Tracking Error Versus VaR: What Each One Is Actually Measuring

A portfolio manager loses 30% for a client in a single year. On the numbers alone, that sounds like a disaster. It might also be the best year that manager has ever had, if the benchmark lost 30% too and the portfolio's tracking error against it barely moved.

That is the whole distinction worth explaining, and it is why the question comes up in almost every long-only interview: tracking error and value at risk both get called "risk," they both produce a single number, and they measure completely different things.

30%

a loss that can still be a good year

if the benchmark fell the same amount

2x

a UCITS relative VaR cap

the cap on VaR against an unleveraged reference portfolio

1

risk measure a long-only mandate actually lives on

tracking error, not VaR

What tracking error is measuring

Tracking error is the standard deviation of the gap between a portfolio's return and its benchmark's return, annualized so it can be compared across periods. It is a relative measure. It says nothing about whether the portfolio made money or lost it in absolute terms; it says only how consistently the portfolio's return has tracked, or failed to track, the index it is judged against.

That is a stranger idea than it sounds, so it is worth stating plainly: tracking error is directionally agnostic. A manager who beats the benchmark by 5% every single year has the same tracking error, zero, as a manager who lags it by 5% every single year.

Both are perfectly consistent relative to the index; neither has drifted at all. Tracking error has nothing to say about which of those two managers you would rather hire. That is a different question, covered further down.

Ex-ante versus ex-post, and why they differ

A tracking error number can be built two different ways, and the two versions of the same portfolio routinely disagree.

• Ex-post tracking error is calculated from realized, historical returns. It answers "how much did this portfolio actually drift from its benchmark, looking backward." It is the number that shows up in a fact sheet after the period has closed.
• Ex-ante tracking error is a forecast. A portfolio manager, usually working with a commercial risk model, estimates how much the current portfolio is likely to drift from its benchmark going forward, based on the portfolio's exposures to style, sector, size and other systematic factors today. It answers a different question: "given how this portfolio is positioned today, how much drift should I expect."

The two disagree for a reason worth knowing rather than memorizing: the risk model behind an ex-ante number is built on historical relationships between factors, and those relationships shift. A portfolio can carry a low ex-ante tracking error going into a period, in a market where correlations between sectors then break down in a way the model did not anticipate, and finish the period with a considerably higher ex-post number.

Neither number is wrong. They are answering different tenses of the same question, and a manager who can explain why they diverged on a specific period is answering a technical question most candidates only define.

What a tracking error number implies

A number on its own is not useful until you know what it implies about the range of outcomes underneath it.

Assuming a normal distribution of returns, a tracking error of 4% means the portfolio's return should fall within about 4 percentage points of the benchmark's return roughly two years in three, and within about 8 percentage points of it in roughly nineteen years out of twenty. That is the standard one-standard-deviation, two-standard-deviation framing risk teams actually use when they translate a tracking error figure into a plain-English expectation for a trustee or a client.

What that means for reading someone else's number

The practical use of this framing is reading a tracking error figure the way it is meant to be read: as a band of plausible outcomes in a typical period, not a promise. A 6% tracking error fund that underperforms its benchmark by 9% in one bad year has not necessarily broken anything; it has landed outside its usual one-standard-deviation band in a way that happens, on the normal-distribution assumption, more often than the tidy math implies.

What should actually worry a client is a sustained pattern of underperformance relative to that fund's own stated tracking error, not one rough year.

The mandate bands, and what each one is really saying

Tracking error is not one number with a universal "good" reading. It is set, deliberately, by what the mandate promised, and the bands genuinely differ by type of fund rather than sitting on one continuous scale.

The bond-versus-equity gap inside that middle row is worth pausing on. A core bond manager running a 4% tracking error would be running an unusually aggressive book by fixed income standards; a core equity manager running the same 4% would be entirely ordinary. Tracking error only means something once it is read against the asset class it belongs to.

Index / passive fund

~0.1%

Enhanced index

~2.6%

Core active (median)

~5.4%

High-conviction (top quartile+)

~8%+

Order-of-magnitude bands built from institutional risk-budgeting data, not universal thresholds: a passive-equity target, one factor-based manager's own tracking error, the ten-year median for active large-cap equity managers, and the top-quartile-and-above range for the least efficient equity markets.

Why the range matters more than the label

None of this is about one band being smarter than another. A pension trustee who hired an index-tracking manager and gets a 6% tracking error back has a real problem, because that manager has quietly become something the mandate never asked for. A high-conviction manager running a 1% tracking error has arguably the opposite problem: being paid a high-conviction fee for what looks, on this evidence, like an index fund with extra steps.

