In this paper, we study the long-run effects of the largest R&D shock in U.S. history. In World War II, the newly-created Office of Scientific Research and Development (OSRD) led an expansive effort to develop technologies and medical treatments for the Allied war effort. From 1940 to 1945, OSRD engaged industrial and academic contractors in more than 2,200 R&D contracts at over $9 billion (2022 dollars), despite no pre-war tradition of funding extramural (externally-performed) R&D. At the height of the war, the U.S. government was funding the research behind nearly 1 of every 8 U.S. patents—more than five times pre-war and modern levels, and nearly twice the level at the peak of the Cold War in the 1950s and 1960s. The immediate effect of these investments was a range of technological advances which were not only instrumental to the success of the Allied campaign, but also of wide civilian value after the war ended. Its longer-run impact was to reshape the U.S. innovation system.
Related: Moonshot: Public R&D and Growth and Public R&D Spillovers and Productivity Growth and Pentagon Plans Vast AI Fleet to Counter China Threat
The figure shows the path through time of two interest rates. The lower line shows the path of the Fed’s policy rate going back to April 2022, and its projected path forward over the coming two years. The upper line shows what we call the “QT-equivalent policy rate,” which accounts for the effect of QT. The QT-equivalent rate is the policy rate needed without QT to have an effect on financial conditions equivalent to that produced by the actual policy rate with QT. What we find is that today’s policy rate of 5.375% is, under the current QT roll-off schedule, having roughly the same impact on financial conditions as a policy rate of 5.763% without QT. That’s a gap of 39bps, or 39 hundreds of a percent. But the effect of QT rises sharply going forward, as $95 billion in assets continue to roll off the balance sheet each month, before reaching a high of 100 basis points—or one full percent—in May 2025.
Related: The Grind Ahead and Inching Toward Equilibrium and Macro Outlook 2024: The Hard Part Is Over
Earlier this year, the New Tenant Rent Index (NTR) was showing significant disinflation in rent prices that have since begun passing through to decelerations in CPI shelter prices—and recently released NTR data through the third quarter suggests that even more stabilization is yet to come. Growth in Gross Labor Income—the aggregate wages and salaries of all workers in the economy—continues to decline as the labor market slows toward normal pre-pandemic growth rates. Given how tight the relationship between cyclical growth in employment/wages and housing inflation is, a deceleration in NTR has naturally followed the slower labor market of the last year.
Related: The Most Important New Disinflation Indicator and Where Is Shelter Inflation Headed? and Striking Similarities (and Differences) Between Inflation Today and In the 1970s
Affordability was worse in September than in August, as house prices and mortgage rates both increased. In September 2023, houses were the least “affordable” since 1982 when 30-year mortgage rates were over 14%. We already know affordability will be even worse in October since mortgage rates have increased further. For September: a year ago, the payment on a $500,000 house, with a 20% down payment and 6.11% 30-year mortgage rates, would be around $2,427 for principal and interest. The monthly payment for the same house, with house prices up 4.0% YoY and mortgage rates at 7.20% in September 2023, would be $2,822 - an increase of 16%. However, if we compare to two years ago, there is huge difference in monthly payments. In September 2021, the payment on a $500,000 house, with a 20% down payment and 2.90% 30-year mortgage rates, would be around $1,665 for principal and interest. The monthly payment for the same house, with house prices up 15.1% over two years and mortgage rates at 7.20% in September 2023, would be $3,125 - an increase of 88%!
Related: Higher For Longer and The 2024 Housing Outlook and America's Missing Empty Homes and With Housing, Millennials Have Much to Complain About
For decades, the biggest foreign holders of US Treasuries were central banks and sovereign wealth funds around the world. Foreign official institutions were buying Treasuries because many countries, in particular emerging markets, were intervening to limit the appreciation of their domestic currencies because a domestic currency that is too strong hurts exports. In other words, the foreign official sector was not buying Treasuries for yield reasons but for FX reasons to support the US dollar and thereby domestic exports. With the Fed raising rates and the dollar going up, that has now changed. Foreign central banks no longer need to buy US Treasuries and US dollars to depreciate their currencies. And foreign private buyers find US yield levels attractive despite high hedging costs. The bottom line is that with the Fed raising rates and the dollar going up, yield-insensitive central banks have been selling Treasuries to limit the weakening of their domestic currencies.
