Each data set is associated with a specific paper. Users can view and download all related return series at once, or select a subset of series as an Excel file.
All return series are from hypothetical portfolios from academic articles (or working papers) by researchers associated with AQR. Some series are returns of long-only portfolios in excess of the U.S. one-month Treasury bill rate, while others are differences in returns or return premia of long/short factor portfolios.
The granularity for each simulated return series varies in accordance with the research effort with which it is associated. We link each data set to a specific paper which makes it clear what portfolios and factors are represented by the data in each spreadsheet. In some cases, the simulated returns are more specific (e.g., long/short equity within U.S. stocks) or more coarse (e.g., long/short quality in stocks, globally). Additionally, simulated returns are provided monthly and, in some cases, daily returns are also available.
It is important for users of the AQR data sets to understand that all data sets provided contain the returns to portfolios described in the relevant papers, and not the live returns or backtests of AQR-specific portfolios, funds or products. (We ask users of the data set to please include the following citation: “AQR Capital Management, LLC.”)
For each article in most cases we provide two sets of data series:
We aim to update each data set monthly with a lag of about two months.
Finally, many of the files also contain additional global factors, such as those of Fama and French, which are constructed similarly to their original studies.
 The return series of the updated data sets are highly (but not perfectly) correlated with those used in the original studies, even over the same sample period. This is due to the data from the providers changing over time as well small discrepancies that occur. While we try to use the same design decisions, indicators, investment universes, and data sources as in the original work, small discrepancies are inevitable and certain design decisions are made so that we can update the data on an ongoing basis. Similarly, small discrepancies may occur in the “original” data sets provided along with each study.