At DataTrek we have a nerdy fascination with asset price correlations. Returns are ultimately more important, of course. But how different S&P industry sectors trade relative to the overall equity market runs a close second in our eyes.
Why? One key reason: correlations are signposts on the journey through a capital markets cycle. Consider the last 9 years as a case study.
Post-Crisis recovery (2010 to 2013, peak correlations):
Over these three years, correlations averaged 0.83 to 0.88, or an r-squared of 69 – 77%. Some months even saw +0.90 correlations and 1.00 is, of course, as high as they go.
Since cross-cycle correlations average 0.70 (r-squared of 50%), it is clear investors were treating equities as an asset class rather than a collection of individual stocks with varying fundamentals.
Normalization (2014 and 2015, falling correlations):
Over this 24-month period correlations declined to annual averages of 0.75 to 0.82 for r-squares of 56 – 67%.
Markets were sensing the worst had passed, and began to differentiate between various sectors. Correlations (r-squareds, really) fell closer to that 50% normal level).
Trough correlation (2017, the low point of this cycle):
Last year correlations declined dramatically to a monthly average of just 0.55 for an r-squared of 30%.
Spurred by hopes for a tax cut, deregulation and further economic growth, investors more actively picked winners and losers. R-squareds fell to below normal levels.
Current conditions (2018 year to date, end-of-cycle fears):
Now we face investor concerns about the next recession/crisis. For the YTD sector correlations have averaged 0.67, or a 45% r-squared.
More importantly, the last 2 months have averaged 0.77, or an r-squared of 59%. That is slightly above average, but heading in the “wrong” direction after a trough in 2017.
The important question now:
“Is the current uptick in correlations another sign of an impending bear market/recession, already signaled by faltering asset prices?”
The jury is out on this. Other market signals like the performance of Financials (more on that in the Data section below), a flattening yield curve, and market expectations for Fed rate hikes all point in that direction, to be sure. The bearishly inclined will lump higher correlations into their list of concerns.
We would offer up 3 points about rising correlations that are relevant regardless of market direction:
#1. Higher correlations drive market volatility. When sectors move in closer lock step, diversification does less to limit daily price swings for the S&P. And since market measures of volatility like the VIX key off actual price movements, this also puts a floor under the VIX. The days of a 10-15 CBOE VIX Index range are past us.
#2. Correlation tends to be “sticky” absent an overwhelming catalyst. We’ve only had one dramatic shift in the last decade: the 2016 election of a Republican president and Congress (2017, as noted above). Aside from that, they tend to move slowly (2010 to 2016, for example). Since we see no analogous event to 2016’s election on the horizon, we assume the current volatility regime will continue for the foreseeable future.
#3. High sector correlations are not necessarily negative for US stock returns. The 2010 – 2013 period of 0.83 to 0.88 average correlations saw the S&P gain 33.1%. To our point #1, however, annual total returns were lumpy (2010: 14.8%, 2011: 2.1%, 2012: 15.9%). And, of course, these came during a time of heavy-duty Federal Reserve policymaking with rates at zero and bond buying to boot. Those initiatives kept the dollar weak, which also helped US equities.
Summing up: we see higher S&P sector correlations and incremental price volatility as a ‘new normal”. Even a timely resolution in the US/China trade dispute will not likely change that. Too many other issues wait in the wings, from high financial leverage to worries over global economic growth and US corporate earnings. History shows the S&P 500 can still work even with higher correlations. But even if it does, gains will come in fits and starts rather than calmly.