Notes on being a psychologist in a room of bankers

Discount rates are financial inputs, but they also carry judgment.

A model may display 8%, 10%, or 12% as if the number arrived fully formed. It did not. Behind it sit estimates of business risk, leverage, market prices, taxes, and the opportunity cost of capital. Once a rate enters a workbook, however, its visual precision can make those choices feel more settled than they are.

That is where my training in behavioral science still follows me into finance.

What a WACC actually does

A weighted average cost of capital combines the required return on equity and the after-tax cost of debt using capital-structure weights. Under the usual assumptions, WACC is used to discount unlevered free cash flow to the firm: PV = Σ FCFF_t / (1 + WACC)^t.

It is not a universal discount rate. Cash flow to equity is generally discounted at the cost of equity. Debt cash flows use a rate consistent with their credit risk. A project whose risk differs from the company average may also need a project-specific rate or a framework such as adjusted present value.

The distinction matters because a familiar number can become an anchor. If a team inherits a WACC from a template, the valuation may look internally consistent while the rate, cash flow, and capital structure describe different risks.

Why the behavioral lens matters

Before Columbia, I studied behavioral science. The discipline asks a slow question: do the reasons people give match the decisions we can observe?

Finance gives that question a set of artifacts. Assumptions live in a model. A central case receives a label. A sensitivity table decides which variables deserve attention. A memo template determines which evidence appears first. None of these devices is inherently biased; all of them frame the decision.

The useful response is not to psychoanalyze a committee. It is to make the frame inspectable. Where did the anchor come from? What evidence would move it? Which uncertainty is represented by a scenario, and which has been left outside the model?

Two things I notice now

The first is the base-case problem. "Base" can mean an expected case, a management plan, a lender case, or simply the case displayed first. Those are not interchangeable. A model becomes easier to challenge when it states what the base case represents, identifies the evidence behind it, and shows a break-even or downside case alongside it.

The second is false precision. A model with a hundred inputs may still be governed by three or four drivers. The remaining detail can improve accuracy, but it can also hide which assumptions carry the conclusion. This is not the conjunction fallacy, and one-variable sensitivity tables do not capture correlation. The practical tools are driver analysis, coherent scenarios, and, where the inputs justify it, a simulation that models dependencies explicitly.

Precision is earned when the evidence supports it. More cells alone do not earn it.

A practical check

A useful practice after finishing a model or memo is to write a short decision audit:

• Which assumptions drive the result?
• What is the strongest fact against the current conclusion?
• What would an outside-view reference class suggest?
• Which uncertainty is missing because it was difficult to quantify?
• What new evidence would change the recommendation?

Then I read the original work as if I had not built it. The gaps are often not calculation errors. They are pieces of context that remained in the analyst's head instead of reaching the page.

Behavioral science does not replace valuation. It helps govern the judgment that valuation cannot remove.