Thursday, July 16, 2009

Why You Should Pay No Attention to Consensus Estimates

While perusing the news headlines on Seeking Alpha this morning I learned that Fairchild Semiconductor (FCS) reported second quarter earnings per share of -$.03, soundly beating the consensus estimate of -$.11. Hooray! The company only lost $.03 a share in the worst economic downturn in 80 years. This is surely something to celebrate, as evidenced by the fact that the stock was indicated up in pre market trading. Now, I have no idea whether the stock will finish or up down on the day. I also don’t know if only losing $.03 a share in this economic environment is actually indicative of shrewd management and effective cost containment. It could be that FCS did an admirable job in Q2 navigating its way through a very tough market. I am certainly not an expert on the semiconductor industry and am not picking on FCS. I’m just using Fairchild’s Q2 results as an example of what I see as the absurdity of putting any weight behind how a company performed relative to consensus forecasts.

Being a contrarian investor, whenever I hear that the “consensus” believes something I tend to shudder. In fact, I often take universal agreement as a contrary indicator, meaning that I tend to feel that whatever it is that the herd is predicting will either not play out or, even worse, the exact opposite will result. During earnings season investors are inundated with consensus forecasts for earnings and the market is unfortunately prone to celebrating companies who trump analysts’ estimates but to scorning those who fall short of expectations. Given that market participants often react very strongly to announcements of quarterly revenue, margin, and earnings per share figures, doesn’t it make sense to ask simple questions regarding the historical accuracy of analyst estimates? In other words, if the market as a whole puts so much emphasis on the prognostications of analysts, isn’t it important to know if the numbers that companies are being compared against are even meaningful? What good is using a benchmark to evaluate the quality of a company if that benchmark is completely skewed?

In his 1998 book, “Contrarian Investment Strategies: The Next Generation” fund manager David Dreman tries to answer the very questions posed above. (After a brief search I could not locate any more recent data but I suspect that nothing has changed in the last 11 years, especially given the fact that increased market volatility has only made forecasting more difficult.) For any of you who rely on consensus estimates and forward earnings multiples in order to make investment decisions, the results of Dreman’s study with Michael Berry of James Madison University should be eye opening.

The study compared brokerage analyst quarterly earnings forecasts between 1973 and 1996. In total, the duo surveyed over 94,000 estimates and in order to make sure that they were not relying on too few analysts, required that at least four analysts’ estimates were included for each company. Furthermore, more than 1500 companies were included to ensure that companies with a wide spectrum of market caps were included. According to Dreman:

“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 44% annually.” (Emphasis mine)

Dreman goes on to point out that despite the explosion of readily available information on companies since the 1970s, analysts’ ability to forecast did not get any better. In fact, in the last eight years of the study the average error was an incredible 50%. The researchers also attempted to eliminate the effect on the averages of large errors:

“[We] eliminated all companies that reported earnings in the + or – 10-cent range to prevent large percentage errors from this group distorting the study…Even using this ultra conservative method, the average forecast error was still 23%...more than quadruple the size that market pros believe could set off a major price reaction.”

As referenced above, Berry and Dreman use 5% as a benchmark for the maximum amount a company could beat or miss by without setting off a major market reaction. In other words, they assumed that for a company expected to earn $1 per share, the company could earn anywhere between $.95 and $1.05 (approximately) without making the stock price move dramatically. In addition, the team used the +/- 5% range as a barometer of analyst accuracy. For an analyst to be deemed prescient, he or she would have to be able to predict the quarterly earnings within a 5% range. Whether you think that measuring stick was fair or not, the percentage of analysts who met that criteria was pathetically low. Specifically, only 29.4% of analysts were able to forecast within +/- 5%. Even more troubling, only 46.8% were able to do so within a +/- 10% range and only 58.3% fell within a +/- 15% window. Taking that company that is supposed to earn $1 again, the evidence shows that, on average, less than 60% of analysts would be accurate within $.15 cents in either direction. When you consider how much a stock can move if the company makes $.85 or $1.15 as opposed to $1, the average analyst error is not insignificant.

