dartmonkey

I think the consolidated non-insurance earnings have historically been small enough that they been lumped in with corporate overhead and interest and just disappear:

I'll answer the questions one by one, if I get to them all! But just to be clear, when I exclude the unrealized gains on the bond portfolio (which average out to about zero anyways), I get an average 13.6% on equity investments over the past 6 years, not too different from the initial estimate of 13.2%. Last year's 17% was maybe a bit of an outlier. There is one additional step in the RBC analysis of iinsurance companies, where they look at how much of an impact those gains have on ROE. This obviously depends on how much equity investments are in comparison with the company's book value (also called equity, confusingly...) So for instance last year's 17% return on their $21,139.9 worth of non-fixed income investments was $3,584.2m, and that, in relation to their average book value last year, $24,621.2, gives them a 14.1% effect on ROE, i.e. that 14.1% gets added to the effect of underwriting on equity (7.4%; i.e. the 7% they make on underwriting, times the 1.07x leverage they get from having 1.07x as much net premiums written as average equity), and so forth. The total ROE I get for last year after tax would be 22.7%, which seems a little high. And assuming a drop of their combined ratio to 0.94 next year (along with stagnant net premiums written, with the softer insurance market) and an average 13.6% return on non-FI investments (not quite as good as last year's 17%, although we are off to a great start), I would get an 18.1% ROE this year, instead of 22.7%.

Yeah, I had thought it was more of a trading thing but I guess they add even the unrealized gains every year and it adds a lot of noise. The average in the last 6 years is close to 0 - $19m in fact, so I just excluded it, since adding it to interest and dividends returns from bonds would mean the bond returns are all over the map, whereas the bond yield is the one thiing that is pretty consistent from year to year. I think it is reasonable to expect that on average the market gains and losses will be about zero, which is maybe slightly on the conservative side, since they do trade in and out occasionally based on macro guesses, but for the moment I'll just leave those gains or losses on bonds out of the model. As expected, it makes the equity returns less variable:

If there is a no-uptick rule, this only means the previous trade has to be at a lower price, it doesn't mean you can't repurchase shares when the general trend is upwards, as it has been for 3.5 years (I was going to say 5-6 years, but the share price in October 2022 was about $450, almost exactly the same price as in October 2014...

Do you include gains/losses in the bond portfolio in the fixed income return? No, I kept it with non-FI investing gains. I thought about putting it with fixed income, since it is from trading fixed income securities, even if it is clearly not part of the 5.0% fixed income return. Perhaps I will switch it, since it is clearly reported in the annual statements and it doesn’t really make sense to have bond sale gains in the numerator compared to non-fixed income investment assets in the denominator. What do you think?

I think 9.3% returns on total iinvestments is probably too optimistic - the average for the last 6 years is 6.1%. The last 6 years show how variable this is, from year to year, compared to bond returns: I calculate the last line by backing out the fixed income return from the overall return, and calculating the return on equity investments (which includes associates and consolidated non-insurance companies.) As you can see, the fixed income return was low in 2020-2022 and has recovered to about 5% now. But the equity investments are, as you would expect, all over the place. For instance, in 2021, returns were great, because they booked a huge gain in Digit Insurance. 2022 was terrible because it was a terrible year for stocks, with the S&P down 19%. And so on. So I think extrapolating last year's great 9.3% return, and particularly its 18.8% equity investment component, is too optimistic. But even so, with the leverage in place, low corporate costs and low interest costs, and great investments like Eurobank and Fairfax India trading at low valuations, I am pretty confident in their prospects. I think we already have My model is still embryonic, but if I assign 13.2% to equity returns (the 6-year average) instead of 18.8%, with fixed income returns unchanged at 5.0%, we should still have overall returns of 7.6%, a bit better than the 6-year average of 6.1%. And with already 2 major investment successes for 2026 (Eurolife - $350m, Poseidon - $866m, total of $1216m), after only 2 1/2 months we are well on our way to repeat something similar to last year's $3151m in total investment gains (9.3%) than the 6-year average.

That's an interesting way of looking at it. If we speculate that Fairfax's offer might have been the best offered, but not as high as the VWAP in 52 weeks or, more likely, not as high as the highest share price in the past 26 weeks. The bidding process has been going on for years, so if Fairfax offered INR100 a share March 2024, when the shares were trading for INR85, then the recent share price of up to INR115 would disqualify their offer. Now that shares are back down to INR74, their offer might appear good again, but if the above rule is correct, we might need to wait another 6 months before that INR115 price gets to be 6 months in the past. Which maybe also means that they won't be focussing on repurchasing their own shares after all, but it may mean all the time and effort they spent trying to get this bank is not yet completely lost.

