First, a meta question I always think about when considering the topic of automation and abundance. Why are we so afraid artificial intelligence will automate tasks that many people just fundamentally hate? You can maybe be afraid AI kills us. But I don't think that's the actual reason everyone is anxious, except my wife who has a very real fear of humanoid robots. So maybe that's a subset of people. But many of these tasks are so boring and antithetical to the human spirit that Hollywood made three critically acclaimed movies just in the year 1999 around this exact narrative motif.

_{Fight Club (1999)}

_{The Matrix (1999)}

_{Office Space (1999)}

_{Three depictions of the late-1990s office worker, as imagined by Hollywood: hollow in Fight Club, trapped in The Matrix, and taking a bat to the machinery in Office Space. Long before AI, popular culture was already asking why we spent our lives this way.}

Ezra Klein, co-author with Derek Thompson of Abundance, offers a reasonable answer in a recent conversation about AI and the public good: the fear is fundamentally about distribution. People can imagine AI creating enormous abundance. They simply do not trust that any of it will go to them.

Abundance is the result of an economic loop

A positive productivity shock is not a pile of money waiting to be divided. It creates a saving. It does not determine where that saving goes.

AI has already shown that it can create a real surplus. In a study of 5,172 customer-support agents, a generative-AI assistant increased issues resolved per hour by 15 percent on average. The gains were largest among less-experienced workers. It was one company and one measurable workflow, but the premise is no longer hypothetical: some work can become more productive. Brynjolfsson, Li, and Raymond (2025)

The company can keep the saving as a larger margin. It can lower prices, raise pay, serve more customers, shorten hours, or employ fewer workers. Every possibility begins with the same technical achievement.

A productivity gain becomes abundance only when it successfully moves through a loop. Lower costs become lower prices. Lower prices create new customers. Larger markets justify more investment, greater capacity, specialized labor, and further invention. Workers eventually earn more in an economy where their pay also buys more. That purchasing power creates demand and the loop repeats.

A productivity dividend is the first step. Abundance comes from repeating the loop.

None of this is automatic. Competition affects whether savings reach consumers. Finance affects whether opportunity becomes investment. Workers need routes into the new work. Productive activity also happens somewhere, and people must be able to reach it. That makes housing one of the links in the loop.

Housing policy affects who benefits from a productivity shock

The distributional debate usually stops with labor and capital. Will the saving become profit or pay? That accounting is incomplete because workers spend their pay, and they often have to buy access to the place where their labor is most valuable.

Suppose AI makes the Bay Area much more productive. A renter already there may receive a raise and then pay more in rent. An incumbent homeowner may receive the same labor-market gain while the value of his house rises too. A teacher, nurse, technician, or young software engineer living elsewhere may never move because the wage premium does not cover the cost of entry.

Richard Hornbeck and Enrico Moretti show what a local productivity boom looks like after housing adjusts. They study manufacturing-productivity shocks across 193 American metropolitan areas and trace the local effects for three decades. More productive cities paid more. They also became more expensive. Hornbeck and Moretti (2024)

_{Figure 1. Long-run point estimates for the 1980-2000 local response to a 1 percent increase in metropolitan manufacturing revenue productivity during 1980-1990. Annual earnings rose 1.45 percent, rent 1.47 percent, and home values 2.46 percent. Estimated purchasing power rose 0.62 percent for renters and 1.11 or 1.60 percent for incumbent homeowners under two alternative treatments of home equity. Housing and other local prices offset about 57 percent of renters' nominal earnings gain. All displayed coefficients are statistically significant at the 5 percent level; robust standard errors are omitted for legibility. Historical manufacturing evidence, not an AI forecast. Applied Punk visualization of Hornbeck and Moretti (2024), Table 2, column 2.}

In the long-run estimates, a 1 percent increase in local manufacturing revenue productivity was associated with annual earnings that were 1.45 percent higher. Renters still gained. After accounting for housing and other local prices, their purchasing power rose an estimated 0.62 percent. Using the authors’ point estimates, higher local prices offset about 57 percent of the nominal earnings gain.

The boom drew in people too. Local employment rose an estimated 4.16 percent. That result does not isolate housing’s role, but it shows that a productivity shock changes the number of participants as well as the value of what existing residents own.

Incumbent homeowners gained more because they received the labor-market gain while owning an asset that became more valuable. The two homeowner estimates are not uncertainty bounds. They reflect different assumptions about whether housing wealth can be consumed through borrowing, moving, or heirs living elsewhere.

