Unless you invest in index funds, you ultimately hope to beat the market. However, realistically, few individuals can beat the market on a risk-adjusted basis over the long haul. Benjamin Graham understood this difficulty and put forth that for an investor to beat the market he first must have a sound theory, then have opinions and projections that are not only correct, but also different from those of the market.
How often have you heard of a stock price falling on the announcement of increased earnings? Or, in these times, an Internet stock leaping up after reporting negative earnings per share? In these instances, actual earnings did not turn out as expected. Expectations play a key role in determining if a stock’s price “gains” or “loses” when actual earnings are reported.The market is forward looking. Stock prices are established through expectations and adjust as these expectations change or are proven wrong. Earnings per share EPS estimates involve the interaction of many company, industry, and economic forces. They embody an analyst’s opinion of factors such as sales growth, product demand, competitive industry environment, profit margins and cost controls. Earnings are a key variable used to value stocks and slight changes in expectations for future earnings or the earnings growth rate can strongly impact the stock price.
We have witnessed a strong increase in the number of services that track and analyze expected earnings per share estimates. Services such as I/B/E/S, First Call, Standard & Poor’s, and Zacks provide consensus earnings estimates by tracking the estimates of thousands of investment analysts. Tracking these expectations and their changes is an important and rewarding strategy for stock investors.
Services tracking earnings estimates normally track the expectation of approximately 4,000 “sell-side” analysts tracking around 6,000 stocks. They are referred to as sell-side analysts because they are employed by brokerage firms providing stock research for their customers. These brokerage firms do not primarily buy stocks for their own portfolio, but rather “sell” stocks to their brokerage customers. In contrast, buy-side analysts perform research for firms actively managing portfolios such as mutual funds and pension funds.
Sell-side analysts have two primary functions: to provide EPS estimates and to provide buy/hold/sell company investment recommendations. The brokerage firm uses the research developed by their analysts to attract and retain brokerage customers and create investment banking opportunities.
Stocks with earnings estimates may have from one to 40 analysts tracking and analyzing them. The consensus estimate refers to the average EPS estimate.
Stock Investor Professional (AAII’s fundamental stock screening and data program) uses I/B/E/S as its source for consensus earnings estimates. Of the 9,276 stocks tracked by the program, 5,173 have current fiscal year earnings estimates. However, nearly 20% of these firms are tracked by only one analyst—not really a broad consensus.
Figure 1 provides a screen shot of the EPS estimates for Hewlett-Packard Company HWP. The consensus EPS estimate for Hewlett-Packard’s current quarter ending April 2000 is $0.82. For the current fiscal year ending October 2000, analysts are expecting earnings to come in at $3.42 per share. Analysts also normally provide a three-year to five-year annual growth rate estimate, which for Hewlett-Packard is currently 15.0%.
The high and low estimates provide a feel for the range in analyst opinion for a given company. For the current fiscal year, the range covers a low of $3.30 to a high of $3.60. The long-term growth rate estimates range from 12.0% to 18.0%. The standard deviation also provides a statistical measure of the dispersion of the earnings estimates. Hewlett-Packard’s standard deviation is $0.06 for the current fiscal year.
The number of estimates provides a feel for the depth of coverage for a company. Twenty-eight analysts are providing an estimate for the current fiscal year. This is a relatively large number of analysts. On average, four analysts are providing current year estimates for the firms with estimates.
In using earnings estimates, the first rule to keep in mind is that the current price usually reflects the consensus earnings estimate. There is nothing to be gained by screening simply for high levels of expected earnings growth. Studies show that over the long run, stocks with high expected earnings growth tend to underperform stocks with low growth rates and low expectations. It is difficult to meet and exceed high expectations over an extended period of time. In pursuing a screening strategy using earnings estimates, it is best to focus on surprise and revision.
It is common to see price declines for stocks that report earnings increases from the previous reporting period because in many cases, while the actual earnings represent an increase, the increase is not as great as the market had expected. Earnings surprises occur when a company reports actual earnings that differ from consensus earnings estimates.
Listed companies are required to file quarterly reports with the Securities and Exchange Commission SEC within 45 days of the fiscal quarter end. Most companies announce earnings within one month after the end of the quarter.
The fiscal periods for most companies coincide with the calendar quarters. Institutional investors and analysts work at a frenzied pace for about three weeks starting mid-month in January, April, July, and October as major companies report their earnings from the previous quarter. During the earnings reporting season, financial newspapers and Web sites provide daily reports on earnings announcements. Firms with significant earnings surprises are often highlighted.
