Deep Ecological Inference

Nic Fishman, Colin McAuliffe

Keywords: ecological inference, representation learning, multi-task learning, bayesian deep learning

Abstract: We introduce an efficient approximation to the loss function for the ecological inference problem, where individual labels are predicted from aggregates. This allows us to construct ecological versions of linear models, deep neural networks, and Bayesian neural networks. Using these models we infer probabilities of vote choice for candidates in the Maryland 2018 midterm elections for 2,322,277 voters in 2055 precincts. We show that increased network depth and joint learning of multiple races within an election improves the accuracy of ecological inference when compared to benchmark data from polling. Additionally we leverage data on the joint distribution of ballots (available from ballot images which are public for election administration purposes) to show that joint learning leads to significantly improved recovery of the covariance structure for multi-task ecological inference. Our approach also allows learning latent representations of voters, which we show outperform raw covariates for leave-one-out prediction.

One-sentence Summary: We extend ecological inference with an efficient loss function, and build models to infer probabilities of vote choice for candidates in the Maryland 2018 midterm elections.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Reviewed Version (pdf): https://openreview.net/references/pdf?id=h3PF2VQb0