Michaely and Womack’s study is a comparison between underwriter “buy” recommendations and nonaffiliated “buy” recommendations for IPOs within their first year of trading. Here is a brief synopsis of what they found.
1. Analysts who work for underwriting firms follow stocks that their firms do not bring to market as well as those they do. Consider a single analyst and suppose we were to divide the stocks for which he issues “buy” recommendations into two groups: (1) companies brought to market by his firm, and (2) all others. Are analysts consistent in the way they follow stocks? Which group do you think does better over the long term? Michaely and Womack find that for twelve of the fourteen analysts in their sample, the second group, “all others,” does better. For those suspecting that analysts recommend stocks of firms brought to market by their firms, more favorably, this has the makings of a “smoking gun.”
2. In the first month after the quiet period, underwriter analysts issue 50 percent more “buy” recommendations than nonaffiliated analysts do; and they issue them sooner. Moreover, it seems that underwriters do attempt to prop up the share prices of poor performers. Is there a second smoking gun here? It appears so. Consider the thirty days preceding each buy recommendation on a new IPO stock. During this time, the prices of stocks recommended by non-underwriter analysts have gone up (by 4.1 percent). But the stock prices of firms recommended by underwriter analysts actually have gone down on average (by 2.4 percent).
3. Investors’ reaction to a buy recommendation actually depends on who issues the buy. The excess return associated with underwriters’ “buy” recommendations is 2.8 percent, less than the 4.4 percent associated with nonaffiliated analysts. Yet two years after an IPO, the stocks for which nonaffiliated analysts issued a “buy” recommendation were over 50 percent higher than the stocks for which underwriter analysts issued a recommendation. Taken together, this means that although investors appear to recognize that the recommendations of underwriter analysts need to be taken with a grain of salt, they do not fully discount the effect. If the recommendations of underwriter analysts were superior, because they have access to private information, this should not happen. The underwriter buys should be better predictors of future returns.
4. The IPO stocks that did worst in the 1990–1991 period studied by Michaely and Womack were those that lacked a non-underwriter analyst to corroborate the underwriter analyst recommendation.
5. The cost of subscribing to First Call is substantial. So, it should not be surprising to learn that the market does react when an analyst changes his recommendation. Michaely and Womack report that the immediate average price increase to a buy recommendation is 3.6 percent. The immediate reaction to a sell recommendation is a 10.5 percent drop—and, the reaction to a “removed from buy” recommendation is an even stronger 12.7 percent drop. Now the interesting thing is that although investors respond to the change in recommendation, they actually underreact. In the twelve months following the recommendation change, newly recommended “buys” continue their upward march, whereas new “sells” and “removed from buys” continue their downward decline.5 After twelve months, “removed from buys” are down 15 percent, whereas “sells” are down over 60 percent!
In a sense, we should expect overly optimistic recommendations from the analysts who work for underwriting firms. After all, how easy a time would an underwriter have selling a company’s stock if its own analysts withheld their endorsement? But as Michaely and Womack demonstrate, the key behavioral issue is that investors fail to take full account of the bias. To be sure, it’s not always easy to sort things out.
Not every “buy” recommendation turns out to be a dud. And investors know that these analysts might have access to superior information.6
The Earnings Game: Optimism in Analysts’ Expectations
Now let’s consider the earnings game, the second type of game played by corporate executives and analysts. In his 1997 Wall Street Week with Louis Rukeyser guest appearance, Edward Keon stated: “Analysts as a rule historically have tended to be too optimistic in their forecasts, although that’s changed in the last five years. For the S&P 500 companies, the analysts have been very accurate, and for quarterly forecasts they’ve been too pessimistic.” What is the historical evidence on optimism?
Robert Hansen and Atulya Sarin (1998) have conducted an illuminating study of seasoned equity offerings (SEOs). For the period 1980 through 1991, they examined the behavior of analysts’ earnings forecasts concerning SEOs. Hansen and Sarin note that analysts’ expectations generally tended to be excessively optimistic in this period, by about 2 percent. This corroborates Keon’s characterization and earlier work by Patricia O’Brien.
