The Greeks Are an Operating Budget
Five numbers, taught as calculus and spent as budget lines. Each one commits a different resource — trading bandwidth, capital, carry, premium, balance sheet — and in most programs two of them end up with no owner at all.
Every explanation of the Greeks is a calculus lesson. Delta is the first derivative, Gamma is the second, here is the Black-Scholes formula, here is a chart of an option's value curving away from its payoff.
All correct, and almost useless for running a hedging program, because it describes what the numbers are rather than what each one costs you to carry. A carrier that has sold equity-linked guarantees is not solving a differential equation. It is deciding how many people watch a screen, how much premium it is willing to spend, how much slippage it will tolerate in a bad quarter, and which committee owns the answer when something moves.
Read as an operating budget, the five stop being a list and start being five separate decisions with five different owners.
Delta: trading bandwidth, and the number everyone gets slightly wrong
Delta is the hedge ratio. If the liability behaves like 0.65 units of the index, you hold 0.65 units of futures against it and you are flat to small moves. It is the only Greek most programs hedge continuously, and what it costs is operational bandwidth: somebody watching, somebody authorised to trade inside the window, and a reconciliation that closes daily rather than at T+1.
There is also a piece of received wisdom about Delta that is wrong, and wrong in a way that matters specifically to insurers.
Delta is routinely described as the probability the option finishes in the money. It is not. Delta is N(d1). The risk-neutral probability of finishing in the money is N(d2), and the two are different numbers.
They are close enough to pass unnoticed on a one-month at-the-money equity option, which is where the shorthand gets taught. On a slightly in-the-money index call struck at 96.7% of spot, one year out, at 20% volatility, Delta is 0.6514 and the probability of finishing in the money is 0.5856 — a gap of nearly seven and a half points. Stretch that to the tenors insurers actually write and the two diverge much further.
The shorthand is harmless where it is taught and misleading where insurers live. Long-dated, high-volatility guarantees are precisely the region where Delta and the probability of exercise stop resembling each other.
If a paper anywhere in your file describes Delta as a probability, it is not a typo. It is a sign that the person who wrote it was working from a summary rather than from the model.
Gamma: two problems, and one of them is units
Gamma is the rate at which Delta changes, which makes it the number that decides how often the Delta hedge is wrong between rebalances. It costs capital or premium — either you hold the convexity by owning options, or you absorb the losses that come from chasing a moving Delta.
The first problem with Gamma is presentational and it defeats more risk limits than it should. Gamma scales as 1/S. On an index trading near 7,700, a plain vanilla Gamma reads 0.00023. Put that in a limit table and nobody can tell a comfortable number from an alarming one; it looks like rounding error, and limits nobody can read are limits nobody enforces.
Restated in units a human can act on, the same figure says: a 1% move in the index shifts Delta by about 0.018. That is a sentence an operations lead can hold a tolerance against. Express Gamma per 1% move, or per 100 index points, or in contracts of rebalancing per 1% — anything but the raw second derivative.
Where the call spread turns around
The second problem is real rather than presentational, and it belongs to anyone hedging a capped index crediting strategy.
A capped FIA credit is economically a call spread: long a call at the participation strike, short a call at the cap. The carrier that sold it is short that spread. Its net Gamma is not merely small — it changes sign inside the spread.
Here is a 100/110 spread, one year, 20% volatility, priced across the index level:
| Index level | Spread value | Net Delta | Net Gamma |
|---|---|---|---|
| 80 | 0.93 | 0.1020 | +0.00691 |
| 90 | 2.28 | 0.1629 | +0.00457 |
| 95 | 3.14 | 0.1802 | +0.00228 |
| 100 | 4.06 | 0.1855 | −0.00011 |
| 105 | 4.98 | 0.1797 | −0.00215 |
| 110 | 5.84 | 0.1650 | −0.00359 |
| 120 | 7.28 | 0.1221 | −0.00461 |
Long the spread. A carrier that has sold the crediting strategy is short it — flip every sign.
Below the crossing point the position behaves one way; above it, the correction runs the other way. A rebalancing rule written as "buy into strength, sell into weakness" is right on one side of that level and backwards on the other. This is what practitioners mean when they call call-spread Gamma explosive, and it is the single most useful thing to understand about hedging a capped credit.
It gets one degree worse, in a way worth writing into the procedure. The crossing point is not fixed. It sits at
S* = sqrt(K1 * K2) * exp( -(r - q + sigma^2 / 2) * T )which for the spread above is 99.8 with a year to run, 103.6 at three months, and 104.4 with a month left — drifting upward toward the geometric mean of the two strikes, 104.9, as expiry approaches. The level at which your hedge reverses direction is itself moving, and it moves fastest in the final weeks.
A procedure that names a fixed level will be wrong by a few percent at exactly the point in the contract year when the exposure is largest.
Theta: the line item you either pay or collect
Theta is time decay, and it is the clearest budget item of the five because it is the one you settle in cash.
