Chapter 21 · Dividends — Chapter 5 Finale

When Stocks Leak Value

Dividends break the Martingale — here is how to fix the math so your models still work.
The dividend paradox

Everything built in Chapters 18–20 rests on one foundation: the discounted stock price e^(−rt)S(t) is a Martingale under the risk-neutral measure ℙ̃. When a stock pays a dividend, it literally transfers cash out of the company and into your account — making the stock price drop on the payment date. A falling stock cannot be a Martingale. The models break. This chapter explains the fix.

The dividend paradox — stock price drops on ex-dividend date, breaking the Martingale property A stock price path that trends upward as expected, then drops suddenly at the ex-dividend date, then resumes upward trend. The drop breaks the Martingale property. EX-DIV DATE Stock drops by dividend amount Expected path Dividend δ Price drops → not a Martingale → pricing formula breaks

The price drop at the ex-dividend date breaks the Martingale property. The fix: track total wealth (price + reinvested dividends) instead of price alone.

The resolution is a subtle but important shift in perspective. Stop tracking the stock price. Start tracking the wealth process — the total value of the stock plus all dividends, immediately reinvested back into more shares. The stock price leaks, but the reinvested wealth does not. The wealth process is the true Martingale.

1 · Continuous dividends — the leakage rate δ

For indices and ETFs — like NIFTY 50 or a mutual fund — dividends flow continuously at a known annual rate δ. Every instant, a tiny fraction δ·dt of the stock's value is paid out. The stock price itself drifts downward relative to a non-dividend stock, but the reinvested wealth — where every dividend payment instantly buys more shares — still grows at the risk-free rate r under ℙ̃.

Stock price alone (leaks)

Under ℙ̃:   dS = (r − δ)S dt + σS dW̃

The stock grows at r − δ, not r. It leaks δ per year. e^(−rt)S(t) is NOT a Martingale anymore.

Reinvested wealth (Martingale)

Let ψ(t) = e^(δt)·S(t) (reinvested account).

Under ℙ̃:   e^(−rt)ψ(t) IS a Martingale. The leakage is cancelled by the reinvested dividends.

Three paths — dividend stock price, no-dividend stock, and reinvested wealth — under risk-neutral measure Three curves over time. A no-dividend stock grows at rate r. A dividend stock grows at r minus delta, slower. But the reinvested wealth grows at r, matching the no-dividend stock. No-div stock (grows at r) Div stock (grows at r−δ) Reinvested wealth (Martingale) δ gap Reinvested wealth tracks the no-dividend path — it is the true Martingale

The dividend stock (red) drifts below the no-dividend path. But reinvesting every dividend immediately (gold dots) keeps total wealth on the Martingale track.

V(0) = S(0)e^(−δT)·N(d₊) − K·e^(−rT)·N(d₋)
d₊ = [log(S(0)/K) + (r − δ + ½σ²)T] / (σ√T)
d₋ = d₊ − σ√T

Identical to the standard BSM formula — except the current stock price S(0) is discounted by e^(−δT).
This captures the "leakage": the stock will be worth less by the time the option expires.

Economically: S(0)e^(−δT) is the "forward price" of the stock adjusted for dividends — what you would expect to receive at time T if you started with S(0) today and leaked δ per year. Plugging this into BSM instead of S(0) is all that is needed to correctly price options on dividend-paying stocks.

2 · Lump sum dividends — the quarterly jump

Most real companies pay dividends quarterly, not continuously. On the ex-dividend date, the stock price falls by the dividend amount — a discrete jump. This looks alarming on a chart, but your total wealth does not change: the price drop in your shares is exactly offset by the dividend cash landing in your account.

📋 The Lump Dividend in Numbers
Before ex-div: S = ₹100 You hold 10 shares = ₹1,000 wealth
Dividend paid: a = 2% ₹2 per share paid to you in cash
After ex-div: S = ₹98 + ₹20 cash received = ₹1,000 total ✓
Wealth jump: ZERO Portfolio value unchanged — risk-neutral pricing still valid
Lump dividend — stock price jumps down at ex-dividend date but portfolio wealth is unchanged Two panels. Left shows stock price path dropping sharply at the ex-dividend date. Right shows portfolio wealth path staying smooth through the same date, because the cash received offsets the price drop. Stock price — jumps down Portfolio wealth — smooth Ex-div a·S drop Price "leaks" the dividend Ex-div Wealth unchanged — Martingale holds

The stock price jumps down (left). The wealth account stays smooth (right) — because the cash received exactly offsets the price drop. The risk-neutral formula remains valid on the wealth process.

