Warren Buffett Is Just an Average Employee

Pay ratios.

Ahahaha sure:

Berkshire Hathaway Inc.’s Warren Buffett is scoring particularly well on a new rule requiring companies to disclose the ratio of a chief executive officer’s pay to that of the median employee.

His annual compensation of $100,000 was just 1.87 times the median employee’s pay of $53,510, a figure calculated from a sample of about two-thirds of Berkshire’s total employees, according to a filing released Friday. He also gives back about $50,000 to the company “for minor items such as postage or phone calls that are personal,” meaning his take-home pay would be less than that median figure.

First of all: Unlimited talk, text and data cell-phone plans will run you less than $1,000 a year; landlines are even cheaper. Warren Buffett spends … $49,000 a year on … postage? Like, for personal letters? It is a little mysterious.

But also: Warren Buffett makes $100,000 a year? Really? I mean, yes, it is famously his salary. But Buffett increases his wealth each year in two ways: He gets paid for doing his job, and also he has billions of dollars invested in Berkshire Hathaway and most years Berkshire’s stock goes up. In 2017, Berkshire Hathaway’s stock was up about 22 percent, meaning that the value of Buffett’s shares increased by about $15.1 billion, to $84.1 billion. So in a sense he made $15.1 billion in 2017, or $15.1001 billion if you include his salary, or $15.10005 billion if you deduct the stamps. That’s a pay ratio of about 282,435 to 1. 

Is that the right way to count? Meh. Every year Institutional Investor’s Alpha publishes its “Rich List of the World’s Top-Earning Hedge Fund Managers,” and every year people write stories saying that the list reveals how much hedge fund managers “are paid” or “take home,” and every year I point out here that it is actually mostly a list of how much those managers’ investments appreciated. It is not really how much they are “paid.” But people like to interpret it that way, and you can understand why, since for practical purposes that appreciation is a big part of their economic reward for running their hedge funds. Most years Buffett would be way above anyone on the hedge-fund-manager list, if he counted. 

Obviously the Securities and Exchange Commission has good reason for counting only actual pay, and not stock appreciation, in its pay-ratio rules. And obviously Berkshire Hathaway has good reason for doing the calculation according to the rules. (And perhaps the median Berkshire employee also owns some stock that appreciated last year.) I don’t know exactly what those rules are meant to measure, though to be fair the SEC doesn’t know either. (“Congress did not expressly state the specific objectives or intended benefits of Section 953(b),” sniffed the SEC when it begrudgingly adopted the pay-ratio rules in 2015, five years after Congress mandated them.) But in any case, it’s clear that Warren Buffett’s annual economic reward for running Berkshire Hathaway is more than $100,000. That’s barely enough to cover his postage.

Elsewhere in compensation, Blackstone Group LP Chief Executive Officer Stephen Schwarzman got some:

The firm will provide Mr. Schwarzman free access to a car and driver for life, privileges he previously was guaranteed for a limited number of years. If and when he retires, Mr. Schwarzman will be reimbursed for travel on behalf of Blackstone and entitled to legal representation for matters related to the firm, reflecting his possible continuing role as a public face of Blackstone.

As part of a new agreement, the buyout firm will allow Mr. Schwarzman’s estate to invest in or alongside Blackstone funds without fees in the decade after his death, an unusual perk, in return for his contributions to the firm. A previous contract allowed him to make those investments without fees only during his lifetime.

It’s strange. At the very tippy-top levels of founder-CEO compensation, there are basically two opposite philosophies. There is a maximalist approach in which the CEO should be showered with all the goodies you can think of, which proves his importance and his firm’s love for him. And there is the sort of Zen approach in which the CEO is paid a dollar, or $100,000, can’t even expense his postage, and grows his dynastic wealth just by owning most of the company. The latter approach looks more appealing; there is something unavoidably lame about a board spending time negotiating with a billionaire CEO to extend his lifetime fee waiver to a lifetime-plus-10-years fee waiver. But from the perspective of an ordinary mortal the approaches kind of come to the same place. The perks, the salaries, they all look like rounding error compared to the basic economic engine of owning billions of dollars’ worth of stock in a growing company. Once you’re there, your choice about how else to get paid is mostly aesthetic.

