Conditional Volatility and Information in ARCH Models

Summary

This exchange clarifies why volatility estimated from past returns can be called conditional. In ARCH terminology, conditional variance is variance given information available from the process’s history. That history matters when volatility clusters: recent observations can help estimate the volatility expected for the next period. A simple standard deviation over a sample of past returns instead summarizes variation across that sample and does not, by itself, condition the estimate on the latest information in an ARCH model’s sense.

The response suggests that an unconditional standard deviation might serve as an input or starting point for estimating conditional volatility. It does not lay out a specific estimator, derive the bias or inefficiency attributed to persistence, or explain the cited paper’s full context. The distinction is therefore conceptual, and the practical details would depend on the model and the information set used.

Key ideas

• ARCH conditional variance is defined relative to past information about the process.
• A sample standard deviation summarizes returns over its chosen sample window.
• Volatility clustering means recent history can inform estimates of future volatility.
• An unconditional volatility estimate may be used as an input to a conditional model.

Tags

# Conditional or unconditional volatility?


# Conditional or unconditional volatility?












I am reading a paper (reference below) that states "The conditional volatility for each underlying security (or for a market index) can be estimated using the standard deviation of the stock’s periodic returns. However, since volatilities are persistent, as we have learned from the ARCH literature, such an estimator of volatility will be biased and inefficient, as shown by Chou (1988)....."

Why is this a "conditional" volatility? To me it seems like you can't possibly get any more unconditional than taking a simple standard deviation over past returns.

"Investigating the Behavior of Idiosyncratic Volatility" by Xu and Malkiel in the Journal of Business (2003).

## Answer by GAM (score 4)

https://quant.stackexchange.com/a/7041

In terms of ARCH conditional variance is the variance conditional on past information (i.e. the history of the process). This is useful for modeling a process that exhibits volatility clustering. Perhaps he means that starting with the standard deviation (unconditional volatility) of stock returns one can then use that as an input to estimate the conditional volatility.

Edit: Some notes on ARCH models and conditional volatility

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.