Early in the development of exchange-traded funds, advisors spent an awful lot of time explaining the essential differences between open-end ETFs and closed-end funds. One of the oft-touted distinctions of first-generation ETFs was their seeming inoculation against premiums and discounts — that is, the price gaps between a fund's market and net asset values.

Open-end ETF advocates argued that premiums and discounts weren't likely to develop because real-time pricing of the portfolios and the indexes (first-generation portfolios were all index funds) broadcasts arbitrage opportunities which would be immediately exploited by institutional traders. Ensuing purchases of cheap assets against the sale of overpriced securities would likely close any pricing gaps.

For the most part, that's indeed how things worked. The indexes used by first-to-market ETF manufacturers were mostly well-known and completely transparent, making it easy to trade baskets of securities emulating their benchmarks. The readiness of index sponsors to create and redeem fund shares upon large traders' orders made arbitrage easy. The mere existence of such a facility, in fact, often put the kibosh on any embryonic premiums and discounts.

Premiums and discounts are a persistent feature of closed-end fund trading because there's no redemption or creation mechanism to translate investor demand for fund shares into an asset. Open-end funds create new shares in exchange for the deposit of assets — either cash or securities — by investors. Thus, the fund's worth is influenced not only by changes in the market value of the portfolio securities, but also by asset inflows and outflows.

Closed-end funds provide no means for investor interaction with the portfolio. The supply of fund shares is limited to the secondary market. Open-end ETFs rely upon a cadre of institutional “authorized participants,” or APs, to keep a fund's portfolio value in line with its share price.

The fund broadcasts the value of its portfolio, at a minimum, every 15 seconds during the market day, together with a description of the fund's “creation basket” — the roster of securities and cash needed to be deposited in a “like-kind” exchange for a block of fund shares.

The per-share portfolio value — often referred to as its intraday indicative value — is disseminated under a unique ticker symbol, usually tied to the fund's ticker. That makes it easy for APs to spot budding arbitrage opportunities. Sometimes, the magic doesn't work. Even the most liquid ETFs' market prices can, at times, veer dramatically from their NAVs.

Take the iShares Barclays Capital Aggregate Bond Fund (NYSE Arca: AGG) as an example. AGG's market price got far ahead of its intraday indicative value during the credit market meltdown of 2008. Normally at such times, APs would be tempted to buy, or to assemble from inventory, the underlying bonds necessary to build creation baskets and then exchange them for fund shares. By selling the newly created fund shares in the secondary market, the share price could then be driven down to approximate the indicative value.

On the other hand, if fund shares traded at a discount to indicative value, APs would be ordinarily incentivized to scoop up the cheap shares in the secondary market, swapping 100,000-share blocks for the expensive portfolio securities, which could then be sold. At the very least, removing supply from the fund share market would bolster prices, squeezing the fund's share price closer to its indicative value.

Bond ETF Premiums

Well, that's the way the things should work, anyway. Arbitrage, however, became a very “iffy” proposition when the corporate debt market imploded. In the autumn of 2008, liquidity dried up as default panic hit the market, literally freezing bond trading over a very cold winter.

Premiums of 1 to 3 percent developed — and persisted — in the funds tracking the Barclays Capital Aggregate Bond Index — a proxy for the U.S. corporate bond market. In addition to the iShares product, the Vanguard Total Bond Market ETF (NYSE Arca: BND) and the SPDR Barclays Capital Aggregate Bond Fund (NYSE Arca: LAG) were particularly beset.

The lack of reliable, real-time pricing in the over-the-counter bond trade contributed to the development of premiums. In addition, the thinness of the market forced trading desks to take much longer to build the large bond portfolios necessary for creation baskets. Holding on to paper for the time necessary to fund a creation unit meant more risk for the desks. Essentially, higher premiums represented compensation.

Bond fund premiums persisted throughout 2009 and have only recently shrunk to more “normal” levels (see below) as credit markets have stabilized.

