The City the Machine Made, and the City It Ate
In 1840, Chicago had 4,000 people. In 1890, over a million — the fastest urban growth in history. In 1950, Detroit had 1.85 million. In 2000, under a million. Same city, same factories, same people — but the machine changed. From cotton mechanization pushing Black farmers north, to railroads pulling industry into cities, to computers rewarding coastal hubs and punishing the Rust Belt. Every technological revolution redraws the map — but not evenly, not inevitably, and not without politics.
Agricultural Mechanization and the Great Migration: Not Just the Tractor, the Law
The most studied case: displacement of tenant farmers and sharecroppers in the U.S. cotton South, 1930–1970, triggering the “Great Migration” of African Americans (and whites) to northern cities. Fligstein’s regression analysis of 741 southern counties across 1930–1940 showed the primary cause wasn’t tractor arrival but the prior reorganization of southern agriculture through New Deal subsidies. The Agricultural Adjustment Act (1933) paid large owners to reduce cotton acreage, enabling them to buy tractors and systematically displace tenants. Subsidy payments had statistically significant effects on net out-migration even controlling for direct mechanization measures — “individuals realized as early as the 1930s that cotton agriculture would change its social organization and therefore left as they understood the implications.” For Blacks, disproportionately concentrated in the tenant class (four of five vs. one of two whites), effects were especially severe: Black tenancy fell 33% per county on average during the 1930s, white tenancy only 12%.
The mechanical cotton picker’s role was belated and spatial. Heinicke and Grove show the West (California, Arizona) adopted first (10% threshold by 1951), then Delta states (1953), then Southeast (1960). This sequence was driven by cotton yields, terrain, climate — not “southern social institutions” per se. High Western yields made mechanization profitable earlier; the Southeast didn’t reach profitability until the late 1950s. The 10–90 diffusion lag was actually fastest in the Southeast (9 years) vs. 11 in the West and 14 in the Delta — once conditions were right, adoption was rapid everywhere. The study estimates ~90% of the California-Louisiana mechanization gap in 1951 was attributable to machine-cost differences from environmental/yield factors. By 1970, the South’s rural labor force had largely departed. The concentrated displacement burst occurred mainly post-1959, accentuating 1960s urban tensions as displaced poor Black southerners joined the rural-to-urban stream. Net Black out-migration from the South totaled 1,243,000 in the 1940s alone, continuing through the 1960s; by 1980 over 4 million southern-born Blacks lived outside the region. Tolnay documents how this influx doubled the northern Black workforce and transformed northern cities like Chicago, Detroit, New York, Philadelphia.
Distinguishing technological from other causes: Tolnay emphasizes the Great Migration combined “push” factors (agricultural reorganization, mechanization, racial violence, Jim Crow) and “pull” factors (WWI industrial labor demand, restrictive immigration policies opening jobs to Black workers, kin-chain migration). Crucially, Alexander’s analysis of 1930s Pittsburgh marriage records shows that by later phases, most Black migrants to northern cities came from southern *cities and towns*, not directly from rural areas — over two-thirds of southern-born Blacks in his sample were census-urban born, versus under one-third of the overall southern Black population. This “urban differential” suggests agricultural mechanization first pushed people into southern cities, where they gained wage-labor experience, before moving to the industrial North.
Tobacco mechanization: a contrasting case. Perkinson and Hoover’s study of tobacco harvest mechanization in North Carolina provides a useful counterpoint. Their projections found even under substantial mechanization (47% reduction in hired harvest labor), potential out-migration was limited to ~2% of exposed households per year — far less than the cotton South. Tobacco workers had multiple income sources; mechanization losses represented only 4–7% of household income. Conclusion: “concern over massive migration resulting from mechanizing the tobacco harvest… does not seem to be substantiated.” This demonstrates that migration effects of agricultural mechanization depend heavily on pre-existing economic structure, local alternative employment, and household reliance on the mechanized crop.
Industrialization and Urban Growth in the Nineteenth Century: Rural Beginnings, Rail-Driven Shift
The classic narrative that industrialization “pulled” rural workers to cities isn’t wrong, but evidence reveals a complex, phased process. Early American industrialization was predominantly rural. Atack, Margo, and Rhode document that in 1850, only about a quarter of manufacturing establishments were in urban places, even though 15.4% of the population was urban. Early New England textile mills were deliberately sited in the countryside, drawing on local waterpower and young women’s labor from nearby farms. Kim explains this rural pattern persisted because early factories had limited division of labor, a relatively homogeneous workforce (locally recruited women and children), and low labor-matching costs making urban concentration unnecessary.
