©2005 All content on site protected by copyright
Demographics and the Stock Market, by Victor Niederhoffer
It's exciting to see an effort to take a grand macroscopic look at the effect of an important variable like demographics on individual stock market returns. Professor Stefano DellaVigna of Berkeley makes such an effort in his paper Attention, Demographics, and the Stock Market. His startling conclusion is that portfolios with good demographics outperform those with poor demographics by about 8% a year for the 30 years ending in 2003. The key, according to DellaVigna, is that investors pay too much attention to short term trends in age factors in the 1- 5 year horizon but not enough to predictable changes in the age distribution in the 5- 10 year horizon.
But there are so many hurdles to leap over with such a grand motif that the author's conclusion that predicted age structure can lead to systematic outperformance seems highly tenuous. Some of these hurdles:
One of the key links of the paper is an attempt to show that the profitability of certain industries will be affected by the changing age distribution of the population. Among the affected industries:
The next step is to estimate the concentration ratio in each industry to come up with a theoretical sieve as to which industries would be expected to maintain high profitability. Another step is to calculate the time horizon that investors might react to the predicted changes in age structure. Finally, the companies' exposure to each of these variables is estimated using Compustat and SIC date. Each one of these steps requires a tremendous amount of data mining and retrospection, and implicitly contains the end result of a myriad of hurdles that would have led to negative results thrown out and good paths followed.
The paper cries out for an out-of-sample test, with the companies that are supposed to be helped or harmed clearly spelled out so the reader can make his own verifications, refutations and extensions. Failing that, since most academics have a certain reluctance to give away the skinny, a series of out-of-sample tests, perhaps starting at year-end 1999, using only contemporaneously-available data, would be most educational. A study that is statistically significant with past data is not necessarily likely to be predictive for the future. Indeed, the opposite is true.
The author believes his results show that "forecastable future demand changes (in the 5- 10 year horizon) due to demographic variables predict abnormal annual stock returns." I conclude that by overreaching and artful misuse of seemingly rigorous statistical procedures, that this paper belongs in the bailiwick of the Minister of Non-Predictive Studies. However, Professor DellaVigna is to be complimented for focusing attention on a key variable to consider with respect to the analysis of individual companies.
Dr. Kim Zussman Adds:
Adding to the frustration is how obvious such trends look in retrospect:
The harder questions are about what's coming:
Jason Ruspini Responds:
Demographics might also have an effect on rates. All things being equal, it seems that if both total population and the ratio population ex-retirees,children and the unemployed / total population are growing, the supply of credit should also grow.
Technology almost always has an effect on time, either by making things take less time (e.g. travel, communication, manufacturing processes), or in the case of medical technology, simply giving us more time. If legislation caused the average retirement age to suddenly jump higher, I wonder if that would have a material effect on rates. Numerous other factors should drown it out, thus my cheapish "all things being equal."