South Korea Real Rate Gap

The real rate gap is the difference between the ex-ante real short-term interest rate and the natural rate of interest, measuring the stance of monetary policy relative to the economy's structural equilibrium.

Formally, the real rate gap is defined as:

where

is the real interest rate, the MSB 91-day yield minus expected inflation, which within the HLW model is constructed as a 4-quarter trailing average of core CPI inflation, and

is the natural rate.

The real rate gap directly enters the IS curve of the HLW model as the mechanism through which monetary policy affects the real economy:

The coefficient $a_r < 0$ ensures that a positive real rate gap reduces the output gap, consistent with the standard monetary transmission mechanism.

A positive gap ($r_t > r^*_t$) indicates that monetary policy is tighter than neutral, where the real interest rate exceeds the rate consistent with balanced growth, restraining aggregate demand and putting downward pressure on inflation. A negative gap ($r_t < r^*_t$) indicates accommodative policy.

Estimated jointly with the natural rate of interest ($r^*$, KRNR), trend growth (KRTRGDP), and the output gap (KRGDPGAP) within the HLW (2023) 3-stage MLE state-space model. The full model specification, data inputs, and estimation procedure are described in detail in the KRNR methodology section.

(1) Real rate gap computation. The real rate gap is not a Kalman filter state variable but is computed post-estimation as the difference between the observed real rate and the one-sided filtered natural rate:

where $i_t$ is the MSB 91-day yield (quarterly average of daily observations), $\pi^e_t$ is the 4-quarter trailing average of core CPI inflation (serving as the expected inflation proxy within the HLW framework), $g_t$ is the Kalman-filtered trend growth, and $z_t$ is the filtered residual natural rate component. All components except the nominal interest rate and observed inflation are model outputs.

(2) Expected inflation specification. The HLW model uses backward-looking inflation expectations by design:

that is, a 4-quarter trailing average of past core CPI inflation rates. This is consistent with the original Laubach and Williams (2003) and HLW (2017, 2023) specification, where the IS and Phillips curve equations are jointly estimated under the assumption that agents form expectations by extrapolating recent inflation experience. This differs from KRED's standalone ex-ante real rate series KREARPR, which uses a forward-looking expected inflation estimate combining UCSV trend inflation (Stock and Watson 2007) with BOK Consumer Survey data via Nelson-Siegel. The backward-looking specification in HLW ensures internal consistency within the structural model, where the same $\pi^e_t$ enters both the real rate construction and the Phillips curve, so that the IS-curve parameter $a_r$ is estimated conditional on this specific expectations process.

(3) Estimation uncertainty. The real rate gap combines observable and unobservable components. The nominal rate $i_t$ and inflation $\pi_t$ are observed, but $r^*_t$ is a one-sided Kalman-filter estimate carrying substantial uncertainty. The gap therefore inherits the full estimation uncertainty of the natural rate:

In the committed record the gap at 2026Q1 is 0.37 percentage points and the filtered standard error of $r^*$ at that quarter is 2.18 percentage points, so the gap lies well inside one standard error of zero. The sign and magnitude of the gap should be interpreted in light of this uncertainty, which the filtered standard error understates because it is conditional on the estimated parameters and excludes parameter uncertainty (Hamilton 1986). The first quarters lean on the filter's full-sample start value, and every quarter is revised when the model is re-estimated, so the gap is not a real-time measure (Orphanides and van Norden 2002).

The real rate gap is the HLW framework's measure of the monetary policy stance, addressing the question of whether monetary policy is stimulative or restrictive relative to the economy's equilibrium.

The real rate gap is closely related to the Taylor (1993) rule framework. In a standard Taylor rule, the policy rate responds to the inflation gap and the output gap

Rearranging, the implied real rate gap is

The HLW real rate gap provides a model-based estimate of this deviation, jointly estimated with the output gap and the natural rate, allowing researchers to assess whether the BOK's policy actions have been broadly consistent with a Taylor-type rule or whether systematic deviations have occurred.

The real rate gap provides a narrative lens for Korean monetary policy history. Following the 1997–98 Asian financial crisis, interest rates were sharply elevated (positive real rate gap), contributing to the recession but also stabilizing the currency. The 2001–2005 period saw accommodative policy (negative gap) supporting the recovery. The 2008 global financial crisis led to aggressive easing (large negative gap), followed by normalization. During 2020, the BOK cut rates to 0.50% in response to COVID-19, and the real rate gap indicates whether this was sufficient to achieve an accommodative stance given the simultaneously declining $r^*$. In the committed one-sided record the gap stays positive in every quarter of 2020, so by this measure the cut did not take the stance below the natural rate, although the gap stays within one filtered standard error of zero. The 2022–2023 tightening cycle took the gap from negative values in every quarter of 2022 to positive values from 2023Q4, and it remained positive through 2025.

Prolonged negative real rate gaps can fuel asset price inflation and financial imbalances, since they imply that monetary policy is persistently accommodative relative to the structural equilibrium. In the committed one-sided record the gap is negative in every quarter of 2021. The real rate gap provides a structural framework for assessing this channel, where the duration and magnitude of a negative gap quantifies the degree of monetary accommodation that may feed financial stability risks. This connects to the broader debate on 'leaning against the wind' in monetary policy (Svensson 2017; Adrian and Liang 2018).

