US studies may overestimate effect sizes in softer research