One of the early applications of computers in economics in the 1950s was to analyze economic time series. Business cycle theorists believed tracing the evolution of several economic variables over time would clarify and predict the progress of the economy through boom and bust periods. A natural candidate for analysis was the behavior of the stock market prices over time.
Assuming stock prices reflect the prospects of the firm, recurring patters of peaks and troughs in economic performance ought to show up in those prices. In 1953 Maurice Kendall, a British statistician, presented a controversial paper to the Royal Statistical Society on the behavior of stock and commodity prices.
Kendall had expected to find regular price cycles, but to his surprise they did not seem to exist. Each series appeared to be “a ‘wandering’ one, almost as if once a week the Demon of Chance drew a random number… and added it to the current price to determine the next week’s price.
In other words, the prices of stocks and commodities seemed to follow a random walk. When Maurice Kendall suggested that stock prices follow a random walk, he was implying that the price changes are independent of one another just as the gains and losses in the coin-tossing games are independent. The figure below illustrates this. Each dot shows the change in the price of Microsoft stock on successive days. The circled dot in the southeast quadrant refers to a pair of days in which a 1 percent increase was followed by a 1 percent decrease.
If there was a systematic tendency for increased to be followed by decreases, there would be many dots in the southeast quadrant and few in the northeast quadrant. It is obvious from a glance that there is very little pattern in these price movements. In fact, the coefficient of correlation between each days’ price change and the next is only +0. 022, i. e. , there is only a negligible tendency for price rises to be followed by further price rises.
More generally, one could say that any publicly available information that might be used to predict stock performance, including information on the macro economy, the firm’s industry, and its operations, plans, and management, should already be reflected in stock prices. As soon as there is any information indicating a stock is underpriced and offers a profit opportunity, investors flock to buy the stock and immediately bid up its price to a fair level, where again only ordinary rates of return can be expected.
These “ordinary rates” are simply rates of return commensurate with the risk of the stock. But if prices are bid immediately to fair levels, given all available information, it must be that prices increase or decrease only in response to new information. New information, by definition, must be unpredictable; if it could be predicted, then that prediction would be part of today’s information. Thus, stock prices that change in response to new information must move unpredictably. Randomness in price changes should not be confused with irrationality in the level of prices.
If prices are determined rationally, then only new information will cause them to change. Therefore, a random walk would be the natural consequence of prices that always reflect all current knowledge. Indeed, if stock price movements were predictable, that would be damning evidence of stock market inefficiency, because the ability to predict prices would indicate that all available information was not already impounded in stock prices.
The first use of the term “efficient market” appeared in a 1965 paper by Eugene Fama4, who defined it as: a market where there are large numbers of rational, profit-maximizes actively competing, with each trying to predict future market values of individual securities, and where important current information is almost freely available to all participants…”
- Weak-form EMH : asserts that stock prices already reflect all information that can be derived by examining market trading data such as the history of past prices, trading volume, or short interest. This version of the hypothesis implies that trend analysis is fruitless. Past stock price data are publicly available and virtually costless to obtain. The weak-form hypothesis holds that if such data ever conveyed reliable signals about future performance, all investors would have learned long since to exploit the signals. Ultimately, the signals lose their value as they become widely known, because a buy signal, for instance, would result in an immediate price increase.
- Semistrong-form EMH : states that all publicly available information regarding the prospects of a firm must be already reflected in the stock price. Such information includes, in addition to past prices, fundamental data on the firm’s product line, quality of management, balance sheet composition, patents held, earnings forecasts, accounting practices, and so forth. Again, if any investor had access to such information from publicly available resources, one would expect it to be reflected in stock prices.
- Strong-form EMH : states that stock prices reflect all information relevant to the firm, even including information available only to company insiders, this version of the hypothesis is quite extreme. Few would argue with the proposition that corporate officers have access to pertinent information long enough before public release to enable them to profit from trading on that information.
Indeed, much of the activity of the Securities and Exchange Commission (SEC) is directed toward preventing insiders profiting by exploiting their privileged situation. Rule 10b-5 of the Security Exchange Act of 1934 limits trading by corporate officers, directors, and substantial owners, requiring them to report trades to the SEC. Anyone trading on information supplied by insiders is considered in violation of the law. Implications of the EMH Technical Analysis. Technical analysis is essentially the search for recurring and predictable patters in stock prices.
