Actuarial pricing, capital modelling and reserving

Pricing Squad

Issue 26 -- July 2018

Welcome to Pricing Squad

Pricing Squad is the newsletter for fellow pricing practitioners and actuaries in general insurance.

Today's issue invites you to the 2nd Annual Pricing in Personal Lines Insurance conference and once more warns you about reckless use of GLMs.

2nd Annual Pricing in Personal Lines Insurance

You are invited to the Pricing in Personal Lines conference. Enhance personal lines pricing models with data analytics, machine learning and AI to improve predictability and increase profitability.

17-19 September 2018, Canary Wharf, London.

Sign up if you want to hear all the fantastic speakers, the least amongst whom is my modest self. I shall run a workshop called Trust is good, control is better - how to quickly sense check a complex pricing model and challenge its creator.

The workshop combines an interactive session and a demonstration of free high-level modelling tools. We are providing a workbook and will discuss various problems as a group.

Amongst its many virtues, the presentation features this beautiful dragon.

GLM Utopia

Overheard in a pricing meeting:

- The new model is vastly more predictive than the current algorithm - says the consultant.
- Yes - agrees Adam, the pricing manager. Everyone else nods in agreement.
- So we should adopt it.
- But it is very different from the existing model. We will struggle to justify such changes.
- Well Adam, think about it this way. If you were starting the business today from scratch, you would adopt the new wholesale, correct?
- Yes.
- Surely, if this is your first choice you should be implementing the model now as well.
- I do not think that follows. For example, I am married but I would not choose to marry again. It does not mean that my best strategy is to divorce today.

Good point Mr Pricing Manager!

Replacing current rates with a completely new rating algorithm in the name of a GLM utopia = revolution in rating tables.

It will be two years before the dust settles and Adam can ascertain the impact. The odds are that by then Adam be in square one. Perhaps with a new clever consultant.

Instead, he can challenge the modeller to pick one (yes, just one) rating factor with the highest benefit. Let the modeller explain their choice. Implement this change. See what happens. Repeat.

This way Adam will a) help the modeller to identify and prioritise data forces inside the GLM; b) force the modeller to articulate a small number of rationales which can be openly challenged by everyone; c) achieve about 50% of the total benefit from that single factor alone; d) enable himself to evaluate the change quickly.

Be sceptical about utopias.

Copyright © 2018 Jan Iwanik, All rights reserved. You are receiving this email because you subscribed to updates from We publish data and analysis for informational and educational purposes only. You can unsubscribe from this list by emailing us.