How to Compare Two Uses of Money Without Fake Certainty

The cost you cannot see

Every choice about money has two outcomes. One of them happens, and you can watch it: the account statement, the receipt, the thing you bought. The other one does not happen, so there is nothing to watch. It leaves no record, sends no notification and never appears on any statement.

That second outcome is the cost of the choice. Economists call it opportunity cost, and the name is unhelpful because it sounds like a category of expense. It is not an expense. It is the difference between what happened and what would have happened instead — and the reason it is so easy to ignore is not that people are careless with money. It is that the evidence for it does not exist.

This article is about how to construct that missing evidence honestly, and about the several ways the construction goes wrong.

Opportunity cost is not a property of an option

Here is the first thing that trips people up. “What is the opportunity cost of this investment?” has no answer, and it has no answer for a structural reason rather than a factual one: opportunity cost is not a property of the option you are looking at. It is a property of the pair.

Ask what a €20,000 purchase costs you and the honest response is another question: instead of what? Against leaving the money in a current account, the answer is one number. Against a portfolio you would actually have bought, it is a different number. Against paying down a debt, different again. None of those is more correct than the others in the abstract. They are answers to different questions, and the question is set by the alternative.

This has an immediate practical consequence, and it is the one worth carrying away from this section:

A comparison against an alternative you would not actually have chosen is not a conservative estimate. It is a precise answer to a question nobody asked.

If you would never have put the money into equities, comparing against an equity return does not “put the purchase in context”. It manufactures a cost out of a decision you were never going to make. The alternative has to be a real one — an option you would genuinely have taken, at an assumption you would genuinely have used.

What makes a comparison like-for-like

Once you have chosen the alternative, the comparison itself can still be broken in a way that is much harder to see, because it produces a number that looks fine.

The requirement is that the two paths differ only in what you are trying to measure. If Option A ends higher than Option B, that difference should be attributable to the options, not to the fact that one of them was quietly given more money or more time.

Three things have to match:

The same starting amount. Both paths begin from one pool of money — the amount you actually have — before either path’s own costs are taken out of it.

The same recurring amount, at the same times. If you would add €300 a month to whichever option you chose, both paths get €300 a month, on the same dates. Not €300 to one and €400 to the other, because the extra €100 is a bigger budget rather than a better option, and the arithmetic cannot tell the difference.

The same horizon. Comparing ten years of one against fifteen years of the other measures five extra years of compounding and reports it as a quality difference between the options.

Each of these sounds too obvious to state. Each is violated constantly, usually in a spreadsheet, usually by accident, and usually in the direction the author was already hoping for.

Why the funding rule is stricter than it looks

There is a subtler version of the same failure, and it is worth its own section because it is the one that survives review.

Suppose one option requires €500 a month and the other requires €200. It is tempting to model exactly that: €500 into one, €200 into the other, because that is what each one costs. But then the second path has €300 a month that the model simply loses. It vanishes. In reality that €300 went somewhere — into a savings account, into spending, into the other option — and wherever it went, it did something.

A model that drops it is not neutral. It quietly gives the expensive option credit for money the cheap option also had, and the resulting gap is partly the option and partly the accounting. Fixing it properly means deciding what the unused budget does, which is a whole second set of assumptions: does it sit in cash, does it earn anything, is it spent?

That is why the honest version of this comparison holds the recurring amount equal and says so, rather than modelling unequal cash flows and hoping nobody looks. The Opportunity Cost Calculator takes the stricter position deliberately: one allocation amount, one schedule, both paths. It is a real limitation, and it is a smaller one than the alternative.

Upfront costs, and the capital that never compounds

Now the part that produces the most counter-intuitive arithmetic on this whole subject.

A one-time cost paid at the start — an entry charge, a setup fee, a premium over what an asset is worth, a commission — does not simply reduce your outcome by its own size. It reduces the amount of capital that begins compounding, and that reduction compounds too.

Pay €5,000 out of €50,000 and you are not €5,000 behind at the end of twenty years. You are €5,000 behind plus everything those €5,000 would have become. At a 5% annual rate over twenty years, that is a little over €13,000. The cost was €5,000. The gap it opens is nearly three times that, and the multiplier is just the horizon.

This is exactly the same mechanism as fee drag, arriving all at once instead of in slices. It also explains why the same upfront cost is a very different decision at three years and at thirty, which is not usually how such charges are presented.

Why a higher annual rate can start behind

Put the previous two sections together and you get a result that surprises most people the first time they see it.

Take two options funded from the same €50,000. Option A charges €5,000 upfront and is assumed to grow at 7% a year. Option B charges nothing and is assumed to grow at 5%. A has the higher rate — by two full percentage points — so it should be ahead.

It is not. It starts €5,000 behind, and at the beginning a two-point rate advantage on €45,000 is worth a few hundred a year. Compounding takes time to overcome a deficit that exists from day one. In this case A does not pass B until month 68 — nearly five years and eight months — and only then begins pulling away, finishing about €7,000 ahead at ten years.

