Posted by Tim Harford

Why did audience members fail to flee a deadly fire… despite being told to escape?

Flames are spreading through a Cincinnati hotel. The staff know it, the fire department is coming, and the people in the packed cabaret bar have been told to evacuate… and yet people hesitate to move. Why don’t we react to some warnings until it’s too late?

A version of this episode was first released in the summer of 2020. This update includes a conversation between me and the Cautionary Tales showrunner Georgia Mills, reflecting five years on, on the experience of the covid pandemic – and making Cautionary Tales during lockdown.

[Apple] [Spotify] [Stitcher]

Further reading and listening

I first found out about the Supper Club Fire from The Ostrich Paradox by Howard Kunreuther and Robert Keyer. The fullest account I could find is in Amanda Ripley’s book The Unthinkable, including an interview with Walter Bailey.

Another source on the fire was Drue Johnston and Norris Johnson “Role Extension in Disaster: Employee Behavior at the Beverly Hills Supper Club Fire.” Sociological Focus, vol. 22, no. 1, 1989, pp. 39–51., www.jstor.org/stable/20831497.

For the details on the Torrey Canyon spill, the two key sources are Oil and Water by Edward Cowan, and The Black Tide by Richard Petrow.

For a more contemporary discussion of plan continuation bias I recommend Meltdown by Chris Clearfield and Andras Tilcik.

On ambiguous threats see Michael Roberto, Richard Bohmer and Amy Edmondson, ‘Facing Ambiguous Threats’ Harvard Business Review November 2006 https://hbr.org/2006/11/facing-ambiguous-threats

Posted by Tim Harford

One death is a tragedy. A million deaths is just a statistic. If Stalin ever said such a thing, he wasn’t the first — but the ghoulish claim has stuck to him because he is one of very few politicians with more than a million deaths on his conscience.

The list of government actions that deliberately or negligently led to the deaths of more than a million people is short and ugly. There are civil wars, famines and a cluster of atrocities surrounding the second world war, but not many governments have been so evil or so reckless as to pass that horrendous target.

Incredibly, there is now a case for adding the Trump administration to the list. Elon Musk boasted in early February that, “We spent the weekend feeding USAID into the wood chipper.” Musk is gone but the White House budget request for next year pencils in a cut of two-thirds to global health and humanitarian funding. Foreign aid has a mixed reputation, it is true, but that cut would plausibly cause a million deaths in the next 12 months alone.

Clearly there is a difference between murdering someone and refusing to save their life. It is one thing to kill a child by pushing them into deep water and another to let a child drown because you don’t want to get your suit wet. Where exactly the difference lies, we can leave to the philosophers. Either way, the child is dead.

The figure of a million deaths is both an estimate and a forecast. The estimate may be wrong for all the usual reasons, and it may also be wrong either because the US changes course, or because philanthropists or other official aid agencies find ways to rescue what is being lost.

Despite the caveats, a million deaths is a staggering number. It comes from Charles Kenny and Justin Sandefur, two respected researchers at the think-tank the Center for Global Development. They reckon that if the cuts to humanitarian assistance happen — from $8.8bn to $2.5bn — then 675,000 people are likely to die from HIV within a year, and 285,000 from malaria or tuberculosis.

“This estimate is based on models, and models have their weaknesses,” Kenny told me. “But we think these numbers are at the conservative end of estimates.”

The biggest contributor to this figure is the number of people projected to die from HIV, and it is not an outlandish estimate. A study published in The Lancet in March reached a similar conclusion, projecting that foreign aid cuts by the US and other donors would lead to 10.8 million extra HIV infections and 2.9 million extra deaths over the next five years. Other researchers have produced even more dramatic projections of the death toll.

Nor is there much mystery as to how these deaths might happen. Pepfar (the President’s Emergency Plan for Aids Relief, established by George W Bush) supplies antiretroviral drugs that keep 20 million people alive. The effectiveness of these drugs is well understood. They suppress the virus and prevent transmission, including from mother to baby.

The main uncertainty is whether the drugs will continue to be supplied, despite the enormous disruption that has already occurred. The US secretary of state Marco Rubio has maintained that life-saving aid is continuing, but clinics have closed, and people are finding it impossible to get the medication they need to keep them alive.

Some Americans believe that one-quarter of government spending goes on foreign aid. This is a misperception on a staggering scale. The total US foreign aid budget has been closer to 1 per cent of government spending — and the humanitarian aid budget of $8.8bn was not much more than one dollar in a thousand of everything the US government spends.