Positive earnings surprises occur when actual reported earnings are significantly above the forecasted earnings per share. Negative earnings surprises occur when reported earnings per share are significantly below the earnings expectations. The stock prices of firms with significant positive earnings surprises show above-average performance, while those with negative surprises have below-average performance.
There are a number of ways to measure the significance of an earnings surprise. Figure 1 provides some basic information about Hewlett-Packard’s most recent quarterly earnings report (above the earnings estimate grid).
Hewlett-Packard reported its January quarter-ending earnings on February 16, 2000. At the time of report, the consensus estimate for the quarter was $0.78, but reported earnings turned out to be $0.80—a two-cent positive surprise. To gauge the significance of this surprise you must compare it against some base figure. One popular measure calculates the percentage change from the consensus estimate to the actual reported EPS (i.e., reported EPS minus estimated EPS, divided by the estimated EPS.). For Hewlett-Packard, the percent surprise is 2.6%.
Surprise % = EPS (actual) – EPS (estimate) / EPS (estimate)
Surprise % = $0.80 – $0.78 / $0.78
Surprise % = 0.026 or 2.6%
The other popular technique used to measure earnings surprise is the standardized unexpected earnings SUE score. An earnings surprise is considered more significant if it is far outside the statistical range of estimates expected at the time of the announcement. The SUE score makes use of the standard deviation of expected earnings. The standard deviation of Hewlett-Packard’s quarterly estimate was $0.05 on February 16, 2000. The SUE score for Hewlett-Packard’s surprise was relatively low at 0.4.
SUE Score = EPS (actual) – EPS (estimate) / Standard deviation
SUE Score = $0.80 – $0.78 / $0.05
SUE Score = 0.4
Hewlett-Packard’s positive surprise of two cents was well within the five-cent standard deviation and actually a bit of a disappointment. Hewlett-Packard announced earnings after the market close on February 16 and was weak the next day, closing down three points on a day the market was up slightly (see Figure 2).
In contrast, on January 13 Intel announced quarterly earnings of $0.69, six cents above the $0.63 consensus estimate. This represents a more significant 9.5% surprise and a much more meaningful 3.0 SUE score (see Figure 3). As the price chart in Figure 4 reveals, this significant positive surprise sparked a strong price increase, with the stock gaining 12 points on the news.
Changes in stock price resulting from an earnings surprise can be felt immediately, and the surprise also has a long-term effect. While it may be difficult for individuals to buy on the initial surprise event, studies indicate that the effect can persist for as long as a year after the announcement. This means that it does not make sense to buy a stock after the initial price decline of a negative earnings surprise. There is a good chance that the stock will continue to underperform the market for some time. Also, it may not be too late to buy into an attractive company after a better than expected earnings report is released.
Not surprisingly, large firms tend to adjust to surprises faster than small firms do. Larger firms are tracked by more analysts and portfolio managers, who tend to act quickly.
Firms with a significant quarterly earnings surprise also often have earnings surprises in subsequent quarters. Many times when a firm has a surprise, it often is a sign that other similar surprises will follow. This is sometimes referred to as the “cockroach effect”—like cockroaches, you rarely see just one earnings surprise.
Since both positive and negative earnings surprises have lingering long-term effects, a rewarding investment strategy is one that avoids stocks you believe will have negative earnings surprises or those that have had negative earnings surprises. Similarly, selecting positive earnings surprise stocks before and even after the earnings come in may be profitable. Even a strategy of simply selling after negative earnings surprises and buying after positive earnings surprises probably has some merit.
In the April 2000 issue of the AAII Journal, Wayne Thorp develops a screen following the investment philosophy of Richard Driehus in which earnings surprises play a significant role.
Estimate Observations From a Noted Contrarian
In the September/October 1999 issue of Computerized Investing, we profiled the investment philosophy of David Dreman, chairman of Dreman Value Management. Dreman studied the psychological underpinnings of the overall stock market and its impact upon valuation levels. In his book, “Contrarian Investment Strategies: The Next Generation” (Simon and Schuster; 800/223-2336; www.simonsays.com), Dreman devotes a considerable amount of space to examining earnings estimates and surprises and their effect upon stock price.
In a study of quarterly earnings estimates, Dreman found an average error of 44% annually. Only 29% of the estimates were plus or minus 5% of the actual earnings announcement, 47% were within 10% plus or minus of the actual earnings figure, and 58% were within a 15% plus or minus band. That means that there is a high probability that you will experience an earnings surprise for a company that you own.
