Growing and pruning the archive: an agent-based model to build letter correspondence networks

Abstract The selection and availability bias concerning digital archives can be a problem for empirical studies. In the context of historical correspondence networks, we still have limited information about how missing data can affect and hinder the variables of our interest. Along these lines, we introduce an agent-based model to reconstruct historical communication using letters. We use this model to simulate the letter-sending process and its subsequent archival. We use it to build counterfactual networks and learn how the missing data caused by the archival of the letters might bias the network statistics.