Another more positive way of seeing this is that even Horne seems to agree that the value of FFH is increasing rapidly. After all, in February 2023, he said the shares were worth C$730, and now 3 years later he says they're worth C$1730. In other words, they have gotten 137% more valuable in 3 years, or 33% annualized. If fair value growth keeps up this pace, then shares will be fairly valued at today's price (C$2267), even by Horne's exacting standards, in less than a year. Hmmm, maybe that's why investors think they're worth $2267? I guess even a loser company with no competitive advantages is worth something, when it's growing at 33% a year, right?

https://global.morningstar.com/fr-ca/actions/rsultats-de-fairfax-financial-holdings TL;DR French: Essentially he acknowledges that 2025 was a good year but it won't be repeated, and he maintains his C$1730 target (down 24% from today's price) for this firm with no competitive advantages, in an environment with softening insurance prices and because despite a good year, Fairfax's investment returns over the long term have been irregular.

Not sure I completely understand this logic (but I am not an accountant either). I also am not an accountant, and also do not understand why you wouldn't just use diluted share count for everything, whether it is earnings per share or book value per share. But I can confirm that Yahoo Finance uses diluted shares in calculating earnings per share and effective shares when calculating book value per share, supporting Viking's idea of doing it that way.

Buying a lot of shares under book would actually increase book value per share. You are right, thank you for the correction. The demoninator would increase more than the numerator, since buying back $1 worth of book value only costs $0.78. It wouldn't have a huge impact - buying back $100m worth of shares would take the book value per share from $22.94 to $23.17, so I got the direction wrong. I was probably subconsciously thinking of P/B, which would go down, from 22.94/17.90= 0.78 to 23.17/17.90 = 0.77.

I guess that's why the share price is up today in a down market - the market believes (probably correctly) that the capital they had reserved for buying IDBI will now go to repurchasing shares. Thinking about what this would do to the book value, given that FIH is currently trading at 0.75x BV, buying a significant number of shares beneath book value would actually decrease the book value per share and postpone incentive fees paid to Fairfax. They hav bought back share for $29m, $127m, $36m, $37m, $8m and $10m in the last six years. They only had $8m cash on the balance sheet at year end, despite the sale of Saraushtra for $75m which closed in November, so it's not clear where they would get the cash for significant levels of buybacks. The BIAL IPO would solve that problem, but then again, if that happens, the share price is likely to rise a lot, so it's not clear to me how they could take advantage of this price without asset sales.

Marked to market (own<20%) dividends would be included in dividend income. When they use equity or consolidated accounting, it would must reduce the carrying/book value. OK, that makes sense. It does make it a bit more difficult to model, though, since if you took the marked to market equity investments at market value, and then you also included dividends and interest as income, you would be double counting. I guess the solution is to NOT subtract out the marked to market equity portfolio, but to value it as a multiple of interest and dividend, for the part of earnings that are paid out, and as investment gains, which presumably reflect the retained earnings, although in a much more volatile way than if we knew exactly how much dividends came from the marked to market equity portfolio. I really like the RBC model you have posted here on multiple occasions, calculating everything as a multiple of equity, but the one thing I would like to do better is the investment yield, which they have as 7.0% (presumably for 2026?). With the 2.8x leverage they estimate, that 7.0% number is doing a lot of heavy lifting, since most of the final 25.5% in pre-tax operating ROE comes from the 19.9% investment return. It would be great to break up that 7.0% into its fixed income component which we know pretty well (I think they have forecast 5.0% for 2026) and the equity component that we need a fairly wide range for. Looking at 2024, we had $1067 investment gains on $23.0b in equity, so 4.6% (return on total equity, not return on equity investments, as everything in these calculations is based on Fairfax's equity i.e. book value), and with 0.71 investment leverage (to use the RBC term; this is in addition to the 2.00x fixed income leverage), this provided a 3.3% boost to total investment ROE, in addition to fixed income's 10.4%, for a total of 13.7%. In 2025, the equivalent numbers are $3151m investment gain on $26.3b total equity, or 12.0%, with 0.67x investment leverage, for a 8.0% contribution to the investment ROE, on top of 9.3% from fixed income, for a total of 17.3%. You posited a 5-15% range for equity investment returns, and I think that is fair; my number is much smaller just because it is as a percentage of total equity (including fixed income assets). Looking at investment returns as a percentage of equity investments, these were 6.5% in 2024 and 17.9% in 2025, within your range. My next step in doing this is to go back a few more years and see what we have in investment gains as a percentage of equity. Or maybe it makes more sense to use equity investment gains as a percentage of equity investments, and then convert that to equity investment gains as a percentage of all equity, at the last step, since the proportion of equity that is in equity investments will be changing over time. Minor technical question: do you think RBC is using previous year end equity as the denominator, or some sort of average equity, like the sum of beginning of year and end of year equity?

