The Survey | Survey Methods
Community-focused Maui recovery research
The survey focuses on people who lived, worked, or owned a business in West Maui or Kula during the Maui wildfires in August 2023. Participant recruitment is facilitated by community and government partners. Participants receive a $20 gift card for the first survey and a $10 gift card for each monthly follow-up survey that they take. It is not a closed cohort, i.e. eligible people can join anytime. The survey form is available online in six languages: English, Spanish, Tagalog, Ilocano, Tongan, and Vietnamese. Recruitment primarily relies on multilingual email outreach but is supplemented by physical flyers available in community spaces and at local events. The Council for Native Hawaiian Advancement provided critical support for community outreach and initial project planning.
The survey questionnaire is designed to follow key indicators such as income, employment, housing type, rent, residential, and job location over time. Participants are asked about their situation before the wildfires, followed by monthly updates on the same set of questions starting in August 2024. The survey questionnaire is available below.
We have used a wide range of data verification checks and individual participant follow-ups to eliminate fraudulent responses from the data. We are confident that this data represents real fire-impacted people. So far more than 900 people have participated in the survey, with an average monthly sample of about 425 responses. All housing and income-related data is reported on a household basis whereas employment and needs data is based on individual-level records.
The survey data has been weighted to maximize representativeness of the fire-affected population. The person-level data has been weighted by age, employment and income-to-poverty ratio and homeownership and poverty status have been used for weighting for the household-level data. Iterative proportional fitting (raking) has been used to implement survey weighting. This method continues to adjust the weights until the sample distribution approximately matches the population distributions for the selected weighting variables. The design effect ranges between 1.1 and 1.5 for the different sets of survey weights.