3
Although the potential for partisan outreach has been evidenced in a number of
recent turnout studies (e.g., Rosenstone and Hansen 1993; Shaw, de la Garza, and Lee
2000), other research has pointed to the inefficacy of widely employed campaign
mobilization techniques (Gerber and Green 2000a; 2000b; 2001; Green, Gerber and
Nickerson 2002; Green and Smith 2003). One reason why these techniques fail is that
overly simplistic models of turnout drive their application. Campaign strategists presume
the political behaviors of key subgroups of the population are invariant across contexts.
As a consequence, in many locations, the wrong people have been targeted. Indeed, the
common campaign practice of carpet-bombing entire neighborhoods, cities and even
states with direct mail, phone bank calls, and house-to-house canvassing is an egregious
misuse of scarce resources. Moreover, while some contacting techniques may be superior
to others, the way people are indiscriminately targeted for contact is also a fundamental
problem, regardless of the technique used.
Our goal here is to assess more accurately the factors affecting turnout, taking
particular care to determine how features of neighborhood context interact with more
conventional predictors of voting. This allows us to pinpoint those who could be most
effectively contacted and is likely to provide substantive insight that can enhance the
effectiveness of traditional party mobilization tools.
Our approach emphasizes a unique set of data: official lists of registered voters
for 14 major counties in seven swing states (including Florida) from the 2000 presidential
election. The individual-level data are augmented with tract and precinct level contextual
data drawn from the United States Census Bureau. These data facilitate analyses that are
both theoretically and practically appealing. As a theoretical matter, we use voter lists in
Convention