South Korea Leading Composite Index

The Korean business-cycle indicators KRLEAD, KRLAG, KRCAPUTIL, KRINVRATIO, and KRCONSTORD are raw monthly series recorded as reported without any transformation, a panel that reads the current state of real activity across the manufacturing sector and the aggregate cycle. The panel spans a leading and a lagging composite cyclical index, a manufacturing capacity-utilization rate, a manufacturing inventory ratio, and construction orders, each a distinct vantage on the same underlying expansion and contraction.

The composite indexes descend from the empirical study of business cycles as recurrent but non-periodic fluctuations in aggregate activity (Burns and Mitchell 1946), and from the selection of series that systematically lead the turning points of that cycle (Mitchell and Burns 1938; Moore 1961). The practice of reading activity from a barometer of several coincident and leading series predates them (Persons 1923), and the composite-index apparatus was later standardized (Shiskin 1961) and set within the modern measurement of business cycles (Zarnowitz 1992).

Capacity utilization and the inventory ratio measure the intensity of resource use and the stock-to-shipment balance in manufacturing, the production magnitudes formalized in the measurement of output and productive capacity (Fabricant 1940; Kuznets 1941). The dynamic view that treats such series as the propagation of impulses through the productive structure originates with the impulse-propagation account of the cycle (Frisch 1936), and their empirical reading in the measurement-without-theory tradition follows (Koopmans 1947).

Construction orders lead the building cycle, an early and volatile signal of investment intentions, and are compiled as a component eligible for a composite indicator of activity (Nardo 2003; Stock and Watson 1989). Four of the five indicators are seasonally adjusted at source and are recorded here without further adjustment, while construction orders are not seasonally adjusted and retain their calendar structure.

KRED applies no transformation to KRLEAD, KRLAG, KRCAPUTIL, KRINVRATIO, or KRCONSTORD, neither deflating, re-scaling, smoothing, nor annualizing them, so each published number is the compiled figure exactly as measured. The seasonal adjustment carried in four of the five indicators is performed at source and reproduced verbatim, not applied by KRED, and construction orders are left in their unadjusted form.

The measurement basis for the composite indexes is the classical business-cycle program, which dates the recurrent expansions and contractions of aggregate activity from a reference chronology (Burns and Mitchell 1946) and builds leading and lagging groupings from series screened for their timing at turning points (Mitchell and Burns 1938; Moore 1961; Bry and Boschan 1971). The aggregation of the screened components into a single composite index follows the standardized index construction (Shiskin 1961; Nardo 2003), and the dynamic-factor reading of a common activity signal is its modern counterpart (Stock and Watson 1991).

Capacity utilization and the inventory ratio are production ratios in the output-measurement tradition (Fabricant 1940), and the combination of heterogeneous component series into an index invokes the classical index-number apparatus (Fisher 1922; Törnqvist 1936; Diewert 1976). Construction orders are compiled as a levels flow and enter the panel without seasonal adjustment, so that their calendar structure is preserved by construction rather than by estimation. In every case the published figure exists as a compiled measurement rather than a modelled quantity (Kuznets 1941).

The business-cycle indicators are among the most-watched real-activity quantities, because they trace the position of the economy within the expansion-contraction cycle in close to real time. The composite indexes summarize the recurrent but non-periodic movement of aggregate activity into a single reading (Burns and Mitchell 1946), and the leading grouping is built precisely to turn ahead of that cycle at its peaks and troughs (Mitchell and Burns 1938; Moore 1961).

The indicators behave differently by design, and so carry distinct information. Capacity utilization and the inventory ratio gauge the pressure on productive resources and the stock-to-shipment balance, the production magnitudes whose measurement organizes the study of output over the cycle (Fabricant 1940; Kuznets 1941). The impulse-propagation view explains why such series oscillate rather than drift (Frisch 1936), and their reading in the measurement-first tradition treats the regularities as objects in their own right (Koopmans 1947; Persons 1923).

