Tracking Error Explained: Why Zero Can Still Cost You

What is tracking error, and why can a perfect score mislead you?

Tracking error measures how variable a fund’s periodic returns are relative to its benchmark. It does not measure how much money the fund ultimately leaves behind. A fund can have zero measured tracking error in a constructed example and still finish the year meaningfully below its index.

That distinction matters when you compare two index funds. The lowest tracking-error number describes the steadiest relative return path, not necessarily the cheapest implementation or the best ending balance. Tracking difference answers the separate question of how much the fund gained or lost relative to its benchmark over a stated period.

Our original twelve-month example makes the difference concrete. Two hypothetical funds start with $10,000 and follow the same invented benchmark. Fund A has zero measured tracking error but finishes $130.60 behind the benchmark. Fund B has approximately 0.36% annualized tracking error and finishes only $0.51 behind. Neither is a real fund, and these figures are not market quotes, forecasts or backtest results.

The lesson is not that higher tracking error is better. It is that one number cannot tell you both the consistency and the cost of following an index. Read the two measures together, then check the return definitions, benchmark, dates and trading costs before drawing conclusions.

How does tracking error differ from tracking difference?

Mackenzie Investments’ December 2025 explainer defines excess return, also called tracking difference, as a fund’s net asset value total return minus its index total return. It describes tracking error as the annualized standard deviation of excess returns. We checked this document on October 29, 2026.

We use the sign convention fund minus benchmark throughout. A negative tracking difference means the fund underperformed; a positive number means it outperformed. Some reports present a shortfall as a positive cost instead, so always read the definition rather than interpreting the sign from memory.

Measure	Question answered	What a small number means
Tracking difference	How far apart were the total returns over the stated period?	The ending returns were close.
Tracking error	How variable were the periodic fund-minus-index returns?	The relative return path was consistent.
Expense ratio	What ongoing expenses does the fund disclose?	Disclosed ongoing costs are low, not necessarily all implementation costs.

Fidelity’s investor-education page, checked October 29, 2026, makes the same variability-versus-performance distinction. Its daily-data definition is common in fund reporting. Our teaching example deliberately uses monthly observations so every input fits in one table; its result must not be compared directly with a provider’s daily-data statistic.

If you are still choosing between fund structures, our ETF versus index fund comparison addresses a different question. Tracking quality describes implementation after you identify the exposure and vehicle you actually want.

What happens to $10,000 in our two-fund example?

The following data are entirely hypothetical. Every return is a monthly total return in the same currency, observed over the same twelve periods. Distributions are assumed reinvested. There are no contributions, withdrawals, investor taxes or trading costs outside the fund returns. This isolates the mathematics rather than pretending to reproduce a particular ETF.

Fund A earns the benchmark return minus 0.10 percentage points every month. That is a ten-basis-point monthly active-return shortfall: one basis point equals 0.01 percentage points. Fund B alternates between outperforming by 0.10 percentage points and underperforming by 0.10 percentage points. The labels describe artificial return series, not management styles or products.

Month	Benchmark	Fund A	Fund B
1	2.00%	1.90%	2.10%
2	−1.00%	−1.10%	−1.10%
3	3.00%	2.90%	3.10%
4	0.00%	−0.10%	−0.10%
5	1.00%	0.90%	1.10%
6	−2.00%	−2.10%	−2.10%
7	2.00%	1.90%	2.10%
8	1.00%	0.90%	0.90%
9	−1.00%	−1.10%	−0.90%
10	3.00%	2.90%	2.90%
11	0.00%	−0.10%	0.10%
12	2.00%	1.90%	1.90%

To find each ending balance, multiply $10,000 by one plus each monthly return, in sequence. For the benchmark, the first two months produce $10,000 × 1.02 × 0.99 = $10,098. Adding 2% and −1% would instead suggest $10,100, already showing why arithmetic sums are not compounded returns.

Twelve-month result	Benchmark	Fund A	Fund B
Compounded total return	10.3098%	9.0039%	10.3048%
Ending balance	$11,030.98	$10,900.39	$11,030.48
Tracking difference	Reference	−1.3060 percentage points	−0.0051 percentage points
Annualized monthly-sample tracking error	Reference	0.0000%	0.3618%
Dollar shortfall versus benchmark	Reference	$130.60	$0.51

The displayed balances and shortfalls are rounded independently from unrounded calculations. Subtracting rounded balances can differ by a cent. The benchmark is not an investable product with no costs; it is the reference series used to demonstrate the measures.

