Sources of ETF tracking error

Sources of ETF tracking error

Published September 28, 2026

Sources of ETF tracking error

On this page

Learning Outcome Statement

Describe sources of tracking error for ETFs.

On the exam: You get a client mandate, benchmark, ETF replication method and implementation facts such as fees, sampling, cash, taxes, rebalancing and pricing timing. You choose which facts create, reduce or do not constitute tracking error in an item set or MCQ.

An ETF can hold the right market exposure and still fail to match its benchmark return. The gap is created by what happens between the portfolio and the return you measure, including costs, cash, sampling, rebalancing and timing. You care because the exam separates a persistent shortfall from unstable active returns.

SourceMechanismUsual effect
FeesFund expenses reduce returnUsually negative
Transaction costsTrading consumes portfolio valueNegative, sometimes variable
TaxesTax payments reduce returnNegative versus pre-tax benchmark
Cash dragCash misses market exposureNegative when markets rise
SamplingHoldings differ from benchmarkDifference and variability
RebalancingTrades follow index changesTemporary difference
Timing mismatchCutoffs measure different movesMeasured variability
Securities lendingLending income raises returnOffsets negative effects
Efficient implementationTrading reduces slippageReduces negative effects

Fees, taxes and cash usually pull the ETF below a pre-cost benchmark, while sampling, rebalancing and timing mismatches can make the difference move across periods. Lending income and efficient implementation work in the other direction. Read these effects on two separate dimensions: a recurring drag shifts the average active return, while a changing exposure or cutoff can make that return less stable. The next table groups each source by its usual direction and its behavior across periods; a source can matter on one dimension without having a fixed sign on the other.

SourcesAverage resultPeriod-to-period result
Fees, taxes and cashUsually lower ETF returnOften persistent shortfall
Sampling, rebalancing and timing mismatchesNo fixed directionGap can vary
Lending income and efficient implementationHigher return or lower trading lossNegative gap can narrow

First compare the ETF and benchmark returns over the same period and on the same basis. Call the ETF return in period $t$, $R_{\mathrm{ETF},t}$, and the benchmark return over that same period, $R_{\mathrm{BM},t}$. Their active return, $\Delta R_t$, is the difference the exam wants you to inspect:

$$ \text{Tracking difference}t = \Delta R_t = R{\mathrm{ETF},t} - R_{\mathrm{BM},t} $$

Let $\sigma$ denote the standard deviation across the observed active returns. Tracking error is:

$$ \text{Tracking error} = \sigma(\Delta R_t) $$

Tracking difference describes the level of active return. A fee that repeats can create a negative tracking difference without adding much tracking error. Variable trading costs, sampling, cash, rebalancing and timing can raise tracking error because they make active returns less stable. When the question asks for average shortfall, use the average active return across the stated periods. Do not replace that average with its standard deviation.

The distinction is easier to hold if you follow the return path from implementation to measurement.

Loading diagram…

Worked example

You are reviewing an equity ETF for a taxable institutional client. The mandate seeks to match a broad equity benchmark, with a constraint against unnecessary cash. The ETF uses representative sampling, pays management fees and trading costs, uses a different pricing cutoff from the benchmark, and earns securities-lending income. The exam asks which facts create tracking error and which fact offsets it.

FactStated condition
InvestorTaxable institutional client
ObjectiveMatch broad equity benchmark
ConstraintAvoid unnecessary cash
ReplicationRepresentative sampling
CostsFees and trading costs
PricingDifferent benchmark cutoff
OffsetSecurities-lending income

Read the facts as a chain from mandate to measured return. The objective sets the comparison, while the cash constraint identifies one possible drag. Sampling changes the holdings, and fees and trading costs affect implementation; the different cutoff affects measurement. Lending income can offset some negative effects. In the figure, trace how these implementation choices flow into active returns, then distinguish the average gap from its variability.

  1. Record the objective and constraint before classifying any return effect. This gives you two tests: benchmark-relative return and avoidance of unnecessary cash. It produces two anchors for the review, because the ETF can match the benchmark and still lose return through cash drag.

  2. Compare the ETF’s sampled holdings with the benchmark holdings to classify representative sampling. The comparison matters because different holdings can create a return gap whose direction changes. It produces one tracking-error source, even though it does not produce a fixed positive or negative return.

  3. Record fees, trading costs, cash and the different pricing cutoff as four separate implementation effects. This separates return drains from measurement timing, because fees and costs can lower return, cash can create drag, and the cutoff can move the measured gap when markets move. It produces four gap drivers.

