Category: Data Source
What predicts success in science?
How does a person’s childhood socioeconomic status (SES) influence their chances to participate and succeed in science? To investigate this question, we use machine-learning methods to link scientists in a comprehensive biographical dictionary, the American Men of Science (1921), with their childhood home in the US Census and with publications. First, we show that children from low-SES homes were already severely underrepresented in the early 1900s. Second, we find that SES influences peer recognition, even conditional on participation: Scientists from high-SES families have 38% higher odds of becoming stars, controlling for age, publications, and disciplines. Using live-in servants as an alternative measure for SES confirms the strong link between childhood SES and becoming a star. Applying text analysis to assign scientists to disciplines, we find that mathematics is the only discipline in which SES influences stardom through the number and the quality of a scientist’s publications. Using detailed data on job titles to distinguish academic from industry scientists, we find that industry scientists have lower odds of being stars. Controlling for industry employment further strengthens the link between childhood SES and stardom. Elite undergraduate degrees explain more of the correlation between SES and stardom than any other control. At the same time, controls for birth order, family size, foreign-born parents, maternal education, patents, and connections with existing stars leave estimates unchanged, highlighting the importance of SES.
That is from a new NBER working paper by Anna Airoldi and Petra Moser.
The Declining Relative Quality of the Child Care Workforce
Although it is widely acknowledged that high-skilled teachers are integral to service quality and young children’s well-being in child care settings, little is known about the qualifications and skills of the child care workforce. This paper combines data from multiple sources to provide a comprehensive assessment of the quality of individuals employed in the child care sector. I find that today’s workforce is relatively low-skilled: child care workers have less schooling than those in other occupations, they score substantially lower on tests of cognitive ability, and they are among the lowest-paid individuals in the economy. I also show that the relative quality of the child care workforce is declining, in part because higher-skilled individuals increasingly find the child care sector less attractive than other occupations. Furthermore, I provide evidence that at least three other factors may be associated with the decline in worker quality. First, the recent proliferation of community college programs offering child care-related certificates and degrees may divert students away from attending four-year schools. Second, those majoring in child care-related fields are negatively selected for their cognitive skills, thereby decreasing the quality of the child care labor pool. Third, I show that the increased availability of outside employment options for high-skilled women had a detrimental effect on the quality of the child care workforce.
That is from a new paper by Chris M. Herbst. Via the excellent Kevin Lewis.
Fluoride revisionism?
I am usually skeptical of such efforts, but Journal of Health Economics is quite a serious outlet:
Community water fluoridation has been named one of the 10 greatest public health achievements of the 20th century for its role in improving dental health. Fluoride has large negative effects at high doses, clear benefits at low levels, and an unclear optimal dosage level. I leverage county-level variation in the timing of fluoride adoption, combined with restricted U.S. Census data that link over 29 million individuals to their county of birth, to estimate the causal effects of childhood fluoride exposure. Children exposed to community water fluoridation from age zero to five are worse off as adults on indices of economic self-sufficiency (−1.9% of a SD) and physical ability and health (−1.2% of a SD). They are also significantly less likely to graduate high school (−1.5 percentage points) or serve in the military (−1.0 percentage points). These findings challenge existing conclusions about safe levels of fluoride exposure.
That new article is by Adam Roberts. Via the excellent Kevin Lewis.
The Political Transformation of Corporate America, 2001-2022
This article reconciles conflicting views about the political landscape of corporate America with new data on the revealed political preferences of 97,469 corporate directors and executives at 9,005 different U.S. companies. I find that average ideology for these individuals has shifted meaningfully to the left over time, changing from modestly conservative in 2001 to roughly centrist by 2022. This finding supports a middle-ground position between conventional wisdom casting “big business” as a conservative stronghold and revisionist views holding the opposite. Counterfactual simulations and a difference-in-differences design suggest multifaceted causes for these changes, and hand-collected data on corporate stances on LGBTQ-related legislation coupled with an instrumental variables design indicate that individual ideology has large effects on firm-level political activity. Overall, this transformation has profound implications for American politics, as the individuals comprising one of the most powerful interest groups—corporate elites—appear to be fracturing ideologically and to some degree even switching sides.
