| 6/18/01 The following is an attempt to explain how I try to time the market, and why I think it is worth the effort. CYCLE AND ULTRA STATUS TUTORIAL 1. INVESTMENT PHILOSOPHY Given the spectacular returns of a buy and hold approach to the market over the last two decades and the adequate long term returns throughout history, why try to time the market? For an investor that can psychologically tolerate a temporary loss on the order of 50% of his capital and who has the time and fortitude to wait years to recover that loss, there is no need to tackle that admittedly difficult task. The rest of us will time the market consciously or unconsciously. Either fear and greed will prompt us to buy and sell at the wrong time, or perceptive analysis of the available data will produce correct decisions often enough to do well without the large drawdowns that, at best, are emotionally difficult and, at worst, can jeopardize a comfortable retirement. Over the long term stock prices are correlated with their earnings. Securities analysts recognize this and try to determine the value of a stock by estimating its future earnings and discounting those earnings based on prevailing interest rates to determine the present value. The trouble is that they usually cannot estimate even year ahead earnings very accurately, much less earnings a decade or two in the future. Contrast the situation a couple of years ago, when technology stock analysts were projecting triple digit growth ad infinitum in order to justify astronomical stock valuations, with the current situation as company after company reports disappointing earnings or warns of future earnings disappointments. It would appear that predicting stock behavior on the basis earnings forecasts is a very difficult task even for the professionals, but that is the approach of choice for most investors. Those of us who viewed the historically high P/E ratios of the mid to late 1990�s as a danger signal missed too much of the greatest bull market of our lifetime. Those who believed the analysts projections of future earnings, have given back much of their paper gain. Neither skepticism nor gullibility regarding earnings projections has served investors well. I have concluded that the fundamentals are undeniably important over the long term, but predicting future fundamentals is an exercise in futility. For the shorter term, the market speculates on the unpredictable future fundamentals; it cares little about the known past fundamentals Over shorter terms, markets are driven by investor sentiment. For every buyer there is a seller, and the only thing that determines the direction of the price movement is their relative eagerness to execute a trade. An eager buyer will buy at the higher ask price and an eager seller will sell at the lower bid price. A few eager buyers or sellers will affect the portfolio value of all holders of the security being traded - in effect creating or destroying paper wealth in the process. There are numerous measures of investor sentiment that give useful clues to the likely direction of securities prices over the short or intermediate term. The trouble is that they are usually not very precise; high levels of bullish or bearish sentiment can persist for long periods while the market continues in the direction of the prevailing sentiment. Although sentiment extremes usually coincide with market turning points, it is often difficult to recognize that extreme until after the fact. Sentiment readings are very useful for assessing the level of risk in the market, but by themselves, inadequate for a good market timing system. Most popular market timing systems attempt to follow the prevailing trend in the market - a buy high, sell higher approach. They work well if the market alternates between substantial uptrends and downtrends, but they are disastrous during markets that fluctuate in a modest trading range. The second most popular timing systems attempt to identify overbought or oversold conditions, and bet on a reversal. They work well in a trading range market, but are usually on the wrong side of a trending market. The problem becomes one of predicting what type of market can be expected in the near term future (a task considered impossible by much of the investment community) and applying the type of timing system that works for the predicted type of market. The difficulty of correctly timing the market is attested to by the ULTRA Timing System results that I report weekly. ULTRA is a service that tracks a large number of timing systems that have attracted some attention over the years. Generally, these systems have under performed during bull markets, but have reduced volatility, and some have done well enough during bear markets to outperform significantly over a long period of time. During the two-tiered markets of the last 3 years, however, few of these systems have performed well. Some use fundamental data (earnings, dividends, interest rates). Most use some price measure and many use some measure of market breadth (advance, declines, highs, lows). The unusual degree and length of the divergence between breadth and large capitalization price, and the advent of valuations never seen before, confused many of these systems. I anticipate that the bursting of the tech bubble will return the market to more conventional behavior, and these systems will again outperform, so I continue to report on their status. Nevertheless, I believe I have found a better way � Cycle Analysis. I operate on the premise that the securities markets exhibit identifiable cycles, and that those cycles are sufficiently predictable to be useful for market timing. That the cycles exist is usually obvious to anyone who takes the trouble to look. Any number of techniques can be used to identify the predominant cycles, and many cycles can be observed on any historical bar chart. Whether those cycles are sufficiently regular to be useful for market timing is more debatable. I have come to the conclusion that, with use of good measurement tools, cycle analysis provides the best opportunity to achieve good investment returns without incurring the large drawdowns that cause many investors to leave the