6 Myths About Reinsurance Portfolio Optimization

In an ever-changing reinsurance industry ripe with an influx of alternative capital and new competition, reinsurers have had to identify new ways to boost the profitability of their portfolios. One approach to enhancing the strategic planning process—portfolio optimization—assists underwriting and executive teams in determining how much share to allocate to each individual contract and in constructing reinsurance portfolios that balance maximizing returns while minimizing risk. In this post, let’s confront 6 myths about portfolio optimization and debunk them once and for all!

1) All optimization methodologies are equal.

There are various approaches to optimization, some of which are more effective than others. Linear methods, for example, optimize for one goal or constraint at a time (i.e., “maximize premiums”) before moving on to optimize for the next. But multi-objective, multi-constraint optimization can simultaneously incorporate multiple objectives such as maximizing expected return while minimizing for VaR and TVaR. This approach yields better sets of results that are optimal for your specific objectives and constraints.

2) Portfolio optimization can’t take into consideration special criteria your organization may have.

Optimizers are able to go beyond objectives setting—they can also incorporate what you would call “constraints,” and they can handle many of them. So, let’s say you want to ensure that Florida is no more than 20% of your gross portfolio. Or perhaps you have several key clients for which adjusting the contracts in the coming year for various strategic reasons is not an option. Or at the very least you wish to put constraints on how many changes to make to a layer. You can do all of these. Or maybe you have a target profit margin that must be maintained and you are willing to take on a certain level of risk in most circumstances. Optimizers do this by enabling you to set both the macro level objectives, as well as the more nuanced micro-constraints, thus letting you run the machine learning algorithms against the specific criteria you are looking to optimize for.

3) Portfolio optimization can only be done to optimize to a single view of risk.

There are a couple of approaches that can be applied here. You could run an optimizer against losses from one catastrophe model, as well as your own internal models, and then compare the results. Or you could blend your model output results beforehand, and then run those results through the optimizer for a combined optimized portfolio.

4) The recommendations of the optimizer are not realistic.

This is true only if you set unrealistic objectives and constraints. For example, if you have a ceiling for how much to participate in a deal because of the price being offered or if a broker has said they can only give you a certain share, you need to capture these constraints in the optimizer. For example, if know that you cannot write more than 5% in a deal, set 5% as your maximum for the layer in the optimizer and then the optimizer won’t give you any recommendations with values greater than 5%. The opposite can also be true. Perhaps the goal is to “maximize the business relationship” with a particular cedent; does that mean to lock the layer and write the same amount as last year or does that mean you need to write more? In both instances, the optimizer has no way of knowing what each of these subjective goals are, so you need to translate subjective goals into some sort of objective mathematical range so that the optimizer can incorporate it. The more realistic the objectives and constraints you provide the optimizer, the more realistic the resulting recommendations will be.

5) Portfolio optimization output provides “the perfect recommendation.”

Optimization does not provide one suggestion for exactly what you need to do. Instead, it provides a set of points that fall on what is known as the Pareto efficient frontier; each point represents a different hypothetical portfolio. You have a set of optimal solutions to choose from, providing you with a better risk and reward balance.

6) AI/machine learning is a replacement for underwriters.

We’ve saved perhaps the biggest myth for last. Similar to the point made above, the goal of AI/machine learning is to provide insights into where improvements can be made to a portfolio, but ultimately the underwriter needs to make decisions based on their risk appetite.

Portfolio Optimization Is Reality

Now that we’ve debunked each of these myths, the next time you hear the term “portfolio optimization,” you’ll have a better sense of what it is and what it isn’t, not to mention a better idea of how it’s transforming strategic planning across the industry.

 

What myths or misconceptions about portfolio optimization have you heard? Send a note to info@analyzere.com or chat with us by tweeting @analyzere

Read the portfolio optimization case study
AI and Machine Learning assisted portfolio optimization
Previously