Tech as a Keynesian Quasi-Boom Within a Keynesian Semi-Slump
A common macroeconomic narrative in the decade after the financial crisis was that we were in a period of persistent artificially low demand, and that looser fiscal or monetary policy would have led to a faster recovery. Peak-to-peak unemployment recoveries have been slowing over time; in the mid-70s recession, it took 18 months for total employment to reach a new high, while the timeline was 47 months for the dot-com crash and 75 for the financial crisis. Meanwhile, inflation averaged 1.4% annualized in the decade after Lehman's collapse. Clearly there was some room for more deficit spending, at least if the narrow problem was finding a way to accelerate economic growth.1
Lower rates were a side effect of this, and one side effect of those was that money flowed into higher-risk sectors of the economy where it was still theoretically possible to get a good return. That's one macro view on the startup boom: that it was less about underlying technology and more about persistently low rates that convinced investors to take risks they didn't otherwise want to take.
But another way to look at this is that the Keynesian model always finds a way: the effect of this flow of capital was to create an island of the economy where growth was high, where there was a significant multiplier effect in which incremental spending led to further spending, and where a persistent flow of dollars kept short-term losses at individual companies from tanking the overall economy.
One symbol of the somewhat self-referential nature of the boom in tech companies was the high valuations accorded to businesses like Brex and Ramp, which help startups manage their own internal spending. Ramp raised at a $1.6bn valuation in April 2021, at $3.9bn in August of that year, and at $8.1bn in March of this year. There aren't a lot of companies out there that can 5x a billion-dollar valuation in twelve months. But when money is pouring into new startups, the number of customers a company like Ramp can work with, and the revenue each customer generates, will both be accelerating.
I wrote a few weeks ago about the downside to this: that unit economics for companies that sell to startups are implicitly dependent on more funding for those startups. But there's an upside to it, too: the size of the tech economy grew thanks to the implicit stimulus of venture dollars. And this didn't just happen because of the profusion of software companies founded in the last decade.
It's worth taking a step back and asking: what is a "tech company"? One fun way to extend the semantic debate over what counts as "tech" is to ask whether Walmart was a tech company during its high-growth period. They didn't sell a lot of electronics or software in their early days. But they were early to using barcode scanners to track inventory, they automated inventory management, and they used this to run a more agile business with lower capital requirements than their competitors. So one possibly over-broad definition of a tech company is that it's a business that achieves persistent like-for-like improvements in the outputs of workers. A company that grows by hiring lots of salespeople instead of lots of engineers doesn't get disqualified under this definition, so long as it's consistently getting more out of them.2
So one way to look at the growth of the addressable market of tech companies is to say that it didn't just consist of increasing the size of the tech industry, but increasing the "techiness" of the rest of the economy. A go-to-market strategy that starts out learning how to land startups will, within a few years, evolve into one that is effective at maintaining long-term relationships with large corporate clients, since that's what some startups turn into. Selling to tech companies can be useful because they have high standards and often a high threshold for reimplementing things in-house that could be outsourced instead in order to ship faster.3
It's a great case study in George Soros' concept of reflexivity ($, FT), where investor sentiment doesn't just track fundamentals but actually affects them, by making earnings temporarily better when valuations are high and worse when they're low. ("Temporarily" because at first, capital inflows can create new revenue opportunities—new startups create new revenue for Zoom, Slack, Docusign, Datadog, etc., but eventually when there's too much capital, money gets invested in marginal projects that won't get a return even under the most optimistic assumptions. Actual wealth is being created in the initial stages, and it's very hard even after the fact to say when the boom led to malinvestment.)
Turning to the original argument, there's a strong counterpoint to the claim that deficit spending could have been higher in the post-crisis period without setting off inflation. And the counterargument is, of course, that we got a lot of fiscal and monetary stimulus in 2020 and now inflation is at generational highs. But when you dig into the drivers of inflation, one thing you see again and again—in housing, in fossil fuels, in shipping, in airlines—is that we went through a long period where companies were reluctant to add capacity because it wasn't clear that demand would be there. In tech, broadly construed, we do see high investment, and the big capacity constraint over time is labor (another way to rephrase "labor is the constraint" is to say "wages went up," which is generally a good thing).
