---
ticker: "KRCGAP"
title: "South Korea Credit-to-GDP Gap"
unit: "pp"
frequency: "Quarterly"
source: "Borio and Lowe (2002); Drehmann, Borio, Gambacorta, Jiménez, and Trucharte (2010); Bank for International Settlements (2016)"
release: "Updated Quarterly"
category: "Credit & Financial Cycle"
country: "KR"
language: "en"
canonical: "https://kred.dev/en/series/KRCGAP"
license: "https://creativecommons.org/licenses/by-nc-nd/4.0/"
latest_value: -8.00
latest_date: "2025-12-31"
first_date: "1972-09-30"
observations_total: 214
observations_shown: 120
---

# South Korea Credit-to-GDP Gap

## Overview

A financial-cycle gauge of how far private credit relative to GDP has strayed from trend, flagging credit booms and imbalances.

## Key Figures

|  | Value | Date |
|---|---|---|
| Latest | -8.00 | 2025-12-31 |
| Change from previous | -1.54 | 2025-09-30 |
| Change over one year | -3.12 | 2024-12-31 |
| Highest on record | 15.63 | 2021-03-31 |
| Lowest on record | -20.99 | 2004-12-31 |
| Period covered | 1972-09-30 – 2025-12-31 |  |
| Observations | 214 |  |

