Statistics for Decision Making

TIME SERIES
FORECASTING

Dr. Rohit Joshi, IIM Shillong

 Outline
• Develop and implement basic forecasting models
• Identify the components present in a time series
• Use smoothing-based forecasting models, including
moving average and exponential smoothing
• Apply trend-based forecasting models, including
linear trend and nonlinear trend
• Complete time-series forecasting of seasonal data
• Compute and interpret basic index numbers
 The Importance of Forecasting
• Governments forecast unemployment, interest rates,
expected revenues from income taxes for policy
purposes
• Marketing executives forecast demand, sales, and
consumer preferences for strategic planning
• College administrators forecast enrollments to plan
for facilities and for faculty recruitment
• Retail stores forecast demand to control inventory
levels, hire employees and provide training
 Time-Series Data

• Numerical data obtained at regular time

intervals
• The time intervals can be annually, quarterly,
daily, hourly, etc.
• Example:
Year: 1999 2000 2001 2002 2003
Sales: 75.3 74.2 78.5 79.7 80.2
 Time-Series Plot
A time-series plot is a two-dimensional plot of time
series data
• the vertical axis U.S. Inflation Rate
measures the variable of 16.00
interest Inflation Rate (%)
14.00
12.00
10.00

• the horizontal axis 8.00

6.00
corresponds to the time 4.00
2.00
periods 0.00
1975

1977

1979

1983

1987

1991

1993

1995

1999
1981

1985

1989

1997

2001
Y ear
 Time-Series Components

Time Series

Trend Seasonal Cyclical Irregular

Component Component Component Component
 Trend Component
• Long-run increase or decrease over time (overall
upward or downward movement)
• Data taken over a long period of time

rd trend
Sales Upwa

Time
 Trend Component
• Trend can be upward or downward
• Trend can be linear or non-linear
Sales Sales

Time Time
Downward linear trend Upward nonlinear trend
 Seasonal Component
• Short-term regular wave-like patterns
• Observed within 1 year
• Often monthly or quarterly
Sales
Summer
Winter
Summer
Winter Spring Fall

Spring Fall

Time (Quarterly)
 Cyclical Component
• Long-term wave-like patterns
• Regularly occur but may vary in length
• Often measured peak to peak or valley to
valley 1 Cycle

Sales

Year
 Irregular Component
• Unpredictable, random, “residual”
fluctuations
• Due to random variations of
– Nature
– Accidents or unusual events
• “Noise” in the time series
 Multiplicative Time-Series Model with a
Seasonal Component
• Used primarily for forecasting
• Allows consideration of seasonal variation

Yi  Ti  Si  Ci  Ii
where Ti = Trend value at time i
Si = Seasonal value at time i
Ci = Cyclical value at time i
Ii = Irregular (random) value at time i
Multiplicative Time-Series Model for Annual
Data

 Used primarily for forecasting

 Observed value in time series is the product of
components

Yi  Ti  Ci  Ii
where Ti = Trend value at year i
Ci = Cyclical value at year i
Ii = Irregular (random) value at year i
 Smoothing the
Annual Time Series

• Calculate moving averages to get an

overall impression of the pattern of
movement over time

Moving Average: averages of consecutive

time series values for a
chosen period of length L
 Moving Averages

• Used for smoothing

• A series of arithmetic means over time
• Result dependent upon choice of L (length of
period for computing means)
• Examples:
– For a 5 year moving average, L = 5
– For a 7 year moving average, L = 7
– Etc.
 Moving Averages …
• Example: Five-year moving average

– First average:
Y1  Y2  Y3  Y4  Y5
MA(5) 
5

– Second average:
Y2  Y3  Y4  Y5  Y6
MA(5) 
5

– etc.
 Example: Annual Data
Year Sales Annual Sales
1 23
60
2 40
50
3 25
4 27 40 …
5 32 Sales 30
6 48
20
7 33
8 37 10

9 37 0
1 2 3 4 5 6 7 8 9 10 11
…
10 50
11 40 Year

etc… etc…
 Calculating Moving Averages
5-Year
Average Moving
Year Sales Year Average
1 23 1 2  3  4  5
3 29.4 3
2 40 5
4 34.4
3 25 5 33.0 23  40  25  27  32
29.4 
4 27 6 35.4 5
5 32 7 37.4
6 48 8 41.0
7 33 9 39.4
8 37 … …
9 37 etc…
10 50
• Each moving average is for a consecutive
11 40
block of 5 years
 Annual vs. Moving Average