That is from a new paper by Reilly Steel.
Letters of recommendation
We analyze 6,400 letters of recommendation for more than 2,200 economics and finance Ph.D. graduates from 2018 to 2021. Letter text varies significantly by field of interest, with significantly less positive and shorter letters for Macroeconomics and Finance candidates. Letters for female and Black or Hispanic job candidates are weaker in some dimensions, while letters for Asian candidates are notably less positive overall. We introduce a new measure of letter quality capturing candidates that are recommended to “top” departments. Female, Asian, and Black or Hispanic candidates are all less likely to be recommended to top academic departments, even after controlling for other letter characteristics. Finally, we examine early career outcomes and find that letter characteristics, especially a “top” recommendation have meaningful effects on initial job placements and journal publications.
That is from a new paper by Beverly Hirtle and Anna Kovner. Via the excellent Kevin Lewis.
U.S.A. facts of the day
The U.S. Senate Committee on Commerce, Science, and Transportation minority staff (Committee), which oversees federal science agencies including NSF, analyzed 32,198 Prime Award grants NSF awarded to 2,443 different entities with project start dates between January 2021 and April 2024.
Committee analysis found 3,483 grants, more than ten percent of all NSF grants and totaling over $2.05 billion in federal dollars, went to questionable projects that promoted diversity, equity, and inclusion (DEI) tenets or pushed onto science neo-Marxist perspectives about enduring class struggle. The Committee grouped these grants into five categories: Status, Social Justice, Gender, Race, and Environmental Justice. For the purposes for this report, “DEI funding,” a “DEI grant,” or “DEI research” refers to taxpayer dollars NSF provided to a research or engagement program that fell into one of these five groups.
Here is the full report. Note that by early 2024, that figure had risen to 27 percent.
Who benefits from working with AI?
We use a controlled experiment to show that ability and belief calibration jointly determine the benefits of working with Artificial Intelligence (AI). AI improves performance more for people with low baseline ability. However, holding ability constant, AI assistance is more valuable for people who are calibrated, meaning they have accurate beliefs about their own ability. People who know they have low ability gain the most from working with AI. In a counterfactual analysis, we show that eliminating miscalibration would cause AI to reduce performance inequality nearly twice as much as it already does.
That is from a new NBER working paper by
Further evidence for the babysitting theory of education
Bryan Caplan will feel vindicated:
This paper asks whether universal pre-kindergarten (UPK) raises parents’ earnings and how much these earnings effects matter for evaluating the economic returns to UPK programs. Using a randomized lottery design, we estimate the effects of enrolling in a full-day UPK program in New Haven, Connecticut on parents’ labor market outcomes as well as educational expenditures and children’s academic performance. During children’s pre-kindergarten years, UPK enrollment increases weekly childcare coverage by 11 hours. Enrollment has limited impacts on children’s academic outcomes between kindergarten and 8th grade, likely due to a combination of rapid effect fadeout and substitution away from other programs of similar quality but with shorter days. In contrast, parents work more hours, and their earnings increase by 21.7%. Parents’ earnings gains persist for at least six years after the end of pre-kindergarten. Excluding impacts on children, each dollar of net government expenditure yields $5.51 in after-tax benefits for families, almost entirely from parents’ earnings gains. This return is large compared to other labor market policies. Conversely, excluding earnings gains for parents, each dollar of net government expenditure yields only $0.46 to $1.32 in benefits, lower than many other education and children’s health interventions. We conclude that the economic returns to investing in UPK are high, largely because of full-day UPK’s effectiveness as an active labor market policy.
Here is more from Note by the way that these externalities end up internalized in higher wages for the parents, so at least in this data set there is no obvious case for public provision of a subsidized alternative.
The decline in retail sales jobs
It is steeper than I had thought:
The final labor market trend we uncovered was a very rapid decline in retail sales jobs, show in the figure below. Retail sales hovered at around 7.5 percent of employment from 2003 to 2013 but has since fallen to only 5.7 percent of employment, a decline about 25 percent in just a decade. Put another way – the U.S. economy added 19 million total jobs between 2013 and 2023 but lost 850 thousand retail sales jobs. The decline started well before the pandemic.
