We have examined what happens when we start from scratch. If we break a 30-year investment period into two 15-year periods, we should start out by dollar cost averaging into a 100% stock portfolio. After 15 years, varying stock allocations according to P/E10 provides enough protection on the downside to compensate for the reduction of the upside potential.

Varying allocations during the first 15 years reduces performance without providing protection. Varying allocations over the entire 30-year period reduces performance without protecting the downside.

I have examined my related posts on the accumulation stage. It is thorough. (I have not provided many examples.) I had not already resolved whether or not to change switching algorithms. When switching is used during accumulation, it is with money that has been invested in the first 15 years. A portion of the balance at year 15 is set aside for switching. That portion protects the downside. The remainder is included in the 100% stock portfolio with dollar cost averaging for new deposits.

During accumulation, there are no withdrawals for the allocation switching account. Previous investigations have assumed that there are withdrawals related to the distribution stage. Does this make a difference? The answer turns out to be NO.

**Initial Survey**

In my first survey, I varied the allocation percentage when P/E10 fell between 12 and 21. The stock allocation was 100% when P/E10 fell below 12. It was 0% when P/E10 exceeded 21. I kept the P/E10 thresholds at 12 and 21 throughout.

The initial balance was $100000. Expenses and withdrawals were set at zero. I determined the 15-year growth multipliers starting from 1923-1987. That is, the final balance equals the growth multiplier times the initial $100000 balance after 15 years.

All balances are in real dollars. That is, they are adjusted for inflation.

I examined the number of historical sequences that produced growth multipliers less than 2.0, the minimum growth multiplier and the maximum growth multiplier.

The trade off region appears to be for percentage allocations between 30% and 70%. The number of growth multipliers below 2.0 is much higher with allocations of 0% to 20%. The minimum growth multiplier generally decreases as the stock allocation increases. The maximum growth multiplier generally increases as the stock allocation increases. These are the effects of holding larger stock allocations: the spread of the data increases.

**Follow-up Surveys**

I examined three middle level stock allocations: 30% and 50% and 70%. In each case, I varied the higher P/E10 threshold between 12 and 26 in increments of one.

In most cases, the number of growth multipliers below two decreases as the P/E10 threshold increases. This is the effect of holding at least a few stocks more frequently as the upper threshold increases.

We exclude upper P/E10 thresholds of 12 and 13 because of the high number of growth multipliers less than 2.0. At 30% and 50% middle range stock allocations, P/E10 thresholds of 18 and 21 look a little bit more attractive than others when we look at the minimum growth multipliers, with a threshold of 21 being slightly better. Looking at the maximum growth multiplier as well, we select a P/E10 threshold of 21.

When we look at the 70% middle range stock allocation, P/E10 thresholds of 18 through 21 are best. The number of growth multipliers below 2.0 generally decreases until P/E10 reaches 23. Once again, I exclude P/E10 thresholds of 12 and 13 because of their high number of growth multipliers less than 2.0. I exclude thresholds of 22 and higher because of the minimum and maximum growth multipliers. In particular, the minimum growth multiplier is very low while the maximum has started to decrease.

**Summary**

I stopped my surveys at this point. The old allocations are still valid. The data behave in a manner similar to what we have seen before. Allocations are not excessively sensitive to P/E10 thresholds.

[In the older investigations, there were withdrawals. In these, there were none.]

It is very important to learn the right lessons from these investigations. We do not depend upon any particular number. We search for ways to reduce sensitivities. Along those lines, we have found that having many thresholds for making small shifts in allocations is best. It reduces the sensitivity to the exact threshold levels and stock allocations. It supports the common sense argument to take a little money off the table (in this case, out of stocks) when valuations are high.

Have fun.

John R.