However, Hansen and Sarin discovered that forecast errors are not uniform across companies. For high earnings-to-price (E/P) stocks, which are where positive earnings surprises tend to get registered, analysts underpredicted earnings by about 3.6 percent. However, for the lowest E/P group, or the high P/E stocks if you prefer, analysts overpredicted earnings by more than 17 percent.
Hansen and Sarin document that companies that issued new stock tended to experience rising earnings during the quarters leading up to the SEO. Figure 18-2, taken from their paper, describes what they found.7 There are three series plotted in this figure. All are expressed as earnings, normalized by price (E/P ratios). Two are forecasts made of annual earnings for the next fiscal year. One is the median forecast by lead analysts, and a second is the median forecasts made by nonlead analysts. The bottom series is the actual earnings figure corresponding to the forecasts, time-shifted back to facilitate comparison with the forecasts. The fact that the forecast values lie above the actual earnings illustrates the extent of optimism in analysts’ forecasts.
Notice that in figure 18-2, earnings (or E/P to be more exact) tended to peak at the time of the SEO. After the SEO, earnings tended to decline. But analysts’ forecasts continued to rise, for at least the
Figure 18-2 Comparison of Earnings Forecasts and Actual Earnings around SEOs
The time paths of earnings forecasts for the next fiscal year made by lead underwriters bringing out seasoned equity offerings, and the forecasts of nonlead analysts, relative to actual earnings. Both forecasts are consistently optimistic, but those of the lead underwriter are especially so. Moreover, actual earnings tend to decline after the offering, despite analysts’ forecasts that earnings will continue to rise. subsequent four quarters. Did analysts overreact to the earnings pattern leading up to the SEO and bet on trends? I suggest that they did.
Were the forecasts of analysts working for the lead underwriter any different than the forecasts of other analysts? Look again at figure 18-2. Notice that both analyst groups were overly optimistic on average, and both continued to project that earnings (E/P) would rise after the SEO.
Hansen and Sarin take pains to point out that forecast optimism is not confined to SEOs. They find that over 70 percent of SEOs are associated with both below-average8 forecasts and below-average earnings per share. The extent of optimism for members of this group tends to be about 4 percent, regardless of whether or not the firm issued new shares through an SEO. Hence the bias is not special to SEOs but instead is more systemic.
The latter finding does not surprise Hansen and Sarin because analysts have reputations to protect. So a critical issue is the way that analysts get rewarded. And how are analysts rewarded? Interestingly, they are rewarded partly based on their ratings in a survey of buy-side institutions. The magazine Institutional Investor runs the survey. But as the October 27, 1997, issue of Fortune magazine points out, some analysts also receive a percentage of any investment banking business they help to bring in: Hence the potential for a conflict of interest.
Masako Darrough and Thomas Russell (1998) suggest that by and large, analysts follow a simple, two-stage heuristic when they develop long-range annual earnings forecasts. In the first stage, analysts forecast that the average company’s earnings will grow at the overly op
timistic rate of17.2 percent a year. In the second stage, analysts gradually adjust their forecasts downward by 21 cents per month on average as the forecast date arrives.
Interestingly, analysts appear not to deviate systematically from this simple rule, even when new information begins to arrive about ten months or so from the end of FY2, the end of the fiscal year for which earnings are being forecast. So in general, analysts underreact to the arrival of new information, an issue that constitutes the theme of chapter 8.
Trying to Induce Pessimism in Analysts’ Earnings Forecasts
During his appearance on Wall Street Week with Louis Rukeyser, Edward Keon mentioned that in recent years, optimism has been disappearing from analysts’ forecasts. In fact, he suggested that quarterly earnings forecasts for S&P 500 stocks have actually become pessimistic. Darrough and Russell (1998) do find that this may be the case for the period immediately prior to the announcement of actual earnings. However, outside of the few weeks leading up to the announcement, they remain skeptical that excessive optimism has disappeared from analysts’ forecasts.
