Chapter 5

TIME SERIES ANALYSIS AND

FORECASTING
The Importance of
Forecasting
 Governments forecast unemployment,
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
 Manufacturing firms forecast demand for
their products for necessary manpower and
raw materials
 Introduction
3

 Forecasting methods can be classified as

qualitative or quantitative
 Qualitative methods generally involve the
use of expert judgment to develop
forecasts
 Quantitative forecasting methods can be
used when:
 Past information about the variable being
forecasted is available
 The information can be quantified
 It is reasonable to assume that past is prologue
 Introduction
4

 The objective of time series analysis is to

uncover a pattern in the time series and then
extrapolate the pattern into the future
 The forecast is based solely on past values of
the variable and/or on past forecast errors
 Modern data-collection technologies have
enabled individuals, businesses, and
government agencies to collect vast amounts
of data that may be used for causal
forecasting
Time Series Patterns
 Horizontal Pattern
 Trend Pattern

 Seasonal Pattern

 Trend and Seasonal

Pattern
 Cyclical Pattern

 Identifying Time

Series Pattern
 Time Series Patterns
6

 Time series: a sequence of observations on a

variable measured at successive points in time
or over successive periods of time
 The measurements may be taken every hour,
day, week, month, year, or any other regular
interval. The pattern of the data is important
to understand the series’ past behavior
 If the behavior of the times series data of the
past is expected to continue in the future, we
can use it to guide us in selecting an
appropriate forecasting method
 Time Series Patterns
7

Horizontal Pattern
 Exists when the data fluctuate randomly around a

constant mean over time

 Stationary time series: denotes a time series

whose statistical properties are independent of time

 The process generating the data has a constant mean
 The variability of the time series is constant over time
 A time series plot for a stationary time series will
always exhibit a horizontal pattern with random
fluctuations
Example
 Consider the data show the number of gallons
of gasoline (in 1000s) sold by a gasoline
distributor in Bennington, Vermont, over the
past 12 weeks. The average value or mean for
this time series is 19.25 or 19,250 gallons per
week. Figure 5.1 shows a time series plot for
these data. Note how the data fluctuate
around the sample mean of 19,250 gallons.
 Although random variability is present, we
would say that these data follow a horizontal
pattern.
 Table 5.1: Gasoline Sales Time
Series
9
 Figure 5.1: Gasoline Sales Time
Series Plot
10
Example
 Suppose the gasoline distributor signs a contract
with the Vermont State Police to provide gasoline
for state police cars located in southern Vermont
beginning in week 13. With this new contract,
the distributor naturally expects to see a
substantial increase in weekly sales starting in
week 13. Table 5.2 shows the number of gallons
of gasoline sold for the original time series and
the ten weeks after signing the new contract.
Figure 5.2 shows the corresponding time series
plot. Note the increased level of the time series
beginning in week 13.
 Table 5.2: Gasoline Sales Time
Series after Obtaining the
12 Contract with the Vermont
State Police
 Figure 5.2: Gasoline Sales Time
Series Plot after Obtaining the
13 Contract with the Vermont State
Police
 Time Series Patterns
14

Trend Pattern
 A trend pattern is gradual shifts or movements

to relatively higher or lower values over a longer

period of time
 A trend is usually the result of long-term factors

such as:
 Population increases or decreases
 Shifting demographic characteristics of the
population
 Improving technology
 Changes in the competitive landscape
 Changes in consumer preferences
Example
 Consider the time series of bicycle sales
for a particular manufacturer over the
past ten years, as shown in Table 5.3 and
Figure 5.3
 Table 5.3: Bicycle Sales Time Series
16
 Figure 5.3: Bicycle Sales Time
Series Plot
17
Example
 Data in Table 5.4 and the corresponding time
series plot in Figure 5.4 show the sales revenue
for a cholesterol drug since the company won
FDA approval for the drug ten years ago. The
time series increases in a nonlinear fashion; that
is, the rate of change of revenue does not
increase by a constant amount from one year to
the next. In fact, the revenue appears to be
growing in an exponential fashion
 Table 5.4: Cholesterol Drug
Revenue Time Series
19 Figure 5.4: Cholesterol Drug
Revenue Times Series Plot ($
millions)
 Time Series Patterns
20

