Power and bias in industrial relations research

Abstract

We survey 20,439 estimates from 64 distinct research areas to assess power, bias and statistical significance in industrial relations research. The average estimate published in industrial relations research lacks adequate power; average power is 33 per cent, and median power is only 14 per cent, much lower than the conventional 80 per cent standard. Low power means that industrial relations researchers will find it more difficult to detect important associations pertaining to workplace relations. Low power also leads to exaggerated research findings. We find substantial publication bias in industrial relations research, though nearly half of the research areas have little or no bias.