Although technicians recognize the value of information that has to do with future economic prospects of the firm, they believe such information is not necessary for a successful trading strategy. Whatever the fundamental reason for a change in stock price, if the stock price responds slowly enough, the analyst will be able to identify a trend that can be exploited during the adjustment period. Technical analysis assumes a sluggish response to stock prices to fundamental supply and demand factors. This assumption is diametrically opposed to the notion of an efficient market.
The EMH predicts that technical analysis is without merit. The past history of prices and trading volume is publicly available at minimal cost. Therefore, any information that was ever available from analyzing past prices have already been reflected in stock prices. As investors compete to exploit their common knowledge, they necessarily drive stock prices to levels where expected rates of return are commensurate with risk. At those levels,stocks are neither bad nor good buys. They are just fairly priced, meaning one should not expect above-than –normal (or abnormal) returns.
Fundamental analysis uses earnings and dividend prospects of the firm, expectations of future interest rates, and risk evaluation of the firm to determine proper stock prices. Ultimately, it represents an attempt to determine the present discounted value of all the payments a stockholder will receive from each share of stock. If the value exceeds the stock price, the fundamental analyst would recommend purchasing the stock. Once again, the EMH predicts that most fundamental analysis will add little value.
If analysts rely on publicly available earnings and industry information, one analyst’s evaluation of the firm’s prospects is not likely to be significantly more accurate than another’s. There are many well-informed, well-financed firms conducting such research, and in the face of such competition, it will be difficult to uncover data not also available to other analysts. Only analysts with a unique insight will be rewarded. Active versus Passive Portfolio Management. Casual efforts to pick stocks are not likely to pay off.
Competition among investors ensures that any easily implemented stock evaluation technique will be used widely enough so that any insights derived from it will be reflected in stock prices. Only serious analyses and uncommon techniques are likely to generate the differential insight necessary to generate trading profits. Proponents of the EMH believe active management is largely wasted effort and unlikely to justify the expenses incurred. Hence, they advocate a passive investment strategy that makes no attempt to outsmart the market.
A passive strategy aims only at establishing a well-diversified portfolio of securities without attempting to find under or overvalued stocks. Passive management usually is characterized by a buy-and-hold strategy. Because the EMH indicates stock prices are at fair levels, given all available information, it makes no sense to buy and sell securities frequently, as transactions generate large trading costs without increasing expected performance. If the market is efficient, why not throw darts at The Wall Street Journal instead of trying to choose a stock portfolio rationally?
It is tempting to draw this sort of conclusion from the notion that security prices are fairly set, but it’s a far too simple one. There is a role for rational portfolio management, even in perfectly efficient markets. A basic principle in portfolio selection is diversification. Even if all stocks are priced fairly, each still poses firm-specific risk that can be eliminated through diversifications. Therefore, rational security selection, even in an efficient market, calls for the selection of a carefully diversified portfolio. Moreover, that portfolio should provide the systematic (market) risk level the investor wants.
Even in an efficient market, investors must choose the risk-return profiles they deem appropriate. Rational investment policy also requires that investors take tax considerations into account in security choice. If you are in a high tax bracket, you generally will not want the same securities that low-bracket investors find favorable. Also, when forming a portfolio, the particular risk profile of the investor should be considered. For example, a General Motors executive whose annual bonus depends on GM’s profits should not generally invest additional amounts in auto stocks.
To the extent that his or her compensation already depends on the well-being of GM, the executive is overinvested in GM now and should not exacerbate the lack of diversification. Investors of varying ages also might warrant different portfolio policies with regard to risk exposure. For instance, older investors who are essentially living off savings might avoid long term bonds, whose market values fluctuate dramatically with changes in interest rates. Because these investors rely on accumulated savings, they require conservation of principal.
In contrast, younger investors might be more inclined toward long-term-inflation-indexed bonds. The steady flow of income over long periods that is locked in with these bonds can be more important than the preservation of capital to those with long life expectancies. In short, there is a role for portfolio management even in an efficient market. Investors’ optimal positions will vary according to factors such as age, tax bracket, risk aversion, and employment. The role of the portfolio manager in an efficient market is to tailor the portfolio to these needs, rather than attempt to beat the market.
