This study develops a repeat sales price index using quantile regrssion. Quantile regression(QR) allows us to mitigate the problem of outliers comparing with the OLS estimation commonly used. This advantage can be highligted when we need to generate a price index for a thin market where there are not enough samples. In order to verify the effect, we constructed price indices for four different size groups of apartment condominiums in Seoul. A bigger group has a smaller sample size. The estimation results show that the quantile regression method is effective to smooth the peak points in indices generated by the OLS estimation. This stabilizing effect is obvious in the case of large size apartment condominiums with fewer samples. We chose three evaluation indices to compare the performance of the OLS and QR indices including statistical reliability index(mean standard error), stability index newly developed, and sensitivity index(signal-to-noise ratio). Overall evaluation suggests that QR indices allow us much gain in stability without much loss in statistical reliability comparing with OLS indices.