Who actually pays the wages

Firms sell to each other constantly. Steel goes to a parts maker, parts go to BMW, Salesforce sells software to all of them. But none of those transactions finishes anything. The steel is still in the car. The parts are still in the car. The Salesforce subscription is buried in BMW’s cost base and therefore in the car’s price too.

Economics splits every purchase into two kinds. Goods a firm buys and uses up making its own output are Vorleistungen — intermediate inputs. Goods bought by someone who does not resell them are Endnachfrage — final demand. Only final demand terminates a chain.

A firm’s Wertschöpfung — value added — is what it sold for minus what it bought from other firms. That difference, and only that difference, is the pool it can pay wages, profit and production taxes out of. BMW sold a €50,000 car but bought €19,000 of inputs, so €31,000 stays inside BMW.

Now add the four pools together.

The sum of value added equals the final price exactly. Not approximately — exactly, and by construction. Every euro one firm counts as revenue, the next firm counts as a cost, so all the internal transactions cancel. What survives the cancellation is the one payment nobody passed on: the car buyer’s €50,000.

If a statistician added up every transaction in that chain — 5,000 + 15,000 + 4,000 + 50,000 — they would get €74,000 for a country where exactly one car was produced. The steel would be counted four times over. That mistake has a name, Doppelzählung, double counting, and avoiding it is the entire reason value added exists as a concept.

So gross domestic product is defined as the sum of every producer’s Bruttowertschöpfung, plus taxes on products and minus subsidies on them. Same €50,000. The German statistical office publishes this under the heading Entstehungsrechnung — the production approach.

Here is the part the chain picture leaves out. The car buyer is not standing outside the economy holding money that came from nowhere. He earned it as a wage, in some other firm, in some other chain — probably a chain that also ends at a person who earned their money at a firm. Households supply labour to firms; firms pay wages and profits to households; households spend it back on firms’ output.

That closure is the reason the same €50,000 can be measured three completely different ways and must give the same answer every time — as value added, as final spending, or as the incomes it paid out. More on that in section 10.

The obvious guesses are a person or the government. Those are two of the four. The other two matter a lot for a supplier or an installer, because they are where a great deal of B2B revenue actually terminates.

The investment endpoint is the one that breaks the „everything is intermediate“ model most cleanly. When BMW buys a €400,000 robot from KUKA, that robot is not a Vorleistung — it is not used up in one car. It is a capital good, it counts as final demand in the year it is bought, and only its annual wear (Abschreibung) is charged against production afterwards. Same with a fire alarm system installed in a factory: the customer is a business, but the sale is an endpoint, not a link in a chain.

Wages are not paid out of the sale

BMW pays a worker in March for a car that sells in September. The causality is not „customer money arrives, then wages go out.“ Firms bridge that gap with retained earnings and bank credit, and banks create the deposits they lend. So the €50,000 does not fund the wages in real time — it validates them afterwards. If the car never sells, the wages were still paid and the loss lands on the firm.

B2B firms sell productivity, not a slice of someone’s spending

Salesforce takes €4,000 out of the bar. That looks like it comes at BMW’s expense, but BMW pays it voluntarily because the software makes BMW’s own €31,000 slice larger than it would otherwise be — better lead handling, fewer people needed per sale, faster cycle. A B2B firm survives exactly as long as the value it adds at its customer exceeds what it charges. That is a harder test than selling to consumers, not an easier one.

There is a hard ceiling in any given period, and it is measurable. German GDP in 2025 was roughly €4,470 billion in current prices. That is the whole bar for the whole country. Every firm’s revenue, every wage, every profit fits inside it, and nothing arrives from outside except what foreigners buy.

The interesting number is not the level, though. It is what happened to it.

Almost all of the nominal growth was inflation. In real terms the German pool has been close to flat for three years, after two years of outright contraction. So the instinct that firms are fighting over a fixed quantity describes Germany right now unusually well — but it describes a condition, not a law.

Four ways the pool actually grows

Credit and government deficits do something different — they do not grow the pool so much as move it in time, pulling future spending into the present. Useful in a downturn, not a source of permanent growth.

The sharper problem is not that firms compete for share. It is that the thing every firm treats as its main cost is the same thing that funds everyone’s revenue.

To any single firm the wage bill is a cost to minimise. To all firms together, wages are the demand. Michał Kalecki put it in one line: workers spend what they earn, capitalists earn what they spend.

How much this actually binds is disputed

This is one of the oldest arguments in economics and it is not settled. The demand-side view says suppressed wages mean suppressed sales, so the economy runs permanently below capacity and the wage share matters directly for growth. The supply-side counter is that wage restraint lowers costs and prices, raises competitiveness, and wins export share — which more than compensates. Both descriptions fit real periods of German history.

The German model since the early 2000s has been the second one: wage moderation, strong exports, a large current account surplus. Taken to its conclusion, that means growing by drawing on other countries’ Endnachfrage rather than expanding your own Binnennachfrage. It worked for two decades. The 2025 figures show its fragility — exports fell against US tariffs, a stronger euro and Chinese competition, and the only growth came from household and government consumption at home.

The same fight happens inside the value-added box

Section 02 showed BMW keeping €31,000. That figure is then split three ways: wages, operating surplus, and production taxes. Those three are in direct competition with each other, and the split moves. In 2025 German employee compensation rose 5.1% while business and property income fell 3.9%. Same pool, different division. The slice competition you are describing between firms also runs vertically, inside every single firm.

Almost none of it is your binding constraint. A one-person or ten-person firm’s addressable market is a rounding error against €4.47 trillion. You will never run out of pool; you will run out of reach, capacity or reputation long before. Aggregate demand only becomes your problem when it moves sharply, and even then it arrives as your customers’ behaviour, not as a number in a press release.

Two things do transfer, though.

Mandated demand behaves differently

A good part of fire protection and security work exists because building regulation requires it, not because someone chose it over a holiday. That demand does not sit in a household’s discretionary budget competing against a new kitchen, and it does not disappear in a weak year. Regulation-driven work is markedly less cyclical than the pool it sits inside — a structural advantage worth knowing you have.

Watch value added per head, not revenue

Revenue tells you how much passes through. Wertschöpfung tells you how much stays. A firm reselling €2m of hardware with €1.9m of purchase cost has €100k to pay everyone from. A firm billing €300k of engineering with €30k of costs has €270k. The second is nearly three times the economic actor despite being a sixth of the size on paper — and the difference is exactly what determines what it can pay in wages.