Costs
Part of How to approach career earnings: the ledger that is always missing
How career earnings figures are worked out
Career earnings methodology and the arithmetic of adding a lifetime: why summing decades of estimates makes the error grow rather than average away.
A career total looks like the easiest quantity in this subject. You are not modeling spending, investment or debt. You are adding up what somebody was paid. That framing is what makes the method dangerous, because the addition itself introduces problems that a single-year estimate never has.
What to take away
- Adding estimates accumulates error; it does not average it out.
- Amounts from different decades are different units and cannot be summed directly.
- The boundary of a career decides the total as much as any figure inside it.
Why the sum is not safer than its parts
There is a comfortable intuition that errors cancel: some years overestimated, some underestimated, roughly right in the end. It holds only when the errors are independent and unbiased. Neither condition is met here.
Errors are correlated because every year uses the same method and assumptions. If that method counts announced contract maximums as amounts received, it does so each year, so bias accumulates instead of washing out.
The errors are also one-sided in practice: visible events are the paid ones, while unpaid or collapsed projects leave no trace to subtract.
A career total built from thirty estimated years is not thirty times more reliable than one year. It is one year's bias applied thirty times, in the same direction.
The measurement convention for reporting a value with its uncertainty exists for situations like this. Guidance on expressing measurement uncertainty shows what a stated interval is supposed to look like.
The unit problem
An amount received early in a career and an amount received late in it are not the same unit. Prices change, currencies move, and tax regimes are rewritten. Adding them without stating a basis produces a number in no unit at all.
There are two defensible treatments and one common bad one.
- State everything in nominal terms of the year received, and say so. Honest, but the total then cannot be compared to any present-day figure.
- Restate everything in one year's terms using a published index, and name the index. Defensible, and it requires the compiler to say which index and which base year.
- Add the nominal amounts and print the total as though it were current money. This is what almost every published career total does, and it is not a defensible option.
The same problem appears with currency. An international career is paid in several currencies over decades. Converting the whole total once at a recent rate produces a figure nobody ever held.
The boundary problem
Before anything is added, the compiler decides what counts, and these decisions are rarely stated.
| Boundary decision | Why it moves the total |
|---|---|
| Does commercial and sponsorship work count as career earnings? | For many people this exceeds the income from the work itself |
| Do business ventures unrelated to the profession count? | Including them makes the total a measure of business activity |
| Does income from a company the person owns count in full? | The company has its own costs, staff and tax |
| Do continuing payments for past work count in the year earned or the year paid? | These arrive over decades and can be counted twice |
| Where does the career start? | The unpaid and low-paid years are usually excluded silently |
| Is the total gross or net of representation and tax? | Almost always gross, almost never stated |
Two compilers making different choices here can produce totals that differ substantially while both being internally consistent. That is a sign the quantity is under-defined, not that one of them is wrong.
The double counting trap
Career totals are unusually prone to counting the same money twice. A signing payment gets counted in the year it was paid and again inside the contract total. Continuing payments get counted when earned and again when received. A share of a project's receipts gets counted alongside a fee that was actually an advance against that share.
Sponsorship gets counted as both business income and personal income where a company sits between them, a structural distinction the IRS sets out in its overview of business structures. None of these is detectable from outside, because the compiler is working from announcements rather than accounts.
What an honest career figure looks like
Narrow, dated and documented. A published scale minimum for a defined category of work in a defined year. A prize structure for a named event. An amount established by a court in a specific dispute. A disclosed figure where a disclosure obligation existed.
Each statement is real about a real period, but none is a career. That is the trade, the same one made everywhere else on this site, as set out on estimate methodology.
Where income accrues across a working life is on career earnings. Why announced contract values inflate the inputs is on contracts and salaries. How these totals get ordered into a table is on richest rankings.
Common questions
If a compiler used only reported figures, is the total at least a floor?
No, for two reasons. Reported figures are often maximums that were not fully paid, so the inputs are not floors either. And a floor on gross pay says nothing about a net position, which is what readers take from it.
Would adjusting for inflation fix these totals?
It fixes one of the four problems. The correlated bias, the boundary choices and the double counting all survive the adjustment untouched.
Is a career total ever more knowable than a wealth figure?
Slightly, because it needs no assumption about spending or investment. It still needs every year's pay, which is not public, so the improvement is a matter of degree and not enough to make it publishable.
