You’ve built the model. You know what the number says. Now let’s talk about what to do when the market disagrees — and when the model itself might be sending you the wrong signal.
Six posts ago, we started with a single question: why does the standard free cash flow formula break when you apply it to a bank?
You now know the answer. You know why banks require FCFE instead of FCFF. You know how to use ROE and the retention ratio to project a bank’s growth. You know how to normalize an insurer’s earnings to strip out catastrophe-year distortions. You know how to build a discount rate using the CAPM formula with an industry beta. And in Parts 5 and 6, you watched the complete model run — first on a fictional bank, then on a real company in our investment screening universe.
This post closes the loop.
Every discipline that teaches you when to buy must also teach you when to sell. Without a clear exit framework, the margin of safety becomes a one-way door. You know how to walk in. This post teaches you when — and why — to walk out.
The Universal Signal Still Applies
Start with what hasn’t changed.
The sell signal built into every DCF-based holding in this system is the same regardless of what kind of company you’re valuing. When the market price exceeds your intrinsic value by 20% or more, the margin of safety has inverted. You’re no longer buying a dollar for 80 cents — you’re paying $1.20 for a dollar, and hoping the dollar turns out to be worth $1.50.
That logic applies to banks and insurers just as it does to manufacturers or retailers. The mechanism is identical. What changes is how you compute the intrinsic value itself — and more importantly, how confident you can be in that number.
For a standard industrial company, the FCFF model gives you a fairly direct read on value. Free cash flow is cash flow. Revenue minus costs minus taxes minus reinvestment. There’s noise in the inputs, but the structure is clean.
For a financial firm, the model has a few additional wrinkles — places where the structure of the business makes the IV estimate harder to trust than it looks. Understanding those wrinkles is the whole point of this post. Because a sell signal you can’t trust is worse than no signal at all.
The Excess Return Lens — What the Terminal Value Is Actually Assuming
Here’s a concept worth naming explicitly, because it’s been running underneath the math in every post in this series.
In Parts 2 and 5, we described what happens when we calculate a terminal value for a bank. The terminal phase assumes that the bank’s ROE — its return on equity, the percentage of profit it earns on shareholders’ capital — gradually converges toward its cost of equity over time.
Why does that convergence happen? Because above-cost returns attract competition. When a bank consistently earns 15% on equity while its cost of capital is 8%, that spread draws in competitors. Over time, in most markets, competition compresses ROE back toward the cost of capital. The model captures that reality in the terminal value by assuming a mature firm settles near, but slightly above, its cost of equity.
The gap between ROE and the cost of equity has a name: excess return. When ROE > cost of equity, the bank is earning above its cost of capital — every dollar it retains creates more than a dollar of value. That’s the excess return. It’s the financial-firm equivalent of what the Intrinsic Value series described as ROIC exceeding WACC.
When ROE > Cost of Equity: value is created. The bank is worth more than book value. The excess return is what justifies a premium.
When ROE < Cost of Equity: value is destroyed. Every retained dollar earns less than shareholders require. The bank is worth less than book value — and the model, if run correctly, will show that. The excess return has gone negative.
This is the sell precursor to watch. You don’t need to wait for price to hit 120% of IV. If you see ROE declining steadily toward — and then below — the cost of equity, the model’s intrinsic value is falling even if the stock price hasn’t moved yet. You’re in a shrinking business while the market hasn’t noticed. Getting out before the market adjusts is what the monitoring is for.
What does this look like in practice? Imagine First River Bank — our fictional regional bank from Part 5 — has been earning ROE around 10% with a cost of equity of 7.6%. That 2.4% excess return justifies a modest premium to book value. Now suppose competitive pressures in its lending market compress net interest margins over three years. ROE drops to 9%, then 8%, then 7.4% — just barely below cost of equity. At each step, the model’s IV falls. At the point where ROE drops below cost of equity, First River is no longer creating value for shareholders. Each retained dollar makes things marginally worse.
You wouldn’t necessarily have to sell the moment ROE dips below cost of equity for one quarter. A single bad quarter can reflect a timing issue. But if the trend is clear — ROE declining steadily, the spread narrowing and turning negative — that’s a structural signal. The business model has changed.
What’s Different — Regulatory Capital Deterioration
For banks, there’s a sell signal that doesn’t exist in most other industries: regulatory capital deterioration.
You’ll recall from Part 1 that banks are required to hold a minimum Tier 1 capital ratio — typically at least 6% in the United States, though most well-run banks voluntarily target 10–12% as a buffer. This minimum capital requirement is the reason FCFE is lower than reported earnings: before the bank can pay a dollar to shareholders, it must set aside enough equity capital to keep that ratio intact as its loan book grows.
When a bank’s Tier 1 capital ratio starts falling — toward the regulatory minimum rather than away from it — two things happen that directly affect your IV calculation:
First: The bank may need to retain more earnings than the model assumes. If capital is thin, management will prioritize rebuilding the buffer over returning cash to shareholders. Dividends may hold steady for a quarter or two (to avoid signaling distress), but buybacks typically slow or stop first. The actual FCFE available to shareholders falls.
