How do you choose the right benchmark for a portfolio?

Start from the portfolio's investable universe and constraints, then pick the published index that most closely matches them on size segment, geography, sector scope and currency. The benchmark should be something the manager could plausibly have held instead, it should be investable, and it should be chosen before the period being measured rather than after.

Why does the benchmark choice change the alpha number?

Alpha is defined as return beyond what the benchmark and risk exposure explain. If the benchmark excludes a segment the portfolio holds, the return from simply owning that segment gets counted as alpha. Change the benchmark and the same portfolio, with the same returns, produces a different skill measurement.

Should a portfolio be compared to a price index or a total return index?

A total return index, because portfolio returns normally include dividends received. Comparing a dividend inclusive portfolio return to a dividend excluding index return quietly credits the portfolio with the market's dividend stream. In India, mutual fund performance is required to be benchmarked against total return versions of indices for this reason.

Benchmark Selection for Portfolios: How the Wrong Benchmark Invents Alpha

A benchmark is not a formality attached to a fact sheet. It is the definition of what counts as skill. Because alpha is measured as the return left over after the benchmark explains what it can, the choice of benchmark mechanically determines how much alpha a portfolio appears to have produced, without anything about the portfolio changing at all.

Get it right and your performance numbers answer a real question. Get it wrong and you spend years measuring the return of a size segment or a sector tilt and calling it judgment.

What a benchmark is supposed to do

A benchmark represents the opportunity set. It answers a specific counterfactual: if the manager had made no active decisions and simply held the market they are mandated to invest in, what would have happened?

Everything above that line is attributable to the active decisions. Everything below it is the cost of those decisions. That is the whole logic, and it only holds if the benchmark genuinely represents the alternative.

The industry has converged on a set of properties a benchmark should have. They are worth stating plainly because most benchmark disputes are a violation of one of them.

Specified in advance. The benchmark is chosen before the measurement period. A benchmark selected after seeing the results is not a benchmark, it is a framing device.
Investable. It should be possible to hold the benchmark itself, at scale, through an index fund or a replicating portfolio. A benchmark nobody could own is not a fair alternative.
Measurable. Its return must be computable on the same frequency and the same convention as the portfolio’s.
Unambiguous. Constituents and weights must be knowable, published, and identifiable at any date.
Appropriate. It must match the portfolio’s actual investable universe, not an adjacent one.
Accepted by the manager. If the manager does not regard it as representing their opportunity set, the measurement will be argued about rather than acted on.

Step one: describe the investable universe honestly

Benchmark selection starts with a description of what the portfolio is allowed to do, and then a description of what it actually does. These are not always the same thing.

Write down the constraints. Which market and exchanges. Which size segment: large cap, mid cap, small cap, or unconstrained. Which sectors are permitted or excluded. Is cash allowed to be large. Are unlisted or illiquid positions permitted. Is there a concentration limit, a single stock cap, or a sector cap.

Then look at the holdings history. If the mandate says multi cap but the portfolio has been almost entirely large cap for years, the effective universe is large cap. Benchmarking that portfolio to a broad index means the entire performance difference caused by not owning smaller companies is being scored as manager decision making, when it was really a standing structural position.

The exercise is uncomfortable on purpose. Most benchmark mismatches survive because nobody writes the two descriptions side by side.

Step two: choose the index that matches, not the one that flatters

Given an honest universe description, the candidate benchmarks usually narrow to two or three published indices. Choose on match quality, along these axes.

Size segment. A large cap benchmark for a large cap portfolio, a broad market benchmark for an unconstrained one. Size mismatch is by far the most common source of fake alpha, because the size segments of a market behave differently for long stretches.

Breadth and concentration. A concentrated headline index and a broad market index carry different sector exposures and different single stock concentration even when they overlap heavily. If the portfolio is deliberately spread across mid caps, a headline index will not represent it.

Sector scope. A sector or thematic portfolio should be benchmarked against that sector or theme where a credible published index exists, with the broad market shown alongside for context rather than as the primary benchmark.

