The U.S. pension risk transfer (PRT) market continues to grow. With around $45 billion of transactions, 2023 was the second busiest year on record for PRT deals, and that pace was sustained during the first quarter of 2024.[1] However, as employers have been eager to transfer pension plan liabilities to the insurance sector, most of the simple deals have been done; insurers that want to continue in this market will need to get comfortable with more complex transactions.
Many of the deals so far have been for plans that only feature retirees. These have been relatively straightforward to price. The acquiring insurer knows exactly how many plan participants it will need to pay benefits to and has all the details of how these benefits will change over time. It only must make assumptions about mortality rates, asset returns and its own expenses.
By contrast, most of the plans now coming to market have large numbers of deferred lives – employees or former employees yet to begin taking their benefits. This adds significant uncertainty for insurers. They can’t be sure when deferred lives will start drawing their pensions; some may retire early or late. This can affect the size of the benefit. Nor do insurers know what type of annuity deferred lives will choose. They may not even have the basic demographic details of the dependents of the deferred lives who may be entitled to benefits from the plan.
With so many unknowns, pricing a PRT deal that includes significant numbers of deferred lives is challenging. Some insurers have simply opted to steer clear of such transactions. In the future, however, almost every PRT transaction coming to market is likely to pose this problem.
The good news is that a PRT deal that includes deferred lives is not a complete leap of faith; several sources of data can help insurers move forward with greater confidence.
For example, the sponsoring employer likely has good data on the decisions made by previous generations of deferred lives. WTW’s surveys reveal that insurers have made use of data from DOL Form 5500 Schedule SB filings (which pension plan sponsors must submit annually) that include documentation of this experience data. The census data of current retiree records for the plan, usually included in the deal, are another rich source, as the data include granular information such as commencement ages and form of payment. Finally, insurers may already have done similar deals; the experience of how these transactions played out is one more source of insight.
Using these data, insurers can begin to model the likely outturns for deferred lives of an individual plan in each of the areas where they are grappling with uncertainty:
In an ideal world, standardized assumptions would help insurers model for deferred lives in each of these areas. But because every plan offers different benefits and has participants with distinctive characteristics on gender, age, professional status and more, this simply isn’t possible. Both plan design and demographics have a huge impact. Early retirement subsidies, for example, will skew commencement ages; lump sum options are often popular. Men are more likely than women to select joint-life annuities. There may also be more than one population of participants to consider – where plans have evolved over decades of mergers and acquisitions. It is even possible that participants’ behavior will change as a direct result of the PRT transaction, as participants move away from a plan sponsored by an employer with which they feel comfortable.
For all these reasons, modeling individually for each deal therefore produces a more accurate assessment of future liabilities. It is also important to build a degree of sensitivity into these models – and to increase tolerance levels where larger numbers of deferred lives are younger, since this will increase uncertainty. New technologies will play a role, as some of the current limitations of software will be overcome, making it easier to model multiple forms of payment, to develop multiple decrement models and to forecast with more accuracy. Finally, the rapid evolution of predictive analytics tools also provides reason for optimism.
