Methodology

Sample

A database of 26,000 randomly-selected national telephone numbers was purchased from Telecom. This database was loaded into the Survey Research Unit's CATI system and numbers randomly allocated to interviewers to call. There are several advantages to purchasing a database from Telecom, including the fact that the numbers are current as at the day of collation (no disconnected numbers); as they are randomly drawn from Telecom's national database there is no need to design a sampling frame; business and confidential numbers are automatically excluded.

The survey was conducted with a random sample of New Zealanders, resulting in 5005 suitable interviews. Less than 1% of the included interviews did not answer all relevant questions. The overall response rate (eligible people contacted who agreed to participate) was 84.15%. The sample was generally representative of the population, except that people aged under 18 years were excluded for methodological reasons. Call monitoring was undertaken to ensure that the number of Maori participants reflected the proportion in the population.

Research Instrument

Initial question development was undertaken by the State Services Commission. The writers suggested minor amendments. The questionnaire was then tested in two stages to ensure validity. The first stage was a small number of focus groups to discuss the questionnaire with people from various backgrounds to check for comprehension. Following this a pilot CATI survey with 200 respondents was conducted, to further test for comprehension and to allow for error-trapping within the software.

Data Coding and Analysis

Data coding is automatic with the CATI system, as interviewers enter data into pre-coded fields. Data was analysed firstly by calculating the frequencies of responses, then by performing cross-tabulations of survey questions by demographic questions, using chi square as a measure of significance. This measure shows the actual number and percentage of responses and compares it to the number and percentage of those responses that may be expected relative to the size of the group, or 'by chance'. Results with a confidence interval of 95% or higher are reported below, except where the number of responses is so low that the results may be questionable.