Recently I put something together which tried to identify which s&p 500 stocks were least correlated with the s&p 500 index itself. I pulled monthly returns for the past 62 months (12/31/04). The results were somewhat surprising. Only 1 of the 500 stocks (APOL) was negatively correlated with the index. Only 8 stocks (ABT, ADM, APOL, DV, FDO, FHN, GENZ, PBCT) had a correlation of less than 0.20. The average and median correlation were 0.55 and 0.56. (Obviously this is not a perfect analysis–it would be better to use more months. Also, the component parts of the index will naturally affect the performance of the index, thus promoting a positive correlation–but only 20 of the component stocks have an index weight of >1%, so hopefully this isn’t too big of a factor.) Three questions: 1) anything else wrong with this analysis, 2) does diversifying within equities make any sense, and 3) are there any stocks or industries that are likely to be negatively correlated with the market?
They are all going to be correllated… especially using recent numbers… the market crash killed everything… all risky assets will be correllated during this time period. The only ones in this environment that will appear to be uncorrellated for that time period would be result of significant events… i.e. huge discovery, corporate takeover, huge scandal, etc. Think about multi-factor models… Corporations may face many different company specific risks… but corporations share many risks… i.e. lending environment, investors willingness to take risks, harsh economic environment, horrible economy… the difficulty of this current enviroment is universal to all companies.
It’s hard to find stocks that are uncorrelated with a broad index since, you know, the index is comprised of these stocks. Perhaps a more meaningful analysis would be to regress a certain sector vs. the stocks in the index excluding that sector.
Have you used this? http://www.google.com/finance/stockscreener Education stocks tend to be lowly or negatively correlated.
This depends on the time frame of your analysis. If you have holding periods of one or more years, you really shouldn’t expect to see any negatively correlated returns, because that would imply that as the economy grows, a negatively correlated stock will tend to shrink. That means that the business doesn’t have a positive expected return, which means it shouldn’t be in your portfolio. On the other hand, if you do daily or weekly returns, you may find more stocks that are less correlated at that time scale, because the ratio of noise to signal is larger there. It still is going to be difficult to find truly negatively correlated stocks, simply because most stocks have a degree of market risk in them simply because they are stocks and a number of investors just invest in the indexes that drive them. You might test just the residuals of some model like CAPM and see if there are patterns in the residuals that are negatively correlated. The challenge with that is that the expected return on the residuals is going to be 0, and the volatility of those assets is likely to eat away even further, but it might at least give you a sense of what stocks or industries are possible diversification pairs.
Might be better of building a multi-factor model and looking to see how stocks are linked to the 500 based on a multitude of factors rather than just beta.
(1) That average/median correlation seems too low. (2) It’s not always better to look at a longer time period. Correlations definitely change over time. (3) If you’re looking for negative correlations, look at the entire matrix instead of just S&P.
>>>It’s hard to find stocks that are uncorrelated with a broad index since, you know, >>>the index is comprised of these stocks. Right, this would probably affect the average. But 450 of the 500 stocks in the index have a weight of less than 0.5%, so the performance of just one of these stocks is fairly immaterial to the index as a whole, right? >>>This depends on the time frame of your analysis. If you have holding periods of >>>one or more years, you really shouldn’t expect to see any negatively correlated >>>returns, because that would imply that as the economy grows, a negatively >>>correlated stock will tend to shrink. That means that the business doesn’t have >>>a positive expected return, which means it shouldn’t be in your portfolio. This seemed to make sense, but the data doesn’t bear it out. Of my 8 least correlated stocks, I have a pretty wide range of total returns. I think a stock can have a negative correlation to the market but still a positive expected return (due to the possibility of outsized gains when the economy falters). ABT 44.9% ADM 52.4% APOL -24.0% DV 79.9% FDO 15.6% FHN -59.0% GENZ -3.1% PBCT 51.3% S&P -8.5% >>>(1) That average/median correlation seems too low. Really, why? My highest correlation is 0.82. Only 81 of the 500 stocks have a correlation higher than 0.70.
“Right, this would probably affect the average. But 450 of the 500 stocks in the index have a weight of less than 0.5%, so the performance of just one of these stocks is fairly immaterial to the index as a whole, right?” You are assuming no correlation between individual stocks.
naturallight Wrote: ------------------------------------------------------- > Really, why? My highest correlation is 0.82. Only > 81 of the 500 stocks have a correlation higher > than 0.70. Try it with a shorter time frame.