PORT 3 – Exchange-Traded Funds: Mechanics and Applications

Almost every ETF investor interacts only with the secondary market. Shares are bought and sold through a brokerage account exactly as a stock would be, and the ETF issuer is not a party to, or even aware of, any individual secondary market trade. Shares simply change hands between one investor and another.

US clearing and settlement

In the United States, every trade executed on a given business day is submitted at day’s end to the National Securities Clearing Corporation (NSCC), which becomes the guarantor of the trade that evening: the buyer’s beneficial ownership is locked in as of execution, regardless of what subsequently happens to the seller. The Depository Trust Company (DTC), of which the NSCC is a subsidiary, keeps the underlying book of ownership at the member-firm level rather than the individual investor level; a brokerage firm is responsible for tracking which of its own clients hold which shares. The DTC then nets all the day’s trades through continuous net settlement, so a firm that both owes and is owed the same number of shares of a security settles net rather than gross. The NSCC has two business days to complete this reconciliation, the origin of the familiar T+2 settlement convention that covers the large majority of ETF trades. Market makers, who by the nature of their role are more likely to end a day genuinely short, are given up to six days to settle, reflecting the additional time it can take to create or borrow shares.

European trading and settlement

European ETF ownership is dominated by institutional investors, and the market itself is fragmented across multiple exchanges, jurisdictions and clearing systems. Fundamentally the process is the same: an investor buys from a market maker or other counterparty in the secondary market, while APs use the creation/redemption mechanism to keep the ETF’s price aligned with fair value. Much of the volume, however, takes place through negotiated OTC trades between large institutions that are reported after the fact rather than displayed as live quotes beforehand. Settlement is cleared to one of 29 separate central securities depositories, adding structural complexity that has no direct bearing on an individual investor’s rights but tends to widen local spreads and raise local trading costs.

Shorting an ETF

Because ETF shares trade continuously through a brokerage account like any listed security, an investor can also sell them short in a margin account, borrowing shares to sell now with the intention of buying them back later at a lower price. Mutual fund shares offer no equivalent: they can only be purchased from, or redeemed with, the fund itself at NAV, so there is no borrowable share for a short seller to sell in the first place. The ability to short is one more expression of the underlying structural difference this reading opened with, an ETF behaves like exchange-traded equity in every respect that matters to a secondary market participant, while a mutual fund behaves like a claim on the manager.

Whether an ETF delivers on its promise comes down to three linked questions: what it costs, how closely it tracks its benchmark, and what happens at tax time. Expense ratios are the most visible of the three and the easiest to compare, but on their own they understate the true cost of holding the fund.

Why ETF expense ratios run low

ETFs generally undercut mutual fund fees for structural reasons rather than through aggressive discounting alone. Because ETF shares are held and transacted through brokerage accounts, the issuer does not maintain individual investor records or communicate directly with retail holders, both meaningful costs for a mutual fund complex. Because most ETFs are index based, they also avoid the security and macroeconomic research overhead that active management carries. The remaining variation in cost across funds comes down to portfolio complexity, how many securities are held and how often the portfolio rebalances, issuer scale, and the intensity of competition in a given category. By the end of 2018, expense ratios for the most heavily used, broad-based, capitalization-weighted indexes had fallen as low as 0.03% for US equities, 0.11% for emerging market equities, and 0.04% for US bonds, figures that sit well below the category averages once complex and specialized funds are excluded from the comparison.

Daily and rolling tracking difference

As of December 2018, index-tracking ETFs represented 98% of the US ETF market measured by assets under management, so assessing how faithfully a fund follows its benchmark is central to evaluating it. The simplest measure is the daily tracking difference, the one-day gap between the ETF’s NAV return and its index return. A more informative measure looks at rolling holding periods, most often 12 months, because it shows the cumulative effect of fees and management decisions over a period long enough to be meaningful to an actual investor, and it can be compared directly against the fund’s expense ratio.

Tracking error is a related but distinct statistic: the standard deviation of the daily differences between fund and index returns, ordinarily reported over a trailing 12-month window. It measures the dispersion of the daily gaps, not their direction, so a fund with a small tracking error could still be persistently under- or overperforming its benchmark by a stable amount every day. Tracking error should therefore be read alongside a central-tendency measure such as the mean or median rolling difference, never in isolation.

EEM’s standard deviation of daily performance differences against its index, its reported tracking error over the same period, was 0.012%.

Funds tracking the identical benchmark can post very different results because tracking error accumulates from several independent sources, each operating through a different mechanism.