game. 2. CYCLE SELECTION During the last 60 years US Stocks (S&P500 index) increased in value an average of about 8.5% per year. Overlaid on this long term trend are cycles of approximately 4 years, 2 years, 9 months, 19 weeks, 9 weeks, 13 days, etc. Longer and shorter cycles exist, but the longer cycles aren�t particularly relevant to those of us who are senior citizens, and the shorter cycles are of interest mainly to those willing to sit in front of a computer monitor during the entire market day. Consequently, I focus on the aforementioned cycles. Cycle analysis is useful for predicting the type of market to be expected in the future. For example, if long and intermediate cycles are in their uptrending phase, the short cycle pullbacks should be of little consequence, and should be ignored. If either the long or intermediate cycle is near a top while the other is near a bottom, or if one is in an up-phase while the other is in a down-phase, the market is likely to fluctuate within a relatively narrow band, and a countertrend system can be used for short term trading. Due to the long-term upward bias in the stock market, the criteria for buying should usually be less stringent than that for selling short. That is not the case for bonds or commodities. The appropriate choice of cycles to use in trying to time the market is dependent upon a number of variables that are different for each investor. One is the tax situation. In a taxable account it makes sense to play the 4-year cycle because that provides a holding period likely to qualify for favorable long-term capital gains treatment. In a tax-deferred account, taxes are not a consideration, and more frequent trading can enhance returns. Account size is a consideration. Although commissions are now small, and bid-ask spreads have recently narrowed, these trading costs can become significant for short term traders � especially for trading in small lots. The most important consideration for selecting a trading cycle, however, is probably the time that the investor has available and is willing to devote to investment management. The shorter the cycle traded, the more time invested. 3. CYCLE CHARACTERISTICS AND MEASUREMENT TOOLS A cycle tends to vary in length with a standard deviation of about one-sixth of the nominal cycle length. Consequently, about 95% of four year cycles will fall between 32 and 64 months; 95% of 9 month cycles will fall between 6 and 12 months; 95% of 9 week cycles will fall between 6 and 12 weeks; etc. Such wide variations dictate that nominal cycle durations alone are insufficient to profitably time the market. Appropriate indicators are necessary to measure what the market is actually doing � indicators that can identify the actual turning points in the various cycles as they are happening. The best indicators I have found for this purpose are the slope of the linear regression and variants, developed by Walter Bressert, of the Stochastic and the Wilder Relative Strength Indexes (RSI). Neither Stochastics nor RSI�s are what their names imply. Both are oscillators that measure the price momentum over a specified time period. The stochastic measures the relationship between the closing price and the high-low range over a specified period. The Wilder RSI is based on the ratio of the up closes to the down closes over a specified period. Some sort of smoothing is generally applied to reduce the noise in the data. The choice of smoothing period is always a compromise between one that is long enough to filter out the noise (or shorter cycles), yet short enough to avoid introducing excessive lag into the indicator � one that is both timely enough to produce profitable trades and reliable enough to produce an acceptable percentage of profitable trades. The time period to use in calculating these oscillators is dependent upon the typical length of the cycle to be measured, and the smoothing period is dependent upon the timeliness/reliability tradeoff that is likely to produce the best profit based on back testing of historical data. The appropriate selection of the regression period for the linear regression approach is perhaps easiest to understand. Consider a linear regression over half of a complete cycle of a sine wave. It should be intuitively obvious that the slope of the linear regression will be at a maximum positive value for a period from the trough to the peak of the cycle (i.e., half of a complete cycle), and it will be at a minimum (maximum negative value) for a period from the peak to the trough of the cycle. Consequently the peaks and valleys of the slope of a moving linear regression line with a length that is half of a complete cycle will coincide with the peaks and valleys of the sine wave. Of course, for an ideal sine wave there is no advantage to using the regression slope to ascertain the peak or trough of the cycle. The peaks and valleys can be observed directly from the perfectly smooth sine wave itself. However, if noise were superimposed on the sine wave, a direct observation would entail some uncertainty in the identification of these turning points. The smoothing inherent in the linear regression slope will reduce this uncertainty but not eliminate it. In practice, when applied to inherently noisy stock price data the linear regression slope does a reasonably good job of identifying the peaks and troughs of any cycle with a period that approximates twice the regression period. Moreover, it is fairly insensitive to the normal expansions or contractions seen in the period of the stock cycle. For example, experience has shown that a cycle of approximately 9 weeks or 44 trading days exists in the stock indexes, but the standard deviation in the period is about 7 trading days. If I select a regression period of half of the nominal cycle (22 days) to track this cycle, and the cycle turns out to be shorter, the linear regression slope will lag the price by a few days. If it turns out to be longer, the linear regression slope will lead the price by a few days. But the pattern of the linear regression slope will be quite similar for any regression period that approximates half the cycle length, say 18 to 26 days. I prefer a little lead in