Skip to Main Content

Basic Search

Skip to Search Results
 
 
 

Left Column

Filters

Right Column

Search Results

Search Results

(Total results 1)

Mini-Tools

 
 

Search Report

  • 1. RADEMACHER, ERIC THE PATH TO ACCURATE PRE-ELECTION FORECASTS: AN ANALYSIS OF THE IMPACT OF DATA ADJUSTMENT TECHNIQUES ON PRE-ELECTION PROJECTION ESTIMATES

    PhD, University of Cincinnati, 2002, Arts and Sciences : Political Science

    Much has been written about the success and failure of survey research organizations attempting to project election outcomes in advance using sample survey data. However, less is known about the relationship between specific data adjustment methods used to generate those projections and the level of accuracy those projections achieve. Organizations often adjust raw survey results using data adjustment models, developed a priori, incorporating one, two or all of the following methods: 1) demographic weighting, 2a) likely voter screening and related 2b) adjustments for turnout, and 3) the allocation of undecided voter preferences. However, few investigations have considered the relationship between election projection accuracy and these various adjustments in depth. This research contributes to the literature on pre-election polling and projections through a systematic analysis of the contribution that each of these three types of data adjustment (weighting, turnout adjustment and undecided voter allocation) make to the accuracy of election projections. This research finds that while some data adjustment methods improve election projection accuracy, not all methods improve projections over the accuracy level achieved using unweighted data. For example, while this research presents evidence that specific turnout adjustment methods often lead to improved election projection accuracy, extensive tests also found specific demographic and geographic weighting had a very limited impact on increased accuracy in many situations and actually reduced projection accuracy in several cases. Variation in accuracy across different types of election contests was found when single data adjustment models were applied. As a result, this analysis recommends experimentation with using separate data adjustment methodologies when projecting election results in four types of election contests: "Top of the Ticket" incumbent races; "Top of the Ticket" open seat races; "Bottom of the Ticket" inc (open full item for complete abstract)

    Committee: Dr. Alfred J. Tuchfarber (Advisor) Subjects: Political Science, General