• Fees and expenses. An index is calculated assuming frictionless trading at the closing price. A fund’s operating costs are a direct, continuous drag relative to that idealized benchmark.
• Representative sampling or optimization. Rather than fully replicate a large or illiquid benchmark, a manager may hold a subset of constituents chosen to approximate the index’s risk and return characteristics. GMF is the extreme case already discussed: as of 30 November 2018 its median constituent market capitalization was $2.8 billion against the underlying index’s $0.695 billion, meaning the fund’s sampling systematically omits many of the smaller names. GMF will tend to underperform when emerging market mid-caps and small-caps outperform large-caps, and to outperform in the opposite regime.
• Depositary receipts and other ETFs. Where local shares are illiquid, a fund may hold depositary receipts instead. The receipts trade in a different market with different hours from the local shares the index prices, and any ETF holdings the fund carries are valued at their own closing price rather than at NAV, both sources of daily discrepancy.
• Index changes. An index provider periodically resets constituents or weights; the fund manager transacts these changes at prices that may differ from those the index methodology assumes, though cooperating with APs to trade rebalances at the market-on-close price can narrow this gap considerably.
• Fund accounting practices. Valuation conventions differ. Some issuers strike currency valuations using the WM/Reuters rate set at 4:00 p.m. GMT, while others strike at the close of ETF trading, and fixed-income portfolios are often valued at the equity market close even though bond indexes may value the same securities at an earlier time drawn from bond market convention. Each convention gap is a source of daily divergence, most visible in volatile long-dated bond segments.
• Regulatory and tax requirements. A fund may face withholding or other tax treatment that an index provider’s return series does not fully replicate. Brazil, for years, taxed foreign investment inflows; the tax did not touch local closing prices and so never entered the index calculation, yet non-local ETFs paid it on every Brazilian purchase, a persistent drag the index itself never carried.
• Asset manager operations. Issuers can generate income the index calculation does not capture, chiefly through securities lending and foreign dividend recapture. These operate as a negative cost, offsetting fund expenses and, in a well-run program, sometimes covering them entirely.

Two separate tax questions apply to every ETF: whether the fund itself distributes capital gains to shareholders during the year, and what happens when an individual investor sells shares. The two are unrelated; a fund’s efficiency at avoiding gains distributions has no bearing on the tax treatment an investor faces at the point of sale.

On the first question, funds are generally required to distribute any capital gains realized over the course of the year, most often in a single distribution at year-end, though some funds distribute quarterly or on another periodic schedule. ETFs are described as both tax fair and tax efficient relative to mutual funds, and it is worth keeping those two labels distinct, because they describe two different mechanisms rather than one advantage stated twice.

Tax fairness

In a traditional mutual fund, an investor’s redemption is met with cash, and the fund often has to sell portfolio securities to raise it. If those securities have appreciated, the sale realizes a capital gain that is distributed to every remaining shareholder, so investors who never traded can still owe tax triggered by others exiting the fund. An ETF investor, by contrast, sells shares to another investor in the secondary market; the fund’s portfolio is untouched, and the transaction is invisible to the issuer. Even when an AP redeems shares directly with the issuer, that in-kind exchange is not a taxable event in markets that permit it. This structural separation, where one shareholder’s activity cannot create a tax bill for another, is what makes ETFs tax fair.

Tax efficiency

Tax efficiency is a related but separate advantage, arising from what the issuer can choose to deliver into a redemption basket. When an AP redeems shares, the issuer selects which underlying tax lots to hand over, and issuers commonly favor the lots with the lowest cost basis and therefore the largest unrealized gain. Delivering those out of the fund raises the average cost basis of what remains, continuously trimming the portfolio’s embedded unrealized gains without the fund ever having sold anything in the open market. Tax lot management of this kind is a direct byproduct of the in-kind redemption mechanism and has no equivalent in a cash-based mutual fund structure.

Distributions and sale-side treatment

Dividend and other security-level distributions still generate tax liabilities that vary by jurisdiction, fund domicile and legal structure, and an investor needs to understand each fund’s specific treatment. On sale, most jurisdictions tax an ETF according to the character of its underlying holdings; a US equity or bond ETF, for example, carries through the same capital gain, dividend and return-of-capital rules that apply to its constituents. Structural nuances remain, however. In the United States, an exchange-traded note tracking a commodity index is taxed differently from an ETF that holds the same exposure through commodity futures contracts, a distinction with real economic consequence that has nothing to do with which product tracks its benchmark better.

