One Billion Buildings
The case for thinking bigger (about why we don't have megacities)
The inspiration for this piece is twofold. First, a Matt Yglesias essay “Why Silicon Valley hasn’t done more for most Americans” that was shared on my Twitter feed (with a great deal of contempt) by Malcolm Harris, author of Palo Alto: A History of California, Capitalism, and the World. “I encourage people to read that post, which is written at a much much lower intellectual level than I could have imagined,” Harris writes. And he’s right!
Second, I just got back from San Francisco and have been thinking a lot about the city since. And so, I’ll try to lay out in this essay, Yglesias’s piece purports to be about why San Francisco isn’t a megacity, but it’s not actually interesting in thinking about the reasons why.
So what is Yglesias’s piece about? He riffs off a Paul Krugman blog comparing American and European living standards—specifically an aside suggesting Europe shouldn’t envy America’s tech sector since it produces oligarchs who undermine politics—and tries to insist our tech sector is both worth envying, but also horrible at distributing its benefits beyond a sliver of workers and elites. The rest of America largely gets ancillary benefits (e.g. Google Search, on-demand convenience). Why? Well, dear reader, it’s because of housing.
To make his case, Yglesias looks to earlier economic booms in American history. Chicago’s steel boom, Detroit’s auto boom, these led to population explosions that pulled waves of migration in for work. Agglomeration around these industries created boomtowns and widely distributed prosperity, he argues, so we should be asking why we don’t see this to the same extent around San Francisco’s tech boom. Yglesias finds it even more puzzling that three of the biggest companies in the world by market share are headquartered in San Francisco, but San Francisco isn’t the biggest metropolitan area in the world (or even the state of California). Instead, we see populations that would’ve made up waves of migration during this boom choose to live elsewhere because San Francisco simply doesn’t build enough housing for them.
If San Francisco had built enough housing, he reasons millions would’ve moved to the Bay Area instead of the Sunbelt and enjoyed higher incomes as tech workers (or as service and construction workers in a boom town) And following this counterfactual’s logic, we also would’ve seen America’s manufacturing capabilities mobilized to build out a Bay Area megacity that would’ve certainly distributed wealth more broadly and deeply across the country. Our housing supply, you see, is the bottleneck—one billion buildings now!
It’s a straightforward enough argument, but is it right? Krugman believes the problem traces back to what the tech sector is, Yglesias believes it boils down to where the tech sector lives. I think both of them miss the mark (Yglesias much more than Krugman).
To his credit, Krugman calls the political influence of our tech oligarch class a “negative externality” but that’s still a little too soft, implying the harms they produce are incidental to an otherwise productive and desirable process (like pollution). The key mechanisms of our tech platforms (zero-sum competition, network effects, a preference for consolidation, etc.) are integral, not incidental, to how the sector generates profits. The fortunes and activities that distort markets and politics are direct and intentional consequences. That the structure of our various tech firms and products tends to stabilize around that of a monopoly or oligopoly isn’t a market failure, but a market feature—part and parcel of the sector’s ambition to serve as infrastructure inseparable from the rhythms of daily life.
Yglesias goes out of his way to treat these outcomes as accidents and policy failures, getting the basic history and periodization (as Harris points out, the Midwest and Bay Area booms are not that similar) wrong—which matters because it has implications for recommendations about what policy surrounding the tech sector should look like!
Inconvenient Facts
The heart of Yglesias’s argument rests on his analogies between the Midwest industrial boom and the Silicon Valley tech boom: they’re both economically and geographically concentrated, so they should produce the same kind of boomtown. Any difference we observe, then, can be explained by policy. But what about the actual economic activity driving each boom?
We can pretty quickly identify a few structural conditions that not only defined the industrial boom, but are unlikely to be replicated in a tech boom.
Industrial production needed masses of workers
One place to start is to ask yourself how profits and prosperity are actually distributed. As anyone who has had a job can tell you, what you get paid is downstream of some compromise made at some point between workers and the people they work for. At its simplest, we might say the shape, form, and substance of that agreement is determined in part by the power of the negotiating parties. Certain workers at certain points of the production process might be in better positions to realize compromises that better suit their interests.
