Contrarian Investment Strategies

by David Dreman

  Jack Grubman was among the masters. Grubman was always negative on AT&T, but Citigroup CEO Sandy Weill “asked” him to take a fresh look at his rating of the company, which he had previously never recommended. “Asked,” in this case, implied funneling millions of extra bonus dollars to Grubman if he went along. Rumor at the time had it that AT&T’s chairman would not let Citigroup participate in a major forthcoming underwriting unless Grubman, who carried enormous weight in the communications sector, upgraded the stock. Grubman, under Weill’s watchful eye, upgraded the stock to a buy near its peak in 1999. Shortly thereafter, Citigroup earned $63 million in underwriting fees when AT&T spun off its wireless unit.

  By 2002, all this had changed. Mom-and-pop investors were being clobbered on WorldCom and other telecom stocks that Grubman insisted on rating highly as markets collapsed. Suspicion rose that he was aiding Citigroup’s investment banking efforts, particularly with the scandal-ridden WorldCom, which generated very large investment banking fees as it continued a major acquisition program. Citigroup was estimated to have made $1 billion in fees generated by Grubman from its investment banking subsidiaries, while shareholders lost $2 trillion in the telecom scandal alone.2

  (Grubman resigned under suspicion in August 2002. He received $30 million in severance pay from Citigroup’s brokerage subsidiary, and by mutual agreement, Citi continued to pay his legal bills.)

  In the ensuing settlement with Eliot Spitzer, the New York State attorney general, Citigroup and ten other banks settled charges of conflicts of interest for $1.4 billion. Four hundred million dollars of that amount came from Citigroup. In a separate settlement with Spitzer and the SEC, Grubman was banned from the securities business for life, as was Henry Blodget. Ironically—or perhaps not—almost all of the surviving brokerage and investment banking firms were bailed out by TARP in 2008 with taxpayer funds that came in part from retail investors including, by then, the struggling “moms and pops” who bought the bad analysts’ recommendations.

  It’s amazing how quickly we forget. By the mid-2000s, our trust in analysts’ forecasting abilities had been completely restored. Needless to say, “the elite,” selected by Institutional Investor from over 15,000 analysts across the country, are sensational stock pickers.

  Aren’t they?

  Financial World measured the analysts’ results some years back.3 The article stated, “It was not an easy task. Most brokerage houses were reluctant to release the batting averages of their superstars.”4 In many cases, the results were obtained from outside sources, such as major clients, and then “only grudgingly.” After months of digging, the magazine came up with the recommendations of twenty superstars. The conclusion: “Heroes were few and far between—during the period in question, the market rose 14.1%. If you had purchased or sold 132 stocks they recommended when they told you to, your gain would have been only 9.3%,” some 34 percent worse than selecting stocks by flipping coins. The magazine added, “Of the hundred and thirty-two stocks the superstars recommended, only 42, or just 1/3, beat the S&P 500.” 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 and start doing the opposite.”5

  In addition to superstars, professional investors rely on earnings forecasting services such as I/B/E/S, Zacks, Investment Research, and First Call, which have online features that give the pros instant revisions of estimates. First Call provides a service that also gives money managers, as well as competing analysts, all analysts’ reports immediately upon their release. Many of the reports deal with forecast changes. More than 1,000 companies are covered.

  The requirement for precise earnings estimates has been increasing in recent years. Missing the analysts’ estimates by pennies can send a stock’s price down sharply. Better-than-expected earnings can send prices soaring. How good, then, are the estimates? We’ve already seen that the performance of the Institutional Investor’s “All-Stars” has been anything but inspiring. But that was for only a one-year period. Nobody’s perfect, after all. Was it just a onetime slip? A fluke? Or do we see a black swan gliding slowly across the waters toward us? This answer will be important to the investment strategies considered in the chapters ahead.

  Forecasting Follies 2: The Long-Term Record

  Updating the work in The New Contrarian Investment Strategy, as well as a number of articles in Forbes and elsewhere,6 I did a study in collaboration with Michael Berry of James Madison University on analysts’ surprises—how much their forecasts missed actual earnings forecasts—that was published in Financial Analysts Journal in May–June 1995.7 It examined brokerage analysts’ quarterly forecasts of earnings as compared with earnings actually reported between 1973 and 1991, which I subsequently updated to 2010. Estimates for the quarter were almost always made in the previous three months, and analysts could revise their estimates up to two weeks before the end of the quarter. In all, 216,576 consensus forecasts8 were used, and we required at least four separate analysts’ estimates before including a stock in the study.9 Larger companies, such as Microsoft or Apple, might have as many as thirty or forty estimates. More than 1,500 New York Stock Exchange, NASDAQ, and AMEX companies were included, and on average, there were about 1,000 companies in the sample. The study was, to my knowledge, the most comprehensive on analyst forecasting to date.10

  How do analysts do at this game, where even slight errors can result in instant wipeouts? A glance at Figure 8-1 tells all. The results are startling: analysts’ estimates were sharply and consistently off the mark, even though they were made less than three months before the end of the quarter for which actual earnings were reported. The average error for the sample was a whopping 40 percent annually. Again, this was no small sample; it included more than 800,000 individual analysts’ estimates.

