Tracking Difference vs Tracking Error: A Gap and Its Volatility

Two numbers from one subtraction

Investor.gov describes an index fund as “a type of mutual fund or exchange-traded fund that seeks to track the returns of a market index.” Tracking is the whole job, so the interesting question is how far off the fund was, and there are two different ways to answer it.

Tracking difference is a level: the fund’s return minus the index’s return over a stated period, in percentage points. Tracking error is a dispersion: the standard deviation of that same difference, measured over shorter intervals and usually annualized. One number tells you where the fund ended up. The other tells you how straight the path was.

The arithmetic that separates them

Tracking difference over a period needs no statistics:

tracking difference = fund return − index return

Suppose the index returned 8.40% and the fund returned 7.55%; the tracking difference is −0.85 percentage points. That is the whole calculation, and it is the number that matters to a holder, because it is what the account actually experienced.

Tracking error needs the path. Take the difference for each sub-period, average it, measure how far each observation sits from that average, and annualize:

tracking error = standard deviation of the periodic differences × √(periods per year)

The two are related but not interchangeable. A fund whose monthly difference is identical every month has a tracking error of zero, because a constant has no dispersion — but its tracking difference is the sum of those identical differences, which is not zero. That single distinction explains most of the confusion in this area.

Worked example: the same gap, two different paths

The twelve monthly differences below are constructed to show the arithmetic; they are not a fund’s reported figures. Both funds end the year 0.85 percentage points behind their index, which is the tracking difference.

For Fund A:

• Sum of the twelve differences: −0.85 pp, which is the tracking difference for the year.
• Mean: −0.85 ÷ 12 = −0.0708 pp per month.
• Sum of squared deviations from that mean: 0.021092.
• Sample variance: 0.021092 ÷ 11 = 0.0019175.
• Standard deviation: √0.0019175 = 0.0438 pp per month.
• Annualized: 0.0438 × √12 = 0.0438 × 3.464 = 0.152 pp, or about 0.15%.

For Fund B, the mean is identical at −0.0708 pp and the sum is identical at −0.85 pp. But the sum of squared deviations is 2.036492, so the variance is 0.1851, the monthly standard deviation is 0.4303 pp, and the annualized figure is 0.4303 × 3.464 = 1.49 pp.

Same tracking difference, ten times the tracking error. Fund A’s shortfall is a cost. Fund B’s shortfall is a cost plus a distribution of outcomes, and in a different year the same dispersion could have landed on the other side of the index.

Why a fund can track tightly and still lag every year

A fund that charges a fixed expense ratio and applies it daily produces an almost perfectly constant drag. Constant drag adds nothing to the standard deviation, so the tracking error can be close to zero while the tracking difference stays negative for as long as the fund exists. Four mechanisms do most of the work.

Fees. Investor.gov’s fee bulletin breaks the total annual fund operating expenses into management fees, distribution or service fees and other expenses, “expressed as a % of the fund’s average net assets, called the expense ratio,” and notes that when fees are paid out of fund assets, “the value of the fund decreases and the value of all the investors’ shares decreases.” That is a steady, predictable gap.

Withholding tax. Dividends from foreign holdings can be taxed at source before the fund receives them, while a published index usually measures the return without that tax. The fund then lags by an amount that depends on the dividend calendar, not on anyone’s decision.

Sampling. Investor.gov notes that index funds “may take different approaches to track a market index: some invest in all of the securities included in a market index, while others may invest in only a sample of the securities.” A sample tracks the index on average and differs from it holding by holding, so it tends to raise tracking error even when it does not change the long-run gap.

Securities lending. A fund that lends its holdings earns revenue that can offset fees, and the revenue arrives unevenly. It tends to reduce the tracking difference and add to tracking error at the same time.

Where the naive reading goes wrong

Using the two words as synonyms. Investor.gov’s own index-fund page lists “Tracking Error” as a risk and describes it as a fund that “may not perfectly track its index,” then separately lists underperformance from “fees and expenses, trading costs, and tracking error.” The regulator’s usage is broad: it names the whole phenomenon. The measurement meaning is narrower, and a provider’s figure will follow the narrower one.

Assuming low error means low cost. A fund can keep its dispersion tiny by holding a permanent cash buffer that costs it a fixed few basis points a year. The error looks excellent and the difference is worse than a competitor’s.

Comparing figures computed at different frequencies. A monthly-based annualized error and a daily-based one are different statistics. The daily calculation sees more variation and changes the annualization factor, so the two numbers can differ by a lot for the same fund and the same year.

Reading a positive difference as skill. A fund can beat its index through lending revenue, a favorable tax position or a sample that happened to outperform. None of those is a repeatable promise, and none changes the fee.

What the primary sources do and do not say

The investor.gov pages fetched for this article describe what an index fund is, how sampling works, why underperformance occurs, and what the prospectus fee table contains. They are precise about the cost side and deliberately non-quantitative about tracking.

What they do not provide is a numeric definition of tracking error as the volatility of the difference, or any published pair of fund and index return series to compute one from. That definition is the working definition used here, stated as arithmetic you can check, and the worked example above is explicitly constructed rather than observed. If you are comparing real funds, take the tracking error and tracking difference from the provider’s own report and confirm the period, the frequency and the subtraction order before ranking anything.

Checklist

• Write down both numbers with their period. “Tracking difference of 0.85%” means nothing without the year it covers.
• Check the subtraction order. Some providers report index minus fund, which flips the sign of every figure.
• Confirm the measurement frequency. Monthly-based and daily-based annualized errors are not comparable.
• Compare the difference against the expense ratio. A gap much larger than the disclosed fee is being caused by something the fee table does not show.
• Treat a positive difference as unexplained until you find the source. Lending revenue and sampling luck are common causes and neither is permanent.

How these sources were checked

Investor.gov’s Index Funds page at https://www.investor.gov/introduction-investing/investing-basics/investment-products/mutual-funds-and-exchange-traded-4 (read 2026-10-23) supplies the definition of an index fund, the sampling description, and the statements on tracking error and underperformance. Its Investor Bulletin: Index Funds at https://www.investor.gov/introduction-investing/general-resources/news-alerts/alerts-bulletins/investor-bulletins-26 (read 2026-10-23, dated 6 August 2018) carries the same text in bulletin form and supplies the date. Its Mutual Fund and ETF Fees and Expenses bulletin at https://www.investor.gov/introduction-investing/general-resources/news-alerts/alerts-bulletins/investor-bulletins-20 (read 2026-10-23, dated 23 July 2025) supplies the expense-ratio components and the statement that fees paid from fund assets reduce the value of every investor’s shares. CNMV’s portal returned HTTP 403 when requested with a browser user agent on 2026-10-23, so no Spanish regulator source is cited.