Look ahead bias in refinitiv esg data

In our new paper, “Is History Repeating Itself?: The (Un)Predictable Past of ESG Ratings” we show that Refinitiv rewrites data points every week. Some changes are related to persistent changes in the underlying raw data, while others are just accidental (?) and apparently momentarily deletions of raw data. We demonstrate that these changes affect asset pricing tests. This is terrible news for the reproducibility of research based on Refinitiv’s ESG ratings. It also introduces a look-ahead bias for backtests that the finance industry so often relies on. I can see how a revised reflection of the past may be useful, but this makes it useless for any study that needs point-in-time data. If Refinitiv wants to offer a data set that reflects the latest assessment of a firm, it should, at the same time, offer a point in time data set.