|
|
Microstructure of China's Interbank Repo Market |
ZHANG Jinfan, GUO Yunhan
|
School of Management and Economics (SME), The Chinese University of Hong Kong, Shenzhen |
|
|
Abstract The repo market plays a crucial role in the stabilization of financial markets and the transmission of monetary policies. Since the global financial crisis in 2008, scholars have increasingly focused on the performance of the repo market. In China, the repo market has experienced rapid growth and has become a key component of the financial system. The existing literature mainly focuses on the impact of the repo market on bond pricing and its regulations. Few studies have examined the microstructure of China's repo market. This paper fills this gap by examining high-frequency intraday repo rate movements in China's interbank repo market using transaction-level data. Our data consist of close to ten million transactions from February 2015 to June 2021. With this unique dataset, we find that the overnight repo rate demonstrates the following robust W-shaped intraday pattern: The rate initially decreases, then rises, declines again, and finally increases once again. Our detailed findings are summarized as follows. First, the repo rate typically starts at a fixed high level at 9∶00 and declines until 10∶00. We compare the repo rate of transactions between all institutions and the rate between banks only and find that they do not significantly differ in the first hour. This result indicates that transactions during this period are mainly among banks, which are less likely to default and use safer collateral. Because of the high opening rate, the market seeks an equilibrium rate during this period, resulting in the observed declining pattern. Second, the repo rate increases from 10∶00 to 12∶00. We attribute this increase to the participation of non-bank institutions, which are riskier and thus face higher repo rates. In China, holidays and weekdays are usually concatenated, effectively turning weekends into working days; these are defined as working weekends in our paper. On working weekends, banks continue to operate normally, whereas most non-bank institutions are closed. Taking advantage of this unique institutional setup, we demonstrate that the increase in the repo rate almost disappears when non-bank institutions are absent from the market. Third, the repo rate declines again from 15∶00 to 16∶30. In addition to explanations based on volatility and liquidity from the existing literature, we identify that this pattern is driven by banks' intraday liquidity management. Banks need to retain intraday liquidity to accommodate potential withdrawals by firms. Only in the late afternoon, when firms are not likely to withdraw, banks will lend their surplus funds in the repo market, leading to a decline in the repo rate. We conduct an event study by examining repo rate movements on the deadline days for corporate tax payment. The results reveal that the decline between 15∶00 and 16∶30 is significantly larger before a tax payment due date than after it, when banks do not need to maintain extra liquidity in the late afternoon. Our results indicate that the oversupply of liquidity resulting from banks' liquidity management contributes to the repo rate decline in this period. Our paper has crucial implications for regulators and market participants. First, our findings are useful for the construction of benchmark interest rates. This paper shows that the repo market is volatile and may not have efficient pricing between 9∶00 and 10∶00 and between 16∶30 and 16∶50. Thus, market organizers should consider excluding this period when constructing the benchmark rate. Second, this paper provides empirical evidence for the importance of improving banks' liquidity management. The empirical results show that banks' intraday liquidity hoarding contributes to the decline in the repo rate between 15∶00 and 16∶30. As the main supplier in the repo market, banks' liquidity management is closely related to the efficiency of this market.
|
Published: 02 September 2023
|
|
|
|
|
|
|