ADVERTISEMENT

The Robots Can Handle the Factors

The Robots Can Handle the Factors

(Bloomberg Opinion) -- Robots vs. humans (1).

At a high level, what investors do is some form of pattern matching. A stock has some characteristics—a head-and-shoulders price pattern, a low price-earnings ratio, a chief executive officer who won’t look you in the eye, a series of complex financial transactions that boost reported income without affecting cash flows—that match, in some salient respects, some characteristics of other stocks that have previously gone up or down. And you have seen enough stocks, and read enough books by other people who saw a lot of stocks, that you have a well-developed sense—a deeply researched thesis, perhaps, or a gut instinct, or whatever—about which stocks with which characteristics will go up or down, so you buy the ones that look like they should go up and sell the ones that look like they should go down. And sometimes this works out and sometimes it doesn’t, but if it works out enough—if you are good enough at matching the patterns, and if you are operating in an environment where the patterns that you are matching remain relatively stable—then you can make a nice living doing this.

Some of these patterns are complex and subtle, but some are not. One useful pattern is literally that if a stock went up yesterday it will probably go up today. Clifford Asness discovered that one, and it got him both an economics Ph.D. and three billion dollars. And he didn’t even really discover it! People have been buying stocks that went up yesterday for as long as they have been buying stocks, or I guess one day less. What Asness did was to formalize this pattern, to name it “the momentum factor,” to examine it statistically (which got him the Ph.D.), and to launch an asset manager, AQR Capital Management, to invest client money in this and other factors (which got him the money). A “factor” is essentially a stock-price pattern that has been formalized and written down by academics and then turned into part of the formula of a fund. Having a high ratio of book value to market value is a factor: The academic literature recognizes that high-book-to-market stocks outperform lower ones, and you can invest in inexpensive exchange-traded funds that focus on high-book-to-market stocks. Having a chief executive officer who is smart and personable is not a factor: It may improve stock returns, but it is hard to study quantitatively or put in a fund. Having a CEO who is bad at golf is also not a factor: It doesn’t really go into too many academic models of stock returns, and there are no bad-at-golf ETFs that I am aware of. And yet it does predict stock returns, and perhaps in the future as finance becomes more sophisticated and specialized someone will start that ETF. 

Here is Robin Wigglesworth on Asness, AQR, the rise of factor investing, and the question “Can factor investing kill off the hedge fund?” Here is Asness’s view on that question:

“It is part of our business to be the Vanguard of hedge funds. It’s not all of our business, by any means. But to take some of the basics and say you should get this for lower fees,” he says. “What [hedge funds] are doing as a group is good, but simple. And they’re kicking up a whole lot of dust around it.”

That is not quite a claim that factor investing should replace sophisticated pattern-matching by expert humans. It is a claim that factor investing can replace simple pattern-matching by expert humans, combined with a claim that a lot of what the expert humans do is actually pretty simple pattern-matching, plus dust. It turns out that a lot of the expertise you have built up through long experience in the markets, a regression model can replicate just by looking at the history of the markets. Not all of it! Someone still has to shake the hands and examine the covenants and do lots of specialized work at the margins. But you don’t need to pay someone millions of dollars to buy stocks today that went up yesterday. A computer can do that. And, automation and economies of scale being what they are, the person who builds that computer can make billions of dollars.

We talked a couple of weeks ago about one kind of market automation, the kind where computers quickly scan a government release about soybean supplies and decide to buy or sell soybeans based on that data. I wrote:

It is hard as a human to make a profit by reading this morning’s USDA report; the only way left to make a profit is by correctly anticipating next month’s report. That is of course how it should be! Taking away the easy rote tasks and freeing up humans for more creative tasks is the point of automation, in any industry. It makes the industry more efficient, if obvious information is instantly incorporated into prices by robots instead of being laboriously incorporated by overpaid humans. 

The essence of progress in finance is that more kinds of information will become obvious, and will be instantly incorporated into prices by robots. “Stocks that went up yesterday will probably go up today” is now obvious—it’s in a published Ph.D. thesis after all—and so it has been automated, and you don’t need to hire a sophisticated charismatic human to buy the stocks that went up yesterday.

This frees up the sophisticated charismatic humans to focus on other, subtler patterns that have not (yet) been reduced to factors, or to devote themselves to something else (cancer research? launching factor ETFs?) because their skills are not needed for stock-price pattern recognition. But as with the soybeans, it can feel a bit rough on the charismatic humans. They developed their pattern-recognition skills through long experience, and they learned to recognize lots of patterns, and then the most useful ones just got handed to machines. 

Robots vs. humans (2).

At a high level, what investors do is some form of pattern matching. Venture-capital investing is pattern matching like anything else, but compared to, say, public equities investing, there is just a lot less statistical data. You can’t be like “53 percent of the time, venture-capital investments that are up more than 0.3 percent one day go up the next day as well, so let’s buy those”: There aren’t generally daily returns, and even if there were you couldn’t trade them. You can be like “13 percent of the time, companies founded by Harvard dropouts return at least 10x to their Series A funders,” or whatever, but there are just so many fewer data points that it is hard to have much confidence in the validity of your conclusions.

