This essay critically assesses the efficient market hypothesis (EMH) by examining empirical evidence that tests its validity.

Brown and Goetzmann point out, however, that the phenomenon is dependent upon the time period studied. It is predominant in the late 1970’s and late 1980’s. Reversals of returns are also observed for 1981 and 1987 and Malkiel (1995) found reversals in two years following the sample period. Persistence appears to be correlated across managers and suggests correlated dynamic portfolio strategies. It is also suggested that winning funds may be investing on a macroeconomic factor that is priced.

In contrast to this test are the results of a test conducted by Kahn and Rudd. They test for performance persistence in active equity and fixed-income funds and they test these separately. They calculate both in-sample and out-of-sample for monthly returns. Performance is calculated out-of-sample and is divided into two periods. Performance is classified in three ways. It is defined as cumulative total returns, cumulative selection returns or as information ratios. Their first test used regression analysis. They regressed Period 2 performance against Period 1 performance. A positive estimate of the coefficient with a  significant t-statistic is evidence of persistence . Their second tests used contigency tables to test for performance persistence. Their methodology is thus well prepared and leaves little to chance. Regression analysis showed  significant evidence of persistence in equity when it was defined as information ratios only. None of the contingency tables showed evidence of persistence for equity. Regression analysis found persistence for fixed-income funds when persistence was defined as selection returns and information ratios. The same results were found using contingency tables.

The tests by Brown and Goetzmann would indicate performance persistence is present.  The tests by Kahn and Rudd suggest it is a fixed-income phenomenon with very limited occurrence in equity returns. Overall, evidence is inconclusive on performance persistence and results seem to depend on the time period studied and the methodology used. More conclusive evidence of historical returns being used to predict future returns may be found by examining the occurrence of momentum effects.

Momentum Effect

Momentum effects are a related phenomenon to persistence performance. They describe a situation whereby winning stocks continue to outperform losing stocks over an intermediate horizon. The reason for this momentum in stocks is generally explained by a sluggishness in the reaction of stock prices to analyst’s earnings forecasts as well as, to a lesser extent, prior returns.

Chan, Jegadeesh and Lakonishokconduct a series of tests to validate this phenomenon. They use all domestic, primary stocks listed on the New York(NYSE), American(AMEX) and Nasdaq stock markets from January 1977 to January 1993. In each test stocks are ranked by different factors and then assigned to ten equal-sized portfolios. The ranking factors include compound returns over the past six months, the most recent standard unexpected earnings (SUE), and the abnormal return around the most recent past announcement of quarterly earnings. Under each ranking returns for the prior six months and future three years are listed.

In nearly all cases the spread for returns between winner and loser portfolios remain positive for all horizons examined. The factors driving return momentum seem evident from the data given. Returns are closely aligned with both SUE’s and abnormal earnings announcements. The spread in SUE and abnormal earnings announcements (between portfolio 1 and portfolio 10) is positive under all rankings for nearly all dates examined.

Cross-sectional regression is used to separate the effects of earnings surprises and historical returns on future returns. For both momentum variables, their effect, when taken separately, is strongly and positively related to future earnings. When future returns are regressed against both simultaneously they are both significantly different from zero with earnings surprises having greater explanatory power.

Perhaps the most compelling evidence for momentum effects or performance persistence is that presented by Jegadeesh and Titman7. They argue that if a stock overreacts or underracts to information profitable trading strategies based on past returns exist. They select stocks based on their past returns for the past one, two, three and four quarters. They performed thirty- two tests and used stocks from the 1965-1989 period. Their strategy involved buying past winning stocks and selling past losing stocks. Their ‘relative strength’ strategies realised significant abnormal returns for the 1965-1989 period. The strategy which selected stocks based on prior six month returns and then held them for six months was the most profitable. The tests indicated that the profits were not due to their systematic risk nor to lead-lag effects resulting from a delayed price reaction to common elements. The evidence showed profits were due to delayed price reaction to firm-specific information.    

In summary, momentum effects do appear to be genuine in the short run and seem to result primarily from a delayed reaction in stock prices to earnings announcements and secondarily from historical returns.

Value Effect  

Over long horizons ‘value’ stocks appear to give higher future returns relative to ‘glamour’ stocks. Value stocks include stocks with high earnings, cash flow or tangible assets relative to current share price. Glamour stocks are the opposite of these. Fama and French (1992)8 found that stocks with low price-earnings ratios outperformed the market. De Bondt and Thaler (1985)9 made portfolios consisting of value stocks and glamour stocks. The stocks were based on book-to-market value, cash-flow-to-price, price-earnings ratio, sales growth and past returns. They used historical data on U.S. returns. They found that value stock portfolios outperformed glamour portfolios over investment horizons of one to five years.

