Van Nieuwerburgh and Veldkamp (V&V) help us to understand the importance of information costs (as learning capacity) on portfolio decisions. They model the portfolio (diversification) aspect along with the learning (information) costs necessary to hold a diversified portfolio.
The authors stress the difference between pure information costs and the ability of investors to handle information (the capacity side). This allows for the following
"evidence suggests that the degree of diversification only slightly improved over the last decade, in spite of a large drop in (fixed and proportional) transaction and information costs. While the ease of access and the speed of dissemination of financial information have dramatically improved over the last(A note to my own students: in class we have always combined these two into a broad information cost catergory).
decade, the processing capacity of the investor has not."
Following along in this discussion:
"The interaction of the information portfolio problem and the asset portfolio problem creates a trade-off between diversification and specialization through learning. The result is that investors hold some fraction of their assets in a well-diversified fund, about which they learn nothing, and hold the other fraction in a small set of highly-correlated assets that they specialize in learning about."
"For the investor with zero information capacity, it is optimal to hold a diversified portfolio; our theory collapses to the standard model. As the investor's information capacity increases, holding a perfectly diversified portfolio is still feasible, but no longer optimal...."
This "if-investors-are-concentrated-then-it-must-be-for-a-reason" idea is summed up as the following:
"If investors concentrate their portfolios because they have informational advantages, then concentrated portfolios should outperform diversified ones (corollary 3). In contrast, if transaction costs or behavioral biases are responsible, then concentrated portfolios should offer no advantage"The authors point out that there is evidence to support this:
"Ivkovic, Sialm, Weisbenner (2004) find that concentrated investors outperform diversified ones by as much as 3% per year. This excess return is even higher for investments in local stocks, where natural informational asymmetries are most likely to be present." (I would also add Choe, Kho, and Stulz)
The paper also write that their model can partially explain the problems with CAPM:
"We find that the risk premium on an asset is low when its correlation with the risk factors that the economy learns about is high. Asset returns are also described by a CAPM; the CAPM that would hold if each investor had the average of all investors' signal precisions."
Definitely an interesting article! Especially for a largely theoretical paper ;)
Van Nieuwerburgh, Stijn and Veldkamp, Laura, "Information Acquisition and Portfolio Under-Diversification" (March 2005). EFA 2005 Moscow Meetings http://ssrn.com/abstract=619362