Few people in the foreign aid industry would argue every cent saves lives. Many projects do not try to save lives directly. Projects aimed at providing access to clean water, or to schooling, or to contraceptives, might indirectly save lives but they are not included in Kenny and Sandefur’s calculations. Projects often operate in challenging places. Corruption or waste are always risks. There are endless conversations in the foreign aid community about how the whole business might be reformed and made more cost-effective.

In some cases it might be desirable for national governments to find their own funding. For example, the Center for Global Development researchers estimate that nearly 200,000 South African lives are being saved each year from HIV alone, thanks to US foreign aid. That is impressive, but South Africa is not a subsistence economy. It is an upper-middle-income country. The South African government should have the capacity and the funds to supply its own antiretrovirals, and in the long term it might be better if they did.

Yet this is no way to reform anything. The cuts are so abrupt that life-saving services are falling apart before our eyes.

A few former USAID staffers have been working to salvage something from the wreckage. In response to requests from private philanthropists, the team — known as PRO (Project Resource Optimization) — has compiled a list of what they euphemistically call “critical funding opportunities”. These are high-impact projects that might need only a few hundred thousand dollars to complete, or to keep on life-support in the hope that stable funding can be found.

“We originally called ourselves the Lifeboat Project,” says Robert Rosenbaum of PRO. “And I think that metaphor holds better than any other.” A philanthropist told them to find a name that wasn’t quite so dark.

The darkness is justified. When a huge ship sinks, lifeboats can save lives, but you need enough lifeboats, and you need enough time. We have neither. Above all, the lifeboats only work if help is on the way. With the average US voter unaware just how many lives they were saving, and at just how small a cost, it is unclear whether that help will arrive.

A million deaths may be a statistic, but it is also a million tragedies. Most of these tragedies could still be prevented.

Written for and first published in the Financial Times on 26 June 2025.

Loyal readers might enjoy the book that started it all, The Undercover Economist.

I’ve set up a storefront on Bookshop in the United States and the United Kingdom. Links to Bookshop and Amazon may generate referral fees.

Posted by Tim Harford

Lying on the cold metal table, Voyne Ray Cox knew the drill. This was his ninth round of cancer treatment – which is why he was certain that what happened next couldn’t be right. He heard a sizzling sound and saw a blue flash. And then – agony. It was like someone had thrust a hot skewer through his shoulder. He cried out in pain, but the operator was down the corridor and she couldn’t hear him. She blasted him again and again with the red-hot radiation beam.

Ray wasn’t the first patient to be burned by the Therac-25 therapy machine, and he wouldn’t be the last. Its dual-purpose design, controlled by a software programme, was supposed to offer hospitals more bang for their buck. But as patient after scorched patient suffered ulcerated skin and yawning lesions, it should have been clear that something was horribly wrong. Why did it take so long for anyone to put this awful puzzle together?

This episode was originally released on Pushkin+. New exclusive Pushkin+ episodes are released each month.

[Apple] [Spotify] [Stitcher]

Further reading

Nancy G. Leveson and Clark S Turner “An Investigation of the Therac-25 Accidents” Computer 1993

Nancy Leveson “Medical Devices: The Therac-25” in Nancy Leveson Software: System Safety and Computers 1995.

Steven M. Casey Set Phasers On Stun

Supplementary sources include

Charles Huff “Therac-25 Case NarrativeOnline Ethics Center 2003 and Adam Fabio “Killed By a MachineHackaday 26 October 2015

Helen Nissenbaum. 1994. Computing and accountability. Commun. ACM 37, 1 (Jan. 1994), 72–80.

“Lethal Doses Radiation That Kills” The Plain Dealer, Cleveland 16 December 1992

James Reason Human Error

And

Tom Standage 1843 Magazine What we can learn from the world’s first computer bug1843 Magazine 4th Sep 2019

Posted by Tim Harford

Smoking kills. A few people had suspected as much before the second world war, but it was not until 1950 that the scientific evidence began to accumulate that smokers were at dramatically higher risk of lung cancer than non-smokers. Other health risks of smoking would be identified over the years that followed.

Pity the poor smoker. Addicted to a popular product that had seemed harmless, he was now being told that his habit was killing him. (It often was a he, although cigarette companies had also marketed cigarettes to women with the feminist slogan, “torches of freedom”.)

What to do? An editorial in the British Medical Journal captured one solution. “It is said that the reader of an American magazine was so disturbed by an article on the subject of smoking and cancer that he decided to give up reading.”

