Category: Economics

NPR plays overrated vs. underrated with Banerjee and Duflo

Cardiff Garcia led the charge, along with Stacey Vanek Smith, here is one excerpt:

GARCIA: Overrated or underrated – being married to an economist.

BANERJEE: Underrated.


BANERJEE: Excellent.

DUFLO: Underrated, of course.

SMITH: (Laughter). What about overrated/underrated – dating an economist?

BANERJEE: I mean, if you’re married to one (laughter) overrated.

SMITH: I mean, did you guys, like, split checks? Or, like, how did that work? Did you run into any economic quandaries early on?

BANERJEE: We were pretty flexible. We’re not really money people, so we didn’t – never spent a minute thinking about it.

Here is the full program.  For the pointer I thank Michelle Dawson.

The future of distance work arbitrage! (those new Rwandan service sector jobs)

In the central African country of Rwanda, single mothers employed at a Japanese eatery have found a new source of income after their jobs took a hit from the novel coronavirus pandemic: babysitting Japanese kids online.

Despite the seven-hour time difference, the cross-cultural service sees women play and sing with children 12,000 kilometers away in Japan via the videoconferencing app Zoom. The mothers will sometimes stream themselves shopping, chopping vegetables and cooking, to the delight of the kids’ parents as well.

The service is provided twice a day for an hour each in a mix of local languages, English, and Japanese.

“There’s a groove that you can’t experience in neighborhood eurythmic classes,” said Toyochika Kamekawa, 36, from Takahama in Fukui Prefecture. His 2-year-old son regularly takes part in the online sessions and sings songs he has been taught, accompanying himself on his toy drum.

The initiative was started up by Rwanda resident Mio Yamada, 38, who hires single mothers to work at her Japanese restaurant in the capital city of Kigali, and her acquaintance Yushi Nakashima, 30.

Yamada, who studied Swahili at university and now has three sons, moved to Rwanda with her husband in 2016 and opened her restaurant the following year.


“I think my son will come to some realization (about the economic disparity between countries) when he’s older and compares his allowance with the sitters’ wages.”


Some of the songs performed by the sitters touch on these darker themes. In one that foreshadows the conflict, the lyrics implore a child to stop crying with the words that when the war begins, they will be given milk from a cow that isn’t sad.

Here is the full story, via Air Genius Gary Leff.

Caleb Watney emails me

Hey Tyler,

Wanted to let you know that Alec Stapp and I are launching a new blog today called Agglomerations where we’ll be covering a wide range of tech + innovation policy issues.

My first piece today ties to disentangle technology, innovation, and industrial policy, as I feel the conversations there have been quite muddled as of late. I hope you enjoy it.

On vaccine timing, from the comments

Current US excess all-cause mortality is running around 10K above trend line on a good week. For these purposes we need neither know nor care if this is do to Covid directly, fear of catching Covid, or the effects of lockdown policies (e.g. increased rates of suicide from business failure).

Suppose we vaccinated the entire population. We would have to have a fatality rate of 1/32,000 to match a single week of the excess deaths the CDC is reporting every week.

Phase III, according to Moderna, is going to last around 4 months. If we skipped phase III and just jumped to phase IV, we could look to avert around 170,000 deaths with perfect effectiveness. Let’s say that between it being a lousy vaccine and people giving up on social distancing early we only save half of those lives. That gives us around 85,000 lives saved.

This means that we are looking at a 1/4,000 as our rough fatality rate for vaccines to be safer than waiting.

To date, I know of no vaccine that passed phase II clinical trials that has ever had this sort of fatality rate. Further, the vaccine adverse event rate is typically an order of magnitude or more higher than the fatality rate so absent truly astonishingly high rates of clinically significant adverse events it is highly unlikely to hit something greater than 1/4000 but not see some evidence of it with just a few hundred patients under your belt in the early phase trials.

My pre-test probability that any vaccine fall into a mortality rate somewhere between 1/100 and 1/4000 absolute risk increase is pretty low. So low, that all the papers I have read name them specifically rather than tabulating (e.g. Merck V710) and even those tend not to have dramatic increases in mortality (e.g. V710 had increased rates of Staph infections, but no increase in mortality).

Maybe something will be different this time and that will happen. Maybe we will be stuck with an inferior vaccine (though my guess is if we declared all of the top candidates ready for phase IV we could get a wide variety of use). But from where my point of view as I understand things from the coroner, we would need some massively high pre-test priors to make a dent in the weekly excess mortality rates.

