This paper describes how changing your spending for poor (or good) markets, helps you receive more in the long run.
A Dynamic and Adaptive Approach to Distribution Planning and Monitoring.
Paper and cover photo posted here with permission by the Financial Planning Association, Journal of Financial Planning, April 2009, by David M. Blanchett and Larry R Frank Sr.
The Journal of Financial Planning is published by the Financial Planning Association® (FPA®) and all information published within is the sole property of FPA.
Note: The research that supports this paper, and papers of other researchers that evaluate how changing conditions affect the withdrawal rate, is what started my thinking about the experiment design of such research. And these thoughts are what ultimately lead to the series of papers after this paper based on Dynamic Updating of the time remaining variable based on age as well as looking at the retirement income problem in three dimensions instead of one simple allocation.
The design in the past is based on a single Monte Carlo simulation (stochastic) which is what is common for retirement income evaluation. However, because only a singe variable is measure as output (the withdrawal rate), whenever any input variables are changed within the simulation, the only manner these changes may be evident is through the single output variable. This higher initial withdrawal rate, that results from changing input variables (risk, return, cash flow, time, etc) within the simulation, generates the illusion, or misinterpretation, that cash flow (income) may be greater at the beginning (because the withdrawal rate is higher as compared to no adjustments) when in truth, the improved cash flow may not be actually available until later.
This is NOT a criticism of such simulations, but a criticism of the single simulation model point of view (where the model is the single simulation). The model is much more dynamic than this in order to reflect real life aging dynamics. Here's a more developed model: Monte Carlo simulations should be performed sequentially for each change in time period remaining (capture the aging process and changing life expectancy that results). These rolling sequential simulations should be serially connected using just the first year results of each simulation period. For example, the cash flow and portfolio balance for the 20th simulated year within a single 33 year long simulation for a 60 year old couple, is NOT the same expected cash flow and portfolio balance for the simulation for the now 80 year old couple 20 years later with now a 15 year expected longevity (until one of them passes). Or stated differently, the 32nd year simulation for the 60 year olds' last year remaining will calculate a final years' cash flow and terminal portfolio value that is quite different from what 92 year olds would need for an expected longevity of yet another 8 years! All the prior cash flows and portfolio balances are also off because the proper simulation period length does not match the real rolling expected longevity periods of real people.