Contrarian Investment Strategies: The Next Generation

David Dreman

You pay a high price for a cheery consensus - Warren Buffett

David Dreman is the chairman of Dreman Value Advisors, and his Kemper-Dreman High Return Fund is one of the all time highest returning funds in America since its debut in 1988. His strategy is based on an understanding of investor psychology, in particular using the insight that the market overprices popular issues and oversells unpopular ones, in short he is a contrarian investor. His book Contrarian investment Strategies: The Next Generation is another very well written book with specific strategies and reams of data to back it all up. If you liked "What Works on Wall Street" and "A Random Walk Down Wall Street" then this is another book with a very similar angle of attack. It is a fairly big book, running over 400 pages, rivalling Malkiel's great tome, I would encourage you to read it yourself because when the book is that long it becomes a bit hard for me to summarise it all without missing out on a lot of good stuff.

The sure thing almost nobody plays

Imagine you are looking around a deluxe, well appointed casino. Off the lavish entry foyer, there are two ample gaming wings, one hued in reds, the other in muted greens. The red wing looks more enticing, but just for now we'll have a look at the green wing as it is much less crowded.

At first you think you must be in the wrong place. The tables are almost empty, everyone is plainly dressed, the table limits aren't particularly high, but everyone seems to be sitting there behind this enormous pile of money.

Then it comes to you, everyone is winning. In fact, as you walk around you can hardly find a losing player. You know that in most casinos the odds are stacked in favour of the house, but here everyone seems to be gaining at a rate of 60% to 40%. You count again and the results are the same.

A pit boss appears at your shoulder. You ask him if this is really true, that the odds really favour the players?

"Yes, indeed. The odds in the green room usually run 60 to 40. It's been that way since we opened."

"But...most of the players must go away winners!"

"They sure do. At those odds, we calculate that 9,999 out of 10,000 make money. At our high-stakes tables at the back, they do even better, with winners running about 20,000 to 1. It's a good thing we get so few players, or they'd break the house."

Realising that you have not a minute to lose and knowing that the few dollars in your wallet aren't going to get you very far, you hurry out to the foyer to go draw out some money.

Before you do though, you glance into the red room. The action level is much higher. Excited punters make huge bets, they frantically change their cash to chips and there is laughing and crying and entreaties to the gods of luck, but no one seems to be winning. In fact the longer you watch the more it seems that people are losing money with every hand they play. You turn to another pitt boss and ask about the odds at these tables.

"60 to 40 in favour of the house", she says, "This room coins gold for the casino, the chances are 9,999 in 10,000 rounds that we wind up winners."

Shaking your head at the stupidity of the players here, you leave quickly to go draw out your life savings.

You return to the casino with your fistful of money, excited, eager for action, all the time figuring how you'll do even better at the game. But then a strange thing happens, you walk into the red wing and start to play.

It all sounds surrealistic, but it isn't. Generations of stock market investors have shunned investment strategies that steadily make money, putting all their dough into speculative long shots and hot stocks with shaky foundations. Despite their best intentions and hard work, mob psychology is such a powerful driving force that it is irresistible. Although following the mob in investing is a sure way not to make good money, it is impossible not to be part of it, impossible to ignore the opinions of thousands of colleagues and fellow investors, to be always in the minority and hold a position that most people consider to be wrong.

Professional money managers in fact are the worst offenders. SEC records show that it was institutional investors that did the most selling in 1987 and had the most money put into cash and bonds. Most "dumb money" amateur investors actually remained fully invested, and reaped the rewards as the market picked itself back up and moved back into positive territory. The "dumb money" mostly sold out of stocks at the high of 1968 and bought at the market bottoms of 1970 and 1974, which is exactly the opposite of what the pros did. In 1990 when the market panicked over Kuwait, almost entirely the selling was done by desperate institutional investors. This panic was largely ignored by amateurs.

John Bogle, chairman of the Vanguard Group, has stated that 90% of fund managers underperformed the market in every 10 year period since records began in the 1960s. Burton Malkiel, of A Random Walk Down Wall Street fame examined every American equity funds and found that they underperformed their benchmarks from 1971 to 1991 and Morningstar and Lipper Analytical Services, two of the largest fund manager tracking services measured the results for 3-, 5- and 10-year periods to the third quarter of 1997 for the six major classes of stock funds, finding that the Standard and Poor 500 Index outperformed all categories of mutual funds for all periods, only small capitalisation funds scored marginally better than the S&P500 benchmark.

Not just fund managers, but also the self-appointed gurus that sell subscription based advice regarding market timing have a poor record. In fact many pundits, including the author have for years tracked the sentiments expressed by mail order gurus and found a highly useful correlation - a contrarian one. If over 70% of these services are bullish, the market may be approaching a top. When over 70% are bearish, it may be a good time to buy. In 1972 at the market high 71% of gurus were bullish, at the low of 1974, 70% were bearish and also just before the bull market really got underway in 1982.

Chapter two of the book takes a swipe at technical analysis and charting, citing extensive academic research and computer testing of market data using pattern recognition, spectral analysis and tracking the success of actual gurus like Joseph Granville, who set himself up almost as a cult leader (even down to flowing white robes and a glass panel which he put under the surface of his pool so he would look like he could walk on water) - only to bomb out so badly in the next decade that the annual return his loyal subscribers got by following his advice was negative 27.4%PA. Looking at simple charting, sophisticated quantitative methods, astrology and cycles he notes that not a shred of evidence has even been provided that technical analysis actually works. So he states as his number one rule:

Chapter three reads very much like Malkiel's "A Random Walk Down Wall Street" and the opening parts of O'Shaughnessey's "What Works on Wall Street" and is an attack on fundamental analysis, timing, fund managers and even the efficient market hypothesis and the Capital Asset Pricing Model and Modern Portfolio Theory. The latter ones he dismisses as "bankrupt theories", but it is nice that he is being so fair about things, so far in the book he should have covered all bases and hopefully offended everybody.

Chapter four goes to town on experts. Criticising not only professional investors but experts from many fields, extending his contempt to all sorts of eminent respected academics and leaders. He talks about the difficulties of forecasting, giving accounts of just how successful experts are not at forecasting of any kind, and making the usual "dismal science" remarks. To apply a soothing balm to all these ruptured tempers he gives a variety of reasons regarding just how difficult it really is to do economic forecasting, but still he points out the enormous difference between expert opinion and historical results.

Predictable Expert Errors

Dreman opens up chapter four with a long and fairly amusing section on how often world renowned experts seem to screw up at every turn. He leads on to make what struck me as a profound insight.

Experts err predictably and often

In chapter four Dreman introduces a few more rules.

Rule two is a reflection on the futility of trying to analyse complex systems. Don't be fooled into thinking that by subscribing to the ultimate data service you will have any meaningful edge over the herd. Having more information seems to increase confidence but not accuracy. Face it, in a system with hundreds of variables no mortal human is going to succeed in fully understanding it. You may as well focus on important matters that are well known, rather than try to understand the more esoteric finer points.

Rule three deals with the trends everyone thinks they see in stock prices and economic variables. There doesn't seem to be anyone out there that can predict any of them with any significant accuracy. Just abandon any strategy that requires forecasting or extrapolation of past movements and tendencies, they don't work.

Rule four relates to the enormous amount of information we try to process in investing. Those who have access to the most market data often don't do any better than people who know little. He warns against information overload, and cautions that all this data often doesn't help investor performance at all.

