Improving data accuracy – InFact now weighs by education

With education-based weighting, we’re taking InFact surveys to the next level. The results become even more representative, and you get even deeper insight into how opinions and attitudes vary between different groups – by gender, age, geography, and now also education. In short: more precise data and a stronger foundation for smarter decisions.

To deliver representative surveys, we at InFact have always been committed to ensuring that our survey samples reflect the population as accurately as possible. That’s why we adjust for biases in who responds, so that the composition of the data matches the actual population we want to describe.

Like other Norwegian research agencies, we have until now adjusted for – or “weighted by,” in technical terms – gender, age, and geography/region. Now we’re taking a new step at InFact: we’re introducing education weighting to further enhance the quality of our analyses.

What is weighting – and why do we do it?

When we conduct a survey, the answers will never perfectly mirror the population. Some groups respond more than others – for instance, younger people tend to participate less, while older respondents are often more eager to answer. To correct for this, we use a procedure called weighting.

Weighting means assigning each response a “weight” that reflects how large a given group is in the population. If we have too many respondents from one group, their answers are weighted down; if we have too few from another group, their responses are weighted up. This ensures that the overall results better represent the full population we aim to describe.

How is it done?

We use official population data from Statistics Norway (SSB) to determine the distribution of various groups in the population. If we want a nationally representative survey among everyone over the age of 18, the numbers show, for example, that women make up 50.2% and men 49.8% of the population.

In our sample, however, the distribution might be slightly different – for example, 50.9% women and 49.1% men. To correct this, we calculate weights so the sample aligns with the population. In this case, each woman’s response is weighted by a factor of 0.986, while each man’s response is weighted by 1.014. If the sample already matches the population, both men and women are given a weight of 1.0.

Why is education weighting important?

Education has proven to be a systematic source of bias in surveys:

People with higher education are more likely to participate in surveys than those without higher education. This can affect the results of questions related to politics, civic engagement, culture, and media consumption – areas where education level often plays a significant role.

By including education in our weighting procedures, we reduce the risk of our findings being skewed towards the highly educated, and we ensure the results better reflect the broader population. With education now part of our weighting, our surveys become even more representative and reliable.

In other words, you can be even more confident that the numbers you receive from us truly reflect reality – regardless of disparities in which educational groups actually responded to the survey.

Also enables breakdowns by education

These new procedures also unlock new possibilities in our analyses. Previously, we have delivered crosstabs with breakdowns by gender, age, and geography/region as part of our standard outputs. This has made it possible to view both the overall numbers for the entire sample and how different subgroups – such as men and women – responded to each question in the survey.

Now we will also include education as a standard dimension, so that you, as the client, can receive results broken down by educational level. This provides detailed insight into how opinions, attitudes, preferences, and behaviors vary between different educational groups – enabling more nuanced interpretation and analysis.