The shift to urban manufacturing occurred 1850–1880, driven by the “Transportation Revolution” — especially railroads. Difference-in-difference estimates show gaining rail access was positively and significantly associated with increases in steam-powered factories in urban areas. Railroads dramatically reduced transport costs: per-ton-mile shipping rates fell 80% on railroads during the century, while wagon haulage declined only from 70¢ to 15¢ per ton-mile. This expanded market access, making city factories profitable via agglomeration economies.
Kim’s instrumental-variables analysis of firm-level manufacturing data 1850–1880 finds urban wage premiums were substantial and rising — IV estimates show urban workers earning 60%, 25%, 129%, and 118% more than rural workers in 1850, 1860, 1870, and 1880 respectively. Urban firms also had 32–180% higher total factor productivity. The urban wage premium applied across skill levels: both skilled male workers and unskilled female/common laborers earned higher wages in cities, and urbanization actually *reduced* the gender wage gap in 1850 and 1860.
The scale of urban transformation was dramatic. Chicago grew from 4,000 in 1840 to over 1 million by 1890 — one of history’s fastest-growing cities. By 1900, fully a quarter of the U.S. population resided in the nation’s 100 largest cities (pop. 38,000+). Industrialization fundamentally changed city composition: early 19th-century cities were commercial centers dominated by merchants and artisans; by century’s end they were industrial powerhouses with a large working class, increasingly composed of European immigrants. By 1880, 73% of manufacturing employment was urban, vs. 41% in 1850.
Distinguishing technological effects: The waterpower-to-steam shift was long thought the key driver of manufacturing urbanization. But Kim’s analysis suggests steam contributed at most 8–10% of the urbanization increase 1850–1880. More important was the shift from artisanal to factory organization, relying on division of labor and labor-market matching economies. As Kim notes, “the most consistent explanation for why industrialization arose in rural areas and then shifted to urban areas is likely to be based on division of labor and labor-matching costs.” Railroad expansion and immigrant influx (disproportionately settling in cities) further reinforced urban industrial concentration.
Deindustrialization, Suburbanization, and Urban Shrinkage in the Twentieth Century
If industrialization drove urban growth, deindustrialization drove urban decline. Rieniets provides a comprehensive global overview, arguing “shrinking cities have been as much a product of the industrial age as growing cities.” Between 1950 and 2000, over 350 large cities worldwide experienced significant population decline. In the U.S., eighteen of the twenty-five largest cities reported inner-city population loss by 1970; nine continued shrinking through 2000. Losses were staggering: St. Louis -49%, Youngstown -51%, Pittsburgh -50%, Buffalo -49%, Detroit -49%. Driven by deindustrialization (manufacturing automation and later offshoring eliminating jobs) and suburbanization (highways, FHA mortgages, rising car ownership enabling the middle class — disproportionately white — to leave central cities).
Distinguishing deindustrialization from suburbanization: Rieniets distinguishes two processes: suburbanization (cities lose population but metros keep growing) and deindustrialization (entire metro regions decline). Detroit: city shrank 1.85M (1950) to <1M (2000), but surrounding suburbs (Wayne, Oakland, Macomb) nearly tripled 1.2M to 3M. Suburbanization enabled by technological change (automobile, highway construction, industrialized homebuilding) and government policy (mortgage subsidies, highway funding). In contrast, industrial cities in northern Britain (Glasgow, Liverpool, Manchester), Germany's Ruhr, and the U.S. Rust Belt experienced both inner-city loss from suburbanization *and* overall metro decline from deindustrialization. Manchester, the "first industrial city," lost population from 1931 (peak 766,000) until the 1990s, when it revived via a service-sector shift.
Pre-1930 urban decline: a comparative baseline. Beauregard’s analysis of aberrant cities losing population 1820–1930 (when urban loss was extremely rare) provides historical perspective on decline causes independent of post-industrial dynamics. Only 15 large U.S. cities lost population in any decade before 1930; only Charleston did so twice. Causes were overwhelmingly specific, often catastrophic: fires, epidemics (yellow fever, cholera), trade route shifts (steamboats bypassing fall-line ports), hinterland resource exhaustion, or failure to transition from commerce to manufacturing. Nantucket collapsed when whaling declined (due to petroleum refining) and an 1846 fire destroyed its business district. These cases highlight technology could produce urban decline even in the 19th century, but decline was typically episodic and localized, not structural and widespread.
