Session 3

Exponential Smoothing
Simple Exponential Smoothing
• The simple exponential smoothing model can be written in
the following manner:
Ft+1  αX t  1- α  Ft Note: Moving Averages give equal
weight to past values
while Smoothing gives more weight to
where: recent observations.
Ft+1  forecasted value for next period
α  The level smoothing constant (0  α  1)
X t  Actual value of time series now  in period t 
Ft  Forecasted  i.e. smoothed  value for time period t
Weights for Alpha level smoothing = .1
Time Calculation Weight for Xt
t .1 .1
t-1 .9 × .1 .090 •  = .1
t-2 .9 × .9 × .1 .081
t-3 .9 × .9 × .9 × .1 .073 Regardless of the smoothing
constant chosen, the weights will
. eventually sum to 1. Whether the
. sum of the weights converges on 1
quickly or slowly depends on the
. smoothing constant chosen.
Total = 1.000
Weights for Alpha level smoothing = .9
Time Calculation Weight for Xt
t .9 .9
t-1 .1 × .9 .09 •  = .9
t-2 .1 × .1 × .9 .009
t-3 .1 × .1 × .1 × .9 .0009 Regardless of the smoothing
constant chosen, the weights will
. eventually sum to 1. Whether the
. sum of the weights converges on 1
quickly or slowly depends on the
. smoothing constant chosen.
Total = 1.000
Similar to the Naïve Model if  equals .9
Holt’s Exponential Smoothing:
• Used for data exhibiting some trend over time

• Is just as simple to apply as simple smoothing

• Involves two smoothing factors, a simple smoothing factor and a

trend smoothing factor
Holt’s Exponential Smoothing:
Ft+1  αA t + 1- α  Ft  Tt 
Tt+1  γ  Ft+1 - Ft   1 – γ  Tt
HFt+m  Ft+1  mTt+1
where:
Ft +1  Smoothed value for period t  1
  Smoothing constant for the level  0    1
X t  Actual value now  in period t 
Ft  Forecast  that is,smoothed  value for time period t
Tt 1  Trend estimate
  Smoothing constant for the trend estimate  0    1
m  Number of periods ahead to be forecast
H t  m  Holt’s forecast value for period t  m
MANAGING SEASONALITY
Holt Winter’s Exponential Smoothing
• Adjusts for both trend and seasonality

• Is just as simple to apply as simple smoothing

• Involves the use of three smoothing parameters, a simple

smoothing parameter, a trend smoothing parameter and a
seasonality smoothing parameter
Holt Winter’s Exponential Smoothing
Ft  α  A t /St-p   1- a  Ft-1  Tt-1 
St  β  A t /Ft   1 – b  St-p
Tt  γ  Ft – Ft-1   1- γ  Tt-1
WFt  m   Ft  m Tt  St  m-p
where:
Ft  Smoothed value for period t
  Smoothing constant for the level  0    1
X t  Actual value now  in period t 
Ft 1  Average experience of series smoothed to period t  1
Tt 1  Trend estimate
St  Seasonality estimate
  Smoothing constant for seasonality estimate  0    1
  Smoothing constant for trend estimate  0    1
m  Number of periods in the forecast lead period
p  Number of periods in the seasonal cycle
Wt  m  Winters’ forecast formperiods into the future
 Holt Winter’s Exponential Smoothing
• Initial Values if Data is quarterly:
• Level is
A5/S1

• Trend is (A5/S1) - (A4/S4)

Decomposing a Time Series
DECOMPOSING TIME SERIES
• Decomposition methods usually try to identify two separate components of the basic
underlying pattern that tend to characterize economic and business series. These are the
trend-cycle and the seasonal factors.
• The seasonal factor relates to periodic fluctuations of constant length that are caused by
such things as temperature, rainfall, month of the year, timing of holidays, and corporate
policies.
• The trend-cycle represents longer-term changes in the level of trend-cycle the series.
• The trend-cycle is sometimes separated into trend and cyclical components, but the
distinction is somewhat artificial and most decomposition procedures leave the trend
and cycle as a single component known as the trend-cycle
DECOMPOSING TIME SERIES
• Time series data can exhibit a variety of patterns, and it is often helpful to
split a time series into several components, each representing an underlying
pattern category.
• Three types of time series patterns: trend, seasonality and cycles.
• When we decompose a time series into components, we usually combine
the trend and cycle into a single trend-cycle component (sometimes called
the trend for simplicity).
• Thus we think of a time series as comprising three components: a trend-
cycle component, a seasonal component, and a remainder component
which is the random component (containing anything else in the time
series).
DECOMPOSING TIME SERIES
• Forecaster should consider some common methods for extracting
these components from a time series.

• Often this is done to help improve understanding of the time series,

but it can also be used to improve forecast accuracy.

• Decomposing Method now less popular with development of ARIMA

methodology by Box-Jenkins
Additive vs Multiplicative Series
• If we assume an additive decomposition, then we can write

yt= St+ Tt+Rt

where yt is the data,
St is the seasonal component,
Tt is the trend-cycle component, and
Rt is the remainder component, all at period t.

• The additive decomposition is the most appropriate if the magnitude

of the seasonal fluctuations, or the variation around the trend-cycle,
does not vary with the level of the time series.
Multiplicative Series
• Alternatively, a multiplicative decomposition would be written as
yt = St×Tt×Rt

• When the variation in the seasonal pattern, or the variation around the trend-cycle,
appears to be proportional to the level of the time series, then a multiplicative
decomposition is more appropriate. Multiplicative decompositions are common with
economic time series.
• If the seasonal fluctuations increase and decrease proportionally with increases and
decreases in the level of the series, then a multiplicative model is appropriate.
• Multiplicative decomposition is more prevalent with economic series because most
seasonal economic series do have seasonal variation which increases with the level of
the series
Seasonally adjusted data
• If the seasonal component is removed from the original data,
the resulting values are the “seasonally adjusted” data

• For an additive decomposition, the seasonally adjusted data

are given by yt – St

• For multiplicative data, the seasonally adjusted values are

obtained using yt /St