Interbank repo markets are arguably just as important as unsecured markets. Despite this, the global research community has not analysed the microstructure of interbank repo markets in the same detail as unsecured markets, because loan-level repo data have not been available. This paper provides and assesses an algorithm for extracting loan-level repo data from over-the-counter securities transactions data, and applies it to securities transactions data from Austraclear. This approach is similar to how loan-level unsecured data are typically obtained from payments data. False detection and false omission rates are estimated to be 3 per cent or less. While separate prudential data indicate a larger repo market than the algorithm data, likely reflecting repos transacted through foreign (i.e. non-Austraclear) infrastructure, the two datasets have a robust positive relationship.
The algorithm data, capturing non-RBA repos of up to 14-days maturity from several 2-month data samples between 2006 and 2015, reveal various market features. From 2006 to 2015, the distribution of repo-rate spreads (to the cash rate) drifted up and tightened, and the market shifted towards overnight maturities. Loan-level repo rates depend on the loan size and the types of counterparties, but not how long the repo is open. In 2015, the market’s network structure comprises a tightly integrated core, and a segmented periphery with few counterparties. Repo haircuts do not display obvious patterns, appearing randomly distributed around zero.