ETF tracking difference vs. tracking error: why zero error can still lag

This guide concerns ordinary index-tracking funds, not whether an active manager should match an index. Use the exact fund share class and benchmark named in its documents, rather than a familiar index with a similar label.

Use fund net asset value, or NAV, total returns and the matching index return series. Fix the currency, dates, valuation times, and income treatment before calculating a gap.

A distributing fund's raw NAV can fall when it pays income. A total-return series accounts for that distribution; an unadjusted NAV chart alone does not measure the same result.

[Investor.gov's distribution guide](https://www.investor.gov/introduction-investing/general-resources/news-alerts/alerts-bulletins/investor-bulletins/fund-distributions-investor-bulletin) explains that transfer of value.

[Vanguard's tracking explanation](https://www.vanguard.co.uk/professional/vanguard-365/investment-knowledge/etf-knowledge/what-affects-index-tracking) separates the size of a return gap from its variability.

Here, tracking difference, or TD, equals fund total return minus benchmark total return over the same period. A negative value means the fund lagged. Check the provider's sign convention before interpreting its figure.

Suppose the index earns 8.00% over one year and the fund earns 7.70%. TD = 7.70% − 8.00% = −0.30 percentage points, or −30 basis points. One basis point is 0.01 percentage point.

On equal hypothetical starting investments of 10,000, the ending values are 10,800 and 10,770. The fund finishes 30 currency units behind. No personal taxes, trading charges, or external cash flows are included.

Do not call that a 30% shortfall. Nor is it the same calculation as dividing the two ending values: 10,770 ÷ 10,800 − 1 ≈ −0.2778%, which uses benchmark ending wealth as the denominator.

For several periods, compound each return series first and then subtract the matched cumulative returns. A sum of monthly percentage-point gaps is not generally the cumulative TD.

Take an imaginary benchmark returning exactly 0% in each of four months. The fund returns below are invented to isolate the arithmetic, not observed ETF results or a forecast.

Fund A's gap never changes, so its sample tracking error is zero. It still loses money each month relative to this flat benchmark. Zero variability does not mean a zero return gap.

Fund B has the same average monthly gap but moves above and below it. Its tracking error is positive. The two statistics answer different questions even before annualizing them.

• Fund A monthly returns: −0.10%, −0.10%, −0.10%, −0.10%.
• Fund B monthly returns: −0.30%, +0.10%, −0.30%, +0.10%.
• Both funds have an arithmetic mean monthly gap of −0.10 percentage point, or −10 basis points.

Let a_t be fund return minus index return for each matched period. Compute the mean of these signed gaps, then their sample standard deviation: s = sqrt(sum((a_t − mean(a))^2) ÷ (n − 1)), with n greater than one.

[NIST's measures of scale](https://www.itl.nist.gov/div898/handbook/eda/section3/eda356.htm) describe the sample-variance denominator. This example uses n − 1; record a provider's own convention when reproducing its metric.

In basis points, B's gaps are −30, +10, −30, +10. Their mean is −10, giving deviations of −20, +20, −20, +20 basis points.

Squared deviations sum to 1,600 basis-points squared. Divide by 3 and take the square root: s ≈ 23.0940 basis points per month, or 0.230940% in monthly return units.

Using the conventional square-root-of-time scaling, annualized TE = s × sqrt(12) = 80 basis points, or 0.80%. The 12 comes from months per year, not the four observations in this toy sample.

That scaling assumes an appropriate stable variance and no serial covariance. Correlated gaps or changing conditions weaken its interpretation. Four invented observations provide no evidence about a real fund's future tracking.

Keep signed gaps when computing the standard deviation. Averaging absolute gaps or taking the standard deviation of absolute gaps calculates a different statistic.

A's four-month total return is (0.999)^4 − 1 ≈ −0.3994004%. B's is (0.997 × 1.001)^2 − 1 ≈ −0.4001988%. Because the benchmark is flat, these are also their four-month tracking differences.

Both means were −10 basis points per month, but their compounded outcomes are slightly different. Neither result is exactly −0.40%, and neither equals B's annualized 0.80% tracking error.

An endpoint return can establish TD but cannot reconstruct TE. Many monthly paths can lead to the same endpoint. Obtain the dated return observations rather than inventing a volatility number from one annual gap.

The arithmetic versus geometric return guide explains why multiplying growth factors differs from adding or averaging returns.

Suppose NAV rises from 100 to 110 with no distribution, while the benchmark also gains 10%. NAV-based TD is zero for that interval.

Now suppose you buy at 102 and sell at 109 over the same interval. Your pre-cost return is 109 ÷ 102 − 1 ≈ 6.8627%, not 10%. A purchase premium and sale discount change the result without changing those NAV returns.

These are hypothetical executable prices. Actual commissions, bid-ask spreads, currency conversion, and personal taxes can create further differences; do not subtract an assumed spread again if actual fills already include it.

[Investor.gov's ETF overview](https://www.investor.gov/introduction-investing/investing-basics/investment-products/mutual-funds-and-exchange-traded-2) distinguishes exchange prices from NAV.

A snapshot premium or discount is a price-versus-NAV comparison at a moment in time. Tracking difference compares returns over an interval. They are not interchangeable measurements.

Fund expenses, trading, cash holdings, portfolio sampling, and income or tax timing can affect index replication. A return gap alone does not identify which cause dominated.

[Vanguard's index-tracking overview](https://www.vanguard.ca/en/tools-and-resources/etf-fundamentals/management/index-tracking) also discusses effects that can offset a gap, such as securities-lending revenue.

If reported NAV returns already include fund expenses, do not deduct the expense ratio from those returns again. TD is an observed outcome, not a separate invoice; TE is not an additional annual fee either.

Match gross or net dividend conventions and hedged or unhedged benchmarks. An apparent surplus against a differently taxed or differently hedged index is not proof of superior replication.

Positive TD is possible, but it does not establish repeatable outperformance. Compare the fund's explanation and implementation, rather than assuming every gap equals its advertised expense ratio.

Select peers tracking the same benchmark and share-class objective. A comparison against a different index measures different exposure, not just tracking quality. Keep these fields beside the figures.

Historical, or ex-post, TE comes from observed returns. A forward-looking, or ex-ante, estimate comes from a risk model. Do not present an estimate as an observed result or replace missing returns with invented zeros. [!TRYMARK] TryMark ETF tracking audit At the next factsheet review, target a reproducible fund-versus-index comparison. Recompute TD and TE when benchmark version, return dates, distribution adjustments, or fee treatment changes. [!WARNING] Low tracking error is not low investment risk A fund can follow an index closely while both fall sharply. TE is not a maximum-loss bound or a probability guarantee. Check the underlying exposure and drawdowns separately from tracking quality.

Use total-return conventions to match income and drawdown measurement to examine losses. This guide does not recommend an ETF.

• Exact benchmark, currency, share class, and income treatment.
• Start and end dates, NAV or market-price basis, and valuation times.
• TD sign convention and whether returns are cumulative or annualized.
• TE observation frequency, sample length, denominator, and annualization method.
• Fees already reflected, missing observations, and the fund's explanation for gaps.