Marcy Shieh and Blake Reynolds (University of Wisconsin-Madison)
Abstract: The way images and text are presented to us can have a significant impact on how we are affected by the message contained in an advertisement. Therefore, we ask how does the formatting of campaign mail influence voter turnout? Using campaign mail from the 2018 primary and general elections in Texas, we examine the layout of campaign mailers. To do this, we leverage machine learning techniques to identify specific attributes in each mailer, including the text to image ratio, image brightness, size of logo, style of words, and the physical size of mail piece. We then test whether these attributes of campaign mailers affect the turnout of voters assigned to receive each piece of mail. Finally, we calculate the marginal effect of the attributes by using relative lift over the voters’ a priori turnout probability determined by a habitual turnout model and assign composite variables for each type of mailer aggregated for each voter over all mail pieces received prior to each election.