Timeframe
2018–2019
Funding
Vice-Chancellor’s Strategic Development Fund
COMPASS staff
Barry Milne
Lara Greaves
Martin von Randow
Cinnamon Lindsay
Olivia Healey
Collaborators
University of Auckland
Jennifer Curtin
Catherine Frethey Bentham
Pauline Gulliver
Danny Osborne
Thomas Lumley
Marama Muru-Lanning
Description
This pilot project aimed to form a panel of respondents that would be engaged to complete surveys on an ongoing basis. While this was not a novel concept, the intention was that we would assemble this panel probabilistically, with sampling that aimed to represent the New Zealand population on as many characteristics as possible.
All of the online panels that existed at the time were commercially run and non-probabilistically assembled, on an opt-in basis with individuals volunteering their services. Due to their mismatch with the target population, the statistical validity of findings taken from such non-probabilistic panels had been called into question, and we hoped to do better. Probabilistic panels had elsewhere been shown to produce robust and reliable findings that could be generalised to greater populations.
We were funded to establish a probabilistic panel, and branded it as the People’s Online Panel for New Zealand (POPNZ), including the creation of a logo and marketing material. POPNZ was intended to enable researchers across different disciplines and organisations to answer key research questions without incurring the time and money costs involved in recruiting a new sample.
The key components of POPNZ were going to include:
- A random sample of 3,000–4,000 people sampled from 18–74-year-olds on the New Zealand Electoral Rolls, stratified by age, sex, and ethnicity.
- Support for people to respond offline as well as online, to ensure representation of people without an internet connection and those that were simply net-averse.
- Eight surveys planned for the first year, with four intended for the entire panel and the rest focusing on specific subsets, e.g. an age group, ethnicity, or region.
- Surveys of public good research, i.e. not for commercial gain, on health, social, public opinion, and political topics.
- Privacy and confidentiality systems for all panel members along with ethical approval for all surveys.
- Single-topic surveys as well as ‘omnibus surveys’ where several researchers would contribute questions, and also the capacity for longitudinal surveys.
- An ideal time to complete each survey of 15–20 minutes.
We investigated using the New Zealand Electoral Rolls for sampling but found that this was inconsistent with the New Zealand Electoral Commission’s data use policies. We then tried address-based sampling, but this did not produce a high enough response rate for a viable panel; thus, we were unable to establish a probabilistic online panel.