As Seeking Alpha readers well know, leveraged ETFs have been vigorously attacked in the financial media. I think it’s unwarranted and have previously posted on Seeking Alpha articles addressing such topics as path-dependency, holding periods, and most recently, tracking error. The latter article indicated that tracking error is not really troublesome here. But that dealt with end-of-day prices. What about other portions of the day? Although leveraged ETF pricing seems to finish fine, how reasonable is the intraday journey from open to close?
Frankly, I started this experiment expecting to see some troublesome issues. Anecdotally, I had noticed that while tracking error seems reasonable based on closing prices, it did appear that daily open prices seemed more likely to be all over the place. But as I examined some data more systematically, it became apparent that I'd have to change my views. Open, high, and low prices can get erratic, but this seems to have nothing to do with leverage. I saw similar patterns for SPY, the penultimate traditional ETF.
Keeping ETF prices in line
There's no regulatory mechanism keeping ETF prices anywhere near the levels of the index they are supposed to track. It would be perfectly legal for the SPY ETF to trade at a large, say 10% or 20%, premium or discount to the S&P 500.
The reason this doesn't happen is because the arbitrage opportunities engendered by such a scenario would undoubtedly be exploited, causing the premium or discount to quickly wither. This is based on the in-kind exchange mechanism inherent to ETFs, which allow major investors to create or redeem ETF units by contributing or taking from the ETF baskets of securities that match the portfolios. If ETF shares trade at a premium, it pays to contribute new securities and get new ETF units that can quickly be sold for a profit. Institutions don't necessarily do this all the time, but arbitragers anticipating such actions can buy or sell ETF units and thereby cause premiums or discounts to wither.
We pretty much know that in the usual course of business, for ETFs that hold reasonably liquid securities, there tends to be little premium or discount when measured in terms of closing prices. The question is how this plays out during the course of the trading day.
It's an important one for leveraged ETFs because derivatives are important components in their portfolios (the sole components in the case of leveraged short ETFs). How erratic are the intraday premiums or discounts? Put another way, how hard do the institutions and/or arbs have to work during the course of a trading day to enable the ETF to close at a price that is at least in the ballpark relative to asset value.
It's easy to notice that however close to NAV closing prices may be, open prices are apt to be all over the place. Table 1 examines price changes between open prices and the prior day's closing levels for the S&P 500, SSO (the ProShares ultra S&P ETF) and SDS (the ultra short offering). The price changes for SSO have been divided by 2 and those for SDS have been divided by minus 2 in order to factor out the targeted leverage and make the figures comparable to those for the S&P 500.
The left-hand side of Table 1 shows that, on average, there isn't a significant difference between percentage changes seen for the S&P 500 and those observed for the leveraged ETFs (other than the ones we factored out, the ones that are supposed to be there based on the intentional leveraging strategy). However, the right side of the Table shows something different. There, we look at absolute values in order to prevent big positive deviations and big negative deviations from cancelling each other out. Now, we see that day one close to day two open price changes are much larger for the leveraged ETFs than they are for the S&P 500.
Table 2 examines a pair of leveraged sector ETFs, for finance, and uses the unleveraged iShares Finance ETF (IYF) as a benchmark. Here, we see large absolute deviations not just for the leveraged ETFs, but also for the unleveraged iShares product.
IYF and the S&P 500 are not fully equivalent for benchmarks in a study like this. The S&P 500 is, obviously, the index being targeted. IYF does not have such status. It's an ETF that, like the ProShares offerings, is targeting a proprietary finance index created by Dow Jones.
I can't recast Table 2 to be exactly equivalent to Table 1 since details of the Dow Jones sector index are disseminated only to those who pay the license fees. But I can examine how Table 1 might look if it were made more comparable to Table 2 by using SPY, the best known S&P 500-tracking ETF. Table 3 compares open-versus prior-day-close price changes for SPY and the S&P 500.
Voila! The average absolute changes for SPY look a lot more like those of the leveraged ETFs than they do for the S&P 500 itself.
This indicates that the more erratic changes we see in daily open prices are most likely based on the nature of an ETF in general, rather than anything special about leverage and the derivatives that are used to implement it.
Before settling on that notion, there's one more thing that needs to be considered. We've seen that on average, the price changes tend to be in line. But we don't know yet how consistent the ordering is. In other words, if there are two price changes for A (2.58% and 0.33%, for example) and two price changes for B (2.44% and 0.61%), did the 2.58% change occur on the same day as the 2.44% change, or did A's 2.58% change match up with B's 0.61% change.
Tables 4 and 5 address this by considering correlation and rank correlation of the price-change observations.
Table 4 shows the same general levels of correlation between SPY, the ultra ETF and the ultra short ETF with the index itself. As to finance, we don't have actual data for the relevant Dow Jones sector index. But we do see that the daily changes for the leveraged ETFs correlate well with those of the iShares unleveraged offering. If you're wondering how the correlations might look for actual, rather than absolute values, the daily price changes between SSO and the S&P 500 are correlated 0.866 and with SDS, it's 0.863. In finance, actual daily changes between UYG and IYF are correlated 0.988 and between SKF and IYF, it's 0.990.
Intra-day high and low prices
Tables 6 and 7 examine intra-day high and low prices, specifically the percentage changes between the high and that day's close, and the percentage changes between the low and that day's close. And recognizing that positive and negative price deviations will likely cancel each other out for the most part, the tables examine absolute values only.
Table 6 considers the S&P 500, SPY and the leveraged ProShares leveraged S&P 500 ETFs, SSO and SDS.
All the price changes are generally comparable and there are no clear patterns that would enable us to say one is more likely to be consistently changeable, on average, than others. Table 7 paints the same picture for iShares IYF and the ProShares leveraged Finance ETFs.
Tables 8 and 9 look at the correlation data and show scores ranging from reasonable to excellent.
Table 9 - Finance ETFs
Based on the data examined here, it appears that leveraged ETFs are more volatile on an intraday basis but only to the extent they're supposed to be: e.g., double leveraged ETFs are supposed to be twice as volatile as their benchmarks. Once we factor out the intentional volatility, we've seen no evidence to suggest that leveraged ETFs are any more volatile on an intra-day basis than is the case with unleveraged funds. In other words, the in-kind exchange mechanism for leveraged ETFs seems to works well enough to keep intra-day trading in line with where we should expect based on the benchmark and the leverage target.
This is a very small study so it cannot serve as the final word for the total universe of leveraged ETFs. But for now, unless or until I or someone else comes up with evidence to the contrary, I believe leveraged ETFs are entitled to rest on a presumption of innocence when it comes to the reasonableness (in light of the leverage assumed) of intraday trading patterns.