You can’t see the true volatility risk premium on the day, because it depends on how volatile the next month turns out to be. The real-time stand-in is the VIX minus the S&P 500’s realized volatility over the past 21 sessions. From January 31, 1990 to October 8, 2026 (9,235 days), that real-time reading averaged 4.41 points, almost exactly the true premium’s 4.42, but the two agreed on sign only 80.04% of the time and their day-to-day correlation was 0.09.
Part I measured the premium after the fact. This part asks what you could have known at the close, what the VIX term structure adds, and what the two exchange-traded notes the original 2014 post was built around, XIV and VXX, actually were.
Two ways to measure the premium
Both measures start from the VIX close. The difference is which month of realized volatility gets subtracted: the one just finished, or the one about to start.
| Measure | Known on the day? | n | Average (pts) | Median (pts) | Std dev (pts) | % positive |
|---|---|---|---|---|---|---|
| VIX minus past 21-day realized vol | Yes | 9235 | 4.41 | 4.66 | 4.71 | 89.72 |
| VIX minus next 21-day realized vol | No (needs the future) | 9235 | 4.42 | 4.94 | 6.56 | 86.75 |
Over a long stretch the two land in the same place, which makes sense: past and future realized volatility have the same long-run average. Day to day they diverge. Past realized volatility is a fair forecast of the next month’s (correlation 0.655 in this sample), and the VIX is a slightly better one (0.718). But the premium is the small difference between two big, noisy numbers, and the noise in that difference barely lines up between the two versions.
The real-time measure was at or below zero on 10.28% of days (949), against 13.25% for the true premium.
Does a negative real-time reading warn of a negative premium?
Not in this data. When the real-time reading was most negative, the actual premium that followed was still positive more than 82% of the time.
| Real-time VRP (VIX minus past 21-day RV) | n days | Avg VIX | Avg actual VRP that followed | Median actual VRP | % actual VRP positive |
|---|---|---|---|---|---|
| Below -5 | 270 | 33.44 | 4.97 | 6.75 | 82.22 |
| -5 to 0 | 679 | 21.77 | 3.93 | 5.15 | 82.77 |
| 0 to 5 | 4063 | 17.03 | 3.71 | 4.28 | 87.45 |
| 5 to 10 | 3534 | 19.14 | 4.68 | 5.27 | 86.36 |
| 10 and up | 689 | 27.09 | 7.48 | 7.89 | 90.28 |
The reason is timing. A deeply negative real-time reading usually comes right after a shock, when the past month’s realized volatility has already exploded. By then the VIX is high (33.44 on average in the bottom row) and realized volatility tends to cool off over the following weeks. The negative premium the shock created is already in the past.
The top row is the one that lines up best: readings of 10 points or more were followed by the largest actual premium (7.48) and the highest hit rate (90.28%). The middle buckets are hard to tell apart. These are overlapping daily windows, and the two extreme rows each cover only a few hundred days.
What the VIX term structure adds
Cboe publishes the same kind of SPX-option volatility index at several horizons: VIX9D (9 days), VIX (30 days), VIX3M (3 months) and longer (Cboe term structure). Normally the curve slopes up. From January 4, 2011 to November 6, 2026 (3,985 days with all three), VIX9D was below the VIX and the VIX below VIX3M on 72.80% of days. The VIX above VIX3M, an inverted curve (backwardation, in futures language), is the classic stress reading.
| VIX / VIX3M | n days | % of days | Avg VIX | Avg real-time VRP | Avg actual VRP next 21 days | % actual VRP positive |
|---|---|---|---|---|---|---|
| Below 0.85 (steep contango) | 1230 | 28.67 | 14.58 | 4.25 | 3.56 | 84.88 |
| 0.85 to 0.95 | 2236 | 52.12 | 17.83 | 4.13 | 4.23 | 86.58 |
| 0.95 to 1.00 (flat) | 499 | 11.63 | 23.08 | 4.37 | 4.62 | 83.17 |
| Above 1.00 (backwardation) | 325 | 7.58 | 30.76 | 3.13 | 4.94 | 83.08 |
That covers September 18, 2009 to October 8, 2026 (4,290 days). Inverted days carried a bigger average premium afterwards (4.94 points), not a smaller one, while the share of positive outcomes stayed between 83% and 87% in every bucket. What the curve shape sorts is the level of fear, not whether the premium is about to fail. The steepest contango days had the smallest premium in points (3.56), because a calm market leaves little room between the VIX and realized volatility.