In Forces of Labor, sociologist Beverly J. Silver thinks of this structural power as marketplace bargaining power and workplace bargaining power:
“Marketplace bargaining power can take several forms including (1) the possession of scarce skills that are in demand by employers, (2) low levels of general unemployment, and (3) the ability of workers to pull out of the labor market entirely and survive on nonwage sources of income. Workplace bargaining power, on the other hand, accrues to workers who are enmeshed in tightly integrated production processes, where a localized work stoppage in a key node can cause disruptions on a much wider scale than the stoppage itself. Such bargaining power has been in evidence when entire assembly lines have been shut down by a stoppage in one segment of the line, and when entire corporations relying on the just-in-time delivery of parts have, been brought to a standstill by railway workers’ strikes.”
(Silver’s scheme comes from the late sociologist Erik Olin Wright’s theory of worker power, specifically his 2000 essay “Working-Class Power, Capitalist-Class Interests, and Class Compromise.” which seeks to try to map out the conditions where class compromises between workers and those who own their workplaces (capitalists) are positive, benefitting the material interests of workers and capitalists.)
Industrial workers (like those in the auto industry) had both workplace and marketplace bargaining power. Their labor was essential enough that production could be brought to a screeching halt with strikes, and the laborers were numerous enough that they could form a political bloc with that leverage. This also means we can trace a significant part of the broadly shared prosperity associated with industrial workers of that era (and which Yglesias pines for throughout his piece) comes to us from struggles groups like the UAW—which didn’t emerge because lots of people happened to live near factories, but because the production process concentrated workers who could exercise enough structural leverage of capital to force this or that compromise.
All of this also happens within a legal framework (the Wagner Act, the NLRB) that, for a time, permitted and even encouraged collective bargaining. It’s the result of a political struggle, not an automatic consequence of agglomeration. Still, that brief window is only partially taken advantage of. Taft-Hartley, the anti-communist purges of the Congress of Industrial Organization purges, the Red Scare(s) more generally—these were all victories for those eager to claw back profit shares given up to organized labor by strangling those collectives and the compact they achieved. And the postwar compact between labor and capital, however golden it may have been, was still a racial one—American apartheid was premised on excluding Black people from home ownership and wealth accumulation, as well as creating a dual labor market that confined Black workers to the worst set of jobs and working conditions.
So assuming Yglesias doesn’t want to emulate the exclusionary nature of the Midwest boom, we’re left with the structural elements—the nature of (industrial) production and the proactive redistribution of prosperity via political struggle—that are fundamentally different from those of the Bay Area boom. Does it make much sense to boil this down to where workers choose to live?
Domestically there was a closed circuit
Structural power enjoyed by industrial workers along the production process was also reinforced by the fact that their firms were, in a real way, fixed to one spot. Auto plants were enormous fixed-capital investments beyond the assembly line itself. An automobile required production inputs like steel, rubber, and flat glass before it was finally assembled, with supply chains criss-crossing the country and in turn requiring their own node and network of capital-intensive extraction, processing, and logistical infrastructure.
To take one example relevant to Yglesias’s argument, look to the Ford River Rouge complex in Dearborn, Michigan—first envisioned as a bird sanctuary, then mobilized as an arms manufacturer during World War One, then an automobile factory (and once the world’s largest factory at some 2,000 acres, boasting some 100,000 employees at its peak). The sprawling industrial park was vertically integrated, allowing it to take raw inputs (iron ore and timber from northern Michigan, coal from Pennsylvania, rubber from Latin American plantations), produce its own steel and glass and electricity, and finally assemble an automobile.
There was not much of a credible threat of moving the industrial park or any of its key facilities along the production process—which was partly why firms like Ford reacted so viciously to the threat of collective bargaining (particularly to recognizing the United Auto Workers union). In 1932, thousands of unemployed auto workers marched to present demands to the company and were met with Ford’s private security alongside dozens of police officers who let loose with tear gas, water cannons, and bullets in what would be called the Ford Massacre (hundreds were fired for participation, five people were killed). Five years later, we have the Battle of the Overpass where Ford rolled out private security again to beat up UAW organizers outside of the Rouge River complex (pictures of the beatdown ended up forcing Ford to the negotiating table).
