Sources and Accuracy

Sources

Prior to January 1962, treasury data were obtained from a number of different sources. These sources include the Wall Street Journal, Salomon Brothers, Inc., and the Bank and Quotation Record.

Beginning with January of 1962, the majority of prices came from the Composite Closing Quotations for US Government Securities compiled by the Federal Reserve Bank of New York (FRBNY). In 1984, the quotation sheets were renamed the “Composite 3:30 P.M. Quotations for US Government Securities”. The time at which the quotes were compiled was related to the fedwire deadline the FRBNY set for the transfer of securities. The deadline was set for 2:30 p.m. Eastern Time, but was regularly extended as much as three-quarters of an hour. The FRBNY trading desk began a “closing run” at 3:00 p.m. The reference to “closing quotations” from 1962 to 1984 probably refers to the “closing run” at the FRBNY. With the close of the day on October 15th, 1996 the FRBNY discontinued publication of composite quotations. 

The start of the day, October 16, 1996, our source for daily and monthly price quotations, maturity dates, and coupon rates changed to GovPX, Inc. GovPX receives its data from 5 inter-dealer bond brokers. Live, intra-day bids, offers and transactions in the active over-the-counter markets among these primary dealers are the source of GovPX’s 5 p.m. End-of-day US Treasury prices.

GovPX was acquired by ICAP in 2008. Beginning in February 2009, CRSP released its daily and monthly treasury databases using the new ICAP data. 

Bid and Ask Quotes 

The FRBNY described its listed bid price as “…the most widely quoted price from the range of quotations received”. The ask price was determined by the FRBNY based on what they expect a typical bid-ask spread to be. The rule used to make this derivation was not public domain. 

GovPX described its listed bid and ask prices as the “best price”. To determine their “best price” they observe the prices from the 5 inter-dealer brokers and report the bid and ask prices that produce the smallest bid-ask spread.

A further distinction must be made after the acquisition of the GovPX data by ICAP. The two data sources handle bid and ask quotes differently. ICAP provides the actual bid and ask quotes, thus calculated spreads will fluctuate daily. GovPX imputed quotes from their available data. When looking at a time series of spreads,  using GovPX data prior to February 2009, for the most part, they are constant. Beginning in February 2009 with the actual quotes from ICAP operations reported in their 5pm file, fluctuation in the spreads may be observed. In both cases the midpoints of the real and imputed spreads are very close.

Debt Outstanding

The total amount outstanding (TMTOTOUT, TDTOTOUT) is obtained from the Monthly Statement of the Public Debt of the United States published by the Treasury Department. The amount publicly held (TMPUBOUT, TDPUBOUT) is obtained from the quarterly US Treasury Bulletin. Money Rates are obtained from the Federal Reserve. The following non-derived data: issue date, coupon payable dates, bank eligibility, tax status and call status are obtained from the US Treasury Department.

CUSIP

Prior to 1990, CUSIP was obtained from Standard & Poor’s CUSIP Directory. From January, 1990 through October 15th, 1996, the CUSIP was obtained from the Composite 3:30 p.m. quotations for US Government Securities. GovPX, as of October 16, 1996, provided the CUSIP number. Since February 2009, ICAP provides CUSIP. When in question, the CUSIP is verified by Standard & Poor’s CUSIP Directory.

Accuracy

All data are checked for internal consistency with each release of the file. Secondary sources, such as the Wall Street Journal, are used to check suspect prices.

Considerable resources are expended in checking and improving the quality of the data. Errors are not common. Some of the errors found in checking the data are the results of inaccuracies in the initial data source. The inaccuracies are corrected as soon as possible. Other errors are CRSP coding errors; over time these coding errors are found and corrected. Historical corrections account for the differences in the data from update to update.