Actuarial pricing, capital modelling and reserving


Pricing Squad


Issue 6 -- September 2016

Welcome back to Pricing Squad

Pricing Squad is a newsletter for fellow pricing practitioners and actuaries in general insurance. Enjoy, and let me know your comments and ideas for future issues.

Today's issue is pushing the limits of predictive modelling into the realm of the super-human using GLM-free methods.


Are you a dog person or a cat person?

I can only capture demand elasticity differences for those factors which are actually captured in my data: age, vehicle type, location etc. It is impossible to capture differences by, say, policyholder's mum's hair colour.

Right?

Wrong.

What?

I like to maximise the information I extract from elasticity data. You can do this too, using GLM-free rate change impacts.

Example

In this Excel sample dataset (www.iwanik.co.uk/uploads/public/dog_cat_person_v1.xlsx) 10,000 quotes are equally divided between "dog people" and "cat people", a characteristic your company definitely does not collect.

Cat people are assumed to have elasticity 4 and dog people 2. Otherwise, both groups of pet owners are identical.

You can see in the Excel file that a 10% rate hike shifts the business mix towards dog people. Excel can predict this effect easily because it knows pet preference for each quote. Nothing unusual so far.

But now ignore the "pet preference" column and copy-paste raw quote data into the free Impact Express tool (www.iwanik.co.uk/app-impact-express/free_online_rate_change_tester_step1.php). Then run the analysis.

You can see that Impact Express also correctly predicts a mix shift towards "dog people". (This can be seen in changes of average burning cost, as I assumed different costs for "dog people" and "cat people" to make the effect detectable).

Magic.

The ability to perfectly capture elasticity differences is amazing, given that pet preference is not being fed to Impact Express!

Just imagine the competitive edge you could gain by correctly capturing mix changes for all non-collected customer characteristics.

How is this possible?

In case you missed the GIRO 2016 talk where the boring maths was explained, here is the short version.

GLM-free impacts do not require a GLM, so the normal limitations of a regression model do not apply.

Instead, the GLM-free rate change impacts are predicted by bootstrapping from historical experience, considering influencing by randomised price test. This way, even the most obscure sub-segment of your book will weigh on the impact prediction if it has historically had distinct elasticity.

What if cat people actually do make safer drivers?


Can we help you?

If you are interested in new pricing ideas to radically simplify your current analytical procedures and deliver reduced loss ratio quickly, or if you are simply looking for an actuarial contractor, please get in touch.

Thank you for reading, and have a great day, Jan Iwanik, FIA PhD


Copyright © 2016 Jan Iwanik, All rights reserved. You are receiving this email because you are subscribed to updates from www.iwanik.co.uk. We publish data and analysis for informational and educational purposes only. You can unsubscribe from this list by emailing us.