Vega Bucketing And Term Structure Hedging

Total vega tells you how a book responds if implied volatility across every single expiry moves by the same amount, all at once, in the same direction. Real volatility surfaces almost never move that way — a surprise earnings announcement spikes the volatility of options expiring next week while barely touching options expiring next year, and a shift in the macro outlook can do the reverse, moving long-dated volatility while short-dated options stay put. A book can be "vega neutral" in total and still be sitting on a large, undiagnosed bet on exactly this kind of relative move.

One thermostat versus a room with several radiators

Imagine a house with a single, whole-house thermostat reading an average temperature across five rooms, but each room actually has its own radiator that can run hot or cold independently. The average can read "comfortable" while one room is freezing and another is roasting — because the single number threw away exactly the information you needed. Vega bucketing replaces the single whole-house thermostat with one thermostat per room: instead of one portfolio vega, the book's vega is split by expiry bucket (say, under 1 month, 1–3 months, 3–12 months, over a year), so a spike in one bucket cannot hide behind an offsetting position in another.

Bucketed vega

Rather than a single sum, vega bucketing computes

$$\text{Vega}_j \;=\; \sum_{i \,\in\, \text{bucket } j} n_i \times \text{vega}_i,$$

for each maturity bucket $j$ separately, treating an implied volatility move in one bucket as independent of the others. In plain English: group every position by how far out it expires, add up the vega within each group separately, and never let a large short-dated vega and a large long-dated vega cancel into a falsely reassuring "zero" — they are two different risks that happen to share a symbol.

Worked example 1: a book that looks flat but isn't

A book holds: long 100 contracts of a 1-month option, vega $0.05$ each, and short 20 contracts of a 1-year option, vega $0.25$ each.

Total vega: $100(0.05) + (-20)(0.25) = 5.0 - 5.0 = 0.0$. Perfectly vega neutral on the total.

Bucketed vega:

• Under-3-month bucket: $100(0.05) = +5.0$.
• 6–18-month bucket: $-20(0.25) = -5.0$.

The book is flat in total but carries $5.0 (in vega-point units) of long short-dated vega against $5.0 of short long-dated vega. If a headline event spikes short-dated implied vol by 5 points while long-dated vol is unmoved, the book gains 5.0 \times 5 = \25.0 on the short-dated bucket and nothing offsets it — a real \25 move from a position the total-vega number swore was riskless.

Worked example 2: pricing the term-structure risk directly

Suppose short-dated vol jumps 5 points and long-dated vol falls 1 point (a realistic "front spikes, back barely moves or eases" pattern around an earnings date). P&L from each bucket:

• Short-dated: +5.0 \times 5 = +\25.0$.
• Long-dated: -5.0 \times (-1) = +\5.0$.

Total: +\30.0$, even though the naive total-vega figure of exactly$0.0$ predicted zero sensitivity to any vol move whatsoever. The two numbers disagree because total vega assumes a parallel shift; the real move here was anything but parallel.

Look at how implied volatility varies across tenor in the surface below — the short end and the long end can sit at noticeably different levels and move somewhat independently. A single portfolio vega treats this entire surface as one number; bucketing respects that it isn't.

What this means in practice

Every options desk with more than a handful of positions monitors vega by bucket, not just in total, and typically hedges each bucket separately using options in roughly that expiry range, because the volatility surface's short end and long end genuinely do move somewhat independently — the correlation between them is high but nowhere near 1. This is the direct term-structure analogue of hedging by strike: aggregation across expiries is exactly as dangerous as aggregation across strikes if the market doesn't actually move in parallel.