This is deeply troubling. Researchers are more likely to choose statistical models whose results align with their ideological priors. Seventy-one research teams independently analyzed the same dataset on the effect of immigration on public support for social welfare programs. Teams composed of pro-immigration researchers were more likely to conclude that the effect was positive. Teams composed of anti-immigration researchers were more likely to find a negative effect. Let me repeat: they analyzed literally the same dataset. Full paper in the first comment.