Are Markets Efficient? Not surprisingly, the EMH is not enthusiastically hailed by professional portfolio managers. It implies that a great deal of the activity of portfolio managers – the search for undervalued securities – is at best wasted effort and possibly harmful to clients because it costs money and leads to imperfectly diversified portfolios. Consequently, the EMH has never been widely accepted on the Wall Street, and debate continues today on the degree to which security analysis can improve investment performance. There are mainly three factors that together imply the debate probably will never be settled:
An investment manager overseeing a $5 billion portfolio who can improve performance by only one-tenth of 1% per year will increase investment earnings by . 001 x $5 billion = $5 million annually. This manager would clearly be worth his or her salary. Yet we, as observers, probably cannot statistically measure her contribution. A one-tenth of 1% contribution would be swamped by the yearly volatility of the market. Remember that the annual standard deviation of the well-diversified S&P 500 index has been approximately 20% per year.
Against these fluctuations, a small increase in performance would be hard to detect. Nevertheless, $5 million remains an extremely valuable improvement in performance. All might agree that stock prices are very close to their fair values, and that only managers of large portfolios can earn enough trading profits to make the exploitation of minor mispricing worth the effort. According to this view, the actions of intelligent investment managers are the driving force behind the constant evolution of market prices to fair levels. The selection bias issue: Suppose you discover an investment scheme that could really make money.
We have two choices: either publish your technique in the Wall Street Journal to have your fifteen minutes of fame or keep your technique secret and use it to earn millions of dollars. Most investors would choose the latter option, which presents us with a conundrum. Only investors who find that an investment scheme cannot generate abnormal returns will be willing to report their findings to the whole world. Hence, opponents of the efficient market’s view of the world always can use evidence that various techniques do not provide investment rewards as proof that the techniques that do work simply are not being reported to the public.
This is a problem in selection bias; the outcomes we are able to observe have been preselected in favor of failed attempts. Therefore, we cannot fairly evaluate the true ability of portfolio managers to generate winning stock market strategies. The lucky event method: In virtually any month, it seems we read an article in the Wall Street Journal about some investor or investment company with a fantastic investment performance over the recent past. Surely the superior records of such investors disprove the efficient markets hypothesis. This conclusion is far from obvious, however.
As an analogy to the “contest” among portfolio managers, consider a contest to flip the most heads out of 50 trials using a fair coin. The expected outcome for any person is 50% heads and 50% tails. If 10,000 people, however, compete in this contest, it would not be surprising if at least one or two contestant flipped more that 75% heads. In fact, elementary statistics tells us that the expected number of contestants flipping 75% or more heads would be two. It would be silly, though, to crown these people the head-flipping champions of the world.
They are simply the contestants who happened to get lucky on the day of the event. The analogy to the efficient markets is clear. Under the hypothesis that any stock is fairly priced given all available information, any bet on a stock is simply a coin toss. There is equal likelihood of winning or losing the bet. Yet, if many investors using a variety of schemes make fair bets, statistically speaking, some of those investors will be lucky and win a great majority of bets. For every big winner, there may be many big losers, but we never hear of these managers.
The winners, though, turn up in the Wall Street Journal as the latest stock market gurus; then they can make fortune publishing market newsletters. The point is that after the fact, there will have been at least one successful investment scheme. A doubter will call the results luck; the successful investor will call it skill. The proper test would be to see whether the successful investors can repeat their performance in another period, yet this approach is rarely taken.
Early tests of efficient markets were tests of the weak form. Could speculators find trends in past prices that would enable them to earn abnormal profits? This is essentially a test of the efficacy of technical analysis. The already-cited work of Kendall and of Roberts (1959), both of whom analyzed the possible existence of patterns in stock prices, suggests that such patterns are not to be found. One way of discerning trends in stock prices is by measuring the serial correlation of stock market returns.
Both Conrad and Kaul (1988) and Lo and MacKinley (1988) examine weekly returns of NYSE stocks and find positive serial correlation over short horizons. However, the correlation coefficients of weekly returns tend to be fairly small, at least for large stocks for which price data are the most reliably up-to-date. Thus, while these studies demonstrate price trends over short periods, the evidence does not clearly suggest the existence of trading opportunities. A more sophisticated version of trend analysis is a filter rule. A filter technique gives a rule for buying or selling a stock depending on past price movements.
