Tim Harford points to a paper on the EMH showing that:
…after an anomaly has been published, it quickly shrinks – although it does not disappear.
The anomalies are most likely to persist when they apply to small, illiquid markets – as one might expect, because there it is harder to profit from the anomaly.
It’s always good to remind ourselves that all anomalies are only fully priced in equilibrium, and we’re probably never in equilibrium. The process of moving towards equilibrium involves market participants seeking out those anomalies and exploiting them. So the continued discovery of new market anomalies isn’t evidence against market efficiency: it’s merely observation of the normal equilibrating process.