South Korea Ex-Ante Real Policy Rate (Base Rate)

The ex-ante real policy rate is the Bank of Korea's base rate (the 7-day repurchase rate) minus 1-year expected inflation, representing the expected real purchasing power of the policy rate at the time of the rate decision. As a forward-looking measure, it captures the real monetary policy stance as perceived by economic agents when they make consumption, investment, and saving decisions.

The conceptual foundation lies in the Fisher (1930) equation:

where $i$ is the nominal policy rate and $\pi^e$ is the expected rate of inflation. Since the policy rate is an overnight rate with no meaningful term premium, the Fisher equation can be applied directly without the need for affine term structure decomposition, unlike longer-maturity real rates that require separation of risk premiums.

The choice of the 1-year expected inflation as the deflator is economically motivated, because the BOK's Monetary Policy Committee explicitly targets the 1–2 year inflation outlook in its forward-looking policy framework. The ex-ante real policy rate is therefore the variable most directly relevant to assessing whether the current policy setting is consistent with the central bank's inflation and output objectives.

Two core CPI deflators are provided. The primary (default) uses the DB index (excluding food and energy), aligned with the international standard adopted by the Cleveland Fed, ECB, and OECD. The secondary (hidden) uses the QB index (excluding agricultural products and petroleum), the traditional Korean measure. Both share the same policy rate and differ only in the expected inflation component.

A higher ex-ante real policy rate implies a tighter monetary policy stance, while a lower or negative value implies an easier one.

Computed via the Fisher (1930) equation:

where $i^{base}_t$ is the Bank of Korea base rate (7-day repo rate) and $\pi^{short}_t$ is the 1-year expected inflation rate. Since the policy rate is an overnight rate with no term premium, ACM decomposition is unnecessary and the Fisher equation is applied directly.

$\pi^{short}_t$ (1-year expected inflation) is run separately on each CPI measure, DB core, QB core and headline, and estimated in two stages.

(1) Trend inflation extraction. Trend inflation is extracted with the UCSV model of Stock and Watson (2007) from the year-on-year inflation of the variant's own CPI measure, DB core CPI for the default line and QB core or headline CPI for the hidden lines. The trend is a one-sided expanding-window estimate that uses inflation data only up to each month, but its volatility prior is a fixed calibration chosen on the full sample, so the expected-inflation leg is a quasi-real-time model estimate, not a real-time one (Orphanides and van Norden 2002).

Stochastic volatility applies to both observation and trend innovations, and the precision-based Gibbs sampler of Chan and Jeliazkov (2009) draws 2,000 burn-in and 5,000 posterior samples, cycling 5 blocks (trend states, observation SV, trend SV, mixture indicators, innovation variances) via $O(T)$ tridiagonal Cholesky.

(2) Short-term expectation estimation. Short-term expectations combine the UCSV trend with the BOK Consumer Survey, the one household survey common to all variants, via an expanding-window OLS whose target is the same CPI measure's realized inflation twelve months ahead.

The Monetary Policy Committee sets rates based on the 1–2 year inflation outlook, making the 1-year expected inflation the economically appropriate deflator. The DB-based result is the default line, and the QB-based and headline results are hidden alternatives.

The ex-ante real policy rate is the single most important indicator for assessing the intended stance of monetary policy. In the Taylor (1993) rule framework, the optimal policy rate responds to the inflation gap ($\pi - \pi^*$) and the output gap ($y - y^*$), and comparing the implied ex-ante real rate with estimates of the natural rate of interest ($r^*$) is central to judging whether policy is contractionary, neutral, or accommodative.

Woodford (2003) establishes the theoretical foundation in the New Keynesian framework. The welfare-relevant monetary policy stance is determined not by the nominal rate per se, but by the deviation of the real interest rate from the natural rate. When $r^{ex\text{-}ante} > r^*$, the intertemporal substitution effect reduces current demand relative to potential, creating downward pressure on inflation. When $r^{ex\text{-}ante} < r^*$, policy is stimulative.

The distinction between the ex-ante and ex-post real policy rate (KREPRPR) is substantively important. The ex-ante rate reflects agents' expectations at the time of decision-making and is the theoretically appropriate variable in New Keynesian models and Taylor-type policy rules. The ex-post rate, computed using realized inflation, measures the actual real return experienced after the fact and can diverge significantly from the ex-ante rate during periods of inflation surprises, such as the energy price shocks of 2022–2023.

