The Statistics of Forecasting - Part I

Now I am not sure I should have bitten off this topic, but it is very interesting to me and certainly something that the average investor does not understand - forecasting is really a guess and not something on which to weld your portfolio investment methodology.

The first topic that I want to dive into is correlation. Correlation is used extensively in creating portfolios. The idea is to try to reduce the volatility (risk) of a portfolio by investing in an array of differing investments. If each investment is not identical to the other then they will not produce the same returns as each other in the same market place. It is said that they are not correlated to one another. If we take this a step further and find assets that move in the opposite direction to one another then we have a negative correlation.

As an example a bear fund should go up when the market goes down and this should offset losses associated with the remainder of your stock investments. This negative correlation can be a great protector of value during market corrections and can smooth returns during a volatile market. However there are a lot of misconceptions on this score and it can water down your returns during a roaring bull market.

The first misconception is that the negative correlation holds in all market conditions. People that diversify their portfolio into stocks, bonds, commodities, real estate and other exotic investments discovered, to their dismay, that when the market tanks that these relationships erode and all positions deteriorate at the same time removing the net just at the time it is most needed.

Second is that while the most sophisticated financial models can find correlations to anything these models end up "mining" data. That is to say that they get so sophisticated that they end up finding correlations that are fictitious. Back in the early 1900s Pearson described this as a "lurking variable". There is a false variable that for a moment shows that the two positions are negatively correlated so it is added to the portfolio under the guise that it will reduce the overall portfolio volatility. In fact this lurking variable can disappear at a moments notice and remove that supposed "positive" influence.

The thing to remember is that while we all strive to reduce the volatility of our portfolios it should not come at the expense of losing a rational investment methodology. If things get bad, get out! Do not rely on your supposed diversification or your correlation studies. These will fall apart at the seams right when you need them most. Invest wisely and take risk off the table by selling rather than hoping.