| Year | n days | Days VIX > VIX3M | % of days | Highest VIX/VIX3M |
|---|---|---|---|---|
| 2009 | 73 | 3 | 4.11 | 1.06 |
| 2010 | 252 | 22 | 8.73 | 1.24 |
| 2011 | 252 | 51 | 20.24 | 1.26 |
| 2012 | 250 | 1 | 0.40 | 1.04 |
| 2013 | 252 | 8 | 3.17 | 1.06 |
| 2014 | 252 | 19 | 7.54 | 1.11 |
| 2015 | 252 | 19 | 7.54 | 1.31 |
| 2016 | 252 | 15 | 5.95 | 1.08 |
| 2017 | 251 | 2 | 0.80 | 1.03 |
| 2018 | 251 | 53 | 21.12 | 1.33 |
| 2019 | 252 | 13 | 5.16 | 1.12 |
| 2020 | 253 | 53 | 20.95 | 1.34 |
| 2021 | 252 | 2 | 0.79 | 1.04 |
| 2022 | 251 | 15 | 5.98 | 1.02 |
| 2023 | 250 | 0 | 0.00 | 1.00 |
| 2024 | 252 | 9 | 3.57 | 1.14 |
| 2025 | 250 | 29 | 11.60 | 1.27 |
| 2026 | 214 | 12 | 5.61 | 1.10 |
2026 runs through November 6. The most inverted close in the file was February 28, 2020 (VIX 40.11, VIX3M 29.85, a ratio of 1.344). The latest sessions show how fast it flips: on November 4, 2026 the VIX closed at 16.61 against a VIX3M of 18.80; on November 5 it closed at 28.18 against 25.55, and on November 6 at 25.64 against 24.60, still inverted.
One limit: these are option-implied indices, not the VIX futures that VXX and XIV held. Each futures contract prices where the market expects the VIX to be at that contract’s expiration, so the two curves are related, but futures prices can sit well away from the index. This page doesn’t stitch the futures curve together, so the futures roll yield those notes earned or lost isn’t measured here.
What XIV and VXX were
Both were exchange-traded notes (bank-issued notes, not funds) whose value followed an index of short-dated VIX futures, and neither exists in its original form today.
- VXX, the iPath S&P 500 VIX Short-Term Futures ETN, was issued by Barclays. Bloomberg reported on December 13, 2018 that it would mature on January 30, 2019 under its rules, and that Barclays had launched a replica, VXXB (Bloomberg via Investing.com). Holding VXX was a long-volatility position.
- XIV, the VelocityShares Daily Inverse VIX Short-Term ETN, was issued by Credit Suisse and paid the daily inverse of the same kind of futures index: a short-volatility position. Its closing indicative value on February 2, 2018 was $108.3681. On February 5, 2018 the intraday indicative value fell to 20% or less of the prior day’s close, which under the note’s terms was an “acceleration event”; Credit Suisse said on February 6 that the last trading day would be February 20 and the notes would be redeemed on February 21 (Credit Suisse release, SEC). Reuters calculations cited by CNBC put the drop at $72.59 at 4 p.m. to $4.22 by 5:10 p.m. that evening.
The 2014 rules this post originally discussed switched between those two notes. Since both products are gone, any backtest built on them now leans on simulated pre-launch data and on prices that can’t be reproduced. The strategy test uses SPY and Cboe’s PutWrite index instead, and Part III looks at what short-volatility exposure did in 2018, 2020 and 2024.
What this means for reading the premium
The real-time measure gets the average right and the timing mostly wrong. It tells you roughly how much premium options are charging; it says little about whether the next month will be the one where that premium fails. An inverted VIX curve marks fear that is already here, not a reliable warning of more. None of this is a signal to trade on; it describes how these numbers have behaved since 1990. More studies on the Market Stats page, including VIX surges of 20%.
Every number above was calculated from the Cboe and S&P 500 daily data listed under Sources.
