Being (Conservatively) Roughly Right

by Geoff Gannon

The latest podcast episode is 20 minutes all about Frost (CFR). Frost is a stock I own. And it’s a stock I wrote something like a 10,000 word report about a few years ago (the report is available on the Focused Compounding member site). I was also interviewed at length about this stock last year. So, you’d think that I would have all my facts exactly straight when Andrew and I recorded the Frost podcast.

Our podcasts are recorded without notes. I don’t know what questions Andrew is going to ask ahead of time. And he doesn’t go back and edit them later. We don’t re-record anything. So, what you hear in that podcast is 20 minutes of me talking “on the spot” about Frost. The only things I’m able to say are things that come immediately to mind. So, the facts you hear me reel off are the ones I’ve repeated in my mind over and over again when thinking about this stock. The stuff I get wrong is the stuff I don’t thinks is very important.

I make several factual errors in that podcast. And I think that’s instructive. I put 25% of my portfolio into this stock a few years ago. And I continue to have about 25% of my portfolio in the stock today.

Yet, I say several things in that podcast that simply aren’t true. For example, I say that Frost doesn’t do consumer lending. If we take a look at the company’s most recent 10-K, we can see I’m wrong about that:


The bank makes consumer loans. As you can see, consumer real estate plus “consumer and other” is 12% of the bank’s loan portfolio. This is barely more than energy loans (11.4%) and much less than commercial real estate loans (40.2%). Business loans of some kind are about 88% of the bank’s loan portfolio. Furthermore, the bank’s loan portfolio is only about 40% of its total assets. The securities Frost buys are usually not consumer related – for example, they have just 6% of their bonds in mortgage backed securities. State and local government bonds are 65% of the portfolio and U.S. Treasuries are 29% of the portfolio. The bank has about 40% of its assets in loans and about 40% of its assets in bonds. So, it has maybe 5% of assets in consumer loans (11.4% times 0.4 equals 4.6%) and maybe 3% of its assets in consumer backed mortgages (6% times 0.4 equals 2.4%). In other words, Frost has about 93% of its assets in things tied to businesses and governments rather than households. In my mind this simplified to “Frost doesn’t do consumer lending”. I did, however, remember that what consumer lending the bank did was made up in large part by home equity loans.

Obviously, I didn’t think it was worth spending any time thinking about Frost’s consumer loans or its mortgage backed securities. And, I think I’m right about that. Close to 90% of Frost’s loans are not consumer loans and close to 60% of Frost’s assets are not loans at all. To me, this meant I should focus all my analysis of Frost’s assets on just two categories: business loans and government bonds.

During the podcast, Andrew also asked me if Frost buys back stock. And I said “no”. Here is a passage from the company’s most recent 10-K proving it does buy back stock:

“On October 24, 2017, our board of directors authorized a $150.0 million stock repurchase program, allowing us to repurchase shares of our common stock over a two-year period from time to time at various prices in the open market or through private transactions. No shares were repurchased under this plan during 2017. Under prior plans, we repurchased 1,134,966 shares at a total cost of $100.0 million during 2017 and 1,485,493 shares at a total cost of $100.0 million during 2015.”

So, why did I say Frost doesn’t buy back stock. Here is the company’s shares outstanding at the end of each of the last 5 years.

2013: 61.1 million

2014: 63.0 million

2015: 63.5 million

2016: 63.0 million

2017: 64.7 million

To me, if there’s a trend there it’s a trend toward a rising share count rather than a falling share count. Compare the above trend to the share count at Omnicom (OMC), a company I often use as an example of a constant buyer back of its own shares.

2012: 262.0 million

2013: 257.6 million

2014: 246.7 million

2015: 239.7 million

2016: 234.7 million

To me, Omnicom buys back stock and Frost doesn’t. So, in my analysis of Frost I never really thought about stock buybacks. I did, however, (as I mentioned in the podcast) think a little about Frost using its own shares to acquire other Texas banks.

Finally, let’s talk about valuation in terms of being “roughly right” or not. Near the end of the podcast, you can hear me say that “at a price of about $100 a share, Frost stock is probably trading for about two-thirds of what it’s worth”.

Someone who listened to that podcast about Frost and had done their own updated valuation of the company using the same methodology I did in my original report pointed out that I seemed to be saying Frost now has an intrinsic value of $150 a share – when the numbers (using my own methods) say it has an intrinsic value of $185.

Here is the email in full:

While listening to the Frost podcast you and Andrew put up, I was surprised to hear you say you thought the stock's value was around $150 per share (what you said was that at $100 it is trading at 2/3 of value). 

Did you change your valuation approach from the time you did your Singular Diligence research report? At the time you valued CFR by taking 20x after-tax earnings, which was calculated as 2.65% x earnings asset x .65. 

If I apply that formula now, I take the recent earning assets 29,574 x 0.0265 and get 784 of pre-tax owner earnings. Since the federal tax rate was cut to 21, I think it makes sense to use no higher than a 25% tax rate, which gives us 588 of after tax earnings. Multiply that x 20 and we get a business value of 11,760. Divide that by 63.7m shares and we get a value of $185. 

