Macro Roundup

Acemoglu/Autor/Hazell/Restrepo find that despite a rapid growth in demand for workers with AI expertise outside of those individuals the diffusion of AI into businesses hasn’t impacted overall employment (yet) despite AI replacing humans in some tasks, “…. we find no discernible relationship between AI exposure and employment or wage growth at the occupation or industry level, implying that AI is currently substituting for humans in a subset of tasks but it is not yet having detectable aggregate labor market consequences….”

Overall finding“…We study the impact of AI on labor markets, using establishment level data on vacancies with detailed occupational information comprising the near-universe of online vacancies in the US from 2010 onwards. We classify establishments as “AI exposed” when their workers engage in tasks that are compatible with current AI capabilities. We document rapid growth in AI related vacancies over 2010-2018 that is not limited to the Professional and Business Services and Information Technology sectors and is significantly greater in AI-exposed establishments. AI-exposed establishments are differentially eliminating vacancy postings that list a range of previously-posted skills while simultaneously posting skill requirements that were not previously listed. Establishment-level estimates suggest that AI-exposed establishments are reducing hiring in non-AI positions as they expand AI hiring. However, we find no discernible relationship between AI exposure and employment or wage growth at the occupation or industry level, implying that AI is currently substituting for humans in a subset of tasks but it is not yet having detectable aggregate labor market consequences….Taken together, we interpret these results as indicating that there is a genuine increase in the deployment of AI technologies throughout the US economy and this increase is much more pronounced in establishments that have task structures suitable for deploying AI-powered algorithms. Reflecting the potential for substitution between algorithms and humans, we detect nontrivial reductions in non-AI hiring in more AI-exposed establishments, though these estimates are not uniformly robust with all three of our AI exposure measures. We do not detect any aggregate relationships between AI exposure and either employment or wages, plausibly because AI technologies are still in their infancy and have spread only to a limited part of the US economy. ”

Takeaway, “…Our results consistently show no positive relationship between AI exposure and establishment hiring. Rather, we find evidence of lower hiring associated with greater AI exposure in almost all of the specifications using the Felten et al. measure and in most specifications with the Webb measure. The timing of these relationships is also plausible, concentrating in the time window between 2014 and 2018, during which AI vacancy postings surged. This pattern of results, combined with our estimatesshowing that AI adoption is concentrated in establishments with more AI-exposed tasks, suggests that the recent AI surge is driven in part by task substitution whereby AI automates a subset of tasks formerly performed by labor. We find no evidence for either the view that there are major human-AI complementarities in these establishments or the expectation that AI will increase hiring because of its large productivity effects—though of course we cannot rule out that other applications of AI not captured here could have such effects. In contrast to the establishment-level patterns, we do not detect any relationship between AI exposure and overall employment or wages at the industry or occupation level. There are no significant employment impacts on industries that have a task structure that exhibit greater exposure to AI, and also no employment or wages effects for occupations that are more exposed to AI. We conclude that despite the notable surge in AI adoption, the impact of AI is still too small relative to the scale of the US labor market to have had first-order impacts on employment patterns—outside of AI hiring itself…”

However overall the impact of increased AI exposure on employment or wages at the aggregate occupation or industry level doesn’t show up in data. So isn’t AI having aggregate labor market consequences (yet)

First core finding, employment market for persons with AI expertise took off around 2015-2016“…Our first result is that there is a rapid takeoff in AI vacancy postings starting in 2010 and significantly accelerating around 2015-2016. Moreover, consistent with a task-based view of AI, this AI adoption is driven by establishments with task structures that are compatible with current AI capabilities. For instance, a one standard deviation increase in our baseline measure of AI exposure based on Felten et al.—which is approximately the difference in the average AI exposure between finance versus the mining and oil extraction sector—is associated with 15% more AI vacancy posting. The strong association between AI exposure and subsequent AI hiring is robust to numerous controls and specification checks when using the Felten et al. and the Webb measures, but not robust with SML. Because the SML measure is not strongly predictive of AI adoption, we place greater emphasis on the other two measures, Felten et al. and Webb…”

Second core finding, firms seems to be substituting AI for labor in some cases, however net employment impact is limited“…. We then investigate whether AI exposure is associated with changes in the skill content of posted vacancies. With the Felten et al. and Webb measures (and with the SML measure to a lesser degree),we find that AI exposure is associated with both a significant decline in some of the skills previously demanded in vacancies and the emergence of new skillThis evidence bolsters the case that AI is altering the task structure of jobs, replacing some human-performed tasks while simultaneously generating new tasks accompanied by new skill demands….”

