Participant selection bias does not inflate contact-prejudice effect sizes.

Evidence

Note that the no-choice samples provide a significantly larger mean effect size (mean r .280) than do those samples in which participants had some choice (mean r .190), QB(1) 20.58, p .0001, or full choice (mean r –.218), QB(1) 8.98, p .01. The fact that the no-choice studies were, in general, of higher quality magnifies this difference between these three types of studies. The basic correlation between choice and effect size, r .086, p .05, becomes nonsignificant, r .005, p .89, when we partial out four indicators of research quality. But the key finding is that the studies that allow the participant selection bias to operate do not typically yield the larger effect sizes that would be predicted by participant selection bias. — Thomas F. Pettigrew, Linda R. Tropp, A Meta-Analytic Test of Intergroup…, 103

Categorization

• DEEPTHINKING: A counterintuitive methodological result: studies where participants freely chose their intergroup contact show weaker effects than no-choice studies — undermining the standard ‘selection bias inflates contact effects’ criticism of the contact literature.

Related

• Contact’s prejudice-reducing effect is causal, not a selection or publication artifact. (similarity 0.86)
• Randomized contact experiments yield near-large prejudice reduction effects. (similarity 0.84)
• Contact studies with cleaner control groups show larger effects. (similarity 0.83)
• More rigorous contact studies show larger prejudice-reduction effects. (similarity 0.83)
• The contact-prejudice effect varies across samples by design and context (similarity 0.82)
• The intergroup contact-prejudice link survives file-drawer checks, indicating limited publication bias. (similarity 0.81)
• The contact-prejudice effect may be partly an artifact of publication bias. (similarity 0.81)
• Stronger contact effects in rigorous studies would confirm the contact-prejudice link. (similarity 0.81)
• Subjective literature reviews suffer selection biases that meta-analysis avoids. (similarity 0.79)
• Meta-analysis pools all studies to estimate effect sizes and moderators (similarity 0.79)