Annex A. Methodological notes on the OECD Trust Survey implemented in Brazil
The OECD Trust Survey, carried out by the OECD Directorate for Public Governance, had around 2 000 respondents per country in the twenty-two participating OECD countries: Australia, Austria, Belgium, Canada, Colombia, Denmark, Estonia, Finland, France, Ireland, Iceland, Japan, Korea, Latvia, Luxembourg, Mexico, the Netherlands, New Zealand, Norway, Portugal, Sweden and the United Kingdom. In Brazil, the survey was implemented online with a final sample of 4 140 respondents.
The survey process and implementation were guided by an Advisory Group comprised of public officials from the Controladoria-Geral da União (CGU), representatives of National Statistical Offices from Brazil and Colombia, and international experts, including representatives from Latinobarometer, the Americas Barometer (LAPOP) and the Inter-American Development Bank.
The survey was conducted online by the survey company Netquest between 7 April and 6 May 2022, sourced by Netquest’s and its partners panel in Brazil.
The Trust Survey questionnaire was prepared in English, translated into Portuguese and reviewed by CGU’s public officials and public governance specialists that were also part of the Advisory Group established for the Trust Survey in Brazil.
The OECD Trust Survey uses an eleven-point scale for the response choices on questions about levels of trust and drivers of trust, following reviewed best practices (OECD, 2017[1]) and applications in country studies in Korea, Finland, Norway, and New Zealand (OECD/KDI, 2018[2]; OECD, 2021[3]; OECD, 2022[4]; OECD, 2023[5]). A numerical 0-10 scale with verbal scale anchors is recommended and used here for survey questions on trust and drivers of trust, as it allows for variance in responses, increases overall data quality and complexity, and facilitates translatability across languages. This presents a more nuanced analysis, allowing respondents to provide a “neutral” response that other surveys do not allow.
The online survey uses a non-probability sampling approach, based on ex-ante country-level quotas representative of the Brazilian population by age, gender, level of education, socioeconomic category and region. The quotas were derived from national estimates of these population groups based on probabilistic surveys and census data. The non-probability sample construction was the most feasible option for the OECD Trust Survey given its simplicity, timeliness, and lower cost. The implementation in Brazil largely follows the same quota design as in other participating countries, with one exception. Instead of using income as a soft quota (bottom 20%, middle 60% and top 20%) (Nguyen et al., 2022[6]), in Brazil socioeconomic categories were used following the Brazil Economic Classification Criteria (Critério Brasil-ABEP). The Brazil Economic Classification Criteria classifies households into five groups (A-E) based on their estimated purchasing power. It assesses access to public utility services, level of education and possession of several amenities (such as bathroom, dishwasher, freezer, etc.) (ABEP, 2021). The quotas were derived from national estimates of group prevalence based on probabilistic surveys and census data.
Responses were collected until the country-specific quotas were filled and post-stratification weights were calculated using the “random iterative method (RIM)” based on age, gender, education, region and socioeconomic status. The median interview duration was 19 minutes.
The econometric results presented in Chapter 2 are based on three logistic regression analyses for establishing the main drivers of trust in the federal government, local government, and civil service in Brazil.
Based on the OECD Framework on the Drivers of Trust, respondents’ perceptions of the responsiveness, reliability, openness, integrity and fairness of government and public institutions are expected to be the main drivers of trust in the three institutions (federal government, local government and the civil service). Trust in each of the institutions is recoded as a binary variable (low or no trust: 0-4 and high or moderately high trust: 6-10). Neutral responses (5) and “don’t know” are excluded. The analysis operationalises government competencies and values through 15 variables, measured on a 0-10 response scale and standardised for the analysis. The model also includes six further variables: internal political efficacy, external political efficacy, satisfaction with administrative services, the perceived relevance of knowing a broker to access public services, willingness to formalise any new business and confidence in the country’s ability to tackle environmental challenges.
The following explains the technical details about the econometric analysis.
Stepwise deletion: All the drivers of trust are included in the three baseline regression models and those which are not statistically significant are deleted (stepwise deletion process). The results in this section show all the significant drivers of trust in the three institutions.
Model specification: All models include survey weights and control variables for individuals’ socio-demographic characteristics (age, gender, education, social class and region of residence), interpersonal trust, perceptions of economic and physical insecurity, and whether the respondent voted for the parties in government. Missing data are excluded using listwise deletion.
Technical interpretation: The statistically significant drivers are shown as average marginal means. The technical interpretation of the effect of government effectiveness to reduce greenhouse emissions on trust, for example, is that one standard deviation increase in the perceived likelihood that the government is effective is associated with a 10 percentage point increase in trust in the federal government. Or – taking into consideration all other variables in the model – all else being constant, moving from the typical citizen to one who is one-standard-deviation more satisfied, results in a 10 percentage point increase in trust in the federal government in Brazil.
References
[6] Nguyen, D. et al. (2022), “Survey design and technical documentation supporting the 2021 OECD Survey on Drivers of Trust in Government Institutions”, OECD Working Papers on Public Governance, No. 53, OECD Publishing, Paris, https://doi.org/10.1787/6f6093c5-en.
[5] OECD (2023), Drivers of Trust in Public Institutions in New Zealand, Building Trust in Public Institutions, OECD Publishing, Paris, https://doi.org/10.1787/948accf8-en.
[4] OECD (2022), Drivers of Trust in Public Institutions in Norway, Building Trust in Public Institutions, OECD Publishing, Paris, https://doi.org/10.1787/81b01318-en.
[3] OECD (2021), Drivers of Trust in Public Institutions in Finland, Building Trust in Public Institutions, OECD Publishing, Paris, https://doi.org/10.1787/52600c9e-en.
[1] OECD (2017), OECD Guidelines on Measuring Trust, OECD Publishing, Paris, https://doi.org/10.1787/9789264278219-en.
[2] OECD/KDI (2018), Understanding the Drivers of Trust in Government Institutions in Korea, Building Trust in Public Institutions, OECD Publishing, Paris, https://doi.org/10.1787/9789264308992-en.