Why A "Commodity Supercycle" is Unlikely

diff.substack.com · by Byrne Hobart

In this issue:

  • Why A "Commodity Supercycle" is Unlikely

  • The Rise and Fall and Slight Rise of American Growth

  • Financial Engineering

  • Financial Engineering, Pt. 2

  • The Decline in Federal R&D is Closer to a Shift

  • Reflexive Palantir

  • Hotel Flipping

Why A "Commodity Supercycle" is Unlikely

Goldman Sachs argues that a new commodity supercycle is at hand. Zerohedge agrees. The CRB index of energy, agricultural commodities, precious metals, and industrial metals is up 77% from its late April low, and back to the level it was at in the few years before Covid, despite a drop in global economic activity. If institutions, anti-institution blogs, and prices all agree, then a viewpoint counts as consensus.

I've written before about how inflation will be higher in 2021, which every central bank wants to happen, and that it will be hard to measure because the consumption basket in 2020 was so different from before. 2021's consumption will be hard to compare to 2019's, even if the products are similar: will restaurants permanently reduce headcount because QR code menus and advance ordering makes them slightly more efficient? Will the businesses that survived the pandemic be less labor-intensive than the ones that went under? If Zoom lectures remain optional at colleges, will the lack of a forcing function for going to class mean that more of college is Khan Academy with different material and worse production values? And, if so, does this require a quality adjustment for consumption? (A meal at Burger King is mostly about food, but when you order the prix fixe meal at a fancy restaurant, you're partly paying for entertainment ($, WSJ). Somewhere in that continuum there exist restaurants that used to charge a premium for service, and may end up charging the same premium without the same level of service.)

But the short-term inflation noise is not the same as a long-term trend towards higher commodity prices. A supercycle would be an extreme event; if the next oil bull market is like the last one, that's a 635% increase in oil prices, or $450/barrel oil. This is not impossible—$147/bbl in mid-2008 would have been a shocking prediction in 2002—but it implies a world that is wildly different in some ways, and wildly static in others. It requires persistent increases in resource consumption, comparatively insensitive to higher prices, and that means either that very poor countries reach middle-income status or that middle-income countries pay through the nose in the future to achieve what counts as a rich-world standard of living today.

Supercycles accompany epoch-defining changes. The 70s commodity supercycle—a decade of 21% annualized growth in commodity prices—was accompanied by:

  1. A peak in US oil production, leading to higher imports,

  2. An end to the Bretton Woods system, reducing the value of the dollar and (by definition) raising the dollar value of commodities, and

  3. The Baby Boomers, the largest generation in history up to that point, reaching the peak of their consumption as they got old enough to form households (i.e. buy houses), get cars, drive their cars, buy appliances, use electricity to power those appliances, etc.

The next commodity supercycle was in the early 2000s. In this case, the coinciding factors were:

  1. A supply shortage caused by decades of under-investment in commodity production after the end of the last supercycle.

  2. A ~40% decline in the dollar from the start of 2002 through early 2008 (the Bretton Woods shock caused a 23% decline in two years, although that was relative to other currencies that also lost value).

  3. China's rise from a poor country to a middle-income country, which put a billion people on the path to owning cars, driving cars, buying appliances, using electricity to power those appliances, etc.—but which involved proportionately more investment in heavy industry than previous development strategies. (China, too, had lingering effects from a baby boom: birthrates peaked in 1965.)

Both supercycles did not just require one-time events, but a confluence of one-time events. There have been plenty of economic recoveries that coincided with a temporary increase in commodity prices, but did not set off a long-running bull market in those assets. The 1982-1990 economic recovery, for example, was partly driven by lower commodity prices. It was certainly not a pleasant time to be in the oil or copper business, much less in agricultural commodities.

The 2020s recovery will, in that sense, look a bit more like the 1980s. While China was a major consumer of commodities in 2020, and is most of the reason prices recovered as much as they did, the country's long-term goal is to shift the economy away from resource-intensive activities and towards internal consumption, which will necessitate a larger services sector and less concrete and copper per incremental dollar of GDP.

Meanwhile China, like the US, Western Europe, Japan, and South Korea, faces long-term demographic headwinds. Commodity supercycles coincide with high family formation, but family formation should be net negative in China in the coming decades; the only population cohort that's growing is the 50+ age group. Older people do spend money, so consumption will still grow, but their consumption skews to healthcare, not manufactured goods.

Extreme price movements are still possible, especially in commodities sensitive to low-emissions energy production and transportation like copper ($, FT). That will be especially likely in 2020, because of Covid-induced supply bottlenecks which have affected copper in Chile (for now ($, FT)), uranium in Canada, and other products.

But the supply shortages are a transient effect of the pandemic. Meanwhile, a transition to low-emissions energy is a generation-long endeavour, and poorer countries will choose to keep polluting if the cost of not doing so is high. The same transition could lead to higher oil prices, if production cuts happen faster than consumption cuts, but while oil production is inelastic in the short term, long term production rates are set by the least eco-sensitive entity with access to reserves. Saudi Arabia's Neom hydrogen project gives the country room to maneuver: it can increase oil production now in exchange for planning low-emissions growth at some future date, so it can backfill demand. And some other producers are even less sensitive to environmental concerns.

A supercycle is a compelling idea, but it's not a default state.

The last few were the result of many interacting forces, not one big trend, and certainly not just caused by a recovery from an economic contraction. If anything, we should spend more time thinking about the inverse of a supercycle: a long period where the one-time effects of catch-up growth are tapering off, while the world's population gets relentlessly older, shifting more resources away from goods and to services instead.


