Article summary
Publication bias is common in social sciences because journals favor statistically significant results. However, publication bias depends on a variable's role—key or control—in regression models. This study shows that when democracy is a key variable (rather than a statistical control) it faces stronger publication bias.