In a recent Project Syndicate article Dr. Roubini argues:
[M]y economic research firm has a quantitative model, updated every three months, that ranks 174 countries on more than 200 economic, financial, political, and other factors to derive a measure or score of these countries’ medium-term attractiveness to investors. This approach provides strong signals concerning which countries will perform poorly or experience crises and which will achieve superior economic and financial results.
Weeding out the bad and the ugly based on these scores, and thus picking more of the good apples, has been shown to provide higher returns with lower risk than actively managed alpha or passive beta funds. And, as the rankings change over time to reflect countries’ improving or worsening fundamentals, the equity markets that “smart beta” investors choose change accordingly.
The claim goes beyond the pale, and is absolutely stunning. It is hard to imagine that alphas and betas are not time varying parameters. In fact, studies by Blume; Hawawini, Michel, and Corhay; Levy and others have shown that stock betas can change drastically over two succeeding periods, and some have argued that linear estimators of beta are unrealistic estimates. Thus, one wonders, about the validity of any “smart” (or “enhanced”) beta strategy that can at any specific period pick up the true betas. It has always been a puzzle to many as to how some serious people look at betas and alphas as ex ante criteria for portfolio selection. Is it not reasonable to believe that the intrinsic value of any stock is derived from the firms’ competitiveness characteristics and the market fundamentals for the underlying goods or services that are represented by various stocks?
Based on the ex post data any econometric technique can always identify some alphas and betas that appear to have superior characteristics supported by an array of statistical measures attesting to the explanatory power of the regression. Such models may capture part of the impacts of the real market fundamentals, say a rightward shift of demand curve for the underlying goods and services, or a shift of the cost structure of the firm producing those goods and services, and so on. The data may also contain some memory, due to various lags that can be captured be the estimated equations. However, if a portfolio manager shows you a selection of stats (and there are hundreds of those; R-squared, P-tests, LM, DW, BP, F to name a few) that appear to suggest some superior predictive information content, then one really needs to ask why the investment manager is prepared to share such a valuable information for a small fee, instead of attempting to corner the market!
This is neither a rehash of efficient market hypothesis, nor an argument derived from the possibility of black-swans. It is a subtle recognition of the nature of risk and uncertainty in its Knightian framework. In other words, the expected return from a portfolio is not the same as the expected return from casting of a number of fair dice. The distribution outcomes from casting of a die can be detected by a repetitive casting process and thus the volatility of returns can be formulated as a Knightian risk. However, this is not the case for the expected return of a stock, because each return would be derived from a specific demand-supply configuration for the underlying stock’s goods and services and the position of the short-run average cost of the company producing them at a particular time. This does not lend itself to a repeated sampling. Moreover, we are seldom in an idealistic case of perfect competition, in reality various strategic pricing and capacity decisions together with logistical constraints would also play important roles. Thus, the underlying distributions of the expected betas are unknown –Knightian uncertainty. Can one resort to time series analysis of say cointegration type? Simply because of the unavailability of long enough data (i.e., degrees of freedom restraint), difficulties in detecting of the order of integration and a host of other technical issues that are well known to practitioners that option too would be impractical.
Read more at https://www.project-syndicate.org/profile/551891a0bc1f570d68f3eac8#6VwyDPZ00gQdq8WH.99
No comments:
Post a Comment