Beyond the expense ratio, an ETF investor pays a commission and a trading, or market impact, cost driven by the fund’s bid–ask spread, the size of the trade relative to normal volume, and how easily the market-making community can hedge the position. The market structure and liquidity of the underlying securities is the single biggest driver of that spread: fixed-income securities, which trade in a dealer market, carry wider spreads than large-cap equities, and an ETF holding securities that trade in a different time zone reflects that mismatch too.

The quoted spread for a particular transaction size is generally no wider than the sum of the following components, some of which can also work in the fund’s favor:

• Creation/redemption fees and other direct trading costs, such as brokerage and exchange fees, which can add to or subtract from the spread
• The bid–ask spread of the securities the ETF holds
• Compensation to the market maker or liquidity provider for the risk of hedging or carrying the position for the rest of the trading day
• The market maker’s desired profit margin, itself constrained by competition
• A discount reflecting the likelihood that an offsetting ETF order will arrive quickly, which narrows the spread

In highly liquid, actively traded ETFs, buy and sell orders match so quickly that the creation/redemption process is rarely invoked at all, so the first three components barely register and the quoted spread can end up tighter than the spread on the underlying securities themselves. The quoted spread, however, is calibrated to a specific, usually modest, trade size; a transaction larger than roughly 10% of average daily volume is generally better handled through negotiation with a capital markets desk than by simply crossing the posted spread.

Data as of 7 November 2018, previous 12 months for premium/discount figures. Source: FactSet.

Data as of 7 November 2018, previous 12 months for premium/discount figures. Source: FactSet.

*Includes currency ETFs in addition to the categories listed; total currency ETF assets were $1,799 million. Data as of end of December 2018, based on 60-day averages. Source: FactSet.

The pattern across the table follows directly from what drives a spread in the first place. US equity and US fixed-income ETFs, the two deepest, most heavily hedged categories, post the tightest asset-weighted spreads, 0.03% and 0.02% respectively, because their sheer scale of AUM, $1.87 trillion and $590 billion, brings continuous two-way order flow. International equity and international fixed-income categories widen out to 0.05% and 0.06%, reflecting underlying markets that are only partly open during US trading hours and harder to price simultaneously. Longer-horizon, lower-turnover categories such as asset allocation and alternatives carry both smaller asset bases and the widest spreads in the table, 0.21% and 0.18%, because they depend more heavily on market makers sourcing liquidity through the underlying securities rather than on a deep pool of ongoing ETF order flow.

At the end of each trading day an ETF has a closing NAV, calculated from the closing prices of its underlying holdings, against which its closing market price can be compared. During the trading day exchanges also disseminate an indicated NAV, or iNAV, an intraday fair-value estimate built from the day’s creation basket. The ETF trades at a premium when its price sits above iNAV and at a discount when it sits below.

Timing differences and stale pricing

NAV is a reliable fair-value estimate when the ETF trades on the same exchange, and during the same hours, as its underlying holdings. It becomes a weaker estimate when it does not. A US-listed ETF holding European or Asian stocks calculates NAV from closing prices set hours before the ETF itself stops trading, so the fund’s closing price on the US exchange, reflecting information that arrived after the foreign market closed, is not a mispricing so much as the market’s best current estimate of where the foreign holdings would trade if their market were still open. Fixed-income ETFs face a related problem from the other direction: because bonds do not trade on an exchange, NAV relies on dealer bids or pricing-service estimates rather than a true closing print, and in stressed markets, when few bonds actually trade, those estimates can go stale, in which case the exchange-traded, continuously priced ETF may be a better read on current value than its own NAV. ETFs that themselves trade infrequently add a further layer: comparing a last trade from, say, 1:00 p.m. against a 4:00 p.m. NAV calculated after the market has since moved will show a premium or discount that reflects nothing more than the passage of time between the two prices.

The components of total ownership cost

Trading costs, commissions, bid–ask spread and any premium or discount paid, are incurred once at purchase and once at sale, regardless of how long the position is held. Management fees, tracking error, portfolio turnover and security lending income accrue continuously and scale with the length of the holding period. The practical consequence is that the relative importance of the two cost categories flips as the holding period lengthens: trading costs dominate for short holds, and the management fee dominates for long ones.

The practical implication follows directly. Broad, low-fee, capitalization-weighted equity index ETFs tend to make trading cost the largest single component of ownership cost, so a short-horizon tactical trader should weight liquidity and spread tightness over management fee when choosing between similar funds, while a long-horizon buy-and-hold investor should weight the management fee more heavily, since it is what accumulates.

