Use Earnings-to-Price to Rank Firms with Negative Earnings
Summary
The document addresses how to summarize valuation data when companies, especially young firms, have negative earnings or negligible sales. It explains why price-to-earnings ratios become difficult to interpret in this setting: a negative ratio sorts below positive values numerically, although loss-making firms are generally considered more expensive or less attractive under a low-multiple ranking. Dropping such companies can also distort a sample, while assigning arbitrary large values introduces a different source of bias.
The proposed remedy is to work with earnings-to-price, the reciprocal orientation of the earnings multiple. This preserves the intended ranking: negative earnings yield values fall below small positive yields, which in turn rank below larger positive yields. The discussion focuses on the earnings measure and does not provide a parallel solution for missing or extreme price-to-sales observations. It also does not prescribe aggregation methods, outlier treatment, or adjustments for sector differences, so those choices still require separate consideration.
Key ideas
- Negative price-to-earnings ratios do not preserve the intended valuation ranking.
- Earnings-to-price places loss-making firms below firms with small positive earnings yields.
- Removing unprofitable firms can change the apparent valuation profile of a sample.
- The proposed earnings-yield transformation does not resolve missing or extreme price-to-sales data.
Tags
Full text
# Best way to treat negative P/E and extremely high P/S in data?
# Best way to treat negative P/E and extremely high P/S in data?
I'm working on a data that deals with start-up/similar companies data. A lot of the companies have negative P/E and/or little to no sales. Is there good ways to create meaningful data (like averages, medians) from this?
I obviously should not use negative values in calculations, since low P/E is meant be a positive factor. So far, I've tried two alternative ways, neither of which are great:
- Ignore all negative P/E and huge/undividable P/S data. However, this significantly does skew the data as you are easily dropping over 50% of the companies and thus gives much rosier picture of the market than it should.
- Assign a large dummy value for negative P/E and huge/undividable P/S. For example, any negative P/E becomes positive 50, while any missing P/S becomes 100. However, since these numbers are picked arbitrarily, it can unnecessary skew the data to higher-than-real averages.
Any ideas about better approaches? (And yes, I know this question is not necessarily a perfect fit for quant finance, but there really isn't any better place for it either...)
## Answer by nbbo2 (score 3, accepted)
https://quant.stackexchange.com/a/60559
The best solution according to most quants, is not to use P/E (Price to Earnings) at all but use E/P (Earnings to Price).
When you do this the negative E/P stocks are the lowest E/P, lower than the positive but small E/P stocks, which in turn are lower than the others. So the natural order is preserved.
When you use P/E the negative P/E need to "shifted" into the highest P/E category. Mathematically a negative is considered smaller than a positive, but this is contrary to the intuition about negative P/E: financially these are the most generously valued ("highest P/E") of all. So don't use P/E, the rank ordering of P/E's does not make sense.Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.