  Interestingly, these large errors are occurring in the midst of the information revolution. Yet in spite of this fact, estimated errors remain enormously high—much too high to be of any use in determining the real value of most stocks. Yes, the margin of error simply swamps any chance of accurately determining earnings!

  Since many market professionals believe that a forecast error of even plus or minus 3 percent is large enough to trigger a major price movement, what did an average miss of 54 percent in 1991 or an average error of 40 percent over the past thirty-eight years do? When we look at sharp price drops on sizzling stocks after analysts’ misses of only a few percent, it becomes apparent that even small estimate errors can be dangerous to your investment health. Yet that is precisely how the game is played on the Street by analysts, large mutual funds, pension funds, and other institutional investors, and swallowed by average investors.

  You might wonder whether the results are skewed by a few large errors. To check for that kind of skew, we measured earnings surprises four different ways.11 In all cases, errors were high. What about surprises from companies that report small or nominal earnings? A miss of the same amount would obviously result in a higher percentage error for companies reporting very small earnings per share compared to those reporting much higher earnings per share. If, for example, the estimate was $1.00 and the company actually reported 93 cents, the miss would be 7.5 percent. But if the estimate was 10 cents and the company reported only 3 cents, the miss would be 233 percent.*46

  No matter how we analyzed it, the slightest error in the earnings forecast had a disproportionate effect on the fortunes of a company’s stock, irrespective of other measures of the company’s soundness or management performance.

  Forecasting Follies 3: Missing the Consensus Forecast

  I’m sure that many readers know only too well what happens when a stock misses the consensus forecast by much. E*Trade, a former dot-com favorite, tumbled 42 percent in late April 2009, when earnings came in a little more than 4 percent below estimates. Akomai Technologies dropped 19 percent in July 2009, when earnings came in 2 percent below forecast.
Symantec fell 14 percent in July 2009, when reported earnings were 4 percent under estimates. But this was not a one-way street; Amazon.com rose 33 percent in October 2009, when earnings came in 36 percent above forecast. All this occurred in a market that went up over 20 percent that year.

  Earnings surprises, as you would suspect, have had a major impact on stocks over time. During the Internet bubble, 3Com tumbled 45 percent when analysts’ forecasts missed reported earnings by a scant 1 percent in 1997. Sun Microsystems dropped 30 percent on a 6 percent shortfall. On May 30, 1997, Intel announced that its earnings for the June quarter would be sharply higher than in the corresponding 1996 quarter; however, they would be below the analysts’ consensus forecast by 3 percent. This caused a price drop of 26 points, or 16 percent, on the opening, which reverberated first through technology stocks and then through the market as a whole. The result was that the S&P 500 lost $87 billion in minutes. People who relied on the estimates got clobbered—to say the least.

  Again, one must ask if there are exceptions. How good is this dedicated and hardworking group at its members’ vocation? As we just saw, earnings surprises of even a few percentage points can trigger major price reactions. Current investment practice demands estimates that are very close to—or dead-on—reported results. Normally, the higher the valuation of a stock, the more important the precision is. As noted, Street-smart pros normally expect reported earnings to be within a 3 percent range of the consensus estimate—and many demand better.

  Is this doable? Look at Figure 8-2. We used our large database of 216,576 consensus estimates. To utilize a stock, we required at least four analysts’ estimates on it, bringing the total to 866,000 individual estimates at a minimum. Since a large number of stocks have more estimates—Apple, for example, has forty—I estimate that the total number of analysts participating in the estimates was well over one million for the thirty-eight years to the end of 2010.

  Figure 8-2 summarizes our findings. We gave the analysts more leeway than most give themselves, widening the forecasting range from 3 percent to 5 percent to consider an estimate a miss. Even so, the results are devastating to believers in precise forecasts. The distribution of estimates clearly refutes their value to investors. Less than 30 percent of estimates were in the plus or minus 5 percent range of reported earnings that most pros deem absolutely essential. Using the plus or minus 10 percent error band, which many professional investors would argue is far too large, we found that only 47 percent of the consensus forecasts could be called accurate in the quarter. More than 53 percent missed this more lenient minimum range. Worse yet, only 58 percent of the consensus forecasts were within the plus or minus 15 percent band—a level that almost all Wall Streeters would call too high for any quarter.

