Many of today's retirement strategies assume that the retirement income goal is fixed, or inflexible, rather than variable, or flexible, when modeling participant outcomes. This unrealistic assumption implies retirees have neither the desire nor the ability to change their spending over time.

In reality, most retirees can adjust their spending if they suspect their savings may not last throughout retirement. Alternatively, if markets do well, retirees can potentially spend more than originally planned. The flawed assumption that retirement spending is static has significant implications on a myriad of retirement decisions, particularly determining the optimal risk level forretirement income-generating portfolios.

Retirement Spending: Needs vs Wants

In our paper titled "Retirement Income Beliefs," we discussed the concept of breaking retiree spending into two categories: Needs and Wants.

Needs Spending: These expenses are generally less flexible and tend to be recurring in nature. Retirees generally have less discretion over the level or frequency of these types
of expenses.

Wants Spending: These expenses are generally more flexible and may be more irregular in nature. Retirees tend to
have more discretion in both the level and frequency of
these expenses.

Distinguishing between the required level of certainty for different types of spending has important implications when thinking about how to invest a retiree's assets. For example, a portfolio focused on Needs would likely have a greater focus on downside protection and inflation protection, while a portfolio focused on Wants would have more of a growth focus.

Measuring Outcomes

The probability of success is a widely used metric that measures the effectiveness of retirement income solutions and is very common in financial planning tools. We believe using a metric that focuses on goal completion can be more intuitive. For example, if a retiree had a target annual income goal of $100,000 for 30 years but were to fall $1,000 short in the final year of retirement, the outcome would be treated as a "failure" using traditional probability of success-related metrics, even though nearly 99.9% of the goal was achieved.

Additionally, we think it's important to incorporate retiree preferences, via a utility function, when estimating the respective goal completion metric. Utility is a way to measure how someone feels about achieving a given outcome, where the greater the utility, the greater the implied happiness.

While both participants in Exhibit 1 are achieving the same income level, the notably different levels of utility can have incredibly important implications for the optimal retirement income strategy. For example, Participant B would likely benefit more from allocating additional savings to guaranteed income than Participant A, in order to reduce the possibility of a shortfall and the significant penalty (i.e., disutility) associated with it.

Analysis of Optimal Portfolio Risk Levels

Using Monte Carlo analysis, we can determine the optimal portfolio risk levels (i.e., equity allocations) for retirees' retirement savings for different situations (i.e., savings levels, guaranteed income levels, retirement ages, and Needs percentages) and for different preferences.

Exhibit 2 demonstrates that while Participants A and B have the same level of retirement savings and the same target annual retirement income goal, there are significant differences in the optimal equity allocations, and glide paths, given their differing projected Needs spending. Participant A, who has more flexibility with retirement spending, can take on considerably more risk at younger ages than Participant B and should decrease the risk level of the portfolio considerably throughout retirement.

Overall, our analysis suggests that incorporating spending elasticity into the portfolio optimization process is likely to result in more aggressive portfolio risk levels for younger retirees and more conservative for older retirees, on average. Incorporating additional factors further customizes the optimal risk level.

Conclusion

Retirement is significantly more complex than assumed in most retirement income solutions available to DC participants today. A notable shortfall is the assumption that the retirement spending goal is inflexible, or more simply that a retiree is unable and unwilling to "change course" during retirement depending on how their situation evolves. By incorporating spending flexibility into retirement income models, our research shows a notable impact when determining the optimal portfolio risk level for individuals.