Methods of testing
The semi-strong form can be tested in two ways.
1. Trying to predict future price movements by using currently available public information. For example, do stocks with high dividend yields earn superior returns to stocks with low dividend yields? These tests can be based on time series analysis of returns for individual stocks or cross-sectional data for different stocks.
2. Seeing how quickly stock prices react to new information, to establish if there is a profit window. These tests are known as event studies.
When looking at how stock prices react to information, it is important to adjust the returns to information, it is important to adjust the returns to allow for market movements in the period concerned. Otherwise, any stock price movements may be solely due to general market movements.
The research therefore should focus on the abnormal price movement beyond that due to general market movement. For example, if the stock price would have been expected to rise by 10% due to general market movements and it actually rose by 12% there would be an abnormal gain of 2%. This would be referred to as an anomaly, since it is not expected under efficient markets.
In the long run, abnormal returns would be expected to come to zero, given an efficient market.
Time series tests
Test based on ratio.
Time series test have indicated that it is possible to predict returns and make superior profits by looking at various ratios such as price to book, dividend yields, yield spreads of lower grade bonds over high grade bonds.
Broadly speaking, studies have indicated that companies with low price to book rations and high dividend yields have outperformed over long periods of time. In addition, a high default spread (the amount by which the yield on low grade and large corporate bonds is different) has also been followed by high performance. On the face of it this would suggest that markets are not efficient, since these ratios can be used to earn higher returns even after allowing for transaction costs.
The reason for this may be that high dividend yields and high default spreads indicate a poor economic environment for stocks and bonds. As a result, investors require higher rates of return to compensate for the risk of investing in such securities.
Tests based on earnings reports.
Tests have looked at factors such as quarterly earnings report, aiming to establish whether it is possible to predict future stock price returns based on published earnings reports.
It has been found that it is possible to make abnormal returns when there are positive abnormal surprises in earnings, i.e. Earnings are significantly higher than expected. The positive information contained in the earnings surprise is not immediately reflected in the stock price.
In measuring whether or not an earnings surprise is significant, the early tests looked at percentage differences between actual earnings and expected earnings. More recent tests have looked at the difference between actual earnings and predicated earnings as a percentage of the estimate, based on a regression line predicting earnings using time series analysis.
These studies suggest that it is possible to make abnormal returns through trading based on earnings surprises. This evidence goes against the Efficient market hypothesis.
Other tests have looked at whether it is possible to make superior profits by following rules such as buying stocks in December and selling them in January.
This is to make advantage of tax-based trading in the market around the year-end when investors sell stocks in December and repurchase them in January. This is referred to as the January effect (or anomaly), stating that stock prices fall in December and then rise in January. It has been established that this is possible, but any gain may be eliminated by transaction costs. The effect worked for stocks that had suffered large declines in the previous year, but not for those which had large gains, supporting the tax motivation for this trading.
A weekend effect that has been tested is that returns on Mondays’ are negative, whereas returns on the other four working days are positive.
Cross-sectional tests are based on the assumption that in an effect market all stocks should earn a risk adjusted return given their betas. I.e. the returns should all lie on the security market line. If it is possible to predict returns for stocks based on public information other than that based on the security market line, then it indicates on anomaly from the efficient market hypothesis.
One problem with these tests is that they assume that the asset pricing model being used as the basis for the test, for example, the capital asset pricing model, is a good predictor of returns if the market is efficient. There is therefore a joint hypothesis in the test, that the asset pricing model does predict returns while the other factor being tested does not. If the asset pricing model is not effective, then the results will be distorted.
Tests of price earning ratios have indicated that stocks with low price earnings tend to outperform those with high price earnings. Possibly because the market tends to overestimate the growth prospects of high that stocks
Tests on price earning to growth ratios have given inconclusive evidence. Some people believe that companies with this ratios compared to the expected growth rate in earnings will offer superior returns in the future.
Tests based on the size of the companies have indicated that small firms earn consistently higher risk adjusted returns than large firms. There may, however be measurement problems here because trading in small firms is infrequent. If the small firm actually has a higher beta than that measured for capital asset pricing model purposes, then its predicted position on the security market line will be lower than it should be. Therefore, it will appear to be offering high risk adjusted returns although this is not actually the case.
Tests based on neglected firms that are not following by analysts indicate that they have abnormally high returns. However, this may be due to the fact that the measured beta understates the risk of these firms. If this is adjusted for, the return achieved will not be beyond the required on a risk adjusted basis. For example, investors may require a higher return on such firms to compensate for poor information.
Tests based on the ratio of book value to market value have indicated that firms with a high book value relative to market value earn superior returns, regardless of their size. Stocks with high book to price ratios (or low price to book ratios) are referred to as value stocks while stocks with low book to price ratios are referred to as growth stocks. Once gain, there is an argument that stocks with high book to market ratios are riskier and hence have to earn a higher rate of return to compensate.