Fund A’s relative returns never vary: all twelve equal −0.10 percentage points. Their standard deviation is zero, even though the shortfall compounds through the year. Fund B’s average monthly active return is zero, but its alternating differences have a positive standard deviation.

Fund B does not quite match the compounded benchmark return despite its zero arithmetic mean active return. The interaction between the benchmark path and monthly deviations matters. Equal average gaps do not guarantee equal ending wealth. Our example therefore separates three ideas often collapsed into one: average active return, compounded tracking difference and tracking-error volatility.

How do you calculate tracking error step by step?

First calculate each periodic active return: fund return minus benchmark return. Use returns, not price levels or dollar differences. Express both inputs in the same units. A fund return of 1.90% minus an index return of 2.00% is −0.10 percentage points, or −0.001 in decimal-return notation.

Second calculate the arithmetic mean of those active returns. Third subtract that mean from every observation, square each deviation and add the squared deviations. For the sample standard deviation used here, divide by the number of observations minus one, then take the square root.

Finally apply the stated annualization convention. For this monthly illustration we multiply the monthly sample standard deviation by the square root of twelve. This is a convention for scaling the statistic, not a forecast of the coming year’s relative return.

Fund B has twelve active returns, six at +0.001 and six at −0.001. Its mean is zero. The sum of squared deviations is 12 × 0.001² = 0.000012. Divide by eleven and take the square root to obtain approximately 0.0010445, equivalent to 0.10445% monthly sample standard deviation. Multiplying by √12 gives approximately 0.36181% annualized tracking error.

Using the population denominator of twelve instead would give 0.10000% monthly standard deviation and approximately 0.34641% annualized tracking error. Neither result is an arithmetic mistake; they use different conventions. A report should identify the convention rather than imply the fourth decimal place makes unlike statistics comparable.

In a spreadsheet, put fund returns and benchmark returns in adjacent columns, subtract them row by row, and apply STDEV.S to the active-return column. For our monthly example, multiply that result by SQRT(12). To calculate compounded tracking difference, use the product of one plus each fund return minus the equivalent benchmark product. Keep decimal inputs distinct from percentage formatting.

Do not multiply a monthly tracking difference by √12. The square-root scaling belongs to standard deviation. Do not sum monthly active returns and label the sum an exact annual tracking difference. Do not calculate standard deviation of the fund’s returns alone and call that tracking error: that measures absolute volatility, not variability relative to the index.

Why can zero tracking error coexist with underperformance?

Standard deviation measures distances from the average, not distances from zero. If every active return is the same negative number, each observation equals its own average. Every deviation from that average is zero. The statistic has no way to tell you that the average itself is unattractive.

That is why our Fund A can steadily lag without generating measured tracking-error volatility. Its zero score is the mathematical consequence of a deliberately constant monthly gap. It does not mean a real fund can operate without friction, that a provider reports exactly zero at every frequency, or that an investor should disregard consistent underperformance.

The ten-basis-point monthly drag is an assumed return difference, not a quoted expense ratio. Calling it a 1.20% annual fee would blur a modeling input with a product disclosure and would not reproduce the exact compounded shortfall. The twelve-month arithmetic sum is −1.20 percentage points; the actual tracking difference in our path is approximately −1.3060 percentage points.

For an index investor, a stable shortfall may be understandable if it reflects disclosed costs and difficult implementation. But “understandable” is not synonymous with “competitive.” Compare realistic alternatives on the same basis before accepting the cost of consistency.

What causes funds to diverge from their indexes?

The Mackenzie and Fidelity explainers identify several drivers, including fees, trading costs, replication choices, cash management, tax differences and securities lending. These mechanisms can work in different directions. They are explanations to investigate, not a universal equation assigning every shortfall to management quality.

Ongoing expenses

Fund expenses reduce the assets available to investors. All else equal, a fund charging more to deliver the same exposure has a larger hurdle. But historical tracking difference is not simply the expense ratio printed on a factsheet. Other effects can offset or amplify the expense drag, and a fee change may mean the latest disclosure differs from the cost applied during an older measurement period.