Seasonal Pattern
 Seasonal patterns are recurring patterns over
successive periods of time
 Seasonal patterns are short-term cyclic fluctuations
 Example: A manufacturer of swimming pools expects low
sales activity in the fall and winter months, with peak
sales in the spring and summer months to occur every
year
 Time series plot not only exhibits a seasonal pattern
over a one-year period but also for less than one year
in duration
 Example: daily traffic volume shows within-the-day
“seasonal” behavior
Example
 Consider the number of umbrellas sold
at a clothing store over the past five
years. Table 5.5 shows the time series
and Figure 5.5 shows the corresponding
time series plot
Time Series Patterns
Figure 5.5:
Table 5.5: Umbrella Sales
Umbrella Sales Time
Time Series
Series Plot
Example
 The data in Table 5.6 and the
corresponding time series plot in Figure
5.6 show quarterly smartphone sales for
a particular manufacturer over the past
four years.
 Time Series Patterns
24

Trend and Seasonal Pattern

 Some time series include both a trend

and a seasonal pattern

Table 5.6: Quarterly Figure 5.6: Quarterly
Smartphone Sales Time Smartphone Sales Time
Series Series Plot
 Time Series Patterns
25

Cyclical Pattern
 A cyclical pattern exists if the time series plot

shows an alternating sequence of points below

and above the trendline that lasts for more than
one year
 Example: Periods of moderate inflation followed by
periods of rapid inflation can lead to a time series
that alternates below and above a generally
increasing trendline
 Cyclical effects are often combined with long-
term trend effects and referred to as trend-cycle
effects
 Time Series Patterns
26

Identifying Time Series Patterns

 The underlying pattern in the time series

is an important factor in selecting a

forecasting method
 A time series plot should be one of the

first analytic tools

 We need to use a forecasting method

that is capable of handling the pattern

exhibited by the time series effectively
Forecast Accuracy
 Forecast Error
 Mean Forecast Error (MFE)

 Mean Absolute Error (MAE)

 Mean Squared Error (MSE)

 Mean Absolute Percentage Error (MAPE)

 Table 5.7: Computing Forecasts
and Measures of Forecast
28 Accuracy Using the Most
Recent Value as the Forecast
for the Next Period
Forecast Accuracy
 Naïve forecasting method: Using the most
recent data to predict future data
 The key concept associated with measuring
forecast accuracy is forecast error
 Measures to determine how well a particular
forecasting method is able to reproduce the time
series data that are already available
 Forecast error
 Mean forecast error
 Mean absolute error (MAE)
 Mean squared error (MSE)
 Mean absolute percentage error (MAPE)
 Forecast Accuracy
30

Forecast Error: Difference between the actual

and the forecasted values for period t.

Mean Forecast Error: Mean or average of

the forecast errors.

MFE =
Forecast Accuracy

Mean Absolute Error (MAE): Measure of forecast

accuracy that avoids the problem of positive and
negative forecast errors offsetting one another.

MAE =

Mean Squared Error (MSE): measure that avoids the

problem of positive and negative errors offsetting each
other is obtained by computing the average of the
squared forecast errors.

MFE =

31
Forecast Accuracy

Mean Absolute Percentage Error (MAPE): Average of the

absolute value of percentage forecast errors.

MAPE =

32
 Table 5.8: Computing Forecasts
and Measures of Forecast
33 Accuracy Using the Average of
All the Historical Data as the
Forecast for the Next Period
 Forecast Accuracy
34

 Compare the accuracy of the two forecasting

methods by comparing the values of MAE,
MSE, and MAPE for each method
Naïve Average of Past
Method Values
MAE 3.73 2.44
MSE 16.27 8.10
MAPE 19.24% 12.85%

 The average of past values provides more

accurate forecasts for the next period than
using the most recent observation
Moving Averages and Exponential
Smoothing
 Moving Averages
 Forecast Accuracy

 Exponential Smoothing

 Forecast Accuracy
 Moving Averages and
36
Exponential Smoothing
Moving Averages
 Moving averages method: Uses the

average of the most recent k data values

in the time series as the forecast for the
next period
 Moving average forecast:

=
 Table 5.9: Summary of Three-
Week Moving Average
37 Calculations
 Figure 5.7: Gasoline Sales
Time Series Plot and Three-
38 Week Moving Average
Forecasts
 Figure 5.8: Data Analysis
39
Dialog Box
 Figure 5.9: Moving Average
Dialog Box
40
 Figure 5.10: Excel Output for
Moving Average Forecast for
41 Gasoline Data
 Moving Averages and
42
Exponential Smoothing
Forecast Accuracy
The values of the three measures of
forecast accuracy for the three-week
moving average calculations in Table
8.9:
MAE = = = 2.67