Capital’s immobility, alongside the structural power workers enjoyed over production, was an integral pillar of the capital-labor accord. Yglesias is right that there is a geographic concentration, but it was at various points of the nation itself that meant strikes could shut down production because of a lack of alternative sites when it came to extracting raw materials, crafting components, assembling the product, or eventually distributing the finished goods.
Another piece of here is that the workers are also consumers and domestic markets were designed around this contingency. Henry Ford and his business empire is emblematic here: from our benevolent titan of industry came Fordism, where factories should be organized such that they produce as much as possible for the lowest cost possible so as to compel as much consumption as possible; that higher wages for industrial workers meant higher consumption, higher sales, higher profits, and lower turnover. If owners of capital wanted to preserve their workforce and domestic markets, they had an interest—a reluctant one, constantly contested and extracted over decades of negotiations and strikes and legal battles and political struggle, but real nonetheless—in maintaining wages at a level high enough to help sustain this Fordist wage-consumption loop.
Production yielded wages yielded consumption yielded revenue yielded reinvestment yielded more production, and so on and so on. A relatively closed national economy was one where producers, workers, and consumers were relatively immobile, subject to broadly similar labor laws, spending the same currency, buying the same products from the same firms. And so like a snake eating its own tail, capital—which needs labor to produce and consume—and labor—which needs capital to reinvest—are brought close enough to reach an unsteady peace, so long as they were all stuck in the same room.
The breakdown of those conditions are a key part of why we see such a sharp divergence between wages and productivity that has come to define our society today. Or as the Economic Policy Institute puts it in a primer on the gap between productivity and worker compensation:
Starting in the late 1970s policymakers began dismantling all the policy bulwarks helping to ensure that typical workers’ wages grew with productivity. Excess unemployment was tolerated to keep any chance of inflation in check. Raises in the federal minimum wage became smaller and rarer. Labor law failed to keep pace with growing employer hostility toward unions. Tax rates on top incomes were lowered. And anti-worker deregulatory pushes—from the deregulation of the trucking and airline industries to the retreat of anti-trust policy to the dismantling of financial regulations and more—succeeded again and again.
We can quickly throw in a few other developments worth noting as they’ve made it so the average American and our society at large enjoy lower prices and better goods, but done little to address the deeply unequal distribution of prosperity here. Containerization, which made it cheap enough to ship partially assembled products across oceans; free trade agreements, which opened low-wage labor markets across Latin America and East Asia to American capital; the proliferation of export-processing zones that sparked a global wave of outsourcing (and an erosion of worker bargaining power) by offering low wages, tax exemptions, and little to no labor enforcement in hopes of attracting attracting foreign direct investment.
Where does this leave our compact? The wage-consumption loop frays—I can produce in one country, pay those workers starvation wages, then sell the product to another country with purchasing power (ideally at the highest possible price). A world where capital is mobile but labor now forced to immobility is a world where agreements contingent on the former sharing that condition suddenly dissolve. Where workplace and marketplace bargaining power are insufficient to compel owners of capital to stay in line, where negotiated wages and contracts dissolve—along with the sort of prosperity Yglesias believes is a natural consequence of economic booms.
Results may vary
The core goods of the industrial booms were manufactured then transported—tethered to a place upon which supply chains entangling the globe converge, employing (or contracting [or contracting out the contracting of]) all across the world. What does the web look like for the Bay Area’s products?