They find that AI vacancies are growing rapidly, even outside IT, and especially in AI exposed firms. These firms are responding by posting vacancies for new skills that they did not previously require, and posting fewer vacancies for the skills that they previously required presumably as AI substitutes for labor. These firms did reduce overall hiring in non-AI vacancies while increasing AI driven hiring (ie workers with AI skills)

Economist on AI starting to impact labor markets (studies that are mentioned are attached)

“…A look beneath the aggregate numbers, though, reveals that change is indeed afoot.…work by Daron Acemoğlu and David Autor of the Massachusetts Institute of Technology, Jonathon Hazell of Princeton University and Pascual Restrepo of Boston University, which was presented at the recent meeting of the American Economic Association (aea). The authors use rich data provided by Burning Glass Technologies, a software company that maintains and analyses fine-grained job information gleaned from 40,000 firms. They identify tasks and jobs in the dataset that could be done by ai today (and are therefore vulnerable to displacement). Unsurprisingly, the researchers find that businesses that are well-suited to the adoption of ai are indeed hiring people with ai expertise. Since 2010 there has been substantial growth in the number of ai-related job vacancies advertised by firms with lots of ai-vulnerable jobs. At the same time, there has been a sharp decline in these firms’ demand for capabilities that compete with those of existing ai. An ai-induced change in the mix of jobs need not translate into less hiring overall. If new technologies largely assist current workers or boost productivity by enough to spark expansion, then more ai might well go hand-in-hand with more employment. This does not appear to be happening. Instead the authors find that firms with more ai-vulnerable jobs have done much less hiring on net; that was especially the case in 2014-18, when ai-related vacancies in the database surged. But the relationship between greater use of ai and reduced hiring that is present at the firm level does not show up in aggregate data, the authors note. Machines are not yet depressing labour demand across the economy as a whole.. .. A paper from 2019 by Timothy Bresnahan of Stanford University argues that the most valuable applications of ai have nothing to do with displacing humans. Rather, they are examples of “capital deepening”, or the accumulation of more and better capital per worker, in very specific contexts, such as the matching algorithms used by Amazon and Google to offer better product recommendations and ads to users. To the extent that ai leads to disruption, it is at a “system level”, says Mr Bresnahan—as Amazon’s sales displace those of other firms, say. New work by Ajay Agrawal, Joshua Gans and Avi Goldfarb of the University of Toronto suggests that this state of affairs may not persist for long, though. As the quality of ai predictions improves, they write, it becomes increasingly attractive for ai-using firms to restructure in more radical ways. At some level of accuracy, for example, Amazon’s ability to predict consumers’ desires could encourage the firm to adjust its business model—by pre-emptively shipping goods to consumers before they ever go searching at Amazon in the first place—in ways that are likely to change how many workers and of what sort the firm requires. In that event, the influence of ai on the economy could change dramatically…”

Economist Staff, "Research shows the robots are coming for jobs—but stealthily,"The Economist, January 16, 2021, https://www.economist.com/finance-and-economics/2021/01/16/new-research-shows-the-robots-are-coming-for-jobs-but-stealthily

JEP paper creates framework for thinking about AI’s impact on labor market, how much of a job is driven by “prediction,” “…Machine learning does not represent an increase in artificial general intelligence of the kind that could substitute machines for all aspects of humancognition but rather one particular aspect of intelligence: prediction....We define prediction in the statistical sense of using existing data to fill in missing information. As deep-learning pioneer Geoffrey Hinton (2016) said, “Take any old problem where you have to predict something and you have a lot of data, and deep learning is probably going to make it work better than the existing techniques.”...”
Implication 1, “…First, artificial intelligence may directly substitute capital for labor in prediction tasks. Some tasks, like demand forecasting, are already prediction tasks. Where humans currently perform these prediction tasks, they are increasingly replaced by artificial intelligence. At the same time, other tasks that were not historically viewed as prediction tasks are being transformed into prediction-oriented tasks as machine learning improves and the quality-adjusted cost of prediction decreases. Many parts of the workflow in human resources are being broken down into prediction tasks so that they can then be performed by artificial intelligence tools. For example, in the broad area of human resources, recruiting is the task of predicting—based on resumes, cover letters, LinkedIn profiles, and interview transcripts—which subset of applicants will perform best in the job. Promotion is the task of predicting which existing employees will perform best in a higher-level position. And retention is the task of predicting which star employees are most likely to leave and which of the available incentive options could most effectively be employed to encourage them to stay.…”
Implication 2“…Second, when automated prediction can increase the relative returns to capital versus labor in complementary decision tasks, it can lead to the complete automation of a complementary decision task. For example, human reaction times are slower than those for machines. The returns to a machine predicting a potential car accident a few seconds or even a fraction of a second before a human would predict the accident is higher when the response time of the machine is faster. Thus, automating the prediction task increases the returns to also automating certain decision tasks associated with vehicle control. Sometimes, the artificial intelligence is able to make better predictions than a human could because it has access to different data, such as feeds from cameras, RADAR, and LIDAR around a car.1 Once the prediction task is automated, it increases the returns to automating some of the complementary tasks, such as those associated with vehicle control…”

