In a roughly twenty-minute online session, a participant is told they have been assigned the role of a manager. On screen is a short profile of the worker they are matched with: the worker's gender, the quarter of the year the worker was born in, a favorite hot beverage, a favorite color.1 The worker has already answered twenty puzzle questions from an IQ test and been ranked against nineteen other people, and the manager has just been told where that worker placed. Now the manager decides what to pass along. The choices are the exact result, a vaguer version of it, or nothing. Lying is not one of them.1

That is the setup of an experiment by Alexander Coutts, Boon Han Koh and Zahra Murad, published online by Management Science in July. They recruited 2,620 people in the United States through an online research platform, and about half sat in the manager's chair.1 (The paper calls them advisors.) The authors' model predicted that good news would always be passed along. What they found was that about a quarter of their managers held back some or all of the result, "even when feedback is of immediate instrumental use for decision making by the advisee," and that "positive feedback is obscured nearly as much as negative feedback."1 Among managers told the worker's exact quartile, and that it was in the top half, 29 percent sent something vaguer or nothing at all.1

The explanation that comes to mind first

My read is that most leaders, asked why a manager sat on a performance result, would answer that the news was bad and the manager wanted to spare someone. It is a comfortable explanation, since it makes the withholding a kindness and points toward a familiar fix: teach managers to deliver hard messages better.

The data support that story in only one corner. Bad news was hidden more often than good when the manager's information was vague and the feedback would arrive too late to change anything the worker did.1 Otherwise, timing made little measurable difference. Some workers got their feedback before deciding whether to bet their score in a competition for a bigger payout, and some got it after, and the share of managers holding back was not statistically different between the two groups.1

Last week I wrote about directors who mute their dissent to protect relationships with colleagues. This experiment looks downward, and the motive it turns up is a different one.

What predicted the silence

To get at reasons, the authors ran a second experiment with 1,023 more people in the manager's role, all of them in the version where feedback arrived too late to matter, and asked each of them to pick, from a list, every consideration behind their choice.1 Managers who passed good news along were more likely to cite transparency and a sense of duty toward the worker. Managers who sat on good news more often pointed to worry about what the worker would think of them and to how the worker's score compared with their own, though those worries ranked below two stronger predictors: a weaker sense of transparency and duty, and a view that the feedback was not useful.1 The authors conclude that hiding good news is best explained by "the relatively lower importance some advisors attach to transparency or a sense of duty," and that their evidence does not support their starting assumption that managers act out of care for the worker, apart from those who thought the feedback useful.1 Reasons people report about themselves are soft evidence, as the authors acknowledge, and they found the same pattern in free text explanations from the first experiment.1

A result that never reaches the person it describes usually says more about the one holding it than about the one it was meant for.

What the lab leaves out

These were not real managers. They were paid participants in a single online session, with average bonuses of about a pound and a worker they knew only from a four line profile.1 The authors note that an anonymous setting like this one, where nobody meets twice, may understate how strongly duty can push people to share.1 Real workplaces also bring back what the lab removed, like a working relationship a manager would prefer not to strain, and my guess is that this pushes the other way. The paper does not claim to know the net.

The paper's gender result needs the same care. When the manager's information was vague and the feedback came too late to matter, managers hid a result in the bottom half from women 34.2 percent of the time and from men 17.5 percent of the time, a gap the authors describe as “only marginally statistically significant,” and one that did not reach significance in their follow-up experiment.1 Exploratory analysis traced it to more concern for protecting women's egos, and the authors cite recent evidence suggesting women are no less eager for performance feedback than men.1 A marginal result should be held loosely. It is still worth a look in any calibration process that runs on impressions instead of written ratings.

Make it part of the job

The authors suggest structured feedback forms or rating scales as a way to reduce the role of personal preference, and they write that "institutions should cultivate strong norms about transparency and duty to their workforce."1 They also suggest "Framing precise feedback as a hallmark of effective mentorship," so that sharing a result fits the way managers see themselves.1

I think that argues against reaching first for another workshop on difficult conversations, which assumes the manager wants to tell the truth and lacks the words. A quarter of these managers had the words handed to them on a menu and picked a vaguer option. What changes that, in my view, is treating the exact result as something a manager owes the employee, said plainly and put in writing, and something you ask about the way you would ask about any other commitment, starting with how you tell your own direct reports where they stand. Include the good results, since those are the ones nobody thinks to audit.

If one of your strongest people asked tomorrow how her last quarter ranked, would her manager give the exact answer, and how would you know if he did not? I would like to hear how your organization handles it.