The chapter goes on to quite study after study showing how poor analysts picks are, choosing great growth stocks at the very peak of their pricing just before they fell off a cliff. In many of these surveys the top analysts really did no better than random chance. These studies no doubt did much to advance the cause of the efficient market theorists back in their day but Dreman argues against efficient markets as well.

At the end of the chapter, after arguing that forecasts are not wrong merely 50% of the time, which would indicate mere chance, but often more than 90% of the time he comes up with another statement which pretty much sums up the whole theory of contrarian investment:

The failure rate among financial professionals, at times approaching 90%, indicates not only that errors are made, but that under uncertain conditions, there must be systematic and predictable forces working against the unwary investor.

Those forces are investor psychology, the speculative mindset that believes that trees really do grow all the way to the sky. Dreman's thesis is that by understanding the difficulties of forecasting and stepping back to take an unemotional distant perspective of expert failures and how psychology works against us, one may see an exploitable pattern emerge. His contention is that speculators are actually quite reliable in the amount that they exaggerate things. Exciting growth stocks are so overbought that when there is a surprise in earnings it will almost always be a downward one, but when a stock is universally regarded as a dog, usually the only surprises are on the upside. Excessive optimism and pessimism are actually so routine that they are actually quite reliable and if you can bring yourself to do the opposite of what the crowd thinks you can exploit this effect.

A 1 in 50 billion shot

Institutional Investor magazine several decades ago formalised the process of determining the "best" analysts. Each year the magazine selects an "All Star" team made up of the "top" analysts in all the important industries - biotech, computers, telecommunications, pharmaceuticals, chemicals - after polling hundreds of institutions. There is a first, second and third team for each industry, and each team appears on the cover of the magazine dressed in American football uniforms with their broker's name on each jersey. Making the team is a tremendous boost to the analysts career and competition is fierce, for making the team is enough to secure the analyst a seven figure income and many times that amount in commissions for the firm.

To make the team is just about the highest accolade that an analyst can achieve, it is an indication that this analyst is a true master of his field and well deserves his salary that is measured in millions. People flock to that firm to take advantage of this genius and the firm makes a fortune in commissions.

But then in 1980 Financial World magazine tried to find out just how good the picks were ("The Superstar Analysts", Financial World, November 1980, p. 16). They encountered a wall of silence from the brokerages themselves, for some reason they weren't letting on about the analysts actual batting average. The magazine had to find out indirectly, via major clients who gave this information "only grudgingly".

The magazine concluded that "heroes were few and far between - during the period in question the market rose 14.1 percent. If you had purchased or sold 132 stocks that they recommended, when they told you to, your gain would have been only 9.3 percent,", 34% worse than by throwing darts. The magazine went on, "Of the hundred and thirty-two stocks the superstars recommended, only 42, or just over 1/3, beat the S&P500." A large institutional buyer of research summed it up, "In hot markets the analysts ... get brave at just the wrong time and cautious just at the wrong time. It's uncanny, when they say one thing, start doing the opposite. Usually you are right."

Of course you can't just point a finger at a small group of analysts over just one year and make a broad conclusion from that. So Dreman collaborated with Michael Berry of James Madison University to carry out one of the largest studies of broker's quarterly earnings forecasts ever done, which was published in The Financial Analysts Journal in May/June 1995.

This study examined brokerage analysts' quarterly forecasts of earnings as compared to earnings actually reported between 1973 and 1991, which has subsequently been extended to 1996. Estimates for the quarter were usually made in the previous three months, and analysts could revise their estimates up to two weeks before the end of the quarter. In all, 94,251 consensus forecasts were used, and they required at least four separate analysts' estimates before including a stock in the study. Larger companies such as Microsoft or Exxon, might have as many as 30 or 40 estimates. More than 1,500 NYSE, NASDAQ and AMEX companies were included, and on average there were about 1,000 companies in the sample.

Many market professionals believe an error of +-5% is enough to trigger a major price move, so to be of any worth the forecasts must be inside that range. The average error of the 500,000 individual analysts reports in fact was 44% annually. In fact this is an average over the whole time, but for some reason analysts estimates actually got worse over the period, in the 1990s the average error was more than 50%.

They tested for skews in the data. Are the errors inflated by a few very large errors that dominate the sample? No, they had a look and errors were fairly evenly paced out and forecasts were consistently bad. They also tracked errors by industry, maybe the methods work well for some stable industries but not others. Again a skew was not found. Analysts were no better at predicting the earnings of stable blue chip financial companies than they were at predicting the earnings of highly speculative stocks. Every industry they looked at had forecast errors that were far too high, except for tobacco, which was the only industry with a single digit error, being 4%. The next lowest error was 25% in telecommunications and foods. What about small versus large companies? The results were a little better for very large companies, but not much. Errors were still 23% on average, almost 5 times too high to be usable.

Dreman even went so far as to codify this into a rule:

If a forecast is to be of any use at all it must be within 5%, yet the error is creeping up to more than ten times that amount, one might ask what use at all is this information? Who would trade on this advice? Obviously following the advice of these analysts is extremely bad for your financial health. But this is exactly how people play the game in the stock markets, experts receive major adulation and billions of dollars are sent after these flaky predictions year after year.

So what proportion of estimates were actually on track? Out of 94,251 estimates Dreman found that 29.4% of them were within plus or minus 5% of actual earnings. Less than half (46.8%) in fact were within plus or minus 10% and only 58% of consensus forecasts were even within 15%, a tolerance level that most Wall Streeters would agree is far too high. This creates a serious problem, because companies are not sold on one year forecasts, but on consensus estimates of profits many years into the future. The forecasters were no better at predicting long term earnings than short term, meaning that if taken as a whole the investor who invests consistently in broker recommended stocks has a cumulative probability of beating the market of next to nothing.

Dreman tested results over booms and recessions, and found no large difference in accuracy for different economic conditions. In all cases analysts were off.

Which direction were the analysts errors? Another study is cited where Jennifer Francis and Donna Philbrick examined analyst estimates from the Value Line Investment Survey, 918 stocks for the 1987 - 1989 period and found that analysts were on average too optimistic, overestimating by approximately 9%. In a report to subscribers, IBES, the largest earnings forecasting service which monitors earnings on over 7000 companies found that the average revision to forecasts for companies in the S&P500 is 12.9% from the beginning to the end of the year in which the forecast is made. Analysts revise their estimates 6.3% in the first half and 19.5% in the second half of the year. Despite the changes, however, analysts recommendations seem to still be on average far too optimistic, they revise their optimistic estimates at the start of the year and triple that revision, usually downwardly in the second half, yet after all this they are usually still too optimistic.

Stock prices are usually set by some variation on a discounting model. This requires extreme accuracy in forecasts for many years in advance in order to make a meaningful estimate of value.

Cragg and Malkiel (the latter being the guy who wrote "A Random Walk Down Wall Street") did an early analysis of long-term estimates, published in the Journal of Finance (23, March 1968, "The Consensus and Accuracy of Some Predictions of the Growth of Corporate Earnings" pp 67-84), looking at the projections made by groups of analysts at five respected firms, covering 185 stocks. The researchers found that most analysts estimates were based on linear extrapolation of current trends with low correlations between actual and predicted earnings.

They found that analysts would have substantially improved their accuracy if instead of extrapolating past growth rates they had simply inserted the long term company average growth rate of 4% annually.

