and not convincing-at least to me. But, recalling Einsteins dictum that the theory determines the observations, the evidence could not stand if the dreadnought was to proceed merrily along annihilating traditional investment practice.

Buying low-P/E stocks appeared successful in studies of past performance. As a practical matter, it had worked for me-no small inducement to belief. Consequently, 1 thought it might be interesting to update the findings, which 1 did, first in Psychology and the Stock Market (1977) and second in Contrarian Investment Strategy (in 1980). Watching the collapse ofthe high-P/E concept stocks through the early 1970s and of the entire two-tier market in 1973-1974,1 saw that things had not changed much. The studies again demonstrated the low P/E case, as indicated by the results of the second study, from August 1968 to August of 1977, shown here.

The experimental design tried to deal as thoroughly as possible with the problems of the previous findings.* The sample, the largest to that time, was constructed from the Compustat 1800 Industrial Tapes, which contain data on the largest publicly held companies in the country. Included in the sample of 1,251 issues were 70% of the common stocks listed on the New York Stock Exchange, as well as large companies on the AMEX and over-the-counter. The study covered the mid-1968-1977 period. The total return (capital gains or losses and dividends paid) was measured in each quarter of the study.

1 used a large number of different holding periods to get as many comparisons as possible of the performance of the various P/E groups. Portfolios were recast according to new P/E information as frequently as every three, six, or nine months, or held unchanged for as long as nine years.

Table 7-3 gives the results for the entire period. Even a glance at the record indicates the superior results of the low-multiple groups over the length of the study. Recasting the quintiles annually according to latest 12-month earnings showed dramatic results. Had an investor put $ 10,000 evenly in the bottom 20% of P/Es at the end of August 1968 and switched yeariy thereafter to keep in the lowest P/Es continuously, he or she would have seen their capital increase $10,326 before commissions through August 31, 1977. These returns occurred during an extremely poor period for the market! (The averages returned only 4.8% annually.) By comparison, had someone switched every 12 months on the same basis into the best stocks, the highest P/Es (quintile 1), his or her $10,000 would have decreased to $9,733. Over the life ofthe study, the lowest P/E stocks more than doubled, while the most-favored group was still near the starting gate.

Table 7-3

A Workable Investment Strategy Annualized Compound Rates of Retum August 1968-August 1977 (Full Period Study)

Even buying and holding the same stocks over the length of the study distinctly favored the low P/Es. Ten thousand dollars in the bottom 20% of stocks in August 1968 held unchanged to August 1977 would have become $18,618, while $10,000 in die top 20% only grew by $743, or 8/10 of 1 % annually. Although the superiority of the bottom over the top group is the most impressive, the second lowest always did far better than the second highest.

The findings once again show the clear-cut advantage of a low P/E strategy. The study covered the two-tier market of 1970-1972, the bear market of 1973-1974 (the worst in the postwar period) and the subsequent recovery. It did not matter whether the investor started near a market top or a market bottom; superior retums were provided in any phase of the market cycle.

The cycle we measured showed the widest fluctuations of any in the postwar period. The average multiple of the S&P varied sharply-from a high of 21 to a low of 7-yet the findings consistently and dramatically supported the strategy of buying and holding out-of-favor stocks. The results were certainly more striking than I had expected. If anything, they indicated that pattems of investor behavior-and error-are far more systematic than one would have believed.

Through the 1980s to the mid-1990s, I completed (witii various collaborators) half a dozen separate studies on low P/E strategies, most of which were published in Forbes. * One study measured the 1,800 largest

* One academic study in the Financial Analysts Joumal was done in collaboration with Mike Berry.

Source: Forbes, June 16, 1986.

companies on the Compustat tapes between 1963 and 1985. The lowest P/E quintile retumed 20.7% annually against 10.4% for the highest. Another divided the 6,000 stocks on the Compustat tapes into five equal groups according to market size, for 21 years ending in 1989. Each group was then divided again into five subgroups according to P/E rankings. The low P/E groups handily outperformed the high P/E groups for all market sizes, ranging from the smallest at around $50 million in market value to the largest at nearly $6 billion.*

Another study measured how low price-to-cash flow performed for the 22 years ending March 31,1985, using 750 large companies. There-suits are shown in Table 7-A. The stocks were again separated into five equal groups and ranked each year according to the ratio of price-to-cash flow. As the chart shows, the most out-of-favor stocks-lowest price-to-cash flow-almost doubled the annual performance of the favorites through this extensive period."

But what of some of the past criticisms? Did any of them still have validity? Our experimental design adjusted for these criticisms, and still provided the results shown above.

All this has been reconfirmed by other research in the late 1970s and early 1980s. Three carefully prepared studies by Sanjoy Basu came up with similar results. In his study published in the Joumal ofFinance in June 1977, Basu used a database of 1,400 firms from the New York Stock Exchange between August 1956 and August 1971. He took 750

Table 7-4

Contrarian Cash Flow