The 2x2 Inflation Debate
Plus! Valuementum; Quants and Lumber; Cheap Hacks; Tax Adjustment
This week, paid subscribers also saw:
What Happened to the Outsourcing Threat? asks why labor cost differences didn't cause all the software jobs in the US to migrate to much more affordable countries. It also introduces a novel pop culture-based market-timing tool.
Japan: Stock Picker's Paradise looks at how the Japanese stock market matches the same pattern seen in post-bubble hangovers in other places and at other times: there's high dispersion in prices, low average prices, and financial engineers are finding new ways to close the valuation gaps.
Who Amazon Grows With: Airbnb recently talked up how they cut their AWS bill by almost $100m each year. I wrote about why that's good for AWS.
Thanks for reading The Diff. If you'd like to support it, please subscribe.
Coming next week on The Diff: Payments & Fintech week. I'll be looking at some pre-Internet networks effects businesses, explaining a lovable bank, and writing about Stripe. Subscribers will also be able to ask questions for next Friday's Q&A call, featuring Stripe's Patrick Collison.
Today is the weekly free edition of The Diff, going out to 21,812 readers, up 60 from last week.
In this issue:
The 2x2 Inflation Debate
Quants and Lumber
The 2x2 Inflation Debate
Inflation is an exciting debate again. Google trends shows somewhat elevated interest, at least once you adjust for some seasonality.1 And there’s endless commentary on price levels. But the inflation debate is really two simultaneous debates:
Price levels will rise more in 2021 than they did in previous years, as almost everyone agrees. But is this temporary, or is there a new normal of higher inflation ahead? And, if so, is it "higher inflation" in the sense that it'll be on the high side of 2% rather than the low side, or in the sense that prices will ratchet upwards at an accelerating pace? Once you’ve name-checked the 70s, you might as well name-check Weimar or Zimbabwe. (Hyperinflationary spirals apply to exaggerations as well as price levels.)
The second debate is over whether or not high-but-not-runaway inflation might be a good thing. What if, ironically, record high prices for lumber and a persistent shortage of chips indicates that inflation has been too low for decades, and that we ought to rev things up a bit?
The Fed, at least, does not buy the argument that inflation will accelerate uncomfortably in the coming years. In Wednesday's statement, they attributed the rise in inflation to "transitory factors." This, naturally, does not offer much comfort to people who worry about inflation, because they're often reacting to fears that other people—the ones who control the price of money—are insufficiently worried about it.
If monetary policy and fiscal policy together determine the arc of inflation, then there are other reasons to be worried. Two trillion here, two trillion there, two trillion over there, and pretty soon you're talking real money, even in the context of an economy that can produce $22 trillion worth of goods and services and, in a good year, collect around $2tr in federal taxes and another $2tr in state and local taxes. If these spending bumps happen—and, to be fair, they're spread out over time—and they aren't offset with taxes, then the US will be running larger deficits in the next few years, which is one way to create inflationary pressure.
One mechanism for this is that higher transfer payments change incentives for low-wage workers. Many restaurants have complained, loudly, that they can't find people willing to take traditionally low-paying restaurant jobs. Since restaurants have existed for centuries, through plenty of economic cycles, including inflationary ones, they will presumably get around to raising pay and raising prices to compensate, and we'll all get used to slightly higher prices for meals. $2 more for entrees may be the price of a lower Gini coefficient.
Among restaurants, Wingstop has a lot of experience dealing with commodity swings, since their results are so keyed in to the volatile price of chicken wings.2 On their last earnings call, analysts asked about whether they had trouble handling inflation and hiring workers. They didn't, but they say their suppliers do. If the chicken processing industry has to raise wages to attract workers, that will eventually show up in higher chicken prices, but unlike some other kinds of inflation, it is indeed transitory: the capacity is there in terms of equipment and raw materials, but the willingness to admit that the low end of the labor market doesn't work the way it used to has lagged.
But wages are not the only place where inflationary pressures show up. The other way they show up is when there are capacity constraints, and a "capacity constraint" just means that investments have been withheld because of the perception that there wouldn't be enough demand to make them viable. The simplest theoretical way to model supply and demand is through smooth, intersecting curves, but in practice the curves are choppy—you can't build a tenth of a steel mill—and they're three-dimensional, because investing decisions today affect capacity and future profit margins for as long as those investments are productive.
This is a drum that Joe Wiesenthal of Bloomberg has been banging for a while: lumber prices are up, but timber prices are not, because the problem isn't a shortage of trees, but a shortage of sawmill capacity. One sawmill operator reported earnings yesterday, and said that profits were constrained by logistics problems. The demand is there, but the supply is not. And for the supply to get there, demand has to stay high—which, for now, is inflationary.
