings

Fit LR sd across all undertakings

Fit LR sd across all undertakings

E(Loss) and Var(loss)

E(Loss) α Prem sd(Loss) α sqrt (Prem)

E(Loss) α Prem sd(Loss) α sqrt (Prem)

E(Loss) α Prem sd(Loss) α sqrt (Prem)

E(Loss) α Prem sd(Loss) α Prem

Overfitting? In general: Fits more parameters – Highest level of overfitting – leads to lower parameters

In general: Fits more parameters – Overfitting – leads to lower parameters

Fits fewer parameters – underfitting – overstates volatility (will always give larger parameters)

In general: Fits more parameters – Overfitting – leads to lower parameters

Allows for diversification credit by volume (Diversification credit exists if larger co’s have lower variability)

Yes (more weight to larger companies)

Yes Yes? No

ELIMINATE IF LR’s are not the same for industry (Graph B)

There is evidence that larger companies have smaller sd’s (Graph C)

Page 11: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

11

Default method 2 Motor personal lines and Engineering

If negative relationship, eliminate method 4

Premium risk calibrationAnalyses by segment – sigma vs premium – method 2 vs 4

Page 12: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

12

Default method 2 If graph shows spread, eliminate method 3

Premium risk calibrationAnalyses by segment – histogram of average loss ratios – method 2 vs 3

Page 13: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

13

Usually only method 2 Indication of appropriateness of lognormal assumption

Usually indicates lognormal is too optimistic in tail Practicality of using different distribution?

Premium risk calibrationAnalyses by segment – pp-plots of selected methods

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

M2_Paid

M2_Reported

Theoretical

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

M2_Paid

M2_Reported

Theoretical

Page 14: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

14

Premium risk calibrationAnalyses by segment – initial parameters

SA QIS 2 Method 2 - original Method 2 - Industry benchmno of insurers (callibration) Paid Reported

Average pd/rep

Prop PL 26 8.2% 6.2% 5.7% 5.9%Motor PL 23 8.2% 5.7% 5.5% 5.6%Motor CL 23 8.2% 6.4% 6.1% 6.3%Prop CL 21 8.2% 12.6% 14.9% 13.8%Liability CL 19 13.9% 9.3% 14.9% 12.1%M&T 16 14.9% 14.5% 14.5% 14.5%Health 15 4.0% 18.6% 9.6% 14.1%Engineering 13 8.2% 9.4% 14.9% 12.1%Misc Other 13 12.8% 6.9% 7.8% 7.3%Warranty 6 12.8% 15.6% 16.8% 16.2%SuretyComm 5 11.7% 27.8% 27.0% 27.4%Legal expenses 3 6.5% 8.9% 9.4% 9.2%Aviation 2 14.9% 9.9% 11.4% 10.7%Travel 2 12.8% 9.2% 7.7% 8.4%Crop 2 12.8% 102.6% 109.8% 106.2%SuretyRetail 1 11.7% 17.9% 7.3% 12.6%

Page 15: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

15

Impact of excluding single companies

Paid vs reported

3 methods:o Company specific DFso Industry DFs chain laddero BF proxy

Consistency with reserve risk method

Credibilityo Number of submissions o Number of accident years (data points)o Percentage of industry by premium volumeo Size of segmento Other?o Proposed methods

Premium risk calibrationAnalyses by segment – other considerations

Page 16: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

16

Structure

Volume measures by segment

Total RI %’s per counterparty per segment – from best estimate technical

provisions calculation

No additional NP volatility adjustment factor

Reserve risk calibrationQIS 2 recap

Page 17: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

17

Compulsory data request – end of 2012

Data cleaning and quality checko Similar to premium risk calibration

Method typeso Premium risk type methods (3)o Triangulation type methods (3)

Similar considerations to premium risk calibration

Consistency between premium and reserve risk methods

Reserve risk calibrationData collection and analysis

Page 18: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

18

Natural catastrophe risk calibrationWorking group structure

Non

Life

Und

erw

riting

Ris

k Natural Catastrophe

Catastrophe Task Force

Catastrophe Modellers

Premium Risk

Reserve Risk

Man-made Catastrophe Risk

Workbook

Page 19: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

19

Natural catastrophe risk calibrationBackground

QIS1

• 2009 J statements adjusted for errors, Sum Insured by CRESTA Zone• Data modelled by cat modellers to get an industry view of the 1 in 200 year loss • Modelled factors were chosen by voting process within group

