Cycle Analysis versus Buy-and-Hold

The financial literature is replete with arguments demonstrating the impossibility of “timing” buy and sell decisions with any reliable advantage—that is, the impossibility of any given investor being able to predict the best and worst times to initiate long or short positions, and to do so accurately, period after period.

We agree that most attempts at timing are a fool’s errand, and that a strategy based on attempting to consistenly outperform a benchmark based on market-timing decisions alone is almost certainly bound to fail—that such a strategy would fail to improve portfolio risk/return metrics over extended periods. But time is a constant factor in portfolio-management decisions: To buy now, or to wait? To sell now, or later? When a manager instructs his trader to build a position in XYZ asset, he has made the decision to begin building that position now; alternatively, if he decides to wait, he has again made a timing decision—one that will yield positive or negative results versus the same decision made at any other time, regardless of whether the attempt to time the market was an express part of the manager’s thinking or not.

Any buy or sell decision incorporates an element of time: a decision node that may or may not be simultaneous with other decisions or expected market or economic events, incremental information, etc. Even staying fully invested all the time does not remove time from the investment decision-making process, unless the portfolio is 100% allocated to a single asset or asset class (a practical non-starter). Otherwise, the question becomes not whether or not to be fully invested, but in what. And the answer to that question, whatever it may be, must also incorporate the time element. We cannot even think of a theoretical way to cancel the time element out of the investment equation—and there is certainly no practical way to avoid some degree of market timing in even the most fundamental of investment strategies. All investors are timers by necessity, whether we like it or not.

That said, the degree to which we try to (or simply must) time our buy and sell decisions may vary along a spectrum, as the exhibit below delete next for words on the next page shows.

Examples of the pitfalls of such timing strategies are plentiful as well, and quantitative examples are as easy to generate as they are foreboding, inasmuch as they quantify the differences between a hypothetically lucky and unlucky investors at the extremes.

To take several such examples: If you had bought the S&P 500 index in August, 1998 (or built a portfolio mimicking the index), the position would have increased 15.1% by July, 2004; whereas if you had bought the index just one month earlier, the position would have declined in value by (1.7%)! A difference in timing of one month, over a holding period of almost six years, generated a difference in returns of 16.8%. Or to take a more recent example, and one shorter in scope: If you had bought the S&P 500 in September, 2003, that position would have increased 10.6% through July, 2004; whereas buying three months later would have resulted in losses of 0.9%. We could go on and on with such examples.

Do such examples demonstrate the pointlessness of market timing strategies? They seem to, by implying that if an investor were to initiate a position on such-and-such a date, he would be a winner (and significantly so); but just a little earlier or later?—a loser.

For example, if we take the longest economic cycle on record, which began in March, 1991, and trace it to the S&P 500’s monthly peak in August, 2000, we see that roughly one-third of the index’s gains were realized in just five of those 114 months*, half of the gains in just ten months**, and three-quarters of the index’s gains were realized in 15 months***.

* December, 1991; November, 1996; July, 1997; October, 1998; and March, 2000
** To the previous set of months, append: January, 1997; February, 1998; September, 1998; October, 1999; and August, 2000.
*** To the previous two sets of months, append: April and May, 1997; December, 1998; and June, 1999.

To put in perspective the degree to which the performance of the longest cycle on record was driven by a few key periods: One-third of the gains were realized in 4% of the months; half the gains in 8% of the months; and a full three-quarters of the index’s gains were real-ized in just 13% of the months that made up the longest cyclical bull run in U.S. history. Any effort at market timing that failed (for whatever reason) to position an investor long during these brief intervals of outperformance would have resulted in dramatic underper-formance relative to the benchmark, ceteris paribas.

The point is clear: even the substantial gains and losses that characterize lengthy, pronounced, and seemingly “easy to catch” market trends depend on a few critical periods, into which much of the trend’s returns are packed. Often—but for these short bursts—the lengthier trends of which they are a part might never earn recognition as secular trends (bull or bear markets) at all.

Arguments against the practice of market timing warn that it’s impossible, both theoretically and as a matter of practice, to identify these key periods in advance of their occurrence; both (i) because it is impossible to predict the outcome of random processes, and (ii) because if it were not, someone would have already identified the opportunity and arbitraged it away (the efficient-markets hypothesis).

But is this so? Is any attempt at timing the market really doomed from the outset? And if so, then let us certainly enjoin timing as a practice, moving on to fundamental principles of stock selection, indexing, arbitrage, buy-and-hold, or whatever.

« Previous: Typical Cyclical Profile | Next: Market Efficiency »   

Your suggestions for improving this site are greatly appreciated. (To report an error on this page, please copy the URL before visiting the suggestion box.)