Notes on being a psychologist in a room of bankers

Discount rates are also a belief system.

I have spent a year now in rooms where the discount rate is treated like gravity — a fact of the universe you account for but do not argue with. 8%. 10%. 12%. The number depends on the industry, the risk, the market, the mood of the committee. But once it's set, it's set. You do not generally see someone push back mid-meeting on a 10% WACC.

This is because the WACC is the most effective psychological instrument in the room.

What a WACC actually does

Technically, a weighted average cost of capital blends the firm's cost of debt and cost of equity, weighted by their shares in the capital structure. It is the number you divide future cash flows by to make them commensurable with money today. That is the mechanical description.

The psychological description is different. A WACC is a commitment device. It is a number the team has agreed to not fight about, so that they can fight about other things instead. Without a fixed WACC, every discussion about a potential investment degenerates into a meta-discussion about what returns we're trying to clear. With one, the conversation can proceed to the actual question — is this company worth the number the model spits out — because the number is already anchored.

This is a psychologist's job. Anchors let groups reason. Groups reasoning without anchors melt.

Why the training mattered

Before Columbia, I trained as a behavioral scientist. The work was mostly slow: read a messy situation, ask what exactly is going on, test whether the stated reasons match the observed behavior. When I moved into finance, I thought I was leaving that work behind. I wasn't.

A deal meeting is a controlled experiment in group decision-making under uncertainty. Bankers are some of the best-trained decision-makers in the world, and they are also — in my very limited observation — aware of almost none of their own biases. This is not a criticism. It is a function of the speed at which the work is done. You do not have time to be your own psychologist when you have a committee at 4 p.m.

So the bias lives in the artifacts. In the WACC. In the "base case" that is almost always the expected case and almost never the base. In the sensitivity tables that run one variable at a time because the joint distribution is too expensive to compute. In the memo template, which privileges certain narrative structures over others.

Two things I notice now

First: the "base case" problem. I have sat in at least a dozen deal meetings where the base case was defended as what we really think will happen. It almost never is. The base case is usually the case that makes the deal work. The real base case — what the team actually thinks, after two glasses of wine on Friday — sits somewhere between the base and the downside. Nobody calls it that because nobody wants to be the one to reframe the memo.

Second: the premature precision problem. If a model has a hundred inputs and three of them are driving 80% of the outcome, the other 97 are decoration. Behavioral scientists call it conjunction error — adding detail to make a scenario feel more likely when it is actually less. Models are conjunction error engines. This is not the model's fault. It is the reader's attention.

What I try to do

I have one small practice I stole from a faculty member at DKU. When I finish a memo, I write a second memo — one paragraph — called What I would want to know if I hadn't built this. It forces me to read the first memo as a stranger would. Half the time, the things I'd want to know aren't in the model. They're in my head. The practice asks me to write them down.

It's not a cure. It is a counterweight. A psychologist in a room of bankers gets one counterweight per meeting, maybe two, if the meeting runs long. You learn which ones matter.