Assessing One-Day Spikes in Implied Volatility Data

Summary

The document raises a data-quality question about occasional one-day jumps in a 12-month at-the-money call implied-volatility series for a bank equity. The observed values reportedly rise sharply for a single day and then quickly return toward their prior levels. The central issue is whether these observations reflect genuine market information or errors in the downloaded series, and whether they should be removed before using the input in a model.

The only response suggests treating the observations as possible outliers and checking what happened on the dates in question before deciding. This is a cautious but incomplete recommendation: it offers no diagnostic procedure, event analysis, data-source comparison, or evidence that the spikes are glitches. The document therefore highlights a useful modeling decision—validate unusual observations before filtering—but does not resolve whether the particular observations are erroneous or informative. Removing them without investigation could discard real information, while retaining bad quotes could distort a model.

Key ideas

• A one-day implied-volatility jump may be a data anomaly or a genuine market event.
• Investigate the dates of unusual observations before deciding whether to exclude them.
• The document provides no evidence establishing the cause of the reported spikes.
• Filtering unexplained observations can remove either errors or useful market information.

Tags

# One-day spikes in implied volatility data


# One-day spikes in implied volatility data












I am building a model that takes the 12 month ATM call implied volatility as one of its inputs. I downloaded this implied vol time series data from Bloomberg for CM CN Equity (Canadian Imperial Bank of Commerce) for the last decade or so. However, I've noticed a strange pattern in the data. Occasionally, the implied volatility will double or triple for a single day, and then immediately subside.

In this time series graphs you can see a few of the sudden spikes that I am talking about around the middle, which last a single day. Is there a reason that one-day volatility spikes might be observed in the market, or is it likely a glitch in the data? Should I remove those spikes from a model, or would I be taking out useful information?

## Answer by Cristian Fernando Rodriguez (score -2)

https://quant.stackexchange.com/a/30522

If were you, probably i would take off these data. It seems to be outliers. In any case, check what could be happened in those days.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.