Basic Business

Statistics

Time-Series Forecasting
 The Importance of Forecasting

 Governments forecast unemployment rates,

interest rates, and 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
 Common Approaches to Forecasting

Common Approaches
to Forecasting

Qualitative forecasting Quantitative forecasting

methods methods

 Used when historical data

are unavailable Time Series Causal
 Considered highly
subjective and  Time Series - use past
judgmental data or time to predict
future values
 Causal – factors that are
not past data or time
 Time-Series Data

 Numerical data obtained at regular time intervals

 The time intervals can be annually, quarterly,

monthly, weekly, daily, hourly, etc.

 Example:

Year: 2019 2020 2021 2022 2023

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

U.S. Inflation Rate

16.00
 the vertical axis 14.00

measures the variable 12.00

10.00
of interest
8.00

 the horizontal axis 6.00

4.00
corresponds to the
2.00
time periods 0.00
 Time-Series Components

Time Series

Trend Seasonal Cyclical Irregular

Component Component Component Component

Overall, Regular periodic Repeating Erratic or

persistent, long- fluctuations, swings or residual
term movement usually within a movements over fluctuations
12-month period more than one
year
 Trend Component

Long-run increase or decrease over time

(overall upward or downward movement)

Data taken over a long period of time

Sales

Time
 Trend Component

(continued)
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 trough to
trough
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

 Does Your Time Series Have A Trend
Component?

A time series plot should help you to answer

this question.

Often it helps if you “smooth” the time series

data to help answer this question.

Two popular smoothing methods are moving

averages and exponential smoothing.
 Smoothing Methods

Moving Averages
 Calculate moving averages to get an overall
impression of the pattern of movement over time
 Averages of consecutive time series values for a
chosen period of length L

Exponential Smoothing
 A weighted moving average
 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

(continued)
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
2 40 60
3 25
50
4 27
5 32 40
Sales

6 48 30
7 33 20
8 37
10
9 37
10 50 0
1 2 3 4 5 6 7 8 9 10 11
11 40
Year
etc… etc…
 Calculating Moving Averages

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

Annual vs. 5-Year Moving Average

The 5-year 60
moving average
50
smoothes the
40
data and makes
Sales

it easier to see 30

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

Annual 5-Year Moving Average

 Exponential Smoothing

Used for smoothing and short term

forecasting (one period into the future)
 Exponential Smoothing Model

 Exponential smoothing model

E1 = Y1
Ei = WYi + (1 − W )Ei−1
For i = 2, 3, 4, …
where:
Ei = exponentially smoothed value for period i
Ei-1 = exponentially smoothed value already
computed for period i - 1
Yi = observed value in period i
W = weight (smoothing coefficient), 0 < W < 1
 Exponential Smoothing Example

 Suppose we use weight W = 0.2

Forecast
Time Sales Exponentially Smoothed
from prior
Period (i) (Yi) Value for this period (Ei)
period (Ei-1)
E1 = Y1
1 23 -- 23
since no
2 40 23.000 (.2)(40)+(.8)(23)=26.4 prior
3 25 26.400 (.2)(25)+(.8)(26.4)=26.12 information
4 27 26.120 (.2)(27)+(.8)(26.12)=26.296 exists
5 32 26.296 (.2)(32)+(.8)(26.296)=27.437 Ei =
6 48 27.437 (.2)(48)+(.8)(27.437)=31.549
WYi + (1 − W )Ei−1
7 33 31.549 (.2)(33)+(.8)(31.549)=31.840
8 37 31.840 (.2)(37)+(.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
 NOTE: the
40
smoothed value in Sales
30
this case is
generally a little 20
low, since the 10
trend is upward
0
sloping and the 1 2 3 4 5 6 7 8 9 10
weighting factor is Time Period
only .2 Sales Smoothed
 Exponential Smoothing
(continued)
The weight (smoothing coefficient) is W
 Subjectively chosen
 Ranges 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
 Forecasting Time Period i + 1

 The smoothed value in the current

period (i) is used as the forecast value for
next period (i + 1) :

Ŷi+1 = Ei
 Three Popular Methods
for Trend Fitting & Forecasting

Linear Trend Forecasting

Nonlinear Trend Forecasting

Autoregressive Modeling
 Linear Trend Forecasting

Estimate a trend line using regression analysis

Time  Use time (X) as the

Year Period Sales independent variable:
(X) (Y)

Ŷ = b0 + b1X
2017 0 20
2018 1 40
2019 2 30 In least squares linear, non-linear, and
2020 3 50 exponential modeling, time periods are
numbered starting with 0 and increasing
2021 4 70 by 1 for each time period.
2022 5 65
 Linear Trend Forecasting
(continued)
The linear trend forecasting equation
is:
Time
Year Period Sales Ŷi = 21.905 + 9.5714 Xi
(X) (Y)
Sales trend
2017 0 20
80
2018 1 40 70
60
2019 2 30 50
sales

40
2020 3 50 30
20
2021 4 70 10
0
2022 5 65
0 1 2 3 4 5 6

Year
 Linear Trend Forecasting
(continued)
 Forecast for time period 6 (2023):

Ŷ = 21.905 + 9.5714 (6)

Time
Year Period Sales
= 79.33
(X) (y)
Sales trend
2017 0 20
2018 1 40 80
70
2019 2 30 60
50
2020 3 50
sales

40
2021 4 70 30
20
2022 5 65 10
0
2023 6 ??
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 = β0 + β1Xi + β2 X + εi
2
i

Compare adj. r2 to that of linear model to see if

this is an improvement or test significance of
the quadratic term
Can try other functional forms to get best fit
 Autoregressive Modeling

Used for forecasting

Takes advantage of autocorrelation

 1st order - correlation between consecutive values
 2nd order - correlation between values 2 periods apart

pth order Autoregressive model:

Yi = A 0 + A 1Yi-1 + A 2 Yi-2 +  + A p Yi-p + δi
Random
Error
 Autoregressive Model: Example

The Office Concept Corp. has acquired a number of office

units (in thousands of square feet) over the last eight years.
Develop the second order Autoregressive model.

Year Units
15 4
16 3
17 2
18 3
19 2
20 2
21 4
22 6
 Autoregressive Model: Example Solution

 Develop the 2nd order Year Yi Yi-1 Yi-2

table 15 4 -- --
16 3 4 --
 Use Excel to estimate a 17 2 3 4
regression model 18 3 2 3
19 2 3 2
Excel Output 20 2 2 3
Coefficients 21 4 2 2
Intercept 3.5 22 6 4 2
X Variable 1 0.8125
X Variable 2 -0.9375

Ŷi = 3.5 + 0.8125Yi−1 − 0.9375Yi−2

 Autoregressive Model Example:
Forecasting

Use the second-order equation to forecast

number of units for 2023:

Ŷi = 3.5 + 0.8125Yi −1 − 0.9375Yi − 2

Ŷ2023 = 3.5 + 0.8125(Y2022 ) − 0.9375(Y2021 )
= 3.5 + 0.8125(6) − 0.9375(4)
= 4.625
 Autoregressive Modeling Steps

1. Choose p (note that df = n – 2p – 1)

2. Form a series of “lagged predictor” variables
Yi-1 , Yi-2 , … ,Yi-p
3. Use Excel to run regression model using all p
variables
4. Test significance of Ap
 If null hypothesis rejected, this model is selected
 If null hypothesis not rejected, decrease p by 1 and
repeat