Cars Manufacture the Modern Middle Class

Plus! Not Enough Housing; Durable Network Effects; VPN Integration; Reserve Currency Problems; Movies

This post begins The Diff's series on the auto industry. The car industry is about 3% of global GDP, but, as we'll discuss, its economic impact is larger than that. When countries are growing, cars are often a bigger share—the auto industry was roughly 20% of US economic output in the 1950s, for example, and plenty of other economic activity is downstream from the industry: suppliers, of course, but also borrowers, advertisers (cars are the single biggest category of TV ad spending in the US). The industry has gone through serious transitions before: from craft production to mass production, then from US dominance to global competition, and now from internal combustion to electric vehicles. Let's take a look at what this means.

Cars Manufacture the Modern Middle Class

In 1914, Henry Ford famously offered workers at his Highland Park plant a $5 daily wage, approximately twice what they'd previously earned. Part of his argument was that workers who made Ford cars ought to be able to buy one, and that's probably accurate: throughout his life, Ford was an advocate for high consumer spending (he once changed a proposed advertising tagline, "Buy a Ford and Save the Difference," to "Buy a Ford and Spend the Difference"). Ford had made an immensely popular product, but like many visionaries in the design of physical products, ended up spending more and more of his time managing supply chains instead.1

But it was also part of a wider effort by Ford to enforce particular values on his workers. As it turns out, there were some footnotes to the $5/day offer:

Part of this was tactical, and part of it was ideological: the pay plan was proposed to deal with high absenteeism and turnover (10% daily and 370% annually, respectively). Workers with families depending on them were less likely to quit, so it made economic sense to bid up for them. Assembly lines have a hard time tolerating inconsistent attendance, since it means over-hiring for whatever their true needs are, and assembly-line work also can't tolerate an employee who fails to keep up because he's nursing a hangover. So there were some eminently practical reasons for Ford to hire workers for a "sociological department" that would investigate employees' home lives and report back to the head office on what they found.

But it was also a decision based on Ford's on highly idiosyncratic political views, which, aside from the notorious ones, included: advocating pacifism and non-intervention during the First World War, promoting dairy-free diets (Ford researchers created an early form of soy milk, and also used soybean oil for industrial purposes), writing a wild anti-smoking jeremiad called "The Case Against the Little White Slaver," and an abiding love of the McGuffey Readers. What Ford was trying to do with the $5 day, and what other auto companies were ultimately able to do, was to impose a set of more urbanized, middle-class, and American-normed values on workers who were either agrarian locals or recent immigrants.

An important part of middle-class existence is being highly respondent to institutional cues: showing up at work, getting paid a steady salary, and paying bills on time basically define the economic side of the middle class ideal. The car industry created a middle class out of its customer base both by creating a set of high-paying jobs which, during the heyday of the unions, had high job security, and by getting tens of millions of people used to the idea of making monthly payments on a big capital asset that gave them access to work as well as leisure. An agrarian worker has external cues that tell them what to prioritize when, and their economics are more self-contained; mistakes cost them directly, but don't get externalized as much. And craftsmen have similar incentives: if they start with raw materials and build a finished product, a delay or mistake affects the single piece they're building, but doesn't affect another worker making something similar. When work gets divided up and specialized, though, minimizing variance matters more, and auto manufacturers imposed this economically through how they hired and fired, and in Ford's case did so in a more granular and intrusive way.

This was not unique to the US. The early history of Japan's auto industry is also a history of figuring out a new set of postwar social norms: Japan's 1947 constitution, written under the supervision of US occupying forces, weakened many traditional social structures, and, among other things, explicitly allowed workers to form and join trade unions. It took a while to figure out what this meant: at Industrial Bank of Japan, one group of workers formed a union whose key demand was for their manager not to put his feet up on his desk. (Management conceded.)

At other companies, unions were more explicitly communist and more militant. "Militant" is often a sort of flavorful word for unions, that can mean that they're a bit tougher when they negotiate, so it's useful to go into detail. They'd sometimes ambush non-union members and subject them to days-long "trials" demanding that they sign up, and sometimes occupied factories. And the early unions were powerful. A 1960 mass strike involved 6.4 million people. Meanwhile, law enforcement and companies were more "militant" themselves, sometimes relying on organized criminal syndicates to intimidate, beat, or even kill striking workers.

The arrangement companies like Nissan and Toyota ended up with was captive company-level rather than industry-level unions. These unions gave workers some level of representation, and gave them seniority-based wage scales, but limited their ability to strike. In a US context, this would be a fairly zero-sum transaction: less compensation for labor meant more for shareholders and management. But Japanese management tended to have much lower pay than US managers, and Japanese companies overwhelmingly reinvested profits rather than paying dividends.2

Japan was deeply impoverished at the start of this process: the same IBJ manager who dealt with union troubles before was once sent to a satellite office, where the most desirable place to work was the room where a fire kept stew warm for employees' lunch; every other room in the office was frigid. That level of poverty meant that continuous reinvestment could easily lead to high GDP growth, first in building a steel industry and rebuilding infrastructure, and later by expanding production of cars and eventually electronics. Even if this GDP growth was weighted more towards investment than consumption, it did lead to higher consumption—workers who were underpaid relative to other poor economies were similarly underpaid relative to other middle-income economies, so they wound up trading a lower immediate standard of living for a higher growth rate.

