A Dual-Class Arbitrage in Berkshire Stock?

Includes: BRK.A, BRK.B
by: Andy Kern

Hat tip to Ray for pointing this out to me. Last week, BRK.B was trading at a level significantly lower than 1/30 of BRK.A. In fact the spread represented 7% of the value of BRK.B.

This was a decent arbitrage opportunity, as the two BRK classes have historically been pretty good about reverting to parity fairly quickly. Indeed, within two days the spread had already narrowed to less than 2% (as a % of BRK.B). Below is a chart of the difference over the last year.

It is interesting to see how the spread became more erratic as market volatility increased in the fall. Note that the spread rarely dips into the negative. This is by construction, since BRK.A is convertible in to BRK.B but not vice versa. This means that a mispricing favoring the B shares could immediately be corrected by arbitrageurs converting their A shares to B. But the opposite cannot be done, leaving the arbitrageur at the mercy of the market to correct the mispricing.

In the rare cases when the spread does bounce up to the 7% range, this makes for an opportunity for a quick profit for those vigilant enough and quick-fingered enough. The recent bounce lasted at least 24 hours from what I could tell, which is plenty of time to implement the trade.

A few other notes: BRK is a great stock to play the dual-class arbitrage with because it is so liquid and small investors should have no trouble finding A shares available for short. Of course, you need at least $78,000 in capital so I guess the really small investor is out of luck.

Moreover, the differential in voting rights between the A shares and B shares is inconsequential in the case of Berkshire since Buffett et al. control the voting power of the company anyhow. Thus we can expect the 1:30 ratio to hold irrespective of the differential voting rights of the two shares.

I suppose measuring the spread as a % of BRK.A would be more appropriate, because a true arbitrage strategy would short BRK.A and long BRK.B, holding until the difference was zero.