The second surprise: Notice that Stephens’s annual earnings forecast is four times the most recent quarterly number. Indeed, the consensus estimate for the second quarter was for 40¢. How did that quarter actually turn out? Were the two analysts right? It turns out there was a second surprise. Instead of being off by 100 percent, as they were previously, analysts were only off by 20.9 percent this time.
On April 16, the date of the announcement, Stephens responded by changing its recommendation from a “buy” to a “strong buy.” The market’s reaction was interesting: Prices declined on the news. Actually, prices had begun to decline a couple of days before earnings were announced, and continued to fall slightly during the three days after the announcement. But the dip was minor and temporary.
The value of Plexus shares rose steadily over the next months. The June 2 issue of Barron’s reported that money manager James O’Shaughnessy had included Plexus in his “growth picks.”2 Other analysts, such as Nesbitt Burns and A.G. Edwards, initiated coverage of Plexus. How did their earnings forecasts fare against the actuals?
Subsequent surprises: The answer is one surprise after another. The First Call mean estimate for the quarter ending June 30 was for 51¢ a share. But on July 17, earnings were reported at 56¢, an 11.8 percent surprise. For the quarter ending September 30, the surprise was even larger, 15.4 percent. But in the final quarter of 1998, although there was a surprise, something different happened. The surprise turned out to be negative.
Graphical recap: Figure 8-1 from I/B/E/S displays the time path of analysts’ earnings predictions, actual earnings, and stock price.
Figure 8-1 makes it apparent that earnings surprises had been consistently positive for five consecutive quarters. Notice the path of the stock price. It climbed steadily throughout this period until October. However, during October prices began to fall. And then they fell quite dramatically on December 18, from a postsplit price of $25 a share to $14, the largest percentage decline that day for any stock traded on any U.S. exchange. The reason? A preannouncement of lower earnings from Plexus’s management.
Just prior to December 18, First Call reported that its survey of the five analysts following Plexus featured a consensus earnings estimate of 30 cents a share for the company’s first quarter, meaning the quarter ending December 31, 1997. However, Plexus preannounced that because of slower than anticipated sales growth, earnings for the quarter would instead be between 21 cents and 24 cents.
Analysts’ reactions to the negative surprise (preannouncement): How did the analysts who follow Plexus react? The excerpts below, from the December 19 issue of the Milwaukee Journal Sentinel, capture their views.3 The first quotation indicates surprise at the extent of the price drop.
Figure 8-1 Plexus Corporation Quarterly Earnings Surprise
Plexus had five consecutive quarters of positive earnings surprises. Although the sixth quarter appears to show a positive surprise, the surprise was actually negative, stemming from a preannouncement. The figure shows the adjusted analysts’ forecasts, not the original forecasts. The price path exhibits momentum along the positive earnings surprise trajectory, followed by a sharp decline after the disappointing preannouncement. The general finding features three consecutive quarters of earnings surprises, accompanied by abnormal returns in the same direction, followed by an earnings surprise in the opposite direction with an accompanying price reversal. Source: I/B/E/S International.
“The degree of punishment here is really high,” said Matthew J. Desmond, equity analyst at Red Chip Review, a publication in Portland, Ore., that researches and writes about small-company stocks.4
The next set of remarks describes the nature of the drop-off in sales:
“This is the first real earnings disappointment this company has had,” said Scott Alaniz, securities analyst at Stephens Inc. in Little Rock, Ark. Slightly more than half of the shortfall occurred because Motorola Inc. moved some of the work Plexus had been doing in-house. … Motorola represents about 6 percent of Plexus’ sales, down from about 12 percent before the work was pulled, said Paul S. Shain, research director at Robert W. Baird & Co.
Motorola represents the biggest risk of this type among Plexus’ customers, Desmond said. The company’s other big accounts, such as IBM and General Electric Corp., don’t have the capacity to bring the work that Plexus does in-house.
Next, we learn how analysts revised their earnings forecasts, in reaction to the preannouncement.