The auto industry has outsized importance in countries that are rising to middle-income and higher status: not only is it a sector that can be incrementally indigenized (starting with assembly and then finding local sources for more and more parts), but national prosperity benefits from the one-time effect of increasing car ownership. From the 1950s through the 1970s, population growth and rising automobile penetration were a significant chunk of demand: cars per thousand people rose from 222 in 1945 to 380 in 1955, 467 in 1965, and 640 in 1975. Meanwhile, many of the mid-century and later manufacturing-driven economic growth spurts coincided with above-average population growth (simultaneously creating a new class of economic producers and consumers—and inducing some families to upgrade cars as their needs grow from two-seaters to minivans). Once cars are ubiquitous, that growth slows, and more sales are for replacements and upgrades instead of new vehicles. The industry needs enough scale to satisfy local demand while it's growing, but that means fixed costs are generally too high once that process finishes. And the car industry had high fixed costs, for two reasons:

  1. Mass production achieved economies of scale by producing as many identical components as possible, with high costs for switching a die from producing one component to another and very high costs for shifting an entire production line from one model to the next. Every model switch was basically the second-system effect, run every single year, with inevitable bugs. This cost was only worth paying if a few weeks of downtime, slow production, and subpar quality could be followed by most of a year of predictable high production. They could insulate themselves from some macro problems by pushing dealers to order more cars during slow periods, but this model also required high levels of inventory since so many products were produced in batches.3

  2. Deals with unions made it hard to fire employees, even when business was slow, so car companies mostly tried to power through periods of slack demand, letting finished cars accumulate either on their books or, ideally, on those of dealers—or even using financing to push them onto consumers' balance sheets instead. Inflexible labor deals meant that the cost of idle plants was not too far from the cost of full-capacity production.

So as the country as a whole gets richer, the car industry’s role shrinks: it was 20% of GDP in the US when America had a manufacturing-centric economy with a comparatively short list of accessible consumer goods. Now, there’s a lot more to spend money on—mostly services, but also a wider variety of physical things—and cars are about 3% of GDP directly. So the late stage of the auto industry is an inversion of its earlier role: as it shrinks, it tends to lead to rising inequality, because those high fixed-cost jobs aren’t viable any more once car sales normalize.

The natural question is how much the values systems inculcated by these companies, whether directly or indirectly, stick around after the companies themselves become less significant. Even before the US car industry struggled with competition from imports, it had trouble keeping workers motivated. On the other hand, a lot more of the US workforce is on some kind of tracked career progression than was the case when Ford was starting up—we may not technically have seniority-based roles, but most of us expect to be earning more money over time in whatever we do. So that aspect of the growth of the auto industry might be one of those permanent pieces of infrastructure laid down by a temporary industry boom, like the US’s still-excellent freight railroads (courtesy of 19th century overbuilding) and the early 2000s’ abundant fiber connectivity (available at pennies on the dollar from bankrupt 90s darling telecom plays). This set of middle-class values is not necessarily ideal, from either a moral standpoint or a more aesthetic one. But that makes them a bit like Henry Ford’s $5 day: soulless and boring, but still quite financially rewarding for everyone involved.


Not Enough Housing

Apracitas Economics as a good piece on how post-financial crisis homebuilding has fallen short of long-term trends. Combine that with favorable demographics—more thirtysomethings mean more demand for single-family homes, while twentysomethings create demand for apartments—and it implies a significant housing shortfall. There are near-term supply-chain reasons for this, but the long-term trend has been towards more restrictive housing rules in the most desirable places.

There's a good demographic and financial case to be made that the 2020s will be a fundamentally-driven reprise of the financially-driven 2000s: a period of unusually high housing demand, reflected in some combination of new homes built and existing home pricing. In this version, the possibility of remote work makes the national housing market more elastic: yes, the best places to work are generally the worst places to build, but it's more feasible for people to leave housing-constrained cities to start families somewhere else.

Durable Network Effects

Years after World of Warcraft ceased to be the most compelling online role-playing experience, some players kept coming back, less for the gameplay and more for their friendships. This is the late stage for many online services: users no longer come for the tool, but still stick around for the network. And now it's happening with niche communities on Facebook. Newer communities might be more likely to form on Reddit (though it's hard to get good data on this), but the ones that formed by default when Facebook was at peak relevance still have more user density, and for those users, churn doesn't mean switching to a cooler site but abandoning their friends.