It’s hard to think of a pithier example of what psychologists call the biased assimilation of information. Biased assimilation refers to the various ways in which we avoid unwelcome facts and seek out information that bolsters our own views. The most obvious example — obvious because it is a fairly novel development and because the mechanics of selecting information are so transparent — is the way we follow like-minded people on social media. But people have long sought out news sources that bolster their particular views of the world, as anyone who has ever had a newspaper round and surveyed the contrasting front pages can attest.

There are plenty of more insidious ways in which we absorb some information sources and miss others. Some of them are hidden and algorithmic. Even if you take care to follow commentators or news sources across the political spectrum, social media companies will show you more of what seems to hold your attention. If you are on the centre-right and share a lot of centre-right articles with your friends, soon enough you will stop seeing much stuff from leftwing columnists, no matter how many of them you think you are following.

Less obvious still are our own mental algorithms, which have been subtly skewing our view of the world for far longer than the internet has existed. Two people with contrasting preconceptions about the world can look at the same information but perceive it very differently. Imagine that you happen to encounter a newspaper article discussing what we know about the effects of the death penalty, including a study by researchers Palmer and Crandall. You are told that these scholars found pairs of neighbouring states with different capital punishment laws, and compared the murder rates in each pair. In eight of the 10 pairs, murder rates were higher in the state with capital punishment. This research suggests that the death penalty does not act as a deterrent.

What to think of Palmer and Crandall’s research? Does it seem plausible? If you’re opposed to the death penalty, then it probably does. But if you’re in favour of capital punishment, then you might quickly notice the potential for sloppy errors. Was this research peer reviewed and professionally conducted? Did Palmer and Crandall consider alternative explanations for the pattern they spotted? Should we really buy the idea of paired comparisons between adjacent states? In short, do Palmer and Crandall really know what they’re doing, or are they hacks?

Do not fear that you might offend Palmer and Crandall. They are fictional. They were dreamt up in the late 1970s by three psychologists, Charles Lord, Lee Ross and Mark Lepper. Lord and his colleagues recruited experimental subjects with strong views about the death penalty and showed them summaries of two imaginary studies.

One of these made-up studies demonstrated that the death penalty deterred serious crime. The other, by the fictitious researchers Palmer and Crandall, showed the opposite. As one might expect, the experimental subjects were inclined to dismiss studies that contradicted their firmly held beliefs. This is biased assimilation in action, not at the level of picking which researcher to follow or which newspaper to read, but at the level of picking which information to accept or to reject.

Lord and his collaborators also discovered something more surprising. In some cases, their opinionated experimental participants were shown brief research summaries. In other cases, they were given more detail about research methods, supplemented with graphs and commentary by other fictional academics. The more detail people saw, the easier they found it to reject the unwelcome evidence. Each new detail was an opportunity to dismiss the whole thing.

We are well used to worrying about people consuming a skewed information diet, shorn of context, detail and balance. We imagine that a balanced, detailed news diet would be better. The study by Lord and his colleagues suggests it would not be as helpful as we might hope — not in the face of a committed believer.

That believer is likely to systematically reject contradictory evidence, meaning that as the balanced evidence pours in, the evidence they actually read, accept and remember piles up only on one side. Loading up both sides of the cognitive weighing scale with equal weights is not going to produce balance if the weights keep accumulating on one side and bouncing out of the other.

We shouldn’t overgeneralise from one study, particularly as Lord, Ross and Lepper deliberately recruited experimental participants who were passionately committed to a point of view. For most people, on most issues, a balanced diet of information is likely to be a healthy one.

Yet that underscores the original point: what we believe about the world depends on which ideas we are open-minded enough to entertain. Providing all the detail and balance in the world is useless when you are faced with a reader or viewer who greets all of it with a selective memory and a lopsided scepticism. A curious and open-minded media ecosystem is undoubtedly important, but so too are curious and open-minded citizens.

Written for and first published in the Financial Times on 20 June 2025.

Loyal readers might enjoy How To Make The World Add Up.

“Nobody makes the statistics of everyday life more fascinating and enjoyable than Tim Harford.”- Bill Bryson

“This entertaining, engrossing book about the power of numbers, logic and genuine curiosity”- Maria Konnikova

I’ve set up a storefront on Bookshop in the United States and the United Kingdom. Links to Bookshop and Amazon may generate referral fees.

Profile

tinyjo: (Default)
Emptied of expectation. Relax.

June 2020

S M T W T F S
 1 23456
78910111213
14151617181920
21222324252627
282930    

Most Popular Tags

Expand Cut Tags

No cut tags
Page generated July 25th, 2025 09:18 am
Powered by Dreamwidth Studios

Style Credit