A pioneer of the alternative indexing world weighs in with a minimum-variance product. Martin Steward talks to Rob Arnott about his latest innovation

Rob Arnott, chairman and founder of Research Affiliates, can fairly claim to be the father of the ‘smart beta’, alternative indexing trend that has swept through institutional investment over the past five to six years.

The concepts often date back decades, but there was limited commercial traction for anything that departed from market cap-weighted benchmarks before Arnott started spreading the word about the Fundamental Index concept via Research Affiliates. The inexorable rise of ETFs, and the affiliates model whereby Arnott’s firm licenses its products to big-name asset managers, created the perfect platform for the ideas to go global.

“I feel like I’ve had the privilege to be a catalyst in this whole area of alternative beta, opening doors for other innovators,” Arnott says.

So when this innovator decides to tackle an area of ‘smart beta’ outside of his home turf of fundamental indexation – where stocks are weighted, not by their market capitalisation, but by their economic footprint as defined by sales, cash flow, dividends and book value – it tells us something important about either fundamental indexation, the ‘smart beta’ he is looking to improve, or both.

“The ‘product innovator’ is the Apple-type business model,” says Arnott. “The more common Microsoft-type model is the ‘fast follower’, where you see an area in the market where you think you can do things better. Research Affiliates is 80% Apple and 20% Microsoft.”

During 2010 and 2011 the 20% Microsoft part was scrutinising the low-volatility equities concept. Why zero in on this over the many other ‘smart betas’ on the market?

“When I first looked at low-vol some years ago I thought it was a bit of a gimmick,” Arnott admits. “But as we began to look more seriously we realised that it wasn’t a gimmick at all, but a very important tool for investors, especially when bond yields are negative in real terms. With most strategies we didn’t see obvious and powerful ways to improve upon them – but with minimum variance as it is currently practised, we did.”

There are as many ways to practise low-volatility equities as there are providers, but Arnott splits them into two basic categories. The ‘heuristic’ camp simply carves out a lowest-beta chunk of the market and then weights them with the simple 1/beta or 1/volatility algorithm – an example is the S&P Low Volatility index family. The ‘optimised’ camp tries to build low-volatility portfolios using mean variance optimisation – as is the case for the MSCI Minimum Volatility index family.

Both have the same ultimate objective – low beta; both pursue it elegantly; but both, whether created as an index or an active strategy, also result in pronounced sector concentrations, a tilt towards smaller, less liquid stocks, and a resulting lack of capacity. 
Research Affiliates’ finding is that when the 1/beta weighting is combined with the fundamentals weighting, with its tilt towards economic size, these problems melt away. In the case of the RAFI US Low Volatility index, the weighted average market cap of the portfolio is brought back up to the level of the Russell 1000’s; and the 90-day average daily volume is brought up to a respectable 8.5m shares, versus the Russell 1000’s 12.3m.

“All of a sudden you have vast capacity,” says Arnott.
But there is more to this than simply removing the obstacles to greater scalability – after all, most low-volatility practitioners constrain their portfolio optimisation to deal with these scalability issues, usually by putting limits on individual stocks, sectors or countries.

Research Affiliates’ claim is that correcting these biases using fundamental weights is much better than the current practice of imposing stock, sector or country constraints that, because they are derived from the market cap-weighted benchmarks, re-introduce the fatal market cap-weighting system back into the process. If you believe in the inferiority of market cap-weighted indexation, you have to believe in the inferiority of introducing cap weighting-based optimisation constraints.

“Every low-vol strategy combines its basic weighting algorithm with an alpha engine,” says Arnott. “Most practitioners believe that their source of alpha – their risk-adjusted added value – is an anomaly in the market where the capital market line between beta and return is flatter than it should be or even inverted. That’s certainly a fact, but it’s only part of the picture.”

Almost all low-vol strategies also have the same alpha engine as the Fundamental Index strategy, he argues: the weighting mechanism that they use does not contain price, which means that they weight against major price movements.