Another study, by Oxford professor I.M.D. Little ("Higgledy Piggledy Growth", Bulletin of the Oxford University Institute of Economics and Statistics, November, 1962) found that corporate earnings in fact seemed to follow a random walk, with little correlation between past and future rates. Recent trends provided no insights useful for forecasts.

A number of studies reach the same conclusion, that changes in company earnings fluctuate in an essentially random fashion.

Well, this is certainly interesting, analysts assume the future will be like the past yet it appears that this assumption is untrue. Therefore we might expect substantial forecast errors. And large forecast errors are indeed what we find.

Career pressures and forecasts

As well as general troubles with forecasting, there are a variety of other pressures that act directly on analysts. John Dorfman, then editor of the market section of The Wall Street Journal provided a list of what determines an analyst's bonus, normally a substantial part of his salary. In Dorfman's words, "Investors might be surprised by what doesn't go into calculating analyst's bonuses. Accuracy of profit estimates? That's almost never a direct factor.... performance of stocks that the analyst likes or hates?... It is rarely given major weight."

The ranking of seven factors determining an analyst's remuneration places "accuracy of forecasts" dead last.

What is most important is the amount of sales commissions he generates for the firm. Many firms conduct a formal poll of the sales force, which ranks the analyst primarily on how much commission business the analyst can drum up. At one firm commissions amount to 50 percent of the bonus, another allocates a point system for buy and sell recommendations, a buy is worth 130 points and a sell only 60. Sell recommendations generate far fewer sales.

There is more to it than this though. Analysts run a substantial political risk by downgrading a company. If an analyst recommends selling a company that company may ban him from further contact. If he puts a sell order on an entire industry he may receive an industry blackball, which virtually excludes him from talking to any important executives. If the analyst is a specialist in that industry, and it represents an important part of his intellectual property, issuing a sell recommendation could end his career.

Even if the analyst is right it can still be costly. When an analyst issued a sell recommendation on an Atlantic City casino owned by Donald Trump back in the late 80s, Trump was infuriated and insisted the analyst be fired for his lack of knowledge. Shortly after, the analyst was fired, but for "other reasons". The analyst was correct though, and soon the Casino went broke. Out of the job, the analyst won the equivalent of a few years of salary from an arbitration panel. The analyst was never rewarded for his excellent sell call and was in fact damaged by it.

Sell orders don't just affect the analyst personally though. They can damage the entire firm. When an analyst at Prudential wrote some negative reports about Citicorp in 1992, Citicorp retaliated by refusing to give Prudential any business in underwriting bond issues. The same happened a year later when the same analyst criticised Banc One for its complex derivatives holdings, which eventually cost Banc One millions of dollars in write-offs. Banc One ceased its bond trading with Prudential. Soon after, apparently coincidentally, that analyst left Prudential.

For analysts at firms that are large underwriters, the pressure not to issue a sell recommendation on a company's stock is great. A study by Dugar and Nathan examined 250 analyst reports from investment banking houses, matching them up against 250 from brokerage firms that did not conduct investment banking. Investment bankers issued 25% more buy recommendations and a stunning 46% less sell recommendations.

Peter Siris, a former analyst at UBS Securities summed it up, "There's a game out there. Most people aren't fooled by what analysts have to say ... because they know in a lot of cases they're shills. But those poor [small] investors - somebody ought to tell them."

What is clear is that analysts aren't really needed to be good forecasters, they just need to spin a good yarn and excite the speculators. While the pressure is great for analysts not to issue a sell recommendation, instead using a variety of other terms , it would be very rare for an analyst to be criticised for issuing a conservative estimate. The question then remains, why are analysts so optimistic about stocks, when they do have the option to be conservative.

Over the next few pages, Dreman talks about psychology and overconfidence, mostly issues already in this FAQ in the article on trading psychology.

Forecasters are excessively prone to treat each problem as unique. They ignore base rates and focus instead on the unique factors that make this case what it is. Rather than keep an eye on historical norms and understand what has happened every time in the past when a situation has come up, they focus on why this time it is different. Forecasters focus on the "inside view", to understand this stock as an entity that is absolutely unique, and refuse to draw parallels. It requires essentially a detailed view of the future. The "outside view " however, is statistical and comparative. The outside view notes historical growth norms and looks at how accurate past reports have been and bear past errors and exaggerations in mind while assessing the probability that the company will exceed averages.

Unless you have genuine reason to believe by some concrete guarantee that this stock differs fundamentally from the thousands that came before, it is more sensible and frequently more accurate to assume that the company will probably grow at an average rate and encounter the same problems as all the others. "This time it is different" has been the last gasp of many a speculator just before a correction wiped him out. In general there are so many variables affecting the price of a specific stock that it is practically impossible to really keep track of them all. The specific outcome reflected in prices is so improbable that it makes no sense to assign a single precise number to any forecast.

As forecasters well know, an estimate must be within 5% of earnings in order to not have a massive effect on the price of the stock. Only 29% of forecasts are within this range in any one quarter, so Dreman prepared a probability tree to demonstrate your chances of making it through a 5 year period unscathed by earnings surprises if you buy broker recommended stock.

The chances of surviving without an earnings surprise more than 5%.

  Any surprise   Negative Surprise   Positive Surprise  
1 quarter 29% 63% 67%
4 quarters 1/130 1/7 1/5
10 quarters 1/200,000 1/110 1/58
20 quarters   1/50 billion 1/12,000 1/3,400

Note these are the chances of not receiving an earnings surprise. For example your chances of going for one year, without your stock taking a dive because analysts overpriced it, are only one in seven. This means you have a six in seven chance of being bludgeoned by a negative earnings surprise some time over the next five years. The chances of brokers actually anticipating profits to within 5% for five years are a staggering one in fifty billion. You are ten times as likely to win a major lottery as you are to own a growth stock without it taking a big hit or push from analysts dodgy forecasts over five years.

Even with the broader range of a 10% error in forecast, the story is still pretty sad.

  Any surprise   Negative surprise   Positive surprise  
1 quarter 47% 70% 77%
4 quarters 1/21 1/4 1/3
10 quarters   1/2,000 1/35 1/14
20 quarters 1/4 million 1/1,250 1/200

So don't budget too carefully just yet based on the stocks your broker sold you, you have only a one in four million chance of getting through with all earnings forecasts correct to within 10%. To get through five years without a negative surprise you are looking at odds of one in 1,250. These are not good odds!


People are likely to be prone to overconfidence because of three factors. People have unrealistic optimism about future events, unrealisticly positive self-evaluations and an unrealistic confidence in their ability to control the situation. The odds of predicting the stock market seem almost negligible according to the odds above, but even when analysts know about them, they still believe that their superior ability to understand this particular instance places them above the statistics.

A similar affliction affects managers of companies, they never see the downside in terms of a really worst-case scenario. Usually their "worst case" is actually quite mild.

An analyst knows that the stratospheric valuation of the stock he is recommending risks a violent collapse if earnings falter, but this does not bother him because he is sure he knows the company so well that earnings will not fail, and that the gamble will pay off.

Paying through the nose for growth

Companies with the best prospects, highest growth rates and most exciting concepts normally trade at a high price relative to earnings (Price Earnings Ratio, PER), cash flow (PCR) and book value (PBR) and invariably provide low or no dividend yields. Conversely stocks with poor outlooks trade at low PERs, PCRs and PBRs and often have a respectable dividend yield.