One rule, for example, might be: “Buy a security if its price increased by 1% and hold it until its price falls by more than 1% from its subsequent high. ” Alexander (1964) and Fama and Blume (1966) found that such filter rules generally could not generate trading profits. These very short-horizon studies offer the suggestion of momentum in stock market prices, albeit of a magnitude that may be too small to exploit. However, in an investigation of intermediate stock price behavior (using 3- to 12- month holding periods) Jagadeesh and Titman (1993) found that stocks exhibit a momentum property in which good or bad recent performance continues.
They conclude that while the performance of individual stocks is highly unpredictable, portfolios of the best11 performing stocks in the recent past appear to outperform other stocks with enough reliability to offer profit opportunities. Returns over long horizons: While studies of short-horizon returns have detected minor positive serial correlation in stock market prices, tests8 of long-horizon returns (that is, returns over multiyear periods) have found suggestions of pronounced negative long-term serial correlation. The latter result has given rise to a “fads hypothesis,” which asserts that stock prices might overreact to relevant news.
Such overreaction leads to positive serial correlation (momentum) over short time horizons. Subsequent correction of the overreaction leads to poor performance following good performance and vice versa. The corrections mean that a run of positive returns eventually will tend to be followed by negative returns, leading to negative serial correlation over longer horizons. These episodes of apparent overshooting followed by correction give stock prices the appearance of fluctuating around their fair values and suggest that market prices exhibit excessive volatility compared to intrinsic value.
These long-term horizon results are dramatic, but the studies offer far from conclusive evidence regarding efficient markets. First, the study results need not be interpreted as evidence for stock market fads. An alternative interpretation of these results holds that they indicate only that the market risk premiums vary over time: The response of market prices to variation in the risk premium can lead one to incorrectly infer the presence of mean reversion and excess volatility in prices.
For example, when the risk premium and required return on the market rises, stock prices will fall. When the market rises (on average) at this higher rate of return, the data convey the impression of a stock price recovery. The impression of overshooting and correction is in fact no more than a rational response of market prices to changes in discount rates. Second, these studies suffer from statistical problems. Because they rely on returns measured over long time periods, these tests of necessity are based on few observations of long-horizon returns.
While some of the studies cited above suggest momentum in stock prices over short horizons, other studies suggest that over longer horizons, extreme stock market performance tends to reverse itself: The stocks that have performed best in the recent past seem to underperform the rest of the market in the following periods, while the worst past performers tend to offer above-average future performance. DeBondt and Thaler (1985) and Chopra, Lakonishok, and Ritter (1992) find strong tendencies for poorly performing stocks in a given period tend to follow with poor performance in the following period.
For example, the DeBondt and Thaler study found that if one were to rank order the performance of stocks over a five-year period and then group stocks into portfolios based on investment performance, the base-period “loser” portfolio (defined as the 35 stocks with the worst investment performance) would outperform the “winner” portfolio (the top 35 stocks) by an average of 25% (cumulative return) in the following three-year period. This reversal effect, in which losers rebound and winners fade back, seems to suggest that the stock market overreacts to relevant news.
This phenomenon would imply that a contrarian investment strategy – investing in recent losers and avoiding recent winners – should be profitable. Moreover, these returns seem pronounced enough to be exploited profitably. The reversal effect also seems to depend on the time horizon of the investment. While DeBondt and Thaler (1992) found reversals over long horizons, and studies by Jagadeesh (1990) and Lehmann (1990) documented reversals over short horizons of a month or less, Jagadeesh and Titman (1993) found that stocks exhibit a momentum property in which good or bad recent performance continues.
This of course is the opposite of a reversal phenomenon. Tests of the Semistrong-Form Efficiency Event Studies: The abnormal return (AR) on a given stock for a particular day can be calculated by subtracting the market’s return on the same day (Rm) – as measured by a broad based index such as the S&P composite index – from the actual return (R) on the stock for that day. We write this algebraically as: AR = R – Rm The following system will help as understand tests of the semistrong form: Information released at time t – 1 > ARt-1
Information released at time t > ARt Information released at time t + 1 > ARt+1 The arrows indicate that the return in any time period is related only to the information released during that period. According to the EMH, a stock’s abnormal return at time t, ARt, should reflect the release of information at the same time, t. Any information released before then should have no effect on abnormal returns in this period, because all of its influence should have been felt before. In other words, an efficient market would already have incorporated previous information into prices.