While not radically different than $150, it is almost 25% higher, which represents a materially higher margin of safety relative to the current price the stock is trading.

Did your change your valuation approach, and if so, what was your thought process?”


The answer is that I didn’t change my valuation approach. I was just saying that – looking at the way I look at Frost – you’re still able to get a dollar for 65 cents if you pay the current stock price. Why did I say it this way?

Well, there are several reasons. One, we record the podcast episodes before we air them (obviously). Sometimes, we might be recording an episode a couple weeks before the episode airs. When you’re listening the podcast, a stock I thought was trading at $100 could now be at $90 or $100.

Then we have the discount rate issue. So, when we appraise a stock we have to pick some way of deciding what we mean when we say it’s “worth” such and such an intrinsic value. My approach to being “roughly right” rather than precisely wrong with bank stocks is to assume you always have to value a bank as if the Fed Funds Rate is now normal. When I first appraised Frost, the Fed Funds Rate was between 0% and 0.25%. A “normal” Fed Funds Rate is – in my view – more like 3% to 4%. Now, obviously, if the Fed Funds Rate is 0% when I’m writing the report it’s not going to be 3% or 4% next year.

If rates rose shockingly fast, a stock I bought at $47 a share would probably trade at $200 a share in less than 5 years. My return in the stock would be something like 32% a year.

I knew that was unlikely. If rates rose incredibly slowly, my annual return in the stock could get dragged down to something like 16% a year (if it took rates almost 10 years to get to where I thought they should be).

And then you have the issue that rates might never rise. If rates never rose – the Fed Funds Rate literally stayed at 0.25% or less from the time I first wrote this report – it’s possible my return in the stock could be as low as 8% a year.

So, the timing of rate increases could alter my annual return expectation from 32% a year (fast), to 16% a year (slow), down to 8% a year (never).

How do you discount for this kind of timing issue?

Honestly, it’s almost impossible. For example, earning 8% a year sounds less than stellar. However, what would be a normal “discount” rate if the Fed Funds Rate literally never rose above 0.25% a year. This stock would – in the bad scenario – be returning 7.75% more than idle funds left at the Fed. In a normal year, that’s a perfectly decent return for a stock. Meanwhile, if the Fed Funds Rate rose quickly and ever got as high as 4% a year – well, that could mean that yields on all sorts of securities were a lot more attractive. Stock prices might be falling (earnings yields rising). The correct discount rate to use in adjusting my appraisal value for the stock would then be pretty high.

I don’t do a discounted cash flow analysis when appraising a stock. But, I am aware – and adjusting for – the fact that some economic assets (like a bank’s deposits) won’t be earning a lot of money till later years. Deposits have to be worth somewhat less to the extent you can’t lend them out at decent rates right now.

But, how much less?

It’s very hard to fix these problems. I didn’t try. Instead, I calculated my appraisal value using a normal Fed Funds Rate and then I did math on how much you’d have in capital gains if it took 5 years, 10 years, or 15 years to reach that Fed Funds Rate. I also asked – what if the Fed Funds Rate never rises? The conclusion I came to was that at $47 a share, the stock seemed likely to be priced to return no less than 8% a year even if rates never rose and yet more than 20% a year if rates rose quickly. I didn’t even bother calculating exactly how much more than 20% a year you’d make under the “good” scenario as far as timing. If you calculate a stock might return more than 20% a year while you own it – don’t worry about how much more than 20% a year it might return, instead worry about how realistic that scenario really is. I didn’t think the 20% a year return scenario was any less realistic than the 8% a year return scenario (the one where the Fed Funds Rate never rises). That was good enough for me. A stock where you’re as likely to make more than 20% a year as less than 8% a year is a good bet.

This is what I mean about being roughly right.

Assumptions about the Fed Funds Rate’s eventual “normal” level and how quick it would get there were important. So, are assumptions about the correct discount rate – basically, what your opportunity cost is in the stock. For example, if I really thought my returns would be in the 8% a year to 20% a year range, the next question would be how well did I expect the stock market to do while I held the stock. That’s a quick – though inaccurate, in my case – way of ballparking your opportunity cost. For me, I didn’t expect the market to do even 8% a year long-term. So, the stock looked pretty good on that basis.

But, there are tons of other assumptions that went into the original appraisal price I put in that report. Several of these assumptions are wrong – and a few are intentionally wrong. Take the tax rate. Frost hasn’t always paid a 35% tax rate even before the recent tax bill was passed. Like that email said, the bank probably won’t pay much more than a 25% tax rate in the future. My appraisal used a 35% tax rate even when I knew the bank was more likely to pay less than 35% than more than 35%. I intentionally used too high a tax rate.

I also used questionable – but, I hope conservative – approaches to charge-offs. Frost has recently had maybe something like 45% of interest-bearing assets in loans and 45% in bonds and then another 10% or so parked in some kind of cash (like with the Fed). That was about what the situation looked like when I was calculating Frost’ earning power.

