This page contains information about data sources and limitations for SOI's annual Tax-Exempt Bonds' studies. Please visit the Tax-Exempt Bonds Statistics to access the annual data tables and articles. The Statistics of Income Division conducts annual studies of Governmental bonds and tax-exempt private activity bonds. The statistics available on the Tax Stats web site and in various publications are based on data compiled from the populations of Forms 8038, Information Return for Tax-Exempt Private Activity Bond Issues, and Forms 8038-G, Information Return for Tax-Exempt Governmental Obligations, filed with the Internal Revenue Service for bonds issued during a given Calendar Year. (For example, the most recent data - Issue Year 2012 - are based on Form 8038-G and 8038 returns for bonds issued between January 1, 2012 and December 31, 2012.) Bond issuers are required to file these tax-exempt bond information returns by the 15th day of the second calendar month after the close of the calendar quarter in which the bond is issued. The sampling period covers the issue year (year 1) and extends into the early part of year 3, in order to include as many applicable returns for a particular issue year as possible. (For example, the 2012 tax-exempt bonds studies include returns processed from January 1, 2012, to April 30, 2014, for bonds issued in 2012.) Where possible, data from amended returns filed and processed before the cutoff date were included. No efforts are made to include data from late-filed returns that are processed after the cutoff date. During statistical processing, returns were subject to thorough testing and correction procedures to ensure data accuracy and validity. Additional checks were conducted to identify and exclude duplicate returns. The data also exclude returns filed for commercial paper transactions, as well as issues that are loans from the proceeds of another tax-exempt bond issue (pooled financings). Wherever possible, returns with incomplete information, mathematical errors, or other reporting anomalies were edited to resolve internal inconsistencies. However, in other cases, it was not possible to reconcile reporting discrepancies. Thus, a certain amount of reporting and processing error may remain.