Lies, Damned Lies, And Earnings Estimates

by: Backroom Analyst


Occasionally I encounter a nice debate about investment philosophy that forces me to do a little research.

Earnings estimates or historical growth rates - which should an investor use?

While both are useful, one of these methodologies is far superior.


Recently, I was taken to task about a stock analysis I performed for Biogen (NASDAQ: BIIB). While I think the research was solid, and it is based on mechanical methodologies that have outperformed the market over the last 13 years, I was told by some in the comments section that I had it all wrong. Some of the commentators (all were anonymous) felt that I should have quantified future analysis and predicted earnings in the coming years. Apparently, looking for financially strong companies with proven business histories and future markets was simply not enough. I had to predict future earnings down to the penny, and not doing so was proof that my methods were invalid.

Putting on my adult britches, I decided to see if predicting future earnings was worth the effort. This is what I discovered.

Earnings Predictions Are Really Bad

Research shows Wall Street professionals are really bad at predicting future earnings. David Dreman and Michael Berry discovered this in 1995. In their work, they stated, "On average, 56 percent of the estimates measured as a percent of actual fall outside a plus or minus 10 percent range, a level that many Wall Street professionals consider minimally actable, approximately 45 percent fall outside a plus or minus 15 percent range. These results indicate that on average, large earnings surprises are the rule rather than the exception." (Dreman and Berry 1995).

According David Dreman's website (Dreman Value Management, LLC 2011) this situation has not improved much since 1995, with earnings misses consistently occurring at high levels. The number of surprises is common, and because they tend to be too optimistic. Stotz and Lu (2015) tell us, "…financial analysts have been optimistically wrong by 25% with their 12-monthly earnings forecasts."

What is sad is that regulations since 2000 have not improved the situation. Espahbodi, et al. (2015) tell us that after 2000 era regulations, there was "…an improvement in these forecast properties in the short run," but, "however, forecast accuracy significantly declined and dispersion significantly increased." They conclude that the regulatory environment, "did not collectively improve the information environment despite the reduction in analyst conflicts of interest. The problem seems to be largely due to the quality of financial reports."

It was easy to confirm these conclusions. I ran my own data (Portfolio123 n.d.), and found that quarterly earnings predictions are inaccurate by at least 10% an average of 53.7%. Annual earnings projections were wrong by at least 10% an average of 27.6% of the time. David Dreman and Michael Berry have been right for a very long time, and pretty much anyone can see it for themselves, as several have. What is really sad is that the situation is not improving over time. If anything, it has been getting worse, albeit not significantly.

(Portfolio123 n.d.)

(Portfolio123 n.d.)

Analysts Do Not Care

The research shows analysts really do not care whether they are accurate or not. What does the typical analyst do when they miss so badly? Apparently not much. According to Bashears and Milkman (2011), "We find that when a stock analyst makes an extreme earnings forecast that a future earnings announcement reveals was incorrect, she sticks stubbornly to her opinion rather than updating as much as analysts whose estimates were closer to consensus." I will assume male analysts are just as bad.

The question then becomes, why are analysts so wrong so often? First, analysts really do not care about the timeliness or the accuracy of their predictions. Brown, et al. (2015) surveyed several analysts to see what the most important aspect of their jobs was. While they ranked industry knowledge, and professional standing the highest, the analysts surveyed ranked the accuracy of their analyses the lowest of the nine factors that impacted their jobs. They simply don't care whether they are right or wrong.

Dreman and Berry (1995) did postulate a theory as to why this might be happening; it's herd mentality. Analysts are too afraid to make bold predictions. Why? They might lose their jobs if they are wrong. If they follow the herd, and the entire herd is wrong, then they were no more wrong then everyone else. Dreman and Berry write, "An estimate that is far off the consensus might pose career dangers, whereas an estimate near the group may provide the analyst with a much higher degree of safety, regardless of how inaccurate it may prove to be."

So how useful are earnings predictions? They're not. Dreman and Berry (1995) add that errors in earnings projections are, "too high for investors to rely on consensus forecasts as a major determinant of stock valuation." Predicting earnings is a fool's game run by fools. They are more likely to be wrong, than to be right, and when they are wrong (which is often) they refuse to change their methods. One should look for other determinants to find valuable companies.

Impact on Portfolio Performance

Maybe one can use consensus earnings projections to find companies that earn money over time. The data shows that one should look elsewhere. Here is where I performed a little experiment.

