Study seems flawed– it preselects two different audiences (allow/not-allow offers mapping to something like low/high interest in gambling), then attributes the offers as the cause rather than the effect of selecting different audiences.