The efficient market hypothesis (EMH) holds that security prices at any time "fully reflect" all available information. Fama (1970) classified evidence across three test categories in "Efficient Capital Markets": weak form, semi-strong form, and strong form. Serial correlations for daily price changes in 30 Dow Jones stocks averaged only .026. Jensen (1968) studied 115 mutual funds over 1955–1964. 89 fell below their risk-adjusted market benchmark, with net ten-year returns averaging -14.6% below the market line.
What the Study Found
Fama (1965) found serial correlations for daily DJIA returns averaged .026. A correlation of .06 explains only .36% of variation in the next day's price change. Filter trading rules required commissions of ~.1% per turnaround, wiping out any edge from short-term dependence. Ball and Brown (1968) studied 261 firms over 1946–1966. No more than about 10–15% of annual earnings announcement information had not been anticipated by the announcement month. Jensen found 89 out of 115 mutual funds produced net ten-year returns averaging -14.6% below the market line. Excluding loading charges, 72 out of 115 funds still fell below the line, averaging -8.9%.
Methodology
The paper reviews empirical studies using serial correlation tests, filter rule tests, event study residual analysis, and risk-adjusted performance evaluation. Datasets span 30 DJIA stocks (~1957–1962), 940 NYSE splits (1927–1959), 261 firms (1946–1966), and 115 mutual funds (1955–1964). Risk was controlled using the Sharpe-Lintner expected return framework and market model beta. Jensen used the S&P 500 as market proxy and measured nominal ten-year returns with continuous compounding.
Key Statistics
| Metric | Finding | Context |
|---|---|---|
| Average daily serial correlation | .026 | 30 DJIA stocks, ~1957–1962; Fama (1965) |
| Explanatory power at r = .06 | .36% of price change variation | Minimum correlation exceeding 2 standard errors; DJIA daily data |
| Minimum NYSE floor trader commission | ~.1% per turnaround | Clearinghouse fee wiping out small-filter trading profits |
| Splits with post-split dividend increases | 71.5% (672 out of 940) | NYSE splits 1927–1959; FFJR |
| Earnings information anticipated pre-announcement | ~85–90% | Ball and Brown (1968); 261 firms, 1946–1966 |
| Mutual funds below market line (net returns) | 89 out of 115 | Jensen; 115 funds, 1955–1964 |
| Average net deviation from market line | -14.6% | Jensen; ten-year returns including loading charges |
| Mutual funds below market line (ex-loading) | 72 out of 115 | Jensen; ignoring loading charges |
| Average deviation ex-loading charges | -8.9% | Jensen; ten-year returns |
| NYSE specialist sell above last purchase | 83% of all sales | Niederhoffer and Osborne |
| NYSE specialist buy below last sale | 81% of all purchases | Niederhoffer and Osborne |
| Transaction-to-transaction reversals vs. continuations | 2–3 times as likely | Niederhoffer and Osborne; NYSE price changes |
Why This Matters
EMH became the benchmark against which all active management claims are measured. The three-tier framework — weak, semi-strong, and strong form — remains the standard vocabulary for classifying what information can generate excess returns. The Jensen result established that even professional managers with full research resources cannot systematically exploit publicly available information.