Dreman found that errors in analysts’ forecasts were high across a wide range of industry groups. Companies in industries with high visibility and high growth were seen to have as many errors as other firms had.
Dreman’s studies found that analyst forecasts tend to be overly optimistic. About a quarter of the surprises were positive surprises, while over three-quarters turned out to be negative surprises. Unexpectedly, the tendency toward over-optimism did not vary significantly during periods of expansion or recession. Analysts tend to be overconfident of their forecasting abilities—typically making last-minute adjustments to their forecasts, yet still missing the mark.
Historically, it has been observed that analysts have a tendency to start the year high with their annual EPS forecast and then reduce the estimate as the year goes forward. This upward bias was especially high when there was an investment banking relationship between an analyst’s firm and the company under analysis. A study of analyst revisions of stocks within the S&P 500 found that earnings estimates tend to be revised downward 12.9% during the course of the fiscal year, with about a 6.3% adjustment in the first six months, and a 19.2% adjustment in the second half of the year.
More recent studies indicate that this positive bias may have been reduced and possibly reversed, at least for quarterly forecasts. Companies understand the potential impact on stock price of a positive earnings surprise and occasionally try to guide analysts into positive quarterly surprises. With greater emphasis being placed on tracking quarterly earnings announcements, the current consensus is that annual forecasts tend to be more objective, while quarterly forecasts come under greater pressure for low balling.
As the earnings reporting season gets into high gear ,whisper earnings start to circulate throughout the investment community. Whisper earnings, or whispers, are considered the actual earnings expectations of the analysts tracking the company—free of any company influence. These “actual” estimates are passed or whispered from analyst to analyst and to the analyst’s best customers, but not actually posted to estimate tracking services.
Whisper numbers have garnered much attention recently because of the apparent weight they carry with institutional investors. There have been some incidences where a company has met or even exceeded its consensus earnings but failed to meet the whisper numbers, and its stock price has plummeted.
A 1998 study performed by Susan Watts of Purdue University, Mark Bagnoli of the University of Michigan, and Messod Daniel Benish of Indiana University looked at both traditional consensus estimates and whisper numbers. While they found that consensus earnings tend to understate actual earnings, they also found that whisper numbers tend to overestimate earnings by almost five cents per share. The researchers feel that whisper numbers “undo” the conservative bias that reported consensus earnings have. They further indicate that whisper numbers appear to better reflect the market’s expectations of earnings.
Traditionally, whisper numbers were circulated orally among top analysts and their best customers. More recently Web sites have been established that try to gather and distribute whisper numbers. How should an investor view whispers? Unless you have direct contact with the analysts, whisper numbers tend to come from unknown and unnamed sources. They may come from company insiders, perceptive investors, or those who are attempting to manipulate a stock’s price. Sites that collect whisper estimates rarely disclose their sources.
Do not simply take the estimates at face value. The whisper number reported on Whispernumber.com for Hewlett-Packard’s first quarter was $0.73, further from the reported $0.80 EPS than the
I/B/E/S consensus of $0.78 or First Call consensus of $0.77 (see Figure 5).
Table 3 at the end of the article provides links to additional sources for whisper numbers.
If we accept that earnings surprises are inevitable, is there a way for investors to take advantage of them? Dreman finds that the best way to take advantage of the high rate of analyst forecast error is to simply invest in out-of-favor stocks.
There are times when the market seems to accept that no price is too high for a company with great growth prospects. These high expectations lead to high multiples of price to earnings, book value, cash flow, and dividends. The high multiples can act as a double-edged sword. Investors are willing to pay a higher stock price for a current level of earnings because they expect earnings to grow more strongly in the future. If expectations of future prospects increase, investors will accept a higher multiple of price to current earnings. A high growth stock that exceeds expectations can get a large price boost because of an expansion of the multiple and the higher current earnings figure. However, if disappointing earnings are released, a tremendous price decline can ensue if investors view the negative surprise as a reason to adjust their expected growth rates and corresponding multiples downward.
In his studies of earnings surprises, Dreman found that stocks with low price-earnings ratios reacted more strongly to positive earnings surprises than did high price-earnings stocks. As it turns out, a positive earnings surprise for a stock with high expectations (as measured by factors such as high price-earnings ratios) is not truly a surprise. It is a reinforcing event that does not change perceptions about a stock.
Positive earnings surprises for out-of-favor stocks, however, are perceived by the market as significant events. Dreman terms them “event triggers” because they initiate a perceptual change among investors.