  4. Record securities-lending income as the offset and check whether efficient implementation reduces slippage. This matters because lending income raises ETF return, while efficient implementation can reduce slippage. It produces one offsetting effect, not a new cost.

The conclusion is a likely negative or unstable active return from sampling, costs, cash and timing, partly offset by lending income, so the mandate should monitor both the persistent gap and its variability.

Changed condition

Now change only the replication method. The ETF uses full replication rather than representative sampling, while the objective, constraint, costs, cash, timing and lending income stay unchanged. Predict the effect before reading the table.

FeatureBeforeAfter
ReplicationRepresentative samplingFull replication
Holdings mismatchMore likelyLess likely
Sampling effectMore variableReduced

Your prediction should be a smaller sampling-related contribution to tracking error because the ETF holdings now follow the benchmark more closely. The change does not erase fees, transaction costs, cash drag, rebalancing or timing mismatches. Before reading on, mark which unchanged facts still push the ETF below the benchmark.

Common error

A candidate sees the ETF return below the benchmark and labels that gap tracking error because the fund failed to match its target. That feels right because both measures involve the ETF and benchmark, but a persistent average shortfall is tracking difference, not automatically tracking error. The question asks for average shortfall or variability of active returns, or compares ETF price with NAV at a single instant, which signals a different test.

Check yourself

Practice

An equity ETF is designed to match a broad benchmark. It uses representative sampling, holds cash for operations, pays management fees and trading costs, faces a different pricing cutoff from the benchmark, and earns securities-lending income.

Question 1

Which stated fact most directly creates a persistent negative tracking difference?

Question 2

Which stated fact is most likely to increase the variability of active returns?

Question 3

Which statement describes tracking error rather than tracking difference?

Question 4

Which stated fact can offset the negative effect of fees and trading costs?

An ETF and its benchmark use different market-close cutoffs. Which source of measured tracking error is most directly involved?

Revision snapshot

Tracking difference is the ETF return minus the benchmark return. Tracking error is the variability of that active return.

$$ \Delta R_t = R_{\mathrm{ETF},t} - R_{\mathrm{BM},t} $$

$$ \text{Tracking error} = \sigma(\Delta R_t) $$

Fees, transaction costs, taxes, cash drag, sampling, rebalancing and timing mismatches can cause ETF returns to diverge from the benchmark. Securities-lending income and efficient implementation can offset otherwise negative tracking effects. Do not label every negative ETF and benchmark gap tracking error. Ask whether the question wants average shortfall, variability of active returns, or ETF price versus NAV at a single instant.

Unlock the full lesson and continue reading

Signup to continue reading this lesson and unlimited access questions, flashcards, AI notes, and more

Scan to download app

Scan to download
UNLOCK FREE ACCESS

Practice Questions: Sources of ETF tracking error

  1. 1. Seaport Asset Management is considering an exchange-traded fund to obtain rapid exposure to a broad equity index. During the trading day, the ETF's market price and its published intraday net asset value diverge as shown in Exhibit 1. An authorised participant can transact in the complete underlying basket, create or redeem ETF shares at NAV in large creation units, and sell either the basket or the ETF without binding short-sale constraints. Taxes, financing costs, and bid–ask spreads are immaterial for the arbitrage decision. Seaport also evaluates the execution of a 100,000-share equity purchase. The portfolio manager released the order when the security's decision price was USD 50.00. The order was subsequently filled in two transactions at the prices and quantities reported in Exhibit 1. No shares remained unexecuted, so there is no opportunity-cost component from a cancelled portion. Implementation shortfall is measured relative to the decision-price paper portfolio and includes explicit commissions. A quantitative manager proposes an index-constituent strategy supported by a 12-year backtest. As described in Exhibit 2, the backtest uses the companies that are index members today for every historical observation. Companies removed from the index following bankruptcy, acquisition, or sustained financial weakness are absent from the historical universe. The manager also assumes costless execution despite annual turnover exceeding 180%. Seaport's analysts review the historical-universe construction separately from the missing transaction-cost adjustment. Finally, the risk committee evaluates Seaport's one-day 99% value-at-risk model. A VaR exception occurs whenever the actual daily portfolio loss exceeds the VaR estimate reported at the start of that day. The committee uses 250 trading days as the validation window and recognises that a correctly calibrated 99% model should produce exceptions on approximately 1% of days. The committee will not reject or alter a risk model solely because one exception occurs, but it requires a formal review when observed exceptions are materially inconsistent with the model's stated confidence level.