Two things follow from that, and they matter in opposite directions.

The first is that a higher rate does not mean immediately ahead, so a comparison made at year three would have reported the opposite answer and been arithmetically correct in doing so.

The second is the one that gets forgotten: the whole recovery depends on the 7% actually happening. The rate is an assumption. The €5,000 is not — it is paid on day one, whatever happens next. The certain cost is paid up front and the uncertain benefit arrives over years, and any comparison that treats those two as equally solid has quietly made the strongest assumption on the page.

Crossover points, and what a horizon decides

The month at which one path passes the other is worth reporting, because it makes something visible that a single ending value hides: the answer depends on when you stop looking.

A path that is behind at year three and ahead at year ten has not changed. The comparison changed, because the horizon changed. So a comparison stated without its horizon is not a fact about two options — it is a fact about two options and one particular date, and dropping the date makes it sound like a permanent ranking.

It is worth being precise about what counts as a crossover, because the loose version is misleading. Two paths that begin at the same value and then separate have not crossed. Nobody was ahead, so nobody was overtaken. A genuine crossover needs one path strictly ahead before and the other strictly ahead after — a real reversal, not a parting. The distinction sounds pedantic until you notice that the loose version lets any comparison at all be described as a recovery.

Break-even return: a hurdle, not a forecast

When one path ends lower, there is a natural follow-up question: what would it have needed?

That question has an exact answer. Hold everything else fixed — the same starting amount, the same upfront cost, the same allocations, the same horizon — and solve for the annual rate at which the lower path reaches the higher one’s ending value. In the example above, Option B would need about 5.88% instead of the 5% entered: roughly 0.88 percentage points more.

This is the single most useful number in the whole comparison, and also the most frequently misread. Three things it is not:

It is not a forecast. Nothing about the arithmetic says that rate will happen.

It is not evidence that the rate is achievable. The solver will happily report that a path needs 40% a year. That is a description of how far behind it is, not a plan.

It is not a threshold of adequacy. “B only needs 0.88 points more” and “B needs a whole 0.88 points more” are the same fact. Whether that gap is small depends on what the two options actually are, which the arithmetic does not know.

What it is genuinely good for is testing how fragile your conclusion is. If the lower path needs a rate barely above what you entered, the ranking rests on a rounding difference in an assumption you invented, and it would reverse if you had been slightly more optimistic about it. If it needs a rate far above, the conclusion is robust to your own uncertainty — still an assumption, but one that would have to be badly wrong to matter.

Expected return is not the same as certainty

Every comparison of this shape puts two numbers side by side as though they were the same kind of thing. Usually they are not.

An assumed 6% on a volatile asset and a guaranteed 4% from paying down a fixed loan are not comparable quantities. The second is a rate you will receive. The first is the middle of a distribution you cannot see, and the actual outcome could be 20% or −30%. A model that multiplies both by twenty years produces two numbers in the same font, and the presentation asserts an equivalence the underlying facts do not support.

Two consequences worth holding on to:

A guaranteed rate is worth more than the same expected rate. How much more is a question about you — your tolerance for the range of outcomes, and what a bad one would cost you — and no calculator answers it.

The sequence matters, and constant-rate models cannot show it. Two paths that average the same annual return produce different amounts depending on the order the years arrive in, particularly when money is being added or taken out. Constant-rate arithmetic reports the average and shows none of the range.

What the calculator does not control

The funding rule controls two things: the money and the time. That is a real achievement, and it is a narrow one. Here is what is still uncontrolled after the comparison is done.

Risk. Not modelled at all. Two paths ending at the same value can differ completely in how likely that value was.

Liquidity. Not modelled. Money you can reach on Monday and money locked for five years are shown as the same figure, and the difference between them is sometimes the entire decision.

Tax. Not modelled. Two options with the same gross outcome can differ substantially after tax, and the difference depends on your circumstances rather than on the products.

Effort and competence. Not modelled. An option requiring active management is being compared with one that requires none, at the same figure, as though your time were free and your judgement uniform.

Concentration. Not modelled. An option that adds to an exposure you already carry is a different proposition from one that does not, and the arithmetic cannot see your existing position.

None of these is a footnote. Any one of them can be larger than the projected difference the comparison reports.

Non-financial value, and why a lower path can be right

The last omission needs its own section, because it is the one where an arithmetic answer most often gets mistaken for a complete one.

A comparison of two financial uses of money measures financial value. That is all it measures. If one of the two options is a purchase — a car, a course, a house that you will live in, a year of not working — then the model records its financial residual value, which may be zero, and records nothing else about it.

The €14,000 that a spent €10,000 might have become in ten years is a real number, and it is not the cost of the purchase. The cost of the purchase is that number minus whatever the purchase was worth to you, and the second term is not in any currency this model handles. Sometimes it is obviously larger. Sometimes it is obviously smaller. The arithmetic is silent either way, and silence is not zero.