And STEM jobs truly are on the rise, even though that is what they may be telling you in school:
The figure also shows rapid employment growth in business and management jobs. The fastest growing occupations in that category are science and engineering managers, management analysts, and other business operations specialists. This is especially striking because STEM employment declined slightly between 2000 and 2012.
With a good picture at the link. That is all from David Deming, with further interesting material throughout.
How much does hard work matter?
That is the topic of my latest Bloomberg column, here is one excerpt:
Economists from Princeton, Vanderbilt and the Federal Reserve Bank of St. Louis have estimated just how much hard work contributes to inequality in lifetime earnings. While the answer depends on context, they arrived at an average for the US workforce: About 20% of the variance in lifetime earnings can be explained by differences in hours worked…
The decision to work harder operates on at least two levels. First, you put in more total time, which leads to higher lifetime earnings. Second, you invest more in your human capital, which makes you more productive. Between one-third and one-half of the higher income for the harder workers stems from this human capital channel. One lesson is that if you are going to work hard, you should do so relatively early in your life, so as to reap the human capital benefits for future years.
Another crucial point is that those who work harder do so because they want to. There can be different kinds of heterogeneity in ability, including in learning capability or initial human capital. But in the researchers’ model, 90% of the variation in earnings due to hard work comes from a simple desire to work harder.
And this:
The study focuses on the US, but it has implications for Europe as well. In France, for instance, work is limited to 48 hours per week, with a standard week of 35 hours. That reduces average earnings and inequality in earnings, since it is harder for the top achievers to keep making more money. This research finds that the losers from this regulation are found at all parts of the wage distribution, not only at the top.
How different are Trump judges?
Donald J. Trump’s presidency broke the mold in many ways, including how to think about judicial appointments. Unlike other recent presidents, Trump was open about how “his” judges could be depended on to rule in particular ways on key issues important to voters he was courting (e.g., on issues such as guns, religion, and abortion). Other factors such as age and personal loyalty to Trump seemed important criteria. With selection criteria such as these, one might expect that Trump would select from a smaller pool of candidates than other presidents. Given the smaller pool and deviation from traditional norms of picking “good” judges, we were curious about how the Trump judges performed on a basic set of measures of judging. One prediction is that Trumpian constraints on judicial selection produced a different set of judges. Specifically, one that would underperform compared to sets of judges appointed by other presidents. Using data on active federal appeals court judges from January 1, 2020 to June 30, 2023, we examine data on judges across three different measures: opinion production, influence (measured by citations), and independence or what we refer to as “maverick” behavior. Contrary to the prediction of underperformance, Trump judges outperform other judges, with the very top rankings of judges predominantly filled by Trump judges.
That new paper is by Stephen J. Choi and Mitu Gulati, who seem to be academic “normies” (NYU and UVA, respectively), not MAGAland crazies.
Via the excellent Kevin Lewis.
Model this
Doctors were given cases to diagnose, with half getting GPT-4 access to help. The control group got 73% right & the GPT-4 group 77%. No big difference.
But GPT-4 alone got 92%. The doctors didn’t want to listen to the AI.
Here is more from Ethan Mollick. And now the tweet is reposted with (minor) clarifications:
A preview of the coming problem of working with AI when it starts to match or exceed human capability: Doctors were given cases to diagnose, with half getting GPT-4 access to help. The control group got 73% score in diagnostic accuracy (a measure of diagnostic reasoning) & the GPT-4 group 77%. No big difference. But GPT-4 alone got 88%. The doctors didn’t change their opinions when working with AI.
There are not 13,099 Illegal Immigrant Murderers Roaming Free on American Streets
Migrants incarcerated for homicide are considered “non-detained” by ICE when they are in state or federal prisons. When ICE uses the term “non-detained,” they mean not currently detained by ICE. In other words, the migrant murderers included in the letter are overwhelmingly in prison serving their sentences. After they serve their sentences, the government transfers them onto ICE’s docket for removal from the United States.
