Surveys – Survey Methods
In comparison with surveys carried out within a single nation, cross-nation surveys involve an extra layer of difficulty and complexity in terms of survey management, research design, and database modeling for the purpose of data preservation and easy analysis. To facilitate the work of the Asian Barometer Survey, the Survey Methodology and Database subproject was launched. It serves the important role of overseeing and coordinating survey research designs, database modeling, and data release.
As a regional partner of the Global Barometer Surveys, the Asian Barometer Survey requires all country teams to comply with the research protocols developed and established by the Global Barometer network. These protocols have been tested and have been proven as practical methods for conducting cross-nation comparative surveys on citizens' attitudes and public opinion.
• National probability samples that give every citizen in each country an equal chance of being selected to participate in the survey. Whether using census household lists or a multistage area approach, the method for selecting sampling units is always randomized. The samples may be stratified – or weighted – to ensure adequate and correct coverage of rural areas and minority populations. As such, the samples selected for the Asian Barometer surveys represent the totality of the adult, voting-age population in each country surveyed.
• A standard questionnaire instrument containing a core module of identical or functionally equivalent questions. Wherever possible, theoretical concepts are measured with multiple items in order to enable testing for construct validity. The wording of items is determined by balancing various criteria, including: the research themes emphasized in the survey, the comprehensibility of the item to lay respondents, and the proven effectiveness of the item when tested in previous surveys.
• Intensive training of fieldworkers. The success of Asian Barometer surveys depends upon the effectiveness of field operations, especially the training of interviewers, supervisors and fieldwork managers. We recruit interviewers from among university graduates, senior social science undergraduates, or professional survey interviewers. All managers and supervisors have extensive field experience. Field teams pass through intensive, week-long training programs to familiarize them with the Asian Barometer research instrument, our sampling methods, and the cultural and ethical context of the interview. Guidelines are codified in instruction manuals that spell out procedures for the selection and replacement of samples, the validation of interview records, and interview etiquette.
• Face-to-face interviews in respondents’ homes or workplaces in the language of the respondent’s choice. In multilingual countries, considerable attention is given to the vexing challenge of questionnaire translation. Local language translations are prepared with the goal of accommodating every language group whose members constitute at least 5 percent of the population. To check for accuracy, the local language versions are screened through blind back-translation by a different translator and any discrepancies are corrected. Interviewers are required to record contextual information on the situations encountered during the interview.
• Maintenance of quality control by means strict protocols for fieldwork supervision. To ensure data quality, we require that all interview teams travel together under the direction of a field supervisor. Interviewers are debriefed each evening and instructed to return to the sampled household to finish any incomplete returns. Supervisors undertake random back-checks with respondents to ensure that sampling and interviews were conducted correctly.
• Quality checks are enforced at every stage of data conversion to ensure that information from paper returns is edited, coded, and entered correctly for purposes of computer analysis. Machine-readable data are generated by trained data entry operators and a minimum of 20 percent of the data is entered twice by independent teams for purposes of cross-checking. Data cleaning involves checks for illegal and logically inconsistent values.
A model Asian Barometer Survey has a sample size of 1200 respondents, which allows a minimum confidence interval of plus or minus 3 percent at 95 percent probability.