What this covers
Megan McArdle and Russ Roberts discuss how societies learn—and habitually fail to learn—from catastrophe. McArdle's central claim is that hindsight bias and recency bias trap us into over-preparing for the disaster that just occurred while remaining blind to the next one. Rather than spending heavily on narrow prevention against the last crisis, she argues, governments and institutions should invest in adaptive capacity: pools of skilled workers, manufacturing flexibility, and research incentives that prove useful across multiple possible futures. Roberts pushes back on several points, but both hosts treat the pandemic as a case study in how institutional learning actually works.
The conversation ranges across pandemic response specifics—the CDC's botched early testing, the intention-to-treat problem with compliance, comparative effectiveness of masks and lockdowns—but treats these as windows into deeper pattern. McArdle argues that base-rate thinking should govern preparation for rare events, that institutional memory fades when old-timers retire and leave finance and other fields vulnerable to the next bubble, and that the ecological fallacy undermines state-by-state comparisons of pandemic policies. She contends that just-in-time inventory management, pharmaceutical over-subsidy, and mRNA research failures were actually fortunate accidents that built the strategic reserve of capacity needed when the crisis hit. The discussion also turns contentious on whether government lockdowns added real value beyond voluntary behavior, whether travel bans can work, and whether school closures will inflict lasting social damage.
McArdle argues that societies systematically prepare for the disaster that just happened rather than the next likely one, so the wiser investment is in general adaptive capacity—skilled people, flexible supply chains, and innovation incentives—rather than expensive, narrow prevention against the last catastrophe.
- Hindsight and recency bias lead us to over-prepare for the most recent crisis and under-prepare for all other plausible ones.
- Insurance for adaptability is cheaper and more broadly useful than insurance for prevention of a specific event.
- Maintaining a strategic reserve of skilled people and manufacturing know-how matters more than stockpiling specific goods.
Institutions failed to prepare adequately because we design infrastructure for common problems and dismiss unlikely ones until they happen.
- The CDC botched the early COVID test by processing patient samples in the same lab where it assembled test kits, contaminating them, and made the test needlessly complex when it could have simply copied the WHO's available test—and this was not the first such failure, having also botched tests for Zika.
“They contaminated the test they were making because they were processing samples in the same lab that they were using to put the test kits together. They tried to make it too complicated.”
- Organizations and regions rationally design their infrastructure for the most common problems they face—Texas builds power plants to shed heat, Minnesota installs de-icing equipment—and this is normally a good thing because we should devote the most attention to the most likely events, even if it leaves us exposed to freak occurrences.
“Places design for the most common problems. And, that's actually normally a good thing. We should think most about the things that are most likely to happen to us.”
- When an unlikely event hasn't occurred for a long time we conclude it will never happen again, and when it just happened we conclude we must heavily prepare against it; instead we should assess risks by their actual base-rate probabilities, recognizing that crises like financial panics and pandemics recur on long but regular cycles.
“When an unlikely thing hasn't happened for a while, we think it's never going to happen again. And, when an unlikely thing just happened, we think we needed to take a lot of preparation against it.”
- While the recent occurrence of a rare event does not by itself make it more likely to recur, sometimes a freak event genuinely signals that underlying conditions have changed and the unlikely thing has become more probable, so the challenge is distinguishing true regime change from noise.
“with the caveat that sometimes when a really unlikely thing happens, it actually signals that something has changed and that the unlikely thing has become more likely.”
- Because of hindsight bias, after an event more people will remember having correctly predicted it than actually did, which fools us into believing our judgment and control are better than they are and that we could have prepared against distant eventualities.
“if you ask 100 people in a room about something--about an event, -how likely something is--and ask them to write down a prediction; and then you let them see how it plays out. And, then you go back, more people will remember having correctly predicted that this thing was going to happen than actually is true.”
- We would in some ways have been better prepared institutionally to handle this pandemic in 1930 than now, because before antibiotics a surgical infection meant death, so hospitals and society cared more intensely about infection control and quarantine—a vigilance that atrophied as antibiotics reduced the stakes.
“institutionally, we would have been way better prepared to deal with this in 1930... that's because they really cared about protecting infection control.”
- Research on Wall Street suggests that people who have lived through a financial crisis recognize the warning signs and avoid the next one, so new bubbles tend to form only when the old-timers who remember the prior crisis start retiring.
“if you've lived through a financial crisis, you don't have another one because the people who lived through it recognize the signs. And, it's when the old-timers start retiring that you get new bubbles.”
America excels at adaptation and innovation but bungled specific pandemic decisions through political capture and misaligned incentives.
- The mRNA technology behind the COVID vaccines was originally developed for other purposes and failed as a long-term treatment because the body mounts an immune response to the mRNA; had the U.S. adopted European-style drug pricing that cuts payoffs to highly speculative research, this technology might never have been pursued, meaning we might have nearly as many government-funded cancer treatments but far fewer breakthroughs like the mRNA vaccines.
