This document proposes a method for private distributed collaborative filtering using estimated concordance measures. It defines concordance as a measure of agreement between users' ratings that can be used to estimate similarity in a privacy-preserving way. The method estimates upper and lower bounds on concordance between users to calculate similarity without revealing private rating data. An evaluation shows this approach can accurately estimate similarity coefficients and generate recommendations, especially for larger and denser datasets. Future work is needed to further analyze concordance-based similarity and its effects on trust in distributed recommender systems.