In 1996, Netscape traded at a PER that was 23 times as high as the American financial group that repackages mortgages, the Federal National Mortgage Association (Fannie Mae), who themselves are a formidable growth stock having managed to increase their earnings at a rate of 20% annually over the last 20 years. But investors will pay these differentials, which can be absolutely staggering, because of their confidence in their ability to pinpoint the future. It is interesting to see what happens when an earnings surprise hits the stocks.

Dreman wanted to see how earnings surprises, both positive and negative affect the prices of expensive and cheap stocks. With his collaborators, Nelson Woodard, Eric Lufkin, Michael Berry and Mitchell Stern he looked at the 24 years prior to 1996. He used exactly the same analysts' consensus forecasts as before.

They wanted to measure a number of factors important to investors. What do analyst forecasting errors do to stock prices? Do earnings surprises have the same effect on favoured as unfavoured stocks? Thirdly, they measured just how accurate investors expect analyst forecasts to be. To measure this they looked at the effect of even tiny surprises, considering anything over one cent a share difference as a surprise.

They ranked stocks into five sets of quintiles ranked by PER, PCR and PBR. The top quintile (quintile = 20% of sample) were the most favoured stocks, the bottom quintile were what you might call the dogs.

Prices were measured against consensus forecasts every quarter between 1973 and 1996, 95 quarters in all were studied and between 750-1000 companies in each of the 95 quarters of the study.

Price earnings ratio and surprise

Overall, the bottom (cheap) quintile of stocks responded positively to earnings surprises. Per quarter they averaged 1.5% above market and over the full year beat the market by 4.2% per annum. This means that the combined effect of all surprises, positive and negative, worked in favour of dogs.

Middle quintile stocks performed just below neutral, underperforming by 0.2% per quarter and 0.5% per annum.

Top quintile stocks though underperformed, returning 1% per quarter less, or 3.5% per year.

The conclusion, stocks with a low price to earnings ratio surprised the market by returning 4.2% per year above the general indexes, expensive stocks disappointed by 3.5% per year.

Price to cash flow ratio and surprise

Cheap stocks outperformed by 1.2% per quarter, 4.1% per year. Median stocks underperformed by 0.2% per quarter, 0.5% per year. Expensive stocks underperformed by 0.8% per quarter, 3.2% per year.

Price to book value ratio and surprise

Cheap stocks outperformed by 1.0% per quarter, 4.0% per year. Median stocks underperformed by 0.2% per quarter, 0.6% per year. Expensive stocks underperformed by 0.7% per quarter, 3.0% per year.

So whatever valuation method you used out of PER, PCR or PBR, favoured stocks generally failed to meet expectations, cheap stocks surpassed the averages. Remember, this is for all earnings surprises, good and bad. The aggregate surprise is positive for cheap stocks and negative for expensive ones.

Price earnings ratio and positive surprise

Cheap stocks outperformed by 3.6% per quarter, 8.1% per year. Median stocks outperformed by 2.3% per quarter, 3.1% per year. Expensive stocks outperformed by 1.7% per quarter, 1.2% per year.

PCR and PBR, while not listed, gave similar results. The conclusion is that dogs responded well to positive surprises, rallying nicely. Expensive stocks, already propped up by optimistic speculators already expected a good profit so gained little.

Price earnings ratio and negative surprise

Cheap stocks underperformed by 0.7% per quarter, 0.1% per year. Median stocks underperformed by 2.8% per quarter, 5.0% per year. Expensive stocks underperformed by 4.3% per quarter, 8.9% per year.

In other words, cheap stocks hardly budged when they disappointed the market, but expensive stocks took a savage beating. The results for price to book ratio were similar.

The effect was even more dramatic for cash flow.

Price to cash flow ratio and negative surprise

Cheap stocks underperformed by 0.8% per quarter, 0.2% per year. Median stocks underperformed by 2.8% per quarter, 5.0% per year. Expensive stocks underperformed by 4.8% per quarter, 11.3% per year.

So once again expensive stocks took a beating when they disappointed the market.

Not only does an earnings surprise give a stock a once off boost or fall, the effect seems to continue over time. Stocks that had traded cheaply tend to rally for an extended period of time, outperforming by many years. Conversely expensive stocks go into a period of prolonged underperformance. Over 20 quarters cheap (low PER) stocks outperformed the market by 34.7% when they positively surprised, but interestingly outperformed the market by 18.2% even with negative surprises. Their tenacity has to be admired.

Over 20 quarters, expensive stocks (high PER) underperformed by 27.1% even with positive surprise and underperformed by 44.7% when negatively surprised. On the other hand, there seems to be no lasting effect on median quintile stocks past the initial moves in the quarter of the surprise.

The reevaluation process is not rapid, in fact analysts are usually quite slow to update their recommendations, only slowly lifting a dog to a buy rating and demoting a growth stock to a sell. This means that when a company is undergoing a change of fortune it will surprise the market again and again for months or years.

Dreman did another test. To see how value strategies went in a bear market, he tested price-to-earnings, price-to-book, price-to-cashflow and dividend yield as well. In every bearish period between 1970 and 1996, the market averaged a 7.5% PA fall. Low price-to-book did better, losing 6.2% PA. Low price-to-cash flow did better again, losing only 5.8% PA and low price-to-earnings lost 5.7% per annum. However the star bearish strategy was dividend yields. Stocks with the highest 20% of dividend yields lost only 3.8% PA during bearish times. This is to say that in a bear market you will outperform the market averages by 3.7% per annum.

Contrarian strategies

This next section reads very much like What Works on Wall Street and his conclusions are the same. Therefore my summary will be very brief.

Strategy #1: Low Price/Earnings

Quintile   Dividend return   Capital appreciation   Total return
Lowest 6.0 13.0 19.0
2nd 5.6 11.8 17.4
3rd 4.5 10.1 14.6
4th 3.3 9.8 13.1
Highest 1.9 10.4 12.3
Market 4.3 11.0 15.3

Strategy #2: Low Price/Cash flow

Strategy #3: Low Price/Book value

The numbers are pretty much the same as for low PER, give or take half a percent. The interesting thing to note is that cheap stocks return superior income and superior appreciation. This makes them ideal for investors that want high growth as well as investors that want income.

Another major advantage of a value approach is that the amount of trading is low. Although contrarian stocks tend to outperform the market, the market does grow with them as well and so it does take a fairly long time for them to fully catch up and become as expensive as the general market. As a result, holding times are quite long and a low portfolio turnover is quite feasible.

At this stage Dreman spent half a page or so explaining how expensive trading is, talking about bid/ask spread, brokerage and taxes. If you want to know more there is another article in this FAQ on the topic.

Strategy #4: Price-to-Dividend

Quintile Dividend return Capital appreciation Total return
Highest 8.0 8.2 16.1
2nd 5.4 12.1 17.5
3rd 3.9 11.2 15.1
4th 2.2 11.6 13.8
Lowest 0.7 11.5 12.2
Market 4.0 10.9 14.9

Dividend yield doesn't work as well as PER, PCR and PBR, the returns of the lowest quintile aren't as good and returns come to a large extent in the form of dividends (should this be surprising?) which may or may not be franked. (In America they are unfranked, so Dreman considers dividends in a less favourable light than we Aussies do because we have imputation and they don't). Either way, returns are inferior to those you can get with the PER strategy, a result not inharmonious with O'Shaughnessey's work. I am a bit curious though why the "market" returns more with the PER strategy than this one.... [sic]

There are a couple more differences as well. First of all the highest yielding quintile is not the highest returning one, the second highest yielding quintile is. Appreciation isn't too great for highest yielding quintile, being beaten by the market. However it is interesting that with dividends reinvested high yield stocks actually get better over a longer holding period. Dreman recommends this approach for those that depend on an income as the returns are somewhat less volatile than the bond market (believe it or not with all the interest rate changes over the last decades bonds have been more volatile than stocks!), but for wealth builders there are better strategies.