## Recent observations

| Date | Value | Change |
|---|---|---|
| 1996-03-31 | 0.47 | +0.33 |
| 1996-06-30 | 1.91 | +1.44 |
| 1996-09-30 | 4.97 | +3.06 |
| 1996-12-31 | 4.46 | -0.51 |
| 1997-03-31 | 8.12 | +3.67 |
| 1997-06-30 | 7.51 | -0.61 |
| 1997-09-30 | 7.77 | +0.26 |
| 1997-12-31 | 12.73 | +4.96 |
| 1998-03-31 | 9.74 | -2.99 |
| 1998-06-30 | 7.85 | -1.89 |
| 1998-09-30 | 10.02 | +2.17 |
| 1998-12-31 | 5.34 | -4.68 |
| 1999-03-31 | 9.14 | +3.79 |
| 1999-06-30 | 0.33 | -8.81 |
| 1999-09-30 | -4.13 | -4.45 |
| 1999-12-31 | -12.68 | -8.55 |
| 2000-03-31 | -13.18 | -0.50 |
| 2000-06-30 | -16.38 | -3.20 |
| 2000-09-30 | -16.63 | -0.25 |
| 2000-12-31 | -19.21 | -2.58 |
| 2001-03-31 | -18.21 | +1.00 |
| 2001-06-30 | -18.47 | -0.26 |
| 2001-09-30 | -16.42 | +2.05 |
| 2001-12-31 | -17.48 | -1.06 |
| 2002-03-31 | -14.77 | +2.71 |
| 2002-06-30 | -12.75 | +2.02 |
| 2002-09-30 | -11.39 | +1.36 |
| 2002-12-31 | -11.50 | -0.11 |
| 2003-03-31 | -11.59 | -0.09 |
| 2003-06-30 | -12.51 | -0.92 |
| 2003-09-30 | -13.36 | -0.85 |
| 2003-12-31 | -14.43 | -1.07 |
| 2004-03-31 | -15.98 | -1.55 |
| 2004-06-30 | -18.08 | -2.10 |
| 2004-09-30 | -18.81 | -0.73 |
| 2004-12-31 | -20.99 | -2.18 |
| 2005-03-31 | -20.94 | +0.05 |
| 2005-06-30 | -18.77 | +2.16 |
| 2005-09-30 | -18.02 | +0.75 |
| 2005-12-31 | -16.72 | +1.30 |
| 2006-03-31 | -15.27 | +1.45 |
| 2006-06-30 | -11.51 | +3.76 |
| 2006-09-30 | -8.70 | +2.81 |
| 2006-12-31 | -5.46 | +3.24 |
| 2007-03-31 | -3.72 | +1.75 |
| 2007-06-30 | -2.34 | +1.37 |
| 2007-09-30 | -1.70 | +0.64 |
| 2007-12-31 | -0.72 | +0.99 |
| 2008-03-31 | 2.20 | +2.92 |
| 2008-06-30 | 5.14 | +2.94 |
| 2008-09-30 | 7.44 | +2.30 |
| 2008-12-31 | 9.99 | +2.54 |
| 2009-03-31 | 11.15 | +1.16 |
| 2009-06-30 | 12.41 | +1.27 |
| 2009-09-30 | 12.37 | -0.05 |
| 2009-12-31 | 10.69 | -1.67 |
| 2010-03-31 | 7.77 | -2.92 |
| 2010-06-30 | 6.42 | -1.36 |
| 2010-09-30 | 4.94 | -1.48 |
| 2010-12-31 | 2.68 | -2.26 |
| 2011-03-31 | 2.34 | -0.34 |
| 2011-06-30 | 2.77 | +0.43 |
| 2011-09-30 | 4.59 | +1.82 |
| 2011-12-31 | 3.75 | -0.84 |
| 2012-03-31 | 3.42 | -0.33 |
| 2012-06-30 | 3.77 | +0.35 |
| 2012-09-30 | 4.47 | +0.70 |
| 2012-12-31 | 2.86 | -1.61 |
| 2013-03-31 | 3.97 | +1.12 |
| 2013-06-30 | 3.32 | -0.65 |
| 2013-09-30 | 2.23 | -1.09 |
| 2013-12-31 | 0.63 | -1.60 |
| 2014-03-31 | 0.24 | -0.38 |
| 2014-06-30 | -0.21 | -0.46 |
| 2014-09-30 | 1.06 | +1.27 |
| 2014-12-31 | 1.12 | +0.06 |
| 2015-03-31 | 0.42 | -0.70 |
| 2015-06-30 | -0.14 | -0.56 |
| 2015-09-30 | 0.27 | +0.41 |
| 2015-12-31 | -0.19 | -0.46 |
| 2016-03-31 | -1.01 | -0.81 |
| 2016-06-30 | -1.21 | -0.20 |
| 2016-09-30 | -1.77 | -0.57 |
| 2016-12-31 | -1.83 | -0.06 |
| 2017-03-31 | -2.93 | -1.09 |
| 2017-06-30 | -2.07 | +0.86 |
| 2017-09-30 | -2.67 | -0.60 |
| 2017-12-31 | -3.42 | -0.75 |
| 2018-03-31 | -2.99 | +0.43 |
| 2018-06-30 | -2.40 | +0.60 |
| 2018-09-30 | -1.07 | +1.33 |
| 2018-12-31 | -0.57 | +0.50 |
| 2019-03-31 | 0.09 | +0.66 |
| 2019-06-30 | 1.94 | +1.85 |
| 2019-09-30 | 3.41 | +1.47 |
| 2019-12-31 | 4.60 | +1.19 |
| 2020-03-31 | 7.04 | +2.44 |
| 2020-06-30 | 11.23 | +4.19 |
| 2020-09-30 | 14.04 | +2.81 |
| 2020-12-31 | 15.06 | +1.03 |
| 2021-03-31 | 15.63 | +0.57 |
| 2021-06-30 | 15.39 | -0.24 |
| 2021-09-30 | 15.04 | -0.35 |
| 2021-12-31 | 13.93 | -1.11 |
| 2022-03-31 | 11.82 | -2.11 |
| 2022-06-30 | 11.80 | -0.03 |
| 2022-09-30 | 12.32 | +0.52 |
| 2022-12-31 | 11.57 | -0.75 |
| 2023-03-31 | 9.61 | -1.96 |
| 2023-06-30 | 8.12 | -1.49 |
| 2023-09-30 | 7.27 | -0.85 |
| 2023-12-31 | 3.52 | -3.75 |
| 2024-03-31 | 0.17 | -3.36 |
| 2024-06-30 | -1.97 | -2.14 |
| 2024-09-30 | -3.32 | -1.35 |
| 2024-12-31 | -4.88 | -1.56 |
| 2025-03-31 | -4.93 | -0.05 |
| 2025-06-30 | -5.72 | -0.79 |
| 2025-09-30 | -6.46 | -0.74 |
| 2025-12-31 | -8.00 | -1.54 |