Shain lowered his rating and earnings estimates for Plexus after the announcement Thursday. Now, he said, he expects Plexus to earn $1.10 per share for fiscal 1998, down from his previous estimate of $1.37. He also lowered his estimate for 1999 to $1.50 a share, from $1.75. “The near-term revenue disruption will likely continue into the second quarter, creating difficult comparisons for the remainder of the year,” Shain said.
Finally, we return to the issue of the market’s reaction, or should I say overreaction, which appears to be at odds with the analysts’ views:
But analysts thought there were other reasons for the pounding Plexus’ stock received.
“It’s just a real touchy market. Geez, you sneeze and it turns into a nightmare,” Desmond said. Investors generally are nervous about the technology sector, and they’ve been spooked by other earnings surprises that initially appeared to be isolated.
“People are finding out they’re linked together,” Alaniz said. Plexus was more alluring than many of its peers because it didn’t have a lot of exposure on the computing or communications equipment side, and because it is more diversified with medical, industrial and other customers, Alaniz said. “That’s another reason why this is a real shock,” he said.
First Call reports that after the preannouncement, analysts revised their estimates to 22¢. In the end, earnings for the last quarter of calendar 1997 came in at 23¢ a share.
General Lessons: Most of the general lessons to be learned from Plexus’s experience can be seen in figure 8-1. There are two things to look at: (1) the pattern of earnings, and (2) the simultaneous price pattern reflecting momentum and subsequent reversal. Starting with a large positive earnings surprise in January 1997, what we see is that analysts underreacted to actual earnings. The revised earnings forecast of $1.60 per share for fiscal year 1997, made in January 1997, turned out too low by almost 20 percent. Analysts underreacted in the sense that they continued to be positively surprised for the next two quarters. However, then they seem to have overreacted, predicting 30¢ a share, when the reality was 23¢. Hence, the earnings surprise was negative four quarters later.
Prices climbed steadily during the first three quarters of calendar year 1997, giving rise to momentum in the wake of a positive earnings surprise. Prices declined during the fourth quarter.
Given the comments by analysts quoted above, it seems reasonable to ask whether the news was sufficiently dire as to justify this price response. In other words, did the market overreact to the preannouncement? Recall Matthew Desmond’s earlier remark, that investors were concerned about misinterpreting a permanent earnings decline as temporary. Comparing earnings per share for the two quarters ending in December, one in 1996 and the other in 1997, we find that their actual earnings were about the same. But so were the stock prices. So, did investors revise their expectations about Plexus’s growth all the way back to the rates of January 1997? Certainly the analysts did: their revised forecasts for fiscal 1998 were actually less than $1.60.
Post-Earnings-Announcement Drift
The case of Plexus is highly representative of a phenomenon called post-earnings-announcement drift. The late Victor Bernard (1993) describes the general phenomenon as one where “analysts’ forecasts tend to underreact to earnings information” and “market prices underreact to analysts’ forecasts.” (p. 322)
Scholars Richard Mendenhall (1991) and Jeffrey Abarbanell and Victor Bernard (1992) present evidence that analysts underreact to earnings information when they revise their forecasts. The
ir work, pertaining to Value quarterly earnings forecasts, finds that analysts do not sufficiently revise their forecasts. One positive surprise tends to be followed by another, and then by yet another.
In his survey of the literature on post-earnings-announcement drift, Bernard (1993) discusses evidence concerning market prices underreacting to analyst forecasts. He explains that the postannouncement stock prices for firms that have reported “good news” tend to drift up, whereas the prices for firms that have reported “bad news” tend to drift down. The identification of “good news” firms and “bad news” firms is accomplished using a criterion with the acronym SUE, which stands for standardized unexpected earnings. SUE is computed by taking the quarterly earnings surprise and scaling by the standard deviation of earnings surprises for that quarter.