But Google is smart, so you never know. Here is Dan Primack at Axios on GV’s investing machine:

When most venture capitalists want approval to make a new investment, they go to their partners. When venture capitalists at GV do it, they go to something called "The Machine." …. Axios has learned that the firm, formerly known as Google Ventures, for years has used an algorithm that effectively permits or prohibits both new and follow-on investments.

Staffers plug in all sorts of deal details into "The Machine" — which is programmed with all sorts of market data, and returns traffic signal-like outputs. Green means go. Red means stop. Yellow means proceed with caution, but sources say it's usually the practical equivalent of red.

It was initially designed and used as a due diligence assistant that could be overruled but, according to three sources, it has evolved into a de facto investment committee.

Sure why not. Well, one reason why not is that The Machine is apparently bad at its job:

Venture capital is as much art as science, intuition as calculation. But several GV investors tell me that "The Machine" has ripped out their guts, possibly costing them lucrative opportunities.

Which is not that surprising, given (1) lack of data for pattern recognition etc. and (2) it was apparently designed by people who didn’t know how to do venture-capital investing?

Few of the early GV investors had much, if any, investing experience. So "The Machine" would leverage the firm's strengths (engineering) as a bulwark against its weakness (proven VC chops).

I am not sure that that’s how strengths and weaknesses work, but whatever. My question is, should you be optimistic that in the long run The Machine will eventually get good at VC investing? There are obstacles: the lack of data, primarily, and the subjectivity of what data there is. “In a world where judgment over a founder's qualities is almost as important as the initial business plan,” writes Jamie Powell, “it seems foolish to leave key decisions to an aspiring HAL 9000.” I don’t know, though. That sort of judgment is just pattern recognition too, and the lack of data is a problem for humans as well as computers. Venture-capital humans rely on underdeveloped heuristics too you know. A computer can measure the strength of a founder’s handshake, or scan her retinas to “look” her in the eye. Presumably somewhere there is a captcha that is training Google’s computers to recognize hoodies, or black turtlenecks. Once it gets as good as the humans at doing that, why not use it as the investment committee?

Financial innovation.

What if—just hear me out here—what if a financial firm raised money from investors and loaned it to companies to use in their businesses? And then, instead of syndicating or selling or securitizing or tranching or hedging or trading the loans, what if the firm just waited to see if the companies paid them back? Is that … is that a thing? Can you do that? Here is Mary Childs at Barron’s on direct lending:

Sure, lending is one of the oldest and most basic functions in finance. Since the financial crisis, however, investors have been pouring into direct-lending funds in waves, eager to share the profits of lending to companies either too small or too risky to be bank clients. …

These aren’t the syndicated high-yield loans that investors fell in love with after the financial crisis. They’re the little sibling, not big enough to get a grade from the ratings firms.

Direct-lending deals are negotiated between lender and borrower. They’re generally smaller than bank loans, between $10 million and $250 million, and made to companies with less than $50 million—or perhaps $100 million—in earnings before interest, taxes, depreciation, and amortization, or Ebitda. …

But instead of selling the debt to a wide array of yield-hungry investors, as in the high-yield bond and loan markets, private lenders generally keep it. They collect the interest payments and hope nothing goes wrong, marking the loans monthly, or even quarterly, a nice reprieve from the harsh light of intraday pricing. At the end of the loan’s life, maybe three or five years later, the lenders hopefully get their money back.

It’s a little like what banks used to do—lend smallish amounts of money to smallish companies based on relationships and handshakes, not credit ratings and syndication—until they were, um, prohibited from doing it?

The push into direct lending is partly a story of banks retreating in the wake of regulations passed after the financial crisis, and asset managers happily picking up the slack. For instance, interest surged in direct lending after early-2013 regulatory guidance that clarified what banks could and could not do in terms of leveraged lending.

Oh right it is not quite like the Bailey Building & Loan approach in that the companies are pretty levered. Like the trick is that, like old-timey banks, the direct lending funds just make loans and hope that they’ll be repaid, but they’re maybe a bit less focused on the getting repaid part. “The company can go to a bank, but if it wants more leverage, or looser covenants—why not?—it might prefer a private direct lender.” Delightfully covenant quality in direct lending seems to be weakening even as no direct lender weakens its covenants:

From his role as chairman of the Alternative Credit Council, which represents asset managers in private credit and direct lending, Fiertz has some visibility into the market via 100-plus members overseeing some $350 billion of private credit assets.

“We have all the large direct lenders in there, and every single one of them says they have not compromised on covenants yet,” he says. “Yet that’s impossible, because we account for the bulk of the market and we know, either through the surveys or larger size loans or anecdotes, that covenant quality is deteriorating.”

Ah yes, all the covenants are above average.

Are derivatives sinful?

