While most financial planners are familiar with the leading practitioner-based research in retirement planning – for instance, Bill Bengen’s 4% safe withdrawal rate studies – the reality is that economists have actually developed their own research approach to evaluating financial planning trade-offs, including optimal strategies for retirement spending and asset allocation. Notably, though, the two tracks of retirement planning research employ substantively different approaches – where practitioners most commonly “test” a particular retirement plan or strategy to see if it’s sustainable (or not), while economics researchers try to create models that can be optimized. However, the gap between these distinct lines of research and practice are beginning to blur, as practitioner-oriented research becomes more sophisticated, economic research becomes more practical, and computing technology makes it easier than ever to conduct complex analyses.
In this guest post, Derek Tharp – our Research Associate at Kitces.com, and a Ph.D. candidate in the financial planning program at Kansas State University – takes a deeper look at dynamic programming, a economics-based methodology for conducting financial planning retirement projections, that introduces new opportunities beyond traditional financial planning software in optimizing retirement spending and asset allocation, and may be outright superior in projecting and modeling how retirement spending and asset allocation might change over time, based on whatever future market returns turn out to be.
The core of the dynamic programming (also known as dynamic optimization) approach is really just a methodology of solving larger problems by breaking them down into smaller problems. In the context of financial planning and retirement projections, it’s about taking a long term retirement plan, and breaking it down into a series of sequential retirement years, each of which can then be optimized based on what happened (or didn’t happen) in the preceding years. What’s powerful about dynamic programming, though, is its capabilities to model a greater number of retirements that might all vary at the same time; for instance, dynamic programming could allow spending (consumption), asset allocation, investment returns, and longevity to all vary at the same time, and then truly optimize the financial plan across all of those variables at once – rather than the more manual process of testing “a plan” and then repeatedly tweaking it with “What If” scenarios until a client is satisfied with it, as is more common amongst practitioners today.
A key advantage of dynamic optimization is that a single dynamic programming analysis also has the potential to provide guidance about what to do now and in the future in a way that a one-time Monte Carlo projection cannot. Dynamic programming can do this by effectively providing a three-dimensional road map of how over time a retiree might adjust spending, and adapt the allocation of their portfolio, based on the actual investment returns that are experienced in the future. And, by being more responsive to an individual’s consumption preferences and investment returns that are experienced, dynamic programming not only helps guide retirement spending decisions that are better suited to an individual’s unique goals and desires, but it may increase a retiree’s consumption in retirement, as prior research from Gordon Irlam & Joseph Tomlinson has found that dynamic programming can provide significant enhancements in retirement income over traditional rules of thumb utilized by advisors. Financial planners who want to explore dynamic programming further can explore products such as Gordon Irlam’s AACalc or Laurence Kotlikoff’s ESPlanner.
Ultimately, it’s still not clear that there’s one “right” way to do retirement planning. Whether it is Monte Carlo versus historical… goals based versus cash flow based… or dynamic programming versus non-optimizing approaches… all can provide different insights, which in turn can help guide decision for clients given the risks and sheer uncertainty they face in planning for retirement. But in the end, if the whole point of doing financial planning is at least in part to come up with an actual plan for how to handle an uncertain future, dynamic programming provides a unique toolset that isn’t available in today’s traditional financial planning software solutions… at least, not yet!