Lakonishok, Shleifer and Vishney (1994)10 found similar evidence for the value effect (1994). They examined returns on stocks bought on the basis of their book-to-market value. To control for size they divided the stocks into five size categories and then divided each size category into ten equal-sized groups based on their book-to-market value. High book-to-market stocks gave an average return   7.8% per year higher than low book-to-market stocks. They also note that the higher returns on high book-to-market stocks are not due to them carrying more risk as their variability of return is no greater than that of the low book-to-market stocks.

In each of these tests there is a value effect. Public information, here being measures of value, are not fully reflected in current security prices. The market is not being efficient. The evidence is strong  and undisputed and suggests excess returns can be earned with this information.

Size Effect

A relationship has been observed to exist between firm size (ME), a stock’s price times shares outstanding, and average return.  Small firms  seem to produce higher returns than large firms. Banz11 documented that over the period 1936-1977 in the U.S. small firms gave a return 19.8% higher than large firms. In fact, the size effect was as statistically significant in explaining return as was beta.

Explanations offered to account for this included the  beta estimate being too low for small firms, the Capital Asset Pricing Model (CAPM) was an inappropriate model for measuring expected return and size may be representing an omitted risk variable such as  ‘survival probability’ in economically hard times. When Chan, Chen and Hsiah12 used a multi-factor model, the Arbitrage Pricing Theory model, (APT), the size effect disappeared. The difference in return between small firm stocks and large firm stocks was, on average, 1.5% per year. The additional variable  used was the difference in return between high risk corporate bonds and government bonds. They do argue, however, that small firms are riskier than large firms due to their low production efficiency and high leverage.

In light of the fact that use of the APT model removed the size effect it may seem appropriate to state that the EMH holds up. Additionally, many researchers have argued that transaction costs on small stocks are very high and when these costs are taken into consideration the excess return is removed or reduced significantly. Tests conducted by Fama and French (1993), however, leave little doubt as to the existence of a relationship between firm size and average return. Furthermore, their tests cast doubt on the validity of the asset-pricing model of Sharpe, Lintner and Black (CAPM), which claimed that expected returns on securities are a positive linear function of their Betas.

Beta

 Fama and French tested stocks from non-financial firms for the years 1963-1990. They used stocks listed on the NYSE, AMEX and NASDAQ. When common stock portfolios were formed on size alone, regression analysis revealed that average return seemed to be related to beta, a firm’s undiversifiable or systematic risk. The betas of the size portfolios, however, were almost perfectly correlated with size. When they removed the effect of size they found there was no relation between beta and average return. A reliable negative relation persisted between size and average return when other variables were used in the regression. Other variables included leverage, earnings-price ratio, and book-to-market equity. The average slope from regressions on beta alone was insignificant. Beta remained insignificant when it was combined with the other variables in the regression. Fama and French tested for the period 1941-1990 to check their results against a longer time period. For this test they used NYSE returns for 1941-1990 and found a reliable size effect but little relation between beta and average return.

Tests by Burton Malkiel and Yexiao (1997)13, however, shed new light on the size effect, beta and average return. They found that size may be representative of a firm’s idiosyncratic or specific risk. They measured the idiosyncratic volatility of the stocks in the S&P 500 index for the 1963-1994 period. They found that idiosyncratic volatility was strongly correlated with the market capitalization or size of each firm in the index. They then formed ten  portfolios of companies and ranked them according to their specific risk. For the period studied, 1963-1994, a definite positive relationship was found to exist between specific risk and average return. This result runs contrary to the EMH which argues that only non-specific or sytematic risk is priced and earns a risk premium. It also serves as a possible explanation for the relation between firm size and average return. It is strong evidence that there is no size effect. Rather, specific or firm-specific  risk is being misrepresented by firm size.  

Conclusion

It seems to be irrefutable that capital markets are not perfectly efficient. Empirical evidence shows that  past information has  a  bearing on future returns. The question that is sometimes harder to answer is exactly what information or variable is responsible for the inefficiencies observed. The well established relationship between a firms size and average return is challenged by evidence that claims size should be a firm’s specific risk. Specific risk may in turn be challenged as it may be decided that it is a proxy for the  systematic risks that influence security returns.  

       

           

       

 

---