As discussed in my earlier post, drawing strong conclusions from the F-Score poses some challenges as both low and high F-Scores show outperformance, contrary to what we would expect to see. At the center of this quandary is the head-scratching debate of Net-Nets and positive earnings.

Starting with a framework of low price-to-book stocks, the F-Score does a good job of sorting out the bad apples from the good apples. Central to the F-Score, and to its success, is the recognition of the importance of positive earnings. In fact, Piotroski notes that the two strongest individual explanatory variables of the F-Score are ROA (i.e., positive earnings) and CFO.

As a reader of my blog commented:

It is interesting to note that not all elements were equally important in the original study. The most important factors were positive net income and positive operating cash flow. Each of these two factors explains 70% -80% of the excess return of the F-score. A much simpler implementation of the F-Score, or an approximation, would be to buy stocks with a low price-to-book value and positive earnings (excluding stocks with losses) or positive operating cash flow.

It is perfectly rational to then assume the same should hold true for Net-Nets, which are low P/B stocks on steroids. A basket of Net-Nets with positive earnings should do extremely well.

Nevertheless, in a twist of logic, multiple authors have proposed a contradictory conclusion, when applied to Net-Nets. In an oft cited paper, Oppenheimer’s study concluded that, from 1970 to 1983, money-losing U.S. Net-Nets outperformed Net-Nets with positive earnings. This conclusion was later reconfirmed by Tobias Carlisle, et al., in their 1984 to 2008 study (also using U.S. Net-Nets).

In spite of this research, most reputable investors lean towards Net-Nets with positive earnings. There are exceptions of course, for example, in biotech or special situations. On the whole, though, I am not aware of anyone who claims to only buy money-losing Net-Nets because they outperform.

To test this theory, I use Net Income before Extraordinary Items, as used by Piotroski. I should also note that I use a trailing-twelve-month number (TTM). In Piotroski’s study, he used fiscal numbers with a 5-month lag. Using lags are pretty common in academia, as they eliminate any debate as to whether financial information was adequately distributed before the time of purchase (and in some cases, fiscal data is all that is available). Aside from Cliff Asness’s HML-Devil, that explicitly tested the lag effects on P/B returns, lags are typically used. As practitioners, we of course would not use a lag and would not limit ourselves to only fiscal numbers. We are going to use the most recent available data. Because of this, I use TTM numbers.

Using LSEG’s point-in-time data, I’ve got pretty good assurance that the numbers I’m using were available at the time of purchase. For example, if we take a random sample from the database, Ohmoto Gumi was purchased on 4/20/21 using financial data from 12/31/20, not financial data from 3/31/21. As is the case with HML-Devil, stock price has no lag and is the closing price as of the ‘Start Date’.

Symbol	1793.T
TR.CompanyName	Ohmoto Gumi Co Ltd
TR.gicssector	Industrials
Start Date	4/20/2021 0:00
End Date	4/20/2022 0:00
TR.F.PeriodEndDate	12/31/2020
Start Date (less) TR.F.PeriodEndDate (in days)	110
Start Price	1860
TR.DPSActValue	56.67
End Price	2000
% chg	0.1057365591

Out of the 760 stocks in the database, 4 data points on earnings returned NULL, leaving 756 stocks in this study.

Below is count by year of the two baskets of stocks tested: companies with LTM (last-twelve-months) earnings above 0 and companies with LTM earnings below 0. In most years, the positive earning stocks outnumber the negative earning stocks. Aside from 2018, each year offered a reasonably sized basket in each cohort.

The results may or may not be shocking: Japanese Net-Nets with LTM Earnings below zero significantly outperform Japanese Net-Nets with positive earnings. And by significant, I mean a lot! Over the length of study, negative earners compounded at 22.39% CAGR and the positive earners compounded at a paltry 17.22% CAGR. For comparison, the baseline (all Net-Nets) compounded at 19.60%. From this simple study, it appears that the theory that negative earning Net-Nets outperform positive earning Net-Nets holds true in Japan as well.

When stacked against our earlier study on the F-Score, Japanese Net-Nets with negative earnings nearly matches the return profile of the high F-Score cohort, albeit with higher volatility. Again, this adds to the difficulty in properly evaluating the efficacy of the F-Score, as positive earnings is a fundamental building block of the F-Score.

As noted in the Carlisle study, Benjamin Graham often recommended that it was better to select Net-Nets that had positive earnings and paid a dividend. But the evidence suggests otherwise. For those of you with with a ton of time and curiosity, it would be interesting to comb through the Graham–Newman Corp. letters and count the number of positions Graham held in companies with negative earnings in the prior year.

The reasons why money-losing Net-Nets works so well is hard to figure out. Some have suggested that perhaps management is more inclined to “do something” to unlock shareholder value because the ship is clearing sinking.

But perhaps it’s purely behavioral. It’s just so darn hard to buy neglected and unloved stocks; additionally, ones that are also losing money! This would lead to the role of the discount to price. It may be possible that money-losing Net-Nets outperform simply because they might be cheaper than their more shiny brethren. In this particular study, money-losing Net-Nets were cheaper only 50% of the time. This may be cause for further study.

Whatever the reasons, as one who manages outside capital, I can attest to the difficulty in buying money-losing Net-Nets. It’s hard enough to communicate to partners the virtues of buying dirt cheap stocks, but money-losing ones is a whole other ball game. And with the increased volatility, talk about scaring the bejesus out of your partners! But for those with more permanent capital that can handle the behavioral challenges it offers a compelling argument.

For those who feel more comfortable in the confines of positive earners and dividend payers, I report the results a bit in jest: you will still have quite satisfactory returns. A 17% CAGR is nothing to sneeze at. It is, though, the worst performing strategy presented. It would make sense then to look to add extra criteria to your filtering process.

Thanks for reading!