Information aggregation and communication in organizations

Operating units must communicate their private information regarding decisi ons to be taken in organizations. This paper characterizes the optimal comm unication structures assuming that (i) a decision maker is fired whenever h e makes a decision that proves wrong ex post relative to the status quo; an d ii) direct communication in a group of k units may result in the loss of messages with a probability that solely depends on the group size. Several levels of partitioning with direct communication taking place in each group are required. It is shown that there exists a group size that allows commu nication technology to be exploited optimally: The optimal communication st ructure is such that it is essentially composed of groups of this size only at every level of partitioning.