About the poll

Poll methodology

The Big Ten Battleground Poll is sponsored by a partnership of eight of the Big Ten Universities: the University of Illinois, the University of Iowa, Michigan State University, the University of Michigan, Northwestern University, The Ohio State University, Pennsylvania State University and the University of Wisconsin. The data collection was conducted by LHK Partners Inc., Newtown Square, Pa.

The question wording and order are presented with the results on the Results and Data pages of this Web site. (The questions about favorability for Obama, McCain, Biden, Palin, and Bush were rotated — that is, they were asked in a random order to each individual respondent.)

The poll’s advisory board approved all final design and fielding issues.

The first Big Ten Battleground Poll was conducted Sept. 14-17, 2008 among 5,794 registered voters and those likely to register to vote before the 2008 Presidential election. The second poll was conducted Oct. 19-22, 2008 among 5,719 registered voters and those likely to register to vote. Both surveys sampled at least 560 respondents from each of the Big Ten states (Illinois, Indiana, Iowa, Michigan, Minnesota, Ohio, Pennsylvania and Wisconsin). The survey also sampled from the remaining 40 states (not including Alaska and Hawaii): 831 respondents in September and 814 respondents in October.

LHK Partners Inc. contacted households by telephone, selecting the person age 18 or older currently at home, with the most recent birthday, to participate in the polls. This was intended to assure a random selection of adults within each household.

A probability sample of all telephone households in the United States was first selected. A series of sample control procedures were implemented to optimize the overall representativeness of the final sample of participants. These procedures were monitored daily. The overall samples were first assembled into replicates, i.e., groups of phone numbers that are assembled into “mini” samples that are completely reflective of the overall sample design. Each replicate shares all of the sample characteristics of the overall sample, except that it is smaller. Every attempt is made to assure that once a replicate is assigned to the survey, every phone number in that replicate is given an equal chance to be contacted.

After a phone number was originally dialed and included in the sample, up to seven additional attempts to reach qualified respondents in that household were made. Computerized sample management programs automatically scheduled these subsequent attempts at alternate times to assure that each household was called several times both early and late each evening of the study (and during the afternoon on Sunday). To account for the unavoidable circumstance that all phone numbers within a released replicate could not be assured to receive exactly the same number of calls, all numbers within a replicate were randomized. Accordingly, we assured that each phone number within a replicate had a similar chance of ultimately being contacted.

The samples are of registered voters and those likely to register to vote before the 2008 Presidential election. Samples were drawn for each of the eight Big Ten states of Iowa, Illinois, Indiana, Michigan, Minnesota, Ohio, Pennsylvania and Wisconsin. In addition, a national sample of additional respondents (831 in September and 814 in October) were selected from the remaining 40 contiguous states, excluding Alaska and Hawaii.

The samples were selected by random digit dialing (RDD) of landline phones. Cell-only households were not included in the samples. Within households the person with the most recent birthday was selected for participation in the study with a quota for sex of respondent in order to balance male and female respondents at 49%-51%.

Sample sizes and margin of error for each state and for the national sample:

September survey:

• Illinois: N=628. MOE= +/- 4.0
• Indiana: N=612. MOE= +/- 4.0
• Iowa: N=643. MOE= +/- 3.9
• Michigan: N=628. MOE= +/- 4.0
• Minnesota: N=610. MOE= +/- 4.0
• Ohio: N=619. MOE= +/- 4.0
• Pennsylvania: N=608. MOE= +/- 4.1
• Wisconsin: N=616. MOE= +/- 4.0
• United States: Weighted N=1114. MOE= +/- 3.0

October survey:

• Illinois: N=572. MOE= +/- 4.2
• Indiana: N=586. MOE= +/- 4.2
• Iowa: N=586. MOE= +/- 4.2
• Michigan: N=562. MOE= +/- 4.2
• Minnesota: N=583. MOE= +/- 4.2
• Ohio: N=564. MOE= +/- 4.2
• Pennsylvania: N=566. MOE= +/- 4.2
• Wisconsin: N=584. MOE= +/- 4.2
• United States: Weighted N=1014. MOE= +/- 3.1

The samples were weighted to the Current Population Survey results for registered voters based on age, education and race using cell weights. The national samples were similarly weighted by age, education and race with the Big Ten state samples weighted by their share of the national voter turnout in the 2004 presidential election.

Results presented under Results and Data on this Web site are for the full sample.

The interviews were collected by telephone with live interviewers.