Pessimistic analyst forecasts are an intriguing part of the earnings game. Analysts are highly dependent on the executives of the companies that they follow for their information. This has led to an interesting situation. In a March 31, 1997 Fortune article titled “Learn to Play the Earnings Game (and Wall Street Will Love You),” journalist Justin Fox suggests that companies try to downplay analysts’ forecasts so that their stock prices will jump when actual earnings exceed expectations.
As an extreme example, Fox presents the case of Microsoft. His article begins by pointing out that for 41 of the 42 quarters since Microsoft went public, its earnings beat analysts’ estimates. He describes the reaction to a typical earnings announcement:
The 36 brokerage analysts who make the estimates were, as a group, quite happy about this—the 57 cents per share announced by the software giant was above their consensus of 51 cents, but not so far above as to make them look stupid. …
But then Fox continues by describing the following intriguing encounter:
After a typically grim presentation by CEO Bill Gates and sales chief Steve Ballmer at an analysts’ meeting two years ago, Goldman Sachs analyst Rick Sherlund ran into the pair outside and said, “Congratulations. You guys scared the hell out of people.” Their response? “They gave each other a high five,” Sherlund recalls. But Microsoft, unlike some companies less attuned to the rules of this game, also lets analysts know when they’re too pessimistic. (Fox 1997)
Here is another example, described by Joseph Nocera in the October 27, 1997 issue of Fortune. In January 1997, during the conference call reporting the results of its fourth quarter for 1996, the leading microprocessor firm Intel gave its usual “cautious guidance.” But analysts downplayed the caution and raised their estimates for the first half of 1997. The consensus estimate for the second quarter of 1997 jumped from $0.97 per share to $1.05. The market expected that second quarter results would be announced in June. By May the consensus estimate for the second quarter had risen to $1.08. But then Intel preannounced, warning that its second-quarter results might be disappointing. In response, analysts revised their estimates to reflect this new guidance. In less than three weeks, the consensus estimate dropped from $1.08 to $0.90. Of course, so did the stock price, which on May 30 dropped from $82 to $75.75.
So, what did second-quarter earnings actually come in at? $0.92 a share, above the analysts’ revised estimates. And how did the market respond? On July 16, the price of Intel stock jumped by $7.50, closing at $88.375. The headline in the Wall Street Journal the next day read “Intel’s Profit Exceeds Expectations.” Figure 18-3 depicts the behavior of Intel’s stock price during this period. The two price jumps mentioned above should be quite evident.
Companies have learned that their stock prices go up when actual earnings exceed analysts’ forecasts. To the extent that they provide “guidance,” many firms now seek to lead analysts to be a little pessimistic, thereby enabling companies to produce positive surprises.
Figure 18-3 Share Price of Intel, January 1, 1997–July 31, 1997
Three events during Intel’s earnings game with analysts: (1) In January 1997, analysts downplayed Intel’s cautious guidance and the stock price rose in response to their earnings forecasts. (2) Then in May, Intel preannounced that its second-quarter results would be disappointing, and the stock price immediately plummeted. (3) Analysts revised their second quarter forecasts downward, and low enough so that actual earnings came in above the revised consensus forecast. Actual earnings were announced in July, and Intel’s stock soared on the news that it had beaten analysts’ consensus estimate.
However, analysts’ estimates need not reflect their expectations. The earnings game has led to the concept of “whisper earnings” that presumably reflect what analysts really expect. Therefore, stock prices can decline even when announced earnings exceed the consensus estimate. For example, in mid-July 1997, Microsoft’s announced earnings barely beat analysts’ estimates. Consequently, its stock price fell $8.938, or 6 percent, to close at $140.50. See the July price dip following the June 1997 earnings announcement as depicted in figure 18-4. The figure also depicts the market’s reaction for the typical case in which Microsoft’s announced earnings are more than a tad above the consensus estimate.