There is a plus for the dividend strategy though, in bear markets it was the strategy that lost the least. It is good for very defensive investors who are afraid of losing a lot when markets go south.

Contrarian stock selection

If you are going to apply the method mechanically, Dreman strongly recommends you diversify so you don't take too big a hit by buying total duds. Note also that Dreman studied only large companies and can't guarantee that the methods work on small ones (according to O'Shaughnessey, they don't as a matter of fact). Dreman prefers big companies because they are easier to trade, are under more scrutiny and less likely to manipulate their earnings with creative accounting and also because of their visibility usually attract the market's attention when they turn around much more quickly than a really anonymous stock.

But do you abandon security analysis completely? Dreman doesn't think so. He uses an eclectic approach and looks at the company accounts to check that the company really is a bargain. This is where you can tie in all the Warren Buffett security analysis and business evaluation with the mechanistic value approach.

Dreman doesn't think that security analysis is worthless, instead he recognises that it is extremely difficult to forecast profits years in advance, particularly when "growth" stocks are involved. Bearing in mind these limitations, he exploits what methods of security analysis are available that do not depend on precise forecasts.

There are five indicators that Dreman looks at.

A strong financial position

This means looking at debt and assets. A cheap stock relative to earnings is not cheap if the liabilities are so great that the company is going into bankruptcy. Make sure the company has sufficient reserves to survive the crisis and conservative debt levels.

As many favourable operating and financial ratios as possible.

Return on equity (ROE), return on assets (ROA), profit margins, low expenses. All that stuff.

A higher rate of earnings growth than the S&P500 in the immediate past, and the likelihood that it will not plummet in the near future.

Contrarian companies have already been to the dog house, they have often made enormous losses and this is what put them where they are today. If the company makes even a modest profit therefore, the growth in earnings will probably be astronomical, so this is not an onerous condition.

After some harsh experiences, Dreman learned to keep an eye out for companies that still have a long way to fall. He does look at Wall Street forecasts to a certain extent, to see if forecasts show huge continued losses. It is often the case that a stock will not pick up straight away when it appears the worst is really over, so there are still opportunities for contrarians even when they do take advantage of analysts research.

Dreman pays little attention to precise forecasts, but he has found that researchers are correct often enough about the general direction of earnings that he knows to avoid a stock when the researchers are still overwhelmingly convinced that earnings still have further to fall. There is a difference here. Dreman says that analysts may well be out by 44%, but if they say earnings will continue falling then he may well just hold off purchase until the worst appears to be over. This is a much simpler and less error prone approach than is commonly used and does not depend on analysts having any great precision.

Earnings estimates should always lean to the conservative side

Graham and Dodd's "margin of safety" pops up to say "hello". If you cut the forecasts in half, but even still the company looks alright, you may have a potentially rewarding investment.

An above-average dividend yield, which the company can sustain and increase.

This of course relates to the previous four indicators, the dividend cover (ratio of earnings per share divided by dividends per share, proportion of profits paid as dividends) should be reasonable. A cut in the dividend won't help the share price at all, so if the company pays one then make sure they can keep it going.

Contrarian strategies within industries

Now in theory, contrarianism should work within specific industries as well. The same forces that exaggerate outlooks for individual companies within the broad market should have pretty much the same effect within an industry. Dreman tested that idea with the 1,500 largest companies on the Compustat database and found this to be the case. A similar relationship where the cheapest stocks outperformed the more expensive ones was apparent.

Not only did the cheapest stocks in each industry tend to outperform that industry, they also did well compared to market averages.

This is nice because it allows you to go contrarian and enjoy growth in a diversified portfolio of industries. You don't need to forecast which industry is due for a bounce, you will buy the stocks with the least to lose in the most expensive industries and the stocks with the most to gain in the ones that are going to boom.

Why this works is speculative, but it appears that company fortunes do change over time. Industry laggards tighten their belts, improve their management, and find ways of increasing their market share or developing new products, which results in their continued outperformance of the market for long periods. The player with the worst market share has the most to gain and the least to lose compared to an industry leader that comes under attack from all sides.

One might think that a more macroeconomic approach of going contrarian in industries would work - you buy stocks in the cheapest industry. Dreman's tests found that this isn't really the case, the cheapest stocks in the cheapest industries performed about the same as the cheapest stocks in the most expensive industries, though the cheaper industry was marginally in front.

Although the strategy marginally underperformed absolute contrarianism (cheapest stocks in the stock universe, ignoring any deliberate attempt to diversify or impose restraints on investment in any given sector), it does have the important psychological advantage that it will necessarily get you a few stocks in "hot sectors" so you won't feel left out and periods of underperformance, which do occur in contrarianism often enough to shake off a few disbelievers and fling them into growth investing, will be less severe.

Dreman's tests also found that the contrarian strategy within industries performed well in bear markets, losing less than the market averages when markets fell. As before, the best bearish defense seems to be the high dividend yield strategy.

Dremen spends a couple of pages talking about when to sell, but it all mostly boils down to this one rule. Too many managers take the "round trip" of holding a stock all the way up and holding it all the way down again.

Sell when the stock has reached the same price as the general market, a new contrarian stock will have better prospects. Also sell it if the stock appears to have a deteriorating outlook. Don't hold bad stocks just because you are a contrarianism, you are looking for major declines in the company's fundamental ability to make money. Sell them and be ruthless. The idea of contrarianism is to buy undervalued stocks, not bad stocks.

Also what to do about stocks that go nowhere? Dreman says it is all pretty much a matter of choice, but suggests you give a company 2.5 or 3 years to work out. John Templeton, another one of the great masters of value investing gave a stock six years.

Chapter ten is about heuristics - mental shortcuts that everyone makes that in the physical world generally help us to make sense of complicated situations, or alternatively to jump to conclusions and make a mess of things. The chapter mostly focuses on the latter aspect of heuristics.

The next chapter has some interesting tables. Dreman tested how contrarian stocks performed before they became contrarian stocks and how favoured ones went.

  Price movement before Price movement after
  (-5 to 0 years) (0 to +5 years)
Low PBR  56.2% 153.3%
High PBR  363.5% 88.8%
Market 141.3% 119.1%

Nothing too shocking there, after spending 248 pages telling us that cheap stocks outperform he gives us another table that demonstrates this, ok....

Then he shows a table demonstrating how the "fundamentals" of the companies change in the period before and after the contrarian purchase.

