We propose a penalized likelihood method that simultaneously fits the multinomial logistic regression model and combines subsets of the response categories. The penalty is nondifferentiable when pairs
Not willing to pay was used as the reference category. Pseudo R2 = 0.118 (Cox and Snell), 0.134 (Nagelkerke). Model Chi Squared (df = 4) = 17.008, p = 0.02.