Berkshire and the power of reputation

Good morning. More details are emerging about the congressional Republican’s proposed budget. The House version, in its current form, adds more to the deficit than many investors had expected, and makes bigger cuts to social programmes than Democrats had feared. 10-year Treasury yields climbed 7 basis points yesterday. Email us: [email protected], [email protected] and [email protected]. 

More on Berkshire’s outperformance

Last week we wrote a couple of pieces about the remarkable long-term performance of shares in Berkshire Hathaway. We emphasised that, as Andrea Frazzini and his colleagues at AQR have argued, a lot of Berkshire’s strong performance is explained by exposure to standard performance “factors” such as value, quality and low volatility — mixed in with some leverage and a lot of years.

Edward Finley of Arrow Wealth Advisory, a friend of Unhedged, did his own regression analysis between Berkshire’s A-share performance and six canonical factors from 1996 through to the start of 2025. That’s shorter than Frazzini’s study period (1976-2016) but more up to date. What Finley found is summarised in the table below. It is full of horrible statistical terms, but I’ll walk you through it (my own stats knowledge is thin, so actual experts can email to correct me if I mess it up):

The blue row across the top gives the factors: beta (low-volatility stocks), quality (high profitability), momentum, value (low price/book ratio), and size (small stocks versus large). The market “factor” is just the performance of the whole market. The next column down shows Berkshire’s regression coefficient for each factor. Look, for example, at the beta factor at the left. The regression coefficient of 0.17 means that for every 1 per cent move in the Beta factor in a given month, Berkshire moves 0.17 per cent. The next column, the T-stat, is a measure of the statistical significance of the relationship between Berkshire’s performance and the factor’s. A number above two generally indicates significance. The final column, annual return, just shows how a market neutral long-short portfolio based on that factor (say, buy small stocks and sell big stocks) performed in an average year.

So what mattered for Berkshire over the period, according to Finley’s work, was exposure to the low beta, value and (in a negative sense) size. However, simple market exposure was the most important. It is interesting that quality was not that influential over the study period. 

Now note the cell to the right, the “annual intercept”: one percentage point. That is the annual excess return, over and above a benchmark, that is not explained by factor exposure. Berkshire’s general outperformance (using the Wiltshire US large-cap index as the benchmark) was 3.3 percentage points a year. So a little less than a third of that is not explained by factor exposures. So what does explain it? Skill? Leverage? Skilful use of leverage?

That is not the most interesting bit. Finley broke down the performance review into a series of five year periods, to see how the factor exposures changed. Here is his chart:

Notice the shape: the exposures fan out early and late, but all get mushed together around zero — that is, they do not explain as much of Berkshire’s performance — in the middle years from 2001-2011.

What is interesting about that period is it starts and ends with a market crisis: the dotcom mess and the housing bubble bursting. And, of course, during the period covering both crises, less exposure to market factors (particularly the market factor) is what you’d want as an investor. How did Berkshire achieve this? Changes in its asset portfolio seems like an unlikely candidate, as Buffett moves slowly there. Maybe cash exposure played a role; as we have written, the portion of the Berkshire portfolio in cash was high from 2003 to 2007.

But here is another possibility: that it wasn’t Berkshire but the market’s attitude towards Berkshire that changed. Maybe during the turbulent period, people bought Berkshire not because of its particular economic exposures, but simply because they believed Berkshire and Buffett represented safety and prudence. If so, that would confirm Unhedged’s view that Buffett’s reputation, image and skills as a showman are a major performance factor, over and above Berkshire’s purely financial characteristics.

More on sovereign credit default swaps

After yesterday’s letter about market reactions to a US default “near miss”, a couple of readers wrote in with an explanation for why Treasury credit default swap prices were so much higher in 2023 than in 2011 and 2013:

Why did default insurance get so much more expensive in 2023? Interest rates, and a quirk in sovereign CDS contracts. Brij Khurana at Wellington Management, fixed income consigliere, explains:

If interest rates move higher, longer duration Treasury bond prices move a lot lower. A 1-year bond matures after 1 year, so it typically trades close to par, even when interest rates rise . . . Sovereign [CDS] contracts are [generally] “physically settled”, meaning the buyer of protection delivers eligible bonds to the seller of protection if there is a default. The buyer of protection only needs to deliver the nominal [“face”] value of the bonds it insured, and is, thus, incentivised to deliver the lowest dollar priced bond. Therefore, in a default, the buyer of protection is likely to deliver a longer duration Treasury that was issued with a very low coupon a few years ago, whose price has fallen significantly because of the higher interest rate environment we are currently in.

Think of it this way: I (Aiden) buy protection from Rob on $100mn of Treasuries. The US defaults. He owes me $100mn in cash; I owe him a nominal value of $100mn of Treasuries, not $100mn market value. Of course, I satisfy this with my longer duration Treasuries that I bought a few years ago when interest rates were zero, which have a face value of $100mn but are now worth much less in dollar terms, because rates have risen. I skip away happily. Rob, who is, well, unhedged, goes bankrupt.

The price of the CDS takes this into account. It is determined in part by the perceived likelihood of a default (numerator) and in part by the present value of expected cash flows if there is a credit event (denominator). With high interest rates, the present value in the denominator is much lower, because the buyer can deliver the Treasuries at the cheapest price. The lower denominator mechanically drives the price charged on a CDS higher — as in 2023.

CDS prices are therefore an imperfect gauge of the risk of a default. We would argue that the odds of a default were much higher in 2011, when interest rates were zero, than in 2023, even though CDS prices went much higher in 2023. And there are a couple of other issues with using them, too. As Antulio Bomfim at Northern Trust noted to Unhedged, “the market for credit default swaps on highly rated sovereigns like the US is very thin, and a small number of transactions can move markets considerably . . . [they are a better indicator] if you are looking at emerging market economies, where there is a more meaningful probability of default”. Also, US sovereign CDS contracts are often denominated in euros, meaning that changes in CDS prices also reflect expectations for dollar devaluation in the event of a default.

In the absence of other good indicators of US budget stress, we’ll still use CDS. But we’ll use them with a grain of salt.

(Reiter)

One good read

Got milk?