Linking Markit RED Data with CRSP

Markit RED (RED) is the market standard for reference data in the credit markets. RED provides a unique 6-digit identifier, redcode, for each entity in the database. In addition, it also carries several other common entity identifiers, such as 6-digit entity_cusip, ticker as well as company name strings.

Researchers can use these identifiers to link the CDS data with other data sources, such as CRSP and TRACE for equity and fixed income respectively.

Top of Section

We illustrate below how to link RED data to CRSP data using the new CIZ format of CRSP data. The logic would be the the same for the legacy SIZ format of CRSP data, just with different database syntax.

The primary linking key is through the 6-digit CUSIP. We also try to establish linkage through a secondary linking key, the ticker. However, it is important to emphasize here that the linking quality through ticker is fairly poor. As a result, we include an additional layer of quality check using the string comparison between the two databases' company names.

We strongly advise our users to carefully examine the linking output, and set their own quality criteria suitable for their individual research agenda.

Top of Section

Please make sure to select "SAS" in the kernel list when trying to run the code in Jupyter Notebook environment.

Jupyter Notebook: ResearchApps/LinkRedcodeCusip_sas.ipynb View Larger Download

Top of Section

As CUSIP is a fairly reliable linking key, we set a relatively relaxed requirement for spelling distance between paired company names (a low distance number implies two strings are highly similar). likewise, as a ticker linking often yields noisy results, we impose a stricter requirement of the company name comparison. As shown in the sample output below, there are quite a few records linked through ticker that yield very different company names.

We would like to emphasize here that the code above is simply for illustration, and we highly recommend users adapt the code and the criteria to meet their own research goals.

Top of Section

Please make sure to select "Python" in the kernel list when trying to run the code in Jupyter Notebook environment.

Jupyter Notebook: ResearchApps/LinkRedcodeCusip_python.ipynb View Larger Download

Top of Section

Notice that in Python with the package we used, if two strings are completely the same, the nameRatio yields 100, the opposite of the spelling distance from the SAS version.

As in the SAS version, ticker-linked pair tends to yield quite different company names, and hence lower nameRatio. We encourage our users to explore the code and output to fine tune the criteria.

Top of Section