South Korea Economic Sentiment Index

The Economic Sentiment Index is a raw composite sentiment index recorded each month and carried through without transformation. It combines the survey responses of firms and households into a single standardized figure benchmarked to a long-run reference value of 100, where readings above that mark denote private-sector sentiment above its historical average and readings below it sentiment below that average.

The measure rests on the behavioral premise that the attitudes and expectations of firms and consumers are themselves observable and economically consequential (Katona 1951).

Its raw material is a pair of qualitative business-tendency and consumer surveys in which respondents report only whether conditions are better, about the same, or worse. This up-same-down design was formalized by Anderson (1952), and those trichotomous replies are condensed into continuous figures by balance and probability reasoning (Theil 1952; Carlson and Parkin 1975).

Pooling the business and consumer balances into one headline follows the composite and diffusion-index tradition of Moore (1961). This exercise sits within the program of cyclical-indicator measurement begun by Mitchell and Burns (1938) and Burns and Mitchell (1946), and it was later recast as a latent common factor of comoving monthly series (Stock and Watson 1991).

The series therefore belongs to the family of attitudinal composite gauges, whose construction and meaning have been reviewed across the literature (Zarnowitz 1992; Nardo 2003; Ludvigson 2004).

KRED applies no transformation to the Economic Sentiment Index, neither rescaling, deflation, seasonal adjustment, smoothing, nor annualizing, and stores the published monthly figure exactly as received. The methodology to describe is therefore limited to the survey-and-index basis by which such a composite comes to exist.

The starting point is a pair of tendency surveys of the Anderson (1952) type, in which firms and households answer in qualitative up-same-down terms rather than reporting magnitudes. The theoretical link between the balance of replies and underlying conditions was first set out by Anderson (1952).

Converting those qualitative shares into quantitative readings uses the balance and probability methods of Theil (1952) and Carlson and Parkin (1975). This family of quantification procedures was surveyed and compared by Nardo (2003).

The several component series are then merged into one composite by the diffusion and composite-index construction codified by Moore (1961) and operationalized for monthly real-time signaling by Shiskin (1961). This is an exercise in the indicator-selection program of Mitchell and Burns (1938), whose modern statistical form treats the headline as a single latent factor of comoving aggregates (Stock and Watson 1989, 1991).

Interpreting the resulting composite against the cycle draws on the turning-point and reference-cycle apparatus of Bry and Boschan (1971) and the synthesis of Zarnowitz (1992). The standing caution that such atheoretical measurement is not a structural model is the one stated by Koopmans (1947).

As an economic gauge, the Economic Sentiment Index is read as a timely summary of the private sector's perceived state of the economy. Katona (1951) envisaged this role when he argued that measured attitudes and expectations help drive spending independently of current income.

Its standing as a coincident-to-leading read on activity follows from the cyclical-indicator framework of Mitchell and Burns (1938), Burns and Mitchell (1946), Moore (1961), and Shiskin (1961), and from the dynamic-factor reinterpretation of the state of the economy by Stock and Watson (1989).

Empirically, survey sentiment carries independent forecasting content for spending and output beyond what fundamentals explain (Carroll, Fuhrer, and Wilcox 1994; Ludvigson 2004). This property is what makes the composite economically informative.

Because sentiment co-moves with labor-market perceptions, it is often read alongside the labor-force aggregates. Those aggregates rest on the operational definitions of the labor force (Hussmanns, Mehran, and Verma 1990) and the flow view of unemployment (Clark and Summers 1979), and the series is also read against the broader evaluation of the indicator approach (Zarnowitz 1992).

The series describes perceived conditions rather than furnishing a structural forecast. This interpretive limit was emphasized by Koopmans (1947).

In financial use, the Economic Sentiment Index serves as a high-frequency read on the private sector that markets monitor for early signals on demand. Its predictive content for spending was established by Carroll, Fuhrer, and Wilcox (1994) and reviewed by Ludvigson (2004).

Its appeal to practitioners is the same one that Katona (1951) identified, namely that survey-measured expectations move ahead of realized outlays and so carry information about the near-term path of cyclically sensitive earnings and credit.

Read as a cyclical indicator in the tradition of Moore (1961), Shiskin (1961), and Zarnowitz (1992), the series helps locate the economy relative to reference-cycle turning points. Those turning points are dated by the apparatus of Burns and Mitchell (1946) and Bry and Boschan (1971), and such placement matters for positioning across the cycle.

Because the headline is a quantified balance of qualitative replies, its movements must be read with the attendant measurement caveats (Theil 1952; Carlson and Parkin 1975; Nardo 2003). Those movements must further be interpreted within the single-factor representation of comoving activity (Stock and Watson 1991).

The series is a descriptive sentiment gauge rather than a tradable forecast, a reading anchored in the up-same-down survey design formalized by Anderson (1952).

The Economic Sentiment Index is a not-seasonally-adjusted monthly composite, tested over 282 observations from 2003-01 to 2026-06 under a constant-plus-trend specification. Although a standardized sentiment composite is a bounded-support object in principle, the index stays far from any binding bound over the sample, so the standard unit-root null nearly coincides with the bounded-support null and no bounded correction is required (Cavaliere and Xu 2014).

The augmented Dickey and Fuller (1979) test with the lag augmentation of Said and Dickey (1984) rejects the unit root at p = 0.0013, the nonparametric Phillips and Perron (1988) test rejects at p = 0.0136, and the KPSS test does not reject stationarity at p > 0.10 (Kwiatkowski et al. 1992), so the three procedures agree without dissent on a clean I(0) classification. The DF-GLS escalation of Elliott, Rothenberg, and Stock (1996) is reserved for ambiguous readings and is not required here.

Because the three procedures concur on a clean I(0) verdict, the level is the stationary object and its mean-break date is citable. The Bai and Perron (1998) procedure, computed by the algorithm of Bai and Perron (2003), identifies a single mean shift at 2012-07, read as a level movement of a stationary series rather than a stochastic-trend break (Perron 1989).

The index is not seasonally adjusted, and the seasonal battery is the designed diagnostic here. The HEGY seasonal-unit-root test rejects seasonal unit roots at p = 0.000 (Hylleberg et al. 1990) and the seasonal-dummy F test rejects at p = 0.000, so the seasonality present is deterministic rather than stochastic, consistent with the stationary-seasonality null framework of Canova and Hansen (1995) and the seasonal-time-series analysis of Ghysels and Osborn (2001). The QS statistic on the differenced series does not reject at p = 0.0675, so no residual seasonal autocorrelation remains beyond the deterministic pattern. On the stationary level the Ljung and Box (1978) portmanteau rejects the no-autocorrelation null at p = 0.000 at both the lag-12 and lag-24 monthly rungs, indicating the strong serial dependence expected of a persistent sentiment level.