living together
Economics keeps two telescopes. The first, Marshall's, studies one market with everything else frozen — ceteris paribus, other things being equal. The second, Walras's, lets every market move at once and asks where the whole system settles. The first is called partial equilibrium; the second, general equilibrium; and the distance between their answers is one of the most useful numbers in the world.
The trap is that the partial answer is always easier, always cleaner, and conditional on a freeze the sentence never mentions. "Tax coffee and coffee drinkers switch to tea" is a fine sentence until tea's price rises to meet its new demand, and the coffee market you finished analyzing starts moving again because of your own conclusion. Below are two markets, coupled, computed live. Freeze one, tax the other, and measure exactly how far "other things being equal" drifts from where things actually go.
| round | coffee price | tea price |
|---|
the one-market telescope
The partial answer isn't wrong — it's round one. Tax coffee and the coffee market clears exactly as Marshall's telescope predicts: price up, quantity down, drinkers drifting toward tea. That drift is where the loan comes due. Tea's price rises to meet the arrivals, and dearer tea sends some drinkers back toward coffee, which moves the coffee price the analysis had already filed as finished. Round three is smaller than round two, round four smaller still, and the sum of the whole convergent series is the general answer — the only price at which no market has a correction left to make.
The ledger above shows the series. With substitutes, every round pushes the coffee price the same direction, so the forecast undershoots — the true burden of the tax is larger than the one-market telescope reports. Flip the coupling to complements and the rounds alternate sign: the forecast overshoots instead. Either way the error has a shape, and the shape is set entirely by the couplings you froze.
what to try
Set coupling to 0.00 and tax away. The forecast and the settlement agree to the penny, at every tax — when the markets truly don't touch, holding the rest fixed costs nothing. Partial equilibrium is exact in exactly one place, and this slider position is it.
Push coupling toward +0.90 with the tax high. The rounds ledger lengthens, every round adds in the same direction, and the miss climbs — the tighter the substitution, the more the frozen market steals from the forecast.
Drag coupling negative — coffee and tea as complements, morning partners rather than rivals. Now the rounds alternate and the forecast lands past the settlement. The one-market answer misses in whichever direction the couplings choose, which is the point: the error's sign lives outside the market you studied.
Toggle freeze tea. The system now does what the forecast assumed — tea's price pinned by decree — and coffee settles exactly on the green figure. Ceteris paribus is a perfectly good prediction about a world that enforces it. The question is always who is doing the enforcing.
beyond markets
Raise one team's salaries and the plan is priced against the neighboring teams staying put; they interview elsewhere by Thursday. Cut one meal and the analysis holds appetite fixed; dinner clears the deficit. Add a lane, and the drivers who "would have" taken the old routes re-route — the Braess page on this site is a general-equilibrium lesson wearing asphalt. Announce a metric, and the behavior the metric was calibrated on adjusts around it, which is Goodhart's law stated as a failed ceteris paribus. Even policy has the disease in its purest form: evaluate a rule by how people behaved under the old rule, and the evaluation dissolves the moment the new rule changes the behavior it was fitted to.
The lens hands you one discipline: before trusting a one-market answer, name what you froze, and ask what it does when it thaws. Sometimes the honest reply is "nothing much" — weak coupling is real, and round one is then the whole story. The failure is silent freezing: conclusions that are loans against markets nobody remembers taking out.
the mapping
| Mathematics | Life |
|---|---|
| partial equilibrium | The plan analyzed with everything else held still — one market, one meal, one team, one metric. |
| ceteris paribus | The freezing assumption itself: honest when named, dangerous when silent. |
| general equilibrium | Where the whole coupled system settles once every market has answered every other. |
| the coupling | How strongly the frozen things respond — substitutes push back, complements pull along, zero leaves the forecast exact. |
| the rounds | The forecast as round one of the truth; the miss as the sum of every round you declined to compute. |
| the settlement | The state with no corrections left — reached by the system whether or not the analyst follows it there. |
where the metaphor tears
A general-equilibrium answer needs the whole system modeled, and a misspecified whole can lose to an honest part. The two-market toy converges because its couplings are true by construction; a model that freezes nothing but guesses every linkage has traded a named assumption for a hundred unnamed ones. Sometimes the one-market telescope with its assumption stated is the more scientific instrument.
Walras proved settlements exist under broad conditions. Nothing guarantees the tidy round-by-round convergence this instrument shows — real adjustment can cycle, overshoot, or take the long way for years, and the Sonnenschein–Mantel–Debreu results say aggregate behavior can be nearly anything. "The system settles at the fixed point" is the model's promise, honored here by linear markets and nowhere else on credit.
The lens licenses a question — which couplings are strong enough to matter? — never a refusal to analyze. All tractable reasoning holds something fixed; the choice is between doing it with the freeze named and priced, or invoking total interconnection to dodge every definite claim. A partial answer plus its stated condition beats both the frozen lie and the fog.