Because indices are, to varying degrees, incomplete market portfolios, index construction amounts to active management.2 Providers choose index holdings by size, liquidity, sector, geography, profitability, and the like. Index designs run the full gamut, from highly systematic, rules-based procedures to largely discretionary, committee-based processes. In every case, the explicit selection criteria, weighting rules, and committee decisions directly affect indices’ active shares. Index construction methodologies may seem arcane, but their effects are far from inconsequential.
Capitalization-based indices are inherently biased toward including more liquid, higher-priced growth stocks and stripping low-priced value stocks of their index certification. Conversely, indices that are not price-linked trade into depressed stocks and out of high-flying ones. Over the same 2011–2013 period covered by our internal S&P 500 analysis, we find that securities that our fundamentally weighted RAFI™ indices wish to buy have fallen by approximately 13% more than their peers. Liquidity providers have no incentive to pump the price of index holdings that are not weighted by market cap, because doing so would reduce—not increase—their potential profits.
One of the attractive features of capitalization-based indices is their low market impact. Market capitalization closely tracks liquidity. Moreover, cap-weighted indices are self-adjusting; they do not require rebalancing. The only trades they require are occasioned by index changes. Thus, by design, they minimize this particular type of cost.
Nonetheless, capitalization-based indices should not be viewed as costless, especially in aggregate. Our model indicates it would take over $1.1 trillion in assets for a RAFI index to match the current implicit implementation costs of capitalization-based indices.3
Non-capitalization indices have higher implementation costs on a dollar-for-dollar basis than capitalization-based indices. This we do not contest. But it’s lunacy to believe that the implementation of popular capitalization-based indices is costless, that their negative selection and weighting bias is zero, or that their implicit trading cost as a percentage of aggregate assets is currently below that of well-designed smart-beta offerings.