Rebalancing and trading

An index’s rules can call for changes that a portfolio must implement through transactions. The fund may pay spreads or encounter market impact, while a reference index applies its calculation rules rather than executing orders. Timing also matters: trading before or after the index’s effective change can alter relative returns.

Full replication versus sampling

Holding selected securities instead of every constituent can reduce operational difficulty but create exposure differences. Sampling is not automatically poor tracking. The practical comparison is whether the implementation produces suitable cost and consistency for that index. A broad or difficult-to-trade benchmark may present different challenges from a concentrated liquid one.

Cash and distributions

Cash from dividends, coupons or investor activity may not be invested immediately. If the index rises while part of the fund remains in cash, the fund can lag. If the market falls, that same cash position can help relative performance. The direction is conditional; “cash drag” is not guaranteed to be negative in every observation.

Tax and valuation differences

The withholding assumptions in an index may differ from the taxes actually affecting a fund. Currency conversion times and valuation policies may also differ. A gap that appears to be a management result can partly reflect mismatched reference conventions. Read the exact index name and methodology rather than relying on the broad exposure label.

Securities lending

Some funds earn revenue by lending holdings to borrowers. Revenue retained by the fund can offset expenses and improve relative returns. It is not a free guarantee: the lending arrangement, revenue allocation and associated risks belong in the product documents. Our example assumes return series directly and makes no lending-income claim about any real ETF.

Which return series should you compare?

Start with the official fund NAV total-return series and the fund’s stated benchmark total-return series. NAV describes the value of the portfolio per share. Total return incorporates distributions under the provider’s reinvestment methodology. Matching both sides reduces the chance that you mistake a distribution or return-definition mismatch for poor tracking.

A price-only index and a distribution-reinvesting fund answer different questions. If one includes income and the other excludes it, their difference is not a clean implementation comparison. Likewise, gross and net total-return benchmarks can use different tax assumptions. “Same index family” is not enough; confirm the specific variant.

Check share class, currency, dates and hedging status. Two share classes can have different fees or distribution policies. A currency-hedged class should not be judged against an unhedged reference without understanding the intended benchmark. A fund quoted in a currency does not automatically remove currency exposure from its underlying assets.

Do not treat a chart downloaded from a retail screen as proof that these bases match. Adjusted prices may use conventions different from the official NAV total-return series. If you cannot establish what the adjustment includes, label the comparison limited rather than presenting it as an audited tracking statistic.

A benchmark change is another boundary. Combining observations across different stated indexes can produce a number that hides a change in mandate. Use a consistent measurement window or explain the break. The point is not to make the calculation harder; it is to ensure the result answers the question you think you asked.

Why might your own ETF return differ from NAV tracking?

ETF investors buy and sell at exchange prices, while NAV tracking compares the fund portfolio with the index. Fidelity and Mackenzie explicitly flag NAV-versus-market-price comparison issues. Your execution price, spread, brokerage costs and any change in the premium or discount can affect your result separately from the fund’s NAV performance.

Consider a second, independent hypothetical example. NAV begins at $100 and ends at $110, with no distributions. An investor buys at $100.20, a 0.20% premium to starting NAV, and sells at $109.78, a 0.20% discount to ending NAV. The NAV return is 10.00%; the investor’s price return is approximately 9.5609%, before commissions and taxes.

The approximately 0.4391-percentage-point gap is not evidence that the portfolio tracked badly. It arises from the chosen exchange entry and exit prices. These are invented prices used to show arithmetic, not a claim that such premiums and discounts are normal for any particular ETF.

That example explains why a sensible fund comparison needs two stages. First evaluate portfolio implementation with matched NAV and index total returns. Then evaluate the trading experience for your expected order size, holding period and market. A small annual tracking advantage can be overwhelmed by a costly purchase and sale.

For long-horizon planning, implementation is only one layer. Our sequence-of-returns risk analysis shows why withdrawals and the order of market returns can change outcomes even when summary performance looks similar. Low tracking error does not solve withdrawal risk or make the benchmark itself safe.

Is lower tracking error always better?

For an index strategy whose objective is close replication, lower tracking error generally indicates a more consistent historical relationship with the benchmark, provided the comparison is like for like. That can matter when precise exposure matching is important. It remains only one part of the evaluation.

