Objective research to aid investing decisions

Value Investing Strategy (Strategy Overview)

Allocations for February 2020 (Final)

Momentum Investing Strategy (Strategy Overview)

Allocations for February 2020 (Final)
1st ETF 2nd ETF 3rd ETF

Best Way to Implement Volatility Weighting?

| | Posted in: Volatility Effects

What volatility weighting scheme best exploits equity return volatility persistence based on net outcome? In the June 2015 version of his paper entitled “Dynamic Volatility Weighting in the Presence of Transaction Costs”, Valeriy Zakamulin examines a volatility weighting strategy with features that allow suppression of rebalancing frictions. The idea behind volatility weighting is to construct a portfolio that targets a specified (benchmark) volatility based on predictability (persistence) of asset volatility. Specifically, he compares three strategies:

  1. The theoretically (frictionless and with perfectly predictable asset volatility) optimal strategy, which weights an asset according to the ratio of benchmark variance (square of standard deviation of returns) to predicted asset variance.
  2. An optimized modified volatility weighting strategy, which includes two parameters to suppress trading: (1) a tuning parameter to control the aggressiveness of response to a change in predicted asset volatility; and, (2) a no-transaction buffer around targeted asset weight.
  3. Conventional volatility targeting, which weights an asset according to the ratio of benchmark volatility (standard deviation of returns) to predicted asset volatility.

For all three strategies, he sets benchmark volatility at an annualized 20%. He forecasts annual asset volatility from an exponentially weighted moving average of daily returns over a rolling window of the past year. He considers daily, 5-day and 21-day volatility forecast revision frequencies. He considers four levels of trading frictions (0.0%, 0.1%, 0.25% and 0.5%) and optimizes modified strategy tuning and buffer parameters for each level. He employs the six Fama-French portfolios formed on size and
book-to-market ratio as test assets. Using daily returns for these six style series and for the aggregate U.S. stock market during January 1989 through December 2014, he finds that:

  • Volatility persistence is strongest for portfolios of small-capitalization stocks.
  • Across the range of trading frictions, the optimized modified strategy generally outperforms the ideal and conventional strategies.
  • Daily rebalancing generates the best net performance across levels of trading friction, and the fall-off in performance from daily to 5-day and 21-day rebalancing is substantial.
  • Net outperformance of the optimized modified strategy relative to conventional volatility targeting increases considerably with higher trading frictions.
  • The theoretically ideal volatility weighting strategy performs poorly when trading frictions are high.
  • The optimal modified strategy outperforms a passive (buy-and-hold) benchmark even with high trading frictions.

In summary, evidence suggests potential for improving net performance of volatility targeting strategies via rules that suppress the level of trading involved.

Cautions regarding findings include:

  • There may be data snooping bias in selection of the decay factor and the look-back interval for the exponentially weighted moving average used for volatility forecasting.
  • Optimization of the buffer and tuning parameters used to dampen modified volatility strategy trading also impounds snooping bias. In other words, findings may be sample-specific.
  • The test assets are essentially indexes that are frequently reformed. Construction of tracking funds for these indexes would involve reformation frictions that would reduce series returns.
  • Implementation of the optimal modified volatility weighting strategy may be beyond the reach of many investors, who would bear fees for delegating execution to an investment manager/fund.

See also “Achieving a Low-volatility Stock Portfolio Efficiently”.

Daily Email Updates
Filter Research
  • Research Categories (select one or more)