AI and Employment: So Far, So Good

By Alex Tabarrok.

"In September 2023, the Census Bureau added questions about AI to its Business Trends and Outlook Survey. Census asked hundreds of thousands of businesses whether they had used AI in the previous two weeks to produce goods and services. At that time, 3.7% said yes; by late 2025 the figure had reached about 10%. (In November 2025 Census broadened the question to ask about AI use in any business function, producing a jump in measured adoption to about 18%.)

Twice the Bureau has asked a key question:

In the last six months, how did the use of Artificial Intelligence affect this business’s total employment?

In Dec. 2023 to Feb 24, when ~5% of firms were using AI the answers were 2.8% increased, 2.6% decreased and 94.6% reported no change. Two years later, in the Nov 2025–Feb 2026 supplement, the answers were: 2.3% increased, 2.0% decreased, and 95.7% reported no change. The answers were similar by firm size.

Some sectors reported more action. Information is the one sector where fewer than 92% report no change. But overall, almost all firms report no change and of those reporting change it’s about evenly divided between increasing and decreasing employment.

 

The supplement also asked about tasks. Among firms using AI, 44% say it supplemented or enhanced work an employee already does. Ten percent say it performed a task an employee used to do. Eleven percent say it introduced a task no one had been doing.

Among those using generative AI, 85% of firms cited writing or editing documents and email as the biggest uses, half cite searching for information, 45% summarizing documents, and 13% coding. Sixty-four percent of adopters say they changed nothing about the business in order to use AI, 15% trained existing staff, another 15% built new workflows, and just over one percent hired anyone with AI skills.

Among firms where AI has taken over some employee tasks, the degree of substitution is growing. The share reporting that AI took over “a large number” of tasks rose from 2.4% to 7.1%, while the share reporting “a moderate number” rose from 13% to 22%. But this group is still small: only about a tenth of AI adopters, who themselves make up about a fifth of firms.

I have reported firm-weighted estimates but employment-weighting gives essentially the same result. Thus, we have unusually direct evidence from a very large sample, and it says that the overwhelming majority of firms using AI do not yet report any effect on total employment. Very consistent with what Tyler and I said in our talk to OpenAI."

 

More Defense Spending Won’t Save the Economy

By Benjamin Giltner of Cato. 

"The Trump administration has failed so far to deliver on its affordability promises. Yet, in a recent Department of Defense video on X, Secretary Hegseth boasted that the administration’s $1.5 trillion proposed defense budget would “supercharge” the American economy. It’s not exactly a novel plan.

The secretary’s statement echoes a long-standing argument since the publishing of NSC-68 in 1950: More defense spending is good for the economy. Of course, as with all federal spending, defense budgets certainly do affect Americans—just not in the way Secretary Hegseth thinks. 

Instead of boosting economic growth, increased defense spending stunts the US economy, wastes money, and raises costs for Americans.

True enough, defense spending can create jobs and contribute to the economy. But this misses a more fundamental question: Which type of federal spending is most beneficial for the economy? The federal government can spend and borrow only so much money, and there are only so many resources and workers to go around. Should scientists be hired for defense research or domestic manufacturing? Should land be used for missile production or building a school? With limited resources and people, policymakers need to know how to spend federal dollars efficiently to limit waste and bloat. 

Herein lies the central problem with Hegseth’s argument: Of all federal outlays, defense spending creates the least number of jobs. And the reasoning is simple—it is a “parasitic output.” The finished products from defense spending—tanks, missiles, bullets, and so on—leave the market once they are made. When that $4 million Patriot missile is built, that’s it. That $4 million either sits in storage or explodes in combat. Parasitic output is accounted for as part of a country’s gross domestic product, which is why, among other reasons, measuring defense spending as a contribution to GDP is misleading. 