Yes, it's a hard element to model. We've got (i) underwriting, (ii) dividend and interest income, (iii) the equity portfolio (I never know whether dividends from the equity holdings are included in (ii), does anyone know?), and (iv) earnings from wholly owned non-insurance subs like Recipe or Sleep Country. But what about (v) the occasional rabbit pulled out of the hat when bonds or stocks or an associate equity holding like half of Poseidon are sold? Clearly the $870m Poseidon capital gain and the $825 increase in market value of the stake that is retained are not adequately accounted for by just taking Fairfax's share of Poseidon's earning, and although the timing is not predictable, this is pretty clearly another source of income for the company. Last year, Fairfax had $3151m in pre-tax investment income, a big part of their $6440 in pre-tax income; in 2024, investment income was $1067m out of total pre-tax income of $3875m. Many people I have talked to about this investment way, "well, yes, but you can't count on having that kind of gains every year." Obviously this is true, but we already have another $866 from Poseidon and $350m from Eurolife for 2026, and the year is young. For a $37b market cap company like Fairfax, trading at less than 6x last year's pre-tax net income, if you don't model in a billion or two of gains on sale of investments every year, you get the kind of nonsensical valuation the market is giving Fairfax right now.

Yes, me too. If we have to have the low share price (around 8 times earnings), at least it allows for really moving the needle with share repurchases. They won't have the cash from the Eurolife and Poseidon sales until Q2, but now that they know they will have it soon, they can be much more aggressive with the buybacks, and still have plenty of cash to repurchase whatever insurance stakes they want to. (I don't know what the time limits on these shares is, I haven't been able to find the Offer Documents from the 2021 sale of 9.9% of Odyssey to OMERS and CPP, along with whatever other minority shares they have the right to repurchase, and the press release just said they would be able to buy back the Odyssey stakes, but not until when: "After closing, Fairfax will retain the flexibility to repurchase the interests of OMERS and CPPIB Credit Investments in Odyssey Group over time." With what looks like the winding down of the hard insurance market, there should be lots of cash going from insurance subsidiaries to the holding company anyways, but with an extra billion, along with a share price that has gone nowhere since May 2025, and 2.2m shares authorized in the current normal course issuer bid, we might see the repurchase numbers move a lot higher.

I believe the rules for an associate holding, where Fairfax has between 20% and 50%, mean that it is neither consolidated (as it would be if it owned more than 50%) nor marked to market (as it would be if they owned less than 20%.) Since thery are going from 45% ownership to 22%, the same treatment should apply, and the fact that they just sold for twice their carrying value means that fair value goes up but carrying value stays the same.

Yes, it seems that way, about $890m pre-tax, plus an increase of almost as much in the difference between carrying value and market value of the other half.

Yes and no. Eurobank's change in MV is not captured, but Eurobank does make a pretty significant contribution to EPS and ROE. As an associate holding, Eurobank contributes Fairfax's share of its earnings, so in 2025 Fairfax will have booked about 32% of Eurobank's E1349m in earnings, about US$510m (2024Q4 to 2025Q3, with one quarter delay). The fact that Fairfax's equity stake is only carried at $2750m means that ROE will be doubly juiced by Eurobank's contribution, and those $25/share earnings made a pretty big contribution to Fairfax's $214/share overall earnings. Someone correct me if I have miscalculated, but I think that's right.

I don't think we really disagree here at all. Fairfax has a bunch of hats, and every few years, we find out that there was a rabbit in one of them, or, more exactly, that the rabbit that we knew was in there was a lot bigger than we thought. First Capital Insurance, Pet Insurance, Stelco and more recently Eurolife come to mind, where the value they obtained was substantially better than what most of us expected. All I am saying is that if you're trying to model future earnings, I think there are enough of these sporadic gains that you have to at least pencil in that there are going to be more of them. The operating earnings from underwriting, fixed income and non-insurance subsidiaries are the more predictable parts of Fairfax's business, but buying cheap and selling dear, mostly for companies but also sometimes bonds, is also a part of their modus operandi, and just because we don't get a rabbit every year doesn't mean that they are not growing inside those opaque hats. By the way, do we know when the annual report gets released? I don't see any press releases in previous years for when it comes out, but the date on last year's report was (Friday) March 7 so I presume it's in a week or two?