Fox, the author of the Fortune earnings game article, points out that because of the conservative way that Microsoft recognizes its shipments, it can engage in earnings manipulation. He quotes Marshall Senk, a Robertson Stephens analyst who follows Microsoft, as saying: “Microsoft does a better job of leveraging accounting—I would almost say it’s a competitive weapon—than anybody else in the industry” (Fox 1997).9
Figure 18-4 Microsoft Corporation Quarterly Earnings Surprise
Microsoft is able to manage its earnings so that they consistently beat analysts’ forecasts. Investors have become accustomed to the pattern, and if the margin is too small, as happened in June 1997, the stock price declines on the news.
Source: I/B/E/S International.
Recognition of Heuristic-Driven Bias?
Are analysts aware of heuristic-driven bias? In 1996, Intel’s gross margins rose approximately 10 percent, and its stock price tripled. But on February 5, 1997, analyst Mark Edelstone of Morgan Stanley Dean Witter downgraded Intel to neutral (hold). Why? He changed his recommendation because he was concerned that his fellow analysts had overreacted to Intel’s 1996 success. Edelstone thought that his fellow analysts were underreacting to a series of potential problems that might arise from a major product transition and competition from archrival Advanced Micro Devices.10 If you recall the discussion about Intel in the previous section, events proved him correct.11
Thresholds in Earnings Manipulation
Francois Degeorge, Jayendu Patel, and Richard Zeckhauser (1997) argue that the earnings game has led companies to employ threshold decision making in respect to the way they manage earnings. These authors identify a hierarchy consisting of three specific thresholds:
1. “Red ink,” meaning zero earnings;
2. The previous period’s earnings; and
3. Analysts’ consensus earnings forecasts.
Degeorge, Patel, and Zeckhauser present evidence that suggests managers manipulate earnings in an attempt to surpass all three thresholds if possible. However, suppose that managers anticipate that actual earnings will not surpass analysts’ forecasts with respect to the coming announcement. Then, instead of missing earnings by a penny, better to bank those earnings until a later date, and miss by more than a penny. How much more? Well, if a firm can meet the previous period’s earnings, then those earnings should be the target. Otherwise, don’t miss by a penny, and move down toward zero.
In chapter 3, I discussed the tendency for people to evaluate outcomes relative to a benchmark. As Degeorge, Patel, and Zeckhauser point out, the use of benchmarks leads naturally to a consideration of earnings thresholds.
It could be that managers’ thought processes are threshold based; or that managers cater to the threshold thinking of investors, or both. Degeorge, Patel, and Zeckhauser also discuss nonpsychological reasons for a threshold structure of earnings management. These mostly have to do with the cost-saving features of discrete indicators.
Do managers actually behave in accordance with these three thresholds? The evidence is quite compelling. Degeorge, Patel, and Zeckhauser show that earnings announcements are not symmetric around thresholds. Rather, they are highly skewed. For example, the number of announcements where earnings are a little less than zero is tiny, whereas the number of announcements where earnings are a little above zero is quite large.
Summary
Corporate executives play two major games with analysts, an earnings game and a recommendation game. But investors are also party to these games. Heuristic-driven bias, and to some extent frame dependence, affect the behavior of the players in these games. The result is market inefficiency.
What type of bias and errors do analysts commit? During a guest appearance on Wall $treet Week with Louis Rukeyser, I/B/E/S’s Edward
Keon made a series of comments that bear on both games. With respect to the earnings game, he characterized analysts’ forecasts as “deeply flawed.” He also cautioned that recommendations be taken with a grain of salt when the recommender also wears an investment-banking hat.
The evidence supports both of Keon’s contentions. David Dreman (1995) claims that, on average, analysts’ earnings recommendations are off by 10 percent. In addition, analysts’ forecasts tend to be overly optimistic, except for the period just prior to earnings announcements, when they turn pessimistic. But corporate executives have also learned to play the earnings game. They do so by attempting to induce pessimistic forecasts that they can beat. If necessary, they preannounce; and they manipulate earnings using a threshold-based rule.
Beyond Greed and Fear Page 33