Institutional risk-consulting data on actual active equity managers backs up how wide that spread really is. Looking at ten years of trailing tracking error for actively managed US large-cap equity strategies, the middle 50% of managers run somewhere between roughly 4% and 7%, with the top-quartile-and-above tail stretching past 9%.

Emerging-market equity managers run noticeably wider still, with the widest tail regularly running into the high teens, simply because the underlying index is less efficient and there is more genuine dispersion to be captured or missed. A core bond manager's entire range, by contrast, sits below where a large-cap equity manager's bottom quartile starts. The same word, tracking error, means a completely different scale of conviction depending on which asset class it is attached to.

Information ratio, the number that measures skill

Tracking error alone answers "how far did this portfolio stray." It says nothing about whether straying was worth it. The number that answers that is the information ratio: active return divided by tracking error.

A manager who beats the benchmark by 2% while running a 4% tracking error has an information ratio of 0.5. A manager who beats it by the same 2% while running an 8% tracking error, twice the active risk for the same active return, has an information ratio of 0.25, half as good, even though the two managers reported an identical excess return. The information ratio is what tells you which of them did more with the risk taken.

What a "good" one looks like

An information ratio above about 0.5 is generally read as a genuinely strong, sustained result at the institutional level; above 1.0 is rare enough that it tends to draw scrutiny rather than automatic praise, precisely because a number that good, sustained, is unusual. A manager who can state their own information ratio unprompted, and explain roughly what it implies, has said more in one sentence than a full paragraph reciting the formula.

What VaR is measuring

Value at risk answers a different question from tracking error, in a different unit. It estimates the maximum potential loss, in absolute currency or percentage terms, over a stated time horizon at a stated confidence level. It says nothing about a benchmark, because it is not measuring anything relative to one.

Why that silence matters

That silence is not a minor footnote; it is the single most important thing to know about VaR, and it is the reason the next two sections both exist. A VaR number tells you the threshold. It tells you nothing about what happens once you are past it, and a portfolio manager who only quotes the VaR figure has told you less than they think they have.

Why a pension trustee and a hedge fund want different numbers

The reason this distinction gets tested constantly is that it is not academic. Different owners of capital genuinely want different numbers, because they are underwriting different promises.

A trustee who mandated an index-tracking manager would be right to shrug at that manager's VaR number; it was never part of the deal. A hedge fund investor who asked their manager for a tracking error figure would get a blank look, because there may be no benchmark to track against in the first place. The two measures are not competing for the same job. They were built for different clients who wanted different things from their capital.

The seat where both numbers matter

A large asset manager frequently runs both kinds of fund under one roof, sometimes on the same floor. A long-only equity desk down the hall from an absolute-return credit fund reports to the same firm-wide risk function, and that function has to be fluent in both languages, because it is aggregating exposure across mandates that were never designed to be compared on the same scale. That is the seat where "why does this matter" stops being theoretical.

Where VaR is written into the rules

Most explanations of VaR treat it as a generic risk concept. It is also, in specific and checkable circumstances, a hard regulatory number, and the detail is more interesting than the generic version.

The UCITS link

Under the European rules governing UCITS funds, a fund must calculate its overall market exposure using one of two approaches. The simpler one, the commitment approach, converts derivative positions into their equivalent underlying exposure and adds it up.

The more advanced one is VaR, and a fund is required to use it specifically where it runs complex investment strategies, carries more than a negligible exposure to exotic derivatives, or where the commitment approach simply does not capture the fund's market risk adequately.

A fund using VaR then has to pick one of two flavours, and the two caps are numeric and enforced, not descriptive. Under the relative VaR approach, the fund's own VaR cannot exceed twice the VaR of an unleveraged reference portfolio, which caps the fund's effective leverage ratio at 2. Under the absolute VaR approach, used by funds without a natural benchmark, the fund's VaR cannot exceed 20% of its net asset value.

Both come with standard calculation parameters underneath them: a 99% one-tailed confidence level, a one-month holding period, and at least a year of historical data behind the model.

Why most long-only mandates never touch it

Here is the part that most explainers skip entirely, and it is the genuinely differentiated fact in this whole subject: a plain vanilla, long-only equity fund that does not lean heavily on derivatives typically qualifies to use the simpler commitment approach and is never VaR-limited by regulation at all. It is governed, day to day, by whatever tracking-error budget its own mandate and risk team have set internally, which is a firm's own choice rather than a regulator's number.

VaR earns its formal, numeric teeth specifically where leverage and derivatives get involved, which is exactly why it belongs to the hedge fund desk and the bank trading book far more naturally than it belongs to a long-only equity mandate.

VaR's real weaknesses, stated fairly

None of this makes VaR a bad number. It makes it a number with specific, well-understood limits, and being able to name them is a stronger answer than either praising or dismissing the measure wholesale.

