# Tracking error

Tracking error is the standard deviation of a portfolio’s returns minus a benchmark’s returns over a specified period, showing how variably the portfolio has diverged from that benchmark.

## What it means

A portfolio can outperform its benchmark in one period and underperform it in the next. Tracking error measures the variability of those return differences, called *active returns*. It is a measure of variation around the portfolio’s average active return, not a measure of whether that average is positive or negative.

Tracking error is usually quoted as a percentage, often annualized. A larger number means active returns have varied more widely over the measured period. It does not, by itself, show how much the portfolio lost, whether it beat the benchmark overall, or what will happen next. The benchmark, return frequency and measurement window all affect the result.

## How it is calculated

For a series of historical returns, first subtract the benchmark return from the portfolio return in each period. Then calculate the sample standard deviation of those differences:

`TE_period = sqrt[sum((a_t - ā)^2) / (n - 1)]`, where `a_t = R_p,t - R_b,t`.

- `TE_period` is tracking error per measurement period, in percentage points of return.
- `R_p,t` and `R_b,t` are the portfolio and benchmark percentage returns in period `t`, measured on a consistent basis, such as both including dividends.
- `a_t` is the active return in period `t`, in percentage points; `ā` is the arithmetic mean of the active returns.
- `n` is the number of periods; the sample formula requires at least two.

When returns are monthly, a common annualization is `TE_annual ≈ TE_monthly × sqrt(12)`. For daily returns, the corresponding convention is `TE_annual ≈ TE_daily × sqrt(252)`, using approximately 252 trading days. These square-root conversions assume the time-series behavior makes scaling reasonable; serial correlation can make them misleading. Using `n` rather than `n - 1` calculates a population standard deviation, so results can differ slightly by convention.

## Worked example

Suppose a hypothetical portfolio and its benchmark have these monthly returns:

| Month | Portfolio | Benchmark | Active return |
| --- | ---: | ---: | ---: |
| 1 | 3% | 2% | +1 percentage point |
| 2 | 1% | 2% | −1 percentage point |
| 3 | 4% | 2% | +2 percentage points |
| 4 | 2% | 2% | 0 percentage points |

The mean active return is `(1 − 1 + 2 + 0) / 4 = 0.5` percentage points per month. Subtract 0.5 from each active return, square the results and add them: `0.25 + 2.25 + 2.25 + 0.25 = 5` squared percentage points. The sample monthly tracking error is `sqrt(5 / 3) ≈ 1.29` percentage points. Applying the conventional monthly scaling gives an annualized figure of approximately `1.29 × sqrt(12) = 4.47` percentage points. That annualized number is an estimate based on just four hypothetical observations, not a forecast.

## Why it matters before you trade

A proposed trade may change a portfolio’s exposure to sectors, countries or individual stocks relative to a chosen benchmark. Those differences can change future active returns and, potentially, tracking error. Reviewing the existing portfolio and the proposed basket against the same benchmark makes the comparison meaningful.

Historical tracking error is not a direct estimate of the effect of one proposed trade. Assessing that effect would require the portfolio’s post-trade weights, benchmark weights and assumptions about how their returns vary together. A benchmark that does not reflect the portfolio’s purpose can also make the figure hard to interpret.

## How PreTrAIde uses it

The product facts do not list tracking error as a PreTrAIde calculation or output. Its pre-trade analysis instead provides portfolio breakdowns by country, sector, region, currency and market capitalization, plus a Herfindahl-Hirschman concentration index. These can help a user inspect exposures before placing an order, but they do not calculate benchmark-relative return variability. PreTrAIde’s GO, NO-GO and MODIFY verdicts are decision support, not investment recommendations.

## Common mistakes

- **Confusing tracking error with underperformance:** A portfolio that lags a benchmark by exactly one percentage point every period has negative active returns but zero tracking error.
- **Comparing unlike figures:** Monthly tracking error and annualized tracking error use different time units; different benchmarks and sample windows also change the comparison.
- **Treating it as total risk:** A portfolio can have low tracking error while both it and its benchmark fluctuate substantially.
- **Assuming an exposure breakdown is a tracking-error estimate:** Holdings information alone does not supply the return relationships needed to estimate benchmark-relative variability.

---
Source: https://pretraide.com/glossary/tracking-error