Correlation Data Flashes Sell Signal

By in
Correlation Data Flashes Sell Signal

We’ve been focused on August’s historical penchant for financial asset price volatility in recent notes, and today’s iteration on the theme relates to US stock sector correlations. We have tracked this data every month since 2009, and here is why this seemingly esoteric topic matters:

  • Every sector has different fundamentals, so none trade in precise lock step with the S&P 500. For example, over the last 30 days rate sensitive groups like Utilities and Real Estate show 0.53/0.61 correlations to the index (max is 1.0). Technology and Consumer Discretionary show much higher correlations right now, at 0.94/0.95.
  • These correlation readings shift over time. From 2010 – 2015, for example, the average S&P sector was 0.84 correlated to the index. From 2016 – now, sector correlations have been much lower, averaging 0.67.
  • Sector correlations live in a virtuous/vicious circle with overall market volatility. When vol rises or falls, correlations follow. Part of that is simple math; the lower the correlation, the higher the systemic benefit from a diversified portfolio. The rest of the linkage is market psychology; in choppy markets investors derisk by selling equities en masse and that tends to push correlations higher still.

Here’s why this is important right now:

#1: Recent US equity market volatility has shoved sector correlations back towards the highs of the late – 2018 selloff.

  • Over the last 30 days the average S&P sector price correlation to the index is 0.79.
  • That is the highest reading since mid-January’s 0.88.
  • It is also higher than either mid-November 2018’s reading of 0.76 or Mid-December’s 0.78.

#2: History shows a clear August – September pattern for correlations:

  • In 8 of the last 9 years, correlations have risen from mid-August to mid-September. The one exception was 2012, but only because correlations were already very high (0.86) and fell modestly (to 0.84) over the next month.
  • Going back to 2010, the average increase in sector correlations from mid-August to mid-September was 7%, so if historical trends continue correlations will rise to 0.85 over the next 30 days. That would make the next 30 days feel very much like late December 2018.

#3: The unavoidable conclusion based on this historical analysis: US equity market volatility will rise further over the next month:

  • Tighter correlations mean less sector-driven diversification to dampen volatility.
  • Whenever correlations rise, the CBOE VIX tends to rise or remain elevated.

The sharp-eyed reader may wonder, “how can such obvious August-September seasonality persist in efficient markets?” The simple answer, which we outlined in detail last week, is the absence of a Federal Open Market Committee meeting in August/early September. Nature and markets both abhor a vacuum, in this case informational, and price volatility is the clear observable outcome.

Our thoughts on how to profit from these findings:

  • Derisk portfolios now. Volatility and returns have their own correlation structure, and over the short term it is negative. The change in volatility is what to watch. Persistently high correlations (as from 2010 – 2015) can make for good markets. Stubbornly low correlations (2016 onwards) can also coexist with rising prices. But the move from low to high correlations… That is a recipe for a downdraft.
  • Sectors with lower-than-average correlations are good parking lots for equity capital. Real Estate and Utilities look best on this measure, with the edge going to the former. Fundamentals also favor this approach; lower interest rates as risk-averse investors buy Treasuries should help these groups.
  • Look for entry points like our “4% rule”, which shows that buying on an outsized down day typically delivers decent returns over the next year. In our Markets section we also outlined our 4 metrics for a near-term bottom, and those are also good signals to consider as well.

Bottom line: we are not trying to be alarmist, and long-term investors may just want to stay the course. But the purpose of this section is to go wherever the data points. And right now it says, “be very, very careful”.