QIS2

•Exposures from industry template sent out in 2011, Sum Insured by Postal Code•EQ- Data modelled by cat modellers to get an industry view of the 1 in 200 year loss•EQ Motor- factors were based on model output for EQ Residential LOB and consensus view •HL- factors were based on consensus view by cat modellers

QIS3

• To be determined

Page 20: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

20

Postal codes were grouped in 19 new zones

Exposure data received from industry and modelled

Model output: Simulated losses by LOB and zone for 50 000 events

QIS5 Formula applied

Natural catastrophe risk calibrationQIS 2 calibration

rxc

cZONErZONEcrLOB WTIVWTIVAGGCAT ,,, **

rxc

cLOBrLOBcrEQEQ WTIVWTIVAGGQCAT ,,, **

Page 21: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

21

Market Factor: Rank total losses and determine the 1 in 200 year loss, disregarding multiple events in the same year

LOB/Zone relativities: Express the share of the 1 in 200 year loss for zone (x) over the exposure for zone (x)o Zone (x)’s share is calculated as the Average Loss for zone (x)/ Average Loss

over all zones over a range (100 observations above and below the 1 in 200 year loss)

Aggregation Matrix: Solve the square root formula between each pair of zones/LOBs

Rescale relativities such that Market Factor * Exposure =

Natural catastrophe risk calibrationQIS 2 calibration

rxc

cLOBrLOBcr WTIVWTIVAGG ,,, **

Page 22: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

22

Model vs. Standard Formula o FSB sent scaled exposures for each co that submitted QIS2o Recalculate standard formula and compare to the modelled 1 in 200 year

loss per company

1 in 200 year loss per co

Natural catastrophe risk calibrationPost QIS 2 results

Base Scenario Count(-70% , -60%) 1

(-50% , -40%) 2

(-40% , -20%) 0

(-20% , -10%) 3

(-10% , 0%) 14

(0% , 10%) 15

(10% , 20%) 3

38

Max (Base Scenario, Alternative Scenario) Count

(<-40%) 1

(-40% , -20%) 0

(-20% , -10%) 4

(-10% , 0%) 15

(0% , 10%) 15

(10% , 20%) 338

Page 23: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

23

Natural catastrophe risk calibrationPost QIS 2 results

Mkt Data (%) Calibration Data (%)0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Comparison of Mkt & Calibration Data(%)

Residential Buildings Commercial Buildings Contents Engineering Motor

Page 24: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

24

Natural catastrophe risk calibrationPost QIS 2 results

Residential Buildings Commercial Buildings Contents Engineering Motor

Comparison of Mkt & Calibration Data

Market Data (FSB) Calibration Data incl BI

Page 25: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

25

Horizontalo Template sent out to collect natural catastrophe claims over last 20-30

years from biggest companies, across all perilso FSB will collate and scale data to reflect industryo WG will analyse data and fit a curve to calibrate the 1 in 10 and 1 in 20 year

events

Verticalo Update exposures and re-run model?o Additional 2 CAT models to run and compare model results too Keep factors unchanged

Natural catastrophe risk calibrationQIS 3

Page 26: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

26

Premium and reserve risk segmentationo PL/CL split vs combinedo Aviation vs Marine and transport (vs pure transport)o Miscellaneous classeso Test alternatives in QIS3? – data split, FSB tests afterwardso Other segmentation concerns from industry – for discussion document

Volume measure – premium definition? Method 2 – volume measure square root – inconsistent with standard formula Allowance for loss absorbency of:

o Cash-back bonuseso Sliding scale/profit commission arrangements, both reinsurance, UMA’s,

cells, other?o Limited risk transfer

Allowance for additional risk caused by above? Lapse risk Contract boundaries Premium recognition

Premium and reserve riskFurther issues

Page 27: SAM NLUR Working Group Industry workshop QIS 2 to QIS 3 Johannesburg Country Club 13 June 2013.

27

Factor based method – not as optional as industry might think it is

Postal code information o multiple risk addresses under one policyo solution: expected that industry addresses this, rather than changing

formula to allow for poor quality?o How should this be addressed? FSB should enforce at broker level?o Working group still to test impact of alternative zoning vs Cresta zones

Natural catastrophe risk calibrationData issues


Recommended