Dividend discount model

FCFF is the right cash flow when you can meaningfully separate operations from financing. For most companies you can. For banks and insurers you can't. Their deposits and float are raw material, not financing. Your WACC build already refused them a WACC. "Operating cash flow" and "capex" barely exist as concepts for them. And regulatory capital requirements (CRAR/CET1, whose thresholds Phase 5 will drill) dictate how much of the earnings must be retained before anything can be paid out. What's left that you can observe, and that means something, is the cash that actually leaves for shareholders: dividends, plus buybacks. Hence the dividend discount model. It's the oldest valuation formula. For financials and other mature payers, it's still the right one:

Dividend discount model formula

Gordon DDM:   P₀ = D₁ / (r − g)        r = cost of EQUITY (never WACC — dividends belong
                                        to shareholders alone; M3.04's matching rule)

It's the same engine as the Gordon formula, with the same constraints: g is capped, and growth has to be funded. For a bank, g = ROE × (1 − payout), the retention identity, so you can't choose payout and growth independently. A bank growing its book 15% at 15% ROE can pay out nothing at all, and the mastery check tests this trap.

The retention identity has a diagnostic expansion you've seen before in the sustainable growth rate: g = P × R × A × T. That's the sustainable-growth rate written as net margin times retention times asset turnover times the assets-to-equity multiple. It matters here because a terminal g is a claim about all four, and the four move differently. A g that rises because retention rose is a payout decision management can reverse next quarter. A g that rises because turnover improved is an operating change that has to survive competition. A g that rises only because leverage rose isn't sustainable growth at all. It's the same earnings bought with more risk, and the cost of equity in your denominator should have moved too. When your terminal g changes between drafts, run it through the expansion before you defend it, and say which of the four letters moved.

How does the DDM value a bank?

The bank preview. The full machinery comes later, when you reach bank valuation proper. For now, divide the Gordon DDM by book value and substitute D₁ = B₀ × ROE × (1 − g/ROE):

P₀/B₀ = (ROE − g) / (r − g)

It's the most useful formula in bank valuation. Take a mature bank with ROE 14%, r 12% and g 4.5%: P/B = (14 − 4.5)/(12 − 4.5) ≈ 1.27×. When ROE = r, P/B = 1 exactly. A bank earning its cost of equity is worth book, the banking version of the ROIC = WACC fixed point. And when India's best private banks trade at 2.5 to 3× book, you can now read the message. The price asserts ROE ≫ r plus rapid growth, sustained for a long time. It's the same duration bet as the FMCG at 60×, wearing a banking license. One formula, and Phase 5's P/B-to-ROE sector lens suddenly makes sense.

How does the H-model work?

Multi-stage DDMs and the H-model. Real payers aren't in steady state today. Dividends might be growing 10 to 15% now and fading toward maturity. You can always build the explicit fade: project each dividend, discount it, and put a Gordon value on the tail, the same shape as the cement build. The H-model is the elegant shortcut for a linear fade. Growth starts at g_S and declines linearly to g_L over 2H years, where H is the fade's half-life in years, and then stays at g_L forever:

P₀ = [ D₀(1+g_L)  +  D₀ · H · (g_S − g_L) ] / (r − g_L)
      \_____________/    \________________/
      mature perpetuity   premium for the fade period

Here's the intuition for the second term. During the fade the dividend runs "hot" by (g_S − g_L)/2 on average, for 2H years. That's roughly an extra H × (g_S − g_L) of D₀-scaled dividends, capitalized at the mature rate. It's an approximation, good to a few percent when the fade really is gradual and H ≤ ~8. It overstates slightly, because it credits the average-growth path a touch generously, and it degrades when g_S is extreme or r − g_L is thin.

Worked example: an ITC-style Indian consumer conglomerate

Worked example: an ITC-style Indian consumer conglomerate. It's illustrative, at ~FY24/FY25 scale. Picture an FMCG-led business paying out ~80 to 90%, the classic Indian H-model specimen. Pull the actual AR. D₀ = ₹14/share. Near-term dividend growth g_S = 10%, tracking earnings. Mature g_L = 5%. Is that a whisker above the 4.5% cap? It is, and the cap applies here too, so the right input is g_L = 4.5%. We'll compute at 5% first on purpose and then show the cap's bite, because that's how you build the reflex. Cost of equity r = 10.2% (4.5% + 0.8 × 7.1%, from your WACC build). H = 5, which means a 10-year linear fade.

P₀ = [14 × 1.05 + 14 × 5 × (0.10 − 0.05)] / (0.102 − 0.05)
   = [14.70 + 3.50] / 0.052
   = 18.20 / 0.052  ≈  ₹350

Now cross-check it, because you never trust one formula. The exact two-stage, with the same linear fade computed year by year, gives ≈ ₹341, so the H-model is +2.6% high, the known mild overstatement. A crude two-stage (10% flat for 5 years, then a 5% Gordon) gives ≈ ₹350. That's a tight cluster, so call it ₹340 to 355. Now apply the cap. Rebuilt at g_L = 4.5%, the value drops about 7% to ≈ ₹324, as the denominator widens to 0.057 and the fade premium thins. Forget the exact rupee. The habit you want is to treat g_L as a terminal growth rate that obeys the g-cap like every other one. You also now own the reverse reflex. At a market price of ₹420, holding r and g_L, the implied g_S is ≈ 15% near-term dividend growth against ~10% delivered. So the price wants a step-up in earnings growth, or a cheaper required return, the bond-proxy bid for high-yield franchises. At a 3.3% dividend yield, you can now say what the extra ₹70 buys.

Where does the DDM apply?

Where DDM applies. It's mandatory for financials, since FCFF is undefined there. It works for mature high-payout franchises where dividends+buybacks ≈ FCFE (utilities, tobacco, big consumer staples), because there the DDM and an FCFE model converge, so use whichever is simpler. And it gives a floor estimate for cash-rich, reluctant payers. It misleads on low-payout compounders. Their dividend is a policy choice, not a measure of earning power, so a DDM on a 15%-payout compounder values the policy instead of the business. It also misleads on buyback-heavy US payers unless you count buybacks in D, either through a "shareholder yield" DDM or by switching to FCFE. Banks with buybacks take the same fix. The special cases complete the picture with the excess-return model, which is the DDM rearranged to run directly on ROE − r.