    -10 to -5 years   -5 to 0 years   0 to +5 years  
Growth in cash flow  Low PBR 11.5% 9.9% 12.2%
Growth in cash flow High PBR 21.3% 26.1% 15.8%
Growth in cash flow Market 14.3% 15.5% 12.2%
Sales growth Low PBR 12.4% 10.4% 6.8%
Sales growth High PBR 18.0% 21.0% 14.9%
Sales growth Market 13.8% 13.6% 10.1%
Earnings growth Low PBR 9.3% 6.4% 11.6%
Earnings growth High PBR 18.6% 24.6% 12.1%
Earnings growth Market 12.2% 14.2% 10.6%
Return on equity Low PBR 10.5% 9.7% 8.1%
Return on equity High PBR 15.3% 17.9% 17.4%
Return on equity Market 12.4% 13.0% 12.3%
Profit margin Low PBR 9.8% 7.4% 5.3%
Profit margin High PBR 12.2% 13.6% 12.8%
Profit margin Market 11.1% 10.2% 8.7%

The implications of this table are staggering! Note that the growth stocks that were awarded high multiples in the first place continued to be highly successful growth stocks after the purchase as well. All of the numbers are favourable for the growth stocks throughout, they continued to be superior to the market - and yet they underperformed. Yet the contrarian stocks remained lousy companies, their growth and return on equity was still below the market averages after the purchase, and yet they outperformed.

Note the changes in the numbers. Growth stocks went from great to really good, dogs went from bad to mediocre. The result was that the market, which expected the impossible from growth stocks, yet had the lowest possible expectations for the dogs had to revise its opinions on both of them. What is important is that dogs performed a bit better than expected and gained a lot, growth stocks performed a bit worse than expected and underperformed drastically.

Growth stocks hardly respond to good news as they are already priced to take good news into account, the only thing that affects them is bad news. The converse applies to dogs. This also shows that analysts are reasonably good at spotting growth stocks, the problem lies in them being ridiculously overpriced and having expectations that are often unachievable.

Dreman then comes back to the idea of regression to the mean. That no company can continue to grow at a very rapid rate indefinitely, and it will eventually reach an equilibrium of sorts. Ben Graham actually summed up the problem in Security Analysis:

The truth of our corporate venture is quite otherwise [than investors think]. Extremely few companies have been able to show a high rate of uninterrupted growth for long periods of time. Remarkably few also of the large companies suffer ultimate extinction. For most, this history is one of vicissitudes, of ups and downs, with changes in their relative standing.

Dreman makes this a rule:

But somehow no one ever seems to appreciate this. The tendency to extrapolate trends indefinitely is a powerful one, and leads those who are aware of this to have an important advantage, which brings us to:

Crisis Investing

Contrarianism works in many different ways, but all depend on profiting from people going too far. Of course a crisis is not good for the economy, little good ever comes from a war or a political scandal, yet it is very good for certain investors.

International and domestic crises often create market panics. A contrarian of course appreciates that the market will of course panic too much. Rush in and buy when everyone else is running for the door.

Here is what the Dow Jones did after major international crisis events:

  Market low after crisis Following year Following 2 years
Berlin Blockade 19 July 1948 -3.3% 13.2%
Korean War 13 July 1950 28.8% 39.3%
1962 stock market break 26 June 1962 32.3% 55.1%
Cuban missile crisis 23 October 1962 33.8% 57.3%
Kennedy assassination 22 November 1963 25.0% 33.0%
Gulf of Tonkin 6 August 1964 7.2% 3.1%
1969/1970 stock market break 26 May 1970 43.6% 53.9%
1973/1974 stock market break 6 December 1974 42.2% 66.5%
1979/1980 oil crisis 27 March 1980 27.9% 5.9%
1987 crash 19 October 1987 22.9% 54.3%
1990 Persian Gulf War 23 August 1990 23.6% 31.3%
Average appreciation   25.8% 37.5%

Alright, not many people buy right at the bottom, but even if you were 10% off you would still enjoy a healthy profit and would have lost money in only one year in the eleven, and then only 3.3%.

(Not mentioned in Dreman's book, but I'll take the liberty, contrarian crisis investing is probably a pretty good way to describe Sir John Templeton's approach to investing. He used to go to countries where there was blood on the streets and bombs in the parliament and buy up big. After Warren Buffett and Peter Lynch one would call him arguably the world's third greatest investor.)

After saying that, Dreman gives a few more pages to the importance of fundamental analysis so you end up buying quality companies that are genuinely undervalued. He accepts that a blind mechanical contrarian approach will lead to the purchase of a fair few really awful stocks if you aren't careful. Nevertheless, he still endorses diversification.

(These value lifelines to which he refers are low PER, PCR and high dividend yield).

What is risk?

Chapter 14 is a good solid bashing of volatility based definitions of risk and Modern Portfolio Theory. I thought the chapter summed up how bad this whole concept is that I incorporated most of it in another article in the section on portfolio construction "Modern Portfolio Theory criticisms".

Take a minute or two to read that article, and then you'll know where the next rule fomes from:

Volatility is not a reason to sell a stock, and it doesn't mean the stock has become more risky. Even the best blue chip companies get hit by occasional extraordinary events, tumble sharply and become very volatile for a while. This is not something to avoid, it is a gift of opportunity. To a value investor the more a stock has been sold below intrinsic value, the more attractive - and the more safe it becomes.

Volatility does not tell you anything about a company's financial strength, it tells you nothing about a stock being very overpriced. All it tells you is how rapidly traders changed their mind about a stock in the past. This information is not particularly useful to most rational investors.

Alternative measures of risk

I'm yet to meet an investor who considers a stock moving up to be a bad thing. Yet volatility (standard deviation of returns) makes no distinction at all between upward movement and downward movement. To a volatility based model, a +50% surge in prices is just as much a catastrophic event as a similar crash, and must be avoided.

There is a much better measure of volatility than standard deviation, and while it may surprise many to find that most CAPM theorists reject this alternative measure with surprising acrimony. Semivariance is a volatility measure that only looks at downward volatility. All upward movement is ignored as "risk", price falls are regarded as risk events. In this model a 50% fall is regarded as a bad thing, a 50% rise is ignored in the risk assessment.

While semivariance is obviously a far more sensible measure to use than standard deviation, it is still unrealistic in that it is still just a measure of price volatility, and while semivariance makes more sense as a means of keeping score with "risk adjusted returns", it still doesn't have any predictive value.

More realistic measures of risk

In the old days, before the Second World War, inflation was not a tremendous concern. Inflation rates were close to zero for centuries, only in very short periods of war and upheaval, or during the speculative binges of Tulipomania and the South Seas Bubble did it ever really become noticable. In Victorian times bond yields were only 2 or 3%, yet for their time they were superior investments to the bonds of today, because with inflation at next to nothing and taxes hardly a problem there was still an excellent return on investment.

Inflation and taxes have changed everything. Bonds and Treasury Bills no longer provide an adequate return by any standards. Inflation and taxes are so severe that when taken into account those investments usually regarded as most safe are seen in an entirely different light.

Returns after inflation is taken into account. 1946 to 1996. Percent of times stocks beat bonds and treasury bills.

