No. of Recommendations: 12
Hi Jim (mungofitch),
I know you have regularly pointed to ROE as one of the best "single factors" to pick good stocks. I am just wondering in any of your backtests, have you ever tested an average ROE (like average of the last 2, 3 or 5 years) rather than just using TTM ROE or last Financial Year ROE.
The reason I ask is that I scaped ROE from Marketwatch for the 100 stocks in the Nasdaq 100 (I don't think they liked this as I think they may have blacklisted my IP address now). Anyway, the highest ROE stock I identified was Verisk Analytics Inc (Ticker: VRSK). But looking at say
https://www.morningstar.com/stocks/xnas/VRSK/key-m... here is the ROE in recent years along with Total Equity:
Year ROE Tot. Equity
2025 444.05% 309.8
2024 466.96% 105
2023 59.67% 322.2
2022 41.78% 1767.7
2021 24.16% 2842.5
2020 28.74%
2019 20.77%
2018 29.97%
2017 34.08%
2016 43.72%
It sort of looks like the high ROE is the result of a significant drop in Equity. I didn't look into what caused the drop in Equity, but I was wondering if there was any backtest evidence to support the use of some sort of average ROE rather than a single year that might reduce the impact of any anomolies.
Kind Regards
StevnFool
No. of Recommendations: 21
Long story short, it's hard to get a meaningful ROE figure. The two approaches to mitigate that seem to be
* live with the fact that a naive ROE calculation will give you the wrong idea sometimes, and just buy a whole lot of high ROE stocks to lose it in the averages, or
* spend some time really looking at a company to see what its return on "meaningful" assets is. Either by hand, or with more and more elaborate screening criteria.
The first thing you'd want to check are that it isn't a high ROE because they've simply geared up too far. For example, imagine a stock with moderate equity and nice profitability, a medium ROE. They borrow a gazillion dollars and pay out a big special dividend, the shareholders' equity drops near zero, and the ROE now seems super high even though it's the same underlying business activity. Years ago, this was called a "recap". This is of course bad: they firm is now highly leveraged and much more at risk of failure, and at the very least earnings are likely to drop due to interest costs. Maybe that's what happened with the firm you looked at? The same thing can come from a huge one-time write-off. Also, fresh share issuance increases equity without (immediately) changing returns, so ROE falls. Huge buybacks depress book per share, flattering ROE if there is enough of it done. Frankly, lots of things can throw it off.
The biggest risk is overgeared firms looking "too good" based on ROE. Checking leverage levels as a simple extra step is wise, if you're placing bigger wagers per company. Usually folks look at debt to equity ratios, but not me. Personally I don't find it easy to pay off debts with assets...ever try to sell half a factory machine to make a loan payment?...so I like to look at the ratio of long term debt to normalized earnings. A firm with debt less than five typical years of earnings is not overly geared. That can go up to 10 years for the occasional firm that has extraordinarily steady and long-lived streams of earnings. Hershey is sometimes used as an example here. This can go to extremes: a few firms have such high reliability of earnings streams that they pay out all their equity, and more, and end up with negative book value while earnings keep on chugging. Naively dividing earnings by equity you get a negative ROE, but in reality this is in effect an infinite ROE: they don't need any net assets at all to earn money, and on average they are good investments. Moody's, Coke, etc. But that's rare. In fact, a tiny number of firms with super high P/B also makes a good screen for analogous reasons, but I digress.*
Another check which makes a lot of sense, but which oddly I don't usually do, is to look at ROA instead of (or addition to) ROE. As much as possible, you want to look at what return they're getting on the assets they are actually using to generate their earnings. A really big disused pile of cash would not be included.
In short, a high ROE emphatically does not mean you are looking at a great business, because of flaws and gotchas like those above. However, essentially all great businesses have high (or infinite) ROE, especially over long periods. So, despite the outliers, a large slate of firms with high ROE will have a higher-than-average chance of having above-average quality compared to a randomly selected set.
To your specific question, do I look at recent or multi-year average ROE, I have historically tested things only with recent ROE because that's what I had available. But when I'm doing global equity screening, I look at five year average ROE because the screener I use for that has that field! Really one should check both: high average so you know the recent figure wasn't just anomalously good because of a one-time gain, and high recent so you are skipping firms that just had a permanent turn for the worse, or highly variable year-to-year returns.
Sorry for rambling, I think about ROE way too much. As far as I know, if you wanted to do quant investing and had only one field you could use, that's what I'd pick. Despite the flaws.
From within the Value Line 1700 set, equally weighted portfolios:
1997-2025, top 25% by ROE, 13.07%
1997-2025, bottom 25% by ROE, 9.53%
1997-2025, S&P 500, 9.94%
Jim
* Here's a surprising one for a "deep value" investor
2000-2025, S&P 500, 8%/year
2000-2025, top 8 from VL set by price-to-book-value (i.e., seemingly most overpriced), 16%/year
The reason this works: to a very crude first approximation, people pay similar P/E ratios for companies. If the P/E ratio is comparable to the average but the P/B is very high, the firm must have a high ROE. Firms with consistently high ROE have something that is preventing others from raising money to undercut them, some sort of moat or monopoly: if it only takes a few assets to be in that business, everybody would want to do so.
No. of Recommendations: 0
Thanks Jim,
So taking feedback from both yourself and musselmant, it would seem like the following may be a reasonable strategy:
Possibly for starting with the Nasdaq-100 universe after removing any companies that you have no wish to invest in for various other reasons:
For larger quant portfolio (maybe 40 - 50 stocks):
Just use a single ROE criteria - either most recent year or maybe a 3 or 5 year average - whatever is readily available.
Additional criteria if you want to concentrate with less picks (maybe 15 - 20 stocks):
* Eliminate companies that will take more than 5 years of normalized earning to repay Long Term Debt. Possibly extend to 10 years if earnings are very stable / reliable.
* Look for recent year ROE > say 5 year average ROE. I think this should go some ways to ensuring that there is still a good return on incremental capital.
* Rank what is left based on high ROE. Give consideration to doing this as a sum of ranks of both Recent ROE and 5-year average ROE.