Then there is the person who never appears in the figure. A worker outside the city receives no local wage gain and owns no appreciating local asset. If housing is scarce enough, the move may no longer pay.

This is where housing changes the loop itself. Restrictions do not merely transfer part of an existing gain from renters to owners. They can reduce the number of people who join the productive cluster, the number of matches between workers and firms, and the size of the market supporting the next round of investment.

A landlord does not have to build a model, train a worker, or understand the technology. When valuable activity concentrates where housing cannot expand, ownership of the scarce location is enough to capture part of the gain.

The industrial revolution and the origins of abundance

The Industrial Revolution automated tasks previously performed by skilled artisans and placed the machines under the control of capital owners. Capitalists wanted profit. Landowners wanted rent. Workers had every reason to fear that the machines would make someone else rich.

Why did the result eventually become mass abundance?

Not because the owners became less greedy. Productive capacity kept expanding. Cotton gives us the cleanest numbers. From 1784 to 1820, the estimated cost of processing a pound of medium British cotton yarn fell 88 percent, while cotton throughput rose from 11.3 million to 120 million pounds. Leunig and Voth (2011); Harley (1998)

_{Figure 2. Estimated processing cost for medium cotton yarn and raw cotton processed in Britain. Processing cost fell from 107.2 to 12.6 pence per pound in 1784 prices, while throughput rose from 11.3 million to 120 million pounds. Quantities before 1811 do not adjust for inventory changes, and total processing includes cotton later exported directly or after further processing. Applied Punk visualization of Leunig and Voth (2011), appendix table “Prices and consumption of cars and cotton.”}

The machine lowered costs. Profit attracted capital and new producers. Capacity expanded. Competition pushed some of the saving into lower prices. Lower prices helped unlock mass demand. Larger markets justified more investment, specialization, and invention, which lowered costs again. Evidence from early Lancashire firms suggests that extraordinary profits did not remain with the first producers forever. Entry and expansion passed part of the gain to buyers. Harley (2012)

That loop did not make the transition clean or fair. British cotton was embedded in global trade, slavery, and empire. Particular crafts were destroyed, and productivity initially outran the purchasing power of many workers. Allen (2009); Crafts (2022); Olmstead and Rhode (2018)

What was new was compounding. Manufactured goods and machines could be reproduced. When demand rose, firms could build more capacity, hire and train more people, and make more goods. Private ambition produced a broader dividend because output could expand and competition made it difficult to preserve every saving as a permanent rent.

A productive address is different. Homes can be reproduced. A particular location cannot, and new homes can appear there only if physical capacity and public rules allow it.

AI productivity is geographically concentrated

AI looks placeless because its output is digital. Much of its value may travel nationally through cheaper software and services. The firms, workers, investors, and knowledge networks producing the frontier are still located somewhere.

Brookings estimates that the Bay Area accounted for 13 percent of U.S. AI-related job postings in its 2025 analysis. San Francisco and San Jose were the only two of 195 metropolitan areas classified as AI “Superstars” across talent, innovation, and adoption. Muro and Methkupally (2025)

Concentration is part of how productive clusters work. Workers find specialized firms. Firms find specialized workers. Ideas, capital, and new companies circulate among them. But those benefits depend on people being able to enter the cluster.

Remote work can weaken the link between a job and a particular office. The pandemic showed that technology can redirect housing demand across cities. It rearranged geography. It did not abolish it. Mondragon and Wieland (2025)

This is not a prediction that Bay Area landlords will capture the national AI dividend. Digital products can become cheaper for consumers everywhere. The claim is narrower. To the extent that frontier firms and complementary workers remain concentrated, housing restrictions raise the price of joining that economy.

That affects more than distribution after the boom. A smaller accessible labor market means fewer potential matches between workers and firms, fewer people sharing knowledge, and less capacity for the cluster to support the next round of companies and investment.

Housing restrictions limit abundance

Housing demand can resolve itself in two ways. A place can add homes, or households can bid more aggressively for the homes that already exist. The demand shock can be identical. The supply response determines how much appears as quantity and how much appears as price.

