Abstract
A model is developed to forecast monthly county level voter registration totals by party for Florida, Iowa, Maryland, Oregon and Wyoming. A prediction model of change in voter registration can help election administrators prepare for periods of high workload. Another application is to aid the adjustment of voter registration figures with an eye towards removing deadwood registrants. The aim of doing so is to boost voter registration data’s efficacy in predicting phenomena such as postcensal population growth, election outcomes and voter turnout. Descriptive statistics on the concentration of drops and negative net changes in total registrants are presented to discuss how much information such data can yield to help estimate deadwood in conjunction with prediction models utilizing demographic and other factors