Global Mind Project data in the United States: A comparison with national statistics

Population surveys have traditionally been conducted using mail, telephone, or face-to-face recruitment of randomly selected individuals within demographically stratified bands to acquire data samples that are representative of the population of interest. More recently, the growth of the internet has opened up the opportunity to generate large-scale samples at a much faster rate and lower cost. However, online participant recruitment methods typically result in non-probability samples that are subject to the black-box algorithms of online advertising platforms. This raises the question of whether it is possible to obtain data samples that demographically align with national statistics using web-based methods. Here we describe the online recruitment method used by the Global Mind Project (GMP) that serves advertisements through Facebook and Google Display inviting individuals to complete an anonymous, self-report mental wellbeing assessment which respondents take for the purpose of receiving a mental wellbeing score and personalized self-help report. Ads target respondents by age, sex and regional groups across a broad range of interest groups. Response rates are tracked by target group and dynamically adjusted to obtain the desired quotas. GMP data from the United States (83,589 responses from 2020 and 2023) were compared against time-matched data from the American Community Survey, Household Pulse Survey and the American Trends Panel for questions where there were exact or near-exact matches. These included educational attainment and marital status by age and biological sex, the percentage seeking treatment for mental health problems and number of close friends. Demographic trends match within 5-10% with a slight bias in GMP data towards single people with fewer friends who were seeking mental health treatment providing a demonstration that GMP data obtained in the United States closely matches trends in census data.