“But current practitioners don’t seem to be aware of this – and if you have an alpha engine that you’re not aware of you are going to make mistakes that compromises or gives away some of the alpha,” he reasons. “As well as starting with a selection universe that is determined by cap-weighted benchmarks, they put in sector constraints tied to the cap-weighted index and turnover constraints that create price drift that reproduces the price dynamic of cap-weighting. This represents a wonderful opportunity to avoid those mistakes and build a better mousetrap.”

By using Fundamental Index universes and weights, Research Affiliates can impose the sector constraints that a low-volatility strategy needs without re-introducing the link back to price. Both alpha engines remain intact.

The backtests that result from this suggest that RAFI Low Volatility delivers all the benefits of liquidity and capacity alongside similar returns to typical low-volatility strategies (2-4 percentage points in excess of cap-weighted benchmarks) and similar volatility (25-28% lower than the cap-weighted benchmarks). But performance attribution also suggests a good deal of the improvement over the cap-weighted benchmarks comes from the Fundamental Indexation side of the equation.

Take the 300 lowest-beta stocks from the Russell 1000 index, for example, and returns between 1967 and 2011 improve upon the full index by about 90 basis points. Take the 300 lowest-beta stocks from the standard RAFI 1000 Fundamental Index – but continue to weight them by market cap – and you add another 25 basis points. This shows that, even before you begin to weight your low-beta stocks fundamentally, simply selecting those stocks from a non-cap weighted universe makes a difference.

This is an important observation, as there exist products that blend low-volatility and fundamental weighting but still select from a cap-weighted universe. Standard & Poor’s recently rolled out its S&P Global Intrinsic Value indices (GIVI), for example, which begin by excluding the 30% of market capitalisation with the highest beta and only then weight that sub-universe by ‘intrinsic value’ – a combination of book value and projected earnings.

So now we have just over 100 basis points of excess return from selecting the low-beta stocks from a fundamentally-weighted universe. The next step, re-weighting those stocks by fundamentals, sees excess return explode by another 100 basis points. Take the final step and re-weight those stocks by both 1/beta multiplied by fundamentals, and yet another 80–90 basis points are added, taking you to a full 3.2 percentage points excess return over the Russell 1000. About 2.1 percentage points of that comes from selecting low-beta and then re-weighting by low-beta; leaving about 1.1 percentage points from selecting and re-weighting on fundamentals.

“But when we take the strategy to emerging markets, the picture is different,” notes Arnott.

Very different, in fact. Here, selecting the lowest-beta stocks from the cap-weighted index actually hurt returns between 1999 and 2011, to the tune of 250 basis points. Moreover, the final step of introducing the 1/beta re-weighting scheme also trims 50 basis points off the massive, 14 percentage point gains in excess return achieved by selecting and weighting fundamentally.

These are likely to be controversial results, given the widespread claims for the persistence of the low-volatility effect across asset classes and markets, and indeed the prevailing behavioural-finance hypotheses deployed to explain it. But they certainly indicate that the combination of low-volatility and fundamental weighting is more powerful than either one alone – both operationally and for long-term performance.
So does this mean that RAFI Low-Volatility will supersede the core RAFI indices, among the most successful of all the ‘smart beta’ products? That’s not how Arnott sees it.

 “I don’t tend to think of this as a combination of the Fundamental Index strategy and low-beta,” he explains. “I think of this as better low-volatility. I view the Fundamental Index approach as a core index spanning the total macro-economy. Fundamental low-beta does not span the whole economy. For that reason I don’t see this new strategy as a better version of the Fundamental Index concept because they each serve different functions. I can’t imagine the market abandoning a comprehensive solution, whether it be market-cap or the Fundamental Index strategy, in favour of low-vol in any flavour.”

In other words, Arnott sees the Fundamental Index strategy as a better way than market-cap of buying the market; low-volatility is not just about re-weighting the market according to beta, but also about selecting the low-beta sub-sector of the market. By combining the two strategies, Research Affiliates not only introduces a compelling new product, but raises important questions about what constitutes a representative index in the brave new world of ‘smart beta’.