Implementing Automated Data Validation for Canadian Political Datasets

Abstract

This paper describes a series of automated data validation tests for datasets detailing charity financial information, political donations, and government lobbying in Canada. We motivate and document a series of 200 tests that check the validity, internal consistency, and external consistency of these datasets. We present preliminary findings after application of these tests to the political donations ($\approx10.1$ million observations) and lobbying ($\approx711,200$ observations) datasets, and to a sample of $\approx380,880$ observations from the charities datasets. We conclude with areas for future work and lessons learnt for others looking to implement automated data validation in their own workflows.