When I look for companies, I want to find investments that will double the initial investment price over a five year period. That is a 14.9% annual price return. If I cannot find that, I may as well invest in an index fund and call it the day. To calculate a five-year target price, I use a standard compounding growth formula to determine the future earnings. I then multiply those future earnings by the lesser of the current P/E ratio or the five-year average P/E ratio. Russell 3000 companies that pass that test are kept in my portfolio for one year, at which time I start the process all over again. All companies are equally weighted. I ran a rolling backtest at four week intervals to see how this strategy would have done. I assume a 0.25% slippage rate. 215 one-year trials since January 1999 were performed, and this is how this strategy would have performed using different growth rates:

January 1, 1999-June 10, 2016

Russell 3000

All Russell 3000 Companies

Consensus Long-Term Growth

Historical 10-Year Growth

Historical 5-Year Growth

Historical 3-Year Growth

Historical 1-Year Growth

Rolling Return







Standard Deviation














Up Markets







Down Markets







Percent Time Beating the Index






(Portfolio123 n.d.)

The evidence does show that consensus long-term growth projections do perform better (11.74% v. 11.14%) than the equal-weighted market index. This difference is significant. The evidence, however, also shows that ignoring projected growth rates, and using long-term historical growth rates, especially the 5- and 3-year periods, performs even better (12.27% and 12.29% respectively). These are not insignificant differences, and they produce portfolios with lower volatility, and less downside deviation. It would appear that one should focus more on historical returns, which reveal a company has a real business, than to focus on speculative models. Score one for mechanical investing.


Too often analysts get it wrong; more than 50% quarter to quarter. Using analysts' predictions to determine whether a company will grow in the future might improve one's portfolio performance, but using historical growth statistics works better. Please understand, I am not advocating that one not look at the overall financial health of the company, nor am I suggesting that one should ignore the future prospects of the business. I perform my due diligence. I am suggesting, though, that there is room for this type of mechanical investing if one is so inclined, and it is a process that will yield superior results.

I had an interesting exchange with John Reese recently, and he reminded me that Peter Lynch preferred using historical figures vs. analyst estimates also. In fact, if one studies Reese's guru analyses, they will notice that earnings estimates do not enter any of the models. Read his book (Reese and Forehand 2009), and one will see that his guru models consistently outperform the overall market. The American Association of Individual Investors (2016) also has several guru screens that consistently outperform the markets. Most do not use estimated earnings, and those that do use them because the original predictions were so wrong.

I suggest using a regression model to determine a future earnings estimate. I use the 5-, 3-, and 1-year growth rates for earnings, revenues, and margins. Then, I determine regressed exponential models to calculate future earnings. Yes, I do use Excel, I don't apologize for that. From there, I am able to determine a target price. The advantages of this approach are twofold. First, I only analyze companies that have a persistent history of positive earnings every year. Second, it takes in account declining growth rates, and shrinking margins. It is a strategy that has beaten the market over the last 13 years, and I have the data to prove it.

Passing Companies

These are companies that should double over the next five years whether one uses consensus Wall Street estimates, 5-year historical growth rates, or 3-year historical growth rates:



Crown Castle International


Chipotle Mexican Grill


Cognizant Technology Solutions


Delta Air Lines






General Motors


Mohawk Industries


Martin Marietta Materials


Norwegian Cruise Line Holdings


NXP Semiconductors


Priceline Group




Skyworks Solutions


Tesoro Corporation


Under Armour


Ulta Salon Cosmetics & Fragrance




Happy Investing!


American Association of Individual Investors. Stock Screens: Performance History. 2016. (accessed 2016).

Beshears, John, and Katherine Milkman. "Do Sell-Side Stock Analysts Exhibit Escalation of Commitment?" Economic Behavior & Organization 77, no. 3 (March 2011): 304-317.

Brown, Lawrence D., Andrew C. Call, Michael B. Clement, and Nathan Y. Sharp. "Inside the "Black Box" of Sell-Side Financial Analysts." Journal of Accounting Research 53, no. 1 (March 2015): 1.

Dreman Value Management, LLC. Behavioral Finance. 2011. (accessed June 2016).

Dreman, David N., and Michael A. Berry. "Analyst Forecasting Errors and Their Implications for Security Analysis." Financial Analysts Journal, May-June 1995: 30-41.

Espahbodi, Hassan, Pouran Espahbodi, and Reza Espahbodi. "Did Analyst Forecast Accuracy and Dispersion Improve after 2002 Following the Increase in Regulation." Financial Analysts Journal 71, no. 5 (September/October 2015).

Portfolio123. Portfolio123. (accessed June 2016).

Reese, John P., and Jack M. Forehand. The Guru Investor: How to Beat the Market Using History's Best Investment Strategies. Hoboken, New Jersey: John Wiley & Sons, Inc., 2009.

Stotz, Andrew, and Wei Lu. "Financial analysts were only wrong by 25%." Working Paper, 2015.

Disclosure: I am/we are long CMG, CTSH, PCLN, ULTA.

I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.

About this article:

Author payment: Seeking Alpha pays for exclusive articles. Payment calculations are based on a combination of coverage area, popularity and quality.
Want to share your opinion on this article? Add a comment.
Disagree with this article? .
To report a factual error in this article, click here