Return convention. Use the total return version. Portfolio returns include dividends received, so the benchmark must include them too, otherwise the market’s dividend stream shows up as manager skill.

Currency and geography. Obvious in principle, routinely violated when a portfolio holds a slice of foreign listed exposure with no corresponding benchmark component.

When no single index matches, the disciplined answer is a blended benchmark: a fixed weight combination of published indices reflecting the mandate, for example a stated split between a large cap index and a mid cap index, rebalanced on a stated schedule. Write the weights and the rebalancing rule down in advance. A blend whose weights are revisited whenever results disappoint is not a benchmark either.

Step three: verify the fit with numbers, not intuition

Once you have a candidate, test whether it actually explains the portfolio.

Look at r squared. This measures how much of the portfolio’s movement the benchmark explains. A benchmark that explains very little of the variation is probably the wrong benchmark, or the portfolio is doing something the benchmark cannot see.

Look at tracking error. Tracking error is the volatility of the difference between portfolio and benchmark returns. Very high tracking error against a supposedly matched benchmark is a warning that the universes differ.

Look at beta. A portfolio with a persistent beta far from one against its benchmark carries a systematic exposure difference. Alpha measured without accounting for that is partly just leverage or defensiveness.

Decompose the difference. Split the active return into what came from sector allocation, what came from stock selection within sectors, and what came from being in a size segment the benchmark does not cover. If the third bucket dominates, the benchmark is wrong.

Test stability. Run the fit over multiple sub periods. A benchmark that matches in calm markets and diverges wildly in stressed ones is telling you the portfolio has an exposure that only appears under stress.

Step four: govern it

Benchmark discipline is a governance problem more than an analytical one.

Document the benchmark, the reason it was chosen, and the date of the decision. Review it when the mandate changes, not when the results change. If a benchmark must be changed, record both the old and the new series and report the transition explicitly rather than quietly restating history. Keep a consistent return convention, a consistent frequency, and a consistent fee treatment, gross or net, on both sides.

One more discipline matters for anyone testing strategies rather than measuring live portfolios: the benchmark series you use must be the one that was knowable at each historical date. Index constituent lists change, and a study that measures a strategy against today’s index membership applied to past dates has borrowed information from the future.

What benchmark selection does not tell you

A good benchmark does not make alpha real. Even a perfectly matched benchmark leaves you with a difference that could be skill, could be an unmeasured risk exposure, or could be luck over a short window. Alpha is a residual, and residuals absorb everything you failed to model.

It does not settle statistical significance. Outperformance over a few years, against any benchmark, is usually within the range of chance. The benchmark tells you the sign of the difference, not whether it is meaningful.

It does not capture risk taken. A portfolio can beat its benchmark by holding fewer, more concentrated, more volatile positions. The excess return is real, the risk profile is different, and the benchmark comparison alone will not show that. Risk adjusted measures and tracking error have to be read alongside.

It does not account for constraints the benchmark never faced. Liquidity limits, mandate restrictions, redemption flows, taxes and transaction costs all sit between a portfolio and its benchmark. An index has no cash drag, pays no brokerage, and never has to meet a redemption.

It cannot fix a portfolio with no coherent universe. If a portfolio genuinely roams across size segments, geographies and asset types with no stated policy, no benchmark will represent it fairly. The right response is to define the mandate, not to keep hunting for an index that fits the results.

Portfolio metrics explained: the hub for how portfolios and benchmarks are measured.
Total return index vs price index: why dividends have to be on both sides of a comparison.
Nifty 50 vs Nifty 500: matching the benchmark to the size segment a portfolio actually owns.
What is tracking error: how far a portfolio drifts from its benchmark, and what that implies.
What is alpha in investing: why alpha is a residual and what it quietly absorbs.

This article is educational. Altys Labs is not a registered research analyst or investment adviser, and nothing here is investment advice or a recommendation to buy, sell, or hold any security.