The amazing runup in technology stocks from mid-1998 to early 2000 challenged all investors. No manager who avoided the bull charge because of the seemingly obvious overvaluation of tech stocks by traditional measures will soon forget the frustrations of that time. The subsequent rapid collapse of the tech stocks seems to have vindicated the idea that value eventually must prevail. Sadly, our research indicates this widely accepted belief is wrong.
Motivated by the unprecedented failure of traditional value factors in 1999 and our previous studies of value and momentum factor interaction, we have hit upon a novel way of modeling economic sectors, confirming that different sectors demand different approaches in predicting future individual stock returns.
Our new optimal sector models use varying blends of valuation and momentum. This approach tests well over the past several decades, has theoretical support and goes a long way toward explaining the recent technology stock phenomenon.
Here is a summary of our findings:
* Stock returns are adequately modeled by combining pure momentum and pure valuation.
Virtually all of the many known excess-return factors can be classified as either valuation or momentum based on their statistical correlations and/or fundamental characteristics. This implies that the prices of individual stocks are based on a combination of the worth of current activity (valuation) and the worth of future activity (momentum). We have constructed six-month horizon forecasting models incorporating pure valuation and pure momentum that can be thought of as proxies for the extent to which investors are incorporating current and future activity in valuing stocks.
* Stocks exhibit a fluctuation in the effectiveness of pure valuation and pure momentum over time.
The predictive effectiveness of valuation and momentum are strongly negatively correlated and have been for 30 years. The valuation/momentum cycle seems to be two to three years and is driven partly by observed long and short interest rate differences and partly by current expectations of economic growth. It is not market direction dependent, nor can it be forecast reliably. It seems to be a reflection of the extent to which investors are optimistic or pessimistic about growth prospects.
Our new results establish that this phenomenon is seen contemporaneously in every economic sector. The years 1998 and 1999 favored momentum. Both valuation and momentum worked in 2000, reflecting the shift from extreme optimism to pessimism about future growth.
* Optimal sector valuation/momentum blends vary by sector and over time, and can be predicted.
When measured over multiple cycles, economic sectors vary greatly in the correct or optimal blend of valuation and momentum providing excess return. Slow growth sectors (e.g., utilities) prefer more value, and fast growth sectors (e.g., technology) prefer more momentum. Changing definitions of sectors over time make this measurement process extremely difficult. We carefully reconstructed economic sectors as they were defined at the time over the past 10 years. Using more than 20 years of historical data we discovered a practical way to forecast the optimal blend for the next year. This resulted in a sort of replay of an investment process over the past that produces changing models tuned to the past performance of each economic sector. Tracking the shifts over time of optimal economic sector valuation/ momentum blends reveals the shift in the growth prospects of sectors as they respond to investors' perceptions of their current and future worth.
Momentum discriminated reasonably well among technology stocks and all other sectors in each of the past three years. Valuation failed in 1998, failed horribly in 1999, and worked as well as momentum in 2000.
Our new results reveal the previously unappreciated fact that the optimal blend of valuation/momentum for the technology sector has strongly favored momentum over the past 20 years, not just the past few. This suggests that overall value has never been important in discriminating among technology stocks. Sometimes value does work in tech stocks even better than momentum, but when value fails in tech stocks it's a real killer; when momentum fails, returns are often just less positive.
Valuation's reversed 1999 performance is seen in light of this research as a natural consequence of the fact that technology, however the definition of technology stretches and shifts to include exotic ventures, is inherently a bet on new, risky activities, often currently unprofitable, but with a high payoff potential. The payoff might never come, but investor expectations, not traditional value, drive prices. A helpful way of thinking about the technology sector is as a category for companies that are using new discoveries. Whether it be railroads, electronic power, radio, television, personal computers, drugs or the Internet makes no difference. As the new technologies mature, the companies move to more staid classifications and eventually become utility-like with limited growth prospects and stock prices determined by current opportunities, not by future growth.
Because value plays such a small role in assessing tech stocks, the explanation for their seemingly irrational runup and collapse in the past few years lies in the nature of momentum. Collectively, the factors that define momentum - changes in past prices, changes in consensus earnings estimates, earnings surprise and changes in reported earnings - have weak price sensitivity. This means that as tech stocks advance, rational investors are not looking at price-to-reported earnings or price-to-cash flow ratios, as pure valuation would suggest. Instead, investors properly respond to continually upward-ratcheting prospects. Investors bid up prices dependent only on the improving future scenario. As prices advance, valuation fails by definition. For example, as Amazon.com and America Online Inc. kept increasing their revenue growth, investors increased their projections of future profits even though the companies produced little or no profits. The sky was the limit. Only when growth projections came down did prices start to retreat.
The price retreat for tech stocks was caused neither by the crossing of some overvalued line, nor by some sudden realization that prices were irrationally high. An end to investors' optimism about prospects and a corresponding shift toward pessimism drove prices down. All during the rise and decline of the tech sector, momentum continued to discriminate well among tech stocks, as it has on average in the preceding 20 years.
With most tech stocks down dramatically from their 2000 highs, how do they look now? The tech sector as a whole still shows negative momentum, and likely will underperform the market over the next six months. Yet there is historic evidence that tech stocks can be driven too low. When tech stocks become among the worst in momentum but the best in value, they tend to outperform. Unfortunately, there are only a few stocks in this category now. As the shakeout continues survivors will emerge.
In summary, pure valuation and pure momentum adequately describe the performance of stocks. Sectors vary in the extent to which valuation and momentum are best combined for predictive power. These blends shift over time in a cyclic fashion for all stocks as investors' time horizons vary and within sectors as they evolve with the economy. The sector shift can be measured well enough to forecast. Technology stocks are driven almost entirely by momentum. It is the collapse of growth prospects that precipitated the recent tech sector crash and not the sudden realization that they were overvalued by traditional measures.
John S. Brush is president of Columbine Capital Services Inc., Colorado Springs, Colo.