## Definition

The credit-to-GDP gap measures, in percentage points of GDP, how far a country's ratio of total private non-financial credit to GDP has departed from its own long-run trend. The object of inference is not the level of the ratio but its deviation from trend, defined as the ratio $r_t$ at time $t$ less the long-run trend $\text{trend}_t$ at the same date:

$$\text{gap}_t = r_t - \text{trend}_t$$

This gap is the financial-cycle indicator that Borio and Lowe (2002) proposed and that Basel III adopted as the reference indicator for the countercyclical capital buffer (CCyB).

The input is the BIS-published Korean total private non-financial credit-to-GDP ratio (adjusted for breaks, in percent of GDP). This series takes the published ratio as is rather than reconstructing it from a credit stock and a GDP series, which is what makes it match the Korean credit-to-GDP gap that BIS itself computes and publishes to within a basis point on the common sample. The BIS-published gap is computed under a single cross-country procedure, so it can differ from the CCyB gauges that national supervisors define and operate, and this series follows the BIS definition.

A large positive gap points to credit expanding above trend, while a large negative gap points to a contraction or de-leveraging phase in which credit runs below trend.

## Methodology

The gap is computed in four steps that apply a one-sided (recursive, expanding-window) Hodrick–Prescott filter to the total private non-financial credit-to-GDP ratio.

**(1) Input.** The total private non-financial credit-to-GDP ratio $r_t$ is used as the input without transformation. Because the ratio is already divided by a four-quarter rolling GDP sum, which removes GDP seasonality, and credit is measured as an end-of-quarter stock, no separate preprocessing such as seasonal adjustment is applied.

**(2) One-sided HP trend.** At each quarter $t$ a standard two-sided HP filter with $\lambda=400{,}000$ is fit on the sub-sample $r_1, \ldots, r_t$, and only the trend value at the endpoint is retained as $\text{trend}_t$. The smoothing constant $\lambda=400{,}000$ is the Basel standard for quarterly data, documented in Drehmann et al. (2010) and in the BIS (2016) September Quarterly Review enhancement note.

**(3) Gap computation.** The gap is computed from the adopted trend:

$$\text{gap}_t = r_t - \text{trend}_t,$$

in percentage points of GDP. The burn-in is 40 quarters (roughly ten years), and both gap and trend are left missing for the quarters before that window. Because the input credit-to-GDP ratio starts in 1962 Q4, the gap is available from 1972 Q4 onward.

**(4) Real-time re-estimation.** The defining property of the one-sided filter is that the trend value at $t$ depends only on data through $t$, so the trend itself never revises after the fact. When the series is re-estimated each quarter, the only reason historical trend and gap values change is a revision to the BIS input ratio. Because BIS periodically revises the input ratio for credit-stock break adjustments and GDP rebasing, this module recomputes the full sample on every pipeline run and replaces the entire DB table.

## Applications in Economics

The credit-to-GDP gap is a simple but powerful summary statistic of the financial cycle. Borio and Lowe (2002) show that when a credit build-up runs far above trend while co-moving with rising asset prices, the probability of a financial crisis over the following five years rises to a statistically meaningful degree, and Drehmann et al. (2010) confirm in out-of-sample panel work that the credit-to-GDP gap outperforms any other single indicator in crisis prediction. The value of the gauge lies not in pin-point timing of crises but in summarizing in one number the fact that systemic risk accumulates non-linearly the longer a credit build-up runs well above trend.

