Washington Post Polls: Detailed Methodology

Tuesday, March 31, 2009 8:55 AM

Since 1981, the large majority of Post polls have been conducted by the research firm TNS of Horsham, Pa., formerly known as Chilton Research Services.

TNS is responsible for:

* purchasing and screening the sample,

* programming the survey questionnaire,

* fielding the survey and overseeing interviewer quality, and

* processing and weighting the data.

The Post works closely with TNS during all these stages.

Who is eligible

For the standard national poll, respondents must be aged 18 or over, live in the continental United States, and have a home phone. Currently, all Post interviews are conducted in English, except on special projects.

Creating a sample

As of June 2003, telephone samples come from Survey Sampling, Inc. (SSI) Samples are drawn using list-assisted Random Digit Dial (RDD) methodology. Telephone exchanges are stratified prior to sampling based on: Census state and county; area code; prefix; and working block. Blocks of telephone numbers with more than 1 residential directory listing are defined as working blocks; others are removed. A systematic random sample of working blocks is then selected from the stratified database, and a random two-digit suffix selected for each working block. The final sample is matched against a list of known business phone numbers which are then flagged. It is also screened for non-working numbers using a non-ringing auto-dialer.

The questionnaire

The questionnaire is written by Washington Post staff, including polling unit staff as well as beat reporters and editors. Many questions are pulled from databases of previously asked items.

Programming the questionnaire

The questionnaire is programmed into the computer-assisted telephone interviewing system (CATI) at TNS. The Post's project director and newsroom staff oversee this process.

Fieldwork and Interviewers

Calls are made from TNS field houses in Indiana, Pa., and Eugene, Ore. Newly hired interviewers all go through a two-day orientation/training program on interviewing skills, including practice sessions with live respondents. A senior staff person then intensely monitors their performance for two weeks, providing feedback and additional training. During fieldwork, the field houses maintain a ratio of one supervisor and one quality assurance monitor for every 10 interviewers.

Days in the field

A standard national survey is in the field three to five nights. In the case of extremely fast-breaking news, such as the outbreak of war, The Post may do an overnight survey in order to inform readers about initial reactions to major events. Results of these overnights should be interpreted with extra caution, since the shortened field period constrains normal field procedures.

Dialing the sample

On the average poll, each number in the sample is dialed at least four times. The average is at least seven calls per number. The sample is released in replicates by region. There is no systematic attempt to convert explicit refusals, though interviewers do redial initial hang-ups. Interviewers do not leave voice mail messages, but are permitted to schedule a callback time.

In-house selection method

Respondents are selected using the "at home, last birthday" method, stratified by sex. This means that in 75 percent of the cases, interviewers will ask to speak to the male currently at home who last had a birthday, and in 25 percent, they will ask to speak to the parallel female. This design is intended to adjust for the fact that women are more likely to pick up the phone in most households.

Response rates

See extended discussion on response rates here.

Weighting the Data

The data is weighted to census demographics including age, race, sex and education using a 48-cell design. Target national numbers come from the most recent Current Population Survey.

View all comments that have been posted about this article.

© 2009 Washingtonpost.Newsweek Interactive