Theta and Rho

Motivation: why this matters in quant finance

Theta (

\Theta

) measures how an option's value decays as time passes. Rho (

\rho

) measures how it responds to interest-rate changes. Together with delta, gamma, and vega, they complete the standard sensitivity toolkit.

The two Greeks live on very different scales of practical importance:

Theta is unavoidable. Every long option position is losing money to time decay every day the underlying doesn't move. A long gamma book earns convexity by rebalancing its delta — theta is what it pays for that convexity. On a delta-hedged portfolio, the daily P&L is approximately $\tfrac{1}{2}\Gamma(\Delta S)^2 + \Theta\,\Delta t$ , so the market-maker's entire business is the sign of this expression.
Rho is usually ignored. For short-dated equity options, rho contributes a tiny fraction of total P&L — rates move slowly and the discount factor $e^{-rT}$ is near $1$ . But for long-dated options, swaptions, and interest-rate derivatives, rho is the headline Greek. Fixed-income desks report risk by rho (or DV01) above all else.

This note pairs the two because the Black-Scholes derivations look parallel (both are partial derivatives of the same pricing formula) but their uses divide cleanly by product type and tenor.

The informal idea

Theta answers "how much do I lose tomorrow if nothing else changes?" — pure time decay. Because it measures change per day or per year, conventions matter: a theta of

-0.05

in per-year units means the option loses about

5

cents per year, which is about

0.014

cents per day, not

5

cents per day.

Rho answers "how much does the option value move for a

1\%

parallel shift in the flat risk-free rate?" The key word is flat: rho treats the yield curve as a single number, so it aggregates away all the curve structure that an interest-rate desk actually cares about. This is why rate-sensitive products use curve-based measures (DV01, key-rate durations) instead.

Formal definitions

For an option value $V(S, t; r, \sigma, \ldots)$ , the Greeks are partial derivatives:

\Theta := \frac{\partial V}{\partial t}, \qquad \rho := \frac{\partial V}{\partial r}.

Two sign conventions appear in practice:

Mathematical theta is $\partial V/\partial t$ , positive because $t$ increases as time passes and long options usually lose value per unit $t$ — so the formula yields negative numbers. Most textbooks use this.
Trader's theta is $-\partial V/\partial t$ , flipped so a negative number means "I lose money" matches intuition. Some systems use yet another convention: $\partial V/\partial \tau$ where $\tau = T - t$ (time to expiry), which has the opposite sign.

Always check the sign convention of whichever system you're reading from.

Black-Scholes formulas

Let $\tau := T - t$ be time to expiry. Under Black-Scholes with no dividends, a European call priced at

C(S, t) = S\,\Phi(d_1) - K e^{-r\tau}\Phi(d_2)

has

\Theta_{\text{call}} = -\frac{S\,\varphi(d_1)\,\sigma}{2\sqrt{\tau}} - r K e^{-r\tau}\Phi(d_2),

\rho_{\text{call}} = K \tau e^{-r\tau}\Phi(d_2).

For a European put,

\Theta_{\text{put}} = -\frac{S\,\varphi(d_1)\,\sigma}{2\sqrt{\tau}} + r K e^{-r\tau}\Phi(-d_2),

\rho_{\text{put}} = -K \tau e^{-r\tau}\Phi(-d_2).

Here

\varphi

is the standard normal density,

\Phi

is its CDF, and

d_1, d_2

are the usual Black-Scholes quantities. Both formulas follow from differentiating the Black-Scholes formula with respect to

t

r

; see the Black-Scholes derivation for the underlying pricing function.

Decomposition of theta

$\Theta_{\text{call}}$ has two pieces:

$-S\varphi(d_1)\sigma/(2\sqrt{\tau})$ — the gamma-driven decay. This term is always negative and scales with gamma; it is the premium the long-option holder pays for convexity.
$-rKe^{-r\tau}\Phi(d_2)$ — the discount-driven decay. This reflects that the strike $K$ , paid at expiry, is discounted less aggressively as $\tau$ shrinks.

For a deeply ITM European call on a non-dividend-paying stock at very low rates, the second term can dominate and

\Theta

can briefly become positive (the call becomes worth more as expiry approaches, because the discount-rate effect on the strike overwhelms the gamma cost). This is rare but instructive.

Key properties

Theta.

