What Actually Drives a ReturnStart here

Every return decomposes into four things, and only one of them is the story people tell. A decomposition applied to bonds, equities, funds and structured products in turn.

The decomposition

Any holding-period return, in any asset class, can be split into four components. They are not a model — they are an identity, and they always add up:

• Income — what the asset pays you for holding it: a coupon, a dividend, a rent, a funding rate. It is contractual or near-contractual, and it is the only component that is knowable in advance.
• Growth — how much the underlying cash flow itself grows: earnings, rents, the principal uplift on an index-linked bond.
• Re-rating — what other people are willing to pay per unit of that cash flow. A multiple, a yield, a spread. This is the component with no anchor.
• Currency — the exchange-rate move, which is a full component in its own right and routinely larger than the other three combined over one year.
• Costs — subtracted with certainty from all of it, which is why they get a whole section here.

The value of the identity is diagnostic. When someone describes a return, ask which component they are claiming. Most disagreements about markets are disagreements about one of these four, misfiled as a disagreement about the asset.

Applied: a government bond held to maturity

• Income — the coupon, known exactly.
• Growth — zero. The cash flows are fixed; that is what "fixed income" means.
• Re-rating — the yield move, which is everything for a bond sold early and nothing for one held to maturity. This single line explains why "bonds are safe" and "bonds lost 13% in 2022" are both true statements about different holding periods.
• Currency — zero domestically, potentially dominant otherwise.

The whole return of a held-to-maturity government bond is therefore its purchase yield. Nothing that happens to its price in between changes the outcome, only the path — which is the point the bond-versus-fund comparison turns on. Price the components with the duration calculator.

Applied: an equity index over a decade

• The horizon flips the ranking. Over one year, re-rating swamps everything. Over thirty, it is arithmetically forced towards zero, because a multiple cannot expand indefinitely — and income plus growth is all that remains.
• This is why the same market can be "expensive" and "a good long-run holding" simultaneously. The two statements are about different components.
• Run the earnings yield and growth against the bond yield with the valuation check: it is exactly this decomposition in three inputs.

Applied: a fund, where a fifth component appears

• A fund's return is the index return minus the ongoing charge minus the tracking gap — and the tracking gap is itself a decomposition of tax treatment, cash drag, trading costs and securities-lending revenue.
• Two funds on the same index can differ by 40 basis points a year with identical headline charges, and every basis point of it is in that residual. The tracking-difference decomposition separates it.
• Compounded, the residual is not small. Over 25 years, 0.40% a year is roughly a tenth of the ending value. It is invisible in any single year and decisive across a working life.

Applied: a structured product, where the decomposition is the whole analysis

• Income — the coupon, which is the premium on an option you sold, relabelled.
• Growth — usually stripped out entirely, because the underlying is a price index and the dividends accrue to the issuer.
• Re-rating — capped by construction, which is what funded the coupon.
• Costs — the issuer margin, taken up front and disclosed only partially. See how to read a term sheet.

Applying the identity to a structured product tends to end the conversation quickly, because it makes visible that the "income" and the "capped upside" are two sides of one trade rather than two separate features.

What the decomposition is good for, and what it is not

• Good for: telling whether a claimed return is repeatable. Income and growth can repeat. Re-rating cannot repeat indefinitely, by arithmetic.
• Good for: locating where a disagreement actually is, which is almost never where the argument is being had.
• Not good for: forecasting. Every component is easy to measure backwards and hard to predict forwards. The identity holds regardless; the inputs are the problem.
• Not a valuation method. It tells you what happened, or what would have to happen. It does not tell you what will.