Conclusion
Is the density-fertility relation a statistical artifact? The answer, I conclude, is no. Like all empirical findings, the observed coefficients might misstate the true effects, so the density coefficient could be “wrong.” Moreover, the density coefficient does not identify the mechanism(s)—inheritance, child labor value, or migration—through which density affects fertility. But the density coefficient is not artifactual. Jensen’s claims are specious. His principal claim, that the use of ratio variables with common components creates bias, is based on a statistical myth. Two other claims, that pooling the village data creates a spurious correlation between density and the CBR and that the density coefficient has been inflated by simultaneity bias are shown to be false. His final claim, that omitted-variable bias must inflate the density coefficient, is also wrong. I see no reason, then, to modify the conclusions of the Punjab study.
