It uses a combination of the three types of composite factor systems discussed in the book: Value Composite Factor (50%), Financial Strength Composite Factor (25%), Earnings Quality Composite Factor (25%).
Not spectacular, but low turnover and reasonable liquidity. Not sure how well it will hold up. But hey, it’s free. I don’t trade it myself, but was using it to get practice using P123. I have been an O’Shaughnessy fan over the years, and wanted to test for myself.
Thanks Tom for sharing this but in your ranking under the node “O’Shaughnessy Earnings Quality Composite” you have quite a few custom formulas. The ranking cannot be used without having those formulas public. Is it possible to put the whole formula in the ranking system?
it’s coming back to me now. one of the reasons I was doing using formulas was to make sure I didn’t run into the space limitation. I was able to do it in most of them except for “Change in Operating Assets”. Here are the formulas if you want to add them to your custom formulas.
That’s terrific work! It’s hard to do an O’Shaughnessy model because he presents such a wide-ranging study. So anyone attempting to do such a model must do a lot of picking and choosing, and in this regard, you’ve made some nice choices, particularly the heavy emphasis on value and the large role for earnings quality. And it’s very kind of you to make it public – this can be a great learning experience for many in the p123 community!
If it’s OK, I’d like to offer some feedback.
In the external financing formula (a nice addition to the model), I notice you’re using Annual items. Perhaps you might substitute TTM, to keep them fresh as the year progresses.
The Debt to Cash Flow formula is: DbtTotQ / (Price / Pr2CashFlQ). I’m not sure why the denominator is expressed as a price ratio rather than just cash flow. Also, bear in mind that both debt (especially the short-term component) and cash flow can fluctuate a lot over the course of a year for benign reasons. You could use TTM values for both (debt isn’t usually thought of as a TTM item, but we do compute it for purposes such as this, where we might want to use a running tally (an average) rather than the traditional balance-sheet snapshot.
I’m curious about the change in operating assets factor (lower is better). This definitely can be home to some serious earnings quality problems, but also, it can signify plain-vanilla inefficiency (especially PP&E). I need to think about that some more. Maybe you can focus on the current portion of operating assets (net of payables) and, perhaps PP&E/Sales, or even ROE (I know O’Shaughnessy likes that in conjunction with value).
Finally, I did an exercise I typically do when I fine tune my models; my alternative to the quant “robustness” tests: I look at the passing stocks to see if those are the sort I feel should have made it into the model. I noticed right away that the current list has two out of ten stocks in for-profit education (APOL and CPLA). I don’t have a definitive feeling about that – I (and many others) constantly go back and forth (it’s a business that’s important to our future and typically incredible fundamentals measurable many different ways but balanced by some serious ethical problems among industry participants and aggressive regulatory responses). So it is important that anyone who has a model that’s prone to pick up these stocks (as would be the case with any sort of intense value-quality orientation) take a good look at the group and come up with a go or no-go opinion, and if it’s a no-go, look for ways to model them out. If you’re a go, then nothing more needs to be done. If you come out as a no-go, a comparison of APOL and CPLA, even from the p123 data panels alone (i.e. without the 10-Ks, which you should check anyway), suggest some possible ways to screen situations like these out. Notice that APOL’s sales are falling briskly and that its charge-offs have been heavy. ROE is strong but trending badly. CPLA is a lot less conspicuous in these regards. If CPLA is ethically stronger or further along in cleanup, that might provide a clue that screening rules setting minimum growth and/or ROE trend thresholds could help. Momentum filters might be useful for other such situations but not specifically these, since those coming out as “go” for this group have most recently had the upper hand in the market. And again, you may decide this is OK (obviously, you wouldn’t be alone – a bullish case would be based on valuation and cleanup efforts already in place and presumably discounted by the market). I point this out because the p123 panels (Chapter 7 of the Z to z Guide) are badly underappreciated and underutilized, and could go a long way toward helping designers improve out-of-sample performance by better designing out situations stocks that pass the letter of the law but fail the spirit of the law; we can never be perfect in this regard, but we can often accomplish quite a lot.
On this one, as I recall, I couldn’t find a field for cash flow (I found one for Free Cash Flow). So I tried to back into it thinking if I used (Price / Pr2CashFl) it would be the same as (Price / (Price / Cash Flow) ) and I would end up with Cash Flow.
on this particular factor I was just pulling from pp. 278-279 of 4th edition of What Works On Wall Street.
Here is the quote:
In their paper, “Do Investors Overvalue Firms With Bloated Balance Sheets?”, David Hirshleifer, Kewei Hou, Siew Hong Teoh, and Yinglei Zhang state, “When cumulative net operating income (accounting value-added) outstrips cumulative free cash flow (cash value-added), subsequent earnings growth is weak. If investors with limited attention focus on accounting profitability, and neglect information about cash profitability, then net operating assets, the cumulation of the discrepancies between the two, measures the extent to which reporting outcomes provoke over-optimism. During the 1964-2002 sample period, net operating assets scaled by total assets is a strong negative predictor of long-run stork returns.” They go on to state that, “the level of net operating assets…measures the extent to which operating/reporting outcomes provoke excessive investor optimism…In other words, a high level of net operating assets, scaled to control for firm size indicates a lack of sustainability of recent earnings performance.”……we see that the data confirm their hypothesis. Decile 1, those firms with the highest change in net operating assets, earned an average annual compound interest of just 3.5%, significantly worse than the All Stocks Universe and worse than an investment in U.S. T-Bills……Deciles with the lowest change…especially deciles 9 and 8, significantly outperformed……
Thanks. For those interested in the choices, take a look in the 4th edition, Chapter 14, “Accounting Ratios”, and Chapter 15, “Combining Value Factors into a Single Composite Factor”. That’s what I was referring to in coding this (and looking at the posts on this board of a lot of very knowledgeable people).
Seems to me that the numerator is a total value for the entire company, but the denominator is a per share number.
I think Price / Pr2CashFlQ should come to Cash Flow Quarter per share.
Maybe try this: DbtTotQ / (MktCap / Pr2CashFlQ)
Or this might be cleaner, based on Paul’s post: DbtTotQ / (DepAmortCF(0,TTM) + NetIncCFStmt(0,TTM))
The definition of Pr2CashFlQ says: “Cash Flow is defined as Income Before Extraordinary Items plus Depreciation and Amortization.” Not sure if that is the same thing.