I wonder if his study was simply underpowered, and there may be a real but small effect.

An underpowered study is one that doesn’t get statistically significant results for a real effect because there is not enough data collected to distinguish between a small effect and random chance.