Take Apple, which has close to a $3.7 trillion market capitalization. It’s got about 166,000 employees, a sizable chunk of which work in its retail stores. Relatively speaking, its value-generating core is tiny compared to the wealth it produces. In Silver’s parlance, its core technical workers have high marketplace bargaining power, yes, because their skills might be relatively scarce but they have little to no workplace bargaining power. A bitter lesson we’ve been forced to learn over and over again is that no single group of workers can shut down these tech firms (not for lack of trying!), let alone organize them into a collective group large enough to seriously try at doing so. Partly because of the division of labor (amongst employees, but also between employees and contractors). Partly because of the limits of labor law (which has only been whittled down over the years). Partly because of the sheer size of the workforce. But partly because this tech boom has produced a labor aristocracy—well-paid, individuated, and ideologically aligned (by carrot of compensation or stick of constant layoffs)—a move that makes complete sense from the perspective of these firms, who have nothing to lose and everything to gain by eviscerating the “mass worker” as a subject. When it comes to workers who are not in that rarefied elite—the great mass of service workers who clean, cook, and care for the sector’s labor aristocrats—Silver herself notes that the bargaining power of today’s low-wage service workers is actually closer to that of 19th century textile workers than 20th century auto workers. These workers are left wanting for structural power, dependent on associational power and shifting alliances as opposed to any real leverage over any particular point of production. Do they enjoy higher wages because of their proximity to the “ground zero for global innovation,” as Yglesias suggests?
With the abandonment of the Fordist loop, we can trace out a path a firm like Apple sees its value is captured in Cupertino, but its physical products are assembled by Foxconn workers in conditions familiar to workers trapped on the wrong side of the Wagner Act. Cancer clusters, consistent injuries and maimings, persistent exploitation, death from overwork, and so on and so forth. We can dive further into the supply chain to find more deplorable conditions also familiar to workers before the Wagner Act (such as slaves and child workers in Central Africa mining for minerals that go in our advanced electronics). It is true that, as Yglesias writes, tech’s market “scales pretty easily to become truly global” but mainly because it’s free to decouple industry from any sort of obligation to where its headquartered or where its production is placed.
Still, the labor-capital compact’s destruction isn’t just a story centered on individual policy choices or a long chain of deregulations and reregulations in favor of owners of capital. In The Long Twentieth Century, Giovanni Arrighi (Silver’s husband) offers a structural periodization to understand the decline of American hegemony: the postwar order that made the Fordist loop was one phase of American hegemony organized around material expansion—around the production of material goods, infrastructure, and the employment of workers at scale. What kicked off in the 1970s wasn’t simply a policy pivot, but a seismic shift towards financial expansion as the hegemon showed signs of infirmity and as the engine powering its ascension began to sputter. And so, as manufacturing declines and disappears we see capital pour into finance, seeking returns through increasingly abstract instruments and securities and business models over labor-intensive, geographically fixed material production.
We can use this to reframe the tech boom’s relationship to deindustrialization. What spurred the globalization that enabled tech’s scaling via offshoring, the creation of global supply chains, and the prying open of low-wage labor markets? Financialization! Financialization and the creation of a global order privileging American capital’s ability to benefit from production anywhere, anytime, uncoupled from domestic production. This decoupling helped midwife what would become a new institutional life-world, if you will, complete with venture capital funding schemes, equity-based compensation, and corporate governance regimes organized around shareholder value—a life-world where a certain species of corporation might thrive, focused on capturing enormous value while getting away with the minimal number of domestic workers. And so the tech boom doesn’t happen despite deindustrialization, it spins out of it—born within an economic and political order already restructured (and eager to continue to restructure itself) into forms that maximize returns to capital while minimizing any offerings to labor.
Would things have been different if San Francisco built millions of housing units to absorb waves of domestic migration? Probably not. The firms at the center of the Bay Area boom are structurally distinct from the firms at the center of the Midwest’s boom. They’re not anchored to massive fixed-capital investments in industrial parks, they’re not employing hundreds of thousands of domestic workers along a production process vulnerable to stoppages and strikes, they don’t need people who buy the products to be the people who make them, they don’t necessarily have an interest in positive class compromise that props up wages. It also helps that software doesn’t need to be extracted, processed, shipped, or assembled across the country. All of this is a bit deeper than municipal zoning, no? This is about reorganizing American society and its economy around a different logic of capital accumulation.
