Dual Share Class Relative Value

Thesis (edge)

Some companies issue more than one class of shares. Typically the classes have the same claim on the company's profits, but different voting power: one class lets you vote, another does not, or one gets ten votes and the other gets one. Founders keep the powerful class, the public gets the rest.

Economically, if both classes have the same claim on dividends and on the value of the business, they should be worth nearly the same. In practice they usually trade at a small, persistent gap, and that gap moves around.

The trade is to treat the ratio between the two prices as something that has a normal range. When it stretches unusually far outside that range, buy the cheap class and short the expensive one, betting the ratio comes back.

This is the cleanest possible pair. Same company, same earnings, same management, same everything except a governance detail. There is no question about whether the two things are related.

But note the crucial word in that sentence: nearly identical. The trade lives or dies on what the difference actually is, and the difference is real.

Where it works (regimes)

• Works: when the ratio has a genuine history of moving within a range and returning, and when the gap widens because of a temporary liquidity event, an index flow, or a burst of one-sided demand for one class.
• Works better: when there is a plausible catalyst on the horizon, such as a proposal to unify the classes, which gives you a reason to expect convergence rather than merely hoping for it.
• Fails: when the gap widens for a reason. If a takeover becomes likely, the voting class can suddenly be worth far more, because votes matter in a control fight. That is not a mispricing to be faded, that is the market correctly pricing an option you were short.
• Fails: when one class is included in a major index and the other is not. Index demand can support a persistent premium indefinitely, and there is no force pushing it back.
• Fails quietly: when the ratio simply drifts to a new level and stays there. There is no expiry date, no conversion mechanism, and no arbitrageur who can force the two prices together. Unlike a bond that must mature at par, these two shares can disagree forever.

Signals

• The ratio. Divide the price of one class by the price of the other. That is the whole signal.
• The normal range. Chart the ratio over years. Most such pairs trade in a band. The question is whether today's ratio is unusual by the standards of that band. Note carefully: the fair value is whatever the historical range says, not necessarily parity. A class with no votes trading at a persistent three percent discount is not mispriced, that is its level.
• Stability check. Has the ratio actually reverted historically, or has it trended? A ratio that has drifted steadily in one direction for five years is not a mean-reverting spread, and you should not trade it as one.
• Catalyst awareness. Any news that could change the relative value of the classes, especially anything to do with control, voting or index membership, should override the statistical signal.

Portfolio construction

• Long the cheap class, short the expensive class, matched by value so you are neutral to the company's overall performance.
• Diversify. This is essential. Any single ratio can be dislocated for years. A basket of a dozen such positions across different companies is a strategy. One large position is a bet.
• Size small. The apparent risk is tiny, since both legs are the same company. That is exactly the illusion that leads people to lever it up. The historical record of relative value trades that looked riskless and were levered accordingly is not encouraging.
• Time limits. Because nothing forces convergence, you need your own discipline to replace the missing mechanism. Set a maximum holding period and honour it.

Risk model

• No forced convergence. This is the defining risk and it deserves to be stated bluntly. Nothing in the world compels these two prices to come together. The trade works because gaps have historically reverted, not because they must.
• Control events. A takeover, a proxy fight or an activist campaign can make voting rights suddenly valuable. If you are short the voting class, you lose, and the loss can be large and fast.
• Index flows. A decision by an index provider to include one class and exclude the other creates a permanent, structural demand imbalance. Your mean-reverting spread now has a new equilibrium.
• Borrow. The expensive class is the one you must short, and if it is the less liquid class, the borrow may be expensive, scarce, or recalled at a bad moment.
• Dividend differences. If the classes have different dividend entitlements, that difference accrues every quarter and must be in the model, or you will slowly bleed without understanding why.
• Leverage risk. The classic failure in this space is not being wrong, it is being right too late while levered.

Costs & implementation

• Low turnover, which is a rare blessing. Positions are held for weeks or months, so transaction costs are not the main enemy here, unlike most mean-reversion strategies.
• Borrow cost is the main ongoing cost, and it is charged for the whole holding period. A trade that takes eight months to converge at a two percent expected gain, with borrow at four percent annually, is a losing trade even if you are right.
• Liquidity is often poor in one class. Usually one class is the liquid one and the other is thin. Entry and exit in the thin class can be slow and expensive.
• Data: you need the charter details, dividend entitlements per class, index membership for each class, and a long, clean price history of both. Public data often conflates the classes, which is a real trap.

Failure modes

• Assuming parity is fair value. The classes are not identical and the discount usually exists for a reason. Trading toward parity rather than toward the historical range is a systematic mistake.
• Shorting votes before a control event. The single worst way to lose money here, and it is not predictable.
• Levering because it looks safe. Same company on both legs invites overconfidence. Convergence trades that cannot be forced to converge should be sized as if they might never converge.
• Ignoring borrow carry. A slow convergence with an expensive short is a loss even when the thesis is correct.
• Trading a trending ratio. If the gap has been widening for years, buying it because it is "extreme" means fighting a trend with no support.
• Data confusion. Mixing up the two classes in a price series produces phantom opportunities and real losses.

Our Notes & Suggestions

The right mental model is not arbitrage, it is relative value with no expiry date. That distinction changes everything about sizing. When a bond must be repaid at par on a known date, time is your friend. When two share classes can disagree forever, time is your enemy, and the borrow cost is the meter running.

Read the charter before you look at the chart. What actually differs between the classes is the whole story: voting power, dividend rights, whether one can be converted into the other, whether either is index-eligible. A conversion right in particular changes the trade completely, because it creates a mechanism that can force convergence in one direction.

Finally, be sceptical of the historical range you fit. These pairs are few, the histories are long, and it is very easy to conclude that a ratio is mean-reverting when you have simply looked at a period in which nothing happened. The events that break these trades are rare, which is exactly why they are not in your sample.