Use of Secular Trend Concept for Asset Allocation in 401K
In a previous article, I described how one can use secular trend theory to guide asset allocation. During a secular bull market trend long term investors should be fully invested in stocks. Once a secular bear market trend has begun, asset allocation between stock and non-stock investments should be guided by valuation concerns. I showed a graph of valuation in terms of relative P/R (rPR) for previous secular bear markets that showed the general trend in rPR is downward over the course of a 15-20 year secular bear market. I presented tables showing the probability of various sorts of returns for various ranges of rPR. Somewhat surprisingly, the data showed that the potential for successful long-side investments of a fairly short term nature (~ 1 year) was better for high levels of valuation than it was for lower levels. Long-term returns were all poor. Only when extremely low levels of rPR are reached (typically close to the end of the secular bear market) do long-term stock returns start to look very attractive.
The relatively good prospects for long-side investments at the very highest levels of rPR during secular bear markets were explained by the fact that high levels of rPR occur early in the secular bear market, when it is not yet clear to most investors that a secular bear market has begun. I wished to take advantage of this situation. This article describes specifically what allocations I have made in my own 401K since September 3, 1999 when I moved completely out of equity in anticipation of the beginning of the secular bear market, which I believed had already begun the previous July. I was wrong, the secular bear market did not get underway until Sept 2000, five months after I had finished writing Stock Cycles, but two months before its publication date.
In our company's 401K program employees have a choice of several stock funds, an income fund, and a blended stock/bond fund. There is no pure bond fund. One of the stock funds is an S&P500 index fund. The funds I use for my allocation are American Century Ultra (TWCUX) and the income fund. Ultra shows a higher degree of volatility than the index fund and so is more suitable for the asset allocation scheme I had in mind. The income fund is like a money market fund in that it shows an interest-like return, but it is invested primarily in asset-backed securities and tends to give a return that follows longer term interest rates. Over the past two years it has yielded about 6% on average.
Figure 1 shows a graph of American Century Ultra with arrows marking my 9/03/99 sale and subsequent repurchases of TWCUX. The region defined by the March 27, 2000 top and the April 4 2001 bottom in TWCUX corresponds to Table 1 in my secular trend application article. This table shows that historically, randomly selected investments made at the levels of rPR corresponding to the levels in this region showed positive one year trends in return about half the time. Thus, the idea is to buy in the lower half of this region and sell in the upper half of this region as I have marked in the chart. The problem with this neat idea is this "region" defined by the March 2000 and April 2001 extremes was not defined until after April 4 2001 (and may not yet be defined if the market were to fall below the April 4 low in the next few months). Yet here I was buying before April 4. How were these buying decisions made?
Figure 1. American Century Ultra (TWCUX) and buying (red arrow)
and selling (black arrow) points.
After I moved to 100% invested in the income fund, by 401K was earning a 6% return, while TWCUX (the position I had exited) advanced 35% to its peak on March 27, 2000. As a long-term investor, I should not concern myself with short term movements like this, but nevertheless it was difficult to sit on the sidelines and watch the market do things I never thought it could do (like the NASDAQ soar to 5000+). During the NASDAQ minicrash from March to May 2000, I saw evidence that perhaps I would again be able to buy TWCUX at prices below where I had exited. That May, I decided I would begin to buy on a wide scale as soon at TWCUX dropped decisively below my exit level (adjusted for the return I had gotten since then in the income fund).
Having decided where to start re-entry into the market I needed to ascertain where the bottom of my scale buyng would end up. That is, how low would the market have to go for me to be 100% invested in stocks? In chapter 4 of Stock Cycles I presented a model I developed that described the behavior of the market over shorter periods of time. This model suggested that as long as earnings growth continued unabated (that is, we would have no recession) the stock market would continue climbing. The very optimistic forecasts of Dow 36,000 and higher made by various bullish analysts would actually come true if the ten-year old expansion keep right on going for another 7-10 years. If it came to an end there would instead by a rather severe bear market, which would very likely mark the beginning of a secular bear market predicted by P/R.
In Stock Cycles I used the model to analyze a hypothetical recession that started with an earnings peak at the end of 1999. The model I developed is not predictive, one of the inputs into the model is the index value, so to use the model one must know what the market will do. Of course one doesn't know the future path the market will take. What one can do is put in various hypothetical market scenarios into the model and see if they are consistent with past behavior. So I inputted a hypothetical recession by using the last recession as a template. By doing this I found that at a recession market bottom the S&P500 would have to be in a range of 650 to 1360 for it to be consistent with past behavior. So the worst case for a recession beginning at an earnings peak at the end of 1999 would be about 650 on the S&P500-assuming a recession of 1990 severity. Since earnings did not peak at the end of 1999, they would necessarily peak higher and so I picked 700 as a bottom. Looking back to when the S&P500 had been around 700 I found that TWCUX was in the 16-17 range at that time. So I determined that I would buy on a wide scale starting a bit under 35 and going all the way down to 16.5, if necessary.
On October 12, 2000 the market dipped strongly and I moved 10% of my 401K from income to TWCUX at a net asset value of 34.61. I now could construct my buying scale, which is shown in Table 1 as the horizontal blue lines. Over the next six months I moved 40% of my 401K, in four 10% aliquots, on big down days roughly spaced apart as suggested by the buying scale. My last action (so far) was a 10% move on March 21 at a value of 25.71. On April 4, TWCUX bottomed at 24.95 (not far enough for another purchase) and has risen since. At this point we are in limbo. If the market starts heading down again I will resume buying as suggested by the scale, all the way down to 16.5 if necessary. If the market continues to rise I will need to sell at some point. I now need to determine a selling scale.
When I moved the first 10% into TWCUX at 34.61, there was a possibility that the market would experience a flat correction then move back up in early 2001 as it did the previous year. I did not think it likely, but could not rule it out and so I was willing to "bet" 10% on this scenario. At that time I arbitrarily had decided I would sell $5 higher, or at about $40. Had this flat correction occurred I would be back 100% in the income fund at a value of 40 so I decided I would keep this level as the top of my selling scale and so be 100% back in income should TWCUX reach 40. For the bottom of my selling scale I use the findings of Table 1 in my previous article. Assuming that the April 4 bottom holds and we go up from here, this means that we spent the entire period since the July 1999 peak in rPR inside the range covered by Table 1. Since Table 1 suggests that half of all randomly chosen points are bad times to enter long positions for ~1 year holding periods I figured that this half would be good times to sell. Logic suggests that these points would be more prevalent in the upper half of the trading range defined by the March 2000 peak and April 2001 bottom. Thus, the mid-point of this range (about 35) should be a good point to commence selling. Thus, I defined by selling scale as between 35 and 40 on TWCUX. These levels are identified in Figure 1 by the horizontal black lines.
Having defined both the buying and selling scales the operation now becomes quite mechanical. If the S&P500 falls below 1100 in the coming months I will continue to increase my stock allocation in accordance with the blue lines. If TWCUX rises above 35, I will sell in accordance with the black lines. Once the situation is resolved I plan a follow-up article describing what I did and what comes next.