Flash Crash

Not every instance of market turbulence produces a lasting effect on ETF trading, however. Take the recent “flash crash” of May 2010 as an example. In a harrowing 20-minute interval, as bids evaporated for component stocks and ETF spreads widened, resting sell stops were elected for execution at prices well away from ETFs' intrinsic value. Briefly.

Minutes later, when the value of the constituent stocks could be reasonably relied upon by market makers, ETF spreads narrowed and normal trading resumed.

Of eight iShares products named most frequently in new accounts, including the S&P 500 (IVV), the S&P 500 Growth (IVW), the S&P MidCap 400 (IJH), the S&P SmallCap 600 (IJR), the Russell 1000 Value (IWD), the Russell Midcap (IWP) and the Dow Jones Select Dividend (DVY) products, variances between fund NAVs and their respective benchmarks normalized by day's end. On May 6, the average fund NAV dipped 3.25 percent, pretty much mirroring the 3.26 percent fall in the average index.

Tracking Error

The variance between a fund's price — either its market value or its NAV — and its benchmark is known as its tracking error. The paucity of tracking error in first-generation, open-end ETFs used to be one of the funds' biggest selling points. Because of the low fund expenses involved in operating an index portfolio and a minimal need for a cash hoard to meet redemptions, tracking errors — especially in large-cap domestic equity portfolios — were tiny.

More recently, though, tracking errors have ballooned. A recent report from Morgan Stanley, in fact, showed ETF tracking error widened to an average 1.25 percent in 2009. In 2008, the variance was clocked at only 0.52 percent. It's not just a handful of funds racking up spectacularly big numbers, either. Last year, 54 ETFs exhibited tracking errors exceeding three percentage points. The previous year, only four funds swung so broadly from their benchmark values.

Market volatility (especially in the small-cap space), diversification requirements, and optimization strategies all contributed to last year's spike in tracking error.

The proliferation of ETFs tracking abstruse or illiquid investments has been a factor in the growth of tracking error. For these asset classes, fund managers often use portfolio sampling, or optimization, rather than complete index replication. Using a representative smattering of components, i.e., underweighting smaller, less liquid securities, can cut down on transaction costs and trading delays but increases the odds that the portfolio will vary from the index return.

Case in point: Holding just two-thirds of its benchmark's components, the PowerShares FTSE RAFI Emerging Markets Portfolio (NYSE Arca: PXH) lagged the index by 10.7 percent in 2009 (see Table).

Another bugaboo for funds — especially narrowly focused portfolios — is SEC-mandated diversification requirements. The managers of the Vanguard Telecommunication Services ETF (NYSE Arca: VOX) found more “bug” than “boo” last year. VOX's underlying index is heavily weighted — 49 percent, in fact — in AT&T (NYSE: T), a stock that performed, um, poorly in 2009. While the rest of the domestic equity market soared last year, AT&T ended up slipping 1.6 percent.

The SEC, though, “saved” the Vanguard fund. The agency's rules cap a fund's holdings of an individual stock at 20 percent if the fund wants to earn the sobriquet of a “diversified” portfolio. Requiring an underweight in the laggard stock actually caused the fund to outdo its benchmark by 17.1 percent.

Other regulatory issues can also generate error. Tax rules, for example, required the PowerShares Global Listed Private Equity Portfolio (NYSE Arca: PSP) to periodically liquidate substantial chunks of its holdings last year. As result, the fund limped behind its benchmark by a 13.7 percent margin.

Tracking error clearly can be a good thing or a bad thing. Perhaps it's useful to keep in mind that a fund that overperforms one year is likely to underperform in another. Certainly, variance owing to diversification requirements or other regulatory issues makes fund performance unpredictable from one year to the next. For investors, though, the true measure of a fund's worth is found in its average performance over a period of years. If a fund's average return approximates that of its index, perhaps a little year-over-year dissonance is tolerable.

In any event, investors and advisors need to closely monitor how many of the securities in the benchmark are actually owned by an ETF. A fund with a relatively small ownership stake may be cheaper to trade but its unpredictable return pattern just might trump this advantage.