Increased defense spending also weakens America’s manufacturing industry, an economic sector in rough shape these days. The workers, research, and capital that could’ve been used to strengthen domestic manufacturing are being used to make weapons. Yes, building new weapons may increase employment rates. But such an obsessive focus on defense production means missing out on the wider employment and economic benefits of manufacturing other products with higher returns on investment. 

Additionally, increased defense spending puts upward pressure on inflation. As the federal government pumps more money into the economy with little return, inflation rises. To offset this, governments have three primary options: increase interest rates, raise taxes, or reduce spending in other sectors. All three options are politically unpopular. 

Reducing defense spending is the logical position for policymakers to take. Reforming the weapon acquisition process and walking back US military commitments abroad, for instance, are compelling policy options. But bolder action is needed. A spending cap should be placed on the defense budget, which is in fact how these budgets were made prior to the 1960s. Such a cap would force the military to make use of set funds, laying down an imperative to spend efficiently.

Matching the defense budget to America’s national interests makes sense in theory. And indeed, this is what the current Planning, Programming, Budgeting, and Execution process aims to do. Yet, threat inflation regularly goads Congress into paying any price to safeguard against exaggerated threats.

Proponents of hiking the defense budget argue that proposals to reduce defense spending put money before national security and that less spending in a world characterized by risk is radical. But what is truly radical is the notion that the United States can sustain its exorbitant defense spending indefinitely. It’s also radical to suppose that there are no trade-offs with federal spending. And it is radical to separate economic conditions from national security. 

If the Trump administration is serious about lowering costs for American families, it cannot pretend that defense spending is somehow exempt from basic economic realities. A larger Pentagon budget does not create prosperity out of thin air. Lawmakers will need to scrutinize defense spending more heavily if they hope to fix the country’s economic woes."

We’re Preparing for the Wrong AI Labor Crisis

Mass unemployment is unlikely. AI will reorganize the white-collar corporate workforce, not destroy it.

By Stephen Lewarne. He is a professor of economics at Franciscan University. Excerpts:

"the notion that the economy faces mass technological unemployment doesn’t fit the evidence"

"the broader market continues to show relatively stable aggregate demand for labor. In March, the unemployment rate stood at 4.3%, close to both the Federal Reserve’s estimate of long-run normal unemployment and Congressional Budget Office projections for the coming decade. Total nonfarm payroll employment increased by 178,000 jobs during the month, while healthcare added 76,000 jobs and averaged roughly 29,000 new jobs a month over the prior year."

"U.S. entry-level job postings have fallen roughly 35% since January 2023, with highly AI-exposed entry-level postings declining more than 40%."

"Employer surveys indicate a substantial shift away from GPA-based screening and toward skills-based hiring. Employers increasingly emphasize demonstrated competencies, project-based experience and practical problem-solving abilities." 

[there is] "a labor-market transition more complicated than conventional automation narratives suggest." 

Youth unemployment in Canada near record highs since 2022; unprecedented levels outside of a recession

By Philip Cross of The Fraser Institute. 

The Extraordinary Increase of Youth Unemployment in Canada

• The surge in unemployment for Canada’s young people (ages 15–24) since 2022 has been extraordinary. The upturn reverses a decades-long gradual downward trend.
• Both the speed of the increase and the level of youth unemployment reached are unprecedented for an economy not in recession.
• Youths who were unemployed remained jobless for the longest period ever on record (data starts in 1976).
• The gap between the unemployment rates for youths (13.8 percent) and adults (5.7 percent) in 2025 reached a near all-time high (8.1 percentage points).
• The spread (3.8 percentage points) between the Canadian youth unemployment rate (13.8 percent) and the US rate (10.0 percent) in 2025 approached its all-time high.
• Canada’s youth unemployment rate has been higher than the US rate since 2015, indicating a sustained period of higher unemployment for Canada’s youth compared to our southern neighbour.
• Finally, the unemployment rate for 15–19-year-olds (19.5 percent) was at a near-record high, outside of the COVID-19 pandemic (2020), and the gap (8.6 percentage points) with 20–24-year-olds (10.9 percent) was also at a near all-time high.
• Several government policies contributed to the historic increase in Canada’s youth unemployment.
• Specifically, the increase in immigrant labour (increased supply) and simultaneous increases in minimum wages in most provinces (decreasing demand) are key explanations for the marked rise in youth unemployment.
• It is noteworthy that the weakness in youth employment has been concentrated in the retail trade and accommodation and food services sectors, where 70 percent of youth jobs are situated.

