TWO weeks ago, I tested the strategy of buying baskets of stocks with varying ratios of dividend yield to price-to-book ratio.

I downloaded the list of stocks trading on Singapore Exchange every year on March 31, starting from 1990. The last price taken was that of Jan 17, 2012.

Included in the data is the dividend yield of each stock, its price-to-book ratio (that is its share price relative to the historical cost after depreciation of its assets), and its last traded stock price of the day.

I ranked the stocks based on the ratio of dividend yield over the price-to-book (PTB) ratio. Let's call it D/PTB ratio. So if the dividend yield is 4 per cent and the PTB is 2 times, then the number I get is 2.

Stocks with the highest dividend yield, but trading at a very deep discount to book value will have a very high D/PTB ratio. The stocks are then ranked based on that number, from highest to lowest, and divided into 10 groups of equal numbers. Decile 1 would be stocks with the lowest dividend yield but highest PTB ratios; Decile 10 are stocks with the highest dividend yield, and lowest PTB. No screening was done on quality of assets and corporate debt levels.

Each year, there will be 10 groups of stocks. I then tracked the returns of the stocks in these 10 groups a year later. I used the median price appreciation of the stocks in that group as the return of that group. I also added in the median dividend yield to calculate the total return for that cluster of stocks. Each year, the groups of stocks will be different depending on their dividend yield and PTB as at March 31. Let's assume that 10 investors started 1990 with $100. Investor 1 will always invest in Decile 1 portfolio, and Investor 10 in Decile 10. At the end of the first year, each investor's portfolio would be the median return of the stocks in each decile, plus the median dividend. The amount of money at the end of the first year will be reinvested into the second groups of stocks in the various deciles in the second year. The process goes on for 22 years, from 1990 till Jan 17, 2012.

Here's what I found. Buying stocks with the highest dividend yield, but lowest PTB ratio appears to be a winning strategy. Following that strategy over 22 years would turn $100 into $1,575. That's a compounded annual return of 13.3 per cent a year. This group of stocks is relatively resilient during downturns, supported by the dividend yields. In comparison, stocks with low dividends and high PTB fared miserably. The $100 invested in Decile 1 would have diminished to just $16. That's a decline of 7.9 per cent a year. Meanwhile, the Straits Times Index's capital appreciation over that period was 3 per cent a year. That figure excludes dividends.

Having tested that strategy, I'm curious as to how it compares with the strategy of simply buying stocks with the lowest price-to-book ratios. The use of PTB as a predictor of a company's future stock performance was first highlighted by US finance professors Eugene Fama and Kenneth French in their breakthrough study in 1992. They showed convincingly that the lower the company's ratio of PTB value, the higher its subsequent stock performance tended to be. Fama and French found that no other measure had nearly as much predictive power - not earnings growth, price/earnings, or volatility.

So this week, I repeated the test by grouping stocks based just on their PTB to see how this strategy would fare versus that of D/PTB. Median dividends are also added to calculate returns. Decile 1 gathered stocks with the highest PTB and Decile 10 had those with the lowest PTB.

From Chart 2, you can see that indeed buying stocks with the lowest PTB would allow you to outperform the rest of the baskets of stocks. However, that strategy paled in comparison with that of buying stocks with the highest D/PTB ratios. A $100 invested in the lowest PTB portfolio since 1990 would have grown to $365. That's a compounded annual growth of 6.1 per cent. This is almost comparable to the STI's return if we assumed say a 3 per cent dividend yield a year for the STI component stocks. Remember, high D/PTB stocks turned $100 into $1,575 during the same period. The compounded annual return was 13.3 per cent. So, adding the additional criteria of dividend significantly boosted the performance of the basket of stocks picked. Comparing Chart 1 and Chart 2, you can see that the portfolios of high D/PTB stocks in general fall less than the general market when there is a downturn. This is because the dividends provided a floor for the prices of the stocks.

Another feature to note for Chart 1 and 2 is that, the performance of the highest D/PTB portfolio towered over the rest. It appears that this measure manages to capture the absolute cream of the crop of the market. The outperformance of the lowest PTB portfolio is not as great. And indeed Decile 9 and 8 also managed to outpace the STI, although dividend was not included in the latter's return calculation but was for the former two.

Chart 3 pitted the best portfolios from both strategies against one another. The highest D/PTB strategy is ahead, and started to put a greater distance between it and the lowest PTB strategy from 2004.

Chart 4 showed that only Decile 10 of the D/PTB strategy managed to outperform the STI. The other nine deciles, or 90 per cent, of the market fared miserably.

As for the lowest PTB strategy, not all good performing stocks ended up in one basket. They are spread out into three baskets - decile 10, 9 and 8. So going by the results above, it seems screening stocks based on both dividends and PTB would yield a better outcome than selecting stocks just on PTB.