Citizenship is widely summarized as producing a “naturalization premium” for immigrants, but naturalization is highly selected on language, residence, employment, and motivation. We provide a quantitative meta-analysis of citizenship effects on immigrant earnings, synthesizing 63 preferred estimates from 31 studies. Identification strategy explains more between-estimate heterogeneity than citizenship regime, destination country, effect measurement window, outcome data source, earnings-sample basis, or treatment estimand. A multilevel random-effects model yields a pooled earnings premium of 9.6% (95% CI 5.0 to 14.5), broadly consistent with the study-country-collapsed estimate of 11.2% (95% CI 6.1 to 16.5). This average masks a sharp identification-strategy gradient: a randomized encouragement design implies a precisely estimated near-zero effect of 0.4% (95% CI -0.9 to 1.7); the regression-discontinuity study yields 11.3% (95% CI 5.2 to 17.8); reform-based and panel within-person estimates are positive on average but statistically indistinguishable from zero; observational instrumental-variable estimates average 39.9% (95% CI 12.6 to 73.9). Excluding observational instrumental-variable estimates reduces the pooled premium to 5.8% (95% CI 3.6 to 8.1). A within-study diagnostic from the randomized trial shows the same pattern: experimental estimates are near zero, while an observational panel comparison that ignores randomization and uses self-selected citizenship uptake implies a 9.6% gain (95% CI 3.6 to 16.0) four years after the lottery. The published citizenship premium is positive, but its magnitude and causal interpretation are highly sensitive to research design.