South Korea PPI Inflation (YoY)

The series in this group are year-over-year inflation rates derived from the underlying consumer-price and producer-price index levels rather than from any model of price dynamics. Each rate is the twelve-month proportional change of a fixed-base index, and because the common base period enters the current and the year-ago index identically it cancels in the ratio, so the growth rate is invariant to the choice of base year and to any rebasing or relinking of the level series (Shiratsuka 1997). The twelve-month differencing that defines the rate is itself a seasonal-differencing filter that strips the deterministic within-year price pattern without a model-based adjustment step, at the documented cost that a year-over-year change is influenced by the previous year's developments and lags turning points in the underlying rate, a caveat noted for exactly this construction (Shiratsuka 1997).

The group holds a headline consumer-price rate, two exclusion-based core consumer-price rates that omit the volatile food, energy, agricultural, and oil components, and a producer-price rate that sits upstream in the pricing chain, read together as the observed inflation panel against which the site's model-based trend and limited-influence underlying-inflation gauges are compared (Bryan, Cecchetti, and Wiggins 1997; Clark 2001; Rich and Steindel 2007). The core rates isolate the persistent component that fixed exclusion removes the same way each month, while the headline and producer rates carry the transitory relative-price movements that the core measures are designed to filter (Cogley 2002; Stock and Watson 1998).

Each rate is produced by a single arithmetic transform of the published index level, with no estimation and no seasonal-adjustment model.

The year-over-year rate at month $t$ is $g_t = 100\,(P_t/P_{t-12} - 1)$, the twelve-month proportional change of the index $P_t$ expressed in percent, so the base period common to $P_t$ and $P_{t-12}$ cancels and the rate is unaffected by rebasing or relinking of the level series (Shiratsuka 1997). For a monthly input the lag is twelve months and for a quarterly input it is four quarters, the horizon that spans one full seasonal cycle in each case. The transform is applied identically to the headline, core, and producer index levels, and no smoothing, deseasonalizing, or outlier treatment is layered on top, so the published rate is the raw twelve-month change and inherits the level series exactly (Bryan, Cecchetti, and Wiggins 1997).

Central banks watch year-over-year inflation rates because each mixes the persistent component to which policy should respond with transitory relative-price noise, and the core rates are the attempt to isolate that persistent component (Bryan, Cecchetti, and Wiggins 1997; Cogley 2002). A fixed-exclusion core removes the same items every month regardless of where the shocks fall, so its relation to the headline rate has been a running theme of the evaluation literature, and comparative studies find that no single core measure dominates (Clark 2001; Rich and Steindel 2007). Within an inflation-targeting regime the operational case is that policy should respond to the persistent part of inflation relative to the target, in Korea the 2 percent consumer-price target (Stock and Watson 1998).

The producer-price rate sits upstream in the pricing chain and is a window on the pass-through to consumer prices, though that transmission is incomplete and lagged (Cogley 2002). On the forecasting question, core measures predict total inflation better than headline in some samples (Crone et al. 2013; Meyer, Venkatu, and Zaman 2013), while simple benchmarks that use only lagged inflation are notoriously hard to beat (Atkeson and Ohanian 2001). Because these rates can carry the upward bias of the mean on a right-skewed price-change distribution, on this site they are read alongside robust central-tendency gauges and the model-based trend (Ball and Mazumder 2019; Meyer and Venkatu 2014; Shiratsuka 1997).

Fixed-income and inflation-market participants price nominal yields off a decomposition into expected inflation, real rates, and risk premia, in which the expected-inflation input is a persistent object rather than the latest monthly print, so the year-over-year rate and its core component serve as the reference (Bryan, Cecchetti, and Wiggins 1997; Cogley 2002). For reading monthly releases the informational question is whether a surprise moves the persistent component or only the noise, which is why long-maturity yields react to macroeconomic news through revisions in long-run inflation expectations (Gurkaynak, Sack, and Swanson 2005).

The producer-price rate carries pipeline pressure that leads consumer prices and enters breakeven-rate assessment, though its pass-through is incomplete (Crone et al. 2013). Because no single inflation gauge dominates, reading the core, headline, and producer rates together is useful (Clark 2001; Rich and Steindel 2007; Stock and Watson 1998), and the limits of their forecasting performance are stated plainly (Atkeson and Ohanian 2001; Meyer, Venkatu, and Zaman 2013; Ball and Mazumder 2019).

KRPPIYOY is the year-over-year producer-price inflation rate, the twelve-month proportional change of the underlying price index level in which the common base period cancels, and the rate itself is the candidate stationary object tested over 377 monthly observations from 1995-01 to 2026-05 under a constant-plus-trend specification. The augmented Dickey and Fuller (1979) test with the lag augmentation of Said and Dickey (1984) rejects the unit root at p = 0.012 with a statistic of −3.91, and the KPSS test fails to reject level stationarity at p > 0.10 with a statistic of 0.100 (Kwiatkowski et al. 1992), so the lag-augmented unit-root test and the stationarity test concur on stationarity, while the Phillips and Perron (1988) test is the sole dissenter, failing to reject the unit root at p = 0.065 with a statistic of −3.30. Because that dissent reflects the documented size and power problems of the Phillips-Perron statistic relative to the lag-augmented and efficient alternatives, the disagreement leaves no clean order of integration assigned even though the two better-behaved procedures concur on stationarity (Ng and Perron 2001).

On the level of the rate, the candidate stationary object, the Ljung and Box (1978) portmanteau rejects the no-autocorrelation null at p = 0.000 at both the lag-12 and lag-24 rungs, with Q statistics of 1500.34 and 1516.21. Because the rate is the object under test, a level break search is the appropriate mean-shift diagnostic (Perron 1989), and the Bai and Perron (1998) procedure computed by the algorithm of Bai and Perron (2003) locates five mean shifts, at 1999-09, 2007-07, 2012-03, 2016-11, and 2021-07.

Because the year-over-year construction is itself the seasonal-differencing filter that removes the seasonal unit roots the HEGY class interrogates, the HEGY and Canova-Hansen seasonal-unit-root tests are deliberately not run, since interrogating seasonal unit roots the transform has already annihilated is circular (Hylleberg et al. 1990; Canova and Hansen 1995; Ghysels and Osborn 2001).