Blog Posts — September 27, 2018

Beware the GICS Turnover!

In our recent blog post about the upcoming revamp of GICS sectors (see here), we noted that from a risk perspective changes were small, with a few exceptions that were mostly related to the newly formed Communication Services sector.

From a portfolio management perspective, however, some additional turnover may be required if maintaining tight sector weights is part of the mandate. We ran some simple tests to get a sense of how much turnover might be required to maintain those sector weights under the new sector regime. In our tests we created US portfolios that targeted some of Axioma’s risk factors (those that are often used as alpha generators as well). The portfolios were optimized on a single date — August 31, 2018 — to target 3% tracking error versus the Russell 1000, while maximizing (minimizing in a couple cases) exposure to the given factor, and keeping active sector weights based on the old sector designations at 0%. We then used those “initial” portfolios to determine the minimum turnover required to achieve the same risk and return objective, while restricting active weights on the new GICS sectors. Since the underlying alphas and risk components remained the same, the major reason to trade would be sector differences.

Different factors required different levels of turnover, but even the lowest turnover might be cause for concern to some managers because transaction costs would be incurred for no reason other than someone else’s decision to rearrange where companies should be placed. The table below shows the minimum required turnover to move from old to new weights:

If a stock moved sectors and was already held in the initial portfolio, there should be no trading required, except perhaps to fine-tune the portfolio risk. However, if it was not held in the initial portfolio, or was in but at a low weight, it may have to be bought in the updated portfolio to fulfill sector weight requirements. For example, Verizon (in the old Telecomm sector) was not included in the initial Dividend Yield portfolio, but as the new Communication Services sector became a larger part of the index, the optimizer needed to buy the stock, especially as many of the other new additions to the sector had much lower dividend yields. Interestingly, in the Dividend Yield portfolio the optimizer also bought Alphabet, although it does not pay a dividend. Presumably it did so because there was a lot of sector weight to fill, many of the new stocks in the sector do not pay dividends, and because of its size Alphabet was a diversifying bet in the portfolio. This, in turn, allowed for other high yielding names that may have been higher risk to enter or remain in the portfolio. Another example comes from our Momentum portfolio, which was able to take advantage of the higher weight in Communication Services where Momentum has been strong (possibly in anticipation of this upcoming change) to buy or buy more of Alphabet, Netflix and Twenty-First Century Fox. However, it had to sell some Information Technology stocks that also had decent Momentum scores to fulfill that weight requirement.

Across our portfolios, most of the biggest “buys” were in the new Communication Services sector (the exception was in the Earnings Yield portfolio, where the biggest buys were in Consumer Discretionary, and Alphabet and Facebook were both trimmed). The “sells” were mainly in the Information Technology sector (not surprising, as its weight decreased). Consumer Discretionary—which, like Information Technology, saw a number of names move to Communication Services—saw some of both. Most trades were relatively small—less than 1% of the portfolio—although a few, such as a buy for Facebook in the Profitability portfolio, AT&T in the Dividend Yield portfolio and Alphabet in the low Volatility portfolio, were larger (but still less than 2%). Sells tended to be more common but smaller in magnitude than buys.

Of course, these tests assumed that portfolios would maintain complete sector neutrality, while constraints are not typically quite that tight for most managers. The amount of turnover required in our simple tests varied quite a bit based on the strategy and what was initially held in the portfolio, but portfolio managers should expect to incur some turnover, and asset owners should expect some addition turnover costs, based on the upcoming change. Even without a sector-neutrality requirement, we expect to see heightened trading in names in the three affected sectors as portfolio managers rebalance to the new sector weights.

*Many thanks to my colleague Dieter Vandenbussche for this idea and his help in executing it.