South Korea One-Year Inflation Expectations Survey

KRINFEXPSV is the raw one-year-ahead inflation-expectation series that records, without any transformation, the percentage rate at which household-survey respondents expect the aggregate consumer price level to rise over the coming twelve months. What respondents forecast is not any single price but the proportional change in a consumer price index, an aggregate whose level is defined by index-number theory rather than by any one price.

In the fixed-basket tradition the price level is measured as a base-period-quantity-weighted average of individual prices (Laspeyres 1871), set alongside the current-weighted counterpart (Paasche 1874) and the ideal index that reconciles the two through their geometric mean (Fisher 1922). The economic-theoretic target such an index approximates is the true cost-of-living index, the expenditure ratio that holds utility fixed (Konüs 1939). That cost-of-living index is framed as the concept a published consumer price index is meant to track (Diewert 1998).

Respondents therefore report an expected growth rate of a Laspeyres-type aggregate. By construction this aggregate departs from the superlative formulas (Törnqvist 1936; Diewert 1976) and from the axiomatic ideal characterized by the four-approach synthesis of Diewert (2002) and by Balk (1995). The headline number respondents anchor to embeds commodity-substitution, outlet, and quality-change bias relative to a cost-of-living target (Boskin et al. 1996). The expectation is thus formed over a measured rather than a theoretically exact inflation rate, and over a headline rather than an underlying or core series (Eckstein 1981).

KRED stores the survey reading exactly as collected and applies none of rescaling, deflating, smoothing, annualizing, seasonal adjustment, or revision. The series comes to exist through two layered measurement procedures.

The first layer is the object respondents are asked to anticipate, the change in a consumer price index constructed as a base-weighted fixed-basket aggregate (Laspeyres 1871). This aggregate is neither the current-weighted form (Paasche 1874) nor the geometric mean of the two that satisfies time and factor reversal (Fisher 1922). A superlative alternative such as the log-change expenditure-share index of Törnqvist (1936) approximates a flexible cost-of-living aggregator (Diewert 1976) and has been organized axiomatically (Balk 1995; Diewert 2002). The gap between this fixed-basket measure and the utility-constant index (Konüs 1939) is quantified as substitution bias (Diewert 1998) and enumerated as the broader basket, outlet, and quality biases (Boskin et al. 1996).

The second layer is the aggregation of the expectation itself, elicited from a cross-section of households and summarized to a central reading. This shares its spirit with the limited-influence and trimmed-mean estimators used to summarize a skewed cross-sectional distribution of price changes, and with the framing of measured inflation as trend plus noise (Bryan and Cecchetti 1994; Bryan, Cecchetti, and Wiggins 1997; Cecchetti 1997). Because KRED imposes no model identification, the persistent-trend decomposition is named only to locate where this raw reading sits rather than applied to it (Stock and Watson 2007).

The series is a measure of where households expect the cost of living to head. The economic-theoretic target that any price index approximates is the expenditure ratio that holds welfare constant (Konüs 1939). This ratio serves as the benchmark against which a fixed-base aggregate reading is judged (Laspeyres 1871; Diewert 1998).

Expected inflation matters economically because the level households perceive and project is the measured aggregate rather than the superlative or ideal construct. The very number being anticipated is therefore driven by substitution, outlet, and quality bias (Boskin et al. 1996) and by the differences among index formulas (Laspeyres 1871; Paasche 1874; Fisher 1922).

The persistence of the series connects to the distinction between headline and underlying inflation. Well-anchored expectations should track the durable component rather than transitory food and energy moves, a component identified as the long-run growth-path concept (Eckstein 1981), the output-neutral component (Quah and Vahey 1995), and the trend-versus-transitory decomposition (Stock and Watson 2007).

When the cross-section of price changes is skewed, the survey median is informative about underlying price pressure through the logic of robust central tendency (Bryan and Cecchetti 1994). The properties an inflation concept worth forecasting ought to satisfy are set by the axiomatic survey (Balk 1995) and the four-approach synthesis (Diewert 2002).

The series is the survey input to a real interest rate. Deflating a nominal yield requires an expected change in a consumer price index whose construction follows the fixed-basket aggregation (Laspeyres 1871), the current-weighted form (Paasche 1874), and the ideal index that reconciles the two (Fisher 1922).

Index-linked instruments and cost-of-living escalation clauses settle on a published price aggregate, so the inflation compensation embedded in nominal assets is priced against a measured rather than an exact index. This published aggregate approximates the utility-constant target (Konüs 1939) only up to the substitution wedge (Diewert 1998) and the basket and quality biases (Boskin et al. 1996).

Because hedging demand responds to the durable component of inflation, how much of a given survey reading to treat as a persistent repricing signal becomes the operative question. This judgment is informed by the short-run-versus-trend framing (Cecchetti 1997), the core concept (Eckstein 1981), and the unobserved-components trend (Stock and Watson 2007).

A contract that fixes payoffs to a particular index definition bears formula risk, which is bounded by the superlative-index results (Törnqvist 1936; Diewert 1976) and the axiomatic characterization (Balk 1995). This is the same definitional choice catalogued in the four-approach account (Diewert 2002).

KRINFEXPSV is the raw one-year-ahead inflation-expectations survey print, a not-seasonally-adjusted monthly series tested over 292 observations from 2002-02 to 2026-05 under a constant-plus-trend specification. The survey level stays well inside its natural bounds over the sample, so the bounded-support and standard unit-root nulls nearly coincide and no bounded correction is applied (Cavaliere and Xu 2014).

The augmented Dickey and Fuller (1979) test with the lag augmentation of Said and Dickey (1984) rejects a unit root at p = 0.024, but the nonparametric Phillips and Perron (1988) test fails to reject at p = 0.188 while the KPSS test rejects trend stationarity at p = 0.040 (Kwiatkowski et al. 1992), so the parametric and nonparametric tests disagree and the stationarity-null test points the other way again, and no clean order of integration is assigned. The DF-GLS escalation of Elliott, Rothenberg, and Stock (1996) is reserved for exactly such ambiguous readings and enters only as confirmation of the ADF leg, never as a fourth vote.

Because the protocol assigns no clean stationary level, the serial-correlation and mean-break diagnostics are computed on the first difference, the object those procedures require (Bai and Perron 1998; Perron 1989). On the differenced series the Ljung and Box (1978) portmanteau rejects white noise at both the lag-12 and lag-24 rungs, with p = 0.037 at lag 12 and p = 0.001 at lag 24, and 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.

The series is not seasonally adjusted and the seasonal battery applies. The HEGY test rejects seasonal unit roots at p = 0.000 (Hylleberg et al. 1990), consistent with the stationary-seasonality null of Canova and Hansen (1995) and the seasonal-analysis framework of Ghysels and Osborn (2001), so any seasonality present is deterministic. On the differenced series the QS statistic rejects residual seasonality at p = 0.001 while the seasonal-dummy F test does not reject at p = 0.719, indicating residual seasonal autocorrelation without a stable deterministic seasonal pattern.