Previous Studies, "Due Diligence" , and the Methodology
I started with the same Ibbotson data as my understanding of the Jarrett study (the Trinity study appears to have used a subset of this information). Specifically this is the annual (1926 through 1998) performance of the following assets classes.
I first tried to duplicate the Jarrett Study results using a starting asset base of $1M and varying asset mixes and the withdrawal rates (see specifically set 4 in the Jarrett Study link). I assumed a complete portfolio rebalancing to the target allocation each year (after withdrawal for the years living expenses). Click here for additional details regarding the methodology. My analysis used the following four asset allocations:
The results showed that the asset allocation that optimized the worst case starting year was asset mix #3 (20% Large Company and 20% Small Company, etc.). The analysis supported an inflation adjusted withdrawal rate of 4.4% of the initial portfolio value. This means that in all cases the portfolio had a positive value (albeit small in the worst case) after 30 years. None of the other asset allocations would support a withdrawal rate greater than 4.4% in all years.
I did this strictly as a check on my own work and these results were consistent with the Jarrett study and slightly better than the Trinity Study (which is to be expected since the Jarrett study uses a wider range of asset classes). But this was not the question that I was pursuing so once I got this verification, I did not evaluate this scenario further. Click Here to see a sample of the data.
When you look at the data note that there is a real cost to using the relatively non-aggressive asset allocation of 40% equities the opportunity to generate a substantially larger end value portfolio. But these higher rewards have higher risks.