In our last Factor Spotlight, we delved into the recent behavior of beta to understand why this measure has been less useful in helping investors hedge market risk. We highlighted the apparent de-coupling of beta from the general market based on the decreasing correlation of the Beta factor return to the Market factor return using US and global equity risk models from Axioma and MSCI Barra.
This trend has been underway for some time, but has steepened over the past 3 years and has gone into a particularly sharp decline since the March 2020 COVID downturn.
This week, we start to dive into market trends that are potentially pushing beta to its demise.
Context Around Beta
Most market investors know and understand beta like the back of their hands. Conceptually, beta is intended to explain a stock’s price movement relative to the broad market. If a stock has a beta of 1.5 and the market moves by 10%, we would expect the stock to move by 15%.
However, it may not be as straightforward as we think.
As you can see from the interpretation of beta, the measure is only meaningful in the context of the universe used for the calculation. A key assumption here is that beta is actually calculated relative to a broad universe.
But, what happens when the benchmark or universe used is no longer representative of the of the broad investable market?
Concentration Could be Throwing off Beta
It’s no secret that common benchmarks, which are market cap-weighted, are becoming increasingly concentrated as the FAANGs of the world continue on their trillion+ dollar trajectories. Currently, the top 10 names in the Russell 3000 (proxied using IWV) and the MSCI All Country World (proxied using ACWI) make up over 22% and 14%, respectively, of the benchmark by weight.
Given beta is being measured relative to these highly concentrated universes, it’s possible that the standard measure of beta may become misleading as a measure of general market risk.
Looking at market cap-weighted universe concentration over time shows a trend that appears to coincide with the breakdown of beta vs market correlation.
Our measure of concentration is the effective number of assets in the universe, calculated as the inverse of the Herfindahl-Hirschman Index (sum of squared security weights). A lower effective number of assets means the universe is becoming more highly concentrated, whereas a higher number means the universe is becoming more diversified.
As the charts above show, the effective number of assets in the US and global universes has declined rapidly, especially since 2016.
For the US universe, the effective number of assets peaked in June 2014 with 249 assets and hit a trough in August 2020 with 90 assets, which represents over a 60% decline.
For the global universe, this peak to trough decline was over 50% (428 effective assets in Sept 2013 to 207 effective assets in August 2020).
Like the sharp decline in beta vs market correlation over the past 3 years, especially in 2020, we also see the effective number of assets have declined sharply over this period for both universes.
Is There a Better Way to Measure Market Risk?
Intuitively, the heavy concentration in today’s market universes could be a good explanation for the general feeling that beta is “broken”. Though the effective number of assets does appear to be on a slight rise again, it may be risky for investors to continue relying on standard calculations of beta as the best tool to measure and hedge market risk. Perhaps, reframing how we view and calculate market risk could be a key to restoring beta’s usefulness for investors.
In next week’s Factor Spotlight, we will continue to investigate the dynamics of beta and look for ways to measure market risk in a more holistic manner.
US & Global Market Summary
US Market: 05/31/21 - 06/04/21
- The market crept higher in a truncated holiday week, with the S&P 500 hitting a new intraday high at the end of the week. May’s payroll report came in below expectations, giving investors confidence that the Fed wouldn’t be trimming its bond purchases in the near term.
- Job creation in May doubled April’s disappointing numbers, but still missed growing by 559k workers (vs. the consensus estimate of 650k). The unemployment rate fell to 5.8%, the first time it has been below 6% since the beginning of the pandemic.
- The US Consumer Price Index was up +0.8% vs. consensus expectation of +0.2%, and +4.2% YoY in April, as it becomes increasingly harder to find a good or service that hasn’t seen a price increase in recent months.
- On Saturday, the G7 reached a landmark agreement to back a global minimum corporate tax rate of 15%.
Normalized Factor Returns: Axioma US Equity Risk Model (AXUS4-MH)
- Volatility continued its strong multi-week move, up 0.62 standard deviations as it officially became an Overbought factor at exactly 1 SD above the mean.
- Market Sensitivity also saw a substantial move back towards the mean as it exited Oversold space.
- Value had appeared to be heading back to the mean after bouncing off a recent trough of -1.76 SD below the mean on 5/4. It’s since taken a bit of a U-turn and is now drifting back deeper into Oversold territory.
- Momentum drifted lower, shedding its Overbought label as it now sits at +0.9 SD above the mean.
- Weakness in Earnings Yield continued, as it dug deeper into Oversold territory at -1.39 SD below the mean.
- Profitability continued its sharp decline, as it garnered an Oversold designation at -1.04 SD below the mean.
- US Total Risk (using the Russell 3000 as proxy) increased by 12 basis points this week.
Normalized Factor Returns: Axioma Worldwide Equity Risk Model (AXWW4-MH)
- Similar to the US, global Volatility saw a substantial jump into Overbought territory at +1.02 SD above the mean.
- Market Sensitivity also saw strong gains, up +0.34 standard deviations as it headed back towards the mean.
- Exchange Rate Sensitivity headed higher towards Overbought space.
- Earnings Yield continued to fall, approaching an Extremely Overbought label at -1.71 SD below the mean.
- Profitability was the biggest loser abroad as well, tumbling by 0.82 standard deviations into negative normalized space.
- Global Risk (using the ACWI as proxy) increased by 7bps.