For places that have been able to limit the virus’s entry (e.g. the island nations), this is all moot. For places that had massive initial spikes (e.g. Italy), this may also be moot. But for the US which flatten the initial curve and has maintained a steady excess death rate, early adoption seems to have very good Bayesian odds

You could of course make a QALY argument about Covid fatalities against vaccine fatalities, but again you are trying to hit a pretty narrow window: dangerous enough to beat down 10,000 excess deaths a weak, not so dangerous that it would have obviously failed early studies.

Challenge studies may have their place (e.g. ruling out antibody dependent enhancement), but we need to be sure that the immune response works in real world conditions. And at this point I still have terrible data for what actually happens during the course of the infection.

Inaction has one of the highest price tags we have ever seen in modern medicine. I just do not see any evidence that my pre-test priors for a vaccine being that deadly should be remotely high enough to make the math on delay work out (particularly if we can avoid the most susceptible patient populations for VAEs). Certainly my priors for the ability of social action to stop infections is far worse (e.g. HIV, Lyme, Scabies) so I am increasingly convinced that long term it is either a vaccine or the hard road to herd immunity.

CatintheHat addds:

I agree with this rough analysis and the risk is even less because we, probably only have to vaccinate 25 % of the population to stop the epidemic completely which could be done relatively quickly.

For the opposite case, don’t forget this earlier in August post of Alex’s.

I would very gladly run further estimates of this kind.  To be clear, I am not endorsing any particular conclusion, rather I think several hundred smart people should be working on this full time.  My personal suspicion is that the decisive factor will be “the gains from possibly ending a global depression a few months earlier” vs. “the risk that with a lower quality vaccine we don’t end said depression more effectively than we otherwise might have.”  And I hardly see anyone considering that trade-off at all.

There is also a closely related but conceptually separate question of how many vaccines to approve on an earlier basis, and also for how many should we be in a position to do so.

The impact of economic regulation on growth: survey and synthesis

This study provides a survey of research that uses cross-country comparisons to examine how economic regulation affects growth. Studies in the peer-reviewed literature tend to rely on either World Bank or Organisation for Economic Co-operation and Development measures of regulation. Those studies seem to reflect a consensus that entry regulation and anticompetitive product and labor market regulations are generally harmful to growth. The results from this cross-country research, taken in conjunction with economic theory as well as other country-specific studies of economic regulation, support the hypothesis that economic regulation tends to reduce welfare in competitive markets. Given the continued use of certain types of economic regulation, the findings may offer important lessons for policymakers.

That is a new Mercatus working paper by James Broughel and Robert Hahn.

Uh-oh Mumbai

And here is the research:

The spatial layout of cities is an important feature of urban form, highlighted by urban planners but overlooked by economists. This paper investigates the causal economic implications of city shape in India. I measure cities’ geometric properties over time using satellite imagery and historical maps. I develop an instrument for urban shape based on geographic obstacles encountered by expanding cities. Compact city shape is associated with faster population growth and households display positive willingness to pay for more compact layouts. Transit accessibility is an important channel. Land use regulations can contribute to deteriorating city shape.

Here is the full paper by Mariaflavia Harari.  “Transit accessibility” — what a funny phrase to apply to Mumbai traffic!  Try some Marathi slang instead.  And in case you are not familiar with Mumbai, some of the lower parts have some of the most valuable land.

The quality-adjusted rate of price inflation is much higher than measured

Higher ed on Zoom, fear of going to the hospital, and long lines at the DMV bring us to one part of my latest Bloomberg column:

Education, health care and government are pretty big parts of our economy. If you add on the lower quality of restaurant visits, reduced sports performances (your ESPN cable package is worth less), and an inability to take preferred vacations and trips, you have many more negative quality adjustments that don’t show up in measured rates of inflation.


The Bureau of Labor Statistics, the Bureau of Economic Analysis, the Fed and other institutions have declined to make formal adjustments for these changes in the real standard of living. That is the politically practical way to proceed, if not the technically correct decision.

Inflation measures work best when the consumption bundle is roughly stable over short periods of time, and that just hasn’t been the case this year. Because so many government payments depends on rates of indexation, debating “the true degree of inflation” this year would become an unending political football, with all the major institutions involved, including the Fed, losing credibility. (Just exactly how much worse is that Zoom lecture?) Besides, implicit price inflation from restricted opportunities probably should, for purposes of policy and benefits indexation, be treated differently than price inflation resulting from more expensive food and rent.