This isn’t to say that there are zero agglomeration effects at play here, just that they are largely limited to a small, highly compensated group and the financiers backing them. Scholars like AnnaLee Saxenian might point out that Silicon Valley’s geographic concentration has created real clusters and networks, but at the same time this largely was in service of circulating a narrow elite of technical workers between firms, through informal knowledge networks, and through enmeshed venture capital fund portfolios where friends, co-workers, mentees, and ideological comrades enrich one another.
None of this, in of itself, creates the sort of broad multiplier effects for a wider urban population that Yglesias seems to think it does. In fact, if we spend a little time just looking at the Bay Area’s history then that much becomes even more painfully clear.
Geographer Richard Walker is a great resource here (specifically his chapter “Industry Builds Out the City” in the edited volume Manufacturing Suburbs, as well as his essay “The strange case of the Bay Area”) as he’s taken great pains to reconstruct the Bay Area’s industrial geography across the late 19th and early 20th centuries—specifically the period when this region had the sort of manufacturing base Yglesias believes automatically transmits prosperity to those near it. But even under those more favorable conditions, the boomtown/megacity was never in the cards. There was never one industrial core that later saw suburbanization. From the earliest days of San Francisco’s industrialization, manufacturing was dispersed: southern industrial districts went past South of Market by 1890, East Bay surpassed SF’s manufacturing output by 1910 (and employment by 1920).
It’s important to emphasize this isn’t firms relocating to the periphery, which you might tell yourself if you think this was about workers following the money in a boomtown. It was series of sectoral successions, where each new industrial wave erupted in new places with new spatial forms. The industries that made San Francisco dominant (mining equipment, wooden carriages, leather goods, animal processing) were either dead, dying, or undergoing a fundamental transformation by 1900. The ascendant industries (petroleum refining, alloy steel, automobiles, electrical machinery, industrial chemicals) found homes almost entirely outside the city in Oakland, Emeryville, Richmond, and along the Contra Costa shore (they had to find somewhere else to offload the industrial waste and pollution). Industry was not lost to the suburbs, it was made obsolete by new enterprises that sprung up elsewhere. And so the geography of the Bay Area metropolitan area and its productive base found themselves remade and reconstituted with each successive wave of industry.
There were other reasons for moving, of course, beyond finding new places to dump waste. Owners of capital were looking for cheap land, large sites for industrial production, and (most importantly) freedom from San Francisco’s militant laborers. Nearly every major factory in Contra Costa County was, as Walker puts it in his “Industry Builds Out the City” chapter, “dreamed up and financed" by San Francisco capital” as well as “the rail, water, oil, and electricity networks that fed the new industrial district.” Emeryville would be incorporated as a purpose-built industrial enclave run by factory workers for decades, South SF would be politically controlled by its corporate founders, and so industry’s needs were prioritized first and foremost in the design of these vast swaths of the Bay Area metropolis. No surprise, then, that working-class housing followed capitalist investment decisions instead of the other way around! The class geography of the entire region—bourgeois to the north, proletarian to the south; flatland workers, hillside owners in Oakland; workers clustered near factories because of time, income, and transit constraints—was determined by decisions made by owners of capital, not by restrictive zoning codes.
Even in the period of heavy manufacturing for this earlier boom, the Bay Area’s spatial form was overdetermined by capital’s investment logic! Where industrialists chose to build and what political-legal jurisdictions they chose to construct in hopes of protecting those investments, as well as how successive waves of industry sprung up and went on to reshape the region. The megacity never appeared because of some regulatory block, but because it’s not what has ever made sense in the Bay Area if you look at how capital has been accumulated and reproduced over the years there. And if that was the case during the period when the Bay Area had steel mills, shipyards, automobile plants, and other industrial sites that required huddled masses, then there’s no reason to expect a different outcome with a tech boom that requires even less fixed capital, even fewer domestic workers, even greater spatial flexibility, and even greater subservience and sensitivity to the capital’s demands?
