I assumed a complete rebalance to the target allocation
at the beginning of each year (after withdrawal of that year's living expenses). I ignored
any associated tax consequences associated with this action.
I analyzed retirement starting years from 1926 through
1979 inclusive. This is a total of 54 different 30-year retirement scenarios.
For starting years after 1969 that would have included
years later than 1998 (the last year for which I had data) I simply rolled
over back to 1960, 1961, etc. The choice of 1960 was a strictly arbitrary choice.
I did not do a complete MVO (Mean Variance Optimization),
but the playing around that I did indicated to me that Long Term Bonds (government or
corporate) offered very little value so I ignored these two asset classes in my asset
I used a 5% short term (treasuries) allocation in all
cases because nothing else seems practical to me (even index funds tend to have a couple
percent allocation to cash in most cases).
I ignored any investment expenses (fund expenses,
brokers fees, etc.) that might be encountered in practice.
Each withdrawal was made at the beginning of the year and
placed into an account (in practical terms probably an interest bearing checking account)
whose net return was .33 x initial value x Treasury Bill return. I was trying to simulate
the return of that checking account across the year (average balance of approximately 50%
of the withdrawal at the beginning of the year and earning somewhat less than Treasury
The most straight-forward use of this data would be to
treat taxes just like any of your other expenses.