the indicator on average and find that 19 days works well, but I recognize that the lead comes at the price of more frequent false indications of a turn. In order to reduce the probability of acting on a false indication, I like to observe a consensus of several indicators that are designed to measure the same cycle. I do not use the same set for all of the cycles that I try to measure because some work better than others for a given cycle. The Bressert Double Stochastic works best for cycles nominally encompassing less than about price 25 bars. Consequently I apply it to daily charts to measure the short cycle, weekly charts to measure the 9-week and 19-week cycles, and monthly charts to measure the 9-month and 2-year cycles. Conversely, the RSI works best with more than about 15 bars. I apply it to the daily charts for the 9-week cycle, weekly charts for the 19-week and 9-month cycles, and monthly charts for the 2-year cycle. The Linear Regression Slope (LRS) is the most flexible. I use it on daily charts for the short and 9-week cycles, the weekly chart for 9-week, 19-week and 9-month cycles, and the monthly chart for 9-month and 2-year cycles. If it appears that short through 9-month cycles are bottoming or topping simultaneously, buying or selling is a no-brainer. However that occurs rarely. Conflicting directions for the various cycles is the normal condition. In mid-October 2000, for example, it appeared that the short-cycle, 9-week cycle and 19-week cycle were all at or near bottoms, but the 9-month cycle was still trending down and likely to continue in that direction for a while longer. Based on the tenant that it is generally foolhardy to play a cycle in a direction opposite to the trend of the next longer cycle, that constituted a buying opportunity only for an investor inclined to play the 9-week or shorter cycles. As it turned out, the markets rallied into early November and then resumed their downtrend. 4. THE ROLE OF SENTIMENT INDICATORS: In order to reinforce or cast doubt upon conclusions drawn from the analysis of cycles, I draw upon sentiment data. Based upon both theory and observation, when most investors are at extremes of optimism or pessimism it is time to sell or buy. In theory when most investors are optimistic, they are fully invested so the reservoir of potential eager buyers to drive prices up is small. (They may be eager to buy more, but they have no more cash or margin buying power. That resides with the investors from whom they bought their securities at lower prices, and who are now unlikely to be eager buyers at higher prices.) There are, however, plenty of bulls to be converted to bears to produce the selling pressure necessary to drive prices down. Conversely, when most investors are pessimistic, they have sold out and selling pressure dries up. There are plenty of bears to be converted to bulls and they have cash available from selling on the way down. There are a number of indicators of sentiment that historically support the theory. One of the oldest is the Investor�s Intelligence survey of advisory services. Peaks in bullishness by this group tend to correlate with market tops and peaks in bearishness with market bottoms. The same is true of the American Association of Individual Investors survey of their membership. I use these measures primarily as an indication of risk. The lag in the survey data of about a week and a half prevents their use as a short-term timing tool. They can be very useful, however, in confirming likely intermediate cycle peaks and valleys. The option buyers (a group that is generally wrong) provide better short-term indications. The Volatility Index (VIX) measures the implied volatility of the eight OEX (S&P100) puts and calls closest to the money and with an average time to expiration of 30 days. It tends to spike up sharply to about 4 standard deviations at significant bottoms. While its peaks usually pinpoint bottoms in the market, it tends to provide only danger signals in the general region of a market top with a reading of about �1 standard deviation. I have recently also begun watching the OTC counterpart to the VIX, the VXN, which is based on the NDX (Nasdaq 100) rather than the OEX. It appears to have similar characteristics � moving opposite to the market. Based on data to date, it promises to be an even better indicator than the VIX The other option indicator works well in both directions. It measures the ratio of call trading to put trading on the Chicago Board Options Exchange. When activity in calls relative to puts peaks, the stocks tend to be near tops. Conversely, when call activity relative to puts is abnormally low because option traders are pessimistic, stocks tend to be near a bottom. Both the VIX and the put-call data are reported on a daily basis, and reflect what traders are actually doing rather than interpret their opinions. Hence they tend to be somewhat more reliable and timely than the opinion surveys. Other measures of sentiment exist, but I use the four mentioned here primarily because the data is easily accessible. 5. SUMMARY Buy and Hold investing is fine for risk tolerant investors. For those of us who are not very risk tolerant, market timing offers an alternative that can dampen the swings and provides an opportunity to outperform the market if executed well. Correctly timing the Market is difficult, but perhaps less difficult and time consuming than superior stock selection. Measurement of Market Cycles, supplemented by measurements of investor sentiment, provides the most effective approach that I have found to time the market. The most effective indicators that I have found to identify the cycle turning points are the Linear Regression Slope (LRS), the Bressert Double Stochastic (DStoc), and a Modified Relative Strength Index (RSI). The appropriate choice is dependent on the scale of the chart (daily, weekly, monthly) and the cycle to be measured. The cycles to play are a highly individualistic choice, dependent on the investor�s available time and assets, interest, aggressiveness, and tax situation. Whatever the choice of cycles for the planned holding period, invest only in the direction of the next longer cycle. |