Unlike simple comparisons of the nominal policy rate to headline inflation, the real rate gap accounts for the time-varying equilibrium real rate. A 1% real rate may be deeply accommodative when $r^*$ is 2%, but restrictive when $r^*$ has fallen to 0%. This distinction matters for Korea, where KRED's one-sided HLW record puts the natural rate at 3.86% at 2001Q4 and 0.11% at 2026Q1, with filtered standard errors of 1.69 and 2.18 percentage points. KRRRGAP measures the policy stance within the HLW model's internal logic, while KREARPR is a model-derived real rate whose expectation leg is forward-looking, built on UCSV-based expected inflation rather than the HLW's backward-looking moving average. For a model-derived real rate with a forward-looking expectation leg, see KREARPR.

The real rate gap is a powerful indicator for fixed income strategy, FX positioning, and cross-asset allocation.

The real rate gap is a leading indicator of future monetary policy actions. Large positive gaps (tight policy) indicate the BOK has room to cut rates, as the current policy stance is more restrictive than neutral. This is bullish for short-duration KTB positions and IRS receivers. Negative gaps suggest eventual tightening, supporting short KTB positions and IRS payers. The gap's predictive power for BOK actions can be enhanced by combining it with the implied rate path from the KRED implied rate series, which captures market pricing of the next three Monetary Policy Committee meetings.

The real rate gap primarily affects the short end of the yield curve, as it measures the distance between the current policy rate and its equilibrium. When the gap is wide and positive, the front end is expected to rally as the BOK eases, favoring curve steepeners (long front-end, short back-end). When the gap is negative, the front end is expected to sell off as the BOK normalizes, favoring curve flatteners. The signal is most useful when combined with the output gap (KRGDPGAP) and the term premium (KRTP series) to distinguish between front-end-driven and back-end-driven curve moves.

The real rate gap differential between Korea and other economies is a structural determinant of KRW dynamics. A positive Korean real rate gap relative to the U.S. (Korean policy is tighter than equilibrium relative to U.S. policy stance) tends to support KRW through capital inflows attracted by above-equilibrium yields. Conversely, a negative differential makes KRW carry less attractive. For carry trade strategies, the real rate gap is more informative than the raw interest rate differential because it adjusts for the structural equilibrium. A high Korean rate that is below $r^*$ (negative gap) may not be sustainable, while a moderate rate that is above $r^*$ (positive gap) represents a durable carry advantage.

In Korea, the real rate gap is closely correlated with the housing price cycle. Persistently negative gaps have historically coincided with housing market overheating, particularly in the Seoul metropolitan area, as accommodative monetary conditions drive leveraged real estate investment. Fixed income investors can use the real rate gap as an input for assessing macro-prudential policy risk. When the gap has been deeply negative for an extended period and housing prices are rising, the probability of macro-prudential tightening (LTV/DTI restrictions, targeted taxes) increases, which can independently affect the interest rate environment and credit conditions.

KRRRGAP is the real-rate gap of KRED's quarterly Holston-Laubach-Williams model on Korean data, the model's observed ex-ante real rate minus its one-sided Kalman-filtered natural rate and never a gap to a smoothed path, so the statistical object is a persistence summary of that series and not an integration order of the data. The gap inherits the natural rate, whose integration order is imposed by the random-walk state equation rather than estimated, and a one-sided filter at steady state is an exponentially weighted one-sided average whose gain, set by the median-unbiased signal-to-noise ratios, dominates the serial correlation of the filtered leg.

The augmented Dickey and Fuller (1979) regression with the lag augmentation of Said and Dickey (1984), the nonparametric Phillips and Perron (1988) test and the KPSS stationarity test of Kwiatkowski et al. (1992) were deliberately not run on this path, because a verdict from them would re-import the very inference the model was built to avoid, the signal-to-noise ratios having been fixed by the median-unbiased procedure of Stock and Watson (1998) built on the exponential Wald statistic of Andrews and Ploberger (1994). The GLS-detrended test of Elliott, Rothenberg, and Stock (1996) was not run for the same reason, and neither was the modified-statistic battery of Ng and Perron (2001), since both confirm an integration reading this archetype does not take. The level break search of Bai and Perron (1998) was likewise not run, because a break in the mean of a filtered latent state reflects the filter's response to its inputs rather than a change in the economy (Perron 1989). A level portmanteau such as Ljung and Box (1978) is not applied either, since on a filtered path it reads the filter gain rather than evidence about the data-generating process.

What the matrix mandates in their place is a descriptive persistence summary and a statement about revision. Over the 117 quarters from 1997-03-31 to 2026-03-31 the first-order autoregressive coefficient of the gap is 0.790 and the implied half-life is 2.94 quarters, so the gap decays far faster than the near-random-walk natural rate it is measured against, a property of this series and not an order of integration, read with the small-sample downward bias of autoregressive estimates in mind (Andrews 1993). The revision cell records that the signal-to-noise ratios and the model parameters are re-estimated at every update, the parameters by a seeded basin-hopping search, and that each load rewrites the whole stored history, so the gap inherits every revision of the filtered natural rate although no smoothing enters it, while its observed real-rate leg changes only when the input data are revised, and no dated vintage panel exists from which the size of those revisions could be measured (Orphanides and van Norden 2002).

Three limits are stated rather than implied. No order of integration is assigned to this series, by ruling and not by an inconclusive test, so nothing here is evidence that the gap does or does not carry a stochastic trend (Hamilton 2018). The gap is one-sided but it is not a real-time estimate, since the natural rate it subtracts rests on parameters, signal-to-noise ratios and a start value, a Hodrick and Prescott (1997) trend of log GDP, that are all estimated on the full sample, which makes it a quasi-final estimate in the sense of Orphanides and van Norden (2002). And the frequency is quarterly, so the half-life above is measured in quarters and is not comparable with a half-life measured at another frequency (Holston, Laubach, and Williams 2023; Laubach and Williams 2003; Hamilton 1986).