Returns Stocks beat...
Holding period Stocks Bonds T-bills    Bonds T-bills
1 year 7.5% 0.9% 0.4% 65% 67%
2 years 15.6% 1.7% 0.9% 73% 78%
3 years 24.3% 2.6% 1.3% 78% 90%
4 years 33.6% 3.5% 1.7% 82% 84%
5 years 43.7% 4.4% 2.1% 84% 82%
10 years 106.5% 9.0% 4.3% 94% 86%
15 years 196.7% 13.8% 6.5% 100% 94%
20 years 326.3% 18.8% 8.8% 100% 100%
25 years 512.6% 24.0% 11.1% 100% 100%
30 years 780.3% 29.4% 13.5% 100% 100%

Lets put that another way. Your chances of beating stocks with bonds and T-bills (T-bills can be approximated as "cash" since these short term government securities are what cash management trusts invest in):

Holding period Bonds T-bills
1 year 35% 33%
2 years 27% 22%
3 years 22% 10%
4 years 18% 16%
5 years 16% 18%
10 years 6% 14%
15 years 0% 6%
20 years 0% 0%
25 years 0% 0%
30 years 0% 0%

Even over a period as short as one year, when the standard wisdom would point you away from stocks and into cash and bonds you still have a fairly low probability of beating stocks. A more rational definition of risk would be "the chance of getting an acceptable return on investment". We usually think of a risky investment as a long shot that pays well, a one in a hundred chance of making a huge return. If we were to think of the least risky investment as the one with the highest chance of doing well, we'd go for the one with the best track record. Cash and bond investments are absolutely pathetic, even if you had no worries about taxes and fees, after 30 years you would have made only 13.5% or 29.4% respectively. Take out taxes and we'd see that even this small premium is gone. An "investment" that offers no return on capital, one that in fact over long periods doesn't even really guarantee return of capital (on a basis of at least maintaining purchasing power) is not safe.

Over a moderate time frame, of 5 years, you have an 84% chance of beating bonds, an 82% chance of beating cash. These are not speculative odds, they are excellent probabilities and not what you would expect from a "risky" investment.

So exactly what effect do taxes have on these investments? This chapter has many tables, but one that is particularly interesting is similar to the one above, with taxes taken into account.

Returns after inflation and taxes are taken into account. 1946 to 1996. Percent of times stocks beat bonds and treasury bills.

Returns Stocks beat...
Holding period Stocks Bonds T-bills    Bonds T-bills
1 year 4.4% -1.8% -1.7% 67% 67%
2 years 9.1% -3.6% -3.4% 76% 80%
3 years 13.9% -5.4% -5.1% 84% 86%
4 years 18.9% -7.1% -6.7% 86% 86%
5 years 24.2% -8.8% -8.3% 88% 82%
10 years 54.3% -16.8% -16.0% 90% 86%
15 years 91.7% -24.1% -23.0% 100% 92%
20 years 138.1% -30.7% -29.4% 100% 100%
25 years 195.7% -36.8% -35.3% 100% 100%
30 years 267.3% -42.3% -40.7% 100% 100%

You may note that Dreman has had to make some assumptions about taxes and they work out a little different for Australians. The footnote concerning how he calculated taxes showed he taxed dividends, bonds and tax at 50% until 1988 and 35% from then on. Australians would be ahead for stocks, we have dividend imputation and the yanks do not. Our marginal tax rates are still almost 50% (48.5% at their highest, 21.5% for most of us), they have not been reduced, so our taxes would have affected this sample in favour of stocks again. He assumes capital gains taxes of 25%, I guess they must have something similar to our 50% assessment rule for stocks held more than a year. In all, the calculations probably would have improved the returns of stocks relative to bonds and cash if Australian rules were taken into account.

A more workable definition of risk

A realistic definition of risk recognises the potential loss of capital through inflation and taxes, and would include two factors: Taking these into account you will see that bonds and cash offer an enormous risk to long term investors. Taking time frame into account, something volatility measurements do not do, changes the equation entirely.

A long term holder of bonds has had about the same success as someone that bought on the day before the market fell in October 1987 and sold a week later, absolute devastation! Yet the volatility based definition of risk still reigns supreme. There are millions of pensioners around the world who have decided that now they have retired they cannot afford to have stocks in their portfolio. If you retire at 55 and live to 85, your "safe" investment strategy would have had the same effect, with a little less drama perhaps, of a series of stock market crashes!

In the remainder of the chapter Dreman shows worst case scenarios for stocks, how you would have done by buying at exactly the wrong time, compared to long term portfolios of "safe" investments. Stocks streaked ahead in each of the tables, worst case for stocks was still significantly more cheerful on average for every time period than best case for safe investments. The number of years where bonds and cash did outperform stocks after inflation and taxes were so few and far between that they hardly were worth mentioning. Even over a one year period stocks were ahead.

Small stocks

Dreman spends the first half of chapter 15 demolishing the "small-company effect". This famous phenomenon was brought into being by two valient young Ph.D. students, Rold Banz and Marc Reinganum, trying to defend the efficient market hypothesis against a barrage of attacks from value investing barbarians, Dreman among them. During the 1970s a number of papers appeared, showing that cheap stocks outperformed expensive stocks.

These students measured the returns of stocks traded on the New York Stock Exchange between 1926 and 1979. Stocks were sorted by market capitalisation into quintiles (five groups, each containing 20% of the sample, in case you've forgotten what a quintile is), over 5 year periods for the length of the study. Over the entire 54 year period the average rate of return of small companies was 11.6% compared to 8.8% for large stocks.

The experimentors were unsure about why small companies did better, but they were very quick to claim that this was entirely contradictory with the small PE effect. Fortune magazine, then a champion of efficient markets and Modern Portfolio Theory published their results in 1980,

[T]he small-stock phenomenon has indirectly refuted the most serious challenge yet to the efficient-market theory. A number of researchers have demonstrated that portfolios of stocks with low price-earnings ratios have regularly outperformed the market averages. That finding, trumpeted by David Dreman in his book Contrarian Investment Strategy, is wholly inconsistent with an efficient market. It turns out, however, that low P/E stocks appear to offer superior returns only because small stocks have lower P/Es, on average, than large ones.

Soon a fund was set up. Dimensional Fund Advisors was formed in late 1980 to take advantage of the small cap effect. On the board serving as advisors were ten high powered University of Chicago professors or alumni, including Roger Ibbotson, Eugene Fama, Merton Miller, Myron Scholes and Rolf Banz. Rex Sinquefield, a colleage of Ibbotson and one of the founding partners of Dimensional Fund Advisors, when asked what he thought of active portfolio management, sneered, "crap". Sinquefield was hyping his new product at the time, based on the amazing new discovery of small-cap stock outperformance.

Sinquefield went on, "It's not a fluke. Our computer tells us it happens but none of us knows why ... We're plugged into the academic world, where all the really good research is taking place. We have a kind of tie-in to Mecca". One could certainly not fault the marketing machine that they brought into play, by 1996 Dimensional Fund Advisors had $11 billion under management, a good part of that in their small-cap product. Along with their marketing cronies, the professors happily went along with the whole scheme, suddenly becoming wealthy beyond their previously wildest expectations.

So how did they perform, with their Mecca ties and brute computer power? Not too well as it turns out. From its inception in 1981 to the end of December 1997, the DFA 9-10 Fund trailed the S&P500 by 54%. Some of the academics involved with the fund were shocked, one said "This has never happened before, we're stunned."

DFA continues to market their small-cap funds on the basis of the same historical returns, reporting that small stocks returned 45% annually at the bottom of the depression. So what went wrong?

Not long after the 1980 article in Forbes, Dreman had a closer look at Banz and Reinganum's, as well as Roger Ibbotson's studies on the small cap effect.

The first criticism Dreman has is that not a single small-cap stock actually featured in the study. The New York Stock Exchange is today, and always was, a depository for big, blue-chip stocks. There simply are no small-cap stocks traded on this exchange, and to use a NYSE study to justify the purchase of penny stocks is clearly ludicrous. At best, this was a study of how the smallest 20% of really big stocks perform. What they had measured were huge companies whittled down to small companies by massive falls in stock prices, not small growth stocks as the authors conjectured.