Raven Saks estimates exactly this adjustment. She compares the response to the same 1 percent labor-demand shock in metropolitan areas at the 25th and 75th percentiles of a housing-supply regulation index. The less regulated place is roughly Denver. The more regulated place is roughly Newark. Saks (2008)

_{Figure 3. The same 1 percent labor-demand shock produces a smaller employment response and a larger short-run house-price response at the 75th percentile of housing-supply regulation than at the 25th percentile. The long-run employment estimates have overlapping bootstrap intervals, so this is evidence about the direction and mechanism of adjustment, not a precise estimate of aggregate growth costs or an AI forecast. Applied Punk redraw from vector coordinates in Saks (2008), Figure 5, Federal Reserve 2005 working-paper version.}

The figure is the missing link between housing and the labor market. When construction responds less, a boom draws in fewer workers and creates more pressure on the price of existing homes. The long-run employment paths are not statistically separable with precision, but the pattern is the one the mechanism predicts.

Nathaniel Baum-Snow and Lu Han estimate average neighborhood supply elasticities of 0.29 for housing units and 0.51 for floor space. More importantly, variation within metropolitan areas exceeds variation between them. The ability to add housing is not one citywide fact. It changes from neighborhood to neighborhood with developable land, existing density, construction costs, delays, and regulation. About half of the estimated unit response comes through new construction, with changes to existing structures doing the rest. Baum-Snow and Han (2024)

AI is new. This mechanism is not. Peter Ganong and Daniel Shoag found that rising housing prices and tighter land-use regulation redirected lower-skill migration away from high-income places and weakened regional income convergence. Housing affects who can follow opportunity. Ganong and Shoag (2017)

Housing restrictions are not the only constraint, zoning is not the only determinant of supply, and every restriction is not pointless. New residents require transportation, utilities, schools, and public space. The relevant question is whether institutions respond to growth by adding capacity or by rationing access through higher prices.

If housing expands, the first productivity gain can bring in more workers, support more firms, deepen the market, and help generate another gain. If housing remains fixed, more of the gain becomes rent and land value. Fewer people enter. The cluster is smaller than it could have been. The next turn of the loop begins with less. We do not yet know the size of that effect for AI. But we know which way the constraint works.

The Industrial Revolution’s miracle was not that machines made their owners rich. That part was easy. The miracle was that productive capacity kept expanding. Competition lowered prices, cheap goods created mass demand, investment expanded capacity, and workers eventually earned more in an economy where their wages bought more.

AI can begin the same process. It can make intelligence cheaper and raise the value of the people and places that use it well. But the first productivity gain is not the final outcome.

AI will create a real dividend. Whether it becomes broad abundance depends in part on whether productive places admit more participants or auction access among those already inside.

Sources and data

• Robert C. Allen, “Engels’ Pause: Technical Change, Capital Accumulation, and Inequality in the British Industrial Revolution” (Explorations in Economic History, 2009), source.

• Nicholas Crafts, “Slow Real Wage Growth during the Industrial Revolution: Productivity Paradox or Pro-Rich Growth?” (Oxford Economic Papers, 2022), source.

• C. Knick Harley, “Cotton Textile Prices and the Industrial Revolution” (Economic History Review, 1998), source.

• C. Knick Harley, “Was Technological Change in the Early Industrial Revolution Schumpeterian?” (Explorations in Economic History, 2012), source.

• Tim Leunig and Hans-Joachim Voth, “Spinning Welfare: The Gains from Process Innovation in Cotton and Car Production” (CEP Discussion Paper 1050, 2011), source and Figure 2 data.

• Alan L. Olmstead and Paul W. Rhode, “Cotton, Slavery, and the New History of Capitalism” (Explorations in Economic History, 2018), source.

• Ezra Klein and Derek Thompson, Abundance (Avid Reader Press/Simon & Schuster, 2025), source.

• Ezra Klein, “AI and the Public Good with Ezra Klein” (May 20, 2026), audio.

• Erik Brynjolfsson, Danielle Li, and Lindsey Raymond, “Generative AI at Work” (Quarterly Journal of Economics, 2025), source.

• John Mondragon and Johannes Wieland, “Housing Demand and Remote Work” (Federal Reserve Bank of San Francisco Working Paper 2022-11, revised 2025), source.

• Richard Hornbeck and Enrico Moretti, “Estimating Who Benefits from Productivity Growth” (Review of Economics and Statistics, 2024), source.

• Raven E. Saks, “Job Creation and Housing Construction: Constraints on Metropolitan Area Employment Growth” (Journal of Urban Economics, 2008), source.

• Peter Ganong and Daniel W. Shoag, “Why Has Regional Income Convergence in the U.S. Declined?” (Journal of Urban Economics, 2017), source.

• Mark Muro and Shriya Methkupally, “Mapping the AI Economy” (Brookings Metro, 2025), source.

