macrocephalopod

Impressive! No replication crisis after all... maybe a double log-scale would ease the work of ggrepel(). If this creates confusion it’s always possible to add horizontal & vertical thresholds, eg at t=3.

this is clearly great work by @achenfinance but it doesn’t really impact on the replication crisis because this is afaik replicating studies using the same data, not generating new data and trying to replicate (as you would in eg a psychology replication)

But I agree, the validity of asset pricing anomalies outside their original papers is far from being obvious. Several papers have documented that a large majority of them vanish when they are tested on new data or when minor changes are applied... *this* replication crisis holds.