Metadata: Satisfaction with policy

Flemish Statistical Authority

www.vlaanderen.be/statistiek-vlaanderen/contact

02/09/2025

The data for these statistics are collected as part of the SF survey organised by the Flemish Statistical Authority. The SF survey is a survey of the population aged 18 and over in the Flemish Region and is organised several times a year in different editions. The survey gauges the opinions, beliefs and behaviours of the population with regard to social and policy-relevant themes. See the technical document for an overview of which themes were surveyed during which edition (https://assets.vlaanderen.be/image/upload/v1710164286/Technische_nota_SV-bevragingen_b3fjbr.pdf (in Dutch)(PDF file opens in new window)).

The fieldwork is always organised by a market research agency under the coordination of the Flemish Statistical Authority. The SF survey is a so-called “mixed mode” survey. This means that respondents can complete the questionnaire in different ways. The selected sample persons are first contacted by letter via the post with a request to complete the survey via the Statistics Flanders website. However, this invitation also announces that a paper questionnaire will be sent later if the sample person is unable or unwilling to complete the questionnaire online. After approximately one week, non-respondents receive a reminder letter with another invitation to complete the questionnaire online. After approximately two weeks, a second reminder is sent to complete the questionnaire online or via an enclosed paper form that can be returned free of charge.

The questionnaire is designed to be easy to complete on a smartphone (‘mobile first’ design) as well as on paper. Questions are therefore deliberately kept short and written in understandable language, with cross-references between questions avoided as much as possible.

For more information, see https://assets.vlaanderen.be/image/upload/v1710164286/Technische_nota_SV-bevragingen_b3fjbr.pdf (in Dutch)(PDF file opens in new window).

For all editions of the SF survey, the sampling frame consists of all residents of the Flemish Region aged 18 or older who are registered in the National Register and are part of private households (i.e. not communal households such as residential care centres or prisons). The sampling frame includes both registered Belgians and non-Belgians. A random sample of 6,000 residents is drawn each time.

Satisfaction with policy is monitored in the SF survey via the following question:

‘How satisfied are you in general with the policy of …?’

- your local government

- your provincial government

- the Flemish government

- the federal (Belgian) government

- the European government

Respondents can answer with a score from 0 to 10, where 0 stands for very dissatisfied and 10 for very satisfied. Respondents are also offered the answer category “Don’t know/no answer” for each question.

Due to the change in methodology, the results of the SF survey are not comparable with the results of the Survey on Socio-Cultural Shifts in Flanders (SCV survey), which was conducted annually by Statistics Flanders from 1996 to 2018. The SF survey is a “mixed mode” survey (online and on paper), while the SCV survey was a “face-to-face” survey (with interviewers visiting respondents at home).

Persons who do not tick a number from 0 to 10, who tick “Don’t know/No answer” or persons who tick several answers at the same time are not included in the calculation of the results.

The statistics are always calculated for all respondents and for the following subgroups: men, women, 18 to 34-year-olds, 35 to 49-year-olds, 50 to 64-year-olds, 65-year-olds and older, people living alone, respondents living with a partner without children, respondents living with a partner and child(ren), respondents living without a partner but with child(ren), respondents living with parents, low-educated persons (persons without a diploma or with at most a lower secondary education diploma), medium-educated persons (persons with at most a higher secondary education diploma or a post-secondary non-tertiary education diploma) and high-educated persons (persons with a higher education or university diploma).

The data are weighted using post-stratification weights based on age, gender, nationality and degree of urbanisation of the place of residence. The weights are calculated using raking in R. Confidence intervals are calculated taking into account unequal selection probabilities resulting from the post-stratification weights using complex sampling procedures. Confidence intervals for percentages are first calculated on the logit of the probabilities and then transformed back to the scale of the percentages.

The statistics are calculated on the basis of the various editions of the SF survey. Such a survey provides a valuable picture of opinions, attitudes and behaviours among the population. However, there are methodological considerations that should not be overlooked when interpreting the results. Responses are susceptible to selection bias (those who participate often differ systematically from those who drop out or refuse to participate), and this can distort the results. The way in which questions are formulated can also mean that answers reflect wishes and opinions rather than actual behaviour. Because each survey is a snapshot in time, current events and social debates can also influence the answers. For these reasons, it is important to interpret the results with the necessary nuance and, where possible, to link them to other data sources.

The various editions of the SF survey have consistently achieved a response rate of between approximately 30% and 35%. Although this is a fairly good response rate compared to many other surveys, the majority of respondents still do not complete the questionnaire. Respondents who do not respond may differ significantly from those who do respond.

Due to the inevitable differences in response rates based on background characteristics, the final sample is weighted for the calculation of the results. This means that underrepresented groups are given more weight in the calculation of the statistics, while overrepresented groups are given less weight. In this way, the proportions between the different population groups in the final sample are restored and brought into line with the Flemish population. The following four characteristics are taken into account when weighting the SF survey: gender, age, nationality (Belgian/non-Belgian from the EU/non-Belgian from outside the EU) and degree of urbanisation. The weighting therefore ensures that the respondents are representative of the Flemish population in terms of these characteristics. However, this does not automatically mean that they are also representative of the questions on which the statistics are calculated. Caution is therefore still required.

Statistics may also be distorted by random selection errors, as only a random sample of the population of the Flemish Region is approached rather than the entire population. This uncertainty must therefore be taken into account when interpreting the results of the SF survey. The smaller the sample on which the figures are calculated, the greater the uncertainty. Figures calculated on fewer than 100 respondents are therefore not published because they are not sufficiently reliable.

However, this uncertainty due to sampling can be calculated and translated into margins of uncertainty, also known as confidence intervals. A 95% confidence interval was always calculated for the statistics. This means that if the measurement were to be repeated 100 times in the same population with different samples, the true value among the population would lie within the interval in approximately 95 samples. The influence of the weights is also taken into account when calculating the confidence intervals. After all, the purpose of the weights is to reduce bias in the estimated percentages, but they can also increase uncertainty around these estimates and widen the confidence intervals.

When presenting the results, the evolution of the total score and the differences according to background characteristics for the most recent year are shown (e.g. by gender, age, household position and level of education). In order to assess the differences between the total scores of different years or the scores of different groups within a particular background characteristic (e.g. between men and women), the confidence intervals of the measurement are examined. When the confidence intervals do not overlap, we speak of a statistical difference. However, this does not necessarily mean that the difference is also significant in terms of content, as this involves a subjective evaluation.

The questions themselves can also cause bias. Respondents may systematically misinterpret questions, may not make enough cognitive effort to answer the question accurately, or may give socially desirable answers. In addition, answers may vary randomly depending on the time and context in which the questionnaire is completed. Good questionnaires are designed to minimise the risk of such bias as much as possible. However, this is not an exact science. Compared to other question formulations, a particular question may work well for one respondent in a particular context but less well for another respondent in a different context.

Geographically comparable and comparable over time within the SF survey

Statistics Flanders: SF survey

To the statistic