Call Rate − KOFR Spread

The call − KOFR spread is the difference between the unsecured overnight call rate and the Korea Overnight Financing Rate (KOFR), capturing the secured–unsecured funding basis in the Korean overnight money market. The call rate is the volume-weighted average rate on unsecured overnight interbank lending, while KOFR is the volume-weighted trimmed-mean rate (excluding the top and bottom 5% of transactions) on overnight repos collateralized by government bonds and Monetary Stabilization Bonds, designated as Korea’s risk-free reference rate under the 2020 Act on the Management of Financial Benchmarks (Bank of Korea 2021).

Under normal conditions the spread is small, typically a few basis points, reflecting benign interbank credit conditions. A widening spread tends to coincide with funding stress, a repricing of counterparty credit risk, or segmentation in short-term funding markets.

Computed as the daily spread of the call rate minus KOFR, with the period prior to KOFR observation supplemented by a backcast.

(1) Spread Computation. The two series are inner-joined on their common observation date range, and the simple difference $\text{Call Rate}_t - \text{KOFR}_t$ is computed.

(2) KOFR Backcast. Because KOFR was officially published beginning November 2021, values for the prior period are backcasted to 2006 via a spread-based OLS regression with Newey and West (1987) HAC standard errors and automatic bandwidth selection following Andrews (1991). The backcast regression uses five explanatory spreads against the base rate (call rate, MSB 91-day, CD 91-day, KTB 6-month, and KTB 1-year) and achieves $R^2 = 0.72$ with out-of-sample RMSE $= 3.3$ bp. The final KOFR series concatenates the backcasted segment (2006–2021) with the observed segment (2021–present).

The call − KOFR spread captures the same economic object in the Korean overnight market as the LIBOR–OIS spread in the United States or the €STR-related spreads in the euro area. Taylor and Williams (2009) provided the first systematic documentation that the sharp widening of secured–unsecured overnight spreads during the 2007–2008 global financial crisis reflected interbank funding stress in real time, and established this class of spread as a core indicator of money-market dislocation. Brunnermeier (2009) situates such spread widenings within the broader framework of mutually reinforcing funding-liquidity and market-liquidity spirals, analyzing how interbank credit freezes combine with collateral-value declines and margin-call spirals to amplify crises.

Decomposing the spread’s components clarifies its information content. Michaud and Upper (2008) showed that interbank rate spreads decompose into credit risk and liquidity risk components. Gefang, Koop, and Potter (2011) demonstrated via a dynamic factor model that short-maturity spreads such as the overnight basis are driven primarily by liquidity risk, while longer-maturity spreads load more heavily on credit risk. Schwarz (2019) carried out a more rigorous identification of the credit-versus-liquidity decomposition, confirming that the two channels contribute differently under different market conditions. In the Korean context the call rate is the unsecured interbank overnight lending rate and KOFR is the secured repo rate on government bonds and MSBs, so the spread captures the pure overnight secured–unsecured funding basis.

The call − KOFR spread serves as a real-time indicator of Korean interbank credit conditions and short-end money-market functioning, extending the use of the LIBOR–OIS spread that Sengupta and Tam (2008) propose as a summary indicator of money-market health. Because secured–unsecured spread widenings coincide with episodes of interbank stress (Taylor and Williams 2009), real-time observation of this spread contributes to early detection of funding-market dislocations.

The theoretical underpinning for monitoring this spread draws on the funding liquidity framework of Brunnermeier and Pedersen (2009), who demonstrate that deteriorating funding conditions in overnight markets propagate to broader asset markets through margin spirals and liquidity dry-ups. In the Korean context, Afonso, Kovner, and Schoar (2011) provide the empirical template by showing that during periods of stress the interbank market shifts from random matching to directed lending, with counterparty-risk-sensitive lenders withdrawing from unsecured lending entirely. A widening call − KOFR spread reflects precisely this behavioral shift in the Korean overnight market.

Unlike the term LIBOR–OIS spread, the call − KOFR spread is purely overnight, isolating same-day funding frictions without term-premium contamination. This property makes it complementary to the term premium and shadow rate indicators available elsewhere on this site, allowing simultaneous observation of term-structure decompositions and overnight money-market credit conditions. Gorton and Metrick (2012) argued that secured–unsecured spreads in overnight markets are the earliest signals of systemic funding runs, preceding broader credit-market dislocations by days or weeks.

For financial institutions’ liquidity risk management, the historical distribution of this spread provides an empirically grounded benchmark for calibrating overnight funding-cost assumptions under stress scenarios. Copeland, Martin, and Walker (2014) showed that the microstructure of secured funding markets, including the distribution of haircuts and counterparty concentration, amplifies spread dynamics beyond what aggregate data would suggest, reinforcing the value of high-frequency spread monitoring for risk managers operating in the Korean money market.

Over 4876 observations from 2006-09-25 to 2026-06-09, the unit-root reading of the call-minus-KOFR unsecured-secured overnight funding spread is ambiguous, since the Dickey and Fuller (1979) and Phillips and Perron (1988) tests reject a unit root at p = 0.0 and p = 0.0 while the Kwiatkowski et al. (1992) test also rejects stationarity at p < 0.01, the near-unit-root disagreement of a persistent bounded signal, so no clean order is assigned. On the level the first-order autocorrelation is 0.681 with an implied half-life of about 1.8 observations, a descriptive persistence summary and not an integration claim, the Ljung and Box (1978) portmanteau on the level rejects white noise at lags 10 and 20, Q = 9308.38 and Q = 14144.90 at p = 0.000 and p = 0.000, and the Bai and Perron (1998) procedure, by the Bai and Perron (2003) algorithm, finds 4 breaks in the mean at 2010-04-06, 2017-03-29, 2020-03-20, 2023-04-24 (Andrews 1993; Perron 1989).

This signal, the call-minus-KOFR unsecured-secured overnight funding spread, has no daily seasonal period, so the seasonal-unit-root and seasonal-stationarity machinery of Hylleberg et al. (1990) and Canova and Hansen (1995) is deliberately not run (Beaulieu and Miron 1992; Ghysels and Osborn 2001).