## Data Availability and Access for “Haircut, Interest Rate, and Collateral Quality in the Tri-Party Repo Market: Evidence and Theory” by Sangyup Choi, Inkee Jang, Kee-Youn Kang, and Hyunpyung Kim

The transaction-level repo data used in this paper are non-public, restricted-access data. They are not included in the replication package because the original transaction-level records and the cleaned analysis dataset cannot be publicly redistributed by the authors. However, the data are not proprietary to the authors.

The repo transaction data are managed by the Korea Securities Depository (KSD). Researchers interested in accessing the underlying daily transaction-level repo data may submit an information disclosure request through the Korean public information disclosure portal:

https://www.open.go.kr/com/main/mainView.do

When submitting the request, researchers should request daily transaction-level repo data managed by the Korea Securities Depository. Access to the data is subject to review and approval of the information disclosure request, as well as any applicable data-use restrictions imposed by the relevant institution. The public information disclosure portal currently provides services primarily in Korean.

Related information on securities depository statistics is also available from the Korea Securities Depository’s information disclosure page:

https://www.ksd.or.kr/ko/information-disclosure/resource-center/securities-depository-statistics

This page is also currently provided primarily in Korean.

The main analysis dataset used by the replication code is `repo_complete.dta`. This file is a cleaned and merged analysis dataset constructed from the transaction-level repo data and security-level information from FnGuide. The merge is performed using collateral-level identifiers, including ISINs, and is used to construct collateral characteristics such as asset class and remaining maturity. Access to FnGuide security-level data may require a separate subscription, license, or institutional access.

Because redistribution of the transaction-level data is restricted, `repo_complete.dta` is not included in the public replication package. The submitted Stata code documents the estimation and output-generation steps applied to the analysis dataset. An approved replicator with access to the underlying transaction-level repo data and the required FnGuide security-level information can reproduce the tables and figures by constructing or obtaining `repo_complete.dta`, placing it in the `dta/` directory, and running `combined.do`.

The replication package includes:

1. `combined.do`, which generates all the tables and figures in the paper except for Figure 4 from `repo_complete.dta`;
2. `repo_relation_Final.m`, which generates Figure 4;
3. output directories for tables and figures;
4. this README file, which describes the required input dataset, data access procedure, variable definitions, expected running time, and the structure of the replication code.

Expected Running Time

combined.do: The expected running time is approximately two hours on a standard desktop computer. 
repo_relation_Final.m: The expected running time is less than 10 seconds on a standard desktop computer.