• Nathaniel Baum-Snow and Lu Han, “The Microgeography of Housing Supply” (Journal of Political Economy, 2024), source.

Will artificial intelligence replace you? I don’t know. But a job is not a single thing that vanishes the moment a machine learns one part of it; it is a bundle of tasks, and those tasks do not contribute equally to your wage, autonomy, or bargaining power. So “Will AI take my job?” is the wrong first question: the right one is which tasks it takes, what happens to the worker who performed them, and whether the resulting power shifts toward labor or capital.

That question has been hiding inside automation for two centuries. The Industrial Revolution transferred tasks performed by human bodies to machines. The digital revolution transferred routine information work to computers. Artificial intelligence now reaches into writing, diagnosis, analysis, and judgment. The object of automation is changing. The labor-market mechanism is not. A machine performs a task, a worker’s time is released, the job is rebuilt, and the value of the remaining work changes.

A Tale of Two Jobs

Handloom Weavers in Early Nineteenth-Century Britain

Britain’s handloom weavers show what happens when automation reaches the task around which an occupation’s economic value is organized. Weaving was not one minor duty inside a larger job. It was the productive task supporting the worker’s skill, wage, and bargaining power. The power loom did not need to reproduce everything a weaver knew. It only needed to produce cloth more cheaply and at greater scale.

Employment initially rose. The estimated number of handloom weavers increased from roughly 184,000 in 1806 to 240,000 by 1820 and remained near that level through 1831. Wages moved in the opposite direction. Estimated weekly earnings fell from 240 old pence in 1806 to 99 in 1820 and 72 in 1831. Employment eventually followed, falling to roughly 3,000 workers by 1862. (Wood 1910, Table 41, pp. 127–128)

The sequence matters. The occupation remained large after its economic foundation had begun to erode. Wages registered the change before employment did. Watching the number of jobs alone would have missed the first half of the story.

Figure 1: Estimated weekly earnings and employment among British cotton handloom weavers, 1806–1862. Lines connect annual historical estimates; earnings are nominal pre-decimal pence per week and employment is measured in thousands of workers. Source: Wood (1910), Table 41, pp. 127–128.

Radiologists Two Centuries Later

Radiology appeared to offer artificial intelligence a clean target. In 2016, Geoffrey Hinton predicted that deep learning would soon outperform radiologists. The logic seemed direct: radiology had converted diagnosis into digital images just as machine learning was becoming extraordinarily good at recognizing patterns in images. (Handel 2022)

But image recognition is not the occupation. Radiologists compare scans with patient histories, determine which anomaly matters, communicate uncertainty, guide procedures, manage risk, and assume responsibility for the report. Even reading images contains different tasks. Clearing a routine study, locating a suspicious region, resolving an ambiguous finding, and deciding what happens next are not the same work.

A Swedish mammography trial made that decomposition visible. Investigators randomly assigned 105,934 women to standard screening or an AI-supported workflow. The AI assigned risk scores, helped determine whether an examination received one or two human readings, and marked suspicious findings. The AI-supported group required 61,248 human readings, compared with 109,692 under standard double reading. That was a 44.2 percent reduction in screen-reading workload. Cancer detection rose from 5.0 to 6.4 cases per thousand women screened, with no statistically significant increase in the false-positive rate. This was one workflow, not a forecast for every radiology department. (Hernström et al. 2025)

Figure 2: Human screen-reading workload and cancers detected in the randomized MASAI mammography trial. The trial measures one screening workflow, not the radiology profession. Source: Hernström et al. (2025).

Nor did the measured Medicare-enrolled radiologist population collapse as medical-image AI improved. U.S. Medicare-enrolled radiologists rose from 30,723 in 2014 to 36,024 in 2023. Through 2019, our CPI-adjusted calculation from Malhotra et al.’s plotted median diagnostic-radiology compensation series shows real pay rose about 2 percent outside academia and 5 percent in academic practice. From 2019 to 2023, it fell about 10 percent and 2 percent, respectively. That break overlaps COVID-19, which disrupted imaging volumes and compensation, and the inflation surge that followed. The sources do not estimate the effect of AI. They show something narrower: measured headcount and pre-pandemic real pay did not experience a handloom-style collapse. (Christensen et al. 2024; Malhotra et al. 2025; U.S. Bureau of Labor Statistics 2024)

Figure 3: Medicare-enrolled radiologist headcount rose while CPI-adjusted median diagnostic-radiology compensation in both settings remained stable or increased through 2019. Panel B is an approximate calculation from Malhotra et al.’s plotted nominal medians, deflated by annual-average CPI-U and indexed to 2014 = 100. The post-2019 decline overlaps COVID-19 and higher inflation and is difficult to disentangle the causal effect of the decline related to AI. Sources: Christensen et al. (2024); Malhotra et al. (2025); U.S. Bureau of Labor Statistics (2024).