A carrier that hedges a crediting strategy by buying the matching options has paid its Theta up front, in premium, at a price known on day one. A carrier that hedges the same exposure dynamically with futures has not paid it — it collects the decay on the liability instead, and pays it back later in rebalancing losses whenever realised volatility exceeds what was priced.
Those are the same economics arriving through different doors, and neither is more sophisticated than the other. What separates them is who bears the variance. The first is a fixed cost. The second is a variable cost that peaks in the quarter you can least afford it, because slippage is worst in exactly the conditions that made you want the hedge.
The failure is not choosing one. It is not knowing which one your product pricing assumed.
Vega: the one futures cannot touch
Delta and Gamma can be managed with futures. Vega cannot. Volatility exposure can only be offset with something that itself carries volatility exposure — options, or variance products — which means hedging Vega always costs premium, and premium is most expensive when you most want it.
A carrier that has sold guarantees is short volatility. That is not a view; it is what selling an option means. On the trade above, Vega is 3.3% of the option's premium per volatility point: a three-point spike in implied volatility adds roughly a tenth to the cost of the hedge.
The point worth carrying is that product design is a Vega decision, and it is usually made by people who are not thinking about Vega. A cap does not just cheapen the option — it truncates the payoff, so volatility beyond the cap is worth nothing and the position's sensitivity to it collapses. Monthly averaging does something similar for a different reason: an average is less volatile than a point, so an averaged credit is less sensitive to a volatility spike than a point-to-point credit on the same index.
Caps, participation rates and averaging are argued in product meetings as pricing levers. They are also, simultaneously, the largest Vega decisions the firm makes.
Rho: the orphan
Rho is the sensitivity of the option to interest rates, and it is the Greek most likely to belong to nobody.
The structure is familiar. The equity derivatives desk owns Delta, Gamma and Vega. ALM owns the interest rate exposure of the bond portfolio and the reserves. The interest rate sensitivity of the option sits between them, and on a short-dated hedge it is small enough that neither side is wrong to leave it.
On a long-dated guarantee it is not small. On the one-year trade above, Rho is 41.9 index points per percentage point of rate move — around 5% of the option's value. Lengthen the tenor and that grows roughly with time. When rates fall, the present value of a long-dated guarantee rises, and it rises at the same moment equity volatility is usually spiking, which is why a program that hedged equities carefully and rates not at all gets hit twice in the same quarter.
Rho does not need a dedicated hedge in every program. It needs an owner, and a line in the document naming who that is.
What actually goes in the document
All of the above reduces to five entries a hedging program should be able to produce, in language somebody could act on at the desk.
- Delta. The tolerance band, the rebalancing frequency, and who may trade inside the window without convening anybody.
- Gamma. The limit, stated per 1% index move rather than as a raw second derivative — and, for capped strategies, the acknowledgement that the sign reverses inside the spread and the reversal level drifts with time.
- Theta. Which of the two funding models the product was priced on, and the budget that follows from it.
- Vega. How much is hedged, how much is accepted, and the recognition that the cap and the averaging method already made part of this decision.
- Rho. An owner. Derivatives or ALM, named.
That is five short paragraphs. A program that has them can tell you, on the day something moves, whether what just happened was inside the plan. A program that has the Greeks in a monthly report and nowhere else has five numbers and no decisions.
The calculator
Every figure in this piece comes from a Black-Scholes-Merton workbook you can download and check. Five tabs, 134 live formulas, no macros, no protected sheets, nothing computed elsewhere and pasted in as a value.
It prices a vanilla call and put with all five Greeks, and a call spread with the net Delta and net Gamma table reproduced above. Participation and cap are entered as percentages of the strike, so you can set your own product terms and watch the sign flip move; leave the strike at 100 with participation 100% and cap 110% and it reproduces the table above exactly. It does not do Asian, barrier, digital or cliquet structures; those are path-dependent and need simulation, and a closed-form sheet that claims otherwise is lying to you.
It is on its own page: Option Greeks Calculator.
Scarborough Road works with insurers and asset managers on derivatives governance, hedging operations, and the investment data infrastructure that demonstrates the governance was followed.
General commentary on hedging governance practice. Not legal, actuarial or investment advice. Figures are illustrative outputs of a Black-Scholes-Merton model under stated assumptions — constant volatility, constant rates, continuous trading, lognormal returns — none of which hold exactly; they are not quotes, valuations or a representation of any market. Verify against your own pricing and risk systems, your own products and your domiciliary requirements. No specific insurer's program, systems, counterparties or limits are described.
Reference — Black, Scholes (1973); Merton (1973), extended for a continuous dividend yield. Illustrative parameters: S = 7,742.83, K = 7,489.72, r = 4.5%, q = 1.5%, sigma = 20%, T = 1 year; call spread K1 = 100, K2 = 110 on the same rates and volatility.