Historically, credit booms are a reliable leading signal of financial crises and a harbinger of deeper and longer recessions. Schularick and Taylor (2012) document on long-run cross-country data that credit booms are the single most reliable leading indicator of financial crises, and Jordà, Schularick, and Taylor (2013) show that recessions preceded by credit booms are systematically deeper and longer than ordinary contractions. Aikman, Haldane, and Nelson (2015) take this cross-country evidence into the macroprudential frame and argue that buffering against the credit cycle, rather than chasing it with interest-rate policy alone, is the cheapest insurance against the deep tails. KRCGAP enters this literature as the Korean instance of the Basel III countercyclical-capital-buffer reference indicator that Drehmann et al. (2010) formalized.

Applied to Korea, the gap entered clearly positive territory at each major crisis. It moved quickly into positive territory ahead of the 1997 Asian financial crisis, recorded a large positive value again through the 2008 global financial crisis, and entered notably positive territory during the 2020 COVID-era surge in household and corporate borrowing. The short-term-credit-market stress triggered by the Gangwon Legoland project-finance default in October 2022, documented qualitatively in the Bank of Korea Financial Stability Reports, sits in the descending phase of this cycle. The latest print is back in negative territory. This means the Korean economy is in a de-leveraging phase, since the BIS total private non-financial credit-to-GDP ratio that KRCGAP uses (the BIS aggregate of households, non-financial corporates, and non-profit institutions serving households) remains close to 200 % of GDP on the BIS Total Credit Statistics while credit is growing more slowly than trend.

Two cautions matter when reading the credit-to-GDP gap. First, the indicator measures a deviation from trend rather than a level, so even an economy with a structurally elevated credit-to-GDP ratio like Korea can show a negative gap. Second, the one-sided HP trend gradually adapts as the sample lengthens. A long stretch of positive gaps eventually pulls the trend up and compresses the gap signal, an adaptation that Drehmann et al. (2010) discuss explicitly. Hamilton (2018) raises a more general critique of HP-filter applications, pointing out that filter endpoints and trend definitions are arbitrary and that the resulting cycles can be artifacts of the filter itself. The Basel convention of $\lambda = 400{,}000$ and recursive one-sided application is a policy choice that addresses both the cyclical-frequency and look-ahead concerns, even if the broader HP-filter critique itself remains open.

The BIS standard gap and national supervisors' domestic gauges can differ in their credit definition and their trend procedure. Drehmann and Tsatsaronis (2014) document that national supervisors can and do construct domestic CCyB-decision gauges that differ from the BIS standard in the perimeter of the credit definition (which of households, non-financial corporates, and non-profits to include) and in the trend procedure, and the Basel III text is explicit that reading the BIS gap is a starting point for supervisory judgment rather than a mechanical trigger. The KRCGAP series follows the BIS definition and reads alongside, not in place of, the domestic supervisory gauges of the Bank of Korea and the Financial Services Commission.

## Applications in Financial Markets

This series is the first-order variable the Bank of Korea and the Financial Services Commission consult when reviewing whether to activate the CCyB, and it is a natural monitoring target for the fixed-income market and for bank capital management. Under the Basel Committee on Banking Supervision (2010) framework, a phase in which the gap accumulates in positive territory above 2 percentage points reads as a signal that credit conditions are easing alongside rising risk-weighted assets at financial institutions, and a move toward 10 percentage points indicates a credit expansion large enough that, under the Basel III guidance, the CCyB reaches its maximum (typically 2.5 % of risk-weighted assets). Conversely, a deeply negative gap suggests that the credit cycle is in a contractionary phase and that macroprudential tools have room to ease.

The exact match to the BIS-published figure has operational significance for external policy communication. International institutions and foreign investors use exactly the same BIS figure when assessing Korea's position in the financial cycle, and the BIS (2016) enhancement note is the standard source for the definition they are reading. The Bank of Korea's separately defined and published domestic CCyB gauge differs in its input credit definition (the perimeter of households, corporates, and non-profits) and in its trend procedure, so there are periods when it sits at a different level from this series (Drehmann and Tsatsaronis 2014). That this series follows the BIS definition must be acknowledged as such in use.

