What’s true of the parts isn’t automatically true of the whole
These are two mirror-image fallacies, which is why they make sense to cover together.
Fallacy of Composition – assuming that what’s true of the parts must be true of the whole.
Fallacy of Division – assuming that what’s true of the whole must be true of the parts.
Both get the relationship between scale and property exactly wrong.
Here’s composition in everyday shape.
“Each player on this team is excellent. Therefore this team will be excellent.” Not necessarily. Team chemistry, role overlap, and system fit can make great individual players perform poorly together. Any sports fan who’s watched an all-star team get dismantled by a tighter opponent already knows this. Being made of great parts doesn’t automatically mean the whole is great.
“This brick is light. Therefore a building made of bricks is light.”
Obviously not. Scale changes the property. The brick is carryable. The building is not carryable. Same material, completely different relationship to weight, because there are a lot more of the bricks now.
“Sodium is a metal and chlorine is a poisonous gas. Therefore table salt is a poisonous metal.”
Nope. The compound has properties nothing like its components. Sodium chloride is neither metal nor gas. It’s salt. Chemistry behaves like this constantly. The combination of the parts produces something different from the parts, which is kind of the whole point of chemistry.
Division runs the same mistake in reverse.
“This university is world-class. Therefore every professor here is world-class.” Not necessarily. Institutions have range. Some of the faculty are stars. Some are average. Some are duds. The prestige of the whole doesn’t distribute evenly to every part.
“This country has a high average income. Therefore most people here are wealthy.”
Averages can mask enormous distribution differences. A country with ten billionaires and ninety people living on nothing has a high average income. Ninety out of a hundred people are broke. The average is doing exactly what averages do, which is describe the whole without telling you anything about individuals.
“The economy is doing well. Therefore everyone in the economy is doing well.”
This one has serious real-world consequences when it’s used to dismiss evidence of economic hardship. GDP growth, stock market gains, and low unemployment can all be accurate descriptions of the macroeconomy while millions of specific people are underwater. Aggregate health isn’t individual health. The number for the whole isn’t the number for the person standing in it.
Economic policy debates are full of both fallacies simultaneously.
Composition errors lead to assuming that what works for one business – cutting costs, laying off workers – will work for the whole economy. But if every business simultaneously cuts costs and lays off workers, aggregate demand collapses and everyone does worse. The individually rational move is collectively disastrous. This is known as the paradox of thrift, and it’s been documented for almost a century, and it still gets ignored every recession.
Division errors go the other direction.
Macro-level economic strength gets used to dismiss individual complaints about hardship. “The numbers look great, so your experience must be wrong.” The gains might be concentrated at the top while the median household goes nowhere. Both things are true at the same time. The aggregate number isn’t lying. It just isn’t describing the individual.
The key question is whether the property actually transfers across scale.
Some properties do. Most interesting ones don’t. Physical properties change at scale. Emergent phenomena exist at the level of the whole that have no equivalent at the part level. Statistical measures like averages describe populations, not individuals within them.
When somebody moves from parts to whole, or whole to parts, ask whether the property actually transfers.
If it doesn’t, the argument jumped a gap that wasn’t there.