Ajay Agrawal, Joshua S. Gans, and Avi Goldfarb, "Artificial Intelligence: The Ambiguous Labor Market Impact of Automating Prediction," Journal of Economic Perspectives, Spring 2019, https://pubs.aeaweb.org/doi/pdfplus/10.1257/jep.33.2.31

Implication 4“…The fourth and final type of direct impact of artificial intelligence on labor happens when automated prediction sufficiently reduces uncertainty as to enable new decision tasks that did not exist before. The new tasks can be performed by capital or labor, depending on the relative costs of each (Agrawal, Gans, and Goldfarb 2019). Some tasks that are not economically viable when uncertainty is high become viable as prediction technology reduces the level of uncertainty. This relates to the reinstatement force in Acemoglu and Restrepo (in this issue) where a freeing up of labor as a result of automation increases the returns to technologies that use labor for new tasks. At this early stage in the development and use of machine learning, there are few tangible examples of new tasks that have already arisen because of recent advances in prediction technology…”

Implication 3, “…Third, automating the prediction task, in some cases, may have no impact on the productivity of capital performing a complementary task but may increase the productivity of labor.For example, ODS Medical developed a way of transforming brain surgery for cancer patients. Previously, a surgeon would remove a tumor and surrounding tissue based on previous imaging (say, an MRI scan). However, to be certain all cancerous tissue is removed, surgeons frequently end up removing more brain matter than necessary. The ODS Medical device, which resembles a connected pen-like camera, uses artificial intelligence to predict whether an area of brain tissue has cancer cells or not. Thus, while the operation is taking place, the surgeon can obtain an immediate recommendation as to whether a particular area should be removed. By predicting with more than 90 percent accuracy whether a cell is cancerous, the device enables the surgeon to reduce both type I errors (removing noncancerous tissue) and type II errors (leaving cancerous tissue). The effect is to augment the labor of brain surgeons. Put simply, given a prediction, human decisionmakers can in some cases make more nuanced and improved choices.…”

Stanford working paper argues that the most valuable application of AI won’t be substituting AI for labor but will drive capital deepening as workers grow more productive, “…Artificial Intelligence Technologies (AIT) are already in large scale use in two kinds of applications. We examine those applications, consumer/product matching engines and user interfaces, as well as the early diffusion of AITs and some categories where laboratory AIT and application are closely linked. In the leading applications, AITs are core production technologies with hundreds of billions of dollars in revenue. These production applications of AITs involve no substitution of machine for human work at the task level. The tasks undertaken by AIT in the leading applications were previously done by capital, not by human workers. Whatever your view of AITs, it should be updated to deal with this basic fact about their initial highly valuable applications. There is capital deepening, not factor substitution, at the task level….”
Core of argument, “…Looking not only at the leading applications but at the early diffusion,AIT-based systems appear to be on the same track as most other highly innovative ICT-based systems of the last 25 years. They have spread out over a range of activities, to be sure, but have their largest impact in a narrow range of industries and functions, notably in consumer-oriented mass market production, distribution and marketing. The invention of consumer-oriented services like the WWW and consumer oriented devices like smartphones and tablets represents the diffusion of ICT to consumption. AITs, however, so far are participate in a very different 21st century trend, capital deepening in a narrow range of consumer-oriented markets and functions with divergence across firms in the rate of technical progress.Examining the strengths and weaknesses of AITbased production illuminates the narrowness and depth of ICT application more generally…. The tendency of AIT to be scale using and to be complementary with existing capital assets at the firm level suggest that AIT use will continue these trends rather than change them. The capital deepening arises out of a cluster of characteristics that arise at the firm level: scale economies, complementarity of new AIT capital with existing capital assets, particularly with big data and with modularized production processes, and the ability to specify a quantitative goal for production in an environment with low stakes in the case of error. These important economic and technical elements of AIT-using systems invention are far from task-level substitution….”

