AUM growth is not flows: reading ETF asset changes correctly

The most commonly quoted number about any ETF — its assets under management — is also the most commonly misread. AUM moves for two reasons at once: the market value of what the fund holds changes, and investors create or redeem shares. A headline like “assets grew 8% this quarter” says nothing about which of the two did the work.

The decomposition

Over any period, the change in AUM can be split into a market effect — what the starting assets would have become given the fund's return — and a flow effect, the net value of creations and redemptions. A simple estimate of net flows is the ending AUM minus the starting AUM grown by the period return. If a fund starts at 100 million, returns 5%, and ends at 110 million, roughly 5 million of the 10 million change is market effect and 5 million is estimated net inflow.

This is an estimate, and honest analytics label it as one. Distributions, fee accruals, currency effects for funds holding foreign assets, and timing mismatches between NAV and share-count observations all leave a residual. When official creation and redemption figures are available and licensed, they are the better source; the estimate is the fallback, not the equal.

Why the distinction matters

• A fund can post strong AUM growth entirely from market performance while quietly bleeding redemptions — a sign of distribution weakness that raw AUM hides.
• Flows, not AUM, are the measure of commercial momentum. They show decisions investors made, not markets they happened to be exposed to.
• Comparing issuers by AUM growth mixes their market exposure with their sales success. Comparing by estimated net flows isolates the part an issuer can act on.

Traps in the estimate

Flow estimates go wrong in predictable ways. Using a shares-outstanding method with split-unadjusted share counts produces phantom flows on split dates. Mixing a NAV from one source with a share count from another, observed at a different time, produces noise that looks like daily flow. Summing flows across share classes denominated in different currencies without conversion produces numbers that mean nothing at all.

The remedy is procedural rather than clever: state the method, keep it consistent, label estimates as estimates, and never silently switch methods between periods or between funds in the same comparison. When an input is missing, the right output is “unavailable,” not zero — a zero is a claim that nothing happened, which is rarely what you know.

ETF Engine's analytics follow these rules: every flow figure carries its method and as-of date, estimates are labeled, and missing data is shown as missing. It is less tidy than a complete-looking table, and considerably more useful.