Using Volume Bars to Address Volatility Time in Portfolio Strategies

Summary

The document asks whether volume-based bars can make fixed-parameter strategies more responsive to changing market activity. It starts from the observation that equal calendar periods may represent very different conditions, then suggests volume bars as a possible alternative if volume and volatility move together. The example is a time-series momentum strategy whose fixed lookback may span unlike regimes in different years.

The main obstacle raised is cross-asset alignment: the nth volume bar for one asset does not necessarily correspond to the nth bar for another. That makes it unclear how to construct comparable return series or a meaningful correlation matrix for portfolio construction. The note offers no solution or empirical test, so it serves as a research question rather than a demonstrated method. Its premise that volume tracks volatility is also an assumption that would need testing.

Key ideas

• Calendar periods can represent different levels of market activity across regimes.
• Volume-based bars are proposed as a way to measure market time by activity.
• The proposal assumes that trading volume and volatility are sufficiently related.
• Asynchronous bars across assets complicate return comparison and portfolio correlation estimates.
• The document poses the issue without presenting a tested alignment method.

Tags

# Portfolio construction with volume based bars


# Portfolio construction with volume based bars












In the book Efficiently Inefficient, Lasse Pedersen interviews Myron Scholes:

LHP: Why do spreads tend to widen during some periods of stress?

MS: Well, capital becomes more scarce, both physical capital and human capital, in the sense that there isn't enough time for intermediaries to understand what is happening in chaotic times. Finance is in volatility time, not calendar time.

This italicized observation is interesting to me because a lot of basic trading strategies use fixed parameters, for example his 12-month time series momentum in Chapter 12. If finance is in volatility time, 12 months in 2009 is not the same as 12 months in 2017.

A potential answer to this (assuming volume and volatility are highly correlated) is to use volume based bars instead of time based bars, for example: https://www.tradingtechnologies.com/blog/2012/10/23/constant-volume-bars-vs-time-bars/

For a systematic portfolio strategy though, this is challenging because the nth bars for different assets aren't comparable. For example, if you calculate volume based bars for SPY,QQQ,DIA, and then calculate a return series from these bars, I don't see a way to calculate a meaningful correlation matrix.

Is there a way of dealing with volatility time in a systematic context?

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.