Picture a chief operating officer two weeks before budget lock. Two requests from the same HR leader sit on her desk. The first asks for a larger training line, because the skills gap list has grown and a vendor has a catalog to match. The second, filed under a different heading, asks her to rework how promotions are decided and how managers are rated on developing their people. She approves the first in an afternoon and puts the second in a folder marked next year.

A working paper issued this month by the National Bureau of Economic Research suggests she may have funded one half of a bundle and shelved the other. David Deming, Raffaella Sadun and Christopher Stanton document that some firms are more likely than others to invest in their workers' skills, and that the difference tracks how the firm is run. Their abstract puts it plainly: "better managed firms, even in the same industries and labor markets, are more likely to train."1

What the management score picks up

The size of the association deserves a precise statement, and so do its limits. In the 2022 manufacturing wave of the World Management Survey, which covers 1,344 firms in 11 countries, one more point on the five point management scale (roughly 1.7 standard deviations in that sample) was associated with a 19.5 to 24.3 percentage point higher probability of providing advanced training, even after controlling for firm size, industry and country. The lower figure is for new employees and the higher for existing ones.1 The part of the score that accounts for most of that association is people management, which covers talent development, promotion and performance evaluation.1 Those are the items on the second request, the one in the folder.

The authors describe their evidence as descriptive, and they note that dynamic firms may already have better management in place to prepare for growth.1 A company that expects to grow could be both training and managing better for the same reason. Nothing in the paper shows that rewriting promotion criteria will produce a training program. My read is that the paper still changes the order of the questions, because a training line and a set of management practices turn out to travel together, which makes the budget line a poor place to start the diagnosis. The authors trace both back one step further, to the skill of top managers, which they argue shapes the structure and people practices that training investment then follows.1

Why the skills stay home

Finance has a standard objection to all of this spending: people who get trained leave, and the investment leaves with them. The authors answer with a model that they use to illustrate the idea. In their framework, top managers with different skills build organizations around their own needs, and that makes a general skill worth more inside the firm that shaped the work than anywhere else, which is how the firm earns a return on what it taught. In the model the trained workers earn more too, so the firm and the worker split the gain.1

My read of that model is that the scarce thing is the organization a skill plugs into. Spreadsheet modeling and structured negotiation are portable on paper, and what a person can do with them depends on who decides what, which reports can be trusted and what happens to someone who acts on the training. A course can supply the skill. The organization around it decides what the skill is worth, and a competitor cannot buy that from the same catalog.

A training line reads like a receipt for how a company is run: whom it promotes, how it rates managers and what it asks of the people it develops. Raising the line without touching those decisions mostly buys a larger receipt.

Two kinds of programs

In a 2023 employer survey, the paper also sorts 954 American firms with at least 100 employees, all of which already ran some upskilling or reskilling program, into two groups. In 260 of them the flagship program was local in scope and oriented toward frontline employees. In the other 694 it was much more structured in delivery and reached farther up the hierarchy, with 54 percent including business unit leaders against 40 percent in the first group.1 The abstract adds that training clusters around either managerial or frontline development.1 Where a company's main program sits, and how far up the organization it reaches, is an organizational choice, whether or not anyone recorded it as one.

An earlier piece on the cost of cutting meetings, Nine Thousand Dollars a Head, looked at teaching that happens with no budget line at all. The same logic applies here from the other side. Whatever a company teaches on purpose and whatever it teaches by accident come from the same organization.

So before the next training request is approved, it is worth asking what a graduate of that program will run into on Monday morning, and whether any practice in the building rewards using what was taught. If the honest answer is no, which request on the COO's desk is the real one?