Here’s the thing: I applied the bank’s historical charge-off rates on loans to all of its earning assets. If you read the Frost report carefully, you can see I applied a 0.48% charge-off rate to all “earning assets”. Well, Frost might normally have 45% in loans with a charge-off rate of 0.48%. But, then it might have 30% in Texas state bonds and 15% in U.S. Treasury bonds and maybe 10% in some kind of cash. It’s reasonable to assume the bank will lose $1 a year for every $200 it lends out. But, is it reasonable to assume you will lose $1 a year for every $200 worth of Texas State obligations, U.S. Federal obligations, or money parked at the Fed you have?

Probably not. But, the stock still looked attractive even if you did make those assumptions. And, I could never be sure what the mix of these assets would be in “normal” times. In a huge boom, Frost might eventually lend out 70% of its deposits instead of more like 40%. It’s never going to lend out 100%. But, if you penalize the bank with charge-offs as if it is lending out 100% and the investment case still holds up, that’s a good sign.

So, there we were conservative. I’d say we were wrong to do it that way. But, it would be pretty involved explaining just how wrong we were. The important part is that I know the report erred on the side of conservatism in that case. We assumed more stuff would be subject to charge-offs than really will be.

What about the charge-off rate assumption itself?

This one is super tricky. And it points out why you want to try to be “roughly right” in the sense of a little conservative but still reasonable – instead of precisely wrong. The typical methods for coming up with a charge-off rate assumption would risk not looking far enough back in time.

So, in the 20 years prior to when I wrote the report on Frost the bank averaged a charge-off rate of 0.27%. That’s the charge-off rate from 1994-2014. It would seem reasonable to use that charge-off rate. But, I wasn’t so sure. Yes, 20 years is a long-term average. But, there’s two problems there. One, you only had 2 recessions in those 20 years. One was deep (2008) but the other (2001) was about as shallow a recession as you’re ever going to get. The financial crisis (in 2008) was terrible nationwide. But, it wasn’t that bad for Texas. Texas had a much worse time of it in the late 1980s through the early 1990s. Those problems had been resolved by 1994. This meant that Texas banks – not just Frost – had very low charge-off rates from 1994-2014. I thought this might just be lucky. I could certainly use longer-term FDIC data on all banks around the country to come up with a much higher charge-off rate assumption. But, I knew from studying other banks that this was dishonest. Different banks have very different charge-off rates because of the categories of loans they make and also just how conservative the bank is. For example, from 1996-2014, I had data showing Frost’s charge-off rates were much lower than other banks in the same state. So, what did I do?

I used the mean – rather than median – charge-off rate for as many years as I had data. This allowed me to include the problems Texas banks had in the late 1980s. However, it meant I was assuming a 0.48% charge-off rate when the median charge-off rate for the last 20 years was just 0.23%.

Normally, you want to be careful about that. I don’t like using numbers where extending the length of the series from 20 years to 26 years or using the mean instead of the median makes a big difference in the result you get. Here, by using a 26-year mean instead of a 20-year median we got a charge-off rate more than double the one we’d normally use. Very often, I don’t even have data going back more than 20 years on a stock. Here, I intentionally extended the series further back into the past because I wanted to include the worst financial crisis in Texas’s history (as well as the 2008 financial crisis). Once the series I was averaging included both the late 1980s and 2008, I felt better about the assumption. Now, I was including 3 recessions over 26 years.

But, what’s the right charge-off rate? Is it 0.23%? Is it 0.48%? And then what do you apply that charge-off rate to. Obviously, you should apply it to loans. But, loans might be 35% of Frost’s earning assets at the bottom of a bust and 70% of Frost’s earning assets at the top of a boom. So, do you multiple 0.48% by 0.35 or by 0.70 or by some number in between.

I try to make reasonable assumptions. But, when you get to the point where I’m not sure which of a couple reasonable alternatives to choose for my assumption – I just use the more conservative one.

So, I didn’t worry about whether a 0.48% charge-off rate or a 0.23% charge-off rate was right. I used 0.48%. And then I didn’t worry about whether Frost would be lending out 35% or 70% of its deposits. I just pretended it would lend out everything (for the sake of penalizing the bank’s earning for charge-offs). And finally, I didn’t know if the bank’s tax rate would be 25%, 30%, 35%, etc. With the recent tax bill, we now know 25% is unlikely to be a low assumption. But, a couple years ago, it seemed like it might be. I knew 35% wouldn’t be too low an assumption – so I used 35%.

Your final appraisal value is going to be the sum product of a lot of these smaller assumptions you make. So, if each time you face a choice between the more aggressive and the more conservative assumption you always pick the more conservative assumption, your final appraisal value will have the cumulative conservatism of all these little assumptions built into it.

Is that accurate?


But, it’s better than risking the possibility that your final appraisal value might be the sum product of a lot overly aggressive assumptions.

So, it’s fine to be roughly right. But, always try to be conservatively roughly right.

You can learn more about Geoff Gannon by emailing him:, following him on Twitter: @GeoffGannon, or listening to his podcast. His stock specific write-ups appear on a subscriber supported website: Focused Compounding.