In 1961 Alfred Oberlander, who ran the Rockaway district office of the Metropolitan Life Insurance Company, sorted his salespeople into three units. His six best agents went to his best assistant manager. An equal number of average producers went to an average assistant manager. The low producers went to the least able one.
He asked the top unit to bring in two-thirds of the premium volume the whole agency had sold the year before. The office started calling it the super-staff. By Oberlander’s account, its first twelve weeks “far surpassed our most optimistic expectations,” and the agency as a whole improved by 40%.
The bottom unit went the way everyone expected, too. Its sales fell and more agents left. When one of them resigned, Oberlander told her he was sorry she was leaving. Until then he had said nothing to her at all.
“No you’re not,” she said. “You’re glad.”
So far the sorting looked like a sorting of talent. Then there was the middle unit.
The manager who wouldn’t be average
Its assistant manager, by Livingston’s account, refused to believe she was any less capable than the super-staff’s manager, or that the top agents had more ability than hers. She told her agents they had more potential than the super-staff and lacked only their years of experience in insurance. Then she challenged them to beat the stars.
They never matched the super-staff’s dollar volume. But each year, the middle unit raised its productivity by a higher percentage than the super-staff did.
The story comes from J. Sterling Livingston, who taught at Harvard Business School, in a 1969 Harvard Business Review article called “Pygmalion in Management.” He named it for George Bernard Shaw’s play, where Eliza Doolittle, the flower girl turned lady, puts it in one line: “the difference between a lady and a flower girl is not how she behaves but how she’s treated.”
What a boss expects of you leaks into how you’re treated, and can show up in your results.
Livingston thought the leak was mostly unintended. “Managers often communicate most when they believe they are communicating least,” he wrote. The Rockaway manager denied ever signaling low hopes to the bottom unit, yet those agents took their assignment as a request to resign. People who expect to fail, Livingston argued, make fewer sales calls and avoid closing when a close risks another rejection, and the forecast comes true. High hopes have their own condition: people have to find them “realistic and achievable.”
It is still one office, told by the man who ran it. The best agents also got the best manager, so nobody can say how much of the gap came from belief and how much from talent and coaching. Later studies separated them with labels assigned at random.
Labels drawn at random
One came from the Israel Defense Forces. In 1982 Dov Eden and Abraham Shani led instructors in a combat command course to expect more of certain trainees, who had in fact been assigned at random, and those trainees went on to score higher on objective achievement tests.
Later studies kept the effect and shrank the promise. A 2000 meta-analysis by Brian McNatt, of 17 studies of adults in management settings, found effects that “can be fairly strong” in some contexts but vary widely. They were stronger in the military, with men, and for people of whom little was expected at the start, and McNatt called for more research in ongoing jobs rather than training courses. In schools, where Robert Rosenthal had run the experiments Livingston drew on, a 2005 review by Lee Jussim and Kent Harber concluded that teacher expectancy effects are real but “typically small,” may fade rather than build up, and that teachers’ expectations may predict results more because they are accurate than because they fulfill themselves.
Then Eden tried to put the effect to work. Across seven field experiments, he and his colleagues ran workshops to teach managers a Pygmalion style. They found “little evidence that the workshops influenced leaders or followers.”
A false label handed to an instructor raised scores. Teaching managers the lesson changed little.
The handover note
Say you take over a team of six, and your predecessor leaves you a note on each person. Next to one name, a support engineer with a year on the team, it says: weak, probably won’t last.
Nothing in the note tells you to do anything. You don’t have to. He gets the routine tickets because that seems safe. Your replies to him are shorter than to the others, because you expect less from the exchange. Livingston would recognize the pattern, and so would the engineer.
The other path doesn’t require pretending the note is wrong. Notes can be accurate. You give him one real problem, sized so he can plausibly solve it, and tell him plainly that you expect him to. Then you watch what he does with it, not what the note said he would do.
If he leaves in six months, the note will look like a good prediction. Nobody will be able to tell whether it was a forecast or a set of instructions.