Why Most Analysts Think They're Better Than They Are

In 1977, the educational researcher K. Patricia Cross surveyed college professors and asked them to rate their own teaching. 94% put themselves above average. Two-thirds placed themselves in the top quarter. The numbers made no sense, and that was the point.

I've thought about that study a lot over the years, usually when I'm sitting across from an analyst who has decided they've cracked it.

The same arithmetic problem shows up in markets. A disproportionate number of analysts cannot, by definition, generate alpha. Someone has to be on the losing side of the trade. And yet a striking number of people in this business talk as if they've already worked the market out, as if the only thing left is to execute.

Last year I spoke with a 31-year-old hedge fund analyst who told me, with a straight face, that there wasn't much left for him to learn. He wasn't joking. He had a process, a few sources, a model he trusted. In his own mind, he'd arrived.

The psychology literature has a name for this. Or several. Dunning and Kruger's 1999 work showed that people with the least competence in a domain tend to overestimate themselves the most, largely because the skills needed to be good at something are the same skills needed to recognize you're not. Barber and Odean, looking at retail brokerage accounts through the late nineties, found that the most overconfident traders traded the most and earned the worst net returns. Confidence, it turns out, is expensive.

Here's what struck me during my years at UBS, sitting across the table from some of the most respected investors in the world: they were almost paranoid about what they didn't know. One PM I worked with kept a running list of things he'd been wrong about, going back years. Another would interrupt his own pitch to flag the assumption he was least sure of. These weren't false modesty moves. They were operating procedure.

The analysts who plateau tend to make the same quiet decision. Usually around year five or six, once the basics feel automatic, they stop treating analysis as a craft and start treating it as a pipeline. Pull the data, run the screen, write the memo, move on. The work gets faster. It also gets shallower. A process job dressed up in a blazer.

The edge was never the model. It wasn't the source network, or the Bloomberg terminal at 6 a.m., or the letters after your name. The edge is a disposition. It's the refusal to believe you've arrived, even after you've been right a few times in a row. Especially then.

If there's a single question worth asking yourself at the end of every quarter, it probably isn't "what did I get right?" It's "what do I still not understand about the things I got right?"

That's the question that keeps the job interesting. It's also the question that keeps you employed.