Dave Rosenberg’s newsletter mentioned an interesting fact today that we often forget: There is substantial survivorship bias in the DOW, S&P 500, and other stock indices. Cleaning out the failures from the S&P 500 such as WaMu, Wachovia, Bear Sterns, Lehman, Fannie & Freddie and replacing them with companies that survived was responsible for 40% of the market rally off the 2009 lows – and if the index was re-calculated without those changes, it would be trading closer to 900 rather than 1100.
The main thing to get from this information is that you cannot rely too much on market indices as infallible indicators – and that they work better in growth cycles than recessions. I wonder how this affects the CAPM in practice, and whether “market” data used in regressions for historical Beta’s shows the same bias – or if there are indices that overcome this bias such as total market value of NYSE listed stocks.
The NASDAQ index uses all of its equities and is based on median security values weighted by market value, but this still automatically reduces the weight of failed companies while raising that of survivors… I’m thinking of an indicator based on total market value (de-listing would always knock it down and new listings would always raise it up). This would of course add new biases while eliminating the survivorship bias, but it would be useful because the information is different yet relevant.
This is what I love about being in the academic environment – I have access to great professors who are researchers and Economics PhD’s that I can come to with questions like this. Hopefully they’ll keep answering my questions when I graduate. I’ll post a comment here if I get some interesting answers back.