Related: Preferred Habitats and Timing in the World’s Safe Asset and Resilience Redux in the US Treasury Market and Setser On Foreign Demand For Treasuries
To see how deeply Huawei and the Chinese government are now entwined, look no further than the launch in August of the new Mate 60 Pro smartphone. Huawei timed the release of the phone to coincide with US Commerce Secretary Gina Raimondo’s visit to China in part because of direct encouragement from a senior official at the top of the regime, according to a person familiar with the situation who asked not to be identified discussing sensitive matters. Huawei never disclosed technical details, but a teardown of the handset conducted by TechInsights for Bloomberg News found it was powered by SMIC’s advanced 7-nanometer processor. That suggests China is roughly five years behind the current most advanced technology. Export controls imposed by the Biden administration in 2022 were aimed at keeping China at least eight years behind.
Related: Huawei Building Secret Network for Chips, Trade Group Warns and China AI & Semiconductors Rise: US Sanctions Have Failed and China Imports Record Amount of Chipmaking Equipment
This paper’s main message is that historical mobility was lower than previously estimated in linked data. To show why, I account for two measurement issues: unrepresentative samples and measurement error. First, I account for unrepresentative samples by adding Black families, who historical studies routinely drop. Second, I address measurement error by using multiple father observations to more accurately capture his permanent economic status. Using linked census data from 1850 to 1940, I show that accounting for race and measurement error can double estimates of intergenerational persistence. Updated estimates imply that there is greater equality of opportunity today than in the past, mostly because opportunity was never that equal.
Related: Chetty and Saez Debunk the Claim That Income Mobility is Declining in the U.S. and The Inheritance Of Social Status: England, 1600 to 2022 and The Economics of Inequality in High-Wage Economies
Here is my take: In the US in 1982, the top of the first Forbes 400 list was Daniel Ludwig with nominal $2 billion. That was 85,000 times the then-median nominal family income of $23,430. In 2023, the top of the Forbes 400 was Elon Musk with nominal $251 billion. That was 2,500,000 times the now-median nominal family income of $98,705. Now: ($251B/$99K)/($2B/$23K) = 29.8 How the f*** is the ratio of the top to the median to explode by a factor of 30 while the Auten/Splinter measures show “little change in after-tax top income shares”? Until someone comes up with an explanation for how this could be—how a 30x multiplication since 1982 of the ratio of the top of the Forbes 400 to median household income is consistent with “top income shares are lower and have increased less since 1980 than other studies… increasing government transfers and tax progressivity have resulted in… little change in after-tax top income shares…”—I am going to presume the chances are 99% that there are big things wrong in the numbers in Auten/Splinter.
Related: Income Inequality in the United States: Using Tax Data to Measure Long-Term Trends and The Economics of Inequality in High-Wage Economies
According to the new BEA estimates, the share of GDP spent on health in 2022 declined to 17.1% from 17.5% in 2021 and 17.9% in 2020. This marks the second consecutive year of decline in health spending as a share of GDP. The estimated share in 2022 represents the smallest share of the economy spent on health since 2014 and a smaller share that was spent on health in 2019, before the start of the pandemic. Adjusting for differences in overall and medical inflation, the share of GDP spent on health in 2022 was largely unchanged from 2020 and 2021. As such, the decline in health spending as a share of GDP in 2021 and 2022 was primarily driven by high inflation in sectors outside of health and a lagging medical inflation.
Related: A Huge Threat to the U.S. Budget Has Receded. And No One Is Sure Why and Why Medicare and Social Security Are Sustainable
Remittances rose considerably in the aftermath of the balance sheet expansion following the global financial crisis of 2007-08; they went from 0.2% of GDP and 1.3% of government receipts in 2007 to 0.6% and 3.4%, respectively, in 2015. Remittances then fell due to the 2015-18 tightening cycle, but they rose again in 2020 as the Fed slashed interest rates and resumed its balance sheet expansion (additionally, GDP fell in 2020, which partly explains the positive jump). Between 2021 and 2022, remittances as a percent of GDP dropped from 0.5% to 0.3%. Once the Fed returns to earning a positive net income, it will pay down the value of the deferred asset until it reaches zero, at which point the Fed will resume sending remittances to the Treasury. As of Nov. 8, 2023, the Fed had accumulated a deferred asset of $116.9 billion. In April 2023, the New York Fed estimated that the Fed will return to positive net income in 2025. Combining those New York Fed projections with the latest data on net income, we estimate that the Fed will carry this deferred asset until mid-2027, after which it will resume transfers to the Treasury.
