Statistical analysis of the Fall 2003 "State of the State" survey

Mid-decade redistricting was one of the major political issues facing the state of Texas in Fall 2003. In short, redistricting refers to the redrawing of state electoral maps. It is often initiated by the controlling political party to gain additional power in the House of Representatives. There were many mixed views about the mid-decade redistricting in Fall 2003. The reason for this was because the Republican Party of Texas proposed redistricting after maps had recently been redrawn during the 2000 Census. The need to assess Texans' views of the redistricting plan was one of the main reasons for conducting a survey.

In Fall 2003, the students of the Political Analysis class, POLS 3310, at Texas Tech University, under the guidance of Professor Cherie Maestas, carried out the "State of the State" survey. In addition to addressing views about redistricting, this survey also collected information on how Texans felt about other political issues on the state, such as educational reform, recall elections, and the rights of undocumented workers.

This manuscript illustrates the procedures used in designing, pretesting and performing the "State of the State" survey. Moreover, this project involves the statistical analysis of the survey results. The data were collected by means of telephone interviews. Here, telephone numbers throughout the state of Texas were randomly selected using computer-assisted telephone interview (CATI) software from the Earl Survey Research Laboratory at Texas Tech University. This procedure is often referred to as a random digit dialing (RDD). Interviews were conducted between October 22 and November 2, amassing a total of 504 responses used for preliminary statistical computations, the results of which were distributed to the media as a press release on November 13, 2003.

Poststratification methods were used to yield statistical analyses of the survey results that were more representative of the entire population and to adjust for nonresponse. It was of particular interest to make adjustment for the under-representation of Hispanics in the state of Texas. Bootstrapping techniques and linearization methods were used to estimate standard errors that were needed for chi-square tests of independence. Finally, logistic regression modeling was used to study the effect of the demographic variables: race, gender, income, age, and marital status, on educational attainment, defined as the completion of a baccalaureate degree program.