Computerization and the New Geography of Urban Success and Failure
The 1980s Computer Revolution created a new spatial pattern of urban divergence still reshaping American cities.
Skill-biased technological change and urban fortunes. Berger and Frey show the Computer Revolution marked a sharp break in new jobs’ skill content. Using data on new occupational titles appearing in U.S. census classifications 1970–2000, they document that pre-1980, new jobs appeared mainly in routine-skill occupations. Post-1980, new jobs shifted dramatically toward abstract, analytical, interactive skills — jobs complementary to computing. This shift in the *type* of new jobs fundamentally altered the geography of opportunity.
Cities with abundant “abstract skills” before the Computer Revolution created substantially more new jobs after it. Berger and Frey find virtually no relationship between a city’s abstract-skill endowment and new job creation in the 1970s, but from 1980 onward the relationship became strikingly strong and positive. A one-standard-deviation increase in computer adoption 1980–2000 was associated with a 0.3–0.6 SD increase in new job creation across U.S. commuting zones. Cities historically specializing in routine work — Buffalo, Cleveland, Detroit — fell behind as computers substituted for routine tasks, while skill-abundant cities like San Francisco, Boston, and New York pulled ahead.
The robotics phase: regional specialization and persistence. Leigh and Kraft’s “robotics census” reveals a more nuanced automation geography. While commentary treats automation as universal, their mapping shows robotics activity is spatially concentrated and regionally differentiated. They identify two robotic region types: “supplier-dense regions” on coasts (Boston, San Jose/Silicon Valley, New York) specializing in analytical knowledge — R&D and production of new robotics tech — and “integrator-dense regions” in the traditional manufacturing belt (Detroit, Cleveland, Chicago, Cincinnati) specializing in synthetic knowledge — practical integration of robots into production. Strikingly, some regions most affected by deindustrialization now host strong robotics communities. Detroit and Chicago rank first and second in robotics establishments. This suggests automation doesn’t simply destroy manufacturing employment everywhere; it can create “path renewal” opportunities in declining regions, as Pittsburgh’s transition from steel production to steel technology exemplifies. But Leigh and Kraft caution: jobs created are often “job-preserving” or “job-changing” not “job-creating” at scale, and robotics represents only ~0.05% of total U.S. non-farm employment.
Population composition effects. Computerization reshaped city population composition by intensifying skill sorting. High-abstract-skill cities attract highly educated workers, further increasing attractiveness through dynamic feedback: skilled workers attract firms, creating more skill-intensive jobs, attracting more skilled workers. Meanwhile, routine/manual-specialized cities saw employment shares decline and struggled to retain/attract college-educated workers. This “Great Divergence” among U.S. cities — growing gaps in incomes, skills, opportunities between thriving and struggling metros — has its roots, Berger and Frey argue, in differential ability to adapt to computer technology post-1980.
Synthesis: Distinguishing Technological from Other Causal Forces
Across all three technological revolutions, historical evidence supports several general conclusions about distinguishing technological effects from other causes:
1. Technology operates through institutions. Agricultural mechanization didn’t cause the Great Migration directly; it was mediated by large farm owners’ political power securing subsidies enabling tenant displacement. The same technology (mechanical cotton picker) diffused at different rates across regions depending on environmental conditions and existing investment patterns, not purely on technological availability.
2. Complementary technologies matter. The railroad was arguably more important than the steam engine in driving industrial urbanization. Similarly, the computer revolution’s urban effects depended on simultaneous development of complementary skills, organizational changes, and new job definitions.
3. Pre-existing conditions shape outcomes. Cities with abundant abstract skills before the Computer Revolution benefited disproportionately; those with routine-skill specializations suffered. This path dependence means identical technologies can produce divergent urban outcomes depending on local skill endowments, industrial legacies, and institutional frameworks.
4. Migration is pushed as much as pulled. The classic economic model emphasizing wage differentials as migration’s driver is incomplete. Fligstein shows conditions at the *point of origin* — southern agriculture’s social reorganization — were more decisive than northern pull factors in explaining the Great Migration’s timing and scale. Tolnay similarly emphasizes “to understand large-scale migration, it may be most useful to examine transformations in basic social and economic relations at the point of origin.”
5. Race, class, and gender mediate technological impacts. Agricultural mechanization fell hardest on Black tenant farmers. Industrial urbanization initially relied on young rural women’s labor, then shifted to immigrant labor. Computerization disproportionately benefited abstract-skill workers, more likely college-educated and (in U.S. context) concentrated in coastal metros. In 1930s Pittsburgh, Black women migrating from southern cities benefited from prior urban experience securing non-domestic jobs, but Black men did not — suggesting racial discrimination in labor markets can block skill advantages step-migration would otherwise confer.