For example, on April 16, 1997, Plexus announced that its earnings for the quarter ending March 31, 1997, were 24¢ a share. Since the analysts’ consensus estimate had been for 20¢, the earnings surprise of 4¢ constituted a 20.9 percent surprise. I/B/E/S reported that this surprise had a SUE value of8.44. This means that the typical surprise, which is a single standard deviation, was about 2.5 percent. The size of the April 16 Plexus surprise was 8.44 times as big as its typical surprise for earnings announcements.
In academic studies, the earnings forecast used to compute SUE is based on a simple forecasting rule. Take the quarter for which the forecast is being made, say, the first quarter. Now look at the history of all previously available first-quarter earnings for a given firm and compute the average growth rate for those earnings, on a year-over-year basis. This gives the average growth rate for first-quarter earnings from one fiscal year to another. To form a forecast for next year’s first-quarter earnings, take this year’s actual first-quarter earnings and multiply by one plus the average growth rate.
Suppose we were to arrange all companies into ten groups (deciles) according to their SUE values. Imagine forming two stock portfolios: one based on the highest of the most recent SUE values, and the other based on the lowest. The highest SUE values are, of course, the ones associated with extreme good earnings news, whereas the lowest SUE values are associated with extreme bad earnings news.
Once formed, how did those portfolios perform relative to the stocks of companies of comparable size (measured in terms of market value of equity)? Over the sixty days that followed the earnings announcement, the stocks of the highest SUE firms returned 2 percent more than their comparably sized peer group. The stocks of the lowest SUE firms returned 2 percent less than their peer group. Hence, a trading strategy of shorting the lowest SUE portfolio and using the proceeds to take long positions in the highest SUE portfolio would have earned 4.2 percent more. Call this 4.2 percent an abnormal return. Moreover, this trading strategy was even more successful for small and medium-size firms, which earned about 10 percent more than their peer group. I discussed these results in chapter 4; see figure 4-2.
Victor Bernard and Jacob Thomas (1989, 1990) and James Wiggins (1991) find that the returns to the trading strategy just described have a marked pattern. First, the abnormal returns are not earned all at once. Instead, they are spread over time. The largest effect occurs just after the announcement associated with an earnings surprise. Historically, it has been 1.32 percent, relative to the peer group. Next comes a delayed effect associated with the subsequent announcement one quarter later, and yet another delayed effect associated with the announcement two quarters later. On average, these have been 0.70 percent and 0.04 percent, respectively.
Think about what this pattern implies. It pays to hold stocks that have experienced recent large positive earnings surprises, because the market does not fully adjust to the good news. Instead, the market adjusts over the three quarters that follow an announcement.
There are two additional interesting features about the SUE-based trading strategy. First, it would not have worked if attempted for yet one additional quarter: Such an attempt would have produced a negative abnormal return of –0.66 percent. Second, the abnormal returns all tend to be concentrated in the first three trading days that follow the earnings announcements. They are not spread out evenly throughout the quarter.
Clearly, Plexus was a high SUE firm in 1997; its stock was a high performer for the first three quarters of the year, then turned down for the last quarter. In this respect, Plexus’s experience is representative of the general pattern described by Bernard (1993). The one area where it does not fit the pattern concerns the three trading days following announcements. The large price moves in Plexus’s stock were hardly concentrated during the three trading days following the announcements. In fact, those days were more volume events than price events.
Why does the trading strategy based on SUE lead to positive abnormal returns? One possible reason might be risk. If this explanation were valid, then the stocks of high SUE companies would have to be riskier than the stocks of low SUE companies. If true, the abnormal returns are nothing more than fair compensation for holding long positions in the stocks of high SUE companies.
But this explanation is difficult to justify. To begin with, Bernard (1993) reports that the SUE-based trading strategy generated positive abnormal returns for twenty-two consecutive years, from 1965 through 1986. (If that’s risk, let’s have more of it!) Moreover, the general phenomenon is concentrated in the three days following announcements. For the risk-based explanation to be valid, the stocks of high SUE companies would have to experience an increase in risk only on those days.