And that is only part of the mistake in the numbers you may have heard. Here is more from Alex Nowrasteh:
The third untrue claim is that these 13,099 migrants convicted of homicide committed their crimes recently. Those migrant criminal convictions go back over 40 years or more. Confusion over the period covered by a dataset afflicts the interpretation of other criminal datasets too. If there really were 13,099 migrants convicted for domestic homicides in 2023, then they would have accounted for about 99 percent of all homicide convictions in the U.S. last year despite being about 4 percent of the population. That is obviously not the case because no group of people is criminally overrepresented by a factor of 25 above their share of the population. Even when the 13,099 homicide convictions of migrants are spread out over the entire Biden administration, migrants would have accounted for about one-third of all homicide convictions from 2021 through 2023. That’s obviously not true. The problem comes from erroneously increasing the numerator (the number of homicide convictions) for a single year and decreasing the denominator (the total number of homicide convictions in just one year) rather than spreading out the convictions and the total number of all murders over a 40-plus year period.
As a side observation:
Here is the whole essay. Tweetstorm here. Via Naveen.
Predicting future promotions from police cadets’ facial traits
From the results:
Facial traits are the primary driver of subject perceptions of leadership ability, and those perceptions successfully predict promotional success later in the cadets’ careers. When selecting for leadership potential based on police cadet photographs, respondents predict correct promotional choices at levels well above chance as measured by an AUC score of .70. Further, respondents’ evaluations successfully discriminate both between no promotion and lieutenant promotion, and sergeant versus lieutenant promotions.
That is from a new paper by Ian T. Adams, Scott M. Mourtgos, Christopher A. Simon, and Nicholas P. Lovrich. Via the excellent Kevin Lewis.
Reducing Pollution in India with a Cap and Trade Market
India has some of the worst air pollution in the world. India regulates pollution but it uses a command and control approach with criminal penalties, a system in tension with enforcement given low-state capacity. The result has been widespread corruption, inefficiency, and poor enforcement of pollution controls. In a very important paper, Greenstone, Pande, Ryan and Sudarshan report on an experiment with a market for particulate matter in Surat, India. In fact, this is the first particulate-matter market anywhere in the world.
The experiment created two sets of firms, the treatment set were required to install continuous emission monitoring systems (CEMS) which measured the output of particulate matter. The control set of firms remained under the command and control system which required the installation of various pollution control devices and spot checks. Firms were randomly assigned to treatment or control. Pollution at treatment firms was capped and permits were issued for 80% of the cap so firms could pollute at 80% of the cap for free. Permits for the remaining 20% of the cap were sold at auction and trading was allowed. Treatment plants which polluted more than their permits allowed paid substantial fines, about double the cost they would have paid to buy the necessary permits.
The one and half year experiment revealed a great deal of importance. First, the CEMS systems and the switch to financial penalties reduced the cost of enforcement so that essentially all firms quickly came into compliance. Second, trading was vigorous, which indicated that firms have heterogeneous and changing costs. Moreover, by allowing for a more information rich market the costs of achieving a given level of pollution fell. Pollution costs were 11% lower in treatment firms compared to control firms at the same level of pollution. The value of trade in lowering abatement costs illustrates Hayek’s idea that one of the virtues of markets is that they make use of information of particular circumstances of time and place. In fact, since the costs of achieving a given level of pollution were low, the authorities decreased the cap so that the treatment firms reduced their pollution levels significantly relative to the control firms.
The CEMS systems were a fixed cost but because abatement costs decreased, the overall expense was reasonable. The need for monitoring systems and procedures highlights Coase’s insight that property rights in externalities must be designed and enforced, the visible and invisible hand work best together.
Using estimates on a statistical life-year in India of $9,500 (about 1/10th to 1/30 the level typically used in the US) the authors find that the benefits of substantial pollution reduction exceed the costs by a factor of 25:1 or higher.
I have emphasized (and video here) that there are significant productivity gains to reducing air pollution which would make these benefit to cost ratios even higher. Less pollution can mean more health and more wealth.
The authors are especially to be congratulated because this paper began in 2010 with discussions with the Gujurat Pollution Control Board. It took over a decade to implement the experiment with the authors helping to design not just the market but also the technical standards for CEMS monitoring. Amazing. The success of the system is already leading to expansion across India. Bravo!
Hat tip: Paul Novosad.