In their landmark study of a field
experiment, Gerber and Green (2000) found that
get-out-the-votecallsreduceturnout by five percentage points. In
this article, I introduce statistical methods that canuncover
discrepancies between experimental design and actual
implementation. The application ofthis methodology shows that Gerber
and Green's negative finding is caused by inadvertent deviationsfrom
their stated experimental protocol. The initial discovery led to
revisions of the original data by theauthors and retraction of the
numerical results in their article. Analysis of their revised data,
however, reveals new systematic patterns of implementation
errors. Indeed, treatment assignments of the reviseddata appear to
be even less randomized than before their corrections. To adjust for
these problems, I employ a more appropriate statistical method and
demonstrate that telephone canvassingincreasesturnout by five
percentage points. This article demonstrates how statistical methods
can find and correctcomplications of field experiments. (Note: This
paper was submitted to APSR in August 2002, and was accepted for
publication in August 2003. However, its actual publication has been
delayed because the original authors have not finished writing their
response.)
|