This story is also a testament to Meta's ability to navigate changes in how people use social media, and in its own popularity: Groups were central at Facebook F8 in 2019:

There are tens of millions of active groups on Facebook. When people find the right one, it often becomes the most meaningful part of how they use Facebook. Today, more than 400 million people on Facebook belong to a group that they find meaningful.

In retrospect, this was a way for the company to concede that the days of rapid growth on the core app were over, and that controlling user attrition was a paramount concern. Still, most of the cash flow a good business creates will be in that terminal stage, when there's no longer a good way to drive growth but there are many levers to control the pace of decline. That's the stage when cash flow can be maximized. As a standalone company, the Facebook app would be paying a nice dividend, shrinking its outstanding share base, or both. Since it's under the Meta umbrella, that cash flow goes to areas where they still see an opportunity for growth.

VPN Integration

One feature of the VPN industry (Diff writeup here ($)) is that its dubious reputation means that VPN-specific brands historically didn't face much competition from larger tech companies. That's starting to change: Microsoft is testing out joining Chrome and Opera in bundling VPN services with its Edge browser. Big tech companies with complicated business models have an interesting privacy calculus: they can offer features that limit users' data leaks to arbitrary third parties, while still tracking the same data themselves. This ends up growing those companies' data moats while making it harder for other businesses to find a viable targeting model—and it still gives them pro-privacy talking points.

In other VPN news, ExpressVPN and Private Internet Access are being sued by a group of movie studios, accusing them of enabling piracy. This is a hard charge to dodge; VPNs don't like to mention this use case directly in their marketing materials, but their affiliates certainly don't mind.

Disclosure: I own shares of MSFT.

Reserve Currency Problems

China has historically exercised tight control over the value of the yuan, both as part of trade policy and to reduce the risk of capital outflows. That's harder as yuan holdings globalize, since foreign sellers are harder for the government to influence. China is in the very unlucky position that the dollar is strong at exactly the same time that the Chinese economy is unusually weak due to lockdowns, and given the country's persistent Zero Covid rhetoric, it's hard to imagine the kind of policy shift that has taken place in other countries where the cost of lockdowns becomes intolerable. China's process of integrating with the global trade system prized stability over growth (which wasn't as visible from the outside because China's manufacturing scale was so vast, and the pace of controlled growth was so high).

For earlier coverage on whether or not the yuan could become a major reserve currency—and a thought experiment on a bold policy move that could make this more likely while addressing another weakness in China's financial system—see this post ($).


One of the surprisingly durable parts of the pre-Covid economy was theaters: only 2% of screens have closed since the start of the pandemic, despite an 81% drop in US box office revenue in 2020, and even with 2022 numbers tracking 45% below 2019's total for the same time period. One part of this might be the fact that it's hard to repurpose a movie theater for anything else—a defunct nail salon can turn into a node in a local delivery company's network, or a location for a chain, but theaters' buildings aren't good for much else (especially the larger ones). And the industry was partly recapitalized by the meme stock phenomenon. Still, the movie theater industry's issues predate the pandemic, and both of those forces are inertia slowing a decline, not actively pushing the industry towards growth.

  1. The obvious comparison here is to Apple; other computer and phone manufacturers have largely learned that the products they sell ought to look nice in addition to working well. And there are a surprising number of Jobs/Ford parallels: both took a product category that had previously been out of reach for individuals and popularized it; both were eventually kicked out of the company they founded but started something new that did even better; both were notoriously brutal perfectionists (Steve Jobs once dropped an iPod prototype into an aquarium, arguing that the air bubbles that came out indicated that it had empty space inside and could be smaller, while Ford tore apart a prototype of an improved Model T with his bare hands ("He ripped the door right off! God! How the man done it, I don't know!")). They even both went through a phase of eating only carrots.

    Building a transformative company requires a bundle of weird, high-variance traits, and sometimes these traits can be expressed in the same way two generations apart. And the other possibility is that genuinely creative people whose creativity puts them in control of a large and complex business will tend to go a bit crazy once they spend their time on management rather than on designing new products.

  2. The source material that talks about this predates the days of ubiquitous buybacks. There are good books from the 80s on the rise of the Japanese auto industry, and buybacks happened at this time, but they were relatively rare and often driven by either a) a specific shareholder that management wanted to get rid of, or b) an unusually savvy management team. So dividends end up being the only real outlet for returns of capital that companies talked about during this period.

  3. A big part of the car industry's evolution over time has been driven by the fact that capital expenditures are visible while working capital needs just feel like the nature of the business. When Ford bought its first conveyor belt, the net cash cost was negative because the company reduced its inventory cost by more than the cost of the conveyor belt itself. There's a good broader lesson there on how important certain kinds of fluff and extras are, and how many can be automated away.