and

2004 Election Panel Study Sampling Design

 

The BYU-UW Election Panel was designed to provide not only a nationally representative sample of the electorate but also to provide substantial numbers of cases in both presidential and U.S. Senate "battleground" states. To accomplish this, the design oversampled both types of battleground states, and added stand alone samples of Ohio and Florida.

The sample defines five strata: non-battleground states, Senate-only battleground, presidential battleground, Ohio and Florida. The Senate-only strata are states with potentially competitive Senate contests but with a non-competitive presidential race. The presidential battleground states may also contain a competitive Senate race, but need not.

In order to use the study you MUST weight the data to take account of the unequal selection probabilities for respondents in the various strata. This weight variable is named "weight" in the dataset.

The first link below describes the final sample design as executed. This is the most important description of the sample for users to read in order to understand the sample design, the weights, and the characteristics of the final sample.

The subsequent "sample development" links provide a full description of the sampling design process, the alternatives considered, the decisions taken, and the evolution of the design over time. These are useful for those more interested in the technical decisions or in the changes that were made over the course of developing the study. Users should ALWAYS consult the FIRST link below for the properties of the final sample, NOT the earlier sample designs which changed as the final design was developed.

  • The final sample design and weighting information can be viewed here. (PDF format)

The memos below describe the development of the sample design. The sample characteristics and weights are different from those in the final design above due to the evolution of the design. Use these only to understand what considerations entered into the design of the BYU/UW Election Panel.

  • A discussion of the initial sample design and tradeoffs can be viewed here. (PDF format)
     

  • The revised sample with Florida replacing the Pennsylvania domain can be viewed here. (PDF format)
     

  • The revised sample design with a smaller Florida sample size can be viewed here. (PDF format)

 

Back to Top

Home | Survey Data | Questionnaire | Press Releases | Sampling Design | Investigators
Copyright © 2004 Wisconsin Advertising Project, All Rights Reserved.