MSE = = = 10.22

MAPE = = = 14.36%
 Moving Averages and
43
Exponential Smoothing
Exponential Smoothing
 Exponential smoothing uses a

weighted average of past time series

values as a forecast
 Exponential Smoothing Forecast:

= α + (1 – α)
 Smoothing constant (α )is the weight
given to the actual value in period t;
weight given to the forecast in period t
is 1 –α.
 Moving Averages and
44
Exponential Smoothing
Illustration: Consider a time series involving only
three periods of data: y1, y2, and y3.
 Let equal the actual value of the time series
in period 1; that is, = y1
 Hence, the forecast for period 2 is:

= α + (1 – α)

= α + (1 – α)

=
 Table 5.10: Summary of the
Exponential Smoothing Forecasts
45 and Forecast Errors for the
Gasoline Sales Time Series with
Smoothing Constant a = 0.2
 Figure 5.11: Actual and
Forecast Gasoline Time
46 Series with Smoothing
Constant α = 0.2
 Figure 5.13: Exponential
47
Smoothing Dialog Box
 Figure 5.14: Excel Output for
Exponential Smoothing
48 Forecast for Gasoline Data
 Moving Averages and
49
Exponential Smoothing
Forecast Accuracy
 Insight into choosing a good value for

acan be obtained by rewriting the basic

exponential smoothing model as:

 If the time series contains substantial

random variability, a small value of the
smoothing constant is preferred and
vice-versa
 Choose the value of a that minimizes
the MSE
Using Regression Analysis for
Forecasting
 Linear Trend Projection
 Seasonality
 Seasonality without Trend
 Seasonality with Trend
Using Regression Analysis as a
Causal Forecasting Method
Combining Causal Variables with
Trend and Seasonality Effects
 Considerations in Using Regression
in Forecasting
 Using Regression Analysis
51
for Forecasting
Linear Trend Projection
 Regression analysis can be used to forecast a time
series with a linear trend
 Simple linear regression analysis yields the linear
relationship between the independent variable and the
dependent variable that minimizes the MSE
 Use this approach to find a best-fitting line to a set of
data that exhibits a linear trend
 The variable to be forecasted (y, the actual value of
the time series period t) is dependent variable
 Trend variable (time period t) is the independent
variable
 Equation for the trendline: = t
 Figure 5.15: Excel Simple
Linear Regression Output for
52
Trendline Model for Bicycle
Sales Data
 Using Regression
53
Analysis for Forecasting
 Trend equation for the bicycle sales time
series:
= 20.4 + 1.1t
 Substituting t = 11 into the above
equation yields next year’s trend
projection,
= 20.4 + 1.1 (11) = 32.5
 Thus, the linear trend model yields a
sales forecast of 32,500 bicycles for the
next year
 Using Regression
54
Analysis for Forecasting
 We can also use more complex
regression models to fit nonlinear trends
= t+

 Autoregressive models: Regression

models such as this in which the
independent variables are previous
values of the time series
=
 Using Regression
55
Analysis for Forecasting
Seasonality Without Trend
 We can model a time series with a seasonal pattern by

treating the season as a dummy variable

 Illustration:

 Consider the data on the number of umbrellas sold in

Table 5.5
 The time series plot corresponding to this data in Figure
5.5 do not suggest any long-term trend in sales
 Closer inspection of the time series plot suggests that a
quarterly seasonal pattern is present
 k - 1 dummy variables are required to model a categorical
variable that has k levels
 Thus, to model the seasonal effects in the umbrella time
series we need 4 – 1 = 3 dummy variables
 Using Regression
56
Analysis for Forecasting
 Illustration (continued):
 The three dummy variables can be coded as

follows: 1 if period t is a quarter 1

0 otherwise
Qtr1t =
1 if period t is a quarter 2
0 otherwise
Qtr2t =
1 if period t is a quarter 3
0 otherwise
Qtr3t =

