Why correlation between securities matters
First a digression: I purport to be a long-term investor and preach friends and family to have a long-term view. One might think this means disregarding the short term "noise" associated with day to day or week to week stock price fluctuations in order to focus on the long term. Actually, I pay a decent amount of attention to the short term noise for two reasons: 1) the noise or volatility carries opportunities and 2) the long term is simply a series of short terms stitched together.
Running a mile is running 4x400m. The sum of my mile time is simply the sum of each 400m leg. When calculating investment performance over a decade, however, one cannot simply sum or average the returns for each of the 10 years. Take an extreme scenario where I generate mean returns of 10% in two cases below.
Enter Volatility. The final values in both cases are different, despite the fact that mean annual returns are 10% in both cases. The geometric mean of returns is lower than 10% in the second case.
This is why I use a quantitative overlay (or a sort of risk model) that measures the 1) correlation and 2) volatility of each stock in the portfolio along with traditional bottom-up stock picking. The goal is always to generate the highest returns possible over a long time period, but to do so in a way that:
Minimizes volatility. Lowering volatility can dramatically improve compound growth (geometric mean of returns). Sleeping well is important - I do not want to experience huge drawdowns.
Minimizes correlation. This refers to the correlation between my returns and market returns, but also the correlation between individual securities in my portfolio. As long as fundamentals are good for a particular name, I want to be able to trim a stock that is working to buy the stock that is lagging. This sort of rebalancing only works if you have uncorrelated bets, otherwise everything goes down and up together.
Has repeatability. I gain comfort in knowing that there are some guard rails to the investment process in terms of sizing, taking profits, and in terms of research process.
Correlation output - measures correlation between each name and each other name in the portfolio, a proxy for a real risk model or factor model
Volatility output - measures deviation of price changes, a proxy for volatility
To be clear the most important thing is finding stocks that have good fundamentals and are trading at reasonable valuations. Once you have found that slice of the market, it is important to have a convicted differentiated insight or thesis on each name you own. This requires reading, listening and generally observing the world around us.
As a second step or filter to aid portfolio construction, we must consider correlation and volatility. We will be ruined if we consider correlation and volatility first.