Assessio collects demographic data from candidates. It is not uncommon for candidates to ask questions about why we do this. This article addresses demographic data, why we collect it and what value it brings to candidates and organizations.
First and foremost: the demographic data from candidates is not used as part of the recruitment process by the (potential) employer. They don’t even have access to the demographic data. Instead, the data is only used by Assessio, and only to ensure and improve our assessments. Valid demographic data are crucial to ensure high quality assessments and especially to minimize adverse impact such as bias and discrimination
It may also be good to know that the data is always deidentified before use, meaning that the data is not to be traced back to an individual candidate. When we ask you to share your demographics, we do so to let us ensure that future candidates will have a fair and unbiased assessment, just as your assessment rely on other candidate’s participation and willingness to contribute.
Demographic data that is collected
The following data is collected from candidates before they complete an assessment at Assessio.
- Gender
- Birth year
- Educational level
- Current work area
- Nationality
- Country of residence
- Native language
Reasons for demographic data collection
There are multiple reasons to collect demographic data. Below are the five most important reasons for Assessio to collect the data. In short, the data serves to create a valid and fair assessment for all –by continuously improving the tests and the norm groups.
Ensure well-composed, high-stake norms
Assessio offers normative psychometric tests, meaning that your scores are compared to a relevant norm or reference group. It is a prevailing view that large norm groups are the most important when assessing the quality of a norm. While sample size is important, the more important factor in securing high quality norm groups is the composition. To have a fair group of reference, it needs to be balanced in terms of gender and age groups and reflect the labor market as good as possible. The more valid demographic data we can collect, the more well-composed norms we are able to create.
Limit item bias and minimize the risk of discrimination
When creating new psychometric scales, it is a common pitfall to miss potential item bias – for instance, that women are more likely to endorse a statement, not because they show more of the targeted trait but because they are women. A famous example is an item in a depression scale; Are you prone to tears? More women than men confirm this statement, but is that because they are more depressed? Item bias essentially risk introducing bias that can result in discrimination, which we ultimately want to avoid. To reduce item bias, we need to research response patterns for different demographic groups.
Minimize and document adverse impact
Adverse impact is the risk of favoring applicants from certain demographic groups. This can happen because of item bias, but it can also happen because of true mean differences between demographic groups. To prevent discrimination as a consequence of the use of our tools, we need to document their potential adverse impact and educate and inform our customers as to how they best avoid discrimination.
Ensuring validity and reliability of our products
As a provider of selection, development and performance tools, we are accountable to ensure and document the quality of our products. To do so, we need to investigate and prove that they are a valid and reliable measure of the construct they are aiming to measure. To do so, we need the data from multiple candidates. One way of researching validity is to prove equivalence with other research findings on specific group differences.
Ensuring equivalence of translations
An important part of ensuring quality of our products is to provide the necessary language versions and let as many candidates as possible complete the questionnaire in their native language. Translating the questionnaire is however, not enough. To ensure that comparisons between language versions are valid and fair, we need to document equivalence and the effects of completing the questionnaire as a non-native speaker. For this purpose, we also need demographics.