Margen
Five dollars was what each participant received for completing a survey, in an experiment published in 2007. Afterwards the researchers offered them the chance to donate part of it to Save the Children. Those who read the story of one girl gave 2.38 dollars on average; those who read famine statistics gave 1.14.
The set-up was deliberately unceremonious. The participant was handed five one-dollar bills and an empty envelope for filling in the survey, then read the appeal and decided how much to put into the envelope for the charity. Whatever they did not donate was theirs to keep, nobody watched their hands, and the money that did come out was genuinely sent to Save the Children.
Deborah Small signed the paper from the Wharton School at the University of Pennsylvania; George Loewenstein, from Carnegie Mellon; Paul Slovic, from the University of Oregon. They split the participants into groups and gave them different messages before asking them for money.
One group read the case of Rokia, a seven-year-old girl from Mali, in West Africa, who was going hungry, and whom the donation could feed. Another group got the same appeal turned into figures: food shortages for more than three million children in Malawi, three million Zambians facing hunger, four million Angolans — a third of the country — driven from their homes and eleven million Ethiopians in need of food aid.
The gap between 2.38 and 1.14 is almost exactly half. And the appeal with the figures was not false, or badly written, or asking for anything else: it described the same hunger, multiplied by twenty-one million people.
A third group got both things together, Rokia and the figures, and that is where it gets strange. That message had emotion and context, a visible case and the proof that it was not an isolated one. It gave 1.43 dollars on average: less than those who only met Rokia. Adding the scale to a concrete case did not reinforce it.
Slovic uses the expression psychic numbing to describe that drop. The first case feels complete because one specific person can receive one specific bit of help: five dollars feed Rokia. Five dollars split between twenty-one million people feed nobody, and you sense that before you think it.
In the fourth study of the same paper, the researchers tried something else. Before presenting the request, they asked some of the people to think analytically: a few arithmetic problems, and then the appeal. They added no new data and did not change the recipient. Donations to Rokia went down. With the opposite group, asked to attend to what they were feeling, the response to the statistics did not go up. The switch works in one direction only: you can turn off someone who was going to give, and you cannot turn on someone who was not.
And it is worth looking at what the combined message contained: exactly the same girl, exactly the same countries and one more line of information. Nobody took anything away. The version with more data raised less than the version with less data, and that is the part no campaign-writing manual expects.
It is worth saying what the experiment measured and what it did not. It measured donations to a charity, with five dollars in hand and a sheet of paper in front. It did not measure work emails, or residents' petitions, or company presentations. What comes out of it is a mechanism, not a law.
The mechanism is this: the reader does not doubt the figure. They can believe it completely and, at the same time, feel less able to do anything. If the message starts with the scale, you have asked them to process the size before letting them meet anyone.
And the mechanism is recognisable outside the laboratory. "We have 386 open tickets" reports the load and does not show what is failing. "Laura has been three days unable to open her payslip" shows the consequence; after that, the 386 are there to say it is not a rare case and the queue is growing.
Nor does it need novelising. A name, a blocked action and a concrete consequence are enough, and the case has to be real and checkable: none of this licenses picking a misleading example or hiding the figure that gives the scale. The figure is not surplus. What it does not do is the work a name does.
Tomorrow, when you write that email, put the named case first. The figure goes in the next sentence.