Every Projection Is a Set of Assumptions: How to Read One

Retirement calculators, mortgage estimators, a bank's illustration, a chart from a financial professional: each produces a number about the future. That number is the least informative thing on the page. What matters is what went into it.

Two kinds of input

Every projection mixes two kinds of information.

Known values are facts you can look up today: a current balance, a loan's rate and term, the amount you contribute each month, the tax rules in force this year. Calculations on known values are ordinary arithmetic. Done correctly, they give the same answer every time.

Assumptions are everything about the future that nobody knows: investment returns, inflation, how much a home will be worth, how fast a salary will grow, what tax rules will apply decades from now, how long you will live, how long you will stay in a house. Some assumptions are well informed. None of them is a fact.

A projection is only as reliable as the least reliable assumption that actually matters to the result. The first step in reading one is to tell the two kinds of input apart.

A projection is not a prediction

A projection says: if these assumptions hold, this is what follows. A prediction says: this is what will happen. The arithmetic in a projection can be exact while the result still turns out wrong, because the future didn't follow the assumptions.

One number, or a range?

A single figure, such as "you'll have this much at 65," hides how much the outcome could vary. A range shows the spread you'd get under different reasonable assumptions. A range is not vaguer than a single number. It is more truthful about what can be known.

When you see one number, ask what the range would be, and what makes it wide or narrow.

Find the load-bearing assumptions

Not every assumption matters equally. A load-bearing assumption is one that changes the conclusion when it moves within a reasonable range. In a long retirement projection, the assumed investment return and the assumed spending usually dominate. In a comparison of renting against buying, how long you stay and how much the home appreciates often decide the answer.

The way to find them is sensitivity: change one assumption at a time, by a reasonable amount, and see what moves. An assumption that barely changes the result can be left alone. One that swings the result deserves most of your attention.

Time magnifies small differences

Over a few years, a small difference in an assumed annual rate barely shows. Over decades, compounding turns the same small difference into a large one. The arithmetic below is exact; the two rates are arbitrary and are not a forecast of anything.

Arithmetic illustration: what a starting amount grows to at two assumed annual rates (not expected returns)

Years	At 4% a year	At 6% a year	Gap between them
5	1.22 × the start	1.34 × the start	About 10%
15	1.80 × the start	2.40 × the start	About 33%
30	3.24 × the start	5.74 × the start	About 77%

The same two-point difference in an assumption produces a gap nearly eight times wider at thirty years than at five. Far-off figures in any projection deserve more humility than near ones.

Ask what would change the conclusion

The most useful question about any projection is not "is this number right?" but "under what assumptions would the decision change?"

If the conclusion only flips under extreme assumptions, it is robust, and you can act on it with some confidence. If it flips under a modest, plausible change, you've learned where the real uncertainty sits, and that is where to focus before deciding.

Questions to ask of any projection

• Which inputs are known facts, and which are assumptions?
• What return, inflation and growth rates are assumed? Are they before or after fees, and before or after inflation?
• Is this a single number or a range? If a single number, what would the range be?
• Which one or two assumptions matter most to the result?
• What happens if each of those turns out worse than assumed?
• How long is the time horizon, and how far out do the assumptions start to dominate?
• What would have to be true for the conclusion to change?
• Does the person or tool producing the projection benefit from one answer more than another? That doesn't make it wrong, but it is worth knowing.

A projection that can answer these questions is a useful tool for a decision. One that can't is a number, and a number on its own tells you very little.