The impact of a negative earnings surprise is reversed. Out-of-favor stocks barely flinch, while highly favored stocks generally have significant declines. With out-of-favor stocks, negative surprises are reinforcing events that do not lead to reevaluations. However, negative earnings surprises are event triggers for highly valued stocks that typically lead to a downward revaluation of the firm’s prospects.
As revealed in Figure 6, the addition of an upward earnings revision element to the Dreman contrarian screen improved its performance over the last few years. Additional information on the revised Dreman screen (Dreman with a Twist) can be found on the AAII Web site within the Stock Screens area (www.aaii.com/stkscrns).
As the reporting period approaches, estimates normally converge toward the consensus. It seems that many analysts like the comfort of being near the consensus and are too conservative when issuing outlying forecasts.
Most investors also dismiss outlying forecasts too quickly and underestimate the probability of perceived unlikely events.
While free earnings estimate services do not reveal the analysts behind the estimates, the premium services often do. When researching a firm with a wide discrepancy in earnings forecasts, it would be prudent to try to determine why there is a high degree of uncertainty in the earnings forecast.
Changes in estimates reflect changes in expectations of future performance. Revisions are often precursors to earnings surprises. A flurry of revisions near the reporting period can indicate that analysts missed the mark and are scrambling to improve their estimates.
Revisions to earnings estimates lead to price adjustments similar to earnings surprises. When earnings estimates are revised significantly upward—5% or more—stocks tend to show above-average performance. Stock prices of firms with downward revisions show below-average performance after the adjustment.
Changes in estimates reflect changes in expectations of future performance. Perhaps the economic outlook is better than previously expected, or maybe a new product is selling better than anticipated. Changes in individual analyst estimates away from the consensus average are especially noteworthy and more meaningful than revisions toward the mean.
Companies like to report positive earnings surprises, so it is not surprising that many companies try to “manage” the estimates slightly downward to create a positive surprise. Interestingly, estimates for the fiscal year do not tend to show the same positive surprise bias.
Figure 1 shows the current earnings estimates for Hewlett-Packard as well as a history of recent revisions. For the week ending February 25, 2000, Hewlett-Packard’s consensus annual earnings forecast for its current fiscal year was $3.42 per share. This represents a one-cent increase from the previous week’s average estimate of $3.41.
Twenty-eight analysts are providing current fiscal year estimates for Hewlett-Packard. Over the course of the month, 19 analysts raised their fiscal year forecasts, while two analysts lowered their estimates. Overall, the consensus forecast increased by 5 cents from $3.37 one month ago to $3.42.
The significance of this change can be measured through a percent change of the forecast, or the net number of analyst changes.
The percentage change is calculated in a fashion similar to percent surprise. A positive revision percentage is considered bullish, while a negative figure may be a bearish signal.
Revision % = Estimate (current) – EstimateEstimate (prior)
Revision % = $3.42 – $3.37 / $3.37
Revision % = 0.015 or 1.5%
The net number of analyst revisions is often referred to as “diffusion.” Diffusion is the comparison of the number of up and down EPS revisions to the total number of analysts following the stock. A bullish signal would be provided by a larger number of analysts increasing estimates than the number decreasing estimates. Hewlett-Packard has strong positive diffusion of 60.7%.
Diffusion = # Rev Up – # Rev Down / Total Estimates
Diffusion = 19 – 2 / 28
Diffusion = 0.607 or 60.7%
We have created two screens: one that looks for upward revisions in annual earnings estimates and another that screens for companies with downward revisions. AAII’s Stock Investor, which contains consensus earnings estimates from I/B/E/S, was used to perform our screen.
The first screen filters out those firms with less than four estimates for the current fiscal year. This filter helps to ensure that revisions actually reflect a change in general consensus, not just a change by one or two analysts. However, requiring a stock to have at least four analysts reporting earnings estimates will knock out most of the very small-capitalization stocks.
For the upward revision screen, the next filter requires that the firm have an upward change over the course of the last month in its consensus estimates for the current and next fiscal year. We are also screening to make sure no analysts lowered estimates for the current or next fiscal year during the past month.
The screen for companies with downward EPS estimate revisions simply reverses the elements of the upward revision screen. It looks for firms with decreases in annual fiscal year estimates for the current and next year, coupled with no upward revisions.
We performed a basic test of these screens and the results are presented in Figure 7. To calculate the performance, we assume that the filter is run at the beginning of each month and an equal amount is invested in each stock passing the filter. This rebalancing and screening is performed monthly. With a price change of 101% since September 1997, the portfolio of stocks with upward revisions strongly outperformed the downward revisions portfolio, which lost 9% over the same time period. This confirms the somewhat lasting effect of the revision. Even though the greatest impact on stock price was felt in the month of the revision, the strategy still worked when comparing the current consensus estimate to the previous month’s estimate.
