Patent classification systems are largely designed for administrative purposes, limiting their value for most research purposes. To address this deficiency, Hall, Jaffe, and Trajtenberg (2001) (link is external) developed a higher-level classification for the National Bureau of Economic Research (NBER) Patent Citation Data File by aggregating U.S. Patent Classification (USPC) classes into economically relevant technology categories. While this NBER classification scheme has proven valuable for researchers investigating US patent grants, comparable information on patent applications remained unavailable. For that reason, the Office of Chief Economist (OCE) developed a probability-matching algorithm to apply NBER classifications to patent applications as well as in-force and expired patents. From matched data, we construct the USPTO Historical Patent Data Files, four research datasets containing time series and micro-level data by NBER sub-category on applications, grants, and in-force patents spanning two centuries of innovation. Our hope is that researchers will make use of these data which, for the first time, enable detailed study of the complex dynamics between new filings, pendency, and abandonment and put into context recent trends in patenting activity, litigation, and technological change.