One implication is that, at least for the time being, price inflation rules just aren’t that meaningful any more:

…price rules and other forms of inflation rules don’t really work in times of pandemic. The very measurement of price inflation becomes arbitrary, and dependent on inertial measurement conventions from normal times, so the numbers don’t have enough actual economic meaning to guide policy.

There is more at the link.

My Conversation with Jason Furman

Yes, the Jason Furman, here is the audio and transcript, please note this was recorded in January.  Here is part of the summary:

Jason joined Tyler for a wide-ranging conversation on how monopolies affect investment patterns, his top three recommendations to improve American productivity, why he’s skeptical of place-based development policies, what some pro-immigration arguments get wrong, why he’s more concerned about companies like Facebook and Google than he is Walmart and Amazon, the merits of a human rights approach to privacy, whether the EU treats tech companies fairly, having Matt Damon as a college roommate, the future of fintech, his highest objective when teaching economics, what he learned from coauthoring a paper with someone who disagrees with him, why he’s a prolific Goodreads reviewer, and more.

And here is one excerpt:

COWEN: The US is losing some of its manufacturing capacity, and certainly a lot of its manufacturing workforce. Are there external benefits to keeping those activities more in the US? Significant benefits?

FURMAN: I don’t think that manufacturing itself should be an important objective of US policy. It’s one type of job. It’s been a good type of job, but there’s other good types of jobs as well. I wouldn’t focus on where physical things are being made as opposed to where services are being made. In fact, if anything, I think the error in policy is probably a little bit too much emphasis on manufacturing and a little bit less on services.

COWEN: What do you think of the national security argument? That, say, when building a ship, we might be dependent on South Korean components. If there were a war in Asia, those might be, for some reason, unreliable. We depend on China for rare earths. We depend on Taiwan, to some extent, for high-quality chips, even though we make our own. Is the supply chain extended too long, and it was a kind of economic fantasy, and it doesn’t make national security sense?

FURMAN: I don’t consider myself an expert in any of those national security questions, so I would be open to thinking about the national security concerns associated with the supply chain. I have an awful lot in specific cases — both when I was in government and just in the world more generally — heard people make national security arguments that I found tendentious and pretty unpersuasive.

There may be some that are persuasive and that are true. There’s an awful lot that aren’t. Our administration, towards the end, worried a bit about semiconductors. When I’ve looked at that, there’s enough of a diversified world supply, enough of an ability to scale up if necessary in the United States, that I don’t think on semiconductors — there, it was protectionism under the guise of national security.

So I think we should accept the possibility of national security, take it seriously, but be really, really wary that a lot of protectionist arguments use that trappings.

Economics throughout, with a touch of Dickens.  Recommended.

How bad was the Spanish Flu pandemic of 1918?

Yes it was a terrible tragedy, but many locales had much worse events fairly recently:

Between 1917 and 1918 New York City’s crude mortality rate increased by 3.173 deaths per 1000 persons.  While tragic, the hollow circles in Figure 1 depict 12 other years where the year-over-year increase in mortality exceeded the magnitude of the 1917 to 1918 change.  During the cholera epidemics of 1832, 1834, 1849, and 1854 the year-over-year increase in mortality was 3 to 5 times larger in magnitude than what occurred in 1918. As another comparison, the mortality rate in New York City was higher in nearly every year between 1800 and 1905 than the mortality rate in 1918.

The same is true for many other American cities, but here is a picture for NYC:

And this:

During the first half of the 20th century, Black Americans in urban areas died from infectious disease at a rate that was greater than what urban whites experienced during the 1918 flu pandemic every single year.

On a different but related topic:

…the evidence suggests that the 1918 pandemic was not a major determinant of U.S. stock market volatility.

That is all from the new and very interesting NBER paper by Brian Beach, Karen Clay, and Martin H. Saavedra, “The 1918 Influenza Pandemic and its Lessons for Covid-19.”