• The 5-year Annual vs. 5-Year Moving Average

moving average 60
smoothes the 50
data and shows 40
the underlying
Sales

30
trend 20
10
0
1 2 3 4 5 6 7 8 9 10 11
Year

Annual 5-Year Moving Average

 Exponential Smoothing

• A weighted moving average

– Weights decline exponentially
– Most recent observation weighted most

• Used for smoothing and short term

forecasting (often one period into the
future)
 Exponential Smoothing
• The weight (smoothing coefficient) is W
– Subjectively chosen
– Range from 0 to 1
– Smaller W gives more smoothing, larger W
gives less smoothing
• The weight is:
– Close to 0 for smoothing out unwanted
cyclical and irregular components
– Close to 1 for forecasting
 Exponential Smoothing Model

Fi 1  WDi  (1  W ) Fi
For i = 2, 3, 4, …

where:
Fi+1 = exponentially smoothed forecast value for period i+1
Fi = exponentially smoothed value already
computed for period i
Di = actual observed value in period i
W = weight (smoothing coefficient), 0 < W < 1
 Exponential Smoothing Example
• Suppose we use weight W = .2
Time Forecast
Sales Exponentially Smoothed
Period from prior
(Yi) Value for this period (Ei)
(i) period (Fi-1)
1 23 -- 23 F1 = Y1 since
2 40 23 (.2)(40)+(.8)(23)=26.4 no prior
information
3 25 26.4 (.2)(25)+(.8)(26.4)=26.12
exists
4 27 26.12 (.2)(27)+(.8)(26.12)=26.296
5 32 26.296 (.2)(32)+(.8)(26.296)=27.437
6 48 27.437 (.2)(48)+(.8)(27.437)=31.549 Ei 
7 33 31.549 (.2)(48)+(.8)(31.549)=31.840 WYi  (1  W )Ei1
8 37 31.840 (.2)(33)+(.8)(31.840)=32.872
9 37 32.872 (.2)(37)+(.8)(32.872)=33.697
10 50 33.697 (.2)(50)+(.8)(33.697)=36.958
etc. etc. etc. etc.
 Sales vs. Smoothed Sales
Fluctuations have been smoothed

60

50

40
Sales

30

20

10

0
1 2 3 4 5 6 7 8 9 10
Time Period
Sales Smoothed
 Trend-Based Forecasting
• Estimate a trend line using regression analysis
Time  Use time (X) as the
Year Period Sales independent variable:
(X) (Y)
1999 0 20
2000 1 40 Ŷ  b0  b1X
2001 2 30
2002 3 50
2003 4 70
2004 5 65
 Trend-Based Forecasting
• The linear trend forecasting equation is:
Time
Year Period Sales Ŷi  21.905  9.5714 Xi
(X) (Y) Sales trend
1999 0 20
80
2000 1 40 70
60
2001 2 30 50
sales

2002 3 50 40
30
2003 4 70 20
10
2004 5 65 0
0 1 2 3 4 5 6

Year
 Trend-Based Forecasting
(continued)
• Forecast for time period 6:
Time
Period
Ŷ  21.905  9.5714 (6)
Year Sales
(X) (y)  79.33
1999 0 20
2000 1 40
80
2001 2 30 70

2002 3 50 60
50
sales

2003 4 70 40
30
2004 5 65 20
10
2005 6 ?? 0
0 1 2 3 4 5 6

Year
 Nonlinear Trend Forecasting
• A nonlinear regression model can be used when the
time series exhibits a nonlinear trend
• Quadratic form is one type of a nonlinear model:

Yi  0  1Xi  2 X  i2
i

• Compare adj. r2 and standard error to that of linear

model to see if this is an improvement
• Can try other functional forms to get best fit
 A problem
• Whirly world, specializes in producing recreational
equipments. To forecast future sales based on its
analysis of its past pattern of sales the firm has
collected the information as below. As a sales
planning manager forecast the sale for upcoming
quarters.
Sales per quarter (X $ 10,000)
Q1 Q2 Q3 Q4
2010 16.0 21.0 9.0 18.0 15.0 20.0 10.0 18.0
2011 17.0 24.0 13.0 22.0 17.0 25.0 11.0 21.0
2012 18.0 26.0 14.0 25.0
2013
2014
 Pitfalls in
Time-Series Analysis
• Assuming the mechanism that governs the
time series behavior in the past will still hold
in the future
• Using mechanical extrapolation of the trend to
forecast the future without considering
personal judgments, business experiences,
changing technologies, and habits, etc.