Ben Graham, The Intelligent Investor, Ch. 16, "Convertible Issues and Warrants," opening line.
There are only two pieces. The investment value (bond floor) is the coupon and principal discounted at the yield of comparable straight credit — it moves with rates and with the issuer's creditworthiness. The conversion value = par ÷ conversion price × share price. The market price normally sits above the greater of the two, and the excess is the price of the option, measured by the conversion premium. The bill is itemized: a convertible's coupon is materially below the same issuer's straight debt, and that gap is the option's annual fee.
In September 1987 Berkshire bought $700 million of Salomon 9% convertible preferred at a $38 conversion price. A month later came Black Monday and Salomon's common was cut roughly in half — the bond side did its job: however far the stock fell, the 9% kept being paid. The other end delivered just as faithfully. The conversion price was never durably exceeded, and when Travelers acquired Salomon in 1997 the return on a decade of holding was, broadly, that 9% coupon. The floor saved him; the option gave him nothing.
The bond floor is a line that can collapse: the credit supporting it deteriorates in step with the stock. In 2023 an onshore Chinese convertible defaulted outright for the first time, after years in which "convertibles do not default" was treated as an institutional fact. In 2008 US convertibles traded tens of points below their theoretical floor — the models were not wrong; there were no buyers.
Charlie Munger's standing test for reading any arrangement, from Poor Charlie's Almanack.
Chinese convertible terms are highly standardized, so they can be read directly as a statement of issuer intent. Forced call: if the stock closes at or above 130% of the conversion price on at least 15 of any 30 consecutive trading days, the issuer may redeem at par plus accrued interest — the aim is never to repay you, it is to force conversion and turn debt into equity. Put: if the stock closes below 70% of the conversion price for 30 consecutive days, holders may require the company to buy the bonds back; this is the one clause on your side. Downward reset: lowering the conversion price looks like a gift to holders, but its purpose is to dodge the put and reopen the conversion route, and existing shareholders pay for it in dilution.
Intco's convertible was issued at par of 100 yuan in January 2020. Its underlying stock soared on pandemic glove demand, and in January 2021 the convertible traded as high as roughly 3,618 yuan — the highest price in the onshore market's history. The company then announced early redemption: any holder who failed to convert or sell before the record date would be paid the redemption price of about 100 yuan. The same piece of paper: a market price above three thousand on one side, a contractual price of one hundred on the other, separated only by an administrative deadline.
Treating the terms as a free cash machine is equally dangerous. A downward reset is never guaranteed — the board may decline to propose one, shareholders may vote it down, and the controlling holder's dilution incentive runs against yours. The put right typically vests only in the final two years and can be exercised once a year. In October 2020 Zhengyuan's convertible rose more than 170% intraday; the regulator issued the Administrative Measures for Convertible Corporate Bonds, and from August 2022 the exchanges imposed daily price limits. Playing the terms as pure speculation gets the rules rewritten.
Buffett's 1994 letter, reviewing the USAir convertible preferred bought in 1989.
"Seemingly offered" carries the whole sentence. Seniority converts into value only in liquidation; until then the company can suspend the dividend at will and you have no vote to stop it. Worse, holders routinely relax their scrutiny of the business precisely because protection appears to be there. That is the real cost of a protective wrapper: what it displaces is not risk, it is diligence.
In 1989 Berkshire put $358 million into USAir 9.25% convertible preferred, and in the same year $600 million into Gillette 8.75% convertible preferred. Nearly identical structures, opposite outcomes. Gillette redeemed in 1991, Berkshire converted, took roughly 11% of the company and watched it compound into billions. USAir suspended its preferred dividend in September 1994 and Berkshire wrote the position down to $89.5 million at year end — about three quarters gone. The only difference was whether the business could keep earning money. There is a second half: USAir unexpectedly recovered, paid the arrears and redeemed the preferred, and Berkshire exited at a profit — Buffett said plainly that the outcome owed nothing to his judgment. Capital returned by luck does not validate a method.
The reverse also holds: hybrids are superb instruments at particular moments. In September 2008 Berkshire put $5 billion into Goldman Sachs — 10% perpetual preferred plus warrants — and the terms were rich because almost nobody else would write that cheque. Which reveals the real pricing logic: terms are set by how badly the other side needs money, not by the structure. An individual buying a hybrid in the open market has no such bargaining environment. Treat retail structured products the same way: snowballs and callable notes also advertise "protection plus enhancement," but most of them are selling tail risk for coupon — the upside is sealed and the downside opens fully in the extreme.
Howard Marks's standing insistence that price comes before quality.
The dominant US buyer is the convertible arbitrage fund: buy the bond, short the stock on delta, harvest volatility and coupon, take no view on direction. That keeps US convertibles priced close to option-theoretic value — and makes them acutely sensitive to funding conditions: when leverage is withdrawn, price and theory come apart at once. China largely lacks that leg. Stock borrow is difficult and expensive, convertibles are held mostly by fixed-income-plus funds and individuals, and exits depend heavily on the forced call. The result is a persistent structural valuation premium and a far heavier emphasis on playing the terms.
2008 lit that dividing line most clearly. Convertible arbitrage fell about a third that year (the HFRI convertible arbitrage index around −34%), and not because the trade was wrong: deleveraging forced liquidations, and in September the SEC temporarily banned short selling in nearly 800 financial stocks, severing the hedge leg outright. Convertibles broke below their theoretical floor, then rebounded roughly 60% in 2009. The strategy did not change; the available funding and the available hedge did.
None of which makes "Chinese convertibles are easier money" a conclusion. From August 2022 the exchanges imposed price limits (+57.3%/−43.3% on the first trading day, 20% thereafter) plus abnormal-trading surveillance, compressing the institutional arbitrage; the 2023 defaults ended the implicit-guarantee assumption. For an individual, the honest place for a convertible is buying convexity you don't have to time, paid for with a low coupon: the advantage is that it cannot expire worthless like an option, the cost is that the upside is capped by the call. It belongs at the edge of a portfolio, not as a substitute for a core compounding asset.