The number I went to Aswath Damodaran's page for was the risk-free rate, and it is there in plain words: "the US treasury rate of 4.75% used as the riskfree rate in US dollars". Net of a 0.22% default spread for the United States, the adjusted dollar rate is 4.53%. I wrote that down and was about to close the tab.
Below it sat the equity risk premium. Not one premium, five of them, all implied from the same market data on the same date.
Here they are, in the order I copied them: 4.14% on trailing twelve month earnings with adjusted payout, 4.09% on the trailing twelve month cash yield, 6.05% on the average cash yield of the last ten years, 3.80% on net cash yield and 3.56% on normalised earnings and payout.
The gap between the lowest and the highest is 249 basis points. Each figure is consistent with its own definition, so none of them is a mistake, and a model that cites "Damodaran, September 2026" could mean any of the five.
That is where I stopped reading and started calculating.
I used the standard build, where cost of equity equals the risk-free rate plus beta times the premium, and for beta I took Damodaran's figure for financial services outside banks and insurance, 0.97 across 176 firms.
At the lowest premium: 4.53% plus 0.97 times 3.56% gives 7.98%. At the highest: 4.53% plus 0.97 times 6.05% gives 10.40%. The spread shrinks a little, to 242 basis points, because the beta is under one.
Then I pushed both rates through a constant-growth perpetuity with 3% growth, which is my assumption and nobody else's. One over 7.98% minus 3% is a multiple of 20.07 times. One over 10.40% minus 3% is 13.52 times. Picking the top premium instead of the bottom one removes 32.6% of the value, and the business has not changed at all.
If you model fintech targets, this is the practical part, so I will put it here rather than at the end. Run every valuation at both ends of the published range, and write the chosen premium and beta next to the output. Two buyers using the same source on the same day can be a third apart and both be right. A buyer who will not show you those two inputs is hiding the one that moves the price most.
None of this would surprise Eugene Fama and Kenneth French. Their 2004 paper in the Journal of Economic Perspectives, "The Capital Asset Pricing Model: Theory and Evidence", opens fairly. They note that "four decades later, the CAPM is still widely used in applications, such as estimating the cost of equity capital for firms and evaluating the performance of managed portfolios."
Then the verdict: "Unfortunately, perhaps because of its simplicity, the empirical record of the model is poor - poor enough to invalidate the way it is used in applications."
I had assumed there was an escape route in the argument that the tests, not the model, were flawed. They close that route too. If bad proxies for the market portfolio spoil the tests, they write, "it also invalidates most applications, which typically borrow the market proxies used in empirical tests." I read that sentence and felt slightly annoyed on behalf of every valuation model I have seen with a cost of equity to two decimal places.
For what a discount rate looks like in dollars, the cleanest case I know is FIS and its merchant business. The company's 2023 annual report says that "in 2022, we recorded a goodwill impairment charge of $17.6 billion related to the held-for-sale reporting unit, reflecting our intermediate-term growth expectations."
A second charge came once a price existed: "in the second quarter of 2023, we recorded an additional $6.8 billion goodwill impairment, reflective of the price at which we agreed to sell a majority interest in the Worldpay Merchant Solutions business." A $1.9 billion valuation allowance followed on the assets held for sale. GTCR bought 55% at an $18.5 billion enterprise value, including $1.0 billion contingent, closing on 31 January 2024. FIS took more than $12 billion in net cash and kept 45%.
The cash flows of the business did not need to change for any of that to happen. What changed was the rate a buyer was willing to apply to them.
The result holds when I move the beta. At 0.76, the published figure for money centre banks, the multiples are 23.61 and 16.32 times, a fall of 30.9%. At 1.5 they are 14.56 and 9.43 times, a fall of 35.2%.
Growth moves it too. At 2% growth the fall is 28.8%, and at 4% it is 37.7%.
A single-stage perpetuity is a crude tool, and I would not value a real fintech without a transition period. I used it only to isolate the one variable.
The top figure comes from a ten-year average of cash yields. Choosing it is a claim that today's market will drift back toward the last decade, and that is a view about regimes, not an arithmetic error. Choosing 3.56% is the opposite view.
So the 32.6% gap is the price of a judgement. I have looked for a published test that tells you which regime applies to a specific company and I have not found one. I still cannot say which of the five I would defend in front of a board.