 

The effects of unemployment benefit duration

What happened to the US labor market after the Emergency Unemployment Compensation Act expired after the Great Recession?

By Tyler Smith of The AEA. 

"In December 2013, when Congress failed to reauthorize the Emergency Unemployment Compensation Act, many prominent economists predicted a substantial decline in employment and labor force participation.  

In a paper in the American Economic Journal: Macroeconomics, authors Marcus Hagedorn, Iourii Manovskii, and Kurt Mitman show, to the contrary, that this abrupt end to unemployment benefits actually led to a surge in employment and labor force growth, especially in states with larger cuts in benefit duration.

The sudden termination of federal support for unemployment benefits and its variation across states provided an ideal natural experiment for understanding the relationship between these benefits and the labor market.

Figure 1 from the authors’ paper shows the reform's impact through two panels tracking employment-to-population ratios and labor force participation rates. The dashed vertical line indicates the expiration of the Emergency Unemployment Compensation Act.

 

The chart displays the difference between states that had high benefits and low benefits before the reform, normalized to zero in the fourth quarter of 2013. Prior to the reform, both panels show a steady downward trend, indicating that high-benefit states experienced persistently deteriorating labor markets relative to low-benefit states.

In 2014, the downward trend abruptly changed direction. Employment in previously high-benefit states surged by 0.008 points relative to low-benefit states within a year, completely reversing the multi-year decline. Labor force participation in generous states also recovered, rising by nearly 0.005 points relative to less generous states. This sharp discontinuity at precisely the moment of the policy change, along with the absence of other major changes in the policy environment, provides strong evidence that the benefit cut drove the employment recovery.

The researchers estimate that a 1 percent reduction in benefit duration increased employment by 0.02 log points after four quarters. Nationally, this translated to 2.5 million additional employed Americans by late 2014, accounting for 75 percent of that year's employment growth.

The findings suggest that while unemployment insurance provides crucial support during economic downturns, extended benefits may delay full recoveries.

♦

“The Impact of Unemployment Benefit Extensions on Employment: The 2014 Employment Miracle?” appears in the October 2025 issue of the American Economic Journal: Macroeconomics."

Why Ford Can’t Find Mechanics

Forget about reshoring manufacturing without more skilled workers

WSJ editorial. Excerpts:

"Government subsidies for college and graduate education have encouraged the young to go to college even though they might be better off learning a trade. This has created a skills mismatch in the labor market. Unemployment among young college grads is increasing, while employers struggle to hire skilled manufacturing workers, technicians and contractors.

Only 114,000 Americans in their 20s completed vocational programs during the first 10 months of last year, compared to 1.24 million who graduated from four-year colleges and 405,000 who received advanced degrees. Yet recent bachelor’s recipients in their 20s were 5.6 percentage points less likely to be employed than those who finished vocational programs."

"one third of small business owners reported jobs they couldn’t fill, and 49% reported few or no qualified applicants for positions they were trying to fill." 

No, AI Robots Won’t Take All Our Jobs

Instead, they will boost productivity, lower prices and spur the evolution of the labor market

By Robert D. Atkinson. He is president of the Information Technology and Innovation Foundation. Excerpts:

"In 2013 Carl Benedikt Frey and Michael A. Osborne of Oxford University produced a research paper estimating that 47% of U.S. employment was at risk of being eliminated by new technologies."

"humans have experienced technological disruptions before, and we adapted to meet them."

"In the first half of the 20th century, tens of thousands of men and boys across America worked as pinsetters in bowling alleys."