Timothy Bresnahan, "Artificial Intelligence Technologies and Aggregate Growth Prospects," Stanford University, May 2019, https://web.stanford.edu/~tbres/AI_Technologies_in_use.pdf

Useful literature review of minimum wage research from Neumark, his core takeaway from the existing studies, "... First, there is a clear preponderance of negative estimates in the literature. In our data, 79.3% of the estimated employment elasticities are negative, 55.4% are negative and significant at the 10% level or better, and 47.9% are negative and significant at the 5% level or better…. In its totality, this body of evidence and its conclusions point strongly toward negative effects of minimum wages on employment of less-skilled workers, especially for the types of studies that would be expected to reveal these negative employment effects most clearly..."
"....The disagreement among studies of the employment effects of minimum wages in the United States is well known. What is less well known, and more puzzling, is the absence of agreement on what the research literature says - that is, how economists even summarize the body of evidence on the employment effects of minimum wages. Summaries range from “it is now well-established that higher minimum wages do not reduce employment,” to “the evidence is very mixed with effects centered on zero so there is no basis for a strong conclusion one way or the other,” to “most evidence points to adverse employment effects.” We explore the question of what conclusions can be drawn from the literature, focusing on the evidence using subnational minimum wage variation within the United States that has dominated the research landscape since the early 1990s. To accomplish this, we assembled the entire set of published studies in this literature and identified the core estimates that support the conclusions from each study, in most cases relying on responses from the researchers who wrote these papers. Our key conclusions are: (i) there is a clear preponderance of negative estimates in the literature; (ii) this evidence is stronger for teens and young adults as well as the less-educated; (iii) the evidence from studies of directly-affected workers points even more strongly to negative employment effects; and (iv) the evidence from studies of low-wage industries is less one-sided...."

Key findings, "...Summarizing the research literature this way, our key conclusions are as follows:There is a clear preponderance of negative estimates in the literature. In our data, 79.3% of the estimated employment elasticities are negative, 55.4% are negative and significant at the 10% level or better, and 47.9% are negative and significant at the 5% level or better. This evidence of negative employment effects is stronger for teens and young adults, and more so for the less-educated. The evidence from studies of directly-affected workers points even more strongly to negative employment effects. Theevidence from studies of low-wage industries is less one-sided, with 66.7% of the estimated employment elasticities negative, but only 33.3% negative and significant at the 10% level or better, and the same percent negative and significant at the 5% level or better.We suggest, however, that the evidence from low-wage industries is less informative about the effects of minimum wages on the employment of low-skill, low-wage workers. Overall, we conclude that the preferred estimates of authors of studies evaluating the employment effects of minimum wages in the United States, since the advent of the New Minimum Wage Research in 1992, paint a clear picture that is at odds with how this research is often summarized. In its totality, this body of evidence and its conclusions point strongly toward negative effects of minimum wages on employment of less-skilled workers, especially for the types of studies that would be expected to reveal these negative employment effects most clearly..."

David Neumark and Peter Shirley, "Myth or Measurement: What Does the New Minimum Wage Research Say about Minimum Wages and Job Loss in the United States?" National Bureau Of Economic Research, January 2021, https://www.nber.org/papers/w28388

"...We argue the revenue potential from increasing tax rates on capital gains may be substantially greater than previously understood. First, many prior studies focus primarily on short-run taxpayer responses, and so miss revenue from gains that are deferred when rates change. Second, the composition of capital gains has shifted in recent years, such that the share of gains that are highly elastic to the tax rate has likely declined. Third, focusing on capital gains tax collection may understate fiscal spillovers from decreasing the preferential tax treatment for capital gains. Fourth, additional base-broadening reforms, like eliminating stepped-up basis and making charitable giving a realization event, will decrease the elasticity of the tax base to rate changes. Overall, we do not think the prevailing assumption of many in the scorekeeping community—that raising rates to top ordinary income levels would raise little revenue—is warranted. A crude calculation illustrates that raising capital gains rates to ordinary income levels could raise $1 trillion more revenue over a decade than other estimates suggest.Given the magnitudes at stake, scorekeeping procedures employed in evaluating capital gains should be made more transparent and be the subject of external professional debate and review..."