ETF structure introduces risks beyond ordinary market risk, arising from what the fund holds, how it is legally organized and how investors use it.

Counterparty risk

Some exchange-traded product structures depend on a counterparty rather than on holding underlying securities outright. Exchange-traded notes (ETNs) are the clearest case: despite trading on an exchange with a creation/redemption process of their own, ETNs are not funds. They are unsecured debt obligations of the issuing institution, promising to pay a return pattern tied to an index, less fees, with the issuer responsible for whatever hedging it judges necessary to meet that promise. In the United States, ETNs are registered under the Securities Act of 1933 rather than under the fund-regulation framework that governs ETFs, precisely because they are general obligation debt securities of a bank and are not a pooled investment managed by a fund company for a fee; comparable ETN structures exist in most other markets where ETFs themselves are listed. Because they are registered as bank debt rather than as fund shares, their counterparty risk in a default is, in theory, total: an issuing bank’s bankruptcy could render its ETNs worthless. Because ETNs can be redeemed back to the issuer at NAV, only an extremely fast, catastrophic failure would be likely to trap investors; that scenario played out once, in 2008, with three Lehman Brothers-backed ETNs, and has not recurred since. Where an ETN’s or a deposit-based currency ETF’s risk sits with a specific bank, that bank’s credit default swap spread offers a rough, continuously updated gauge of the market’s view of default risk, though a CDS spread should never be read as a literal probability of default over the coming year; a one-year CDS rate above roughly 5% is generally treated as a signal that meaningful default risk has entered the picture.

A related but distinct exposure, settlement risk, arises whenever a fund uses OTC derivatives such as swaps to obtain its market exposure, common in European swap-based (synthetic) ETFs, geared or leveraged funds, some currency funds and some active strategies. Mark-to-market gains on an open swap are exposed to the counterparty defaulting before the next settlement. Frequent settlement, typically daily or weekly, and collateral held in low-risk instruments such as US Treasury bills at a custodian bank both limit, without eliminating, the exposure between settlement dates.

Security lending

ETF issuers, like traditional fund managers, commonly lend a portion of the portfolio’s holdings to short sellers, generating a fee and interest income on posted collateral, usually short-term fixed-income instruments, that supplements the fund’s return. Loans are overcollateralized, generally to 102% for domestic securities and 105% for international securities, which keeps counterparty default risk low. As of the writing of this reading, no ETF shareholder had ever suffered a loss traceable to a securities-lending program since ETFs were first introduced in the early 1990s, and most issuers pass the resulting income back to shareholders in full, though disclosure of individual lending programs is not always complete.

Fund closures

Issuers close ETFs for the same general reasons mutual fund complexes close funds: regulatory change, competitive pressure and corporate activity. In 2018 the Israeli securities regulator banned the ETN structure outright, forcing more than 700 products to close and reopen as conventional ETFs. Funds that fail to gather sufficient assets against a growing field of competitors are shut down by their own issuer, and mergers between ETF providers routinely lead the acquirer to close underperforming funds in favor of higher-growth products. A closure typically means the fund sells its holdings and returns cash, an event that can trigger a taxable gain for the investor at exactly the moment a replacement holding must also be found.

Short of a full closure, issuers sometimes impose a “soft” closure: a creation halt or a change in investment strategy that leaves the fund technically open while altering what it delivers. ETN issuers halt creations when they no longer want to add debt to their balance sheet against a given index; UBS did exactly this in September 2018 with its ETRACS Monthly Pay 2x Leveraged Mortgage REIT ETN (MORL), suspending new sales from inventory and blocking new share creation. With the arbitrage mechanism disabled on the creation side, MORL went on to trade at a premium of more than 5% to its indicative value, since APs could no longer create fresh, fairly priced shares to sell into excess demand. Separately, issuers sometimes repurpose a low-asset fund by changing its underlying index rather than closing it outright and launching a new one, a common industry practice that can range from a minor portfolio adjustment to a wholesale change of country, industry or asset class exposure.

Leveraged and inverse ETFs promise a multiple, positive or negative, of their benchmark’s daily return, typically delivered through swaps. The word daily is doing all the work in that sentence: because the fund must reset its notional exposure at the end of every trading day to keep the multiple constant going into the next day, the return an investor experiences over any period longer than one day is not the daily multiple applied to the period return, and the gap between the two grows with volatility.