Trading on a Behavioral Bias: Fuller and Thaler Asset Management
Another possibility is that the abnormal returns from the SUE trading strategy are hypothetical, and disappear if transaction costs are factored in. But the returns are not hypothetical.
Consider the experience of RJF Asset Management, now Fuller and Thaler Asset Management (F&T), one of the few firms that promotes itself as being a leader in behavioral finance. Fuller and Thaler manages a mutual fund called Behavioral Growth, which consists of the stocks of high SUE companies. It held the stock of Plexus Corp. throughout 1997.5 The trading strategy underlying Behavioral Growth is the same as that of F&T’s Small/Mid Cap Growth.
F&T started its Small/Mid Cap Growth Fund in January 1992. On an annual basis, the F&T Small/Mid Cap portfolio returned 28.4 percent, gross of fees, to its investors over its first seven years. In contrast, the Russell 2500 Growth Index returned 11.7 percent, and the S&P 500 index returned 19.5 percent. Figure 8-2 shows the cumulative difference between investing a dollar with F&T’s Small/Mid Cap Growth in
Figure 8-2 Cumulative Performance of Fuller & Thaler Small/Mid Cap Growth Fund, January 1992–January 1999
Cumulative returns to a Fuller & Thaler trading strategy, designed to exploit post-earnings-announcement drift. Results are shown relative to the Russell 2500 Growth Index, which serves as the fund’s benchmark. During the seven-year period, the Fuller & Thaler fund returned 28.4% per year gross of fees, while the Russell 2500 Growth Index returned 11.7%. Fees amount to 1.9% per year.
January 1992 and investing that dollar in the Russell 2500 Growth index. Over seven years, a dollar invested in F&T would have been worth $5.69 in December 1998, whereas a dollar invested in the Russell 2500 would have been worth only $2.13.
Behavioral Growth’s long-run performance is consistent with the point made by Bernard (1993) that a SUE-based strategy consistently outperforms its peer group. F&T’s Small/Mid Cap Growth portfolio outperformed the Russell 2500 index in six out of the seven years, 1992 through 1998 inclusive.
The data in figure 8-2 are based on actual trades net of commissions, but before management fees. These data also suggest that the superior performance of a SUE-based strategy is not due to risk. The monthly return standard deviation of the F&T Small/Mid Cap portfolio is almost identical to that of the Russell 2500, approximately 5.2 percent.
Which Behavioral Biases?
F&T’s president, Russell Fuller, explained his trading philo
sophy in the January 5, 1998, issue of Barron’s.6 Fuller’s view is that both analysts and investors are slow to recognize the information associated with a major earnings surprise. Instead, they overconfidently remain anchored to their prior view of the company’s prospects. That is, they underweight evidence that disconfirms their prior views and overweight confirming evidence.7 Consequently, both analysts and investors interpret a permanent change as if it were temporary; thus, the price is slow to adjust. F&T would buy a stock soon after a major positive earnings surprise and hold it until the positive earnings surprises diminished or disappeared.
Overconfidence and anchoring definitely appear to be part of the explanation underlying post-earnings-announcement drift. But there are other biases that may also contribute. Bernard (1993) mentions the work of Paul Andreassen (1990) and Andreassen and Steven Kraus (1990) on the importance of salience for financial predictions. See also Eli Amir and Yoav Ganzach (1998). These authors document the tendency for investors to place little weight on changes to a series, unless the recent changes are salient and attributable to a stable, underlying cause. The case of Plexus serves as an apt illustration. In January 1997, the time of the 100 percent earnings surprise, the major financial press contained virtually no stories describing what had caused the surprises.
The same pattern continued throughout most of the year. Most of the stories simply reported the magnitude of the surprise but little more. However in December, when Plexus preannounced that earnings would be lower than expected, the loss of business from Motorola was highly salient. As can be seen from the earlier press excerpts, coverage went beyond merely reporting the magnitude of the earnings revision.
Beyond Greed and Fear Page 15