For the rates and credit desks, this gauge reads as a low-frequency overlay on the cyclical credit positioning that drives Korean corporate and bank-funding spreads. A persistent positive gap reads through the Schularick and Taylor (2012) evidence as a phase of rising deep-tail risk to sovereign and financial credit, while a deeply negative gap points to a credit-cycle phase that has historically been the better entry point for cyclically sensitive credit allocation. The 2022 Legoland-PF episode is a reminder that short-dated funding markets can transmit stress even when the gap looks benign. KRCGAP therefore sits in the dashboard as the structural credit-cycle anchor placed over the higher-frequency stress channels (KRFCI, KRCP, KRSAHM, KRADS).

The pipeline re-estimates the entire sample right after each BIS release, so historical trend and gap values update in response to revisions in the BIS input ratio. There is no trend revision arising from the one-sided HP filter itself, which is the operational advantage of the one-sided filter over the two-sided HP filter that Hamilton (2018) critiques. The indicator can therefore be used as is as a real-time gauge suited to the real-time decision horizon of macroprudential policy. The Aikman, Haldane, and Nelson (2015) framing of macroprudential buffering as the cheapest insurance against the deep tails is the reason KRCGAP is placed beside the higher-frequency stress channels, not in place of them.

## Statistical Tests

KRCGAP is a one-sided Hodrick and Prescott (1997) credit-to-GDP gap with the smoothing parameter set to 400000, built recursively through each quarter from contemporaneously available data, so it is a stationary-by-construction object whose integration order is fixed by the filter rather than estimated from the data, the same credit-cycle gap used for early warning by Borio and Lowe (2002) and Drehmann and Juselius (2012). The augmented Dickey-Fuller, Phillips and Perron (1988), and KPSS (Kwiatkowski et al. 1992) unit-root battery, together with the DF-GLS test of Elliott, Rothenberg, and Stock (1996) and the Ng and Perron (2001) refinement, was deliberately not run, because an HP cycle of a driftless random walk is judged stationary by the Dickey and Fuller (1979) test in a large majority of replications, so a rejection would confirm the filter rather than discover mean reversion (Hamilton 2018; Hodrick and Prescott 1997).

As a stationary object the gap is summarized by its persistence and amplitude. The first-order autoregressive coefficient of the gap is 0.977 over 213 quarterly observations from 1972-09-30 to 2025-09-30, implying a half-life of about 29.4 quarters, a figure reported as a descriptive property of the constructed gap and not as an order of integration and read with the near-unity small-sample downward bias of Andrews (1993) in mind, since a level portmanteau such as Ljung and Box (1978) would read the filter gain rather than the data-generating process. The gap has a standard deviation of 9.01 percentage points and ranges from −20.99 to 15.63 percentage points.

The Bai and Perron (1998, 2003) multiple structural-break procedure, admissible here precisely because the gap is stationary by construction rather than an I(1) level, locates four breaks in the mean at 1980-06-30, 1988-03-31, 1999-09-30, and 2007-06-30. The gap is non-revising by construction, as the one-sided trend at each quarter uses only contemporaneously available data, so a vintage exercise across three committed quarterly vintages returns a gap at a fixed past quarter that is identical across vintages, a revision of 0.0000 percentage points, and the reading carries no end-point-uncertainty qualifier (Orphanides and van Norden 2002).

## Frequently Asked Questions

### What is the credit-to-GDP gap?

How far the ratio of total credit to the private non-financial sector over GDP departs from its own long-run trend, in percentage points of GDP. The object of interest is the deviation from trend rather than the level of the ratio.

### Why is the credit-to-GDP trend estimated with a one-sided filter?

A two-sided filter uses future observations, which makes past signals look far clearer in hindsight than they were in real time. One-sided estimation uses only what was knowable at each date and avoids that distortion.

### Does a large credit-to-GDP gap mean a correction is coming?

No. It is a descriptive imbalance measure stating that credit has grown faster than output relative to its own history, and it says nothing about whether or when a correction follows.