“the mRNA vaccines were not developed for vac--my understanding is the technology was actually developed for other things. And, it doesn't work for long-term treatments because your body... sees it as kind of funky”
- The Trump administration subsidized vaccines but underinvested in masks because Donald Trump personally disliked masks, considering them unmanly, and wanted to project that the pandemic wasn't happening so the economy would recover and aid his reelection.
“Donald Trump didn't like masks. He thought they looked unmanly. He wanted everyone to just pretend the pandemic wasn't happening”
- The U.S. should have committed to buying enough vaccine to inoculate every American from each of the top five candidates, using the first good one and giving the rest away, because it would have been cheap relative to the social damage and generated enormous soft-power benefits.
“taken the top five candidates and said, 'We will buy enough vaccine to vaccinate every single person in the United States. And, we're going to give away whatever we don't use.'”
- The government should have directly funded the U.S. N-95 manufacturer (Prestige Ameritech) to buy the expensive machines it was unwilling to risk capital on, because N-95 masks are far more effective than cloth at both preventing infection and transmission, and the manufacturer limited production solely to avoid being stuck with idle capacity after the pandemic.
“the reason they didn't was that you had to buy these very expensive machines and they didn't want to invest in all that capacity and then get stuck with it”
- The pandemic revealed two cultural strengths: Europe was better at the control side, achieving much lower caseloads through policy or culture, while America was better at adaptation, leading on vaccines—a contrast that reflects America's enduring capacity for adaptation that made it indispensable in World War II and economically, a capacity it could expand if it stopped tearing itself apart politically.
“Europe did better on the kind of control-side, the controlling-people side... But, the flip side of that is that we did better in vaccines than Europe”
- The most surprising lesson of the pandemic was that American institutions failed at levels not anticipated and at every level—the CDC botched its own tests, governors like Cuomo forced nursing homes to take COVID-positive patients, and others refused basic measures like masks—revealing failures independent of presidential leadership.
“our institutions failed at levels I did not anticipate. At every level.”
Pandemic control measures work mechanically but their real-world effect depends on voluntary compliance and cultural acceptance, not just policy.
- Just as raising a price must by some mechanism reduce quantity demanded over repeated cases—so claims that minimum wage hikes have no disemployment effect ultimately argue against an underlying mechanism—claims that masks don't work ultimately argue against the laws of physics, since putting on a mask and trying to spit demonstrates it blocks respiratory output.
“what's the mechanism by which you raise the price of something and nothing happens, like, over and over and over again?”
- Medical research distinguishes treating people from merely offering treatment (intention-to-treat), since real-world policies resemble the latter—most patients stop taking hypertension meds as directed within nine months—so lockdowns and mask mandates would be highly effective if perfectly followed, but the right critique is that the actual legal policies do less than expected because compliance is imperfect, not that the interventions themselves don't work.
“there's a concept in medical research called Intention to Treat, where you study the effects, not of actually treating people, but of offering them the treatment.”
- Australia's model of mostly open life punctuated by very short, total city-wide lockdowns of a few days to four weeks—shutting everything down until transmission chains break then immediately reopening—is preferable to a year of diffuse restrictions, and in the context of an imminent vaccine even pure delay radically shifts the death load.
“there's more argument for short-term lockdowns the way Australia does them, where most of the time everything's open, and then if they get cases in a city, it's just everything gets shut down for a couple of... weeks”
- There is evidence that people voluntarily hold their activity at roughly the level producing a transmission rate of one, so government lockdown mandates may add little beyond what individuals do on their own, though the UK experience of plummeting case numbers under serious enforcement suggests that hardcore mandates do have an additional effect.
“people just kind of hold their activity at roughly a level that produces a transmission rate of one... and that the government intervention might not have had that much impact in addition to that.”
- The claim that strict pandemic measures are 'un-American' is false: in the early 1900s New York gave the typhoid carrier Mary Mallon a choice between removing her gallbladder or lifelong jail because she kept working as a cook and infecting people, showing that earlier Americans—who lived under fewer daily restrictions and revered liberty—nonetheless accepted serious quarantine because they valued not getting typhoid.
“In the 1900s, the City of New York... told a woman named Mary Mellon--that she had an option. She could have her gallbladder removed, or she could go to jail for the rest of her life.”
The real reserve that matters in crises is skilled people and manufacturing capacity, not stockpiled goods or short-term supplies.
- While we should not stockpile all the masks a pandemic could consume, we should keep a small reserve so that in the first few days we can rapidly dump supplies on emerging hotspots like New York, rather than having health workers wear garbage bags and reuse respirators.
“we should stockpile some masks just because in the first flush of the few days, you want supplies that you can send quickly to the places that are having hotspots.”
- Rather than coercing technology transfer, the U.S. should attract strategic capacity such as semiconductor fabrication by offering Taiwanese firms, given China's invasion threat, visas, citizenship pathways, and subsidies to relocate their skilled workers and plants here—a carrot approach analogous to what Britain offered Hong Kong, building domestic knowledge and personnel for when it is needed.