The second criticism is that apart from one or two really great periods, these "small" stocks actually didn't outperform large stocks at all, the long run average was not particularly attractive if you took out the massive outperformance of "small" stocks in 1931 - 1935, 1941 - 1945 and 1976 - 1979. Interesting, small stocks did really well in the Great Depression, a time when we know that small companies actually had the greatest of difficulties and a substantial number of them actually went right out of business.

Dreman examined the results more carefully and found that there was a little more than meets the eye. Although Banz and his colleagues insisted that their findings absolutely ruled out the idea that value investment or contrarianism had any value, their computer in fact did, necessarily, place into the bottom 20% of capitalisation only stocks that had been the most severely sold down. Quite how the researchers were able to conclude as confidently that this was a size effect and not a value effect is unclear, but they were very sure nonetheless. Following the great depression a fair number of these stocks subsequently recovered, with results one might expect from a blue chip company just coming back from the brink of bankruptcy.

Some of the great success stories of this group were helped along by what one might call "extraordinary" events. For example, the Electric Boat Company, later taken over by General Dynamics, benefited from an $8 million settlement in the World Court from the German government, because the Imperial Navy had stolen a number of its submarine patents in 1911 and 1912, before the First World War. Without this, the company would gave gone down for the third time.

Another major criticism lies in the lack of liquidity that these great performing stocks had. At the beginning of 1931 many of these companies were in receivership, though they were still listed on the Exchange. By 1935, 30% of these had delisted. It took only four surviving stocks (Atlas Tack, Spiegel May Stern, Evans Products and United Dyewood) to improve the performance of the small cap group by 43%.

Although the academic's computers didn't have any problem constructing big imaginary portfolios of these stocks, in reality it would not have been quite as simple. Many of these stocks hardly traded at all! These four companies simply could not have been bought! The markets were so thin that volume was often less than a few hundred shares a week. Moreover, the spread between bid and ask prices was huge, averaging 45%. If you decided to buy at market your cost went up 45%. How much could you have nought if you decided to raise your offer 45% more? Not very much, an investor could buy perhaps a few hundred shares, if that, for the majority of these issues. The slightest demand on the buy side would have been enough to send prices sky-rocketing. The average volume for the 139 smallest stocks over this period was 240 shares a day, and the number is only this high because a number of issues traded with a larger volume of a few thousand. Many stocks did not trade at all for many days. The data used by the researchers listed a point half way between highest bid and lowest ask as the price for the day when volume was zero. The researchers simply assumed that all orders could be filled at that median price, ignoring brokerage and transaction costs. The CRSP database they used did not contain volume, bid or ask information before 1962, Dreman had to dig up this information from old news papers.

The Great Depression recovery wasn't the only good time, the really major one was 1941 - 1945. This was a time of major contracts and huge business for small companies, many of which had had a very hard time in the 1930s and so carried forward big losses into the 40s, which were written off against war profits so hardly any of them had to pay taxes in the next decade. Larger companies on the other hand, very profitable in the 1930s, were hampered by excess profits taxes up to 100%. Nevertheless, volume was just as bad as in the previous boom period, with average volume of just 482 shares a day, in dollar terms this being $335,000 for the entire day's trading on the NYSE in this group on any given day. The spread between bid and ask was 17%, so it is obvious that even a slight increase in buying would have had a huge effect on prices, which is exactly what was seen.

Without the grand performance of the Depression years and the Second World War, on average small companies did substantially worse than large companies. In short this strategy could not have been used except by a handful of very patient investors buying a handful of shares here and there, you could not have made money with it on a regular basis, and you certainly could not have run a huge managed fund with it.

James O'Shaughnessey found exactly the same thing when he looked at the smallest stocks, although returns were apparantly towering in good years, liquidity was so bad, and bid/ask spread so wide that it was not an exploitable strategy.

Although the academics argued that the "great" performance of small stocks was evidence that there was a "small-cap" effect and not a "low-PE" effect, Dreman published an article in Forbes on 23 July 1990. The following table appeared in that article. Testing PER for five different market size groups he found that cheap stocks outperformed in all groups of market capitalisation. (1970 - 1996)

Market capitalisation Low PER   2         3         4         High PER  Market   
$100 - 500 Million 18.6% 18.0% 15.7% 14.6% 12.5% 16.0%
$500 M - $1 B 18.8% 17.7% 14.1% 11.0% 10.4% 14.6%
$1-2 Billion 15.9% 15.1% 13.7% 12.4% 10.3% 13.7%
$2-5 Billion 15.3% 14.1% 11.6% 11.8% 10.2% 12.8%
>$5 Billion 14.2% 13.7% 11.1% 11.2% 8.7% 11.9%

Although Dreman does find that small companies do better than really large ones, he cautions that survival rates are lower and transaction costs are higher. Liquidity is always bad for small stocks, and you only need a bigger spread of a couple of percent to completely offset the much higher theoretical profits of small companies. Dreman recommends people stick to larger companies, but nevertheless includes five more rules for small-cap contrarian investment strategy.

Dreman repeated the tests with price-to-cash flow, price-to-book value and dividends and found that the effect worked well in each, though was less successful for high yielding small-cap stocks.

The pages that follow are critical of small stock exchanges, in particular the NASDAQ. Dreman talks about thinly traded issues, and anti-competitive brokers making fortunes by making the market on the stocks with huge spreads between bid and ask (the stock specialist broker being the one that provides most of the volume for market bids can sit there all day making a profit every time someone makes a purchase or sale because the spread gives the specialist their profit. The story is particularly gloomy and might well cause anyone that reads it to for ever swear off NASDAQ stocks. If you want to read it you'll have to find the book, but he does introduce another rule:

As he spends the first three quarters of the book talking about how hard it really is to pick a big winner by forecasting profits I might read that as "don't trade them at all", but you can read into it what you will.

Either way the tone of the rest of the chapter is very negative and it is clear that Dreman is arguing that the only people making profits in small stocks are members of a cartel of brokers who mostly just trade for their own account, usually only dealing with the public if they can make a big profit at the outsider's expense. If you think making money in big stocks is just way too easy and seek a real challenge then you could pit your wits against "the club" of brokers who deal in the small market. Otherwise I guess you might be better off just sticking with big stocks on a larger and much fairer exchange like NYSE.

Next, Dreman turns to small stock funds with glowing recent performance history, that then forms the core of a blitz marketing campaign which brings in lots of new investors, usually right at the top of the fund's success. Alluding to regression to the mean he points out that tremendous recent outperformance rarely goes unpunished, and makes another rule:

The second last part of the book is "psychology", and Dreman gives another account of the crazy prices of Initial Public Offering (IPO) stocks and various bubbles and manias from South Seas through to the Internet boom. As you would expect, it all culminates in the assertion that speculative bubbles always burst and that sentiment is ever a fickle beast. The last rule of contrarian investment is rule 41:

The book finishes with a roundup of Modern Portfolio Theory and how ineffective it is, talks about other anomalies and correlations such as the "Super Bowl Indicator", huge returns for stocks when they announce a split and other weirdness. I summed up most of Dreman's MPT criticisms in another article in this FAQ, Modern Portfolio Theory criticisms.