The trial tells us what the technology made possible. It cannot tell us what happened to the labor it saved. The freed capacity could screen more patients, reduce a backlog, create more time for difficult cases, raise the number of readings expected from each physician, or reduce the number of labor hours purchased. Better clinical output does not determine who captures the economic gain.

The weaver and the radiologist therefore give us two outcomes, but not two simple morals. The power loom attacked the task around which the weaver’s value had been built. AI entered one part of a larger professional bundle and released capacity that could be redirected toward the work that remained. Both technologies automated tasks. Only one immediately made the worker easier to replace.

The Task-Based Framework: Displacement and Reinstatement

The handloom weaver and the radiologist point to the same mechanism. Production is divided into tasks, and those tasks can be assigned to workers or machines. Technology changes that assignment before it changes the occupation recorded in employment statistics. Acemoglu and Restrepo call this the task-based framework. Automation occurs when capital takes over a task previously performed by labor. The first economic change is therefore not the disappearance of a job. It is a change in the task content of production. (Acemoglu and Restrepo 2019)

A job title hides that structure. The Labor Department lists 30 tasks for radiologists. They include interpreting images, preparing reports, reviewing histories, communicating results, conferring with clinicians, determining procedures, performing interventions, and managing safety. (National Center for Occupational Information Network Development 2026) No single task is “the radiologist.” The occupation is the bundle, together with the judgment and responsibility attached to it.

Figure 4: Paraphrased task categories drawn from the 30 tasks in the Labor Department’s 2026 radiologist profile. The highlighted category is the part of the bundle most directly reorganized in the MASAI screening workflow. Source: National Center for Occupational Information Network Development (2026).

The model separates three forces that public discussion often collapses into one. Automation creates a displacement effect: holding output fixed, fewer labor hours are required for the tasks the machine now performs. Lower costs also create a productivity effect. Output may expand, increasing demand for the human work that remains. Innovation can then create new tasks in which labor has an advantage. Acemoglu and Restrepo call this the reinstatement effect. Displacement removes labor from existing tasks. Productivity can expand the market. Reinstatement creates new reasons to employ labor. (Acemoglu and Restrepo 2018, 2019)

That is why automation does not simply subtract an activity from a fixed job. Once a machine performs a task, the worker has time to do something else. That time can become more output, more difficult cases, a service that was previously too expensive, shorter work, or fewer labor hours purchased. The occupation is rebundled around the new division of labor. The resulting bundle then becomes the starting point for the next round of automation.

Figure 5: Schematic of displacement and reinstatement in the task-based framework. Automation moves the automation frontier, transferring existing tasks from labor to capital. New task creation extends the task set in activities assigned to labor. Conceptual illustration based on Acemoglu and Restrepo (2018, 2019).

The contrast between the weaver and the radiologist now becomes precise. The power loom automated the productive task around which the weaver’s market value had been organized. In mammography, AI reduced routine readings inside a broader bundle of interpretation, intervention, communication, and responsibility. The mechanism was the same. The location of the automated task inside the occupation was not.

The balance between displacement and reinstatement is not automatic. Demand matters. If one radiologist can read twice as many scans and the number of scans remains fixed, fewer reading hours are required. If lower cost and faster diagnosis expand screening enough, demand for radiologists can grow. Control matters too. A productivity gain can become better service, higher output targets, narrower discretion, or fewer paid hours. Employment, wages, autonomy, and bargaining power can therefore move in different directions.

Exposure studies show where this process might begin. Eloundou and coauthors rated O*NET tasks according to whether a large language model, alone or with complementary software, could reduce completion time by at least half without lowering quality. They then aggregated those task ratings into occupation-level exposure indices. (Eloundou et al. 2024)

Figure 6: Human-rated potential LLM task exposure across 774 base occupation records, with specialty records excluded. Scores describe technical capability, not predicted job loss. Source: Eloundou et al. (2024); OpenAI’s GPTs-are-GPTs data and code repository.

So, Will AI Take Your Job?