What about during recessions? 47.4% average error. Expansions? 44.9% average error. But what about if the study is broken down by industry? Using 62 separate subgroupings between 1973 and 1996 the average error was 50%. For example, the worst errors occurred in metals and mining (71%) and oil (73%). Although, I guess investors can take some solace in the fact that errors for tobacco (4%), food (25%) and communication (25%) stocks were far less than the average.

Now, I am well aware that accurately and consistently predicting quarterly earnings is very difficult and borders on impossible. I certainly feel for my sell-side brethren who are constantly under pressure to see into a very murky future and make recommendations based on a cloudy crystal ball. In general, I see that there are five major factors (there could be more) that contribute to this proven inability to forecast in a precise manner (in no particular order):

  1. Reliance on management guidance: Investors can’t forget that management teams have all the incentive in the world to under-promise and over-deliver. No CEO wants to explain why his company was unable to meet or beat the guidance it provided for analysts and investors. Accordingly, executives would rather err on the side of caution by low balling earnings estimates or providing a very wide range, neither of which help analysts when it comes to precision. Also, no matter how much an analyst knows about a company there are always numerous moving pieces that only insiders can understand and assess. The result of this is that many times there are things going on with customers or production that analysts cannot possibly be expected to account for.
  2. Earnings manipulation: Although you almost never hear of the SEC investigating claims of earnings smoothing or manipulation, it would be very naïve to believe that this does not happen on a regular basis. Whether it is companies like GE consistently beating estimates by a single penny, manufacturing firms recognizing revenue aggressively at the end of the quarter or struggling companies taking large write downs all in one quarter (commonly known as a big bath), there is plenty of evidence that earnings manipulation is a relatively common occurrence. While some of it is probably benign, practices like those listed above make it much harder for analysts to anticipate what quarterly earnings are going to be.
  3. Use of complicated models: Many sell-side analysts maintain robust Excel spreadsheets with earnings and margins extrapolated over five, ten or even twenty year periods. But, if analysts are not able to reliably predict earnings on a quarterly basis, how in the world could they do so years into the future? A model is only as good as the underlying data and the more skewed that data is the less useful the results of the model.
  4. Dynamic economy: We are in unprecedented times when it comes to the volatility within the global economy. Company management teams have very little visibility regarding demand tomorrow, let alone next quarter or next year. Accordingly, analysts are currently completely in the dark and my guess is that their forecasts will reflect that. However, even in more benign times supply and demand are fluctuating constantly and a company’s prospects can literally turn on a dime. Unless an analyst has unusual access to industry participants and is literally in the trenches with the companies he or she covers, it is very unlikely that dynamic conditions will be captured in earnings forecasts.
  5. Quarterly noise: The truth is that three months is not a very long time. For companies that plan their budgeting for a full year, small deviations from the budget or unexpected timing of revenues can have outsized effects on quarterly earnings. In addition, things like capital expenditures, hirings, firings, and the timing of share buybacks or debt issuance can skew individual quarterly results dramatically. This is why prudent managers don’t worry about managing quarterly numbers and focus on the long term. But, what this means is that analysts will always be unable to account for what is nothing more than noise in their quarterly predictions.

In conclusion, the above analysis is not really an indictment of the analyst community. I understand very well why it is so difficult to accurately forecast quarterly earnings and do not hold it against the analysts for having an extremely poor historical track record. Believe me; I could not do any better myself. However, this is explicitly a criticism of a market and system that relies on analysts’ estimates to make investment decisions even though the data unequivocally shows that the forecasts are wholly unreliably and inaccurate. The idea that anyone believes that a stock should go up or down based on the deviation from earnings expectations is absolutely ludicrous. Instead, investors would be wise to come up with their own assessment of a company’s value that does not rely on forward earnings or a company's ability to print whatever number it thinks the Street wants to see.

(Picture courtesy of