ABSTRACT
Probability-based sampling methods, such as random-digit dialing, are a staple of modern polling and have been successfully used to gauge public opinion for nearly 80 years. Though historically effective, such traditional methods are often slow and expensive, and with declining response rates, even their accuracy has come into question. At the same time, non-representative polls, such as optin online surveys, have become increasingly fast and cheap. We show that with proper statistical adjustment, non-representative polling can be used to accurately and continuously track public sentiment, offering a new approach to public opinion research.