This, of course, isn’t even a full accounting of all these details. We can talk about the Bay Area’s massive working class and where they have actually lived during the tech boom, the class structure reproduced by the tech boom, or how a hypothetical megacity wouldn’t substantially enrich these systematically disempowered workers. We could talk about the dreaded environmental reasons why the tech boom may not have produced a megacity (goddamn the NEPA, as some of the Abundance crowd might say)—the sector quite literally poisoned the air, water, and earth (taking extreme care to limit the damage to Black and brown communities, however). We could actually talk about the initial boomtown, tracing how imperial plunder and material extraction (and devastation) along with military spending subsidies all but ensured the class structure of any boom in the region would be one that prioritized capital’s needs (and at even greater cost than normal to labor).
Closing thoughts
So where is the megacity? We have a few rough answers we can offer.
First, the tech economy has no real interest in producing one. The core products have a great deal of physical production offshored and minimal labor requirements relative to wealth generated. When labor-intensive physical production was at home, firms fouled the earth with reckless abandon (visit your local Bay Area Superfund site to get a peek). Benefits are largely limited to a small and narrow technical elite, as well as the financiers behind them, and no amount of housing construction will do anything to change that structural relationship. In the Midwest booms, prosperity was distributed by working in the booming industry itself, not adjacent to it. It’s not clear what, if anything, trickle-down economics—rebranded by consuming in the shadow of Silicon Valley’s mountains of treasure—will do for the masses of workers. What’s a slightly higher nominal wage when the tech boom itself is driving up the cost of living aggressively? In today’s Bay Area, a janitor is precariously housed, systematically disempowered, cut off from any meaningful decisions about their workplace or political order. In tomorrow’s megacity, a janitor will be precariously housed, systematically disempowered, cut off from any meaningful decisions about their workplace or political order.
The second answer is that there’s a superficial understanding of the historical models offered as alternatives to learn from. The Midwest boom shared prosperity through organized labor struggling politically and exercising structural leverage within a national economy that immobilized capital. And even that era, hailed as a golden age by many, was racially exclusionary and viciously contested at every single step. The Bay Area’s own earlier boom was organized around extraction and concentrated wealth spooling out of the Gold Rush through the railroad era right to today’s military-industrial complex (and Silicon Valley’s committed re-engagement with it). Is the dream of a billion skyscrapers on John C Frémont’s golden horns a stillborn one? It’s better to say it’s one that comes with a fever—a delusion entertained by projecting hallucinations onto reality. Not to say we shouldn’t desire a megacity, but that we should think a bit more seriously about why one never existed so we can actually work towards creating one (and figuring out what that entails).
And a third answer: megacities do exist! They were built by developmental states wielding tools like massive public housing provision, state-directed industrial policy, public control of land and credit, and putting the fear of god into a few capitalists here and there so that they ignore the devil on their shoulder (profit-seeking). Can we import policy templates from Singapore or Seoul in hopes of building a megacity? Perhaps. Some of Vivek Chibber’s arguments in essay and book form offer a look at part of the problem here: the conditions which enabled successful industrial transformations (state capacity to discipline capital, bureaucratic autonomy from private interests, etc.) should be understood as specific historical achievements—political settlements that were struggled for, not developments that emerged from letting the private sector do the right thing. We understand this when it comes to developing some of the industrial titans of today (such as Huawei and TSMC) but, for some reason, applying this idea to urban development is treated as suspect at best. Though, of course, the reason is clear, isn’t it? State capacity at the scale necessary to twist capital’s arm is state capacity in position to (and mobilized by an ideological project that) may have funny ideas about property rights, capital mobility, state-run enterprises, competition, the political power of the tech or real estate sector, and so on. At that point, the question shifts from “how do we build more housing?” to “what kind of political power would be necessary to organize the economy’s relationship to land, labor, capital, surplus, prices, and so on?” Some may be uninterested in the latter, some may even view entertaining such questions as squandering opportunities to craft their own version of positive class compromise (to make building more housing in everyone’s interests) because it puts various political actors on high alert—and I’m sure this is true in some instances, but so be it!
I’m skeptical of the idea that booms are always healthy, always distributional in a way that benefits everyone, and easily remedied by policy if for some reason this isn’t the case. I think it’s clear even a cursory glance at the history, economics, politics at play here supports that skepticism.