Second: When a bank’s capital ratio approaches regulatory minimums, regulators have the authority to restrict capital distributions entirely. They can prohibit dividends and buybacks until the ratio recovers. If that happens, the IV the model computes — which depends on the payout continuing — overstates what shareholders will actually receive during the restriction period.
Neither of these is a crisis by itself. Banks manage through capital pressure regularly. But a declining capital ratio is a leading indicator. It tells you that the path from model-output IV to actual shareholder cash is getting longer and less certain.
What to watch: Tier 1 capital ratio and CET1 ratio (Common Equity Tier 1 — the most conservative measure). These are reported in every quarterly 10-Q and are prominently highlighted in bank earnings releases. A ratio trending steadily downward toward the regulatory minimum is a flag. It doesn’t automatically mean sell — but it means re-examine the growth inputs and payout assumptions in your model.
What’s Different — Credit Quality Deterioration
Banks are in the business of lending money. Which means they’re also in the business of losing money when borrowers don’t pay back their loans.
Normally, loan losses are small and predictable. The bank estimates them in advance, sets aside money in a loan loss reserve (sometimes called an allowance for credit losses), and charges losses against that reserve as they occur. When reserves are well-calibrated, this process is mostly invisible — it’s already baked into earnings.
The sell signal comes when it stops being invisible.
NPL ratio — Non-Performing Loan ratio — is the fraction of the loan portfolio where borrowers are behind on payments (typically 90+ days past due). A low, stable NPL ratio means the loan book is clean. A rising NPL ratio means more loans are going wrong.
When NPL ratios rise above historical norms, two things happen in the income statement. First, provision expense increases — the bank has to top up its reserves to cover the additional expected losses. That provision expense flows directly through the income statement, reducing net income. Lower net income → lower FCFE → lower IV.
Second, the bank’s historical ROE average — which the model uses to project future growth — becomes unreliable. The model looks backward to smooth the inputs. If credit losses in the recent past were low (good times) but NPL ratios are now rising (a credit deterioration cycle is beginning), the model is projecting a future that looks like the benign past. That’s optimistic in the wrong direction.
For insurers, the equivalent early warning is reserve development. You learned in Part 3 that insurers set aside reserves to cover future claims. When those reserves prove too low — when claims come in higher than expected — the insurer reports adverse prior-year development. A few quarters of adverse development isn’t unusual. A pattern of it is a signal: the company has been systematically underestimating its liabilities, and the income statement has been overstating its true earnings for years.
In both cases, the signal is: the historical earnings you’ve been projecting from aren’t as reliable as they look. Adjust the inputs, rerun the model, and check whether the margin of safety still exists.
The Payout Ratio Signal Distortion — The Model’s Most Important Blind Spot
This one is the most subtle of the three — and the one most likely to send you the wrong signal if you’re not paying attention.
You’ll remember from Part 6 that the FCFE model for banks treats dividends as the measure of cash returned to shareholders. The model sees FCFE as dividends paid, divided by shares outstanding. Buybacks — share repurchases, where the bank uses excess capital to buy back its own stock — don’t show up in the dividend line.
Part 6 introduced this as a known limitation when explaining the OFG valuation. It caused the model to produce a lower IV than a full-payout view would suggest. We flagged it and moved on.
But that same limitation creates a more dangerous problem when you’re monitoring a holding over time: it can generate false sell signals and false buy signals whenever a bank changes how it returns capital to shareholders.
Here’s how the false signal works.
Imagine a fictional bank — call it Ridgemont Bancorp — that has historically returned capital as follows: 20% payout ratio in dividends and 15% in buybacks, for a total of 35% of earnings returned to shareholders. Your FCFE model has been running on the 20% dividend figure. The intrinsic value it produces is lower than a full-return model would show, but it’s been consistent — you know the discount, and you’ve factored it in.
Now suppose Ridgemont’s management decides to shift strategy. They eliminate the regular dividend and convert the entire capital return to a share repurchase program. Total cash returned to shareholders stays at 35% of earnings — or even increases. But the dividend line goes to zero.
Your model now shows FCFE of zero. Intrinsic value collapses — not because the business deteriorated, but because the model can no longer see the capital being returned.
That’s a false sell signal. The business hasn’t changed. The payout hasn’t changed. Only the form of the payout has changed.
The inverse also happens. A bank that has been returning capital through buybacks announces a large special dividend. FCFE appears to spike. IV jumps. That looks like a buy signal — but it may just be a one-time distribution, not a new sustainable payout level. Running toward a spike in the dividend line can be just as misleading as running away from a drop.
What to do with this: Before acting on any IV-based sell signal for a financial firm, verify what drove the change. If the payout ratio has moved significantly — in either direction — check whether the underlying capital return to shareholders (dividends plus buybacks) has actually changed, or whether you’re watching a format shift inside the same cash flow.
