Historical Data on Repurchase Agreements from the Canadian Depository for Securities

We develop an algorithm that extracts information about sale and repurchase agreements (repos) from disaggregated settlement data in order to generate a new historical dataset for research. Data from Canada’s fixed-income settlement authority, the Canadian Depository for Securities (CDS), is a valuable source of historical information on Canada’s fixed-income markets, especially from 2003 to 2016 when few other data sources were available. However, the CDS does not contain details on the terms of trade for repos, such as the repo rate, term or haircut. In the data, each repo is recorded as two distinct settlements but, critically, the sale and repurchase legs of a repo are not explicitly associated. We use a variant of the Gale-Shapley algorithm to solve a “stable roommates” problem to link repos’ sale and repurchase transactions and compute their terms of trade. We verify our algorithm by running it on a separate dataset that explicitly associates the sale and repurchase legs of a repo. In addition, we verify the computed repo terms of trade by comparing a subsample of the CDS data with a third dataset that reports terms of trade directly. The derived data are useful for researchers to study the evolution of fixed-income market structure and market conditions.