By the end of 2018, there was a decrease of 140,000 H-1B approvals (relative to trend) and an unprecedented spike in H-1B denial rates. Denial rates increased from about 6% in 2016 to 16% in 2018. Our event-study estimates imply that a 10 percentage point increase in H-1B denial rates increases Canadian applications by 30%. A back-of-the-envelope calculation suggests that for every four forgone H-1B visas, there is an associated increase of one Canadian application. We find that firms that were relatively more exposed to the immigrant inflow increased sales. Consistent with the increase in production, we find that a firm hired approximately 0.5 additional native workers per new immigrant. We also find that the earnings per native worker at relatively more exposed firms dropped. This result together with the fact that more exposed firms are intensive in occupations that were more impacted by U.S. restrictions, is consistent with earnings per native worker in more affected occupations declining compared to less affected ones.
Related: America’s Got Talent, but Not Nearly Enough and Top Talent, Elite Colleges, and Migration: Evidence from the Indian Institutes of Technology and The Economics of Inequality in High-Wage Economies
We evaluate progress in the War on Poverty as President Lyndon B. Johnson defined it, which established a 20% baseline poverty rate and adopted an absolute standard. While the official poverty rate fell from 19.5% in 1963 to 10.5% in 2019, our absolute full-income poverty measure—which uses a fuller income measure and updates thresholds only for inflation—fell from 19.5% to 1.6%. However, we also show that relative poverty reductions have been modest. Additionally, government dependence increased over this time, with the share of working-age adults receiving under half their income from market sources more than doubling.
Related: Work Requirements and the Lost Lessons of 1996 and The Unexpected Compression: Competition at Work in the Low Wage Labor Market and The Economics of Inequality in High-Wage Economies
In this paper, we study the long-run effects of the largest R&D shock in U.S. history. In World War II, the newly-created Office of Scientific Research and Development (OSRD) led an expansive effort to develop technologies and medical treatments for the Allied war effort. From 1940 to 1945, OSRD engaged industrial and academic contractors in more than 2,200 R&D contracts at over $9 billion (2022 dollars), despite no pre-war tradition of funding extramural (externally-performed) R&D. At the height of the war, the U.S. government was funding the research behind nearly 1 of every 8 U.S. patents—more than five times pre-war and modern levels, and nearly twice the level at the peak of the Cold War in the 1950s and 1960s. The immediate effect of these investments was a range of technological advances which were not only instrumental to the success of the Allied campaign, but also of wide civilian value after the war ended. Its longer-run impact was to reshape the U.S. innovation system.
Related: Moonshot: Public R&D and Growth and Public R&D Spillovers and Productivity Growth and Pentagon Plans Vast AI Fleet to Counter China Threat
The data are quite clear: over the past 4 years, inflation-adjusted wages are up! This is also true if you start roughly right before the pandemic (December 2019 or January 2020 or thereabouts). And not only are inflation-adjusted wages up, they are up roughly the same amount as they were in the years before the pandemic. CPI-adjusted wages are a touch below: about 3% growth over 4 years, versus roughly 4% from 2015-2019. But PCE-adjusted wages are right on track, at around 5% cumulative 4-year growth. It’s true right now that if we start the data in January 2021, at the beginning of the Biden Presidency, CPI-adjusted wages are down slightly: about 1%. But PCE-adjusted wages are up slightly: also about 1%. But unless there is a major reversal of the trajectory of either wage or price growth, by next year these will both be positive (even if only slightly).
Related: Have Workers Gotten A Raise? and Are Real Wages Rising? and The Unexpected Compression: Competition at Work in the Low Wage Labor Market
To get a clearer picture of the economy, therefore, we need to adjust for the changing composition of the workforce and consider changes to wages in each type of job and industry. A BLS statistic, the National Compensation Survey’s Employment Cost Index, does just this. According to ECI, inflation-adjusted wages have shrunk by 3.7% since the end of 2020. While real wages rose in response to falling energy prices late last year, they have been roughly flat since. Worse, the drop in real wages erased all gains made in the late 2010s. Real wages today stand at 2015 levels, meaning Americans’ paychecks don’t go any further now than they did eight years ago.
