Methodology

Two kinds of study

A FieldGrid survey is sampled fieldwork: people are selected to represent a defined population and interviewed face to face. A FieldGrid poll is open online participation: anyone with the link can answer.

A survey can estimate what a whole population thinks. A poll can only report what the people who took part said. FieldGrid labels every result as one or the other and never combines the two.

Surveys: design

Each survey starts from the decision it is meant to inform. We agree the population, for example adults in one state, and the sample. [TODO: sampling method and frame]

The questionnaire is reviewed for leading or loaded wording before fieldwork. [TODO: who reviews it, and the languages used]

Surveys: fieldwork

Interviews are carried out face to face by interviewers recruited in the areas they work in. [TODO: recruitment, training and certification]

The interviewer explains the study and asks for consent before the first question. Each interview is recorded on a tablet, which adds the location at the start and end, the start and end times, and the interviewer's ID.

Surveys: checks

Records are checked as they arrive. An interview is flagged for review if, for example:

• it is shorter than the questionnaire allows [TODO: minimum length rule]
• it was recorded outside the area the interviewer was sent to [TODO: location tolerance]
• the same answer is given to every question in a long set
• it looks like a duplicate of another interview

A sample of interviews is checked again by people who did not conduct them. [TODO: share back-checked, and revisit or phone] An interview that fails is removed and replaced. [TODO: replacement rule]

Surveys: analysis

Results are broken down by state, local government and group. Any adjustment to the sample is described with the result. [TODO: weighting policy]

Polls

Pulse polls are open to anyone with the link. Each response is checked so that one person counts once. [TODO: how the check works]

Because the people who answer choose themselves, a poll has no margin of error and does not speak for a wider population. Poll results are always labelled as a poll and reported apart from surveys.

What is published with every result

• whether it is a survey or a poll
• the population asked and the number who answered
• the fieldwork dates
• how people were reached and interviewed
• the margin of error, for surveys
• who commissioned the study

[TODO: publication and corrections policy]

Privacy

Answers are reported in groups and never by name. The privacy notice explains what is collected and what you can ask us to do with it.