Our Fund B should not be selected merely because it had the better ending balance in a constructed year. We deliberately gave it deviations that largely offset. In another path, variable relative returns could leave a much larger shortfall. The calculation proves that the metrics differ; it does not establish a rule that inconsistent tracking pays.

Active strategies raise a different question. A manager seeking to depart from an index can generate tracking error intentionally. A low number is not automatically success, and a high number is not automatically skill. Assess the stated objective rather than applying an index-replication scorecard to every fund.

A fund following a risky benchmark can also have very low tracking error while suffering substantial absolute losses. Close replication says little about diversification, valuation, drawdown tolerance or suitability. Choose the exposure first, evaluate its implementation second, and avoid treating a precise statistic as a substitute for a portfolio decision.

How should you compare two published factsheets?

Before ranking two figures, ask whether they are measured over the same window. A calm period and a stressed period can produce different relative-return variability. Ask whether the observation frequency is daily, weekly or monthly, and whether both annualize the same way. Our twelve monthly observations are transparent but too sparse to justify a confident forecast.

Square-root-of-time scaling is especially easy to overinterpret. The usual interpretation relies on assumptions about how variability accumulates. Serial dependence, changing conditions and observation choices can undermine that interpretation. We use √12 as an explicitly disclosed teaching convention, not as evidence that the underlying deviations behave independently.

Next check whether a figure is historical or model-based. A historical statistic describes observations; a prospective estimate depends on assumptions about future exposures and risks. Neither guarantees the future. A provider’s methodology can be legitimate yet unsuitable for direct comparison with another provider’s differently constructed number.

Finally compare both tracking difference and tracking error over meaningful, matching horizons. A single excellent year does not establish durable implementation quality. Conversely, a temporary valuation mismatch can make a short window look worse than the longer economic experience. Investigate explanations without assuming every anomaly will reverse.

What is a practical fund-comparison checklist?

1. Identify the exposure. Confirm the exact index, return variant and investment objective.
2. Match the inputs. Use identical dates, currency, share class and total-return definitions.
3. Read tracking difference. Establish the sign convention and the actual historical return gap.
4. Read tracking error. Establish observation frequency, standard-deviation convention and annualization.
5. Explain implementation. Review expenses, replication, trading, cash, tax and lending arrangements.
6. Separate execution. Consider exchange prices, spreads and commissions outside NAV tracking.
7. Check suitability. Consider benchmark risk, holding horizon and the rest of your portfolio.

If the factsheet omits a definition, ask the provider for the methodology rather than inventing it. If two funds do not disclose comparable data, state that limitation. A comparison marked “not directly comparable” is more useful than a tidy ranking built on incompatible numbers.

The most important warning sign is not a particular threshold. There is no universally good tracking-error level for every market, mandate and method. It is an unexplained mismatch: a figure that changes sharply, a benchmark variant you cannot identify, or an attractive score accompanied by a persistent return gap nobody has accounted for.

How was this analysis produced, and what does it not claim?

This analysis combines the publisher definitions cited above with an original, fully disclosed twelve-month hypothetical dataset. AI assisted drafting and automated arithmetic checks; a separate automated review checked the calculations and methodological claims. No independent human review is claimed. The byline identifies the site’s ETF coverage author, not proof of a personal trade or first-hand fund-management experience.

We calculated compounded returns from decimal monthly inputs, tracking difference as fund minus benchmark, sample standard deviation with eleven degrees of freedom, and annualized monthly tracking error using √12. Separate checks reproduce the population-standard-deviation alternative and the invented premium/discount example. All calculations retain full precision until display rounding.

There are no live stock prices, current yields, real fund rankings, forecasts, expected-return estimates or probabilities in this article. We do not infer persistence from twelve observations or attribute the artificial monthly gap to a real fee schedule. Source documents were retrieved on October 29, 2026; where a source does not show a publication date, that retrieval date is not presented as its publication date.

The purpose is to help readers avoid choosing a fund because a single statistic looks reassuring. The useful question is not “Which score is smallest?” It is “Does this fund deliver the exposure I want, at a reasonable observed cost and with an acceptable degree of variation?” Tracking difference and tracking error answer different parts of that question.

This material is educational, not personalized investment, legal or tax advice. Funds can lose value, benchmarks can fall sharply, and historical tracking statistics do not guarantee future results. Check the relevant prospectus and your own circumstances before investing.