Related: Have Workers Gotten A Raise? and Are Real Wages Rising? and The Unexpected Compression: Competition at Work in the Low Wage Labor Market
The massive outperformance of the “Magnificent 7” mega-cap tech stocks has been a defining feature of the equity market in 2023. The stocks should collectively outperform the remainder of the index in 2024. The 7 stocks have faster expected sales growth, higher margins, a greater re-investment ratio, and stronger balance sheets than the other 493 stocks and trade at a relative valuation in line with recent averages after accounting for expected growth. However, the risk/reward profile of this trade is not especially attractive given elevated expectations. Analyst estimates show the mega-cap tech companies growing sales at a CAGR of 11% through 2025 compared with just 3% for the rest of the S&P 500. The net margins of the Magnificent 7 are twice the margins of the rest of the index, and consensus expects this gap will persist through 2025.
Related: A Few Stocks Drive the Stock Market: Dot.com Vs. Today Vs. the Last 100 Years and Long-Term Shareholder Returns: Evidence From 64,000 Global Stocks and Birth, Death, and Wealth Creation
The divergence between the S&P7 and the S&P493 continues. Investors buying the S&P 500 today are buying seven companies that are already up 80% this year and have an average P/E ratio above 50. In fact, S&P7 valuations are beginning to look similar to the Nifty Fifty and the tech bubble in March 2000.
Related: A Few Stocks Drive the Stock Market: Dot.com Vs. Today Vs. the Last 100 Years and Long-Term Shareholder Returns: Evidence From 64,000 Global Stocks and 2024 US Equity Outlook: “All You Had To Do Was Stay”
Here is my take: In the US in 1982, the top of the first Forbes 400 list was Daniel Ludwig with nominal $2 billion. That was 85,000 times the then-median nominal family income of $23,430. In 2023, the top of the Forbes 400 was Elon Musk with nominal $251 billion. That was 2,500,000 times the now-median nominal family income of $98,705. Now: ($251B/$99K)/($2B/$23K) = 29.8 How the f*** is the ratio of the top to the median to explode by a factor of 30 while the Auten/Splinter measures show “little change in after-tax top income shares”? Until someone comes up with an explanation for how this could be—how a 30x multiplication since 1982 of the ratio of the top of the Forbes 400 to median household income is consistent with “top income shares are lower and have increased less since 1980 than other studies… increasing government transfers and tax progressivity have resulted in… little change in after-tax top income shares…”—I am going to presume the chances are 99% that there are big things wrong in the numbers in Auten/Splinter.
Related: Income Inequality in the United States: Using Tax Data to Measure Long-Term Trends and The Economics of Inequality in High-Wage Economies
[In the official reports] both the goods surplus, which is much smaller in the balance of payments than in the customs data, and balance on investment income, which remains in deficit even with the rise in U.S. interest rates, are suspicious. With reasonable adjustments, China's “true” current account surplus might be $300 billion larger than China officially reports. That's real money, even for China. The model implies China's overall income balance should now be back in a surplus of around $70 billion thanks to the rise in U.S. short-term interest rates. So without the unexplained deficit in investment income and the discrepancy between customs goods and balance of payments goods, and China’s current account surplus would now be around $800 billion, over 4 percent of its GDP.
Related: Managing Economic and Financial Entanglements With China and Can China Reduce Its Internal Balances Without Renewed External Imbalances? and Can China Reduce Its Internal Balances Without Renewed External Imbalances?
Hyperglobalization refers to the exceptional period between 1992 and 2008 during which global exports grew at close to 10% a year in nominal terms while GDP increased by only 6% a year. As a result, the share of exports in national economies grew from less than 20% to more than 30% in a little bit more than 15 years. The hyper in hyperglobalization does not come from the level of trade relative to GDP, which remains high, or from levels compared with the theoretical potential of trade, which are low. Rather it comes from the change in the level of trade, which was positive before the Global Financial Crisis (GFC) and stagnant or slightly negative thereafter. After the GFC, a puzzling wedge emerged. China’s trade-to-GDP ratio plummeted by more than 30pp, from 71% to a trough of about 35%. But its global export market share continued to rise at the same heady pace, reaching nearly 15% of total exports and 22% of manufactured exports by 2022.
Related: China's Current Account Surplus Is Likely Much Bigger Than Reported and Managing Economic and Financial Entanglements With China and Pettis On China's Export Strategy
We evaluate progress in the War on Poverty as President Lyndon B. Johnson defined it, which established a 20% baseline poverty rate and adopted an absolute standard. While the official poverty rate fell from 19.5% in 1963 to 10.5% in 2019, our absolute full-income poverty measure—which uses a fuller income measure and updates thresholds only for inflation—fell from 19.5% to 1.6%. However, we also show that relative poverty reductions have been modest. Additionally, government dependence increased over this time, with the share of working-age adults receiving under half their income from market sources more than doubling.