"In the 1920s and ’30s, elevator companies began installing “robot elevators” with automatic controls, and eventually elevator operators all but disappeared."

"the decline of agricultural field workers due to motorized tractors to the rise and fall of “motion picture projectionists,” who operated projectors in movie theaters. Entire categories of jobs were wiped out, yet automation has never created a mass lumpenproletariat."

"AI doomsayers frequently succumb to what economists call the “lump of labor” fallacy: the idea that there is a limited amount of work to be done, and if a job is eliminated, it’s gone for good."

"the saving from increased productivity is recycled back into the economy in the form of higher wages, higher profits and reduced prices. This creates new demand that in turn creates new jobs."

"According to Goldman Sachs Research economists, broad adoption of AI could boost the country’s productivity growth by 1.5 percentage points per year."

"about 20 million U.S. workers are fired or laid off every year. In other words, the supposed AI job apocalypse, if it occurred, would be the equivalent of only about six weeks of normal labor-market churn."

Are Americans Worse Off Since NAFTA? The Data Say No

By Donald J. Boudreaux.

"What do the following three dates have in common with each other?

1975.

January 1, 1994.

December 11, 2001.

Think for a moment.

Give up?

Each of those three dates is one protectionists routinely identify — either explicitly or by implication — as marking a significant turn for the worse in Americans’ economic fortunes due to international trade.

1975 is the last year in which America ran an annual trade surplus. That year was the last to see the dollar value of Americans’ exports of goods and services exceed the dollar value of Americans’ imports of goods and services. Every year since then — 2025 is on track to be the fiftieth — the United States has run annual trade deficits.

January 1, 1994, is the date that the North American Free Trade Agreement (NAFTA) took effect. Despite some minor modifications and a name change (to United States-Mexico-Canada Agreement, or USMCA) in 2020, this trade agreement remained in effect until just a few weeks ago when Pres. Trump unilaterally moved the US into violation of it.

December 11, 2001, is the date on which China gained membership in the World Trade Organization (WTO), thus further increasing that country’s trade with many other countries, including the US.

If you pay attention to pronouncements, discussions, and debates in the US over trade you’ll regularly encounter laments about America’s “chronic trade deficits” (implying that new economic troubles began as 1975 ended), complaints about the depredations visited on Americans by NAFTA (suggesting that Americans’ economic fortunes began to worsen as 1994 dawned), and, of course, dire warnings of the alleged economic dangers that we Americans encounter by trading with the Chinese (implying that America’s economic health only took another turn for the worse as 2001 was coming to a close).

Fortunately, a great deal of relatively straightforward economic data allows us to put these common claims to the test. As is true of any economic data — which, after all, are drawn from an incredibly complex, dynamic, and ever-changing real-world economy — the data that I present below are incapable of proving anything.

Counterfactuals can always be offered, and questions about classifications, excluded variables, missing data, measurement methods, and errors, are never off the table. Nevertheless, even imperfect data can be sufficiently accurate to be revealing. At any rate, when used honestly such data are often an important part — although never the exclusive part — of any sound analysis or argument about economic policy.

So here are six economic phenomena the trends in which are useful to consult to test the claims about trade described above:

1. workers’ real earnings
2. US industrial production
3. US industrial capacity
4. US capital stock
5. US manufacturing employment as a share of total employment
6. inflation-adjusted average household net worth

These six phenomena by no means exhaust the potential data that are useful for judging the debate between protectionists and free traders, but they’re a good start.

Workers’ Real Earnings

Among the most meticulous and informed scholars who research economic trends is the American Enterprise Institute’s Scott Winship. His 2024 study Understanding Trends in Worker Pay over the Past 50 Years (PDF) is well worth a full and careful read. Among other achievements, it reveals the many sloppy uses of statistics by pundits who argue that ordinary Americans are today no better off economically than were ordinary Americans when the White House was occupied by Gerald Ford. But for our purposes, what Winship shows in his Figure 6 suffices: Today, average inflation-adjusted total compensation (wages and fringe benefits) for workers in the nonfarm business sector is about 250 percent higher than it was in 1975, 100 percent higher than in the year NAFTA took effect (1994), and about 40 percent higher than when China joined the WTO (2001).

