2
In this paper I will describe a field experiment that took place in San Francisco in 2002
that investigated whether get-out-the-vote drives were effective in increased turnout. Then I will
describe two different ways of evaluating the results. In brief, I will look at the correct way, but
the primary focus of this paper involves analyzing the results the “wrong” way, abandoning the
random assignment, treatment-and-control design that provides a powerful device to draw
inferences. Instead, I will analyze the data without experimental controls, looking at turnout rate
by treatment outcome.
•
If I were to assess the efficacy of the turnout drive by looking at the data this way, it
would be a fallacy. Experimental controls are the only way to account for unobserved random
variation that causes an individual to either vote or not vote. Any claim that those who
successfully received a get-out-the-vote phone (GOTV) call voted at higher rates than those who
did not receive a call is incorrect. There is a correlation between whether one is available and
willing to receive a GOTV treatment and whether one is likely to turn out at the polls; this is
called a selection effect. However, while trying to assess the effectiveness of the turnout drive
by looking at observational data would be inaccurate, that does not mean we can learn nothing
•
The author is grateful to the Pacific Gas & Electric Company and Solem & Associates for their cooperation and
support in performing this field experiment; I especially would like to single out Don Solem, Jon Kaufman, and
Anne Jeffrey for special thanks. Also providing data were Voter Connect, Inc., Voter Contact, Inc., and the San
Francisco Board of Elections.
The material in this paper was prepared as part of a dissertation; I am grateful to my committee for their help and
guidance: Henry Brady, Bruce Cain, Susan Rasky, Merrill Shanks, and Raymond Wolfinger. I would also like to
thank the Cal-Berkeley Graduate Workshop in American
Politics for their advice and support: Justin
Buchler,
Margaret Carne, Casey Dominguez, Matthew Jarvis, Megan Mullin, Keith Smith, and Amy Steigerwalt. I am
indebted to
Stephen Ansolabehere, Dorie Apollonio, Louis Ayala, and Melissa Michelson for very valuable
feedback and suggestions. Any errors or deficiencies in this paper are solely the responsibility of the author. This
research was approved by the University of California at Berkeley Committee for the Protection of Human Subjects
on October 4, 2002; thanks to Beth Mistretta of the CPHS for her gracious support and assistance. Papers including
much of the material in this chapter were presented at the 2003 Western and Midwest Political Science Association
meetings.
Convention