South Korea Employment (SA)

KREMP is the seasonally adjusted monthly count of employed persons, recorded without further transformation as the level at which it is published. The employed are defined within the standard labor-force framework as those who worked for pay or profit, or were temporarily absent from such work, during a short reference period (Hussmanns, Mehran, and Verma 1990). This headcount counts persons rather than hours and treats the job-search and availability criteria as the boundary separating the employed from the unemployed and from those outside the active population (Clark and Summers 1979).

As a count of an aggregate real-activity quantity, employment belongs to the family of monthly series that are roughly coincident with the general state of business, and it sits at the center of reference-cycle dating (Mitchell and Burns 1938; Burns and Mitchell 1946). Such comoving aggregates are recast as observable indicators of a single unobserved state of the economy, and employment is one of the four monthly aggregates whose common factor defines the coincident composite (Stock and Watson 1989, 1991). The count takes a codified role within diffusion and composite cyclical-indicator systems and functions as a current-conditions signal read in real time (Moore 1961; Shiskin 1961; Zarnowitz 1992).

Such atheoretical counts acquire economic meaning only against a model of the labor market (Koopmans 1947).

KRED stores the published value unchanged and computes nothing from it, applying no rescaling, deflating, smoothing, or annualizing. Every operation that produces the number therefore occurs in the measurement system that generates the series rather than in KRED.

That measurement rests on a probability household survey in which interviewers record activity status over a fixed reference week and tabulate the employed under internationally agreed status rules (Hussmanns, Mehran, and Verma 1990), and it follows the survey-design logic first treated scientifically for activity questions (Anderson 1952). Raw monthly headcounts move with calendar and seasonal regularities, so the published level is already seasonally adjusted at source, and the series KRED carries is consequently untransformed by KRED and adjusted upstream at once, both true without contradiction. The adjustment removes the recurring within-year pattern in the manner of the moving-average decomposition operationalized for monthly indicators, leaving a series read for cyclical rather than seasonal movement (Shiskin 1961).

Interpreted as a cyclical quantity, the count inherits the indicator-selection logic, the reference-cycle measurement, the algorithmic turning-point dating, and the diffusion and composite construction of the cyclical-indicator tradition (Mitchell and Burns 1938; Burns and Mitchell 1946; Bry and Boschan 1971; Moore 1961). The employment level is read as a dynamic factor that proxies a latent activity state (Stock and Watson 1989, 1991). The historical usefulness of that signal has been evaluated (Zarnowitz 1992), and its interpretation carries the qualification that it acquires economic meaning only against an explicit model (Koopmans 1947).

As a coincident measure of real activity, the employment count is among the series whose clustered turning points define expansions and contractions in the reference-cycle scheme, and it was selected precisely for that rough coincidence with the cycle (Mitchell and Burns 1938; Burns and Mitchell 1946). Such a count enters the single-index reading of the economy's current state, the empirical basis for treating month-to-month employment changes as signals of the cyclical position (Stock and Watson 1989, 1991). The series sits inside diffusion and composite systems that identify turning points in near-real time, and the historical reliability of that signal has been evaluated (Moore 1961; Shiskin 1961; Zarnowitz 1992).

The labor-market interpretation reads employment levels through gross entry and exit flows rather than as a static stock, and it relies on status definitions to fix what an increment in the count economically represents (Clark and Summers 1979; Hussmanns, Mehran, and Verma 1990). A standing caution holds that a coincident count carries cyclical information only against a structural model of the labor market (Koopmans 1947). Because employment determines household income, it also conditions the spending and confidence channels that underlie consumption behavior (Katona 1951).

Markets read the monthly employment count as a high-frequency gauge of the cyclical state that conditions monetary policy and the path of short rates, resting on its established coincident-indicator role and its formalization as a latent-state proxy (Mitchell and Burns 1938; Burns and Mitchell 1946; Stock and Watson 1989, 1991). Because employment drives aggregate labor income, it anchors the consumption channel that underlies household spending behavior (Katona 1951), and its forecasting content has been assessed through the related consumer-sentiment series (Carroll, Fuhrer, and Wilcox 1994; Ludvigson 2004).

Composite and diffusion frameworks embed the count in the real-time dashboards on which positioning across the cycle is built (Moore 1961; Shiskin 1961; Zarnowitz 1992), and the turning points are dated by algorithm (Bry and Boschan 1971). For credit and duration risk, the labor-market flows and the status accounting determine whether a change in the level reflects hiring strength or compositional churn (Clark and Summers 1979; Hussmanns, Mehran, and Verma 1990).

The sample is 323 monthly observations spanning 1999-06 to 2026-04, and the unit-root tests are run with a constant and trend.

The augmented Dickey-Fuller regression of Dickey and Fuller (1979) with the lag augmentation of Said and Dickey (1984) rejects the unit-root null at p = 0.0152, and the nonparametric test of Phillips and Perron (1988) rejects more decisively at p = 0.0082. The stationarity-null KPSS test of Kwiatkowski et al. (1992) does not reject at p > 0.10, so all three procedures concur on a trend-stationary I(0) classification with no ambiguity. The GLS-detrended escalation of Elliott, Rothenberg, and Stock (1996) is reserved for ambiguous outcomes under the house protocol and is not needed on this clean reading.

The multiple-break framework of Bai and Perron (1998), computed by the dynamic-programming algorithm of Bai and Perron (2003), locates five structural breaks, dated 2003-06, 2007-06, 2011-10, 2015-10, and 2021-12; because the three procedures agree on a clean I(0) reading the level is the stationary object, so these mean-level break dates are citable and read as shifts in the trend level across the sample rather than a stochastic trend.

The seasonal-unit-root test of Hylleberg et al. (1990), complemented by the stationarity-side test of Canova and Hansen (1995), detects no seasonality, which reflects the source seasonal adjustment already applied to the published count rather than any inherent absence of a seasonal pattern. The portmanteau statistic of Ljung and Box (1978), refining the original Box and Pierce (1970) form, returns p = 0.000 at the maximum lag, indicating strong residual autocorrelation as expected for a smooth, highly persistent employment level.