Sector Correlation Sums as Metric for Market Timing

Using a sector cross correlation metric for market timing

By Clive Corcoran

How much co-movement is there between disparate market sectors and indices?

Correlations and Daily Changes SPY

Each time the normalized correlation sum moves above 0.65 (i.e. where the black columns first appear) has coincided with broad market sell-off points

Clearly evident on the chart are two uptrend periods from mid 2006 until February 2007 and mid-April 2007 - June 2007 where there is an absence of highly correlated movements of the sectors

The hypothesis is that market sectors become more correlated (i.e. they start moving together in a coherent alignment) as quant top/down strategies begin to cohere and this causes reduced market liquidity leading to corrective behavior

Market liquidity depends on fractious behavior and beliefs and where traders and asset managers become too coherent and aligned in their views the degree of correlated movements increases and the direction is usually downwards

Methodology for constructing a simple metric for tracking the degree of co-movement by disparate market sectors.

Use various sector indices that ordinarily will not be highly correlated.

    • SPY -  ETF for the S&P 500 index
    • IWM – ETF for the Russell 2000 index
    • XBD – Broker/Dealer Sector Index
    • BKX - Banking Index
    • XLP - ETF for the Consumer Staples sector
    • XLY - ETF for the Consumer Discretionary sector
    • RLX - The S&P Retail Index
    • HGX - Housing Index
    • XLB - ETF for Industrial Materials

Use a 20 period window to calculate all of the cross correlations 

Sum the individual correlations (some will be negative, most positive)

Normalize the correlation sum values to create a value between 0 and 1

Plot the correlation sum values according to the left hand scale and plot the SPY values according to the right hand scale

Higher correlations amongst sectors also coincides with increased day to day volatility

    • The chart below shows a similar normalized correlations metric to the one above and also plots the day to day percent changes in the S&P 500.
    • To make the chart simpler and avoid clutter the changes that are between plus one percent and minus one percent have been zeroed and show as a continuous orange line.
    • It can be seen that when the correlations are peaking marked as the points A, B and C this corresponds with the most frequent periods for larger day to day changes in the broad market
    • Enhanced correlations are a background condition and probable cause of increased volatility
    • There is an inter-relationship between correlation, liquidity and volatility which can be explained in terms of market coherence versus the more "normal" condition of market fractiousness
Correlations and Daily Changes SPY

Fractious markets are liquid markets

    • Liquidity is one of the more important concepts in trading and finance and yet it is also one of the most difficult to define.
    • Liquidity is often seen as some kind of macro market variable that can be related back to the money supply or credit that is “in the system”.
    • Almost certainly it eludes any obvious way of being quantified.
    • Better not to view liquidity as having to do with money “sloshing around the system”
    • Rather it has to do with the degree of disagreement amongst traders.
    • This is a condition that can be called fractiousness.
    • A typical trading day shows little consensus or agreement about the near term direction of prices.
    • It is precisely this disagreement that facilitates trading.
    • During a typical trading session there are willing buyers and sellers who will have different time horizons and strategic objectives and these different perspectives will find expression in a flow of transactions
    • This flow is predicated on the fact that the different side to a transaction disagree over the fitness of the prevailing price.
    • Fractiousness conveys the notion that price discovery and the moment to moment movements of markets have an adversarial flavor.
    • How does a more uniform or coherent view of price direction arise?

One of the few things that can be relied on to increase during market crises is the correlation amongst asset classes.

    • This should act as a strong antidote to a certain kind of trading strategy that relies on combining long and short positions based on correlations, and so called convergence/divergence strategies.
    • Some kinds of correlation strategies are relatively harmless – pairs trading
    • Others are so fraught with undisclosed risks that they pose a systemic risk as well as sometimes being ruinous for those that practise them.
    • Correlation amongst assets is the underlying metric that best “measures” a liquidity crisis.
    • Trading strategies that assume normal market liquidity will always prevail are dangerous.

In August many “quant” funds based on correlation strategies suffered dramatic drawdowns.

    • Goldman Sachs’ GEO fund lost more than 30 per cent of its value in just a few weeks from late July.
    • Its flagship Global Alpha fund, which uses quantitative strategies across a range of asset classes also experienced a 27 per cent drop by late August.
    • The funds had to be rescued with a $3bn injection from GS and third parties.
    • David Viniar, Goldman’s CFO
    • “We were seeing things that were 25-standard deviation moves, several days in a row. There have been issues in some of the other quantitative spaces. But nothing like what we saw last week.”
    • "Across most sectors, there has been an increase in overlapping trades, a surge in volatility and an increase in correlations," Goldman Sachs explained in a statement.
    • "These factors have combined to challenge many of the trading algorithms used in quantitative strategies.”
    • “Liquidity conditions were most extraordinary during early August.”
    • The more you need cash, the higher the price you have to pay to get it.
    • When average opinion comes to believe that average opinion will decide to turn assets into cash, then liquidity may be confidently expected to go to zero.

Circularity in Assumptions

    • Correlation strategies assume liquidity but one of the best measures of illiquidity is enhanced correlations.
    • All correlation trading strategies are based on the assumption that normal liquidity will prevail.
    • BUT liquidity is itself dependent on normal market fractiousness not coherence in views
    • Illiquidity is best measured by enhanced correlations amongst asset classes and co-movements (usually down) of normally non-aligned variables.
    • All of the historical diversification benefits can evaporate quickly