Saturday, May 30, 2009

Instability & Volatility in Asset Correlations

In a prior post, I mentioned that I was going to focus some attention on the unstable nature of the relationship between asset classes. This post assumes some familiarity with my VIX Asset Allocation Model. USING THE VIX MODEL AS A TIMING INDICATOR FOR LONGER TERM STOCK ALLOCATIONS

For anyone interested in that topic, this is a link to a starter post: Vix Asset Allocation Model Explained Simply With as Few Words as Possible

In addition, it is important to understand the meaning of positive and negative correlation. 

I have just started to dig into this subject more thoroughly. I understand that the relationship between many asset classes is unstable, so that permanent conclusions can not be drawn based solely on historical experience. One asset class may have a low positive correlation or even a negative correlation with the S & P 500 Index during one period and then exhibit a high positive correlation during a severe bear market which would be a negative of course. I discovered this paper written by Willaim Coaker that documents the instability of the correlations between asset classes, and suggests a need for dynamism in constructing an asset allocation scheme, similar to what I have been preaching in this blog which requires a dynamic approach based on conditions in the market and the real world. This is a link to the Pdf of his paper:

One of the most stable negative correlations was between intermediate bonds and the S & P 500 average in years when that stock average declined. In all 8 years since 1970 to the date of Coaker's paper, intermediate bonds gained in value when the S & P 500 declined. This is what you want in an asset allocation. When the Trigger Event happened in the VIX Asset Allocation Model in 2007, the proceeds raised by my reduction in exposure in stocks had to find a temporary home in what would then be viewed as the most likely candidates for negative correlation with stocks. There were several alternatives considered at the time. The alternatives chosen were money market funds, bond ETFs (TFI, BND, BSV, GVI, SHY, TIP, BWX, etc), a municipal bond fund with Tennessee issues, and short term individual investment grade bonds.

Then, in 2008, I did a shift out of some bond ETFs into individual bond or bond like securities such as floating rate securities with a guarantee, trust certificates, trust preferred, fixed coupon equity preferred, and some individual long term bond names, with my buying of those issues documented for the most part day by day after I started this blog in October 2008. I then started a shift in 2/09 of the overweight short term bond positions and used the proceeds to buy stocks in early March 2009, also discussed in this blog.

This was basically a good response to the signal given by the Vix Asset Allocation model in August 2007, but none of those responses were based on any study of historical correlation patterns. It was based on a intuitive understanding of those historical relationships between asset classes based on many years of investing experience. I am now starting to examine in more depth the volatility of correlations among asset classes to better prepare myself for the next bull and bear markets.

An assumption can not be made that an asset class which was negatively correlated in past downturns would necessarily have the same correlation in the next one. Each downturn has to be assessed at its inception or as soon as possible once the investor realizes that a bear market in stocks is underway. In Coaker's study, natural resource stocks earned positive returns in 6 of the 8 down years for the S & P 500.

In August 2007, it would have been foolish to draw any conclusion about the future correlation of natural resource stocks with the S & P 500 based just on that historical negative correlation at the start of a bear market in stocks. In other words, the odds were not 75% in favor of adding to, or keeping the natural resource stocks in August 2007 after the Trigger Event.

The historical probability has to be assessed under the circumstances then prevalent. I would suggest that two factors would have militated against using the past historical correlation when entering the current bear market for the S & P 500: (1) the alternative asset has itself had a high positive correlation with the bull move in stocks and (2) the correlation is more than highly positive, the beta of the other asset is higher, so that the move in the other asset class is stronger in the same direction as the S & P 500 during that indexes bull move.

In fact, several alternative asset classes slid badly in the current downturn. This would include such assets as commodities, natural resource stocks, REITs, and emerging market stocks. My reasoning for jettisoning natural resource mutual funds had to do with the parabolic rise in oil and other commodity prices. The parabolic nature of the rise, plus the additional and obvious problems inherent in demand reduction caused by an economic downturn, would cause a change in the historical negative correlation pattern between those two asset classes and consequently require an reduction or elimination of the natural resource asset class. I have not studied yet the positive correlation statistics for these alternative asset classes during the 2003 to 2008 period, but I lived that history. All of them were moving up with the S & P 500, and moving stronger in the same direction.

This is a link to a prior discussion of beta and positive correlation for emerging market stocks and bonds. Emerging Market Currencies and Bonds as Non-Correlated Asset Classes/Links to Performance Data on Target Funds & More on Their Many Failures

This positive correlation with high beta in the same direction prior to the downturn in the S & P 500 would militate against assuming low or negative correlation in the current bear market. The more reasonable prediction would be a continuation of the the positive correlation with a high beta continuing in the same direction, except the direction would be down during this bear market.

Maybe the negative correlation in an asset class like commodities and natural resource stocks will hold up better in a period (1) where commodity prices had not had a parabolic rise until something happens that causes them to rise quickly while jolting the stock market into a decline (e.g. the first Iraq war in 1990-1991; the 1973 Arab Oil Embargo, 1973 oil crisis - Wikipedia, the free encyclopedia) and/or (2) a real shortage in oil and/or other commodities causes prices to skyrocket which precipitates a decline in the S & P 500 index while natural resource stocks rise in value. Those situations are different than the events which preceded the current bear market. Natural resource stocks had enjoyed several stellar years of large percentage gains. I remember one natural resource mutual fund that I sold in 2007 that had a five year gain of over 400%. You can not make an assumption of negative correlation in a stock bear market after those kinds of gains for natural resource stocks and oil hitting $140 going into a recession. The correlation with the S & P 500 was both positive and at a high beta.

So, after the Trigger Event, all alternative asset categories have to be independently examined and a prediction then made whether or not those assets will have a negative correlation with stocks at that point in time. A high positive correlation and beta before the Trigger Event with stocks would militate against using the alternative asset class. There may even be strong factors present militating against the use of certain bond asset classes, such as a period of stagflation causing both a downturn in stock and bonds. Also, it is important to keep in mind that the VIX Asset Allocation Model does not provided guidance on the severity or length of the downturn in stocks. So when doing the asset allocation, I have to gauge how serious are the events which caused the disruption in the Stable Vix Pattern, causing the formation of the Unstable Vix Pattern, at higher and more dangerous volatility levels. The answer to that question will also impact my allocation decisions in the future. This topic is generally discussed in these posts:

There is one caveat. The formation of another Phase 2 Unstable VIX Pattern in the future will be dealt with immediately by a number of shifts in asset allocation: SEPTEMBER 2008: FORMATION OF THE DEADLY PHASE 2 OF THE UNSTABLE VIX

I am actually starting to work on those potential responses now, realizing that my response in late September 2008 was inadequate in many respects.

I am just writing in this post some very preliminary thoughts on this subject, immediately after waking up this Saturday morning, with no beverage jolt to the mind to help the process, before the thoughts escape me. I will come back to the topics discussed in this post after more study and thought.

No comments: