pricing game, 100% data sciences
TRANSCRIPT
![Page 1: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/1.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Actuarial Pricing Game (Season 2)A. Charpentier (Université de Rennes 1 & Chaire ACTINFO)
with R. Élie, J. Jakubowicz & F. Picard
Paris, 100% Data Science, November 2016.
http://freakonometrics.hypotheses.org
@freakonometrics 1
![Page 2: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/2.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
A Game with Rules... but no Goal
Warning 1
For participants: forget about individuals results, focus on the global picture
Warning 2
Difficult to pool the results together (some mistakes are probably mine...)
Warning 3
The ‘pricing game’ has rules.But no goal was ever mentioned.Every player played according to his/her goals.
@freakonometrics 2
![Page 3: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/3.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Context (and Apologies)
Two datasets : a training one, and a pricing one (without the losses in the later)
Step 1 : provide premiums to all contracts in the pricing dataset
Step 2 : (somme) premiums provided by competitors were given to players
Warning 4
Motor insurance, but on fleats : one ‘contract id.’, several vehicles
Warning 5
Technical issue : the exposure was still in the pricing dataset
Warning 6
!normalization:n∑
i=1π(j,1)(xi) =
n∑i=1
`i = 2.5M︸ ︷︷ ︸loss ratio = 100%
andn∑
i=1π(j,2)(xi) = 2.35M , ∀j
@freakonometrics 3
![Page 4: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/4.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Part 1: Standard Pricing Game, a Global OverviewConsider an ordered sample {y1, · · · , yn} of in-comes, with y1 ≤ y2 ≤ · · · ≤ yn, then Lorenzcurve is
{Fi, Li} with Fi = i
nand Li =
∑ij=1 yj∑nj=1 yj
We have observed losses `i and premiumsπ̂(xi). Consider an ordered sample by themodel, see Frees, Meyers & Cummins (2014),π̂(x1) ≥ π̂(x2) ≥ · · · ≥ π̂(xn), then plot
{Fi, Li} with Fi = i
nand Li =
∑ij=1 `j∑nj=1 `j
@freakonometrics 4
![Page 5: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/5.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Part 1: Standard Pricing Game, a Global Overview
Dispersion of the premium, per company, π(1,j)(xi), j ∈ {1, 2, · · · , 14, 14∗}.
@freakonometrics 5
![Page 6: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/6.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Market Competition
Decision Rule: the insured selects randomly from the three cheapeast premiumcccccccccccc ccccccccccccc cccccccccccc ccccccccccccc cccccccccccc ccccccccccccccccccccccccc ccccccccccccc cccccccccccc ccccccccccccc
Ins2 Ins3 Ins4 Ins5 Ins6 Ins7
787.93 706.97 1032.62 907.64 822.58 603.83
170.04 197.81 285.99 212.71 177.87 265.13
473.15 447.58 343.64 410.76 414.23 425.23
337.98 336.20 468.45 339.33 383.55 672.91
@freakonometrics 6
![Page 7: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/7.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Part 1: Standard Pricing Game
Market shares of the 12 insurance companies
@freakonometrics 7
![Page 8: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/8.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Part 1: Standard Pricing Game
Premium variability (overall) and loss ratios of the 12 companies
@freakonometrics 8
![Page 9: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/9.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Part 2: Pricing Game with Additional Information
@freakonometrics 9
![Page 10: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/10.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Part 2: Pricing Game with Additional Information
New Lorentz curves:order policies according to the premium
π(2,j)(x1) ≥ π(2,j)(x2) ≥ · · · ≥ π(2,j)(xn)
then plot
{Fi, Li} with Fi = i
nand Li =
∑ij=1 `j∑nj=1 `j
@freakonometrics 10
![Page 11: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/11.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Part 2: Pricing Game with Additional Information
Premium variability (overall) and loss ratios of the 12 companies
@freakonometrics 11
![Page 12: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/12.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Part 2: Pricing Game with Additional Information
Premium variability (overall) and loss ratios of the 11 companies
@freakonometrics 12
![Page 13: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/13.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Part 2: Pricing Game with Additional Information
Premium variability (overall) and loss ratios of the 12 and the 11 companies
@freakonometrics 13
![Page 14: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/14.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Smaller Market Competition (3 companies) : Step 1
@freakonometrics 14
![Page 15: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/15.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Smaller Market Competition (3 companies) : Step 2
@freakonometrics 15
![Page 16: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/16.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Smaller Market Competition (3 companies)
Decision Rule: out of 3 companies the insured selects (i) the cheapest premium(ii) randomly from the two cheapeast premiums, with probabilities 2/3(cheapest) and 1/3 (second cheapest)
Ins4 Ins5 Ins6 Ins4 Ins5 Ins6
787.93 706.97 1032.62 787.93 706.97 1032.62
170.04 197.81 285.99 170.04 197.81 285.99
473.15 447.58 343.64 473.15 447.58 343.64
337.98 336.20 468.45 337.98 336.20 468.45
@freakonometrics 16
![Page 17: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/17.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Smaller Market Competition (3 companies)
Step 1: market shares and loss ratios
@freakonometrics 17
![Page 18: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/18.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Smaller Market Competition (3 companies)
Step 2: market shares and loss ratios ((i) and (ii))
@freakonometrics 18
![Page 19: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/19.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Smaller Market Competition (3 companies)
Loss ratios, Step 2 vs. Step 1 ((i) and (ii))
@freakonometrics 19
![Page 20: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/20.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Behaviors of Insurance, Local Perspective: Ins4
Insurer 4 has informations about premiums provided by In-surer 5 and Insurer 6 (after Step 1).Only 50% of the premiums were observed.
Ins 5. vs. Ins 4. Ins 6. vs. Ins 4. Step 2 vs. Step 1.
@freakonometrics 20
![Page 21: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/21.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Behaviors of Insurance, Local Perspective: Ins5
Insurer 5 has informations about premiums provided by In-surer 4 and Insurer 6 (after Step 1).Only 50% of the premiums were observed.
Ins 4. vs. Ins 5. Ins 6. vs. Ins 5. Step 2 vs. Step 1.
@freakonometrics 21
![Page 22: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/22.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Behaviors of Insurance, Local Perspective: Ins6
Insurer 6 has informations about premiums provided by In-surer 4 and Insurer 5 (after Step 1).Only 50% of the premiums were observed.
Ins 4. vs. Ins 6. Ins 5. vs. Ins 6. Step 2 vs. Step 1.
@freakonometrics 22
![Page 23: Pricing Game, 100% Data Sciences](https://reader031.vdocuments.site/reader031/viewer/2022020410/5881f42b1a28abb4748b52fb/html5/thumbnails/23.jpg)
Arthur CHARPENTIER, Pricing Game 100% Data Science - Institut des Actuaires 2016
Behaviors of Insurance, Local Perspective
Very different behaviors, for various insurance companies
Ins. 2 Ins. 13 Ins. 14
@freakonometrics 23