Social preference models were originally constructed to explain two things: why people spend money to affect the earnings of others and why the income of others influences reported happiness. We test these models in a novel experimental situation where participants face a risky decision that affects only their own earnings. In the social (individual) treatment participants do (not) observe the earnings of others. In the social treatment gambles therefore not only affect absolute but also relative earnings. Outcome-based social preference models therefore predict a treatment difference. We find that decisions are generally the same in both treatments, in line with rule-based social preference models, like procedural fairness.