Common mistake 1 with Exponential smoothing

Fitting a seasonal model to fewer than two full cycles of data.

Common mistake 2 with Exponential smoothing

Extrapolating an undamped trend over a long horizon. It will give you absurd numbers.

Common mistake 3 with Exponential smoothing

Assuming it is worse than ARIMA because it is simpler. Test it - often it is not.

Common mistake 4 with Exponential smoothing

Choosing ETS over ARIMA on AIC. The two families overlap without nesting, so their information criteria are not comparable; only out-of-sample accuracy settles it.

EconStatsCraft › Exponential smoothing

Exponential smoothing

Simple exponential smoothing, Holt’s linear trend and Holt–Winters seasonal models, with fitted smoothing parameters, a white-noise check and a forecast.

What it needs

Series

The intuition

Take a weighted average of the past where the weights decay geometrically. That is the whole idea, and it is remarkably hard to beat. Holt adds a slope that is itself smoothed; Holt–Winters adds a seasonal index that is also smoothed. Each component "forgets" at its own rate, controlled by its smoothing parameter.

When to use it

Short-horizon forecasting of business and seasonal series. It is simple, extremely hard to beat, and wins forecasting competitions regularly.

Where the estimator comes from

Simple exponential smoothing sets \ell_t=\alpha y_t+(1-\alpha)\ell_{t-1}. Unrolling gives \ell_t=\alpha\sum_{j\ge 0}(1-\alpha)^j y_{t-j} - geometrically declining weights that sum to one. Holt adds b_t=\beta(\ell_t-\ell_{t-1})+(1-\beta)b_{t-1} and forecasts \hat{y}_{t+h}=\ell_t+h\,b_t; damping multiplies the slope by \phi^h so the forecast flattens. The modern ETS framework writes all of these as state-space models, which is what gives them likelihoods, information criteria and proper prediction intervals.

The model

\ell_t=\alpha y_t+(1-\alpha)(\ell_{t-1}+b_{t-1}),\qquad b_t=\beta(\ell_t-\ell_{t-1})+(1-\beta)b_{t-1}

What it assumes

The components (level, trend, season) evolve smoothly.
You have specified the seasonal period correctly.
Multiplicative components require strictly positive data.
The series is long enough to estimate the smoothing parameters. With a short series the optimiser often lands on a corner - alpha at 0 or 1 - which is a fit to the boundary rather than a fitted model.
No level shifts. Exponential smoothing adapts to a shift eventually, but it treats it as a sequence of surprises rather than as an event, and the forecast lags through the adjustment.

Statistical properties

SES is equivalent to ARIMA(0,1,1) with θ = α − 1; Holt to ARIMA(0,2,2). The overlap is exact, not approximate.
The state-space formulation makes AIC comparison with ARIMA legitimate.
Damped trend is usually more accurate at long horizons than undamped - one of the most robust findings in the forecasting literature.
Multiplicative seasonality requires strictly positive data.
Needs at least two full seasonal cycles to identify a seasonal component.

How to read the output

α near 1 means the level chases the most recent observation; near 0 means it barely moves.
β controls how quickly the slope updates, γ the seasonal pattern.
A damped trend flattens the forecast out - usually more realistic than extrapolating a slope forever.
Ljung–Box on the residuals is the same adequacy check as for ARIMA.
The letters are the specification: error, trend, season, each additive, multiplicative or none. ETS(A,N,N) is simple exponential smoothing and ETS(A,A,A) is Holt-Winters; the module reports which was chosen, and that choice is a finding.
Alpha near 0 means the level barely responds to new data, so the forecast is close to a long-run average. Alpha near 1 means it is close to the last observation. Either extreme is worth questioning.
ETS and ARIMA overlap but neither contains the other: some ETS models have no ARIMA equivalent and vice versa, so comparing them on AIC is not valid - compare them out of sample.

A worked reading

Holt with α = 0.61, β = 0.08: the level adapts quickly to new data while the slope is very stable. The forecast therefore tracks recent movements but does not let a temporary blip rotate the whole trend line.

Symptom → diagnosis → what to do

If you see this	It means	Do this
α near 1	The level chases the latest observation	Fine for volatile series; check it is not just noise-fitting
α near 0	The level barely moves	The series is close to constant - a simple mean may do
Trend forecast looks absurd at long horizons	Undamped trend extrapolating forever	Enable damping
Ljung–Box on residuals rejects	Structure left over	Add trend or seasonal components, or move to ARIMA
Seasonal amplitude grows with level	Multiplicative seasonality	Switch the seasonal component to multiplicative

Common mistakes

Fitting a seasonal model to fewer than two full cycles of data.
Extrapolating an undamped trend over a long horizon. It will give you absurd numbers.
Assuming it is worse than ARIMA because it is simpler. Test it - often it is not.
Choosing ETS over ARIMA on AIC. The two families overlap without nesting, so their information criteria are not comparable; only out-of-sample accuracy settles it.