US Industrial Production

US industrial production is today (February 2025) at an all-time high. Its previous all-time high was in September 2018, after which it leveled off — perhaps due to Trump’s first round of tariffs — and then fell (of course) during COVID. From the end of COVID until now it’s been largely flat and just below its September 2018 level until finally eking out a new all-time high in February of this year.

One plausible culprit for this leveling-off is the barrage of tariffs that began in early-to-mid-2018 and were mostly retained during the Biden years. Because more than half of American imports are inputs used in domestic production, these trade restrictions dampen the productivity of American producers.

Contrary to claims too numerous to count, the American economy has not been deindustrialized. Americans today (January 2025) produce 153 percent more industrial output than in 1975, 55 percent more than in January 1994, and 18 percent more than in December 2001.

US Industrial Capacity

Nor has the American economy’s capacity to produce industrial output been, as is often asserted, “hollowed out.” America today has an industrial capacity that is 146 percent larger than it was in 1975, 64 percent larger than in January 1994, and 12 percent larger than in December 2001. This reality is especially notable given that as we Americans grow richer, we spend more on services relative to what we spend on goods.

US Capital Stock

In 2019 (the last year for which these data are available), the inflation-adjusted size of the US capital stock was 178 percent larger than it was in 1975, 66 percent larger than in 1994, and 36 percent larger than in 2001.

US Manufacturing Employment as a Share of Total Nonfarm Employment

The accompanying graph shows, from January 1939 through February 2025, manufacturing employment as a share of total nonfarm employment in the US.

 

I constructed this graph from two different data sources available at the St. Louis Fed’s FRED data site. As you can see,  protectionists are correct when they note that manufacturing employment as a share of total employment has fallen since 1975, has continued to fall since 1994, and has fallen further still since 2001. But as you can also see, this trend dates back long before 1975. Excluding the wartime (November 1943) peak, manufacturing workers as a share of all nonfarm workers started falling rather steadily in 1954, 22 years before America began its still-unbroken run of annual trade deficits, 40 years before NAFTA, and 47 years before the WTO accepted China as a member.

It’s true that the average monthly rate of decline in the share of workers employed in manufacturing has been very slightly faster from 1976 through today (February 2025), at 0.171 percent, than it was from 1954 through 1975, at 0.134 percent. But it’s also true that the average monthly rate of decline in the share of manufacturing workers has slowed since China joined the WTO. From January 1976 through November 2001, the share of manufacturing workers fell at an average monthly rate of 0.192 percent; from December 2001 through February 2025, that rate dropped to 0.146 percent.

Manufacturing employment as a share of total employment is falling not because of trade but because of improvements in labor-saving technologies, combined with Americans’ rising demand for services relative to the demand for goods.

Inflation-adjusted Average Household Net Worth

If protectionists are correct that US trade deficits either push Americans further into debt to foreigners or result from Americans selling too many assets to foreigners, the net worth of American households would have fallen over the past half-century, as these years saw an unbroken string of annual US trade deficits. But the opposite has happened. As I explain in this short blog post at Café Hayek, the average inflation-adjusted net worth of a US household today (2024) is 232 percent higher than it was in 1975, 140 percent higher than in 1994, and 78 percent higher than in 2001. In short, we Americans have gotten richer as we’ve run trade deficits and since NAFTA took effect as well as since China joined the WTO.

Again, none of these six trends in the data proves that protectionists are mistaken to argue that freer trade and trade deficits have harmed America economically. But taken together they should at least put the burden of proof squarely and heavily where it belongs — namely, on protectionists. It is they, and not free traders, who advocate government-imposed restrictions on Americans’ economic liberties."