Related: Work Requirements and the Lost Lessons of 1996 and The Unexpected Compression: Competition at Work in the Low Wage Labor Market and The Economics of Inequality in High-Wage Economies
Due to population aging, GDP growth per capita and GDP growth per working-age adult have become quite different among many advanced economies over the last several decades. Countries whose GDP growth per capita performance has been lackluster, like Japan, have done surprisingly well in terms of GDP growth per working-age adult. Indeed, from 1998 to 2019, Japan has grown slightly faster than the U.S. in terms of per working-age adult: an accumulated 31.9% vs. 29.5%. Furthermore, many advanced economies appear to be on parallel balanced growth trajectories in terms of working-age adults despite important differences in levels. Motivated by this observation, we calibrate a standard neoclassical growth model in which the growth of the working-age adult population varies in line with the data for each economy. Despite the underlying demographic differences, the calibrated model tracks output per working-age adult in most economies of our sample. Our results imply that the growth behavior of mature, aging economies is not puzzling from a theoretical perspective.
Related: Fully Grown - European Vacation! and Population Aging and Economic Growth: From Demographic Dividend to Demographic Drag? and Growth in Working-Age Population Ends. That’s Not All Bad
Microsoft is currently conducting the largest infrastructure buildout that humanity has ever seen. While that may seem like hyperbole, look at the annual spend of mega projects such as nationwide rail networks, dams, or even space programs such as the Apollo moon landings, and they all pale in comparison to the >$50 billion annual spend on datacenters Microsoft has penned in for 2024 and beyond. This infrastructure buildout is aimed squarely at accelerating the path to AGI and bringing the intelligence of generative AI to every facet of life from productivity applications to leisure.
Related: The Growing Energy Footprint of Artificial Intelligence and The Race of the AI Labs Heats Up and Will A.I. Transform the Economy, and if So, How?
We re-run our analysis allowing for the changing structure of issuance as the Treasury leans more on bill and shorter tenor issuance in the near-term consistent with the signal from the most recent QRF round. Interest paid on debt increasing to about 3.8% in 2030. The average interest on debt approaches 3.5% at the end of the horizon. Of course, because debt-to-GDP is close to 100% throughout, these two measures are very similar. Debt-to-GDP is expected to increase to 113% of GDP by 2030 while the gross financing need (GFN), a measure of the rolling 4Q ahead deficit plus maturing securities including bills, at first increases to nearly 45% of GDP in 2025 as T-bill issuance accelerates, but declines to about 35% of GDP in the baseline as issuance shifts to longer tenors.
Related: If Markets Are Right About Long Real Rates, Public Debt Ratios Will Increase For Some Time. We Must Make Sure That They Do Not Explode and Resilience Redux in the US Treasury Market and Preferred Habitats and Timing in the World’s Safe Asset
China’s surplus in manufactured goods net of commodity imports has continued to grow relative to the economic output of China’s trade partners, thanks in large part to China’s growth relative to the rest of the world. Even though the value of Chinese exports fell in 2023, this has had no impact on China’s overall balance because the amount of money spent on imports is down as well. The past few years have even seen a renewed surge in China’s surplus (properly measured) relative to China’s own GDP thanks to exceptionally weak growth in consumer spending and the sustained plunge in homebuilding. Federal spending—financed in large part by borrowing—has helped shore up private sector balance sheets and sustain demand, even as some spending elements have contained provisions that should put a floor on sales for American producers. This policy mix helps explain why China’s growing surplus has not attracted much ire, or even notice, in the U.S.
Related: Danish Weight Loss Drugs vs. Chinese Cars: Two Models of Export Booms and Can China Reduce Its Internal Balances Without Renewed External Imbalances? and As Long As The US Is Outlet For China's Surplus Rumors Of Decoupling Are Overstated
The unauthorized immigrant population in the United States reached 10.5 million in 2021, according to new Pew Research Center estimates. That was a modest increase over 2019 but nearly identical to 2017. The number of unauthorized immigrants living in the U.S. in 2021 remained below its peak of 12.2 million in 2007. It was about the same size as in 2004 and lower than every year from 2005 to 2015. The U.S. foreign-born population was 14.1% of the nation’s population in 2021. That was very slightly higher than in the last five years but below the record high of 14.8% in 1890. As of 2021, the nation’s 10.5 million unauthorized immigrants represented about 3% of the total U.S. population and 22% of the foreign-born population. These shares were among the lowest since the 1990s.
