Remember to check the assumptions made for the cost of trading when examining a new index concept.
[This blog initially appeared on our sister site, IndexUniverse.eu.]
Vanguard’s warning of the perils of index data mining is timely. As the number of “smart beta” index concepts increases, each promising superior performance than old-fashioned, capitalisation-weighted benchmarks, the possibility of investors getting hoodwinked also grows.
Just about anything can be used to “predict” something else if you use historical data series creatively enough. According to fund manager David Leinweber, the Wall Street Journal reports, annual butter production in Bangladesh “explained” 75% of the annual returns of the S&P 500 over a 13-year period. If you throw in data for US cheese production and the combined sheep population of the US and Bangladesh, Leinweber says, you get to "forecast" US stock prices with 99% accuracy.
Not everyone got the joke. A number of firms asked Leinweber to share his data on Bangladeshi butter production so that they could build a trading strategy around them, the WSJ tells us. Were any index and ETF providers looking for a new smart beta concept among them, by any chance?
Vanguard has its own axe to grind in all this, we shouldn’t forget. The firm sticks religiously to using traditional, cap-weighted indices as the basis for its passive funds, arguing that anything else is an active bet on market behaviour and should be recognised as such. I’ve argued before that this is as much as a commercial strategy as anything else—Vanguard’s huge size precludes it from even considering index concepts that are in any way capacity-constrained, as many non-cap-weighted approaches are.
But a good first step in assessing an index promising “smart beta” and outperformance is surely to ask oneself if the underlying investment concept makes sense. Does Research Affiliates’ “fundamental indexation” approach in equity markets, which uses companies’ revenues, profits, book values and dividends as a way of determining index weights, hold water as a strategy? It does for me.
Does the same firm’s alternative weighting scheme for sovereign bond markets (which is based on countries’ GDP, energy consumption, population and rescaled land area) work as a stand-alone investment concept? I’m not so sure. Does a smart beta strategy focused on historical stock volatilities work as a predictor of future risks? For me, not at all.
There’s also one topic Vanguard didn’t touch on in its review of the pre- and post-launch performance of newer indices—trading costs. Even if you like a new index idea, how do you know that the costs of buying and selling index constituents have been reflected accurately and fairly in the back-test?
There’s an obvious incentive for the promoters of a new index to flatter its historical "performance" by taking an optimistic view of how much it would have cost to buy and sell the index constituents over time. And while cap-weighted benchmarks are largely self-rebalancing, typically generating only a few percentage points of turnover a year, newer index concepts can easily involve annual internal index turnover of hundreds, even thousands of percent.
Historically, it appears that many index providers have dealt with the thorny problem of internal trading costs very simply—by disregarding them completely.
“Turnover-related costs...[have] been widely ignored in index construction, based on the assumption that [these] are negligible for the typical investor. Index providers essentially follow the basic theory that (equity) markets are free of transaction costs,” Konrad Sippel of STOXX writes in an article to be published in the September/October Journal of Indexes Europe, in an issue focussing on index tradeability (you can sign up for a free subscription here).
“Another reason for not including cost-related elements is of course that these are very hard to measure transparently and consistently, as each investor has different cost structures, depending on their individual circumstances. The introduction of client-specific cost elements would dilute the function of an independent and transparent index,” Sippel goes on to point out.
In other words, trading costs incurred by index-related fund management activity may end up being reflected in tracking error of portfolios run against the index, rather than being internalised in the index’s return itself. But there is no common practice in this area. Some (fixed income) index providers do make a charge to their benchmarks’ return to reflect the cost of bonds’ entry and exit into the index portfolio. Equity indices tend to be calculated on the basis of recorded end-of-day trades in the index constituents, presenting a problem if index-tracking funds can’t deal on the same terms.
There are no easy answers here. Index tradeability is a subject that involves complex questions of market structure, technology and regulation. But it’s an increasingly relevant one as the number of smart beta launches multiplies.
As passive fund management moves further away from its cap-weighted roots, and as more and more markets suffer from patchy liquidity, checking what internal turnover a newly advertised index strategy generates and what the associated trading costs are is vital.