“I prefer to say, like, 'Look, Taiwan, you're in a pretty dangerous area... you know what, visas for everyone, bring your chip fabs here, bring all of your skilled workers.'”
- The decades-long management gospel that just-in-time inventory eliminated wasteful stockpiles proved to have a hidden cost: in a pandemic or supply disruption, large inventories that can be processed turn out to be exactly what you want, which is the original purpose inventories served before fast transport—buffering against shipping interruptions.
“it turns out that in a pandemic, what you would really like is a large supply of inventory that you can then process.”
- The key resilience question is whether to stockpile goods or capacity, and the more important reserve is a base of skilled people and manufacturing clusters; America ramped up as the 'arsenal of democracy' in World War II not because it had physical plants but because it had machinists and people who knew how to do the work and could train newcomers, which is far harder to build from scratch than a factory.
“What we had was, we had a lot of machinists. We had clusters of manufacturing. We had people who knew what they were doing.”
Travel bans and lockdowns buy time for innovation rather than stopping disease spread, which is their true strategic value.
- Travel bans only slow disease spread rather than stopping it, evidenced by the observation that COVID did not spread that much faster than the 1918 flu despite air travel being far faster than boat travel, because diseases are very good at spreading; bans are valuable mainly to buy time for vaccine technology, and islands like Australia are the exception that can shut down early enough to work.
“I'm not even sure travel bans work. They slow it, but I think what you see historically is that the pandemic didn't move that much faster than, say, the 1918 flu--which is interesting, if you think about how much faster air travel is than boat travel.”
- The commentator and forecasting profession, like veterinary medicine in James Herriot's story, affords many opportunities to look like an idiot, because pandemic outcomes shifted repeatedly—blue states envied open Europe, then Europe's caseloads surged worse than the U.S., then Europe regained control while the U.S. did not—so any confident verdict was quickly overturned.
“his profession was going to afford many opportunities to look like an idiot. And, indeed it did. And, I think that that has been true of my own profession as well.”
Pandemic response became tribal symbolism, with both left and right adopting positions driven by cultural identity rather than evidence.
- Attributing every weather event—warmer, colder, or unchanged—to climate change creates the impression of a non-falsifiable thesis, which is poor messaging even if not necessarily wrong.
“it creates the impression of an utterly non-falsifiable thesis: If the weather gets warmer, it's because of climate change; if it gets colder, it's because of climate change; if it stays the same, it's because of climate change.”
- Mask-wearing became symbolic tribal politics: conservatives initially supported masks and mocked the CDC for downplaying them, but when Trump refused to wear one they defended him by arguing masks don't work and became invested in that position, while on the left people wear masks outdoors in low-risk settings out of social conformity rather than evidence—shown by the muted reaction when Connecticut, unlike Texas, lifted restrictions.
“Early on conservatives loved masks... And, then Donald Trump wouldn't wear one... So, they started arguing the masks don't work and then they got really invested in it, and it became symbolic politics.”
Government should eliminate price controls that suppress productive capacity in strategically important markets during normal times.
- Far from being unwilling to spend on rare events, we have a large political appetite for spending heavily on whatever unlikely thing just happened, even though the next rare event is probably not a repeat of the last one—the recently realized risk remains exactly as likely as before, not less, but it is no more likely to recur than other dormant risks.
“we have a large political appetite for spending a lot of money on whatever just happened. On whatever really unlikely thing just happened.”
- The low-hanging fruit all libertarians can agree on is to identify and withdraw from markets where the government intervenes to artificially keep supply and prices low for political convenience, because such interventions suppress the adaptive capacity that would be strategically important during disasters like asteroid strikes, war, or pandemics.
“look for: Where are the ways that our government is suppressing our adaptive capacity by going into markets and intervening to get an outcome that is politically convenient”
- While the government should invest more in asteroid defense, private companies like SpaceX and tech billionaires investing in space demonstrate a significant role for private initiative even in domains traditionally thought to require government.
“private companies like SpaceX are also showing us that there is a big role for private initiative here; and the number of tech billionaires who just apparently want to use their money to... get us, humanity, into space”
- If attacked by a bear, do not run because they run fast, and response varies by species: with grizzlies play dead, while with some bears you can punch them in the nose.
“with grizzlies, you just want to actually apparently play dead... And then like with some of them you can punch them in the nose and they don't like it.”
Pandemic risk may be increasing as human population growth and agricultural intensification expand contact with novel animal pathogens.
- Pandemics may be becoming more likely because a growing, richer human population pushes into marginal areas and intensifies agriculture and livestock farming, increasing contact with wild animals carrying novel diseases that can jump to humans.
“humans are pushing into more and more kind of marginal areas and farming and so forth. And, they encounter wild animals, and those wild animals have new diseases that can pass to humans.”