Figure 6 does not divide occupations into safe and doomed groups. It measures the potential exposure of existing tasks under one study’s assumptions. Radiologists scored 51, but that is not a 51 percent probability of losing the job. The score does not tell us whether a system will be adopted, whether it will replace or assist the worker, how demand will respond, or whether new tasks will emerge. It also measures exposure to language models, not every form of AI.

The task-based framework supplies a map, not a verdict. So, will AI take your job? It depends on which tasks it takes. It depends on whether the remaining tasks become more valuable, whether lower costs create more demand, whether new tasks arrive, and whether the current worker can reach them. It also depends on how labor and capital bargain over the time and productivity released by the machine. (Acemoglu and Johnson 2024)

History permits optimism and caution. Robert Allen estimates that British output per worker rose 46 percent between 1780 and 1840 while the real wage rose only 12 percent. During this Engels’ Pause, productivity advanced for decades before workers shared proportionately in the gain. Between 1840 and 1900, output per worker rose 90 percent and the real wage 123 percent. Crafts’s newer estimates substantially narrow the gap and challenge Allen’s distributional interpretation, although consumption earnings still lagged productivity for part of the period. The abundance was real. So was the delay. (Allen 2009; Crafts 2022)

New work is not hypothetical. Autor and coauthors find that a majority of current U.S. employment is in job specialties introduced since 1940. That does not mean new work will arrive quickly enough, in the right places, or in forms accessible to displaced workers. It shows that the task set can expand. (Autor et al. 2024)

AI is already moving automation into knowledge work. If the same intelligence enters robots, it may return to physical work with greater reach. Goods and services that are scarce today could become cheap. That is a possibility, not a promise. AI may make abundance possible. How that abundance is divided will emerge from the continuing interaction between labor and capital.

Sources and notes

• Acemoglu, Daron, and Simon Johnson. “Learning from Ricardo and Thompson: Machinery and Labor in the Early Industrial Revolution and in the Age of Artificial Intelligence.” Annual Review of Economics 16 (2024): 597-621.

• Acemoglu, Daron, and Pascual Restrepo. “The Race between Man and Machine: Implications of Technology for Growth, Factor Shares, and Employment.” American Economic Review 108, no. 6 (2018): 1488-1542.

• Acemoglu, Daron, and Pascual Restrepo. “Automation and New Tasks: How Technology Displaces and Reinstates Labor.” Journal of Economic Perspectives 33, no. 2 (2019): 3-30.

• Allen, Robert C. “Engels’ Pause: Technical Change, Capital Accumulation, and Inequality in the British Industrial Revolution.” Explorations in Economic History 46, no. 4 (2009): 418-435.

• Autor, David, Caroline Chin, Anna Salomons, and Bryan Seegmiller. “New Frontiers: The Origins and Content of New Work, 1940-2018.” Quarterly Journal of Economics 139, no. 3 (2024): 1399-1465.

• Christensen, Eric W., YoonKyung Chung, Elizabeth Y. Rula, and Jay R. Parikh. “Changes in the Radiology Practice Landscape and Indicators of Practice Consolidation From 2014 to 2023.” American Journal of Roentgenology 223, no. 2 (2024).

• Crafts, Nicholas. “Slow Real Wage Growth During the Industrial Revolution: Productivity Paradox or Pro-Rich Growth?” Oxford Economic Papers 74, no. 1 (2022): 1-13.

• Eloundou, Tyna, Sam Manning, Pamela Mishkin, and Daniel Rock. “GPTs Are GPTs: Labor Market Impact Potential of LLMs.” Science 384, no. 6702 (2024): 1306-1308. Occupation-level exposure data.

• Handel, Michael J. “Growth Trends for Selected Occupations Considered at Risk from Automation.” Monthly Labor Review (2022).

• Hernström, Veronica, et al. “Screening Performance and Characteristics of Breast Cancer Detected in the Mammography Screening with Artificial Intelligence Trial (MASAI).” The Lancet Digital Health 7, no. 3 (2025): e175-e183.

• Malhotra, Ajay, et al. “Recent Trends in Academic Versus Nonacademic Radiologist Compensation and Clinical Productivity.” Journal of the American College of Radiology 22, no. 4 (2025): 486-494.

• National Center for O*NET Development. “O*NET OnLine: Radiologists, 29-1224.00.” Accessed August 15, 2026.

• U.S. Bureau of Labor Statistics. “Consumer Price Index for All Urban Consumers: U.S. City Average, All Items.” 2024.

• Wood, George Henry. The History of Wages in the Cotton Trade During the Past Hundred Years. Manchester: Sherratt and Hughes, 1910.