Negative denominator in ratio calculations

Research
1

Dear all,

How do you deal with negative denominators when calculating ratios? Example: we want to calculate the growth rate of a companies’ EPS. The EPSTTM value is 1 and EPSPTM is -2. If we divide 1/-2 it results in -0,5 (equivalent to -50% growth). This is incorrect. If we convert the denominator to absolute using the abs() function, we have 1/2, which results in 0,5 (50% growth) and it is also incorrect.

I have read an interesting paper on dealing with negative denominators, outlier influence and other relevant issues, but still can’t find a viable solution to calculate growth rates here at P123. Any ideas?

Also, I’d like to know how the readily available ratios from P123 deal with negative denominators.

Here is the paper: http://www.fintel.us/download/Problems%20with%20and%20Recommendations%20for%20Large%20Databases.pdf

Best regards.
Bruno

2

Yes, we do the absolute in the denominator. But in your example it is not 50% it’s:

(1 - (-2) ) / abs(-2) = 3/2 = 1.5 = 150%

3

Thank you Marco! This is just what I needed.

4

Another alternative:

Invert - Always Invert!

You will find that you will get better results by converting the (potentially negative) denominators into numerators. For example, instead of P/E<8, use Earnings Yield (such as E/P>0.13). Instead of EV/Ebitdattm<10, use Ebitdattm/EV>0.1. You might be surprised at the improvement in results for your ranking systems, sims and ports when you invert.

I find that inverting usually provides more accurate results than using (abs) or Eval formulas. The reason? Those negatives DO matter and should not be excluded just for our convenience. By inverting, they are not excluded, but they also do not throw off our objective with the formula.

5

I invert as well. But I have found that negatives only matter for some ratios but not for others.

6

Thank you both for your replies.

In the ratios where negatives matter, won’t Marcos suggestion have the same result?

Also, a possible pitfall of inverting is when you can’t predict if the numerator, the denominator, both or neither will be negative. This could happen when calculating EPS growth, if one or more companies in the universe had a positive EPS in the last period and a negative EPS in the current period. Anyway, I’m sure inverting some indicators will work, such as EV/EBTIDA (EV can’t be negative).

BTW, IntelligentValue the pic was a very unusual way to illustrate your point (LOL).