A similar argument can be made, and has been made, regarding chips. The current shortage is not just due to high demand, but to the recent expectation of low demand, which led chip fabricators to underinvest in capacity. Historically, the industry was cyclical. The winners knew when to build out, and when to cut back, and whoever expanded fastest just as demand was peaking would likely get wiped out in the next downturn.3 But right now, we're arguably in a secular growth cycle for chips: more devices use them, more of those devices talk to one another, and there are very data- and processing-intensive applications that are scaling rapidly. As far as capital allocation decisions are concerned, any cyclical swing that lasts longer than the economic life of capital assets might as well be an indefinite secular trend. And since nobody expects their car or phone to get dumber, not to mention their TV, dishwasher, doorbell, factory robot, fire alarm, watch, or earrings, the chip growth story is a growth rather than cyclical story.
At least, it is as long as the economy is growing. Dumb cars are getting replaced by smarter ones, but the car replacement cycle is very sensitive to overall economic growth.
The pro-inflation argument right now is that a credible fiscal and monetary commitment to growth would lead to higher investment. Companies can certainly get whatever financing they need, and inflationary shocks are bad for equity prices, so corporate America's interests are individually and collectively served by buying back less stock and making more capital investments.
But this argument, too, has limits. High investment now is the way to buy lower prices later. But one persistent problem the global economy has had over the last two decades is a supply glut: some countries engage in export-focused, financially-repressive strategies, which allow them to sell lots of goods abroad with little demand at home. For any one country, that's a good way to create high-paying, productivity-enhancing manufacturing jobs. But if everybody does it, there's not enough demand for the products those factories produce. It may still work—all those trillions have produced an uptick in inflation, but not uncontrollable inflation, so the US government can keep spending money and running deficits for a while. But the end state might be the same problem, only bigger: if the investment prompted by that spending happens mostly in countries that have high savings rates, high investment, and relatively low consumer spending, then the US's inflation-producing policies just enhance the global economy's deflationary supply glut.
Looking at the parts of the economy that the non-financially repressive countries specialize in points to some other issues. One is "shrinkflation," and the general problem of how to measure price levels. I wrote about this last year: some companies have found that it's better to keep prices constant and change products than to keep products constant and change prices. In food, this means using cheaper ingredients or selling smaller packages at the same price. In software, it means moving some features into higher-priced tiers. Mature and consumption-heavy economies have a lot of this, both because the talent that could be focused on optimizing manufacturing is instead focused on optimizing selling and because these economies produce more SQL-queryable bits per dollar of GDP.
Price discrimination is another driver of value-creation, or at least market cap-creation, in rich-world economies. It can cause lower inflation, since markets can be expanded through lower-priced options. And then widespread adoption of these can cause higher inflation. Ride-sharing was focused on black cars before it expanded to cheap rides, but once it was available at every reasonable price point, the cost started creeping up again. New ad products temporarily create cheap clicks that can drive economies of scale for their customers—King Digital was basically just Candy Crush, but the success of Candy Crush let them test hundreds of other game concepts on their desktop site, which were deliberately under-monetized in order to gather data to find the next Candy Crush. But the cheap clicks are an arbitrage opportunity, and markets resist arbitrage opportunities. Over time, those clicks start to approach their fair market value, and new ad platforms become a way to scale with known economics rather than a way to improve underlying economics. The big platforms that create temporary supply gluts in ad inventory also earn dominant data moats, and when those businesses mature they can collect a larger share of the marginal profit their customers earn from ads.
All of this makes the debate much more complicated. And at a sufficient level of complexity, 2x2 matrices break down. The real debate may turn out to be: can we measure inflation in such a way that it captures what we care about and gives policymakers guidance about what constraints they face? Or does the future hold a 2x2x2 matrix: is it transitory or not, is it good or not, and can we even define inflation, anyway?
An exchange-traded fund tracking high-momentum stocks may soon dump its high-growth names for companies like Capital One and Wells Fargo ($, FT). Such news is statistically inevitable: growth stocks can't outperform forever, and when out-of-favor categories snap back, they still take a while to lose their reputation. (Probably because their reputation is essentially market cap-weighted; a lot more people paid attention to energy stocks in 2014 than in 2020, for example, so most people who stopped watching them made up their minds that energy is a terrible sector and haven't checked back in since.) Still, it's interesting to think of statistically cheap companies that can also be categorized as value stocks. There have been value booms and value busts—when value as a category starts working, more money flows in, and eventually every cheap-looking company gets a bid. Value busts make sense during technological transitions, and one of the worst drawdowns in value stocks was in the 1930s, perhaps in part because growth industries were more flexible in adapting to a depression. The changing holdings of a momentum ETF are a good reminder that trends, even durable ones, don't lead to rising prices forever; at some point, however good the news is it's fully priced in.
Quants and Lumber
RCM Alternatives notes that lumber prices are not being driven by momentum-chasing quants, mostly because lumber futures don't trade very much: they do $82.5m in volume per day, compared to $12bn for copper and $418bn for the S&P. Often during sustained price increases, momentum-based quant strategies get blamed for the most extreme price movements (partly because, as the previous story indicates, they start selling the assets that are already down to buy the ones that have been going up.) This does not mean that lumber's extreme price moves aren't due to quant-like behavior. After all, systematic momentum strategies are only profitable if they approximate human behavior in advance of when it happens, so at best you can say that the algorithms make prices reach their peak earlier—but also make it lower, because those same algorithms subsequently sell to humans.