 General form of the equation relating the

number of umbrellas sold to the quarter the
salesyˆ ttake place
b0  b1Qtr1t  b2Qtr 2t  b3Qtr 3t
 Table 5.11: Umbrella Sales
Time Series with Dummy
57 Variables
 Using Regression
58
Analysis for Forecasting
We can use a multiple linear regression model to
find the values of b0, b1, b2, and b3 that minimize
the sum of squared errors. Using the data in
Table 5.11 and regression analysis, we obtain the
following regression equation:
yˆt 95.0  29.0Qtr1t  57.0Qtr 2 t  26.0Qtr 3t
Forecast sales of every quarter for next year
Quarter 1: Sales = 95.0 + 29.0 (1) + 57.0 (0) + 26.0
(0) = 124
Quarter 2: Sales = 95.0 + 29.0 (0) + 57.0 (1) + 26.0
(0) = 152
Quarter 3: Sales = 95.0 + 29.0 (0) + 57.0 (0) + 26.0
(1) = 121
 Using Regression
59
Analysis for Forecasting
Seasonality with Trend
 The time series contains both seasonal

effects and a linear trend

 Consider the data for the smartphone time

series in Table 8.6

 The time series plot corresponding to this data
indicates that there is both linear trend and
seasonal pattern
 The general form of the regression equation
takes the form
=t
 Table 5.12: Smartphone Sales
Time Series with Dummy
60
Variables and Time Period
 Using Regression Analysis
61
for Forecasting
 Using the data in Table 5.12 with the
regression model that includes both the
seasonal and trend components, we
obtain the following equation that
minimizes our sum of squared errors:
yˆt 6.07  1.36Qtr1t  2.03Qtr 2t  0.304Qtr 3t  0.146t
 Using Regression Analysis
62
for Forecasting
 The dummy variables in the equation for
Smartphone Sales time series provide
four equations given time period t
corresponds to quarters 1, 2, 3, and 4
 Quarter 1: Sales = 4.71 + 0.146t
 Quarter 2: Sales = 4.04 + 0.146t
 Quarter 3: Sales = 5.77 + 0.146t
 Quarter 4: Sales = 6.07 + 0.146t
 Using Regression Analysis
63
for Forecasting
 Forecast quarterly sales for next year. Next
year is year 5 for the smartphone sales time
series, that is, time periods 17, 18, 19, and 20.
 Quarter 1, Year 5: Sales = 4.71 + 0.146(17) = 7.19
 Quarter 2, Year 5: Sales = 4.04 + 0.146 (18) = 6.67
 Quarter 3, Year 5: Sales = 5.77 + 0.146 (19) = 8.54
 Quarter 4, Year 5: Sales = 6.07 + 0.146 (20) = 8.99
 Using Regression
64
Analysis for Forecasting
Using Regression Analysis as a Causal
Forecasting Method
 The relationship of the variable to be forecast

with other variables may also be used to

develop a forecasting model
 Advertising expenditures when sales is to be
forecast
 The mortgage rate when new housing construction
is to be forecast
 Causal models: Models that include only
variables that are believed to cause changes in
the variable to be forecast
 Table 5.13: Student
Population and Quarterly
65
Sales data for 10 Armand’s
Pizza Parlors
 Figure 5.16: Scatter Chart of
Student Population and
66 Quarterly Sales for Armand’s
Pizza Parlors
 Figure 5.17: Graph of the
Estimated Regression
67 Equation for Armand’s Pizza
Parlors: y = 60 + 5x
 Using Regression
68
Analysis for Forecasting
Combining Causal Variables with Trend and
Seasonality Effects
 Regression models are very flexible and can

incorporate both causal variables and time series

effects
Considerations in Using Regression in Forecasting
 Whether a regression approach provides a good

forecast depends largely on:

 How well we are able to identify and obtain data for
independent variables that are closely related to the time
series
 Part of the regression analysis procedure should focus on
the selection of the set of independent variables that
provides the best forecasting model
Determining the Best Forecasting
Model to Use
 Determining the Best
Forecasting
70 Model to Use
 A visual inspection can indicate whether
seasonality appears to be a factor and
whether a linear or nonlinear trend
seems to exist
 For causal modeling, scatter charts can
indicate whether strong linear or
nonlinear relationships exist between
the independent and dependent
variables
 If certain relationships appear totally
random, this may lead you to exclude
 Determining the Best
Forecasting
71 Model to Use
 While working with large data sets it is
recommended to divide your data into training
and validation sets
 Based on the errors produced by the different
models for the validation set, we can pick the
model that minimizes some forecast error
measure, such as MAE, MSE or MAPE
 There are software packages that will
automatically select the best model to use
 Ultimately the user should decide which model to
use based on the software output and his
managerial knowledge