About Retro Poll sampling methods
Up to the present our polls are all done by telephone interviews taken on randomly generated samples of the general population residing in the 48 contiguous United States (we attempted one poll including Hawaii and Alaska both of which have small populations, but the effort was not productive). Most of the sample is drawn by random digit dialing (Market Systems Group) that assures appropriate sampling proportions by state, SMSA, and region. We also add to that sample a smaller proportion of randomly drawn phone numbers from the listed national directory of telephone numbers and this includes over-sampling of Hispanic surnames and in census tracts of high African American concentration. These latter adjustments are required to try and achieve a final poll sample that closely approximates the U.S. population since people of color have historically tended to be less willing to participate in public opinion sampling.
In general our poll interviews take between 9 and 15 minutes. Once an individual agrees to participate, completion rates have been high, consistently over 90% in all polls. Our polls are translated into Spanish and bilingual pollsters perform the Spanish interviews. Because of resource limitations we are so far unable to poll in languages other than English and Spanish. Our calling by volunteers is all done by hand without a CATI lab. In our April-May 2004 Poll 383 surveys were completed by a commercial lab using CATI technology and the rest by our volunteers. Volunteer pollsters are asked to make at least 3 return call attempts when contact is not achieved.
| POLLS | RESULTS OF POLLS
| CORRECT ANSWERS AND CITATIONS RELATED TO EACH POLL
MISSION STATEMENT | HOW YOU CAN HELP | ADVISORY BOARD | PRESS RELEASES | NEWS |
SUBSCRIPTIONS | FREQUENTLY ASKED QUESTIONS | OTHER | CONTACT