Skip to main content

Cookies on Macro by Mark.We use essential cookies for sign-in, locale, and security. With your permission, we also use optional cookies for preferences, for analytics, and for advertising measurement with the Meta (Facebook) Pixel. Read our cookie policy.

Macro by MarkMacro by Mark
  • Home
Loading…
Loading…
← Models
MacroConceptsPolicyModels

Unobserved components model

State-space decomposition that separates a macro series into latent trend, cycle, seasonal, and irregular components.

OverviewGraphMathReferences

Graph workspace

Trend-cycle decomposition of a simulated series (3x1 panel)

Graph workspace

The UCM's defining output is the decomposition itself. The top panel shows the observed series and the estimated trend. The middle panel shows the extracted cycle. The bottom panel shows the seasonal plus irregular residual. This three-panel layout lets the user see how parameter changes redistribute variation across components.

Chart axis metadata
xTime (quarters)quarters · 1 to TySeries level (top) / Cycle amplitude (mid) / Seasonal + irregular (bottom)index level · data-dependent
Observed series and its trend component static chart preview showing Observed series, Trend component
Cycle and seasonal components static chart preview showing Cycle component, Seasonal component, Zero
xTime (quarters)quarters · 1 to TySeries level (top) / Cycle amplitude (mid) / Seasonal + irregular (bottom)index level · data-dependent

Current results

Trend share of variance

Explain: Trend share of variance
(ση2+σζ2)  /  (ση2+σζ2+σκ2+σω2+σε2)(\sigma^2_\eta + \sigma^2_\zeta) \;/\; (\sigma^2_\eta + \sigma^2_\zeta + \sigma^2_\kappa + \sigma^2_\omega + \sigma^2_\varepsilon)(ση2​+σζ2​)/(ση2​+σζ2​+σκ2​+σω2​+σε2​)

Share of the model's disturbance variance that drives the trend (level and slope shocks) out of all five disturbances: the q-ratio split UCM software reports. High values indicate the series is trend-dominated. It is not the trend's share of the series' sample variance, which is near 1 in every setting because the trend is a random walk.

0.0905

proportion

Cycle period (implied)

Explain: Cycle period (implied)
2π/λc2\pi / \lambda_c2π/λc​

The implied period of the stochastic cycle in the same time units as the data. For quarterly data with λc=0.196\lambda_c = 0.196λc​=0.196, this is approximately 32 quarters (8 years).

32.06

periods

Cycle amplitude (RMS)

Explain: Cycle amplitude (RMS)
T−1∑tψt2\sqrt{T^{-1} \sum_t \psi_t^2}T−1∑t​ψt2​​

Root-mean-square amplitude of the simulated cycle component (the toy draws the components; it estimates none of them). Larger values mean the business-cycle component is economically significant relative to the trend.

1.69

index units

Signal-to-noise ratio

Explain: Signal-to-noise ratio
(ση2+σζ2+σκ2+σω2)  /  σε2(\sigma^2_\eta + \sigma^2_\zeta + \sigma^2_\kappa + \sigma^2_\omega) \;/\; \sigma^2_\varepsilon(ση2​+σζ2​+σκ2​+σω2​)/σε2​

Ratio of the structural disturbance variances (trend level and slope, cycle, seasonal) to the irregular variance. High values mean the model attributes most variation to interpretable components rather than noise.

3.09

ratio

Controls

gently bending trend
Explain: How bumpy the trend is

Level disturbance SD: 0.3

Size of the shocks to the trend's level; larger makes the trend rougher.

Higher values make the blue trend line wigglier and absorb more of the observed variation.

slowly drifting growth
Explain: Whether the growth rate drifts

Slope disturbance SD: 0.05

Size of the shocks to the trend's slope; larger lets the growth rate wander.

Higher values let the trend bend and change direction over long horizons.

moderately uneven swings
Explain: How much cycle size varies

Cycle disturbance SD: 0.8

Size of the shocks hitting the cycle; larger makes the size of swings vary more.

The red cycle line in the middle panel gains or loses amplitude volatility.

long-lived swings
Explain: How long cycles last

Cycle damping factor: 0.9

How long cyclical swings persist; values near one keep them going a long time.

Higher ρ\rhoρ makes cycle oscillations persist longer; lower ρ\rhoρ makes them die out quickly.

business-cycle length swings
Explain: How often the cycle repeats

Cycle frequency (radians): 0.196

How often the cycle repeats; the default setting gives an eight-year cycle on quarterly data.

Controls how many oscillations appear in the middle panel. Lower values produce longer cycles.

slowly evolving seasons
Explain: Whether the seasonal pattern shifts

Seasonal disturbance SD: 0.2

Size of the shocks to the seasonal pattern; zero keeps that pattern fixed.

Higher values make the green seasonal pattern evolve visibly over the sample.

moderate scatter
Explain: How noisy the data is

Irregular SD: 0.5

Size of the random measurement noise left unexplained by the other pieces.

Higher values increase the scatter of gray dots around the trend and make the gray bars in the bottom panel larger.

Keep going

Read next

Continue with a related concept, model, or example.

Theory model

IS-LM Model

a well-connected theoretical model two steps from here

New angle: theoretical model →
Macro by Mark
Help
Privacy PolicyTerms of UseEthics and Compliance

© 2026 Mark Jayson Nation LLC