Pew Research Center’s cross-national studies are fielded in a range of polling environments, from countries where telephone surveys are regularly administered to nations where face-to-face surveys are the most reliable means of achieving accurate, nationally representative samples. In all cases, the Center adopts the survey mode that follows local best practices and meets Pew Research standards for methodological rigor and data quality. Currently, about two-thirds of international polls conducted by Pew Research Center are administered face-to-face, with telephone surveys generally limited to North America, Western Europe and select countries in the Asia-Pacific region, such as Australia, Japan and South Korea. All cross-national studies are based on interviewer-administered survey instruments, except when surveying in the U.S. using the online American Trends Panel.
Pew Research Center constantly seeks to improve and refine its methods of international data collection through more extensive interviewer training and questionnaire testing as well as via adaptive data-assurance and data-control protocols, including the integration of various forms of paradata. The Center also is intensifying investigations of alternative modes of high-quality data collection as technological infrastructure and local surveying capacity within countries continue to evolve over time. At this time, however, Pew Research has decided not to employ opt-in, nonprobability online panels as a means of measuring international public opinion, due to concerns about the representativeness of such surveys and the lack of established metrics for gauging the accuracy of such polls across countries.
Face-to-face surveys
Pew Research Center relies on face-to-face surveys as its primary mode of data collection in countries extending from Latin America to Africa to Asia and Eastern Europe. Probability-based, national samples are based on multistage, cluster designs. What this means is that we first randomly select clusters of individuals – beginning with relatively large geographic units, akin to states or counties in the U.S. Once these primary clusters are selected, we randomly select smaller geographic units, until we work our way down to city blocks or rural villages. At this stage, interviewers either visit addresses selected randomly from a registry (if available), or they follow a “random walk” in which they visit residences following a standard protocol, selecting addresses at a set interval (e.g., every fifth residence) along the route. Within each selected household, interviewers randomly select one adult to participate so all members have an equal chance to be interviewed – not only those present at the time. If the selected respondent is not immediately available, interviewers will return on multiple occasions. Usually up to three attempts are made to complete the interview with the selected respondent.
Telephone surveys
Pew Research Center regularly conducts telephone surveys in countries where such polls are an established feature of the local opinion research and the telephone infrastructure (landline and/or mobile) allows for nationally representative, probability-based designs to be implemented. Pew Research Center designs its telephone surveys as either overlapping dual-frame (landline and mobile) or single frame (mobile only) depending on local telephone penetration rates as well as the availability of accurate sampling frames that facilitate representative and efficient dialing. The Center actively monitors shifts in the balance between landline and mobile rates across countries, especially as the proportion of people who rely on mobile phones increases over time.
Pew Research Center relies on random digit dialing (RDD) for its international telephone surveys. Adults within landline households are selected using a random method. Interviews on mobile phones are conducted with the person who answers the phone (if age 18 or older). For both landline and mobile samples, up to seven contact attempts are usually made to complete the interview with the selected respondent.
