TRACE error-filter  [pdf] [codes]
This note presents the steps used to error-filter the Trade Reporting and Compliance Engine (TRACE) database. An adaptation to the Dick-Nielsen filter is proposed, particularly useful when processing memory and/or speed are binding. I present the structure of the code and the main filtering statistics. Considering the Academic TRACE information between January 2016 and December of 2019, the filter can match 99% of the 1.8 million targeted reports, filtering 4.2% of the 85.5 million observations. Two codes are provided. The first code merges daily raw files sent by FINRA. The second code filters errors for files post Feb 2012 dissemination change. The pre Feb 2012 error-filter code is available on demand.