Has the ETF arbitrage mechanism failed?

The past several months in the markets have brought forth so many events that were unprecedented in most investors' experience that we have become inured. Dividends are being cut left and right for the first time since the early 1970s, some fixed-income funds have lost more than 80% of their value, and the once-proud owner of the world's largest market capitalisation has debt trading at levels we saw only in junk bonds a scant couple years ago. Along with these changes, we have seen one more event that we were told would never happen: ETFs trading at a persistent discount or premium to their net asset value (NAV).

The arbitrage mechanism
The mechanism that keeps ETFs trading close to their NAV is a process called in-kind redemption and creation. Essentially, the fund must publish, every 15 seconds, an up-to-date version of its portfolio, including the vast majority of securities it holds and the amount of cash necessary to buy the rest. The fund also publishes an estimated cash value of those holdings, known as an Intraday Indicative Value, based upon the most recent prices of the securities in its basket. At any point in the trading day, major banks and trading desks known as authorised participants (APs) can come to the fund with a basket of the underlying securities given in the published holdings, which the fund will exchange for a creation unit consisting of a set number of shares in the ETF (typically 50,000).

Similarly, the APs could also buy up shares in the ETF on the market, then exchange them with the fund in return for the published basket of underlying holdings. So if the market price for the ETF starts to rise too far above the price of its underlying stocks or bonds, the APs will buy the underlying holdings, exchange them for shares in the ETF, and sell enough of those shares to drive the price back down to net asset value. Similarly, APs will buy up any shares of the ETF trading at a discount so that they can turn in large blocks of shares for the more expensive underlying securities. This drives prices for the ETFs close to the prices for the underlying stocks and bonds and produces some of the incredible tax benefits of ETFs to boot.

When panic strikes
However, this arbitrage works only if APs can trade the funds' underlying stocks and bonds. When panic strikes, trading can dry up quickly in less-liquid areas of the market, causing market prices for index funds to deviate substantially from the calculated value of their underlying holdings.

The first time we noticed these deviations on a major scale was in October 2008. After a relatively placid August and (by today's standard) minor falls in the stock market through September, early October saw the fallout from the Lehman Brothers explosion begin to settle on the financial landscape. In eight trading days, the S&P 500 fell 23% from 1,166 points to 899. The intrabank lending markets saw even worse devastation, as institutions suddenly refused to trade with one another for fear of counterparty collapse. With this sudden spike in uncertainty over the future of corporate America, corporate-bond markets froze. No one knew what would happen to default rates or interest rates in the future and prices had already fallen substantially, so market participants simply refused to buy any more.

Bond markets are particularly prone to this problem due to the sheer number of securities out there. Each public company has only a share class or two of common stock and perhaps a small handful of preferred shares traded on exchanges, but it could have dozens or even hundreds of different bonds issued at a variety of coupon rates and maturities. Because there's no exchange publishing prices, the corporate-bond market also depends heavily on major broker-dealers who keep a large inventory of bonds on their books for any potential buyer. After the collapse of Lehman Brothers, those broker-dealers were trying to reduce the size and risk of their assets at the same time institutions and investors were trying to sell out of corporate bonds, so the market had no major buyers and no ways for the few willing smaller investors to publicise their buy prices on the myriad corporate bonds available.

However, there were now bond ETFs trading on the stock exchanges. Investors and institutions eager to sell their risky corporate bonds had to settle for hedging their exposure by instead selling the still-liquid basket of bonds on the market. This drove down the market price for the ETFs while the net asset value stood still due to the lack of new prices on the underlying securities. The APs who would normally jump at discounts of 5-10% refused to take on more corporate-bond exposure by purchasing the millions of dollars of shares necessary to redeem for the underlying portfolio, and they couldn't find any buyers for the underlying bonds anyway. Any brave investors who bought the major aggregate-bond ETFs may have bought at a large discount on paper, but they were likely paying a fair price for the basket of bonds due to the sheer uncertainty about future values and the liquidity they were providing. During the few days when panic sweeps over a market like it did in October 2008, net asset values mean very little due to the lack of fresh prices, and frequently traded ETFs provide a better estimate of their holdings' true value than any estimate based on the old prices.