    • 2. Hawthorne Asset Management is transitioning a USD 50 million balanced portfolio after terminating an external manager. During the transition, Hawthorne uses a broad equity exchange-traded fund to maintain market exposure while individual securities are transferred and trade instructions are completed. The ETF tracks a fully transparent domestic equity index and supports in-kind creations and redemptions by authorized participants. At the close of trading, the portfolio manager reviews the ETF information in Exhibit 1. For purposes of measuring the premium or discount, Hawthorne uses the midpoint of the closing bid and ask quotes as the ETF market price and compares that amount with the reported net asset value per share. The manager notes that authorized participants may exchange creation units for the underlying basket or deliver the basket to receive creation units. These transactions are expected to place pressure on material premiums or discounts, although transaction costs and market frictions can prevent perfect convergence. Hawthorne's risk committee also requires a one-day parametric value-at-risk estimate for the transition portfolio. It assumes normally distributed returns, uses the one-tailed critical value shown in Exhibit 2, and defines VaR as a positive loss amount. Expected daily return is assumed to be zero. The committee understands that VaR is a quantile estimate and does not measure the magnitude of losses beyond the selected confidence threshold. The firm's economics group subsequently raises its inflation forecast from the previously expected rate shown in Exhibit 2 to the newly observed forecast. The portfolio manager is instructed to assume that this revision is credible and has not yet been incorporated into current market prices. Hawthorne may tactically adjust exposure among long-duration nominal government bonds, broad market cash holdings, and inflation-sensitive real assets such as commodities and infrastructure. The objective of the tactical position is to benefit from the forecast change rather than merely minimize short-term volatility. The transition manager emphasizes that ETF liquidity should not be assessed solely from the ETF's displayed trading volume because the liquidity of the underlying basket and the creation–redemption process also matter. Nevertheless, orders should be executed with limit prices when spreads widen, and the manager must consider both explicit commissions and implicit costs. The committee evaluates the ETF pricing discrepancy, authorized-participant response, portfolio VaR, and inflation positioning as separate decisions.

      • 3. Elena Alvarez, age 48, is a senior executive at Novus Technologies. Her annual compensation is USD 600,000, of which approximately 40% consists of restricted Novus shares. She expects to retire at age 62. Her current investable portfolio is USD 8.0 million, including USD 2.4 million of vested Novus shares inherited from her father. Her future employment income and a substantial portion of her financial wealth therefore depend on the same company. Alvarez plans to purchase a residence in two years and will require USD 1.2 million at that time. Her adviser establishes a separate cash reserve equal to 5% of the investable portfolio and plans to fund the remaining purchase amount through maturing short-duration bonds. Alvarez has no debt and expects her employment income to cover ordinary living expenses until retirement. During the planning meeting, Alvarez says she is comfortable with substantial short-term volatility because equities have historically recovered over long periods. She nevertheless refuses to sell the inherited Novus shares, stating, 'These shares are worth more to me than an identical holding purchased today because they represent my father's confidence in the company.' She does not claim to possess superior information or to influence Novus's future share price. The adviser estimates that the portfolio must earn a 5.5% nominal annual return after fees to meet Alvarez's retirement objective. The investment policy statement requires the strategic allocation to maintain the 5% cash reserve, reduce Novus shares to no more than 10% of the portfolio within five years, and limit diversified alternatives to 15%. Alternatives would be implemented through quarterly redeemable strategies; the adviser concludes that an allocation up to the policy maximum would be compatible with remaining liquidity needs. The adviser presents the candidate allocations in Exhibit 1. Expected nominal returns are 3.0% for global bonds, 7.5% for global public equity, 7.0% for diversified alternatives, 7.5% for employer shares, and 1.5% for cash. Each candidate is expected to achieve at least the required 5.5% return before any adjustment for the additional company-specific risk arising from Novus exposure. Alvarez has approved gradual sales of Novus shares if the selected allocation is consistent with the policy statement.

        • 4. In the mean–standard deviation figure, A and B have equal risk but A has higher expected return; C lies on the upper boundary. Which portfolio is least likely efficient?

          • A. Portfolio A
          • B. Portfolio B
          • C. Portfolio C
          Show answer

          Answer: B. Portfolio B

          Portfolio B - Portfolio B has the same standard deviation as A but a lower expected return, so A dominates B. - An investor seeking the highest expected return for a given level of risk would therefore reject B as inefficient. Portfolio A - At the risk level shared by A and B, A offers the higher expected return. - This comparison does not prove A is efficient, but it rules out B—not A—as the least likely efficient portfolio. Portfolio C - C lies on the upper boundary of the feasible set, a position consistent with mean–variance efficiency. - Unlike B, C is not shown to have a lower expected return than another portfolio with the same risk.

