Trapped by the Wrong Inflation Stories · by Byrne Hobart

In this issue:

  • Trapped by the Wrong Inflation Stories

  • Amazon's Triumph of Investor Relations

  • The Scaling Problems Aren't Obvious

  • Money- and Time-Weighting

  • Coinbase Comp

Yesterday's CPI number was a surprise, albeit in the same way that tossing a coin and getting heads four times in a row is a surprise: improbable, but not outside the bounds of the plausible. Overall, the CPI is up 4.2% year-over-year, the highest number since 2008.

Digging into the numbers, there are two forces that drove the inflation number:

  • The year-ago comparable period, April 2020, showed month-over-month deflation of 0.8% as large parts of the economy shut down. May's CPI was 0.1% lower than that. Year-over-year inflation was 1.7% in February, 2.6% in March, and 4.2% last month, but the numbers compared to 2019 are +4.0%, +4.2%, and +4.5%. An acceleration, but a modest one.

  • Digging into the inflation numbers, the two most important categories were 1) used cars, and 2) travel. Used car prices rose 10.0% month-over-month, an all-time record. Lodging prices are up 7.6% m/m, and airfares rose 10.2%. What all of these have in common is that they're temporarily supply-constrained—a used car is, by definition, something you can't make more of on demand—and that they benefit from catch-up spending post-Covid.

It's possible that the supply constraints facing the travel business are, over the next year or two, inflationary: hotels are exactly the kind of low-wage business that will have trouble hiring if the economic minimum wage is up and their expectations are anchored to the statutory minimum wage instead. But they're not part of a long-term trend: assuming these businesses have trouble hiring because the lowest-paid workers are getting pandemic unemployment insurance, then a) those benefits are at least intended to be temporary, and b) even if they're not, it's a one-time step function increase in cost, not a recurring one.

Part of what makes talking about inflation hard is that there are basically two case studies that come to mind:

  1. The US's "gradually, then suddenly" bout of inflation starting in the 60s and accelerating in the 70s, and

  2. Various historical episodes of extreme hyperinflation.

The hyperinflationary episodes were mostly in countries that had foreign currency-denominated debts, and were small economies. They don't apply to the US. A reserve currency issuer experiencing hyperinflation would be historically novel. The closest example might be the UK, but the UK's inflation, while higher than the US's in the 70s, never quite reached hyperinflationary escape velocity. British inflation peaked at 24% in 1975, but within a decade it was only slightly higher than US levels.

The 60s/70s inflationary episode is an interesting example for US discourse to fixate on, for a few reasons. First, one reason it's salient is partly related to what caused it: the Baby Boomer generation remembers this period, but they were also partly responsible for it, since they started forming families (and buying houses, cars, and appliances) right around the same time. The 70s were more of a redistributive decade than a slow-growth decade. Real GDP growth was 3.2% annualized, and real wages for the lowest-paid workers improved. Equity investors and (especially) bond investors had a terrible time, but the underlying growth was there.

A more relevant inflationary period to today might be the late 40s. The dynamics at play:

This was a difficult adjustment for the US economy to make, but it was also a temporary one driven by known factors. Shrinking an army from 12 million to 1.5 million—i.e. adding new workers equivalent to 7.4% of the population, all while cutting government spending on munitions, planes, tanks, etc. was a daunting macroeconomic problem. But it was also a known unknown; nobody who saw shockingly high unemployment numbers, or who saw Boeing engaging in mass layoffs while companies that made civilian economy goods couldn't crank them out fast enough, wondered what the cause could be. Annual CPI growth was running at upwards of 10% in 1948, went briefly negative during the subsequent recession, rose to almost +10% again in 1951 as Korean War spending ramped up—and, by 1960, entered a phase of incredible stability, with sub-2% but greater-than-0% inflation for the first half of the 1960s.

We're in a similar situation today: there's a set of workers who will definitely return to work (perhaps with some of them working in Amazon fulfillment centers for $17/hour instead of working at restaurants and hotels). And there's a large deficit that will shrink as a) time-limited programs wind up, b) tax revenues rise due to growth, and c) the political will for expansive spending programs declines. This is not a view that equity markets embrace, but direct inflation proxies did not respond to yesterday's numbers by acting like 4.2% annualized inflation is the new normal. 10-year yields rose a bit (1.69% up from 1.64%, but still quite low by historical standards). Gold dropped. Bitcoin dropped.

High volatility within a known distribution is not the same thing as unknown volatility within an unknowable one, and in terms of inflation, we've seen a movie very much like this one before.

Disclosure: I own AMZN and Bitcoin.


Amazon's Triumph of Investor Relations

Amazon issued bonds on Monday with yields as low as 0.1% above comparable treasuries, a record ($, FT). Amazon has been saying for years that they manage to long-term free cash flow, not to short-term GAAP profits. And they've emphasized that their growth model is predicated on investing heavily in small projects that can compound for a long time. This model has gradually worn down the skeptics, but it's hard to infer the distribution of opinion purely from a stock price. Amazon's price could be where the market clears when there's a majority of bulls who think it's worth $5,000, and a sizable minority of bears who think it's worth $1k at most, and might run the risk of being worth $0 if it makes a big enough long-term bet that doesn’t pan out. But bond markets show the market's view of the distribution of worst-case scenarios, and according to them, the Amazon strategy is as viable as it looks in retrospect and as long-term focused as it says it is.

The Scaling Problems Aren't Obvious

Politico has an overview of efforts to get more Americans vaccinated. US vaccination rates have stalled; vaccination rates peaked at 3.3m/week in early April and are now 1.9m. Different levels of vaccination have different hurdles: producing one effective dose is an R&D problem, delivering over 115m doses in the first three months is a manufacturing and logistics problem, and vaccinating the last 50% of the country is a marketing problem. As it turns out, at least in this case, the US was better at inventing things and manufacturing them than at selling them, which is not the usual pattern.

Money- and Time-Weighting

It's practically an iron law in finance that time-weighted returns are higher than dollar-weighted returns: a fund that performs well will raise more money, and that subsequent performance won't necessarily be as high as the performance that got them all that attention in the first place. (There are investors who have a good lifetime track record but generated negative lifetime returns in dollars because they peaked in assets under management at the worst possible time.) ARK, for example, generated 5.24% annual returns since inception weighted by dollars ($, WSJ), underperforming the S&P. It's not quite a fair comparison, since the people who didn't put money in ARK when it was low, but added money when it was high, wouldn't necessarily have had any better luck timing the S&P. But it does illustrate the challenging math of long-term returns. They revert to some kind of mean, but the easiest time to raise money is after a run of good luck.

Coinbase Comp

Like Lyft and Stripe before it, Coinbase is moving to one-year vesting for equity and making some other compensation adjustments: instead of negotiating individual offers, it's benchmarking compensation to the 75% percentile of peers and making fixed offers. There are some industries where job offers happen in an extremely narrow band, like the famous bimodal graph of lawyer salaries, where the mean is $100k and the mode is the number that big law firms uniformly offer. (To an outside observer, this looks a lot like collusion. Sadly, the good antitrust lawyers seem to have a conflict of interest here.)

One reason salaries might cluster like this is that it's relatively easy to judge whether someone is past a certain skill benchmark, and hard to measure what their ultimate contribution to the company will be. It's an especially good model if employee productivity varies based on their fit with the rest of the company: in that view, the goal is to overpay in year one in order to identify people who fit in unusually well, and then give them raises that a) beat what the competition would offer, but b) underpay them relative to how productive they are. Paradoxically, the more companies vary in how much productivity they can get out of an individual worker, the more it makes sense for wages to converge on some set high level. · by Byrne Hobart