Core of their calculation. They argue currently elasticity is wrong and that realizations deferred when cap gain tax rates rise are unlikely to be deferred indefinitely, "...We conclude with crude estimates of the wide range in revenue potential from raising capital gains rates to the top ordinary income levels under different elasticity assumptions. Applying the conventional elasticity e = −0.7 leads to the conclusion that this substantial rate increase would raise less than $150 billion in revenue. In stark contrast, using an elasticity that incorporates medium-term dynamics suggests that raising rates to the top ordinary income level could raise more than $1 trillion over a decade. Put differently, e = −0.7 implies that once capital gains rates reach around 30 percent, the additional increase to top ordinary rates loses, rather than raises, revenue. Our preferred estimates suggest the opposite. Pairing rate increases with the elimination of loopholes that erode the capital gains tax base—like stepped-up basis and the tax preference for charitable gifts of appreciated assets—produces larger revenue estimates. We provide a simple calculation that abstracts from many important details. Our point is not to offer an official score, but instead to illustrate the magnitude of potential revenue and how sensitive capital gains revenue estimates are to various assumptions. We suspect that making scorekeeping procedures more transparent and the subject of external professional debate and review will lead to an upward revision in the estimated revenue potential from substantially increasing capital gains tax rates....."

Summer's back of envelope arithmetic, "... Suppose that doubling capital gains rates from 20% to 40% causes realizations to occur half as often: instead of realizing gains every year, individuals realize gains every two years. If assets grow at 10% annually, then in the low-tax regime, $100 of assets yield realizations of $10 in year 1 and $10.80 in year 2 (after paying two dollars of tax in year 1). In the high-tax regime, $100 of assets yield realizations of $0 in year 1 and $21 in year 2. Despite the appearance in year 1 of a large elasticity of realizations in response to the tax increase, total revenues over both years increase from $4.16 in the low-tax regime to $8.40 in the high-tax regime. In this simple example without other behavioral responses, the short-run revenue score is zero and the medium-run revenue score is double the baseline.Clearly, the latter revenue score is more relevant for policy purposes...."
Natasha Sarin, Lawrence H. Summers, Owen Zidar and Eric Zwick, "Rethinking How We Score Capital Gains Tax Reform," National Bureau Of Economic Research, January 2021, https://www.nber.org/papers/w28362

Core findings“…We present four main findings. First, using data on Nobel Prize winners and Fields Medalists, we document that the migrants to the U.S. play a central role in the global knowledge network— representing 20-33% of these frontier knowledge producers. Second, using novel survey data and hand-curated life-histories of International Math Olympiad (IMO) medalists, we show that migrants to the U.S. are significantly more productive than migrants to other countries— even after accounting for one’s talent during their teenage years. Migrants to the U.S. are four to six times more productive than stayers, while migrants to the U.K. are more than twice as productive as stayers. Third, we document that financing costs are a key factor preventing foreign talent to migrate to the U.S. In particular, among developing country IMO participants, 66% dream of studying in the U.S. while only 25% manage to do so. Fourth, our findings suggest that certain ‘push’ incentives that reduce immigration barriers to the U.S.—by addressing financing constraints for top foreign talent—could increase the global scientific output of future cohorts of talent by 42% percent. We conclude by discussing policy options for the U.S. and the global scientific community, with a particular focus on the effectiveness of scholarships vs. other science policy actions such as offering green cards to foreign talent…”

The data, “…“….Our contribution is to quantify how U.S. immigration barriers could impact the advancement of science using hand-curated datasets of talented individuals - Nobel laureates, Fields medalists, and participants of the International Math Olympiads (IMO), a prominent worldwide math competition for talented high school students. Our dataset includes career histories of migration and lifetime scientific output of 2,200 IMO medalists from over one hundred countries. We combine these data with newly collected unique survey data of 610 recent IMO participants, which includes information on which universities they applied to, were admitted to and attended. The survey also asks a series of questions where respondents were asked to make choices between hypothetical university offers in different countries—where offers were either funded or unfunded…”

The second point is International Math Olympiad medalists (so a proxy for talent) are more productive if they immigrate to the United States“…We find that migrants are approximately three times more productive than stayers in terms of citation-weighted publications(see Table 1 column 1).8Distinguishing between migrants to the U.S., migrants to the U.K. (the second largest destination) and migrants to other countries, we find that migrants to the U.S. are four times more productive than stayers (in terms of citation-weighted publications) while migrants to the U.K. or to other countries are only around twice as productive as stayers(Table 1 Column 2)…”

IZA working paper documents two important points. The first is the importance of the US in moving the scientific frontier (using Nobel/Fields winner as a proxy for advancing the scientific frontier), 21% of Nobel prize winners and 33% of Fields medalists are migrants to the United States, “….We find that 21% of Nobel prize winners and 33% of Fields medalists are migrants to the U.S. (see Figure 2).A further 11% of Nobel prize winners, and 30% of Fields medalists are migrants to countries other than the U.S. These results highlight both the important role played by immigrants in the global production of knowledge, and the importance of the U.S. as a migration destination….”

Paper Ike mention Tyler linked to, was in our queue.