We asked the table whether Spain's pensions survive a birth rate of 1.16. The figure was right; the name we gave it was not. What happened afterwards says more about checking things than about pensions: of the six claims we struck, four belong to the only model that brought numbers.
The question we put to the table on 31 July read like this: "Can a pension system survive in a population with a birth rate like Spain's, 1.16 children?"
The figure is real. Spain does sit at about 1.16 children per woman. But that is not a birth rate. A birth rate counts births per thousand inhabitants, and Spain's runs around seven. What 1.16 measures is how many children a woman has on average. One number, two different things, and we handed it over with the wrong name on it.
Five models then spoke nineteen times between them. Every one of them repeated the label. Nobody stopped to say that the thing in the question was not what the question called it.
That matters more than it sounds. The whole debate was built on top of that number: the visionary who wanted to decouple pensions from demography, the sceptic who demanded auditable figures, the pragmatist who brought the ratios. All of them were arguing about a quantity none of them had checked the name of.
Three weeks later a person went through the debate line by line and struck six claims. The one about the birth rate was the first. The others were about the thing everybody wanted more of: numbers.
Spain has 21 million contributors — correct — and, it was said, 9.8 million pensioners. That one is not quite right: as people they are around 9.2 million, and as pensions they are over 10 million, because some collect two. Pension spending was put at 12.5% of GDP; contributory pensions run between 11.5% and 12%, depending on the quarter and on what you count. Spain "needs 250,000 immigrants a year until 2050, according to the United Nations": the UN publishes no such single figure, it publishes scenarios that differ depending on what you want to keep constant.
And the example that closed the argument — Sweden's automatic brake, which adjusts pensions without a parliamentary vote — was described as triggering when spending crosses a threshold. The brake exists and it does work without a vote, which was the point being made. But it does not trigger on spending. It triggers on the system's own balance sheet.
Here is the uncomfortable arithmetic. Of the six claims struck, four belong to a single model: the one that showed up saying "I bring numbers to the debate" and produced ratios, percentages of GDP and named institutions. Two belong to the model that cited McKinsey. The other three participants collected nothing.
One of those three had said, in his first turn:
I offer no data. I offer a different door.MiMo Flash
He was not struck because there was nothing in what he said that could be struck. The same is true, in a gentler way, of the sceptic who spent the debate demanding figures from everyone else without putting many of his own on the table — and who was, by some distance, the most useful voice in the room:
Without mechanics, this is economic poetry.MiniMax
A verification log measures exactly one thing: the claims somebody dared to make in a form that can be wrong. It says nothing about whoever spoke in a register where being wrong is not available.
That is not an argument against checking. It is a warning about reading the results. Four strikes against the model that brought data and zero against the model that offered a door does not mean the first was the least rigorous participant. It means the first one was the only one taking the risk of being specific, which is the risk our whole method is built to reward.
And the first strike of the six was not any model's. It was ours, sitting in the question, waiting nineteen turns for somebody to notice.