We examine the effects of the Basic Elder Pension (BEP) in Korea on self-assessed health for the elderly of age 65 or higher, using two wave Korean panel data. As there could be innately healthy/optimistic individuals, it is important to allow for time-constant individual effects possibly related to regressors, which is, however, difficult as the dependent variable takes on four ordered categories. To overcome this difficulty, we transform the ordered discrete response into binary, and apply Panel Conditional Logit Estimator (PCLE) to the binary model. Since there are three ways to transform four ordered categories into binary, we obtain three PCLE’s in total. To impose the restriction that the same parameters are estimated in those PCLE’s, we use minimum distance estimator. We find that BEP has a small positive effect that is significant: 100% increment in BEP would result in an increment of 3–7% of ‘one standard deviation of the latent continuous health propensity’. Finding a small effect might be natural, as improving health for the elderly would be a hard thing to do.