Longreads + Open Thread

Longreads

• Ever since Ioannidis, it's been hard to take research findings too seriously: the risk/reward for publishing bad analysis of noisy data is too favorable. But does this extend to the private sector? Alex P. Miller and Kartik Hosanagar have a fun paper saying the answer is no: they worked with a company that provides A/B testing software, mostly to e-commerce companies, and find that even though they're using the same statistical tools as academic publishing, they aren't reaching the same kinds of erroneous conclusions. It's a fun piece to read in part because the question is a good one, but coming up with a rigorous way to ask it is nontrivial (at one point they end up coming up with counterfactual p-hacking approaches that could have been implemented, to see if they were). The handwavy argument that industry and academia will have different standards for p-hacking goes like this: even assuming the willingness to do this is randomly-distributed, a marketer who chronically p-hacks will either get caught and fired, or will be working for a company that spends its marketing budget suboptimally because its conclusions have no direct connection to reality. Either way, they'll get weeded out. And that appears to be true.
• Dwarkesh Patel interviews the brilliant pseudonymous independent researcher Gwern. Wikipedia is to Gwern's site what the Encyclopedia Britannica is to the output of someone like Pliny the Elder or Al-Khwarizmi, i.e. one person's effort to exhaustively explore a broad set of topics. Gwern's most influential recent piece was when he popularized the scaling hypothesis (and note that I say "popularized," since, as he notes, other researchers had seen bits of this trend, but hadn't put the pieces together. Read the right research papers and you, too, can live in the future—though you'll only know for sure once that future comes to pass). The piece also talks a bit about pseudonymity, and the economics of writing. Which are in one sense not good—at the time of the interview, Gwern was living on $12k/year of donations and some savings, but he now has a grant offer to move to SF. Which is 1) a reminder that market inefficiencies can persist for a long time, specifically the kind where someone doesn't fully internalize their externalities, but also 2) the financial cost of becoming an influential expert on many intellectual domains is below average US spending per public school student. There are still barriers to becoming a well-educated person, but those barriers are internal.
• Tyler Cowen interviews Neal Stephenson on his new book, Polostan, and adjacent topics. (The Diff briefly reviewed Polostan last month.) One of the best riffs is when Cowen and Stephenson discuss how similar the US and USSR were in the 1930s: two countries that were under extreme political uncertainty coupled with recent economic growth, a healthy respect for science, and a mix of dense urban areas and a vast resource-rich hinterland. (They ended up having such similarities even later; Postwar points out that the US and USSR reversed the usual relationship between the imperial center and its other possessions; historically, those vassal/allies export raw materials and import finished goods, but the US imported a growing share of its manufactured goods from allies abroad and the USSR similarly invested in manufacturing in Soviet satellites while selling them wheat and later oil.
• Kevin Gee has a transcript of an interview between Stan Druckenmiller and Nicolai Tangen. The funny thing about interviews with great traders is that there's a classic bathtub curve for the utility of ideas. Whatever Druck thinks about inflation could flip 180 degrees next week, but he also has some advice that could, for the right person, be life-changing. Consider this, on what you might call trader-trade fit: "I didn't know that much about Nvidia. I just knew that AI--and I had some people here tell me how to play it. So we bought Nvidia, and then we were in the process of doing a lot more work, and then ChatGPT happened. But I've always had the view that markets are smart, they're fast, and they're getting much more so with all the communication and the technology we have today, and that if I hear a concept and I like it, if I wait and spend two or three months analyzing it, I may miss a big part of the move and then psychologically be paralyzed. It's hard to buy a stock you're looking at at 100--it's 160--even if it's going to 400, somehow your head is screwed up and you're waiting for the pullback." This is a personal question! Some people feel a lot more comfortable owning something after it's run a bit, so they'll miss the first part of the move but catch the rest. And for other people, it's the opposite—if they don't buy the first time they hear about it, they'll be buying later on from someone who heard about it around when they did and dove in headfirst.
• Dan Schwarz on how Google launched internal prediction markets, twice, and what went wrong. It's a heartwarming Coasian story about both the problem of transaction costs and the nature of firms. Specifically, one of the positive externalities of prediction markets—a case for subsidizing them, in fact—is that they produce the positive externality of legible probability estimates for important questions. But within a company, that kind of information-sharing is harmful! It interferes with managerial volatility-smoothing, where there's a bit of sandbagging early estimates, and where a project that's temporarily behind might be able to catch up before the next check-in with senior managers. If there's a live ticker tracking all of this, the internal variance of progress is more visible, and most managers' risk-adjusted performance looks worse.
• In Capital Gains this week, we finally have a long-form treatment for a topic that frequently shows up in The Diff: selection effects. The people you talk to, the ideas you're exposed to, the everyday observations you make—all of these are highly selected, and it's important to be conscious of this.
• And in this week's episode of The Riff, we cover booms, how different companies navigated bubbles, market sentiment, and why the Internet not only didn't eliminate middlemen but actually created the most valuable middlemen of all time. Listen with Twitter/Spotify/Apple/YouTube.