Read together, the leading, coincident, and lagging vantages date the cycle and its stage, the chronology that anchors applied cyclical analysis (Zarnowitz 1992; Stock and Watson 1989). The lagging index confirms turning points the leading index anticipates, so their divergence is itself informative about the maturity of an expansion, a structural feature of the labor and output cycle (Clark and Summers 1979).

For market participants the business-cycle indicators are a macro overlay for cyclical positioning, capturing the turning points that reprice risk across equity, credit, and rate exposures. The reading of activity from a barometer of coincident and leading series is the oldest such overlay (Persons 1923), and the systematic screening of series by their lead at turning points made it operational for anticipating the cycle (Mitchell and Burns 1938; Moore 1961).

The vantage a given indicator occupies matters for interpretation. The leading composite is a forward signal that desks use to position ahead of inflection, while the lagging composite confirms the phase after the fact, and the reference chronology that dates the phases frames both (Burns and Mitchell 1946; Zarnowitz 1992). Capacity utilization and the inventory ratio sharpen the read on inflationary pressure and the stock cycle, the production magnitudes long tracked for their cyclical content (Fabricant 1940).

The composite construction and its dynamic-factor descendant supply the smoothed common signal that portfolio and macro strategists prefer to a single noisy series (Nardo 2003; Stock and Watson 1991; Bry and Boschan 1971). Activity indicators of this kind carry documented predictive content for spending and returns, which is why they inform tactical allocation across the cycle (Stock and Watson 1989; Carroll, Fuhrer, and Wilcox 1994; Ludvigson 2004).

KRLEAD is a seasonally-adjusted-at-source leading composite index, and its published level is the statistical object, expected to be integrated of order zero. The sample is 677 monthly observations from 1970-01 to 2026-05, and the unit-root triplet is fitted on the level with a constant.

The augmented Dickey-Fuller regression of Dickey and Fuller (1979), in the lag-augmented form of Said and Dickey (1984), rejects the unit root with a statistic of −6.559 at p = 0.000, the nonparametric test of Phillips and Perron (1988) concurs with a statistic of −4.710 at p = 0.000, and the stationarity-null test of Kwiatkowski et al. (1992) fails to reject level stationarity with a statistic of 0.053 at p > 0.10, so all three procedures agree on a clean I(0) classification. The GLS-detrended escalation of Elliott, Rothenberg, and Stock (1996) and the modified M-tests of Ng and Perron (2001) are reserved for ambiguous outcomes under the house protocol and are not invoked on this clean reading.

Because the level is stationary, the portmanteau and the break search are computed on the level itself, since differencing a stationary series would inject a noninvertible moving-average component and distort both diagnostics (Perron 1989; Hamilton 2018; Bai and Perron 1998). The portmanteau statistic of Ljung and Box (1978), refining the form of Box and Pierce (1970), rejects the white-noise null on the level, with Q = 2696.08 at lag 12 and Q = 3968.17 at lag 24, both at p = 0.000, and the multiple-break procedure computed by the dynamic-programming algorithm of Bai and Perron (2003) locates no break in the level.

The seasonal battery is run with the interpretation that any absence of seasonality reflects the source adjustment rather than an inherent property. The seasonal-unit-root test of Hylleberg et al. (1990), with the monthly mechanics of Beaulieu and Miron (1993), rejects unit roots at the seasonal frequencies with a statistic of 6850.75 at p = 0.000, and the seasonal dummies are not jointly significant with F = 0.465 at p = 0.924, so no deterministic seasonality survives at source. The seasonal portmanteau returns 1179.3 at p = 0.000 on the differenced series, which reflects residual serial dependence at the seasonal lags rather than a seasonal unit root or a deterministic seasonal pattern the source adjustment failed to remove. This stationary-seasonality reading is consistent with Canova and Hansen (1995) and with the seasonal-adjustment literature (Ghysels and Osborn 2001).