Solarwinds, famously, suffered what could have been one of the biggest breaches in history when its Orion product was exploited to give hackers access to networks including the US treasury, the Department of Commerce, Microsoft, and security firms. While most of the Fortune 500 used SolarWinds at the time, most of them were not on the list of victims; it turned out to be a national security rather than economic security issue. As it turns out, at least for now, the organizations that are best at performing such attacks are not doing so in order to demand Bitcoin ransoms or vandalize data.
SolarWinds is also a public company, and recently reported earnings. They reported an 87% renewal rate this quarter, down from the 90s but still good, and their largest customer—a government customer—actually agreed to spend more. SolarWinds expects $20m in post-insurance costs from the breach, and to spend $20-$25m annually on security thereafter. Which is a fairly low cost for an issue that could potentially have given an adversary access to numerous government agencies and most of corporate America. SolarWinds does not seem like an outlier in security-consciousness, and they're responding to the same incentives as everyone else: if breaches are cheap, preventing them will get a low investment, at least before it’s unavoidable. Some companies run into security crises which, if they survive them, lead to a relentless focus on safe programming at the cost of iteration speed and flexibility (think Microsoft after the age when viruses ruled the earth). For other companies, the potential cost of a breach is multiples of their revenue, so they have to be the most secure part of the stack (as in payments). Or the most valuable and vulnerable data as a complement to a high-margin business (Google will keep Gmail secure, not because Gmail itself is a great business, but because it makes search a better business). For many other companies, it's still optimal to underinvest in security and hope that someone else is the best target of opportunity. If legal liability rises to match the potential economic cost of these attacks, though, a lot more companies will invest a lot more in keeping their customers’ data safe.
In other security news, a VPN used by "more than a dozen" federal agencies has been hacked. Results to be determined.
Outside of the asset management industry, there's a widespread assumption that asset managers take advantage of tax loopholes to defer or avoid taxes on their profits. And it's true that many investment vehicles are structured to minimize tax burdens, often by cleverly using parts of the tax code in ways that don’t reflect their original intent. But within the money management industry, that stereotype mostly applies to people who invest in real estate, a business with many more opportunities to indefinitely delay paying taxes. This may begin to change soon: 1031 exchanges, which allows real estate investors to avoid paying capital gains taxes if they reinvest their funds in other real estate projects, may be eliminated ($, WSJ).
The defense of 1031 exchanges is that they encourage growth, because they keep people spending money on new property developments instead of cashing out and enjoying their gains. Which embeds two assumptions:
It's generally better to tax consumption than investment, and
Real estate investment is a particularly worthy kind of investment to avoid taxing.
Assumption #1 sounds true, but is circumstantial. Assumption #2, though, is hard to defend. Real estate speculation does produce jobs, but it also produces macroeconomic volatility and sometimes threatens the financial system. From a macroprudential perspective, where the goal is to reduce the odds of financial crises, it might make more sense to have 1031 exchanges for everything but real estate: sell your company, and you can roll the money into starting a new one; sell a mall or skyscraper, and you get taxed.
But it's always fiendishly hard to predict the long-term incentives created by a change in the tax code. Any tax on realizing gains, for example, is implicitly a subsidy on borrowing against appreciated assets instead of realizing those gains. If that's true, the net effect of eliminating 1031 exchanges would be that real estate portfolios would turn over less often. If we assume that people vary in their ability to make good real estate investments, this would mean that the best such investors wouldn't make as many discrete investing decisions, which would make prices a bit less efficient. Which might be a reasonable tradeoff: making real estate investing a less tax-optimal choice could be a fair trade in exchange for making real estate prices less reflective of their value. But it's still a tradeoff, not a straightforward benefit. Quirks in the tax code become load-bearing over time; even if they didn't make economic sense when they were made, the structure of the economy only makes sense in light of the tax incentives that economic actors have already responded to. If you assume that people are reasonably good at reacting to incentives—or, more plausibly, that over time the people who are good at doing this end up controlling more assets—then any change in those incentives has complicated and unpredictable results.
You can't always read the zeitgeist from Google Trends. Searches for "inflation" follow a particular seasonal pattern: high in the fall and spring, very low in the last ten days of December. This indicates that the population that students are searching for these terms. You'll see a similar pattern in other school-adjacent searches, like "hypotenuse" and "mitochondria" and "Adderall."
It’s hard to hedge exposure to wings, as opposed to the whole chicken, and since wing supply is partly a function of the supply and demand of other parts of the chicken, you can’t just hedge by betting on chicken or feed prices. Google “chicken wing futures,” and the #1 result is—Wingstop’s stock quote. Conceptually, Wingstop is partly a managed-futures vehicle that executes calendar spread trades in an illiquid corner of the agricultural commodities market.
That cyclicality was, in part, driven by supply constraints. When demand was hot and backlogs extended, buyers would double-order so they could be sure they'd get something, and then cancelled redundant orders later. Since a shipment of $x could wipe $2x off the industry backlog, demand as well as supply contributed to volatility.