Why compounding erodes the multiple over time

Daily resetting means each day’s return multiple is applied to a NAV that already reflects the prior day’s move, so returns compound rather than simply add across the multiple.

Because of this compounding effect, leveraged and inverse ETFs are generally unsuited to being held for more than about a month without active rebalancing to realign notional exposure with the investor’s intended net position. Beyond the mechanics of leverage itself, these products sit at the center of what is best described as expectation-related risk: an investor who does not fully understand how a complex product is built can be surprised by outcomes that are, in fact, exactly what the fund’s own methodology promised. Index methodology and the fund’s construction approach, not just its headline exposure, need to be understood before the product is used.

Institutional and retail investors alike use ETFs across passive, systematic rules-based and fully discretionary active strategies, spanning both top-down macro positioning and bottom-up security selection. Uses cluster into three broad categories: portfolio efficiency, asset class exposure management, and active or factor investing. Not every strategy suits the structure; a manager running a concentrated or proprietary approach may find the daily disclosure of holdings that ETFs require incompatible with protecting the strategy, and the underlying securities must be liquid enough to support daily creation and redemption.

Portfolio efficiency

ETFs are frequently used for operational rather than directional purposes. Cash equitization uses an ETF to invest excess cash quickly, keeping a portfolio close to its target benchmark exposure and minimizing cash drag, the mis-tracking that results from holding uninvested cash. The same tool handles small, recurring cash flows from dividends or shareholder activity more cheaply than transacting directly in the underlying securities. For rebalancing, a liquid ETF with a tight spread lets a manager restore a portfolio to target weights in a single trade rather than by adjusting many individual positions, particularly useful for tight rebalancing thresholds and frequent rebalancing schedules. Completion strategies use an ETF to fill a temporary gap in exposure, for instance when an active manager’s current positioning has drifted away from a market segment the asset owner still wants represented, without disturbing the manager’s own mandate. Transition management applies the same logic to a change of external manager: a newly appointed manager can hold an ETF that tracks the vacated exposure while gradually selling the outgoing manager’s unwanted positions and building the new portfolio, and an asset owner can similarly use ETF holdings to bridge, or even to fund, a manager search. Very large asset owners sometimes forgo ETFs for this purpose in favor of a dedicated separately managed account, which can be cheaper at scale, more customizable, and free of the public disclosure requirements that apply to large ETF positions.

Asset class exposure management

The core, longest-standing use of ETFs remains gaining index exposure to an asset class or sub-asset class, equities, bonds, commodities and currencies alike, on a strategic, tactical or dynamic basis. Fixed income has become one of the fastest-growing areas of institutional ETF use since the 2008 financial crisis, largely because reduced dealer balance-sheet capacity has made trading bond portfolios directly more difficult and costly than accessing the same exposure through an ETF. Tactically, advisers and institutions use ETFs, including thematic ETFs that isolate a narrow market segment such as cybersecurity or robotics not well represented by a conventional industry classification, to express shorter-term views, and the most heavily traded ETF in a category is generally preferred for this purpose because trading cost and liquidity, not the management fee, are what matter for a short holding period.

Active and factor investing

Smart beta, or factor, ETFs weight or screen constituents by rules tied to a specific return driver, such as value, dividend growth, quality, low volatility or momentum, rather than by market capitalization, and because the weighting is rules-based rather than discretionary, holdings can be disclosed daily without compromising a proprietary process. Global assets in smart beta equity strategies, combining single- and multi-factor approaches, now represent approximately 20% of ETF assets, a substantial share, while fully discretionary active ETFs, concentrated mainly in fixed income, remain a much smaller 2% to 3% of global ETF assets. A multi-factor approach that blends several return drivers typically shows lower return volatility than a single-factor approach, at some cost to the potential upside from timing any one factor correctly. Beyond pure return-seeking, factor and alternatively weighted ETFs are also used defensively: low-volatility equity ETFs target a lower-risk return profile within a market segment, and fixed-income ETFs that pair a long credit position with a short duration hedge let an investor isolate credit risk while minimizing sensitivity to interest rate moves.

Dynamic, top-down multi-asset and global macro strategies round out the picture. Because ETFs exist across most major equity, bond, commodity and currency benchmarks, an asset manager or asset owner can implement a discretionary allocation call, or a systematic risk- or return-based process that rebalances positions back to target weights, entirely through ETFs, alongside futures and swaps where those are more efficient to trade. Whatever the application, using an ETF well starts with the same discipline: understanding the index construction rules, the costs, the risks and the performance history behind the specific fund, not just the ticker.