The mathematical statement: at time tⱼ, the stock price jumps to S(tⱼ) = S(tⱼ⁻)·(1 − aⱼ) where aⱼ is the fractional dividend. But the portfolio value satisfies X(tⱼ) = X(tⱼ⁻) — no jump. The Δ(tⱼ) shares each lose aⱼ·S in price, but the Δ(tⱼ)·aⱼ·S dividend payment exactly compensates. Portfolio continuity is maintained throughout.

3 · Pricing options with lump dividends — the forward price

If you are pricing an NSE option and know that dividends a₁, a₂, … will be paid before expiry, you cannot simply use the current stock price S(0) in the BSM formula — the option will be mispriced. The correct approach: compute the dividend-adjusted forward price and use that as the effective stock price input.

Adjusted Price = S(0) × (1 − a₁) × (1 − a₂) × … × (1 − aₙ)
Multiply by all "leftover fractions" after each dividend is paid.
Then use this adjusted price in the standard BSM formula in place of S(0).

Why? Each dividend shrinks the effective price the option holder receives by (1 − aⱼ). The option is priced on what remains after all the leakage.
Dividend timeline — three quarterly dividends reduce the effective stock price before option expiry A horizontal timeline showing an option with three quarterly dividend payments at months 3, 6, and 9 before the option expires at month 12. Each dividend reduces the effective stock price. Today S(0) Div a₁=2% Month 3 ×(1−0.02) Div a₂=1.5% Month 6 ×(1−0.015) Div a₃=2% Month 9 ×(1−0.020) Expiry T Month 12 S_adj = S(0) × 0.98 × 0.985 × 0.98 = S(0) × 0.9463

Three quarterly dividends reduce the effective stock price to 94.63% of S(0) by expiry. Use this adjusted price in BSM — not the raw current price.

The backtesting warning: On the ex-dividend date, a stock that pays ₹2 on a ₹100 share will show as a 2% drop in the price series. If your algorithm treats this as a price crash and generates a "sell" signal — or if your mean-reversion Z-score treats this as a deviation — you are trading on an artefact, not a signal. Always use dividend-adjusted price series for backtesting. Most data vendors (NSE, Bloomberg, Refinitiv) provide adjusted close prices that account for all historical dividends.

4 · The three dividend regimes — side by side
FeatureNo DividendsContinuous δLump aⱼ
Risk-neutral drift of S r r − δ r (between jumps)
What is the Martingale? e^(−rt)S(t) e^(−rt)·e^(δt)·S(t) e^(−rt)·X(t) (wealth)
BSM input change None — use S(0) Use S(0)·e^(−δT) Use S(0)·∏(1−aⱼ)
Call option impact Baseline price Lower (leakage reduces S) Lower (jumps reduce S)
Ex-dividend date visible? N/A No — continuous drain Yes — discrete price jump
Backtesting adjustment None needed Use total return index Use adjusted close prices
Try it — compare dividend vs no-dividend option prices

Price a European call option with and without dividends. See how continuous leakage or lump payments reduce the option value. Observe that the put price rises when dividends are present — because dividends make it more likely the stock will fall below the strike.

₹100
₹100
20%
5%
3%
Adjusted S (effective)
Call price
Put price
vs No-dividend call

Green = call price across stock prices  |  Purple = put price  |  Dashed = no-dividend baseline  |  Dividends shift both curves downward for calls, upward for puts

Chapter 21 — dividends and the wealth process:

Chapter 5 complete. You have now covered the full arc of Risk-Neutral Pricing: Girsanov's Theorem, the Martingale Representation Theorem, the Two Fundamental Theorems, and dividends. The next chapter connects this probabilistic machinery back to Partial Differential Equations — showing that the two approaches (PDE and expectation) are two sides of the same coin.