Measurement invariance of the Domain-Specific Risk-Taking (DOSPERT) scale
Dillon Welindt, David M Condon, Sara J Weston
*Journal of Behavioral Decision Making*
Why we asked
Researchers often compare risk-taking between groups, such as men and women. Those comparisons only make sense if the questionnaire works the same way for everyone. That property is called measurement invariance, and we tested whether the widely used Domain-Specific Risk-Taking scale (DOSPERT) has it.
What we did
We recruited 412 people on Prolific, with roughly 50 in each combination of high or low SES, White or non-White, and female or male, so that these characteristics were not confounded. Everyone completed the 30-item DOSPERT, which asks about the likelihood, benefit, and riskiness of 30 behaviors. We compared hierarchical and bifactor models and then tested invariance across groups.
What we found
The bifactor models fit better than the hierarchical ones, so we used them for the invariance tests. No model met generally accepted criteria for configural fit. In other words, the scale did not show the same overall structure across groups. Many models also failed the cutoffs for the stricter metric and scalar levels.
What it means
These results add to earlier findings that the DOSPERT does not behave equivalently across demographic groups. We suggest that the field develop a risk scale that does.