• It says nothing about the size of the loss beyond the threshold. A 99% VaR describes the edge of the worst 1% of outcomes. It is silent on how bad that worst 1% gets, which is precisely the scenario a risk manager is paid to worry about.
• It is backward-looking by construction. Both the common parametric and historical-simulation approaches to VaR are built from a historical window of past returns and correlations. A regime change that has not shown up in that window yet will not show up in the VaR number either, until after it has already happened.
• A single VaR number invites false precision. Two funds with an identical VaR figure can be running very different underlying books, one concentrated in a single risk factor and one spread broadly, and the VaR number alone will not tell you which is which.

Why expected shortfall exists

Bank regulators building the post-2008 capital rules ran directly into that weakness, and changed the measure rather than patching around it. The Basel Committee's Fundamental Review of the Trading Book, published in 2016, shifted the standard measure of bank trading-book risk from value at risk to expected shortfall. Expected shortfall does not stop at the threshold; it averages the losses inside the tail beyond it, the piece of information a plain VaR number leaves out.

A candidate who can explain why regulators made that specific switch, rather than simply naming expected shortfall as a term, is showing they understand the gap VaR leaves rather than reciting its successor.

Active share is not tracking error

One correction is worth making carefully, because the two terms get used almost interchangeably in casual conversation and they measure different things.

Active share is the percentage of a portfolio's holdings that differ from its benchmark's constituents and weights. A fund that replicates its benchmark exactly has an active share of 0%; a fund holding none of the benchmark's constituents at all has an active share of 100%. It is a holdings-based measure of stock selection.

Tracking error, as this whole article has covered, is a return-based measure of systematic drift. The two usually move in the same direction, a fund that looks very different from its benchmark tends to also perform differently from it, but "usually" is doing real work in that sentence.

Confusing the two is a specific, correctable error, not an excuse to abandon either measure. A candidate who says "high active share" when they mean "high tracking error" has just told the room they learned the vocabulary without learning what it is for.

The vocabulary an interviewer assumes you already have

None of these get asked as standalone definitions. All of them get assumed the moment a related answer uses one imprecisely.

• Active risk budget — the amount of tracking error a manager is permitted to run before a client or oversight committee asks why.
• Parametric VaR — a VaR estimate built from a statistical distribution (usually assumed normal) fitted to historical volatility and correlations, rather than from actual historical scenarios.
• Historical simulation VaR — a VaR estimate built by replaying a fund's current positions through a window of actual historical market moves, rather than assuming a distribution.
• Expected shortfall — the average loss across the tail beyond a VaR threshold, rather than the threshold itself.
• Ex-ante and ex-post — forecast versus realized, applied to both tracking error and information ratio.
• Active share — the holdings-based counterpart to tracking error's return-based measure, covered above.

How these numbers show up in a junior's actual day

A first-year on a long-only desk is far more likely to spend a morning reading a risk report than calculating a VaR model from scratch. The practical work usually looks like:

• Checking that a portfolio's forecast tracking error against its stated risk budget hasn't drifted after a rebalance.
• Flagging a position that has quietly become a large factor bet nobody signed off on.
• Preparing the plain-English line for a client report explaining why the fund underperformed by more than its usual band in a volatile quarter.

On the rare desk where VaR is the house number, the daily version looks similar in spirit: checking the model's output against the limit, understanding which position moved the number, and being able to say why in one sentence to someone who was not in the room when the model ran.

The mistake that ends the exchange

How to answer this in the room

1. Open with the relative-versus-absolute split, in one sentence each, before any formula: tracking error against a benchmark, VaR in absolute terms.
2. Name a concrete example, like the 30% loss that can still be a good year if the benchmark fell the same amount. A concrete case lands harder than a definition.
3. Say which one a long-only mandate actually lives on, and why: tracking error, because the promise made to the client was benchmark-relative.
4. Name one real weakness of VaR, specifically that it says nothing about the size of the loss beyond its threshold, and that expected shortfall exists for exactly that reason.
5. Keep active share separate, and be ready to say in one sentence why it is not the same thing as tracking error.
6. If pushed further, bring up the mandate bands, and be ready to say what a given tracking error number implies about the kind of fund it belongs to.

The bottom line

Tracking error and VaR are not two versions of the same idea. One measures how far a portfolio has strayed from a benchmark it was hired to track, however loosely; the other measures how much could be lost outright, with no benchmark in the picture at all. A long-only mandate is built and judged on the first. A trading desk or an absolute-return fund is built and judged on the second.

Knowing which one your fund actually lives on, and being able to say why in one sentence, is worth more in the room than either formula on its own.