Yes, says the pope. No, says the chairman of the U.S. Commodity Futures Trading Commission, who is “a devout Roman Catholic.” (The pope is also Catholic.) CFTC Chairman J. Christopher Giancarlo and chief economist Bruce Tuckman responded to the Vatican’s condemnation of derivatives in a letter to the Vatican, as one does:

Messrs. Giancarlo and Tuckman, in the letter, argued for the social utility of derivatives, saying that rather than preying on the vulnerable, derivatives are actually a boon for “the world’s poorest farming communities.”

The two backed up their rebuttal of the Vatican with detailed descriptions of how, for example, derivatives can help stem boom-and-bust cycles in Madagascar’s vanilla-crop prices.

Look, I am obviously team CFTC here, but I am not a Catholic and have no particular standing in this debate. I do want to suggest, though, that there’s no particular reason to assume that the tenets of revealed religion will always match up perfectly with the demands of economic efficiency. It was probably convenient for people to have money-changers in the temple! Improves liquidity and all that. And still.

Oh Tesla.

“It is not from the benevolence of the butcher, the brewer, or the baker that we expect our dinner, but from their regard to their own interest,” wrote Adam Smith, but it can’t hurt to ask:

Tesla Inc. has asked some suppliers to refund a portion of what the electric-car company has spent previously, an appeal that reflects the auto maker’s urgency to sustain operations during a critical production period.

The Silicon Valley electric-car company said it is asking its suppliers for cash back to help it become profitable, according to a memo reviewed by The Wall Street Journal that was sent to a supplier last week. Tesla requested the supplier return what it calls a meaningful amount of money of its payments since 2016, according to the memo.

“We address ourselves, not to their humanity but to their self-love, and never talk to them of our own necessities but of their advantages,” continued Smith, about the butchers and so forth, but perhaps with tire and windshield and suspension manufacturers talking of one’s own necessities goes over better? I think if I ran an auto-parts supplier I would not be all that keen on helping a $50 billion company run by an eccentric billionaire become profitable. Like, right now the situation is that the suppliers are profitable (I hope!) but not $50 billion companies run by billionaires, and Tesla is unprofitable but a $50 billion company. Who really deserves whose charity? If the suppliers sacrifice their profitability and give it to Tesla, I do not see Tesla sacrificing its market cap and giving it to the suppliers. 

People are worried about bond market liquidity.

Oh man, they so are:

In May, Italian two-year government-bond yields notched their biggest one-day jump since at least 1989. The surge was triggered by Italian politics, but a lack of liquidity appeared to amplify the moves as the gap between the price at which traders were willing to buy and where they were willing to sell surged to above half a percentage point, according to Thomson Reuters data.

Alberto Gallo, who runs more than $1 billion in Algebris Investments’ Macro Credit strategy, said it took “around 10 times longer” to unwind a bet on Italian bonds than normal and that it was hard to get bids or offers on trades of more than $10 million in size.

Et cetera. Usually when you see this story the suggestion is that, when bond prices go down, someone—a bank, say—was supposed to buy bonds at close to the old price, and instead didn’t. “Lack of liquidity” can, in these stories, look a bit like “bond prices respond to information.” And there is some of that here, but there is also this:

When emerging-market debt sold off in late June, Mr. Pradère looked to buy Romanian government bonds just when other investors appeared most eager to sell.

But the prices he saw quoted by dealers on his screen were rarely matched by those from actual sellers when he followed up. After a long search, he eventually gave up on the trade.

That’s actually a little weird! That’s a story about a buy-side investor looking to buy when other investors were looking to sell. He was, in the jargon, offering to supply liquidity. The problem here is not about the lack of a buyer who will step in to cushion falling bond prices: He was that buyer. Instead the problem is that for some reason—dealers unwilling to intermediate that trade, a market structure that makes it difficult for real buyers to match with real sellers without that intermediation, etc.—he couldn’t do it. That or just, you know, he thought the bonds should be cheaper than they were and the sellers thought they should be more expensive than they were. You can call that a liquidity problem too.

Things happen.

Trump Lines Up Powell as Scapegoat If Tax, Trade Moves Sour. US banks urge UK to cut corporate taxes to stop Brexit exodus. BOJ Policy Change Speculation Roils Markets. Theranos Investors, Founder Holmes Resolve Shareholder Suit. Pimco Executive Resigns After Allegations of Inappropriate Behavior. Vitalik Buterin’s Conversation With Tyler Cowen. The Premium for Money-Like Assets. Mice treat sunk costs as real.

To contact the editor responsible for this story: James Greiff at jgreiff@bloomberg.net

This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.

Matt Levine is a Bloomberg Opinion columnist covering finance. He was an editor of Dealbreaker, an investment banker at Goldman Sachs, a mergers and acquisitions lawyer at Wachtell, Lipton, Rosen & Katz, and a clerk for the U.S. Court of Appeals for the 3rd Circuit.

©2018 Bloomberg L.P.