Books

The Predators' Ball: The Inside Story of Drexel Burnham and the Rise of the Junk Bond Raiders: In the mid 1970s, high-yield bonds were almost exclusively formerly investment-grade bonds that had been downgraded, hostile mergers were rare, and Drexel Burnham Lambert was one of many marginally-profitable cobbled-together brokerages. It was a mix of the elite firm that had gotten the Morgan family their start in banking, a small brokerage founded in the 30s with money from a family bourbon business, and funding provided by a tire company and a Belgian banking conglomerate. This is the kind of structure people come up with when they're desperate.

By the late 80s, the junk bond market was worth almost $200bn, with over $20bn of new bonds issued every year. Drexel was the most profitable investment bank in the country, and Michael Milken was the best-paid individual in history. And in February of 1990, Drexel was bankrupt, and Milken was on his way to prison.

This book tells the story of what happened in between; it was published in 1988, so the writing started when Drexel was performing well, and it wrapped up before the business completely collapsed. So it's both a good picture of a moment in time and a book full of quotes from interviews that would not have happened if the writing had started a year later.

Milken is one of those people who seems to have an internal switch that toggles between being fairly normal and being insanely driven to accomplish something. That switch was flipped during his Drexel tenure, when he was notorious for an indifference to local time zones (he worked on the West Coast, but expected people to be in the office by 7:30 AM Eastern), a phenomenal memory, and aggressive trading. Early on, the story is about a prop trader who picked a very good seat; junk bonds are good assets to sell when rates are down (because some bond investors will take on more risk rather than accept lower yields) and when growth is up (because it bails out dubious credits). The 80s had a long stretch when both of these were true.

But they didn't make their money by buying and holding; Drexel had a network of junk bond buyers, and traded both bonds and favors to keep them happy and keep the market liquid. By the late 70s, the population of once-investment-grade bonds had been more thoroughly picked over, but companies started doing original-issue junk bonds to meet demand. And this created a fun feedback loop: Milken would convince companies to issue more bonds than they needed (he had the demand), and then remind them that they were paying rates in the teens for money they didn't need, and that they could defray that cost by parking those funds in other junk bonds. This created a compounding data advantage: Drexel knew who owned which bonds, who was willing to, who was having a good year and might take one for the team, and who was having a tough year and would appreciate access to a good trade.

This is a template for a conventional, highly profitable business, and Drexel built that. But they also went in more unusual directions; the book has some stories making the case that Milken would trade ahead of exchange offers that raised the prices of some bonds, and was not especially keen to follow rules about how much trading could happen in unregistered bonds. And in parallel with running capital for the firm, Milken also had separate investing partnerships. (The book offers tantalizing hints at how much fun the 80s were for people who enjoy both financial engineering and tax law. Did you know that you used to be able to buy a treasury bond, sell the principal payment while keeping the interest, and book the difference between the value of the bond and the present value of the principal as a capital loss? For a while, you could!) Drexel would also negotiate deals where bonds were issued with warrants attached, ostensibly because customers demanded equity exposure but in practice because Drexel wanted it—the more you lever up the underlying business, the more volatile it is and the more warrants are worth.

There are some good books rehabilitating Milken after the fact, and it's notable that the problems described in Predator's Ball were not the ones that eventually got Milken in legal trouble, though it's easier to catch more administrative slip-ups than the sorts of activities that don't leave a paper trail. But it's also true that this period was an unparalleled one for rapid wealth accumulation at investment banks; the compensation packages discussed for bankers and traders in the book sound high even today, and the book underestimated Milken's personal peak compensation; it's number was a third of the actual $550m number. There's a lot of upside in bringing a new market into existence, but the kinds of things you can do running a six-figure trading account for a satellite office of a small broker are not the ones you can get away with when you're the best-paid person at the most profitable firm in your already lucrative industry. Since the collapse of Drexel, Milken has become more of an elder statesman and philanthropist. Which is probably what he would have ended up doing anyway; those hundred-hour weeks get old after a while.

Stay tuned for Monday's post with much more on Drexel and its current legacy.

Open Thread

• Drop in any links or comments of interest to Diff readers.
• Netflix is doing more live entertainment and sports, like last night's fight. What do the economics look like there, and how general is the Netflix streaming bundle?

Diff Jobs

Companies in the Diff network are actively looking for talent. See a sampling of current open roles below:

• A growing pod at a multi-manager platform is looking for new quantitative researchers, no prior finance experience necessary, 250k+ (NYC)
• YC-backed, post-revenue AI company that’s turning body cam footage into complete police reports seeks a senior founding engineer/tech lead who can build scalable backend systems and maintain best practices for the engineering org. (SF)
• A Google Ventures backed startup founded by SpaceX engineers building data infrastructure and tooling for hardware companies is looking for a staff product manager with 5+ years experience, ideally with AI and data intensive products. (LA, Hybrid)
• A well-funded startup that’s building the universal electronic cash system by taking stablecoins from edge cases to the mainstream is looking for a senior full-stack engineer. Experience with information dense front-ends is a strong plus. (NYC, London, Singapore)
• An AI startup building tools to help automate compliance for companies in highly regulated industries is looking for a director of information security and compliance with 5+ years of info sec related experience at a software company. Experience with HIPAA, FedRAMP a plus. (NYC)

Even if you don't see an exact match for your skills and interests right now, we're happy to talk early so we can let you know if a good opportunity comes up.

If you’re at a company that's looking for talent, we should talk! Diff Jobs works with companies across fintech, hard tech, consumer software, enterprise software, and other areas—any company where finding unusually effective people is a top priority.