Scoreboard — schools of thought vs the data

Support rate = of a school's predictions that have actually been tested, the fraction where the data agreed with the school. A school with 1-of-1 tested prediction supported is not more vindicated than one with 3-of-5 — the Tested column matters as much as Rate. This scoreboard is the framework updating on evidence in public.

A central pattern to watch is whether a school predicts positive-sum coordination or mostly zero-sum allocation. Many Marxian or heavily interventionist claims can look defensible on distributional intent, but the scoreboard asks the harder empirical question: did the policy expand the surplus being shared, simply move it around, or reduce it through weaker investment, productivity, supply, or fiscal capacity?

Raw net is the simple forecast scoreboard: a clear win counts +1, a partial win +0.5, a partial loss -0.5, and a clear loss -1. Forecast Q-net starts from that same score, then asks how strong the test design is. A study that can credibly test a causal claim should move the scoreboard more than a broad pattern match like "this country got richer and healthier."

Attribution net is stricter. It keeps the evidence quality discount, then also applies claim-link confidence, first-pass screening cautions, and a causal-attribution discount when a school gets credit merely because it predicted a failure. A failed policy is not treated as proof that every opponent's mechanism was right. It now also applies the strict second-order gate: policy-experiment evidence does not move A-net until required mechanism, incidence, supply, quality, leakage, fiscal, and welfare layers are measured or explicitly cleared.

Schools are ranked by Attribution net, now the strict second-order-gated score. The forecast Q-net and original raw score stay visible for transparency, while tiny margins are muted as too close to call.