The Rise and Fall and Slight Rise of American Growth

Robert Gordon, whose Rise and Fall of American Growth is indispensable on the productivity slowdown of the last fifty years, has a new interview in which he speculates that productivity growth will accelerate in the 2020s. He attributes this to a few forces: efficient electric vehicle manufacturing at scale (fewer moving parts should lead to a lower cost, at least if other solid-state devices are any guide), AI, and working from home. The last is mixed; in-person communication is still higher-bandwidth, so productivity gains might be akin to catch-up growth: fast at first, but with a low ceiling on gains.

Financial Engineering

Valuation is always somewhat contextual: a subsidiary that's a money-losing drag on profits can turn into a spinoff that's valued as a high-growth lottery ticket. And, in a lending context: a Greek bank's bad debt book can be an asset manager's distressed debt portfolio instead ($, FT). The magic here is not that the loans themselves have a different value when they're on one balance sheet instead of another, but that they have a different effect on the owner's valuation: the higher a bank's bad debt expense, the worse its lending standards appear to be, so every increase in bad debt raises the market's estimate that there are more increases coming. A distressed debt portfolio, on the other hand, gets treated as one more chunk of assets under management that can continue to generate alpha. From the bank's perspective, one goal of raising capital by selling bad loans is to make it easier to raise capital by selling equity instead; better to take a loss on the loans than to have investors constantly speculate about what future losses will look like.

Financial Engineering, Pt. 2

I've written before about inefficiencies in the pre-IPO stock market. Unlisted companies may have supply and demand, but supply rarely meets demand. Supply and demand curves are a fuzzy collection of dots, not smoothly continuous lines. Another problem in the market is more prosaic: some holders of private company stock options don't have the cash necessary to exercise their options, even if they'd like to. EquityBee has raised $20m to help solve the latter problem, by essentially letting options holders pre-sell a fraction of their shares:

The way it works is fairly straightforward. EquityBee provides capital to startup employees so they can purchase stock options. The employees get money to cover the cost of exercising their stock options and the taxes. The investors who helped provide the funding so they could do that get a return, or a share of the profit, if there’s "a liquidity event." EquityBee makes money by charging an upfront fee from the investor on the investment day, as well as any carried interest upon a successful exit or IPO.

EquityBee is not just part of the process of making private company stock more similar to publicly traded equity; it's an accelerant. If the company is funding deals by raising from outside investors, this presupposes that there's demand from those investors for arbitrary pre-IPO startups. It's only worth it to allocate time or money to that market if there are meaningful opportunities. We can assume that there's not much interest in random investments; no one wants an index fund of pre-IPO companies weighted by their early employee attrition. But there are enough investors who can put serious money to work buying illiquid private companies that it's worth it to source less liquid shares for them, at the right price

The Decline in Federal R&D is Closer to a Shift

Nintil looks at the change in federal R&D spending as a share of GDP, which went from 1.86% at its 1964 peak to 0.61% today. That seems like a significant drop, but every way of slicing the data makes it look less meaningful: government spending on the R part of R&D has been remarkably steady in that time, while development spending has generally dropped. This might be a sign that there are closer links between abstract research and profit-seeking companies that productize it—links that schools like Stanford have historically encouraged. Looking at spending by category shows that it's also stable excluding space and defense. These kinds of spending have positive spillovers—satellites, computers, and the Internet are all nice to have, and all come from this kind of spending—but they also clarify that the decline in federal R&D wasn't because America lost interest in science, but because America won the Cold War.

Reflexive Palantir

One thing I noted in my Palantir writeup is that the company's high level of equity compensation makes it partly dependent on a high stock price. As long as shares are high, employees are earning above-market comp; a long-term drop in the stock price would cost them headcount and institutional knowledge, and probably force them to raise salaries, too.

In one sense, this makes Palantir a healthier version of GameStop. As long as employees have diamond hands, the price mostly goes up and the company's economic model keeps working. If they start selling, the company is worth a bit less. After the end of the lockup, a few early employees sold, and it's unclear how many will choose to follow that signal.

But the GameStop comparison doesn't mean this is an unsustainable setup. Successful companies often have bubble-like dynamics, and at the right pace they can be sustained indefinitely. If employees, customers, and investors all overrate a company, each of those parties can get its expectations fulfilled by the irrationality of the other two. A company can be modeled as a nexus of low transaction cost contracts, but it can also be modeled as a nexus of moderate, benign bubbles.

Hotel Flipping

Hotel owners have been providing seller financing to buyers ($, WSJ), allowing them to get bad assets off the books without taking a loss, in exchange for potentially taking a loss on loans later on. The WSJ thinks the trade is exchanging a capital loss for credit risk, but there's another scenario in which it's trading a capital loss for duration risk. A long-term loan loses value if rates rise, and rising long-term rates coincide with economic recoveries. It's another reflection of the gap between short-term cash flows and long-term valuations; operating a hotel in 2021 is still mostly a money-losing proposition, but selling one and lending to the buyer looks like a better deal.

In other real estate news, very high-priced Manhattan apartments are coming on to the market as hedge funds migrate out of the city ($, WSJ). One apartment broker—clearly someone with a lot of experience trying to relate to the extraordinarily wealthy—describes it like this: "If you think of New York City as a ballet, right now the city is at intermission... During intermission, some people get restless and don’t come back for the next act. That’s what’s happening now." You might think of it as a movie you give up on, or a sporting event you stop watching once it's clear that your team will lose. But if you’re selling something for eight figures, it's ballet.

diff.substack.com · by Byrne Hobart