Where to read this properly

Holt (1957); Winters (1960)
Hyndman, Koehler, Ord & Snyder (2008), Forecasting with Exponential Smoothing
Makridakis, Spiliotis & Assimakopoulos (2018) - the M4 competition

Related Time series - one series

Time-varying coefficients (Kalman) · Unit root tests · Correlogram (ACF / PACF) · ARIMA / SARIMA · Rolling-origin cross-validation · Theta method · Unobserved components · STL forecasting · Forecast combination · Markov regime switching · Change frequency (resample) · ARCH / GARCH volatility · Decompose & filter · Spectral analysis · Forecast accuracy

All analyses

Time series - one series

Time series - several series

Yield curve (Nelson–Siegel / Svensson) · Dynamic factor model · Dynamic regression (ARIMAX) · Structural VAR · VAR · Granger causality · Cointegration · VECM · ARDL & bounds test

Panel data

Panel regression · Panel cointegration tests · Heterogeneous panel: Mean Group & PMG · Gravity model of trade · Dynamic panel GMM · Panel diagnostics · Panel unit-root tests

Multivariate statistics

Principal components (PCA) · Factor analysis · Cluster analysis

Marketing research

Conjoint analysis · MaxDiff (best–worst scaling) · TURF - reach of a product line · Price sensitivity (Van Westendorp) · Market basket analysis · Perceptual map · Scale reliability (Cronbach’s alpha) · Key driver analysis

Causal

Spatial autocorrelation · Spatial regression · Heckman selection · Instrumental variables (2SLS) · Difference-in-differences · Staggered difference-in-differences · Regression discontinuity · Matching & weighting · Oaxaca–Blinder decomposition

Finance & risk

Black–Scholes option pricing · Bond price, duration & convexity · Merton distance to default · EMI / home loan planner · CAPM / beta estimation · Multi-factor model · Portfolio optimization · Value at Risk & Expected Shortfall · Performance & drawdown · Time value of money & NPV/IRR · Equity valuation (DDM, FCF, residual income) · Cost of capital (WACC) · Binomial option pricing · Option strategies (payoff & profit) · Forwards, futures & swaps · Credit risk & Basel capital · Extreme value theory (tail risk)

Actuarial

Life table · Life annuity · Life insurance pricing · Chain ladder reserving · Buhlmann credibility · Survival analysis · Mortality graduation · Aggregate loss model · Ruin probability · Multi-state Markov model · Loss severity fitting

Tests & breaks

Monte Carlo experiment · F-test, Wald & Chow · Structural breaks & stability

Machine learning

Regression vs machine learning · Ridge / Lasso / Elastic Net · Random forest · Gradient boosting

Open the workspace →

Skip to results

Data & variables

Each example is simulated from a known model, so you can check whether the estimator recovers the truth - and nothing here is anyone's private data.

Filter variables

Drag a variable into a slot on the right - or just click it.

Choose an example dataset above to begin.

Choose a transformation - it opens in the data editor, with a live preview.

Apply to

Multiply by

Columns ctrl-click for several

Group by ctrl-click for several

Within groups

Statistic

Bands

How many

Cut points

Keep rows where

Value

and

Fill with

Value

Also mark which rows were filled

Rules

Base period = 100

Which part

Percent at each tail

Threshold

Survey weight optional

Direction

Names column

Values column

Largest first

Delete them, not just count them

Read this column as

Tidy the text

New name optional

EconStatsCraft rigorous, plain-spoken econometrics

Econometrics that explains itself.

Regression, panel data, time series and the whole battery of tests. Every result is followed by a verdict you can defend.

Load a dataset to see it here.

Exponential smoothing

What it needs

The intuition

When to use it

Where the estimator comes from

The model

What it assumes

Statistical properties

How to read the output

A worked reading

Symptom → diagnosis → what to do

Common mistakes

Where to read this properly

See also

Related Time series - one series

All analyses

Charts

Maps

Explore

Regression

Time series - one series

Time series - several series

Panel data

Multivariate statistics

Marketing research

Causal

Finance & risk

Actuarial

Tests & breaks

Machine learning

Econometrics that explains itself.