“Excessive Extrapolation and the Allocation of 401(k) Accounts to Company Stock” Benartzi (2001)
The overinvestment in familiar assets is a significant bias which has already been discussed on this blog. Using a US-based dataset it looks at employees’ voluntary contributions to company stock, this paper asks a slightly deeper question: is it just familiarity, or does the magnitude of returns matter too?
This paper finds that employee contributions to company stock are strongly linked to the stock’s performance over a 3 to 10 year window (although not shorter). Over 3 years employees contribute 13.64% more to high- than low-return stocks, while the difference increases to 29.33% over a 10 year window.
Naive extrapolation of returns is of course not limited to individual stocks; it is a common phenomenom across asset classes too, where it can lead to the formation of bubbles (Dot.com; property). Shiller’s Irrational Exuberance is the classic reference on this topic. Although people often place too great a weight on small samples of data, in most walks of life acquiring data should increase your confidence over random variables. Things that go up a lot in the past tend to also increase in the future. The stock market is different. Since stocks returns tend to mean-revert – periods of above-average returns are often followed by below-average periods and vice cersa – you should actually become wary if an asset increases very quickly!
This paper also presents evidence that many employees believe their company stock is less risky than the overall market, despite the fact that a large portfolio of stocks should lead to a large reduction in risk from diversification. Again, I think this is partly because diversification is an unintuitive concept with few equivalent processes in the real world, and partly this might be because investors make decisions in isolation without seeing how each should be part of a greater plan. So it’s little wonder than many investors would fail to appreciate the two unintuitive finance concepts of diversification and mean-reversion (the latter of which is stronger than regression to the mean).