The portfolios that these screens produced, however, would be difficult for most individual investors to duplicate. The upward revision portfolio averaged 146 stocks with only 16% of companies surviving the rebalancing from one month to the next. The downward revision screen averaged 195 stocks with 21% of the stocks making it from one month to the next. The positive revision screen had as many as 307 companies pass the filter and as few as 52 passing companies for a given month. The downward revisions screen ranged from 124 to 316 passing companies.
To help create a more manageable portfolio, we retested the upward revision strategy requiring at least a 5% revision increase over the last month. This new requirement dramatically increased the performance over the test period while generally creating more manageable portfolios. The price change performance increased from 101% to 314% through the end of February 2000. The average number of portfolio holdings decreased from 146 to 40. Not surprising, turnover increased so that only 7% of the stocks made it from one month to the next. Figure 8 plots the cumulative performance of the 5% revisions screen.
Table 1 shows the 25 firms with the largest upward and largest downward monthly revisions for the current fiscal year. Whenever your filter involves the percentage change of a variable, there is a risk that firms with very small base numbers will dominate. A change from one cent to nine cents is a 900% change. Therefore, when working with percentage changes, it is helpful to use an additional screen to confirm the significance of the change.
The number of estimates for each firm is shown to help gauge the interest in the firm and the meaningfulness of the overall estimates. The larger the firm, the greater the number of analysts that will track it.
The number of revisions upward and downward indicates how many analysts have revised their estimates in the last month. When compared to the number of analysts making estimates, this is a confirmation of the significance of the percentage change in estimates. You can put more faith in a revision if a large percentage of the analysts tracking a firm have revised their estimates.
Examining the range of estimates provides an indication of the consensus within the group of estimates. A wide range of estimates would point to great disagreement among analysts, indicating greater uncertainty and greater chance for an earnings surprise. The price move can be more dramatic, however, if an earnings surprise occurs for a firm with a very tight range of earnings estimates. Firms with tight ranges of estimates will normally have higher SUE scores when there is a surprise.
Earnings estimates are an important element for investors to keep in mind as they analyze and select stocks. They are a numerical view of expectations, and changing expectations drive stock prices. Even though earnings surprises and revisions are widely used by investors, they continue to highlight stocks worthy of further analysis. Table 2 summarizes the main points to keep in mind when dealing with consensus earnings estimates.
The following is a short description of the screens and terms used in Table 1.
Current Estimate ($): The consensus of analysts’ estimates for earnings per share for the company’s current fiscal year as tracked by I/B/E/S. The average estimate is reported and it reflects the earnings expectations built into the stock price.
Highest Estimate ($): The highest EPS estimate for the current company fiscal year. When compared with the lowest estimate, provides an indication of the level of consensus among analysts tracking the stock. The wider the range, the greater the divergence in opinion and the greater the chance for an earnings surprise.
Lowest Estimate ($): The lowest EPS estimate for the current company fiscal year. When compared with the highest estimate, provides an indication of the level of consensus among analysts tracking the stock. The wider the range, the greater the divergence in opinion and the greater the chance for an earnings surprise.
Std. Dev. ($): The standard deviation of the earnings estimates tracked by I/B/E/S. Helps to indicate the level of consensus among analysts tracking a given stock. Low standard deviation relative to the average estimate indicates a strong level of agreement for a given estimate figure. Reported earnings surprises outside the standard deviation range are considered more meaningful than surprises within the standard deviation range.
Month Ago Estimate ($): The average EPS estimate for the current fiscal year from the previous month. When compared against the current estimate provides a feel for the recent change in the EPS expectation for a given company.
Monthly Change (%): The percentage change in the consensus earnings estimate over a one-month period. Prices of firms with positive revisions tend to perform better than firms with downward revisions.
No. of Est. (#): Number of analysts providing earnings per share estimates for the current fiscal year. Indicates how widely a firm is followed. Widely followed firms tend to react more quickly to estimate revisions.
Monthly Revisions—Revs. Up (#): The number of analysts who revised their earnings per share estimate for the stock upward during the month. When compared to the total number of analysts issuing estimates provides an indication of strength of an earnings revision.
Monthly Revisions—Revs. Down (#): The number of analysts who revised their earnings per share estimate for the stock downward during the month. When compared to the total number of analysts issuing estimates provides an indication of strength of an earnings revision.