Because a stock’s return today cannot depend on what the market does not yet know, information that will be known only in the future cannot influence the stock’s return either. Hence the arrows point in the direction that is shows, with information in any one time period affecting only that period’s abnormal return. Event studies are statistical studies that examine whether the arrows are as shown or whether the release of information influences returns on other days. These studies also speak of cumulative abnormal returns (CARs), as well as abnormal returns (ARs).
As an example, consider a firm with ARs of 1%, -3%, and 6% for dates -1, 0 and 1 relative to a corporate announcement. The CARs for dates -1, 0 and 1 would be 1%, -2% [1% + (-3%)], and 4% [1% + (-3%) + 6%], respectively. As an example, consider the study by Szewczyk, Tsetsekos, and Zantout on dividend omissions. The figure below shoes the plot of CARs for a sample of companies announcing dividend omissions. Since dividend omissions are generally considered to be bad events, we would expect abnormal returns to be negative around the time of the announcements.
They are, as evidenced by a drop in the CAR on both the day before the announcement (day -1) and the day of the announcement (day 0). However, note that there is virtually no movement in the CARs in the days following the announcement. This implies that the bad news is fully incorporated into the stock price by the announcement day, a result consistent with market efficieny. Over the years, this type of methodology has been applied to a large number of events. Announcements of dividends, earnings, mergers, capital expenditures, and new issues of stock are a few examples of the vast literature in the area.
The early event study tests generally supported the view that the market is semistrong-form (and therefore weak-form) efficient. However, a number of more recent studies present evidence that the market does not impound all relevant information immediately. Some conclude from this that the market is not efficient. Others argue that this conclusion is unwarranted, given statistical and methodological problems in the studies. Tests of market efficient can be found in the oddest places.
The price of frozen orange juice depends to a large extent on the weather in Orlando, Florida, where many of the oranges that are frozen for juice are grown. One researcher found that he could actually use frozen-orange-juice process to improve the U. S. Weather Bureau’s forecast of the temperature for the following night. Clearly the market knows something that the weather forecasters do not. Another group of researchers found that, as expected, stock prices generally fall on the date when the sudden death of a chief executive is announced.
However, the stock price generally rises for the sudden death of a company’s founder if he was still heading the firm prior to his death. The implication is that many of these individuals have outlived their usefulness to their firms. The Record of Mutual Funds If the market is efficient in the semistrong form, then no matter what publicly available information mutual-fund managers rely on to pick stocks, their average returns should be the same as those of the average investor in the market as a whole. We can test efficiency, then, by comparing the performance of these professionals with that of a market index.
Consider the figure below, which presents the performance of various types of mutual funds relative to the stock market as a whole. The far left of the figure shows that the universe of all funds covered in the study underperforms the market by 2. 14 percent per year, after an appropriate adjustment for risk. Thus, rather than outperforming the market, the evidence shows underperformance. This underperformance holds for a number of types of funds as well. Returns in the study are net of fees, expenses, and commissions, so fund returns would be higher if these costs were added back.
However, the study shows no evidence that funds, as a whole, are beating the market. Perhaps nothing rankles successful stock market investors more than to have some academic tell them that they are not necessarily smart, just lucky. However, while this figure represents only one study, there have been many papers on mutual funds. The overwhelming evidence is that mutual funds, on average, do not beat broad-based indices. Tests of Strong-Form Efficiency: It would not be surprising if insiders were able to make superior profits trading in their firms’ stock.
In other words, we do not expect markets to be strong-form efficient. The ability of insiders to trade profitably in their own stock has been documented in studies by Jaffee (1974), Seyhun (1986), Givoly and Palmon (1985), and others. Jaffee’s was one of the earliest studies to show the tendency for stock prices to rise after insiders intensively bought shares and to fall after intensive insider sales. To level the playing field, the Securities and Exchange Commission requires all insiders to register all their trading activity, and it publishes these trades in an Official Summary of Insider Trading.
