South Korea GZ Corporate Credit-Spread Index

The excess bond premium is the component of corporate bond credit spreads that the issuing firm's measured default risk does not explain. A corporate bond spread compensates investors for expected default loss together with the price of bearing that loss risk, so controlling for measurable default risk isolates its non-default component.

The cross-bond average credit spread decomposes into a component predicted by default risk and bond characteristics and into its residual (Gilchrist and Zakrajšek 2012):

where $\overline{S}_t$ is the cross-bond average credit spread, $\overline{\hat{S}}_t$ is the average spread predicted from the issuing firm's distance to default and bond characteristics, and $\text{EBP}_t$ is the residual excess bond premium.

The excess bond premium captures the effective risk appetite that the corporate credit market prices in. Even with measured default risk unchanged, a rise in the premium means risk appetite has contracted, while a fall means it has expanded.

Sample updates are currently suspended, with the last observation stamped 30 June 2026, and later periods are not yet reflected. Whether and when updates resume is a separate data-pipeline decision.

The corporate-spread decomposition of Gilchrist and Zakrajšek (2012) is applied to the Korean corporate bond market through the following steps.

(1) Cash-flow-matched credit spread. Each senior-unsecured fixed-coupon bond is priced against a synthetic risk-free security with identical cash flows, where those cash flows are discounted on the Korea Treasury Bond zero curve. The cash-flow-matched credit spread is the bond yield less the synthetic yield, and this construction removes the duration-mismatch bias that Gilchrist and Zakrajšek (2012) identify.

(2) Distance-to-default estimation. The firm distance to default reuses the Merton (1974) structural model in the naïve form of Bharath and Shumway (2008) with the default point of Crosbie and Bohn (2003). Vassalou and Xing (2004) motivate the asset-volatility treatment, and Campbell, Hilscher, and Szilagyi (2008) together with Duffie, Saita, and Wang (2007) validate the measure in distress prediction.

(3) Pricing regression. The log credit spread is regressed on the distance to default and on bond characteristics by full-sample ordinary least squares. The bond characteristics are duration, log issue size, coupon, and age, together with a callable indicator interacted with the level, slope, and curvature of the term structure. Industry and credit-rating fixed effects follow Gilchrist and Zakrajšek (2012), and the pricing regression is re-estimated on the full sample at each vintage following the structural-model evidence of Schaefer and Strebulaev (2008).

(4) Excess bond premium computation. The predicted spread applies a lognormal correction that adds one-half the residual variance before exponentiating. The excess bond premium is the cross-bond mean actual spread less the cross-bond mean predicted spread, and its forecasting content is corroborated by Faust et al. (2013) and Gilchrist and Mojon (2018).

(5) Yield input and composite-average recalibration. The corporate yield entering each spread is a single registered pricing agency's per-ISIN evaluated yield rebased onto the composite-average curve, namely the average across the registered bond-pricing agencies. Within each rating-and-maturity cell, each bond's yield is shifted by the gap between the cell's mean yield from that agency and the composite-average curve, so the published level, rating structure, and term structure are the composite average, and that agency identifies the regression while supplying only the within-cell idiosyncratic pricing that washes out of the published aggregates. This is not a pure per-ISIN composite average, because no per-ISIN composite-average history is publicly available. The assumption that the within-cell ordering assigned by that agency proxies the composite average is verified at the current cross-section and assumed stable over history, and the 2023-01-09 change in the composite's published code is notational, where the new code equals the legacy average to the basis point and adds no level discontinuity.

(6) Publication scope. The index is published from 2020-02, the earliest month carrying at least thirty bond-months and fifteen distinct issuers, and the computation frequency is monthly. The regression is kept full-sample, while the thinner earlier months and the investment-grade-only span before BBB issuance enters in 2023 are flagged rather than dropped. The thin BBB tail is decomposed and disclosed by rating without altering the cross-bond aggregate.

A corporate bond credit spread compensates investors for expected default loss together with the price of bearing that risk. The structural credit model formalises the default component by valuing the bond as a contingent claim on firm assets (Merton 1974), and the default barrier of Black and Cox (1976) together with the optimal capital structure of Leland (1994) extend it. The excess bond premium that remains once measured default risk is removed therefore captures the effective risk appetite that the credit market prices in.

