Correction to Gerber and Green (2000), Replication of Disputed Findings, and Reply to Imai (2005)

Abstract

This essay corrects the results reported in Gerber and Green 2000 and replies to Imai (2005). When data-processing errors are repaired, the original substantive findings from the New Haven experiment remain unchanged. As previously reported, brief phone calls do not increase voter turnout. The “correction” that Imai (2005) offers, which purports to show that these phone calls produce large, significant, and robust increases in voter turnout, is shown to contain statistical, computational, and reporting errors that invalidate its conclusions about the effectiveness and cost-effectiveness of phone calls and mail. A replication of the New Haven experiment reproduces both the findings reported in Gerber and Green 2000 and the biases of Imai's statistical analysis.