Right now, we're arguably in the worst of all possible worlds macroeconomically. When governments don't engage in direct countercyclical policy, the economy has rapid booms and sharp recessions, which are unpleasant for the average person but do have a tendency to hasten selective pressure—if we were having a 2008 or 1932 once or twice a decade, instead of every two generations, the average CEO would be a lot more talented because all the below-average CEOs would lose their jobs. We've collectively decided that steady growth is better than breakneck growth and rapid reallocation of resources. But that steady growth depends on political will to keep demand stable. Instead, what we had was a long period where the economic growth model relied on governments to increase demand, which they didn't, and then got a clear example of what happens when they create too much demand and the supply isn't there to handle it.
But we can always go back to Keynes, who also gave us the wonderful term "Animal Spirits" for the human tendency to take fun risks even when strict analysis would suggest a more cautious decision. Those animal spirits always find a way to build the next bubble in the ashes of the last bust. And sometimes, the bubble ends up being a Keynesian-style economy in miniature, operating in a long-running state of quasi-boom. Which means that if the macro theory of the tech boom is at least partly right, and if inflationary policies have been seriously discredited by recent events, the future steady state of the economy will look a lot like the last ten years.
A Word From Our Sponsors
Import millions of rows of messy customer data in minutes
“Hey, can you send me a spreadsheet with the following columns at minimum?”
Sounds simple enough, but ask 100 customers to do it, and chances are you’ll get 100 different spreadsheets, documents or formats sent back to you. Now try collating all that data into one place, without losing your mind. Or worse, asking your customer to do it.
If this sounds anything like your team’s reality then you need to get some help from Flatfile — who have built the ultimate data importing cheat code.
Their innovative platform lets your customers drag and drop their data into your system without formatting beforehand. Then the magic starts. Smart matching, file hooks and algorithmic data detection means that a messy file that used to take hours can now be imported, cleaned and validated in literally minutes. The best part? You can embed Flatfile straight into your product, or in a prebuilt hosted page.
Blockchain research company Arkham has analyzed transactions from crypto lender Celsius, successfully deanonymizing a large borrower and apparently identifying transactions in which Celsius bought its own token while its CEO was selling. Financial collapses tend to produce a sonic boom of information, as people dig through evidence to figure out what was really going on, and start to tell their side of the story. This can happen faster in crypto, since transactions are pseudonymous but use persistent pseudonyms.
The Uber Playbook
A cache of 124,000 internal documents from Uber covering the period from 2013 through 2017 has been leaked, and, in a stirring example of radical transparency, a consortium of media companies appears to be keeping the full documents private but sharing a few excerpts with the public. The Guardian has a piece here.
If you were not reading any news about Uber from 2013 through 2017, a lot of this will be quite revelatory, but much of it is old news, which modern Uber has tried to distance itself from. The company's legal status early on was dubious, for the obvious reason that no such service had been possible before so it was either unregulated or technically covered by rules intended to apply to traditional taxi services. A good model of the business is that they were bootstrapping their way to legality, and the most cost-effective approach in their view was to operate the business according to what they thought the rules ought to be, and then try to craft rules that lined up with their operations. This is of debatable morality—and some of the methods they used crossed the line—but it's an important viewpoint to understand, because a company whose business plan is to raise money, spend it all on lobbyists, and only launch a minimum viable product once the rules are fully settled is probably going to fail.
A lot of what Uber did sounds like familiar tactics, but not from the business world:
The documents indicate Uber was adept at finding unofficial routes to power, applying influence through friends or intermediaries, or seeking out encounters with politicians at which aides and officials were not present.
This is a fairly apt description of how lobbying works: find the avenue through which you can accomplish policy goals, and try to meet with the people who can make a decision (and nobody else).
Any sufficiently transformative technology will make some existing rules obsolete, and the singularly-focused technology company is most likely to achieve this breakthrough. A bigger company has more lobbying muscle—if GM had decided to make ride-sharing viable, they'd have more members of Congress on speed-dial, but they'd also be risking their core business. The three options for widespread deployment are:
- Expecting legislatures or regulators to be forward-looking about the potential of new technologies—which can happen, but is not something to bank on.
- Making new things a monopoly of big business, not just because of their scale but because of their political clout.
- Rule-bending and rule-breaking startups.
Each of these will produce damning internal communications about corner-cutting, lobbying, and offloading risk to some innocent third party. The real question is which of these models is most likely to work in the future—because that tells us which model is most worth understanding.
