Abstract
Conventional wisdom holds that negative campaign messages decrease voter turnout and
are more persuasive than positive messages. Academic evidence in favor of these two
contentions has been mixed. However, prior studies may be limited by research design.
Observational analyses capture real-world campaigns and behaviors, but cannot
definitively rule out endogeneity as a source of potential bias. Laboratory experiments
possess internal validity, but it is unclear the extent to which the results hold in broader
real-world settings. Attempting to combine the strengths of both methodologies, we
worked with two actual campaigns to conduct randomized experiments in the field. By
randomly assigning subjects to be exposed to positive or negative campaign messages,
we can estimate the effect of message tone on turnout and vote choice. In the end, we
detect no difference between negative and positive messages with regards to turnout or
vote preference.
1
Convention