Spurious correlations.

“The Super Bowl Indicator is a superstition that says that the stock market’s performance in a given year can be predicted based on the outcome of the Super Bowl of that year,” says Wikipedia. “This pseudo-macroeconomic concept states that if a team from the American Football Conference (AFC) wins, then it will be a bear market (or down market), but if a team from the National Football Conference (NFC) or a team that was in the NFL before the NFL/AFL merger wins, it will be a bull market (up market).”

Wikipedia is wise enough to know that the indicator is a superstition, or a joke, but the financial press tends to publish semi-serious articles about it every year. I don’t think that there are any exchange-traded funds devoted to trading the Super Bowl Indicator, but it would not surprise me to learn that some retail money chases it. Humans like meaning, you know? It is nice to think that the world is full of significance, that seemingly unconnected events have deep and mysterious ties to each other. And it is nice to think that you are especially skilled at understanding the patterns, and also that you can get rich by watching football games.

Here’s an article about the robot equivalent:

Chief executives that said “please”, “thank you” and “you’re welcome” more often enjoyed a better subsequent share price performance.

However, on closer examination this turned out to be what statisticians call a “spurious correlation” — and an excellent example of one of the biggest risks of the current fad for using AI for investment purposes. “Discovering correlation but failing to search for causation occurs rather frequently in financial research,” says Evan Schnidman, the head of Prattle, which is based in St Louis, Missouri. “We always go to the primary sources, look for correlations and control as much as we can to weed out spurious correlations in an effort to discover a causal effect.”

Discovering correlation but failing to search for causation occurs rather frequently in most areas of human endeavor, as you can tell by casually glancing at the internet

There is a tendency to attribute to artificial intelligence flaws that are categorically different from the flaws of human intelligence. Data mining is “‘the kryptonite of our industry,’ according to Gary Chropuvka, a partner at Goldman Sachs Asset Management’s Quantitative Investment Strategies,” that sort of thing. But it seems to me that humans like spurious correlations at least as much as computers do; it’s just that humans can’t find as many of them as quickly as the computers do. Arguably the humans are better at rejecting spurious correlations that don’t make intuitive sense, but be careful with that: The Super Bowl Indicator makes no sense but people love it, while the politeness indicator (which turns out to be spurious) has an obvious intuitive story. (“Prattle’s politeness signal could have indicated that companies with politer chief executives do better, which does not sound unreasonable.”) And arguably the computers are at least as good as the humans at rejecting spurious correlations that don’t work out-of-sample: The real controls on statistical overfitting are further statistical techniques, not human intuition. 

This is a case of a more general thing I think, which is that the problems of computers in investing are mostly just the problems of humans in investing, but bigger and faster. Bigger and faster versions of old problems can feel like — can be — qualitatively new problems; things that are cute foibles when practiced by individual humans can be systemic catastrophes when scaled up by machines. Still it often seems to me a little unfair to blame the computers. If artificial intelligence is always going around finding meaning where there is no meaning, we shouldn’t feel too superior; that’s definitely something that it learned from us.

Blockchain structured products!

Really it is hard to think of a more straightforward use case for “smart contract” technology than structured notes. A structured note is some more or less complicated derivative embedded into a note, with triggers and formulas that determine how much the borrower actually pays back. If you code all of those triggers and formulas directly into an automated smart contract for the note then, first of all, no one has to remember to keep track and calculate them during the life of the note. Also, for the people who buy the notes, the formulas are auditable: You can go read the code of the smart contract and make sure that the formulas in the code match up with the formulas you expect. (Obviously you won’t, but you could.)

Compared to other sorts of contracts, structured notes are easy to put on the blockchain because not much happens in them. They tend to depend on easily observable real-world events: Structured notes pay off not “if Party A completes construction on this project to the reasonable satisfaction of Party B,” but more like “if Index X moves more than Y percent above its initial level,” and it’s pretty easy to get that data in a publicly accessible form that everyone will agree on. And they tend not to involve any real-world performance other than moving cash: If your structured note hits a trigger, you don’t have to ship bananas across the world or whatever; you just have to move money into an account electronically. It’s the sort of thing that computers are good at.