Price Discrimination and Amortization
Starlink is offering Internet access to ships and oil rigs for $5,000 a month. This monthly cost includes some incremental expenses—the hardware costs $10,000 rather than the usual $600—but a lot of the pricing is pure price discrimination. The nature of Starlink is that it can provide access just about anywhere; if its target market is merely people in remote places who care about fast Internet, that's a limited market because most people who care deeply about fast Internet just won't live in those places. Starlink's use cases—merchant vessels, oil rigs, and yachts—are all assets with a large fixed cost and comparatively low personnel costs, so they're all cases where incremental spending to get either a little more reliability or one more amenity can be worth a massive premium.
In general, when there's a consumer-facing transaction the company with the most recognizable brand name is the one that has the most pricing power. Nobody buys Corn Flakes because of an affinity for the company that makes the cardboard for their boxes, or because they really like the farm that grows the corn. But this model starts to break down when there's a duel between two companies that each have pricing power and are both needed to produce revenue. Tesco and Heinz have just resolved a pricing dispute ($, FT) that briefly led to Heinz products not being sold at Tesco stores.
The grocery business is an interesting one because there's a mix of products that bring customers in the door (most famously Costco hot dogs or rotisserie chickens) and other products that flip the overall margin on a transaction from negative to positive. It's bad for either Tesco or Heinz for a recognizable brand to be unavailable in the store; they're both losing revenue and irritating customers. One possibility is that as inflation ratchets up, more consumer packaged goods companies are trying to force their way into being low-margin essentials from the grocery store's perspective, rather than profit drivers.
A few weeks ago, a gas turbine being repaired in Canada by a German company was blocked from being returned to Russia because of sanctions. With gas prices higher, the turbine is returning ($, FT), though this is pitched as a one-off expedient. The math of sanctions is that, like any trade restriction, they impose a cost on both sides, but asymmetrically, so the aim of the sanctioning country is to make it more painful for their adversary than themselves. (Retaliatory tariffs, for example, often target industries with political heft, like farmers or brand-name luxury goods—or, when they're imposed on the US, industries with lots of workers in swing states.) The paradoxical result is that the punishment for refusing to trade with a country is that the country refuses to trade something else, or, in this case, finds itself unable to do so.
Diff Jobs is our service matching readers to job opportunities in the Diff network. We work with a range of companies, mostly venture-funded, with an emphasis on software and fintech but with breadth beyond that.
If you're interested in pursuing a role, please reach out—if there's a potential match, we start with an introductory call to see if we have a good fit, and then a more in-depth discussion of what you've worked on. (Depending on the role, this can focus on work or side projects.) Diff Jobs is free for job applicants. Some of our current open roles:
- A company building a universal API for accounting software is looking for mid- to senior-level engineers. (Austin)
- A company building an emerging market bank is looking for payment operations and account operations people. (Worldwide, remote)
- A web3 company building a tool for customer engagement is looking for a senior engineering leader. (US, remote)
- Companies in the alt data space are looking for a data scientist and an equity research specialist. (US, remote)
- A company building a platform for investors to build thematic portfolios is looking for a senior data engineer. (NYC)
Which is not to say that's the only problem, or the only one worth solving. There are structural economic issues that keep growth low. But one way to frame structural issues is that they set the "speed limit" at which higher demand just leads to inflation, instead of to higher real growth. And it didn't seem like the energy that could have been directed to increasing demand was instead used to do things like reducing the impact of zoning, reforming the FDA, speeding up environmental reviews for new projects by an order of magnitude or two, encouraging more research into alternative energy and energy storage, increasing port capacity, either reducing employer credentialism or reducing the cost of credentials, or taking other actions that would raise the ceiling on growth. ↩
This is also a useful definition because it lines up with how we think about the impact of technology in macroeconomic terms. A good way to describe a technologically static economy is to say that output comes from either more labor or more capital. Over long periods, that model doesn't line up with what actually happens; aggregate growth is higher, leading to the addition of total factor productivity to the model. (TFP is right up there with "alpha" in the annals of model residuals that turn out to be the most useful part of the model, and the hardest to increase.) ↩
Small companies with selective hiring are probably more capable than average at implementing simple services on their own—if nothing else, such a project is easier to coordinate when the users and the engineers are the same people, and everyone can fit around a single conference room table or in a Zoom call that doesn't mute by default. But big companies have a stronger internal interest in building things themselves that they could buy externally: building such a product justifies headcount, and headcount is a proxy for power among people who don't directly produce revenue, like IT departments at non-tech companies. ↩