Isolating the residual unexplained by measured default risk from individual securities, bottom-up, was established by a sequence of studies. Gilchrist, Yankov, and Zakrajšek (2009) first built the bottom-up spread index from individual securities, Gilchrist and Zakrajšek (2012) formalised it as the excess bond premium, and Favara et al. (2016) restate the published premium for ongoing monitoring. Philippon (2009) adds the q-theoretic reading of bond-market valuations.

The excess bond premium is a leading indicator of real activity. A rise in the premium signals a contraction in the effective supply of credit ahead of slowdowns in output and employment, a predictive relationship that Gilchrist and Zakrajšek (2012), Gertler and Lown (1999), and Mueller (2009) document for the United States and that Gilchrist and Mojon (2018) extend to the euro area. The cross-country evidence of Okimoto and Takaoka (2017) confirms the credit-spread cycle in Japan.

Credit spreads rank among the most informative financial predictors (Faust et al. 2013; Stock and Watson 2003). Elevated credit-market sentiment reverses into weaker growth (López-Salido, Stein, and Zakrajšek 2017), and tighter financial conditions map to the lower quantiles of the future growth distribution (Adrian, Boyarchenko, and Giannone 2019). The financial-conditions-index literature operationalises this link between financial conditions and growth (Hatzius et al. 2010).

The Korean corporate bond market carries comparable signals. Caballero, Fernández, and Park (2019) tie corporate borrowing spreads to emerging-market activity, Lee and Kang (2015) and Kim, Shi, and Kim (2020) build spread-based forecasting and factor-model financial-stress measures, and the Bank of Korea (2022) together with Jeong (2022) assess the short-term funding and credit-market strain of late 2022 around the Legoland event. The Korean excess bond premium therefore complements the recession-probability and financial-conditions gauges already maintained on this platform.

For investors the excess bond premium measures the compensation for bearing corporate credit risk beyond expected default. Intermediary asset-pricing theory ties this compensation to the risk-bearing capacity of financial institutions (He and Krishnamurthy 2013; Adrian, Etula, and Muir 2014; Adrian, Moench, and Shin 2010; He, Kelly, and Manela 2017), and balance-sheet constraints amplify it through the financial-accelerator and funding-liquidity channels (Bernanke, Gertler, and Gilchrist 1999; Brunnermeier and Pedersen 2009).

A large share of credit spreads reflects non-default factors such as risk premia and liquidity rather than default risk (Elton et al. 2001; Collin-Dufresne, Goldstein, and Martin 2001; Huang and Huang 2012; Chen, Collin-Dufresne, and Goldstein 2009; Longstaff, Mithal, and Neis 2005). This non-default component decomposes further into priced default-event risk (Driessen 2005), the link between callable-bond spreads and the term structure (Duffee 1998), and illiquidity (Bao, Pan, and Wang 2011), so the excess bond premium net of measured default risk offers a tradeable signal of credit-market risk appetite that sharpens in crises (Krishnamurthy and Muir 2017).

In the Korean market the default and liquidity drivers of won corporate spreads are documented (Park et al. 2013; Shin and Kim 2015), and the market's behaviour through rising-rate episodes has been analysed (Kim 2022). A widening premium means the compensation for bearing credit risk has grown and a compressing premium means it has shrunk, so the excess bond premium serves as a positioning signal by which bond managers adjust their credit and duration exposure.

Over 118 observations from 2016-03-31 to 2025-12-31, the Gilchrist-Zakrajsek credit spread carries a unit root, with the Dickey and Fuller (1979) test not rejecting at p = 0.193, the Phillips and Perron (1988) test concurring at p = 0.2006, and the Kwiatkowski et al. (1992) test rejecting stationarity, an agreed I(1). On the level the first-order autocorrelation is 0.940 with an implied half-life of about 11.2 observations, a descriptive persistence summary and not an integration claim, the Ljung and Box (1978) portmanteau on the level rejects white noise at lags 12 and 24, Q = 463.15 and Q = 484.22 at p = 0.000 and p = 0.000, and the Bai and Perron (1998) procedure, by the Bai and Perron (2003) algorithm, finds 3 breaks in the mean at 2021-07-31, 2022-12-31, 2024-05-31 (Andrews 1993; Perron 1989).

This signal, the Gilchrist-Zakrajsek credit 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).