I have been reading a lot about correlations of asset classes increasing since 2007ish. Of course this is something I can easily see with my own eyes as it seems like every day in the market is either a risk on day (buy any kind of stock or commodity) or a risk off day (buy Treasuries). Increasing correlations present another problem for Modern Portfolio Theory (MPT). Most practioners assume that asset classes like large stocks, small stocks, international stocks, value stocks, growth stocks, commodities, and high yield bonds are not that correlated (meaning combining them in a portfolio actually gives you diversification) and that correlations will remain constant over the investment period. This presents a couple of problems:
1. Correlations are not constant, as evidenced by increasing correlations lately.
2. Taking a straight line correlation ignores the most important type of correlation there is, underwater correlation, or how correlated asset classes are during a drawdown.
Two assets may appear to be uncorrelated when you look at a long time period, for example stocks might appear to be relatively uncorrelated with oil. However, during an economic decline stocks will generally suffer, as will oil because of a decrease in demand. So having these two assets could help some during bull markets but could just give me two different ways to lose money during a down market.
One time when increasing correlations could help is during a bull market. If my entire portfolio is correlated to stocks then my entire portfolio might be increasing during a bull market. This may seem to be helpful but my 9 year old son, and just about anyone else, can make money in a bull market. Where diversification is supposed to help is in a down market. Increasing asset class correlations plus a down market is a recipe for disaster.
The solution: Diversify by methodology, time frame, and security basket, basically have different return streams that are uncorrelated on an underwater basis.