Mental Health and the Reluctance to Use Talk Therapy

I know very little about this area, but found these results of interest and worthy of further investigation:

Mounting evidence across disciplines shows that psychotherapy is more curative than antidepressants for mild-to-moderate depression and anxiety. Yet, few patients use it. This paper develops and estimates a structural model of dynamic decision-making to analyze mental health treatment choices in the context of depression and anxiety. The model incorporates myriad costs suggested in previous work as critical impediments to psychotherapy use. We also integrate links between mental health and labor outcomes to more fully capture the benefits of mental health improvements and the costs of psychotherapy. Finally, the model addresses measurement error in widely-used mental health variables. Using the estimated model, we find that mental health improvements are valuable, both directly through increased utility and indirectly through earnings. We also show that even though psychotherapy improves mental health, counterfactual policy changes, e.g., lowering the price or removing other costs, do very little to increase uptake. We highlight two conclusions. As patient reluctance to use psychotherapy is nearly impervious to a host of a priori reasonable policies, we need to look elsewhere to understand it (e.g., biases in beliefs about treatment effects, stigma, or other factors that are as yet unknown). More broadly, large benefits of psychotherapy estimated in randomized trials tell only half the story. If patients do not use the treatment outside of an experimental setting—and we fail to understand why or how to get them to—estimated treatment effects cannot be leveraged to improve population mental health or social welfare.

That is from a new NBER working paper by Christopher J. Cronin, Matthew P. Forsstrom, and Nicholas W. Papageorge.

Glen Weyl update and interview

Here is one excerpt:

I’ve moved on from being a researcher. I’m an advisor to Microsoft’s senior leaders about geopolitics and macroeconomics. So, my whole outlook has changed quite a bit as a result of that.


In Taiwan, we’ve come to work extremely closely with Audrey Tang, their digital minister who’s just a remarkable person and, honestly, a much more interesting subject than me. She has been using quadratic voting for administering national hackathons—where people get together and try to create technological solutions to social problems.

Audrey has used quadratic voting to score those competitions and she’s also used another idea that we’re very into, called ‘data coalitions’ or ‘data cooperatives’—they’re sort of data labour unions—to organize those services. Taiwan’s response to Covid was, to a large extent, driven by these civic technology developments and they were the most successful country in the world. They had the lowest infection and death rate and the smallest impact on their economy. A lot of that was related to their harnessing of these civic technology approaches.

Here is the Five Books link, interesting throughout.

Should hiring schools coordinate on delaying their interviews?

The AEA emails me this (web version here):

The AEA suggests that employers wait to extend interview invitations until Monday, December 7, 2020 or later.

Rationale: the AEA will deliver signals from job candidates to employers on December 2. We suggest that employers wait and review those signals and incorporate them into their decision-making, before extending interview invitations.

…The AEA suggests that employers conduct initial interviews starting on Wednesday, January 6, 2021, and that all interviews take place virtually; i.e. either by phone or online (e.g. by Zoom). We also ask that all employers indicate on EconTrack when they have extended interview invitations (

Rationale: In the past, interviews were conducted at the AEA/ASSA meetings. This promoted thickness of the market, because most candidates and employers were present at the in-person meetings, but had the disadvantage of precluding both job candidates and interviewers from fully participating in AEA/ASSA sessions. Since the 2021 AEA/ASSA meetings (which will take place Jan 3-5, 2021) will be entirely virtual, we suggest that interviews NOT take place during the AEA/ASSA meetings to allow job candidates and interviewers to participate in the conference.

Perhaps not surprisingly, they don’t offer much economic analysis of this recommendation.  I have a few remarks, none of which are beyond the analytical acumen of the AEA itself:

1. This proposal could well be a tax on the more conscientious departments, which will abide by the stricture while the more rogue departments jump the gun, giving them a relative advantage in finding job candidates.

2. It is common practice for the very top departments to make phone calls to advisors early, well before Christmas, and in essence tie up their future hires before the rest of the market clears (even if the ink on the contract is not dry until later on).  Whatever you might think of this practice, have any of those departments vowed to stop doing this?  If not, is the new recommendation simply an exhortation that other departments ought not to copy them, thus giving them exclusive use of this practice?  And did the AEA — which essentially is run by people from those top schools — ever complain about this practice?

3. In the more liquid market, as this proposal is designed to create, the better job candidates are likely to end up going to the more highly rated schools.  That is the opposite of how the NBA draft works — this year the Minnesota Timberwolves (a very bad team) pick first.  So maybe the more liquid market is best for the most highly rated schools — is that obviously a good thing?

4. Many job candidates don’t get any early offers at all, and this is likely to be all the more true with Covid-19 and tight state budgets.  Aren’t they better off if the market clears sooner rather than later?  Then they can either move on to other jobs searches, take jobs with community colleges, look for postdocs, or whatever.  Why postpone those adjustments?  Is their welfare being counted in this analysis?  Aren’t some of them the very neediest and also most stressed people in the economics job market?

5. Let’s say instead the market is done sequentially, where first you “auction off” the candidates in highest demand, ensuring that say a department rated #17 does not tie up an offer (fruitlessly, at that) to one of the very top candidates.  Won’t that #17 school then bid harder for the candidates one tier lower, thus making that part of the market more liquid?  I know it doesn’t have to work out that way, but surely that is one plausible scenario?

6. In finance, there are some results that you get less “racing” behavior with batched rather than continuous trading auctions. Again, that doesn’t have to be true, but surely it is no accident that many high-frequency traders oppose the idea of periodic rather than continuous securities auctions?  What exactly are the relevant conditions here?

7. Would many economists recommend that say the top tech firms not make any offers before a certain date, so as to keep that labor market “more liquid”?  What exactly is the difference here?

8. Might it be possible that a permanent shift to non-coordinated interview dates, and less temporally coordinated Zoom interviews and fly-outs, would permanently lower the status and import of said AEA?

I do not wish to pretend those are the only relevant factors.  But here is a simple question: does anyone connected with the AEA have the stones to actually write a cogent economic or game-theoretic analysis of this proposal?  Or does the AEA not do economics any more?

Emergent Ventures winners, 10th cohort

Sebastian Garren, to found John Paul II Preparatory School’s South Campus in St. Louis, a hybrid on-line and in-person educational alternative for K-12, also stressing Western history and the classics.

John Durant, for career development and writing, and explorations into notions of angels.

Mishka Orakzai of Peshawar, Pakistan, to support her thiscodeworks project intent to make snippets of code more available.

Krishaan Khubchand, 20 years old, studying law at Birkbeck, to study mega-projects and capital allocation, he is also a Progress Studies fellow.

Vignan Velivela.  He started as a robotics engineer at Cruise Automation, is a member of the Explorers Club (wikiBBC) for his work on the lightest planetary rover at Carnegie Mellon, worked on a peer-to-peer lending startup in India that was acqui-hired by PayTm, went to college (BITS Pilani) in India studying EE and Economics, and now is co-founder of AtoB.

Wasteland Ventures (no web page), to support their efforts in talent search and development.

And two Emergent Ventures anti-Covid prizes have been awarded to:

Witold Wiecek, Bayesian statistician and consultant, for his work on the Bayesian modeling of the COVID-19 epidemic, and the design of an optimal vaccine portfolio, in cooperation with the Accelerating Health Technologies team.

Arthur W. Baker (no web page, and not this guy) for his efforts on incentive design for vaccines, in cooperation with the Accelerating Health Technologies team.

Here are previous winners of Emergent Ventures grants and prizes.

A Real World AI Economist?

From Salesforce:

The moonshot goal of this project is to build a reinforcement learning framework that will recommend economic policies that drive social outcomes in the real world, such as improving sustainability, productivity, and equality. To achieve this, we’ll need to advance AI, challenge conventional economic thinking, and create AI that can ground and guide policy making. While none of these tasks are easy, together, they make for a true moonshot.

This moonshot is both ambitious and necessary, and more timely than ever given economic challenges around the world. Importantly, the AI Economist is a powerful optimization framework that can objectively automate policy design and evaluation. This will allow economists and policy experts to focus on the end goal of improving social welfare.

Given the social and ethical implications that economic policies can have, we believe it is essential to have transparency in the process. By open sourcing the AI Economist, not only do we empower collaboration from all over the world but we also enable unfettered review of policy simulations.

The key ingredients are:

  • A high-fidelity simulation that should be grounded in data, and aligned with economic theory as well as with social and ethical values. Simulations should not be prohibitively expensive to run, and should be maintainable and modular.

  • AI policy models should be effective in a wide range of scenarios, explainable, and robust to economic shocks.

  • The simulation and policy models should be calibrated against real-world data and, as much as possible, validated in human-subject studies.

I suppose I am skeptical, but fortunately progress does not depend on pleasing me.  There is much more information at the original link.

For the pointer I thank Mike Doherty.