Mark Freeland wrote a very good analysis of the returns of different asset allocations during the inflationary 1970s He estimates the expected results of withdrawals from various allocations of equities 100% equities to 40%/60%.
The best results were a 79/9/16% ( latter is cash) which allows withdrawals for four years without selling equities at the bottom, or when they are below their high
This "cash stash" keeps you from the largest draw on performance, selling into a declining market at the low. This approach is similar to James Cloonan at American Association of Individual Investors, who proposed five years expenses in cash and withdrawing from equites only when the SP500 was within 5% of it's high.
I looked carefully to see it four or five years is enough and I am not sure it it, because the market took 3.2 years to recover from 11/68 high but then less than a year later, ( 1/11/73) it peaked again and took 7 years to recover (7/16/80). The 68 peak to the 80 recovery was over 11 years.
Similarly from a high in 3/2000, the recovery took 7.1 years but less than 5 months later crashed again and didn't recover until 3/2013, a combined total of 13 years.
Consequently, I believe that the customary 4 or 5 years is not enough to cover these prolonged declines. Seven years of expenses in cash or short term bonds seems to be the minimum.
Maybe the withdrawals would have been during those short period of market highs, and the withdrawals if below the peak would have been at least pretty close to it. You also have to withdraw enough to replenish your "cash stash".
Comments
In essence, I started with a traditional 60/40 portfolio and swapped some or all of the bonds for more equity and cash, figuring (hoping?) that the cash would add stability allowing for a higher equity allocation.
Intuition says that if one has too many market corrections/bear markets in a span of a few years, preserving one would be better off with cash. I tend to agree with this intuition. However, it turns out that this risk sometimes vanishes if there is a full recovery between the periods of market decline.
Looking at the 1968-1980 period (assuming withdrawals and reallocations are done on calendar year boundaries), at the end of 1972 (right before the 1973-74 bear market) the allocation methods (including withdrawals) all are in the black. The portfolio values (nominal dollars) range from $101,402 (all stock) to $102,728 (90/10 with a cash buffer). A 28% cash buffer (7 years @4%/year) ended in the middle of the pack with $101,964.
Since all methods ended 1972 above their high water marks, they would rebalance to their "normal" allocations. It would be as if 1969-1972 hadn't happened, except that some portfolios would have a few hundred dollars more or less than others.
The reason why starting with 1969 instead of 1973 seems to make a difference is that this moves the ten year mark. If one ends after 1978, one ends just before the market has a big run up: 18.52%, 31,74%, -4.70%, and 20.42% from 1979 through 1982. It's the arbitrariness of the time periods and not the method that's the problem here.
Finally, a mea culpa. In looking more closely at the two sets of results, I realized that I erred in building the results for 40/60 in 1973-1982. That method (stocks/bonds, annual rebalancing) came out pretty well. The 30% bond return in 1982 (vs. 20% for stocks) and 8% return in 1981 (vs. -5% for stocks) surely helped.