Globally there is over $534 billion invested in 1,208 smart beta ETFs, according to the Investment Company Institute.
Last year alone attracted $47 billion with 208 new products launched by 67 providers.
The growth over the last five years has been an impressive compounded 31%. With numbers such as these the incentives to develop new products are clear.
Back-tested results promise strong returns, however for investors looking for a more rigorous due diligence process, this and subsequent articles set out to provide a framework to follow when selecting smart beta products.
Rule number 1: Be serious about the data!
It is vital to first determine if the factors really do stand on solid academic grounds.
Every January, the American Finance Association, the world’s leading society for research in financial economics, attracts thousands of academics (and some practitioners) to listen to the newest findings on various topics of the field of finance.
The conference starts with the presidential address and this year’s speech was delivered by Campbell Harvey, Professor at Duke University.
Harvey came out with a very provocative statement, claiming:
‘Half the financial products (promising outperformance) that companies are selling to clients are false.’
This statement is based on a recent paper ‘…and the Cross Section of Expected Returns’ that he co-authored with Prof. Liu and Prof. Zhu, respectively at Texas A&M and Duke. The article was published in the 2016 issue of the highly respected, Review of Financial Studies.
Harvey makes the argument that many published papers suffer from a disease that statisticians call datamining.
Rule number 2: Mind the p-hack.
It refers to the practice of analysing financial and economic data without a guiding, a priori hypothesis.
In this case, the risk of ‘p-hacking’ (a term coined by Harvey as a proxy of torturing the data until you find the results that you want) is high.
This risk is increasing by the day since the cost of analysing data continues to decrease thanks to advances in computing power and the widespread availability of financial and economic data.
One possible cure for datamining is to conduct out-of- sample testing. That is checking whether the investment strategy, which promises staggering outperformance, works in different time frames or different countries and geographical areas.
Harvey thinks that this is not enough and proposes a stronger prescription: researchers should raise the bar for identifying statistical significance by increasing the acceptance threshold. Traditionally a t-statistic >2 was deemed sufficient.
He argues that newly discovered factors should have a t >3, before significance is concluded especially in the case where the factors are not founded on strong a priori economic hypotheses.
If strong economic arguments can be made in support of the factors, then, a t-stat less than 3 but still greater than 2 is acceptable. This prescription, he adds, will need to be re-evaluated through time if the yearly growth rate of discovery of new factors continues.
Rule number 3: Apply the Harvey cure.
Administering this new ‘medicine’ to the 296 published papers where factors have been analysed, roughly half (actually, 158 papers) would be considered false discoveries under one of the three adjustments considered by Harvey et al. (2016).
So which factors should investors focus on and implement in their portfolios?
Figure 3, an extract from Harvey et al. (2016) plots some of the most prominent factors discovered in the past.
The only ones that are significant across all three adjustment methods are: value (Basu 1983; Fama-French 1992), momentum (Carhart 1997), short run volatility (Adrian and Rosenberg 2008), durable consumption goods (Yogo 2006) and good old market beta ( Fama and Macbeth 1973).
So, the next time you are given a pitch on a new smart beta product, ask what the t-test backing the strategy is!
 2016 Fact Book, Investment Company Institute
 Fuhr, D. Assets Invested in Smart Beta Equity ETFs/ETPs Reach Record $497 Billion by November 2016. ETFGI, February 2017
 Harvey et al. (2016) consider three adjustments in the framework of analysis: Bonferonni, Holm and BHY. We refer to Bonferonni.