Empirically, estimating $r^*$ is challenging because it is unobservable and time-varying. For the U.S., Laubach and Williams (2003) estimate $r^*$ using a Kalman filter embedded in a small structural model. For Korea, similar exercises (incorporating Korea-specific factors such as demographic aging, potential GDP growth deceleration, and the high household debt burden) suggest a declining natural rate trajectory, implying that a given nominal policy rate is more restrictive today than it would have been a decade ago.

The ex-ante real policy rate is also informative for assessing the monetary policy transmission mechanism. In economies with well-anchored inflation expectations, changes in the nominal policy rate translate approximately one-for-one into changes in the real rate, giving the central bank effective control over real economic conditions. When expectations become unanchored (as can occur during stagflationary episodes), the link between nominal and real rates weakens, complicating monetary policy implementation (Blanchard 2023).

International comparisons of ex-ante real policy rates provide insight into relative monetary policy stances across countries. For instance, comparing the Korean ex-ante real rate with those of the U.S., Japan, and other Asian economies helps assess whether Korean monetary policy is relatively tight or loose by global standards, with implications for capital flows, exchange rates, and competitiveness.

The ex-ante real policy rate anchors the short end of the real yield curve and has immediate implications for money market instruments, short-duration bonds, and floating-rate products.

For fixed income investors, the ex-ante real policy rate determines the real carry on short-term instruments. When the real rate is positive, cash and money market funds generate a positive real return, reducing the opportunity cost of maintaining low-duration positions. When the real rate is negative, investors are effectively penalized for holding short-duration assets, creating incentives to extend duration or take on credit risk in search of positive real returns, a dynamic that Rajan (2005) warned could lead to excessive risk-taking.

The spread between the ex-ante and ex-post real policy rates (KREARPR versus KREPRPR) is a direct gauge of the degree to which inflation expectations are anchored to realized price dynamics, although its expectation leg is a quasi-real-time model estimate rather than a real-time one. A consistently narrow spread indicates well-anchored expectations, while a widening spread signals potential dis-anchoring, an early warning for bond portfolio managers who may need to adjust inflation hedging strategies.

In the credit markets, the real policy rate affects corporate borrowing costs and hence credit fundamentals. When real rates are high, debt servicing costs for leveraged corporations increase in real terms, elevating default risk. Monitoring the ex-ante real rate trajectory is therefore relevant for credit analysts assessing the sustainability of Korean corporate debt burdens, particularly in rate-sensitive sectors such as real estate and construction.

This is a model-derived series, the ex-ante real policy rate from a Fisher decomposition whose expected-inflation leg is the survey-anchored short-run expectation built on the one-sided trend of the unobserved-components stochastic-volatility model (Stock and Watson 2007), so the unit-root reading describes the fitted rate rather than a directly observed price, and the serial-correlation and break diagnostics run on the first difference. The expected-inflation leg is a quasi-real-time model estimate and not a real-time one, because its volatility prior is fixed on the full sample (Orphanides and van Norden 2002). The series is monthly and is tested over 298 observations from 2001-08-31 to 2026-05-31.

The ex-ante real policy rate gives an ambiguous reading on the level. The Dickey and Fuller (1979) test in the Said and Dickey (1984) form does not reject a unit root at the five-percent level, p = 0.083, and the Phillips and Perron (1988) test agrees at p = 0.136, while the Kwiatkowski et al. (1992) test does not reject stationarity either, p > 0.10, so both nulls are retained and no clean order of integration is assigned. The GLS-detrended test of Elliott, Rothenberg, and Stock (1996), an optional escalation in the matrix, was not run for this model-derived member, so the ambiguity is reported as it stands. The observed nominal leg carries a level variance 1.77 times the real rate's against 0.992 times for the expected-inflation leg, with a correlation of 0.668 between the real rate and the nominal rate. On the first difference the Ljung and Box (1978) portmanteau rejects white noise at the monthly lags 12 and 24, Q = 34.71 and Q = 47.77 at p = 0.001 and p = 0.003. The automatic portmanteau of Escanciano and Lobato (2009) is not run, because the matrix reserves it for daily members. The Bai and Perron (1998) procedure, computed by the algorithm of Bai and Perron (2003), finds no break in the mean of the differenced series, a reading consistent with the parameter-instability inference of Andrews (1993) on that object (Perron 1989).

This is a monthly model-derived series with no posited seasonal quantity, whose inputs are deseasonalized where relevant, so the seasonal-unit-root and seasonal-stationarity machinery of Hylleberg et al. (1990) and Canova and Hansen (1995) is not applicable and is not run (Beaulieu and Miron 1992; Ghysels and Osborn 2001).