In this blog, Andrew Rear, Chair of the Mortality Assumptions in Pensions Working Party, explains what prompted the IFoA Risk Alert on Disclosure of Information relating to models
Most actuaries use models most days. We put a lot of care and attention into our models, and they are an important source of competitive advantage. If I develop a model which is a better representation, or shines new light, on an issue, then that is my intellectual property and I am probably hoping I can gain some commercial benefit from it. Because my model is better than your model.
But what if we are on opposite sides of a transaction where we are both modelling the same thing and you don’t think my model is necessarily better? Indeed, you’re not even sure it is a reasonable approach. Your client asks you to help bridge the gap, so you ask me for some details of my model so you can understand what’s going on. I’m reluctant to provide that because you are asking about my intellectual property. Intellectual property that I have spent time and money developing, to give me a competitive advantage over actuaries like you. And now you want to see my work, for free?
But if I won’t show you my model, then what are you to say to your client, other than ‘I don’t know and don’t trust this model (though I’ve got no actual evidence it is wrong)’. That’s hardly satisfactory, for you, your client, or indeed for the profession’s image as a whole. In practice, actuaries in this situation usually find a way through, often through expert modelling actuaries talking to each other.
The Risk Alert on model disclosure is the result of a working party established by the Regulation Board in response to a situation where actuaries in several firms had not been able to ‘find a way through’. We set out to answer the questions of whether there is a legitimate responsibility to disclose to another actuary, and if so, what level of disclosure is appropriate.
Based on counsel’s opinion, we found that a responsibility to disclose information about a model to another actuary does arise if the principals of the two actuaries intend to come to a consensus or mutual understanding of the item being modelled. For example, if pension scheme trustees ask their actuary to confirm that it is reasonable to accept a funding calculation proposed by an employer, the employer’s actuary has a responsibility to facilitate that, unless the employer herself refuses (in which case the trustees probably should not use the calculation proposed). Once the actuaries then do start communicating with each other, they have a professional responsibility to do so in a way which is accurate, not misleading and contains an appropriate level of information.
That right doesn’t extend for ever: there needs to be some legitimate ‘chain of model reliance’ linking the two actuaries. This is not the case in every transaction involving competing models: for example If I believe that something you own (a company, a book of liabilities) is worth more than you believe it is, then we have a basis for a transaction, not an automatic right for us to interrogate each other’s models. There are also legitimate limits to the required disclosure: if we don’t protect intellectual property then there would be no incentive to innovate and we would all still be using standard tables from the green book. The working party tested where these limits lie with some real-life disclosure Q&A sessions on some proprietary models, documenting the outcome as guidelines in the risk alert. The guidelines should allow your fellow actuary to understand your model, why it is reliable and why it produces the results it does; it should not allow them simply to recreate it using your intellectual property.
The Risk Alert is not new regulation. It sets out the Regulation Board’s understanding of current regulation, and of current practice in the vast majority of cases. Hopefully what it does do, however, is provide a helpful benchmark for actuaries to use if they find themselves unable to find a way through to an understanding of each other’s position. It doesn’t change the incentive for actuaries to push the boundaries of actuarial science to find better and better ways to model any scenario they face. Because my model is, to the best of my professional knowledge, better than yours.