3: Dividend Yield

Description of Factor:

This factor is the indicated annual cash dividends divided by the closing monthly market price. The rationale behind this variable is the expectation that stocks with a high yield will outperform stocks with a low yield. However, the trend over the last several years for companies not to pay dividends excludes many stocks from this screen.

Analysis

Equal Weighted: This factor has a relatively strong predictability for quintile 1, but does not perform as well for quintile 5. Quintiles 4 and 5 on average had higher returns than did quintile 3. The average annualized return for quintile 1 is 23.52%, while the bottom quintile’s average return is on 15.83%. Seemingly this would be a good candidate for a long-short strategy just looking at these returns. However, in analyzing the returns each year, the results are not as good. In 2 out of the last 5 years the quintile 1 ranked the lowest. In addition, quintile 5 had inconsistent ranks. In 2003 if our portfolio was long the quintile 1 and short quintile 5, it would have lost 48%. Quintile 1 also has a moderate turnover with only 10% monthly turnover, which would make this a good candidate for a scoring factor.

The average excess returns over the market for quintile 1 is 3.04% and for quintile 5 (2.53%). For quintile 1 the beta is .65 with an R² of .64 and quintile 5 has a beta of 1.2 with an R² of .58.

Value Weighted: Value weighted portfolios performed very similarly to equal weighted portfolios, with annualized average returns of 25.43% for quintile 1 and 15.25% for quintile 5. There is a decent step down function, until we get to quintile 5, where the annualized average return is higher than quintile 4. The betas of quintile 1 and 5 are almost equivalent (.90 and 1.07 respectively), allowing for a better hedge for market exposure. The returns of a value weighted long-short strategy are greater than those of an equal weighted in 2003. However, the value weighted portfolio long short strategy would produce negative returns 1998 and 1999, losing 82% in 1999.

Conclusion:

Using dividend yield as a predictive factor we could go with either equal weighted or value weighted quintiles. The value weighted quintile produces a better market hedge since the net beta is near zero. However it is not prudent to follow a long-short strategy on this factor alone due to the large fluctuations in returns from year to year. Furthermore, while we would generate positive returns most of the years, the years where quintile 5 outperforms quintile 1 would be disastrous with huge losses. As a value predictor, we could potentially use quintile 1 in conjunction with other factors to form a trading strategy.