Caution: Correlation Breakdown

I’d like to add some perspective to what I see as a major fallacy of post ‘credit bubble’ econometrics.  It is something I feel economists and other financial pundits have whiffed on completely.  And it is a rather simple and common concept.  Correlation breakdown.  Now we typically think of correlation in market terms as relationships between or within asset classes.  But correlation is a relationship between any two things and covariance between multiple things.  Essentially how do changes in one thing result in changes to some other thing or things.  Having been responsible for very large nominal daily exposures over the years, correlation and covariance are concepts I paid great homage to.  The idea being I can understand how my overall portfolio of assets will behave given particular inputs by understanding the correlations within my portfolio and between my portfolio and exogenous inputs.  The key is understanding the relationships.  Now relationships amongst assets are no different from human relationships in that they change over time.  Not appreciating the dynamic nature of relationships will guarantee you an ugly ending in humans as in assets.

Now traders tend to be very aware of these concepts in the narrow focus of trade analytics.  Where I feel even traders fall short is accepting the economists assumptions about econometrics.  Specifically, when looking at leading indicators economists are very reluctant to accept that relationships change on the macro level.  And because traders tend to accept the economists at their word on economic issues they are being led astray.  Bad information will always lead to bad decisions over time.

Take for instance the relentless focus over housing.  This is something I simply cannot understand.  For decades it was well accepted that the appreciation in one’s home made them feel wealthier.  So even though that increased equity was not necessarily liquid (although during the boom of the 2000’s equity was liquidated), that home owner would tend to spend more as his/her home gained value.  This is know as the wealth effect.  And through quantitative and qualitative econometrics it was proven that the wealth effect was real.  I agree.  However, the entire paradigm of home ownership was changed during the last (and continuing) economic crisis.   That is to say, the scars and the continuing fear from what took place has essentially muted the wealth effect.

But let’s look specifically at the relationship between the indicators and wealth effect so we can pull this altogether.  The indicators, being existing home sales, new permits and new starts.  The markets look to the release of these figures with great anticipation based on an assumption that strong indications on these inputs should drive the market up.  But that’s a big leap so let’s walk through the relationship.  The old concept is that strong housing indications mean there is strong demand for housing.  Strong demand for housing means housing prices should be moving up given a steady supply.  Increasing housing valuations mean home owners feel wealthier.  Increased wealth effect on home owners means more spending and consumption.  For economists interest, increased consumption will lead to increased GDP given consumer spending makes up around 70% of GDP.  For traders, increased consumption means higher expected future cash flows, which mean higher equity valuations.  And thus the idea is to buy, buy, buy on strong housing indications.  As a small caveat housing starts will indicate more work for construction and thus incomes move up but that is insignificant in comparison to the old wealth effect so let’s leave that out of this discussion.

Ok so there we have the relationship between the stock market and the housing market.  You can see it is not a short step from strong housing indications to increased cash flows, the basis for higher stock prices.  But for a long time it was a real effect.  However, I would suggest that post 2007, this wealth effect has fallen away for several reasons.  First, is that the housing market is fragmented relative to history.  Meaning that correlations between subsector demographics of housing have broken down.  That is, just because one category of housing is acting a certain way is not indicative that other housing sectors are acting that way.  Second, the correlation between housing indicators and price valuations has broken down.  There are huge factors impacting those correlations such as mass amounts of foreclosed homes owned by banks that are in the shadows rather than on the market.  As market supply is bought up the shadow supply makes it way onto the market diminishing the affect of demand.  That is supply keeps a price equilibrium with demand as banks are more interested in selling their supply of foreclosed homes at steady prices than appreciating that shadow supply.  Third, consumers have significantly more non-revolving debt than they ever had.  Non-revolving debt means non credit card, big ticket debt such as auto and student loans.  This debt involves fixed large monthly payments.  This, let’s call it, ‘debt effect’ would wash out any wealth effect created by home appreciation.  Forth, there is a real fear factor that has chilled out home owners from relying on home valuations as a feeling of wealth.  Our homes simply do not instill a stable and confident sense of wealth the way they used to do.

So I have taken one indicator of several for which I believe the old correlations have broken down between the indicator and the market.  If I am correct, then using the housing indicators as market indicators is not just a waste of time it is dangerous.  As traders ignoring correlation breakdowns can skew one’s VaR (value at risk – the basis of which is correlation and covariance).  Granted traders daily VaR is likely not significantly impacted by housing indicators.  However, for portfolio managers and algo traders indicator correlation breakdown will lead to poor allocation decisions.  When enough market action is occurring based on correlations that have broken down, well…. Houston we have a problem.  I would suggest that industry folks and average Joe’s/Jane’s alike should think very hard about the old standard indicators (not just housing).  Whether you’re trading on these indicators or just trying to figure out the risk level in the overall economy, read these indicators with caution.  Unrecognized correlation breakdown is a game ender my friends.