  Of what value are estimates that seriously miss the mark more than 50 percent or 70 percent of the time? After the horror stories precipitated when forecasts were off even minutely, the answer seems to be—not much. We have seen that estimates carefully prepared only three months in advance, by well-paid and diligent analysts, are notoriously inaccurate. To complicate matters, many stocks sell not on today’s earnings but on expected earnings years into the future. The analysts’ chances of being on the money with their forecasts are not much higher than the chance of winning a major trifecta. Current investment practices seem to demand a precision that is impossible to deliver. Putting your money on these estimates means you are making a bet with the odds heavily against you. A Psychological Guideline is in order here.

  PSYCHOLOGICAL GUIDELINE 11: The probability of achieving precise earnings estimates over time is minuscule. Do not use them as the major reason to buy or sell a stock.

  Forecasting Follies 4: Industry Forecasts

  “But maybe there’s a reason for this,” believers in their forecasting prowess might argue. “Analysts may not be able to hit the broad side of a barn overall, but that’s because there are a lot of volatile industries out there that are impossible to forecast accurately. You can make good estimates where it counts, in stable, growing industries where appreciation is almost inevitable.”

  That’s a plausible statement. We fed it into our computer, which digested the database and spat out the answer a few minutes later. We divided the same analysts’ consensus estimates into twenty-four industry groups*47 and then measured the accuracy for each. The results are shown in Table 8-1. The industry error rates were smaller than those for forecasting individual companies but still almost six times as high as the 5 percent accuracy level most analysts consider too lenient. The average error was 28 percent and the median 26 percent. We also found that over the entire time period, almost 40 percent of all industries had analyst forecast errors larger than 30 percent annually, while almost 10 percent of industries showed surprises larger than 40 percent.

  As the chart shows, analysts’ errors occurred indiscriminately across industries. Errors are almost as high for industries that are supposed to have clearly definable prospects, or “visibility,” years into the future, such as computers or pharmaceuticals, as they are for industries where the outlooks are considered murky, such as autos or materials. This result is so consistent that we should call it a Psychological Guideline.

  PSYCHOLOGICAL GUIDELINE 12: There are no highly predictable industries in which you can count on analysts’ forecasts. Relying on their estimates will lead to trouble.

  The high-visibility, high-growth industries have as many errors as the others. The fact that analysts miss the mark consistently in supposedly high-visibility industries—and give much higher valuations—suggests that those industries are often overpriced.

  Finally, let’s settle one remaining item of business regarding analysts’ forecasts. Are they less accurate in a boom period or a recession, when earnings are presumably more difficult to calculate? Could this difference be a possible reason why their forecasts are not any better?

  Forecasting Follies 5: Analysts’ Forecasts in Booms and Busts

  Our 1973–2010 study covered seven periods of business expansion and six periods of recession. If you think about it, you might expect to see analysts’ forecasts too high in periods of recession, because earnings are dropping sharply, owing to economic factors that are impossible for analysts to predict. Conversely, in periods of expansion, estimates might be too low, as business is actually much better than economists and company managements anticipate. This certainly seems plausible and at first glance provides a partial explanation for the battered analysts’ records. Unfortunately, it just ain’t so, as Table 8-2 shows.

  The table is broken into three columns: All Surprises, which is the average of all positive and negative surprises through the study; Positive Surprises; and Negative Surprises.12 The surprises are shown for each period of business expansion or recession. The bottom row shows the average of all consensus forecasts for periods of both expansion and recession. The average surprise for expansionary periods, 39.2 percent, is little different from the average surprise through the entire period, 39.3 percent, or the average surprise of 43.9 percent in recessionary periods. Moreover, the averages of positive surprises in expansions and recessions are also very similar, 23.3 percent versus 26.0 percent, as are negative surprises, –66.0 percent versus –70.0 percent.

  The statistical analysis demonstrates that economic conditions do not seem to magnify analysts’ errors. They are about as frequent in periods of expansion or recession as they are at other times. What did come out clearly is that analysts are always optimistic; their forecasts are too optimistic in periods of recession, and this optimism doesn’t decrease in periods of economic recovery or in more normal times. This last finding is not new. A number of research papers have been devoted to the subject of analyst optimism, as chapter 7 showed, and, with the exception of one that used far too short a period of time, all have come up with the same conclusion.13 This is an important finding for the investor: if analysts are generally optimistic, there will be a large number of disappointments created not b
y events but by initially seeing the company or industry through rose-colored glasses, as we saw in chapter 7.

  Forecasting Follies 6: What Does It All Mean?

  We have found that large analyst forecast errors have been unacceptably high for a very long time. An error rate of 40 percent is frightful—much too high to be used by money managers or individual investors for selecting stocks. Remember, stock pickers believe they can fine-tune estimates well within a 3 percent range. But the studies show that the average error is more than thirteen times this size. Error rates of 10 to 15 percent make it impossible to distinguish growth stocks (with earnings increasing at a 20 percent clip) from average companies (with earnings growth of 7 percent) or even from also-rans (with earnings expanding at 4 percent). What, then, do error rates approaching 40 percent do?