More subtly, the money shipments are one-way: The way a structured note works is that investors buy it on Day One and never have to put up any more money again; the only movements of money after that are from the issuer to the investors. Financial derivatives generally seem like promising smart-contract opportunities, but in the average interest-rate swap, say, both sides might have to make payments at various points during the life of the trade, and so it is hard to guarantee that the contract will work seamlessly without locking up a lot of money on both sides to collateralize the contract. But with a structured note there’s no need to worry about the creditworthiness of the buyers: They put up their money at the beginning and never owe anything again. Of course you have to worry about the creditworthiness of the seller, but if they’re marketing structured notes presumably they have some way of convincing investors that they’re good for the money.

Anyway now there are blockchain structured notes:

Blockchain, the technology promising to usher in simplicity and transparency to finance, is entering one of the industry’s most obscure corners.

London-based Marex Solutions created what it says is the first structured product to be registered, cleared and settled using the distributed ledger technology underpinning Bitcoin. The two-month pound-denominated notes pay a coupon of up to 13 percent per year based on the performance of the FTSE 100 Index, according to a term sheet.

The objection is I suppose that while this is a straightforward use case for the blockchain, it is not an especially interesting one. The upside of automating a purely financial contract with cash flows that only go one way is that you can do it without having to worry too much about interactions between your contract and the real world. The downside is that you are not doing much to change the real world. 

People are worried about stock buybacks.

But Jesse Fried and Charles Wang of Harvard are not. From the Harvard Business Review:

There is little evidence that buybacks and dividends by the S&P 500 are hurting the economy by depriving firms of capital they would otherwise use for investment and paying workers. Far from being starved of resources, S&P 500 companies are at near-peak levels of investment and have huge stockpiles of cash available for even more. Our analysis shows that the proportion of income available for investment that went to shareholders of the 500 over the past 10 years was a modest 41.5%—Less than half the amount claimed by critics. One must also recognize that some of the capital flowing to S&P 500 shareholders is then reinvested in smaller public companies and private firms, fueling growth and employment outside the S&P 500. And payouts don’t appear to meaningfully contribute to income inequality.

They do some odd arithmetic here, arguing that buyback numbers are overstated because they don’t take into account the fact that companies are also issuing stock as well. This is odd because that stock issuance is largely in the form of executive compensation, and in fact a central point made by buyback critics is that buybacks are a sneaky way to transfer cash from companies to executives by giving the executives stock and then buying back stock. (Dan McCrum points out this and other issues in Fried and Wang’s piece at FT Alphaville.) Still the bigger points are that (1) big public companies don’t seem to be especially cash-constrained and (2) some of the cash probably ends up at smaller companies where it can be more productively invested.

People are worried about bond market liquidity.

Nah, but here is a story about how no one wants to buy new-issue bonds anymore:

Fewer orders are coming in for new bonds, relative to what’s for sale. Companies that sell notes are paying more interest compared with their other debt, according to data compiled by Bloomberg, and once the securities start trading, prices by one measure have been falling about half the time. It’s the latest signal that the investment-grade debt market is losing steam after years of torrid gains, as rising rates and talk of tariffs weigh on the outlook for corporate profit.

New-issue bonds play an important part in one of the standard stories about bond market liquidity: Because big investment banks control new-issue allocations, the story goes, and because new issues are an important source both of bonds (with little liquidity, new issues are the only time bonds come available to trade) and of profits (because new-issue bonds regularly pop when they start trading, buying them is an easy way to make money), the big banks have outsized power in the market for trading seasoned bonds. Sure you could go to some electronic platform to trade your bonds, cutting out the middleman, but the middleman is the one who controls all your new-issue allocations, and so you don’t want to make him mad or you won’t get any new bonds. And so all-to-all trading platforms never get as much traction as they should, because investors remain reliant on the banks, and the banks don’t provide as much liquidity as they used to.

But if no one wants the new bonds, and if buying new bonds isn’t reliably profitable, then this story kind of goes away. That doesn’t solve all of the problems of bond market liquidity — there are still a lot of bonds, a lot are still bought and held forever, etc. — but if your bond-market-liquidity worry focused on banks’ power in the new-issue market, then the decline of the new-issue market might look like good news.

Things happen.