This is the big question. Clearly, we won't be getting $3.15b in investment gains every year, like in 2025, and that is surely why Mr Market is not particularly impressed with overall earnings that put Fairfax at about 8x 2025 earnings. But maybe $1.5b is too pessimistic? The big investment gains were from Orla and the return swaps. Fairfax shares are down about $200 from their end of year price of $1908, so that could be a headwind, but Orla is up massively already in just 2 months, with the share price up almost as much in January and February (from $18.46 to $28.74) as it was in all of 2025 (from $7.96 to $18.46.) Yes, they sold a quarter of the position at $17.64, but the other 3/4 is looking at huge gains, roughly C$700m so far. Good gains already from UA, and their KW position may be counted as an investment gain when they take it private at a premium. Throw in one good asset sale (there aren't many years when they don't pull some kind of rabbit out of the hat), and with a little help from the return swaps, we could easily be closer to $3b than $1.5b.

It is also conservative because it is not the standard way of doing the return calculation, which would be 8.5b/65b = 12.3%, not 11.8%, since the increase in value of the portfolio was not because new money was injected into the portfolio but just because of the accumulation from returns on the initial amount. Yes, it would be interesting to see the same table but with an amount for debt and float, to see if they are using more or less leverage. As for how the debt should be treated, the RBC analysis that SafetyinNumbers has mentioned a few times (for instance, Nov. 13) seems like a good way of accounting for the interest. Fairfax's debt allows them to increase investment returns (as a percentage of equity), but there is an interest cost that has to be deducted somewhere, along with overall operating costs and taxes (it was 0.3x equity at that point, with a 6.8% annual cost, so it added a -1.8% drag on the +19.9% equity returns in 2024.

I agree with all this but not only is there a survival advantage, there’s also a return advantage. Which is better, investing in 2 companies that each have one business, uncorrelated to the business of the other company, with each of those businesses having a 15% ROE, or investing in one company which houses those 2 uncorrelated businesses under the same roof? I would say the second, because not only can they survive a hit to one of the businesses, but they can shift capital between the two, as threats/opportunities arise. This is the benefit that investors like Buffett or Watsa have over managers of companies that are mostly in one business, even a wonderful business like Costco or Starbucks or Cisco, that they always have to invest in, in good times or bad.

Agree 100%. For any parameter in an analysis, there will be a range of possible values, going from pessimistic to optimistic. You should not build pessimistic assumptions into your analysis, you should use some sort of central, expected value estimate. That would probably lead you to invest in FFH as a good investment right now. And then, after you've made the investment, if you want to be pessimistic (something might always go wrong) and have low expectations about how your investment will perform, so that you are more likely to be happily surprised by the outcome, maybe this is a sensible way to lead your life. But it is crazy to put overly pessimistic assumptions into any analysis - you would never buy anything, that way.

I was wondering the same thing so I calculated it, 5 years from 2020-12-31 to 2025-12-31 and 5 years from 2021-02-19 to 2026-02-19 (at $1710, right now). Annualized, that gives 43.7% for the first comparison, and 35.4% for the second, just a few weeks later. In the second case, it is considerably lower because the shares jumped up from 311.56 (adjusted for dividends) to 382.06 in those first few weeks of 2021, and have dropped down from 1908.02 at the end of 2025 to 1710 now, lopping off a few percentages points at each end. So a 40% annual return for the last 5 years is a pretty good estimate, and even if we only get half that for the next five years we will still double the share price and get our the prize for winning the bet (beer at the Keg, if Fairfax still owns it by then?)

My calculator says 2^(1/5)=14.9% so the bet is over/under 15% return for 5 years, and I would be a hard 'over' on that one. My only concern would be that, if there is a big market correction as part of the next 5 years, then Fairfax would be drawn down like the others, and would likely substantially outperform the market (maybe by 5% annually) but not hit the 100% return in 5 years. So to hedge the first bet, I would want a similar bet that the market will be less than 10%/year ahead, i.e. up less than 1.1^5-1=61%. IOW Fairfax makes 100% if the S&P makes 61%, or Fairfax outperforms the S&P by at least 5% p.a.