Related: Monopsony, Efficiency, and the Regularization of Undocumented Immigrants and Immigrants & Their Kids Were 70% of U.S. Labor Force Growth Since 1995 and Immigrants’ Share of the U.S. Labor Force Grows to a New High
This fall, The Associated Press illustrated how school attendance has cratered across the United States, using data compiled in partnership with the Stanford University education professor Thomas Dee. More than a quarter of students were chronically absent in the 2021-22 school year, up from 15 percent before the pandemic. That means an additional 6.5 million students joined the ranks of the chronically absent. The problem is pronounced in poorer districts like Oakland, Calif., where the chronic absenteeism rate exceeded 61%. But as the policy analyst Tim Daly wrote recently, absenteeism is rampant in wealthy schools, too. Consider New Trier Township High School in Illinois, a revered and highly competitive school that serves some of the country’s most affluent communities. Last spring, The Chicago Tribune reported that New Trier’s rate of chronic absenteeism got worse by class, reaching nearly 38% among its seniors.
Related: NAEP Long-Term Trend Assessment Results: Reading and Mathematics and ACT Scores Fell for Class of 2023, Sixth Consecutive Decline and Looking For Flynn Effects on a Recent Online U.S. Adult Sample: Examining Shifts Within The SAPA Project
The headline estimate for the United States is a roughly 5pp decline of the labor share between 1929 and 2022. The decline after World War II is even larger, at around 7pp. The great majority of U.S. industries exhibited labor share declines in recent decades. The United States is not unique, as we observe labor share declines in most countries of Europe and Asia and in emerging markets. It helps to organize factors affecting the labor share in five categories: technology, product markets, labor markets, capital markets, and globalization. The factors that have contributed to the labor share decline are intertwined. My view is that the most plausible causes have technological origin. Developments such as the information age and automation, manifesting through changes in the cost of capital and the structure of product markets, caused the labor share to decline. If technological advancements continue to favor capital indefinitely, the natural outcome is a transition to a world in which capital on its own produces the entire global income.
Related: The Unexpected Compression: Competition at Work in the Low Wage Labor Market and Income Inequality in the United States: Using Tax Data to Measure Long-Term Trends and The Economics of Inequality in High-Wage Economies
Labor market tightness following the height of the Covid-19 pandemic led to an unexpected compression in the US wage distribution that reflects, in part, an increase in labor market competition. Disproportionate wage growth at the bottom of the distribution reduced the college wage premium and reversed almost 40% of the rise in 90-10 log wage inequality since 1980, as measured by the 90-10 ratio. The Unexpected Compression as measured by the fall in the 90-10 log wage ratio was nearly half of the Great Compression of the 1940s. The rise in wages was particularly strong among workers under 40 years of age and without a college degree. The post-pandemic rise in labor market tightness—and the consequent wage compression— represent a profound shift in US labor market conditions, seen most clearly in the rise of the wage-separation elasticity among young non-college workers.
Related: Perspectives on the Labor Share and Income Inequality in the United States: Using Tax Data to Measure Long-Term Trends and The Economics of Inequality in High-Wage Economies
Improvements in labor-saving (automation) technologies are negatively related to the wage earnings of workers in affected occupation–industry cells. For instance, an increase in our exposure measure from the median to the 90th percentile is associated with a 2.5 pp decline in the total earnings of the average worker over the next five years. These earnings losses are concentrated on a subset of workers, since exposed workers experience a 1.2pp increase in the probability of involuntary job loss over the next five years. Importantly, the magnitude of these wage declines or job loss probabilities are essentially unrelated to observable measures of worker skill—measured by age, level of wage earnings relative to other workers in the same industry and occupation, and college education. Perhaps surprisingly, but consistent with our model, new labor-augmenting technologies also lead to a decline in earnings for exposed workers, though the average magnitudes are smaller. An increase in our exposure measure from the median to the 90th percentile is associated with a 1.3pp decline in earnings growth and a 0.5pp increase in the likelihood of involuntary job loss. However, unlike in the case of labor-saving technology, the effects of exposure to labor-augmenting technologies are fairly heterogeneous: it disproportionately affects white-collar workers (defined as those with college degrees, or those employed in non-manufacturing industries or in occupations emphasizing cognitive tasks); older workers; and workers that are paid more relative to their peers (other workers with similar characteristics in the same industry and occupation).
Related: Perspectives on the Labor Share and AI Isn’t Good Enough and The Economics of Inequality in High-Wage Economies