A category average is read as what the category did over the period. It is what the surviving members of the category did over the period, which is a different statistic with a predictable direction of error.

When a scheme is compared against its category average, that average is almost always built from whichever schemes exist on the day of the comparison - not from every scheme that existed when the measurement period began. A fund merged into another, or wound up entirely, is not on today’s list and is not in today’s average. What follows concerns the size of that omission in Indian equity categories, and why it does not average out.

The Attrition Is Larger Than It Looks

The drop-off is not a rounding effect. S&P’s SPIVA work on the Indian market found that of 116 large-cap equity funds available as of June 2011, 37 had merged or liquidated over the five years to June 2016 - a survivorship rate of 68%. Roughly one scheme in three did not finish the period it started.

Large-cap is the well-populated, well-capitalised end of the market. Categories with smaller schemes and shorter track records attrit faster, which matters because those are also the categories where investors lean hardest on the category average to judge a manager.

Funds Do Not Disappear at Random

If exits were random, the omission would widen the error bars without moving the centre. They are not random. An asset manager merges its weakest, smallest and worst-performing schemes into stronger ones far more often than the reverse: a scheme with a shrinking asset base and a poor five-year record is a natural merger candidate, and one with a strong record and growing assets is not.

The funds most likely to vanish are therefore disproportionately the ones that would have dragged the average down. A survivors’ average does not merely lose observations - it loses the specific observations sitting in the left tail. The resulting number is biased upward by construction, and the bias is largest exactly where attrition is highest.

What the Gap Looks Like on a Cohort

Here is the mechanism on an illustrative set of ten schemes:

Ten illustrative mid/small-cap schemes from five years ago - eight still exist, two were merged or closed. Numbers are constructed to demonstrate the mechanism, not drawn from real schemes.

The eight survivors return 14%, 16%, 11%, 9%, 18%, 13%, 15% and 10%, averaging 13.25% - the number a rating site would display today. The two merged away, at -2% and 3% before they disappeared, pull the true starting-cohort average to 10.70%. The 2.55-point gap arises purely from which funds are still around to be counted, not from any difference in how the calculation is performed.

Note what that does to a manager assessment. A fund returning 12% over the period looks like a below-average performer against the survivors’ 13.25% and an above-average one against the cohort’s 10.70%. The same fund, the same period, two opposite conclusions, decided entirely by which denominator was used.

Where the Bias Bites Hardest

Not every category average is badly wrong, and not every closed fund was a laggard - some mergers are ordinary business consolidation carrying no performance signal. The claim is narrower and structural: a category average is a survivors’ average unless a report explicitly corrects for it, as the SPIVA scorecards do by publishing survivorship alongside performance.

It follows that the gap is widest in the categories with the lowest survivorship. Those tend to be the smaller-capitalisation and thematic categories, where scheme count is lower, asset bases are thinner and consolidation is more frequent - and where the marketing case for active management leans most heavily on beating the category average.

What This Means for Allocators

Ask for the survivorship rate before using the average. A category average quoted without one is an unqualified statistic, and the qualification moves the number.

Prefer sources that report attrition alongside performance. SPIVA does this; most retail factsheets and rating screens do not, and the difference is not presentational.

And treat the denominator as part of the claim. The average is real. The cohort it came from is smaller than the one you started with. The funds missing from it are not missing at random.

Back to Writing