Price to Book Ranking vs. Benchmark Question

Research
1

Hi,

I’m trying to create a screen showing how stocks perform by P/B, P/E, P/S by quintile. I’m using Price to Book (P/B) for my example using the following criteria:

Main Settings:
Universe: S&P 500 Index
Benchmark: S&P 500 Eq Weight

Rules:
Frank(“Pr2BookQ”,#All,#ASC)>X
Frank(“Pr2BookQ”,#All,#ASC)<Y

Where X and Y are the thresholds:

Lowest 20%: X=80 Y=(blank out formula)
20-40%: X=60, Y=80
40-60%: X=40, Y=60
60-80%: X=20, Y=40
Highest 80%: X=(blank out formula) Y=20

Rolling Backtest:
Frequency: Every 4 Weeks (about 208 samples)
Holding Period: 1 Year
Time: 1/1/2000 - 11/14/2015

The idea is that the lowest P/B stocks perform better than the highest P/B stocks, but when I take the average it doesn’t equal the S&P 500 Eq Weight index

Quintile Annual Performance:
Lowest 20%: 14.38%
20-40%: 13.45%
40-60%: 11.67%
60-80%: 9.21%
Highest 80%: 9.08%

Average Performance: 11.56%
Benchmark: 9.34%

Could you tell me why my average performance is 2.22% higher than the equal weight benchmark? Does the S&P 500 Eq Weight Benchmark not include Dividends but the others do?

2

sp500 ew is index, that is why it does not include dividends.

3

Your X-Y range is non-inclusive. That is, you need to make it FRank(…)>=X. Just using “=” results in some stocks being excluded from every quintile. Also, it’s possible given the way normalization works on the site, that the way you are breaking up the quintiles won’t result in exactly the same number of stocks in each.

Finally, why are you doing this in a screen. Just create a ranking system with P/B as the only factor and then run the binned performance test with however many (or few, as in 5) bins you like and with a single test you’ll be able to see graphically numerically all of the quintiles.

4

Thank you for the advice. I’ll try the ranking system you suggested.