How to read financial research better than Financial Times

FT has recently appealed to the research by Hendrik Bessembinder who compared stock returns to US T-bill returns and analysed value added above this risk-free benchmark

As is traditional for financial journalism, FT has been reading it with its ass. The popular point taken from the research is that few stellar performers make all the profits on Wall Street, which implies that the rest are useless at best. With Bessembinder's 2017 paper already 4 years in my Investment Markets courses at New Economic School and NSU-HSE, let me share my takeaways from it:

1. Champions are indeed few (between 1926 and 2015 in USA just 86 out of 27,000 listed stocks generated half of the $31.8 trillion added by stock market above T-bill returns, 1,000 stocks making 100% of it). The trick point, however, is that predicting the champions is the feat even ultimate insiders can't perform: the founders of Google first wanted to sell it for 1 million, then for 5 billion (but were offered only 3); Tesla has been selling its stocks at the IPO last February 6 times cheaper than now. Therefore, stock-picking trying to find the next Microsoft is not a sound investment strategy.

2. Despite the loud claims that "1,000 out of 27,000 generated all value added by stock markets", it were really 12,000 stocks earning 120% of value added, while other 15,000 were eating through those extra 20%. Thus, stock-picking could provide you with above market returns, if only by not picking garbage stocks. Now, whether your strategy can really separate the wheat from the chaff is another question.

3. Even without stock-picking passive index investing participates in both champions and loser, with average return well above T-bills (what almost noone mentions) and is therefore the best pick for your average Joe.

4. Real stock returns distribution is heavily skewed towards "-100% since inception" outcome, making useless the linear regressions and risk metrics built on the assumptions of iid symmetrical distributions. Too few understand that, while the majority of hegde-funders and risk managers destroy value and generate risks by overrelying on false methods.

Ingersoll, Spiegel, Goetzmann, and Welch (2007) note that performance evaluation measures such as the Sharpe Ratio or Jensen’s Alpha were designed to be used in a world where asset returns confirm to simple distributions such as normal or lognormal. The evidence reported here indicates that longer-term stock returns in particular do not conform to these simple distributions, implying the need for reassessment of standard methods of evaluating investment management performance.