A recent research paper, Estimating the True Cost of Retirement published by Morningstar and written by David Blanchett, has reinvigorated research and discussion of an age old problem for financial planners: How do you estimate retirement expenses?
In this research Blanchett uses the RAND Health and Retirement Study (HRS) to look at longitudinal data on actual retiree spending. A longitudinal study observes the same variables over a long period of time, in this case the spending of specific households in retirement. This study finds two noteworthy items planners should consider when estimating retirement spending.
The first is a verification of the idea proposed in the late 90’s by Michael Stein that spending levels increase during early and late retirement in what Stein called the Go-Go, Slow-Go, No-Go Retirement, and what Blanchett calls the Retirement Spending Smile. The second is that actual retirement spending seems to decrease over time even though the costs for the goods purchased by retirees increases faster than general inflation. This finding of decreasing spending has some important exceptions when looking at low spending, high net worth clients, and is exaggerated for high spending, low net worth clients.
Michael Kitces's blog includes a helpful article summarizing Blanchett’s research, “Estimating Changes In Retirement Expenditures And The Retirement Spending Smile.” One commenter on Kitces's blog points out a research paper published by JP Morgan:The lifecycle of spending. The JP Morgan research found that average U.S. household spending peaks at age 45, and after that the only category that doesn’t decline is health care. The research even splits the periods to show decline in spending from 45-75 and from 65-90.
Blanchett notes the traditional rule-of-thumb of replacing 70-80% of the client's pre-retirement income is a reasonable starting place for retirement spending. His research indicates retirement projections accounting for the "retirement spending smile" require an asset balance at retirement that can be 20% less than projections that do not reflect the spending curve. This research is fascinating, as are the implications.