        • 5. A portfolio invests 60% in an asset with 10% volatility and 40% in an asset with 15% volatility. Their correlation is 0.30. The portfolio volatility is closest to:

          • A. 8.49%
          • B. 9.36%
          • C. 9.68%
          Show answer

          Answer: C. 9.68%

          9.68% - Portfolio variance is \(0.60^2(0.10^2)+0.40^2(0.15^2)+2(0.60)(0.40)(0.30)(0.10)(0.15)=\mathbf{0.00936}\). - Taking its square root gives portfolio volatility of approximately 9.68%. 8.49% - Omitting the covariance term leaves a variance of \(0.0036+0.0036=0.0072\). - Its square root is approximately 8.49%, so this answer misses the assets’ positive correlation. 9.36% - 0.00936 is the calculated portfolio variance, not its volatility. - Reporting 9.36% confuses that variance figure with a percentage volatility; volatility requires taking the square root.

        • 6. An adviser recommends rebalancing, but a client retains the existing allocation solely because changing it feels unnecessary despite materially altered objectives. The behavior most likely reflects:

          • A. Representativeness bias
          • B. Status quo bias
          • C. Endowment bias
          Show answer

          Answer: B. Status quo bias

          Status quo bias - The client keeps the existing allocation because changing it feels unnecessary, even though materially altered objectives support rebalancing. - This preference for leaving the current situation unchanged is status quo bias. Representativeness bias - Representativeness bias involves judging an investment by its resemblance to a familiar category or pattern, rather than using relevant evidence. - The client makes no such comparison; the decision rests on reluctance to change the allocation. Endowment bias - Endowment bias involves valuing an asset more highly because it is already owned. - The client does not express a higher valuation of any holding; the resistance is to changing the allocation itself.

        • 7. A portfolio invests 70% in an asset with 20% volatility and 30% in an asset with 10% volatility. Their correlation is 0.20. The calculated portfolio volatility is closest to:

          • A. 14.89%
          • B. 17.00%
          • C. 14.32%
          Show answer

          Answer: A. 14.89%

          14.89% - Portfolio variance is \(0.70^2(0.20^2)+0.30^2(0.10^2)+2(0.70)(0.30)(0.20)(0.10)(0.20)=\mathbf{0.02218}\). - Taking its square root gives portfolio volatility of 14.89%. 17.00% - The weighted average \(0.70(20\%)+0.30(10\%)\) equals 17.00%. - This slip averages the individual volatilities instead of calculating portfolio variance with the correlation term. 14.32% - Omitting the covariance term gives variance \(0.70^2(0.20^2)+0.30^2(0.10^2)=0.0205\), whose square root is 14.32%. - Because the assets have positive correlation of 0.20, their covariance term must be included.

        Flashcards: Sources of ETF tracking error

        1. 1. ETF (premium/discount) % =

          Show answer

          \= (ETF price - NAV per share) / NAV per share

        2. 2. Risks of investing in ETFs

          Show answer

          - counter party risks - ETNs have high (bank may default) - Large CDS speads indicate high - settlement risks - OTC derivative contracts - security lending - ETFs can lend securities to short sellers for a fee (overcollateralized) - fund closures - potential adverse tax consequences - soft closures: entail creation halts & changes in investment strategy - can be triggered by changes in regulation, competitive pressures or issuer merger - expectation-related risks - ETFs based on complex strategies (inverse or leveraged) may be different than investor expectations - compounding effects of leveraged ETFs make them unsuitable for buy-and-holder investors w investment horizons >1month

        3. 3. Do shareholders of ETFs and/or mutual funds bear the costs of creation/redemption?

          Show answer

          only mutual funds

        4. 4. Describe the creation/redemption process of ETFs

          Show answer

          - Authorized participants (APs) create additional shares by delivering the creation basket to the ETF manager - Redemption is by tendering ETF shares and receiving a redemption basket

        5. 5. If market price does not equal NAVPS, what does the premium/discount reflect?

          Show answer

          an investor's view of management capabilities, liabilities and governance

        Everything you need for CFA prep

        Get full Oncourse access with lessons, practice questions, flashcards and AI study tools.

        GET STARTED FOR FREE
        75% OFFLimited time offer
        GET 75% OFF