From discount to premium
Smaller, spiky discounts on the major bond ETFs persisted throughout October and into November as stock markets continued to fall and major banks tried to further reduce their corporate-bond exposure in order to decrease leverage. But as the US government began to prop up the bank lending and commercial paper markets and volatility began to ease off, a new trend started: the aggregate-bond indices started to trade at a premium to their net asset values. Once again, this was less a fault of the ETFs than a result of the liquidity in the market for the underlying securities. At the end of 2008, bond markets were calmer as the large spreads on corporate debt seemed to offer sufficient compensation for the newly higher risk of defaults. However, the major banks and trading desks that provide liquidity to the corporate-bond market were still trying to reduce their exposure to risky assets, which entailed lower bond inventories, wider bid-ask spreads, and less trading.

In this environment, institutions and investors willing to take on some of the generously priced risk in corporate bonds would have a hard time assembling a portfolio of actual bonds in a short period, but they could buy a large and fairly diversified portfolio quickly using the liquid fixed-income ETFs. Not only that, but the continued trading of the ETFs through the worst of October and November showed that investors could always get out of their bond indices if they needed to, even if it would be at severely impaired prices. This flexibility doesn't come for free, and anyone looking for the relative security of bonds coupled with a liquid market has had to pay up, as seen in the premium prices on bond ETFs and Treasury bills from December 2008 into February of this year. Once again, the APs have ignored price differences between NAV and market value that they would formerly have swiftly arbitraged. Although newly created ETF shares would be fairly easy to sell at a premium, these major trading desks would take a while to accumulate the large bond portfolio necessary to create those shares in the thinly traded bond market. During that time, they would have to carry a larger inventory of corporate bonds and the corresponding risk of another sudden market collapse. With the higher risks persisting today and the longer holding periods necessary to build up and sell off the bond stake, APs demand a larger premium before they will perform the arbitrage.

Liquidity premium
Once again, the NAV may have been a less accurate estimate of these ETFs' fair value than their market price. Liquidity, the ability to sell out of an asset when everything goes haywire, is a major risk factor even though it is invisible during normal markets. The shares of these ETFs and their underlying bond holdings may seem equivalent, but the ability to sell off the ETF shares in any market makes them intrinsically less risky and thus commands a premium price when investors fear a sudden return to volatility. As bond markets return to normal, the liquidity premium will shrink and ETF investors who bought the fund at elevated prices will find their returns lagging the underlying bonds (and similarly, the net asset value returns). But that is the cost of avoiding being stuck holding the assets in case of total collapse.

Ultimately, the quoted net asset value is only as accurate as the asset prices that feed into it. As markets freeze up or remain illiquid, stale prices can cause an index or quoted net asset value to stray from the market's sense of its "true" value. For hard-to-replicate indices and ETFs in illiquid markets, this can even lead to the index futures or ETFs trading at a persistent premium as investors pile into the more liquid single security and are willing to pay a slight premium over the market value of the underlying assets for the ability to buy and sell their exposure more easily.

These deviations from NAV do not necessarily suggest that the arbitrage mechanism on ETFs has failed, nor that investors are buying ETFs at unfairly high or low prices. Many of the largest ETFs are among the most liquid market-traded securities out there today, with the most frequent trading and deepest order books. Individuals, pension funds, and trading desks alike move masses of money in and out of these funds, setting prices by the minute even when the underlying securities remain in deep freeze. In times of unusually high market risk, the mismatch between market prices and NAVs for some ETFs is not just foreseeable, but it is also probably some of the best evidence we have that these funds are working to provide market participants with more liquidity than ever before available.