The historical record demonstrates that agricultural mechanization, industrialization, and computerization each reshaped migration and urban geography, but always through the mediating lens of existing social structures, political decisions, and complementary institutional and technological conditions.
✦ ArchUp Editorial Insight
The reviewed research reveals that each technological revolution — agricultural mechanization, industrialization, computerization — didn’t “cause” urban patterns deterministically. The city isn’t a passive vessel into which technology pours. The city is the reactor where technology interacts with power, race, class, geography, and institutions to produce outcomes technology alone doesn’t predict. The Great Migration wasn’t “cotton mechanization = migration.” It was “New Deal subsidies + planter power + Black tenant class + Jim Crow + northern wartime demand + kin chains = migration.” Remove any element, the outcome changes. The same mechanical picker in California produced a wholly different result than in Louisiana — not because the machine differed, but because yields, terrain, and climate differed.
The counterintuitive finding: the “black box” isn’t the technology — it’s the institution. Berger and Frey show the Computer Revolution didn’t reward “skill” abstractly — it rewarded cities that *already* had abstract-skill concentrations before 1980. Cities that built universities, attracted researchers, fostered innovation culture in the 1960s-70s reaped the 1980s harvest automatically. Cities that bet on routine manufacturing (Detroit, Cleveland) found themselves on the wrong side of technological history not because manufacturing is “bad,” but because their specific skills became algorithmically substitutable. Pre-existing path decides technological fate.
Systemic implications for urban planning are profound. 1950s planners who built highways through Black neighborhoods thinking they were “modernizing” the city applied a technology (car/highway) through an institution (structural racism) that produced sustained urban decline for decades. Today’s planners betting on “smart cities,” “data,” “AI” as urban solutions repeat the same error unless they ask: who owns the data? Who writes the algorithm? Who benefits from efficiency? The smart city that surveils poor people’s movements to direct policing isn’t “smart” — it’s an old technology (surveillance) in new packaging. The circular city that recycles 90% of waste but whose recycling workers are undocumented migrants without rights isn’t “circular” — it’s extraction in green clothing.
The open question: can urban planning anticipate technology rather than react? Manchester’s experience — waiting until the 1990s to pivot from steel to services — shows reactive cities fall behind. Cities that anticipate (build infrastructure for future skills, design adaptable neighborhoods, distribute technological gains before they concentrate) survive. But anticipation requires political courage to distribute costs today for tomorrow’s benefits — the rarest resource in planning. Until we have that courage, every technological revolution will redraw the map the same way: exacerbate inequality, reward those who already have, punish those already marginalized, and call the outcome “historical inevitability.”
References:
1. Fligstein, Neil. “The Transformation of Southern Agriculture and the Migration of Blacks and Whites, 1930–1940.” International Migration Review, 1983.
2. Heinicke, Carl, Grove, William. “‘Machinery Has Completely Taken Over’: The Diffusion of the Mechanical Cotton Picker, 1949–1964.” The Journal of Interdisciplinary History, 2008.
3. Tolnay, Stewart. “The African American ‘Great Migration’ and Beyond.” Annual Review of Sociology, 2003.
4. Alexander, J. T. “The Great Migration in Comparative Perspective.” Social Science History, 1998.
5. Perkinson, L. B., Hoover, D. M. “Tobacco Mechanization and Potential Out-Migration.” Journal of Agricultural and Applied Economics, 1977.
6. Atack, Jeremy, Margo, Robert A., Rhode, Paul W. “Industrialization and urbanization in nineteenth century America.” Regional Science and Urban Economics, 2022.
7. Kim, Sun. “Division of labor and the rise of cities: evidence from US industrialization, 1850–1880.” Journal of Economic Geography, 2006.
8. Rieniets, Thomas. “Shrinking Cities: Causes and Effects of Urban Population Losses in the Twentieth Century.” Nature and Culture, 2009.
9. Beauregard, Robert A. “Aberrant Cities: Urban Population Loss in the United States, 1820-1930.” Urban Geography, 2003.
10. Berger, Thorsten, Frey, Carl B. “Did the Computer Revolution shift the fortunes of U.S. cities? Technology shocks and the geography of new jobs.” Regional Science and Urban Economics, 2016.
11. Leigh, Nancy G., Kraft, Brian R. “Emerging robotic regions in the United States: insights for regional economic evolution.” Regional Studies, 2017.
