Public Research - Does High Interest Rate Volatility Predict Market Turbulence?

A simple quantitative model to capture the predictive relationship between the volatility of the Short Term Interest Rate Index and the S&P 500 forward returns.

Target Index
S&P 500
Explanatory Variable
Cboe Short-Term Interest Rate Index

Introduction

There have been a great deal of studies assessing the stylized facts of Equity volatility:

  • The tendency of volatility regimes to persist
  • The higher volatility regimes' association with lower forward returns

We’re extending these by uncovering cross-asset lead-lag relationships that - anecdotically - have been a decision making factor in the discretionary trader’s arsenal.

Our focus here is on the predictive relationship between CBOE's Short Term Interest Rates Index (IRX) 5-to-200 day rolling standard deviation and the S&P 500 Index (SPX)'s 30 day forward returns.

Loading...

First, we split the data into in-sample and out-of-sample periods, with a cutoff date of 01-01-2018.

If we group the rolling 60 day standard deviation by deciles and aggregate the forward 30 day returns on the other axis (for the in-sample period), we see that higher level of IRX (Short Term Rate Index) volatility is associated with lower futures equities returns. To take advantage of this effect, we specify a threshold: if our rolling metric is above 0.3, we're 100% short the SPX index, below, we're 100% long.

We can also observe a similar effect if we take the rolling mean (instead of standard deviation) of the first difference of IRX.

Loading...

Performance

The presented backtest is constructed with 0.05% fix transaction costs (including slippage) and by "paper trading" on the the next day's close to avoid any lookahead bias.

Loading...

In-SampleOut-of-sample
Sharpe0.610.47
Alpha0.110.10
Beta0.14-0.07
CAGR (%)6.54%4.00%
Volatility (annualized)19.19%19.61%
Average Daily Turnover0.87%0.68%

Sign up for our newsletter

Get the latest public research & company updates

As a final exercise:

  • We diversify between the rolling 5, 10, 15, 40, 60, 80, 100, 120 standard deviation with equal weight
  • Plus, we diversify between the rolling 60, 80, 100, 120, 140, 160, 180, 200 mean with equal weight
  • We use a walk-forward methodology: creating a 1-depth Decision Tree model (imitating the "manual split" we described above) approximately every 3 years using all data up until the split, and exclusively use their out-of-sample “predictions”.
  • On the resulting portfolio, we apply volatility scaling on a rolling basis, where the current weights are scaled based on the portfolio's average volatility up to the day, divided by the recent 20 days’ past volatility.

For the diversified model with basic risk management, we get the following equity curve:

Loading...

The out-of-sample performance of the diversified, dynamic long-short strategy has an Alpha of 0.08, Sharpe Ratio of 0.61 with Beta of 0.15.

Conclusion

We believe predictive risk management models could greatly benefit from looking beyond usual suspects - the volatility of equity indices.

Can the predictive relationship presented here be considered independent / uncorrelated to be included in a model? The Spearman rank correlation between SPX's and IRX's rolling 60 day standard deviation is only around 0.3.

Our preconception was that when the shocks are sudden, like the COVID crash of 2020, the Federal Reserve's actions can only be lagging indicators, but with the inclusion of the rolling mean (capturing not just the volatility, but the direction of change) of the IRX index in the model, even the pandemic market turmoil could have been anticipated.

The famous quote by Markowitz "Diversification is the only free lunch in investing." applies not only to portfolio construction, but also to strategy creation.

The presented model serves as inspiration, bares no statistical significance (because of the low turnover), and should not be considered investment advice.

This effect was previously (at least indirectly) explored by:

Equity Return and Short-Term Interest Rate Volatility: Level Effects and Asymmetric Dynamics by Ólan et al.

Interest Rate Volatility and Stock Returns: A GARCH (1,1) Model by Latha et al.

Our research publication series started with the desire to share lesser-known, low-frequency, interesting, intuitive predictive relationships that we believe deserve more attention.

Learn more about our proprietary, mid-frequency research ↗

We are a boutique investment research firm bringing a unique perspective to the quantitative, systematic macro landscape.

Our offices

  • Berlin
    114A Friedrichstraße
    10117, Berlin, Germany
  • London
    145 City Rd
    London EC1V 1AZ, United Kingdom