Sign. $\Theta \le 0$ for ordinary long vanilla options away from the edge cases described above. Short options have $\Theta \ge 0$ .
Peak near the money. $|\Theta|$ is largest near-the-money and near expiry, because gamma is largest there.
Square-root time scaling. The gamma-driven piece scales as $1/\sqrt{\tau}$ : theta accelerates into expiry. An option losing $2\%$ of its value per day a month out is losing $4\%$ per day a week out.

Rho.

Sign. Calls have $\rho > 0$ , puts have $\rho < 0$ . Higher rates raise the risk-neutral drift of $S$ , favouring calls and penalising puts.
Magnitude. $\rho \propto \tau$ , so rho scales linearly in time to expiry. Long-dated options have enormously more rho than short-dated ones.
Vanishing ATM approximation. For short-dated ATM options, the Brenner-Subrahmanyam-style approximation gives $\rho \approx K\tau e^{-r\tau}/2$ — small in absolute terms.

The theta-gamma relation (PDE consequence)

The Black-Scholes PDE states

\Theta + r S\Delta + \tfrac{1}{2}\sigma^2 S^2 \Gamma = r V.

So for a delta-hedged position worth zero ( $V=0$ , $\Delta=0$ ):

\Theta = -\tfrac{1}{2}\sigma^2 S^2 \Gamma.

Long gamma earns theta with opposite sign. Market-makers charge for convexity by selling options with negative theta; they earn it back when the underlying realises enough volatility that

\tfrac{1}{2}\Gamma(\Delta S)^2

over the rebalance period exceeds

|\Theta|\Delta t

. This is the realized-vs-implied volatility trade in its cleanest form.

Worked example — a one-month ATM call

Take $S = 100$ , $K = 100$ , $\sigma = 0.20$ , $r = 0.03$ , $\tau = 1/12$ .

Then $d_1 = (\ln(S/K) + (r + \sigma^2/2)\tau)/(\sigma\sqrt{\tau}) \approx 0.0866$ , $d_2 \approx 0.0289$ , $\varphi(d_1) \approx 0.3974$ , $\Phi(d_2) \approx 0.5115$ .

Theta (per year):

\Theta_{\text{call}} \approx -\frac{100 \cdot 0.3974 \cdot 0.20}{2 \sqrt{1/12}} - 0.03 \cdot 100 \cdot e^{-0.03/12} \cdot 0.5115 \approx -13.77 - 1.531 \approx -15.30.

Per calendar day (÷365): $\Theta \approx -0.042$ . So the call, priced around $\$ 2.36 $, loses about$ $0.04 $per day — roughly$ 1.8%$ of its value daily with one month to expiry.

Rho (per unit $r$ , i.e. per 100%):

\rho_{\text{call}} \approx 100 \cdot (1/12) \cdot e^{-0.03/12} \cdot 0.5115 \approx 4.252.

For a 1 basis point move ( $\Delta r = 0.0001$ ): $\Delta V \approx 0.000425$ . Negligible compared with theta's daily $\$ 0.04$.

Common confusions and pitfalls

Annualised vs daily theta. A quote like "theta is $-0.05$ " is ambiguous. Always confirm whether the number is $\partial V/\partial t$ in per-year units (divide by $365$ or $252$ to get daily) or already a daily figure. Many risk systems report daily.
Theta can be positive. Deep ITM European puts on non-dividend stocks, and certain barrier options, can have positive theta. Don't hard-code "theta is always negative."
Weekend decay. Option prices in real markets reflect $\tau$ measured in calendar days, but pricing models and intraday traders often confuse calendar-day decay (weekends count) with trading-day decay (they don't). The right convention is product-specific.
Rho and cost-of-carry. For equity options with dividends, $\rho$ splits into a risk-free-rate sensitivity and a dividend-yield sensitivity (sometimes called "phi"). Currency and futures options have two rho-like Greeks — one per rate. Check which $r$ the formula is differentiating against.
Rho ≠ DV01. Rho shifts a flat rate; DV01 shifts the whole curve by one basis point. For interest-rate derivatives you almost never want rho.

Where this goes next

Delta Hedging and Hedging Error — makes the theta-gamma trade explicit.
Black-Scholes PDE — where the $\Theta + \tfrac12\sigma^2 S^2 \Gamma$ relation comes from.
Interest Rate Models — why fixed-income desks use DV01 and key rates instead of flat- $r$ rho.