This is why the central position of this article is a negative one:

A higher projected financial value does not automatically make an option the better overall decision. The comparison is only as useful as the alternative, the assumptions and the trade-offs left outside it.

A worked example

Two options, funded identically from €50,000, over ten years. No recurring allocations, so the arithmetic stays checkable.

Common amount available today   50,000.00     50,000.00
                                Option A      Option B
Less upfront cost                5,000.00          0.00
Initial capital deployed        45,000.00     50,000.00
Annual value change                  7.00%         5.00%
Ending value after 10 years     88,521.81     81,444.73

Option A ends €7,077.08 higher. Reading down each column tells you how each path got where it did — what was available, what the upfront cost removed, what was deployed, and what the modelled change added.

What the table does not tell you is how much of that €7,077 was “caused by” the upfront cost and how much by the rate difference. Those two numbers do not exist. The growth line for A already reflects the smaller capital its upfront cost left to deploy, so any split between them depends entirely on which one you charge first — and charging them in the other order gives a different answer from the same arithmetic. A decomposition that looks causal without saying which counterfactual it assumed is not a finding. It is a presentation choice wearing the clothes of one.

Three further facts about this example, each of which changes how it should be read:

• A is behind until month 68. A ten-year comparison and a five-year comparison give opposite answers, with nothing changed but the date.
• For B to end level, it would need about 5.88% rather than 5%.
• If A’s 7% turns out to be 5.5%, A ends below B. The €5,000 was still paid.

Common ways this goes wrong

Comparing against an alternative you would not have chosen. The most common error and the least visible one, because the arithmetic is flawless.

Giving the two paths different budgets. Usually accidental, and it credits the better-funded option with money rather than merit.

Treating an assumed return as a fact. The rate is the one input carrying all the uncertainty, and it is typed with the same confidence as the amount.

Reading a crossover as a permanent ranking. It is a fact about one horizon.

Reading a break-even rate as a plan. It says what would be needed, not what is available.

Decomposing the gap into causes. Feels rigorous, requires a counterfactual ordering nobody stated, and produces numbers that change when the ordering does.

Comparing a certain saving with an uncertain return. Same units, different kinds of number.

Forgetting that nominal is not real. Every figure in this kind of comparison is in future money. What it buys is a separate question.

Letting the arithmetic answer a question it did not ask. The model reports a financial difference. Whether it is worth the difference in risk, liquidity, tax, effort and everything the model never measured is your call and stays your call.

Where this actually comes up

Entry charges and premiums. Any one-time cost that reduces the capital that starts compounding: an initial charge, a spread, a markup over an asset’s underlying value, a purchase commission.

Packaged exposure against direct exposure. A product that gives access to an underlying asset, at a cost, against holding the asset itself. The comparison is whether the difference in outcome justifies the difference in cost, and the answer is horizon-dependent for exactly the reasons above.

Prepaying against investing. Reducing a certain future cost against pursuing an uncertain future gain. The arithmetic is easy and the equivalence between the two rates is the hard part.

Capital tied up alongside a purchase. Where using a product properly requires holding something else as well, that second position is capital committed to the same decision, and leaving it out of the denominator flatters the comparison.

Spending against investing. Financially one-sided by construction, since the spending has no modelled residual value. Which is precisely why the answer needs the term the model cannot supply.

The formulas, the assumptions and the limits

capital deployed        S - U            per option
monthly factor          (1 + r)^(1/12)   per option, never r / 12
ending difference       D = V_A(T) - V_B(T)
opportunity-cost gap    OC = the size of D, without its sign
accounting bridge       deployed + allocations + growth = ending value
crossover               a genuine sign reversal in D over the months
reversal rate           solve r so the lower path reaches the higher ending value

What it assumes: one constant effective annual value change per option for the whole horizon, already net of recurring product costs; one amount available, one recurring allocation, one schedule and one horizon shared by both paths; each upfront cost paid once at the start and never recovered; nothing withdrawn from either path; and no rounding until display.

What it excludes: inflation and purchasing power, tax of every kind, risk, volatility, sequence of returns, probability, liquidity, delayed starts, unequal recurring allocations, changing rates, debt amortisation, leverage, and every form of value that is not denominated in money.

Where an answer does not exist, the model says so in words rather than printing one. The reversal rate is reported as unavailable rather than as an invented number when no reliable answer can be found, and no figure is allowed to reach the page as an infinity or a negative zero.

Run it on your own figures

The Opportunity Cost Calculator applies exactly the model above to numbers you enter: what you have, what you would add and how often, over what horizon, and for each option its own upfront cost and its own annual value-change assumption.

It leads with the projected difference between the two paths, stated as a direction rather than a ranking. It reports whether one overtakes the other and when, what the lower one would need to close the gap, and an accounting view of each path that is explicitly not a causal attribution. It names, beside the result rather than at the foot of the page, the four things it did not control.

It runs entirely in your browser. UBWHY does not receive the values you enter, stores none of them, puts none of them in a link, and fetches no market rate from anywhere.