Once the Official Summary is published, the trades become public information. At that point, if markets are efficient, fully and immediately processing the information released, an investor should no longer be able to profit from following the pattern of those trades. Seyhun, who tracked the public release dates of Official Summary, found that following insider transactions would be to no avail. Where there is some tendency for stock prices to increase even after the Official Summary reports insider buying, the abnormal returns are not of sufficient magnitude to overcome transaction costs.
Interpreting the Evidence How should the ever-growing anomalies literature be interpreted? Does it imply that markets are grossly inefficient, allowing for simplistic trading rules to offer large profit opportunities? Or are there other, more subtle interpretations? Risk premiums or inefficiencies? The price-earnings, small-firm, book-to-market, and reversal effects are currently among the most puzzling phenomena in empirical finance. There are several interpretations of these effects. First note that, to some extent, these three phenomena may be related.
The feature that low-price, low capitalization, high book-to-market firms, and recent stock market “losers” seem to have in common is a stock price that has fallen considerably in recent months or years. Indeed, a firm can become a small firm, or can become a high book-to-market ratio firm, by suffering a sharp drop in its stock price. These groups therefore may contain a relatively high proportion of distressed firms that have suffered recent difficulties. Fama and French (1993) argue that these anomalies can be explained as manifestations of risk premiums.
Using an arbitrage pricing approach, they show that stocks with greater sensitivity to size or book-to-market factors have higher average returns and interpret these returns as evidence of a risk premium associated with these factors. Fama and French argue that a three-factor model, in which risk is determined by the sensitivity of a stock to the market portfolio, a portfolio that reflects the relative returns of small versus large firms, and portfolio that reflects the relative returns of firms with high versus low ratios of book value to market value, does a good job in explaining security returns.
While size or book-to-market ratios per se are obviously not risk factors, they perhaps might act as proxies for fundamental determinants of risk. Fama and French argue that these patterns of returns may therefore be consistent with an efficient market in which expected returns are consistent with risk. The opposite interpretation is offered by Lakonishok, Shleifer, and Vishny (1995), who argue that these phenomena are evidence of inefficient markets – more specifically, of systematic errors in the forecasts of stock market analysts.
They present evidence that analysts extrapolate past performance too far into the future and therefore overprice firms with recent good performance and underprice firms with recent poor performance. Anomalies or data mining? Some wonder whether the anomalies mentioned and many others more are really unexplained puzzles in financial markets, or whether they instead are artifacts of data mining. After all, if one spins the computer tape of past returns over and over and examines stock returns along enough dimensions, some criteria will appear to predict returns simply by chance.
In this regard, it is noteworthy that some anomalies have not shown much staying power after being reported in the academic literature. For example, after the small-firm effect was published in the early 1980s, it promptly disappeared for much of the rest of the decade. Similarly, the book-to-market strategy, which commanded considerable attention in the early 1990s, was ineffective for the rest of the decade. Still, even acknowledging the potential for data mining, there seems to be a common thread to many of the anomalies that lends support to the notion that there is a real puzzle to explain.
It seems that value stocks – defined either by low P/E ratio, high book-to-market ratio, or depressed prices relative to historical levels – seem to have provided higher average returns than “glamour” or growth stocks. On e way to address the problem of data mining is to find a data set that has not already been researched, and see whether the relationship in question shows up in that new data. Such studies have revealed size, momentum, and book-to-market effects in other security markets around the world. Thus, while these phenomena may be a manifestation of a systematic risk premium, the precise nature of that risk is not fully understood.
So, Are Markets Efficient? An overly doctrinaire belief in efficient markets can paralyze the investor and make it appear that no research effort can be justified. This extreme view is probably unwarranted. There are enough anomalies in the empirical evidence to justify the search for underpriced securities that clearly goes on.
The bulk of the evidence suggests that any supposedly superior investment strategy should be taken with many grains of salt. The market is competitive enough that only differentially superior information or insight will earn money; the easy pickings have been picked. In the end, it is likely that the margin of superiority that any professional manager can add is so slight that the statistician will not be able to detect it. It can be concluded that markets are very efficient, but with rewards to the especially diligent, intelligent, or creative may be waiting.