The Assumption That Moves the Answer Most
A retirement projection has one input with enormous leverage and very little certainty: the assumed rate of return. Moving it by a single percentage point changes a thirty-year outcome by a quarter or more, which means a plan built on one number is a plan built on a guess — and testing it at a lower number is the cheapest improvement available.
- The answer: Run the projection at your assumed return and again one to two percentage points lower, and look at whether the plan still works.
- The trap: Long-run averages are drawn from a period that included particular conditions, and fees, tax and sequence all reduce the return an investor actually receives.
- The recommendation: Use a real return — after inflation — and after fees. Nominal, pre-fee figures overstate the outcome by a wide margin over thirty years.
Where the AI summary above gets this wrong
"Historically shares have returned about 9% a year, so that is a reasonable assumption for a retirement plan."
That's surface-true. Here's what it misses:
- A nominal, pre-fee, pre-tax figure is not what an investor receives — Subtract inflation, fees and any tax, and the number that actually compounds in a plan is far lower. The gap over thirty years is most of the outcome.
- A long-run average is not a sequence — The same average delivered in a different order produces a very different result once withdrawals begin, which is the point of sequence risk.
01 Why the return assumption dominates
Compounding is exponential, so a change in the rate applies to every year and to the accumulated result of every previous year. That is why a one-point change over thirty years moves the answer far more than a one-point change in the contribution rate.
It also means the assumption is doing more work than the evidence supports. Nobody knows the return of the next thirty years, and the historical figures used to justify an assumption come from a specific period with specific conditions.
The response is not to pick a more accurate number, because there is not one. It is to test the plan across a range and see whether the conclusion changes.
02 Using the right kind of return
Use a real return, after inflation, if the spending figures are in today's money. Mixing a nominal return with real spending is the single most common error in a household projection and it makes every answer look better than it is.
Subtract fees. A 7% gross return with 1% of fees is a 6% net return, and the arithmetic in the investment costs post shows what that difference does over a long horizon.
And subtract tax where it applies. Earnings inside a retirement-phase pension are untaxed, earnings in accumulation are taxed at 15%, and earnings outside super are taxed at your marginal rate — three different net returns on the same portfolio.
Shows: the balance a projection produces at your assumed return and at a lower one, so the size of the assumption is visible. Ignores: withdrawals, contributions, sequence of returns, tax and fees beyond whatever you have already netted off the rate.
03 How to test a plan
Run the projection at your central assumption, and again at one and two percentage points lower. If the plan still works at the lower rates, the return assumption is not what the plan depends on and you can stop worrying about it.
If it fails at the lower rate, the useful question is what changes: working longer, spending less, or accepting a lower standard later. Knowing which of those is the fallback is worth more than a more precise central estimate.
Also test a bad first decade rather than only a lower average. That is a different failure mode with a different fix, described in the sequence risk reference.
If a plan only works at 7% and fails at 5%, it is not a plan — it is a bet on a number nobody can know. I would rather see a household with a plan that works at 4% and a pleasant surprise coming than one built on a long-run average that has to show up on schedule.
FAQ
How sensitive is my retirement plan if investment returns are 1% lower than I assumed?
Substantially. Over thirty years a one percentage point reduction removes roughly a quarter of the projected balance, which is more than most other single inputs can move.
What investment returns should I assume for retirement?
Use a real return, after inflation and after fees, and match it to the tax environment the money sits in. Then test the plan one and two points below whatever you chose.
Is a long-run historical average a reasonable assumption?
As a central estimate it is defensible; as the only estimate it is not. The historical figure is nominal and pre-fee, it comes from a specific period, and an average is not a sequence.
Sources
Regulator references
- ASIC Moneysmart — Choose your investments · ASIC Moneysmart · 2026Choosing investments: risk, diversification and time horizon.Last verified: 2026-09-07
- ASIC Moneysmart — Retirement planner · ASIC Moneysmart · 2026The regulator's own retirement income projection tool.Last verified: 2026-09-07
Calculator unit tests · the assertions this page's worked example is checked against, and their last result
Changelog
- 2026-09-07 — initial publish (new format)
Run this rule against your situation
See what this rule does to your own projection — month by month, to age 90.
Join the Waitlist
Jordan Reeves
Founder of Talk Through Wealth. A software engineer for over a decade before turning to retirement planning, Jordan built the projection engine after watching family members get fragmented, country-by-country advice that never reconciled. He writes about retirement the way the engine computes it: month-by-month, lifetime-long, and skeptical of any rule of thumb that hasn't been run through the math.