Evidence
CEO tenure dropping tells you about what we measure, not what works
Everyone points to activist investors and quarterly earnings pressure, which—fine, those are real. But I think the actual story is simpler and weirder: we stopped measuring CEO performance on anything that takes longer than five years to show up in a spreadsheet.
Here's the thing nobody wants to say out loud. Building something robust takes time. Not mysterious time. Specific time. You can't know if your capital allocation strategy actually works for nine years. You can't tell if you've fixed a structural labor problem until you've hired through two full cycles. You can't know if you've genuinely improved a supply chain or just got lucky with commodity prices until you've weathered a disruption. This is boring and obvious and everyone hates it because it means you can't declare victory at the earnings call.
So instead we hired a class of professional CEOs whose job is to show improvement within a five-year window, then get out before the hard stuff lands on someone else's balance sheet. Strip costs, sell the non-core business, massage the metrics, and move to the next company where you can do it again. It's not that these people are evil—it's that we built a system that makes it irrational to do anything else. A CEO who waits ten years to get real traction on something structural is leaving money on the table by staying for the decade, because they could've jumped ship at year four, reset their equity clock at a new company, and done the same trick three times over.
Transit systems have the opposite problem. We're stuck with the same people for decades whether we like it or not, which means bad decisions don't get forgotten—they compound. You'll see a planning director who made a ruinous choice in 1998 still defending it in 2015 because backing down means admitting they wasted twenty years. Make of that what you will.
4 comments
Log in to comment.
I think you're describing real incentives, but the counterfactual might be weaker than it looks. Jack Welch had a famously long tenure at GE (1981-2001) and was basically doing exactly what you're describing—strip costs, optimize for quarterly metrics, move capital around—except he got to do it for twenty years and had the organizational power to make it stick. He's remembered as genius, not as someone who stayed too long. The system rewarded him handsomely for treating the company as a portfolio optimization problem.
So I'm not sure shorter CEO tenures are actually the *cause* here so much as a symptom of something that would've happened anyway once institutional investors got large enough and sophisticated enough to demand that kind of capital efficiency. The measurement problem is real, but it's not clear that longer tenures would fix it. You'd just get longer-duration cost-stripping, which is what happened at Welch's GE—the structural problems took longer to show up, but they showed up worse.
The transit director comparison is interesting but cuts the other way maybe. The person defending a 1998 mistake in 2015 might also be the only person keeping the institutional knowledge around to eventually fix it. Turnover has costs too, especially in complex systems. Harder to measure those costs, which might be why we don't see them reflected in how we hire.
The thing that catches me is the assumption that you can even measure the thing that matters. In trades training, we've been stuck on this for years: nobody knows how to measure whether an apprenticeship program actually works until five, seven years out, when you see whether people are still in the field, still employed, making decent money, not injured. But programs get evaluated on completion rates and immediate placement. So you get schools pushing kids through in the minimum time, cutting corners on the stuff that takes patience to teach right—how to read a room, how to diagnose a problem instead of just replacing parts, why you do something a certain way instead of the faster way.
The transit planner problem is real, but I'd flip it slightly. The problem isn't that they stick around too long. It's that when you make a bad call on something with a thirty-year lifespan, you don't actually face consequences for it because the industry moves slow and by the time it fails, you've already buried it under five new budget crises. A CEO at least has to watch their cost-cutting blow up their supply chain in year six or seven. A transit planner can retire and let someone else explain why the line they designed in 1995 doesn't work.
The real issue both have in common: we've structured things so that the people making decisions don't stick around long enough to care about the long-term feedback, but also can't be easily removed even when they should be. That's a system problem, not a people problem. And yeah, that's harder to fix than blaming quarterly earnings or civil service incompetence.
I think you're describing a real incentive problem, but I'm not sure the measurement story quite holds. If we actually stopped measuring long-term stuff, wouldn't we see CEOs who *did* stay longer getting rewarded more? Instead the data shows shorter tenure is correlated with higher pay and better landing spots, which suggests the market actively prefers the exit-early pattern. That's different from saying we lack the *ability* to measure—seems more like we choose not to weight it heavily.
Also curious whether the labor-cycle thing you mention is actually the binding constraint. Berger et al.'s work on activist investors shows measurable negative effects on R&D and capex, which are the actual long-term bets. But those effects are pretty concentrated in certain sectors and time periods. Do you think that's capturing the core story, or is there something about CEO incentives that would persist even without activists? Because if it's really about measurement and patience, I'd expect to see more variation across industries than we do—some boards should be willing to pay for the long view.
The five-year window thing is real, but I'd push back on how clean the story is. I've watched this play out in apprenticeship programs, which have their own perverse incentives. You get federal funding based on completion rates and job placement within six months of graduation. So programs optimize for getting bodies through the door and into jobs fast, then they're done measuring. What happens in year two when half your first cohort quits because they're making twelve bucks an hour doing grunt work while the journeyman they're supposed to be learning from is too busy to teach them? Not your problem anymore. Completion rate still looks good.
The thing that gets you is that skill transmission actually does take the time it takes. You can't compress it past a certain point without something breaking. A sparky needs to see maybe two hundred different problem types, different configurations, failure modes, before they stop being dangerous to themselves and others. That takes years. But the program gets its money for the placement, the contractor gets cheap labor, and the apprentice gets a paycheck, so everyone's incentive is aligned to call it a success. Then we act shocked that the actual completion rate to journeyman is somewhere around thirty percent, and we blame the apprentices for "not having the work ethic" instead of noticing the system isn't measuring what matters.