2024 Edelman Trust Barometer P. 65 Data Analysis Explained: French Data Model Predictors Question text TRU_INS: Below is a list of institutions. For each one, please indicate how much you trust that institution to do what is right using a 9-point scale where In 2021, the translation of “government in general” in the French questionnaire was incorrectly one means that you “do not trust them at all” and nine means that you “trust changed to “authorities in general,” affecting data in all studies conducted in France from them a great deal”. 2021 to 2023. When we discovered the inconsistency in early 2023, we suppressed all Trust in Media in general affected data while investigating the possibility of modeling and replacing the missing data. institutions Business in general Over the course of 2023, we built a data model that could reliably predict what the France Non-governmental organizations (NGOs) data would have been if the translation were consistent. This involved identifying items Authorities in general (shown in the list on the right) that are predictive of trust in government in other Western Economic CNG_FUT: Thinking about the economic prospects for yourself and your democracies and testing whether they had similar predictive power in France. optimism family, how do you think you and your family will be doing in five years’ time? To build a model with a sufficient level of accuracy, we iterated through 5,000 different Government TRU_3D_GOV: To what extent do you agree with the following statement? randomly split samples, controlling for gender, income, age, and other demographics, of competence France data collected in the 2024 Edelman Trust Barometer and in the 2023 Special Report: Government in general is good at what it does Trust and Climate Change. In both surveys, we used the accurate translation for trust in POP_EMO: Some people say they worry about many things while others say “government in general,” along with the predictor variables. In both surveys, we also included Fears of gig- they have few concerns. We are interested in what you worry about. the trust in “authorities in general” attribute for our French respondents. This allowed us to Specifically, how much do you worry about each of the following? include the mistranslated variable in our model as one of the predictors for trust in economy Permanent jobs with benefits being replaced by freelance, gig-economy or short-term government. jobs that do not offer benefits GOV_PER_DIM: In thinking about why you do or do not trust government in For each split sample, 75% of the sample was used to train the data, and the remaining 25% general, please specify where you think it falls on the scale between the two was used as a test group. Using a random forest classification model, we were able to opposing descriptions. accurately classify 89.41% of respondents as trusters or non-trusters, which gave us the Government Highly effective agent of positive change confidence to repopulate the data in affected reports with data from our predictive model. ethics Honest and fair dimensions The modeled data was applied to the France data from each impacted year (2021, 2022, and Has a vision for the future that I believe in 2023) to predict the overall level of trust in government in general. The Trust Index for those Serves the interests of everyone equally and fairly years has also been updated, given that trust in government is part of that calculation.