This requires one step beyond the model: reading the earnings release and the 10-Q. When did buybacks increase? Did the dividend change? Is total capital return up, down, or unchanged? The model’s number changes when the format changes. The investment decision shouldn’t.
Putting It Together — A Hold/Sell Decision Framework
When your model flags a financial firm as potentially overvalued, or when a monitoring signal fires, here’s the sequence to run through before acting.
Step 1: Is the IV calculation reliable right now?
Check the payout ratio trend. Has the bank shifted between dividends and buybacks recently? If the payout ratio moved by more than a few percentage points since your last model run, check total capital return before trusting the IV change.
Step 2: Is the sell signal from price or from model deterioration?
There are two very different sell situations:
Price signal: The underlying business is performing as expected. The model’s IV is solid. The market has simply repriced the stock above 120% of IV. This is the margin-of-safety inversion — the same sell trigger that applies across every series in this course. Exit because the discipline says so, not because the business is bad.
Model signal: Something has changed in the inputs. ROE is declining. Capital ratios are thinning. NPL ratios are rising. The IV estimate itself is getting less reliable. Here, the response isn’t necessarily to sell immediately — it’s to update the model with the new inputs and see where IV lands. If the updated IV still shows overvaluation, the sell signal is stronger. If the updated IV reveals undervaluation that the old model was masking, the signal may be false.
Step 3: Is the original thesis intact?
Part 6 introduced the Two-Minute Story as a first check before running any numbers. Use the same test here. Can you still explain why you own this in two minutes? Does the story still hold?
The thesis for a bank typically rests on a few key pillars: a durable deposit franchise, a track record of disciplined underwriting, a competitive position in its lending market, and a history of returning capital to shareholders. When any of those pillars starts to crack — not just wobble for a quarter, but show a sustained, structural change — the thesis is breaking, and the model’s IV is running on assumptions that may no longer hold.
A broken thesis is a sell trigger independent of where the price sits relative to IV. If the story you underwrote has changed, hold the model output loosely and ask whether the assumptions inside it are still realistic.
When the Series Signals Align
The cleanest sell in a financial firm looks like this: price above 120% of IV, confirmed by a reliable model (no recent payout shifts, no buyback-to-dividend format changes), ROE still solid, capital ratios healthy, NPLs stable. The business is fine. The market has just priced it above fair value. Exit on discipline.
The messiest situation is the opposite: price approaching — but not yet above — IV, while NPL ratios are starting to rise, and the bank quietly reduced buybacks in the last two quarters. The model shows a small margin of safety. But the inputs are getting less reliable, and the early indicators are pointing in the wrong direction.
In a messy situation, the right response is not to wait for 120%. It’s to update the model with the more conservative assumptions the early indicators suggest, and recalculate. If the conservatively-adjusted IV puts you above the current price, the margin of safety may have already evaporated — even if the base-case model still shows a positive gap.
The tools this series built — FCFE, normalization, cost of equity, terminal value — were never meant to give you a single definitive number to mechanically compare against price. They were meant to give you a structure for asking better questions. The sell decision for a financial firm is a test of whether you’re actually using that structure, or just reading the output without reading the inputs.
A Note to Luca and Lili
Part 7. The last one.
Financial firms have been the trickiest section of this whole course to write — not because the math is harder, but because the places where the model can mislead you are less obvious than they are for an ordinary industrial business.
For Maple Ridge Manufacturing, from the Intrinsic Value series, the model is fairly direct. Cash flow is cash flow. The valuation is a function of what the business earns and how fast it grows. The inputs are messy in practice, but the structure is clean.
For a bank, the structure is a layer more complex. The “cash flow” the model sees is dividends — which may or may not represent the full return the bank is giving shareholders. The capital base is regulated, which means the government has a say in how much can be paid out. The loan book generates credit risk that doesn’t show up until it does — and when it does, it hits earnings hard. The model captures all of this imperfectly.
That imperfection is not a reason to avoid banks. Banks have been among the most reliably profitable enterprises in the history of capitalism. But it is a reason to hold the model output with more humility than you might for a simpler business — and to pay closer attention to the signals that live outside the model, in the footnotes of the quarterly filing and the trend in the capital ratios.
You’ve now got every tool this series can give you. The rest is practice.
— Papa
The One-Sentence Summary
For financial firms, the same universal sell trigger applies — exit when market price exceeds intrinsic value by 20% or more — but before trusting the model’s IV, verify that the underlying payout structure hasn’t shifted (buybacks replacing dividends can make IV collapse even when total capital return is unchanged), check that capital ratios and NPL trends haven’t deteriorated in ways that make the historical growth inputs unreliable, and watch for the excess return signal: when a bank’s ROE falls below its cost of equity and stays there, the model’s intrinsic value is falling whether the market has noticed yet or not.
← Part 6: A Real Bank, By the Numbers
This is the final post in the Valuing Financial Firms series. Explore all seven parts on the Valuing Financial Firms series page.
— Jim