The components of a time series

Finding the trend

Forecasting

Business Decision Making II

Time Series Analysis and Forecasting

Dr. Nguyen Ngoc Phan

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Finding the trend
Forecasting

Outline

1 The components of a time series

Time series components
Time series models

2 Finding the trend

Moving averages
Finding the seasonal variations

3 Forecasting
Forecasting by extrapolation
Deseasonalization

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Time series components
Finding the trend
Time series models
Forecasting

The components of a time series

Definition
A time series is a series of figures or values recorded over time.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Time series components
Finding the trend
Time series models
Forecasting

The components of a time series

Definition
A time series is a series of figures or values recorded over time.

Following is an example of time series.

Year Sales ($’000)
2000 20
2001 21
2002 24
2003 23
2004 27
2005 30
2006 28

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Time series components
Finding the trend
Time series models
Forecasting

Four components of time series

1 A trend
2 Seasonal variations
3 Cyclical variations
4 Random variations

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Time series components
Finding the trend
Time series models
Forecasting

Definition
A trend is the continuous long-term movement over time in
the values of the data recorded.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Time series components
Finding the trend
Time series models
Forecasting

Definition
A trend is the continuous long-term movement over time in
the values of the data recorded.
Example: The time series rises or falls continuously. The
trend, represented by the red line, indicates an upward
movement in value.

Figure: The trend

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Time series components
Finding the trend
Time series models
Forecasting

There are three types of trend in the following example.

Year Output/labor hour Cost/unit ($) # of employees
2004 30 1 100
2005 24 1.08 103
2006 26 1.2 96
2007 22 1.15 102
2008 21 1.18 103
2009 17 1.25 98
(A) (B) (C)

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Time series components
Finding the trend
Time series models
Forecasting

There are three types of trend in the following example.

Year Output/labor hour Cost/unit ($) # of employees
2004 30 1 100
2005 24 1.08 103
2006 26 1.2 96
2007 22 1.15 102
2008 21 1.18 103
2009 17 1.25 98
(A) (B) (C)

1 In time series (A): a downward trend

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Time series components
Finding the trend
Time series models
Forecasting

There are three types of trend in the following example.

Year Output/labor hour Cost/unit ($) # of employees
2004 30 1 100
2005 24 1.08 103
2006 26 1.2 96
2007 22 1.15 102
2008 21 1.18 103
2009 17 1.25 98
(A) (B) (C)

1 In time series (A): a downward trend

2 In time series (B): an upward trend

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Time series components
Finding the trend
Time series models
Forecasting

There are three types of trend in the following example.

Year Output/labor hour Cost/unit ($) # of employees
2004 30 1 100
2005 24 1.08 103
2006 26 1.2 96
2007 22 1.15 102
2008 21 1.18 103
2009 17 1.25 98
(A) (B) (C)

1 In time series (A): a downward trend

2 In time series (B): an upward trend
3 In time series (C): the trend is static around 100.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Time series components
Finding the trend
Time series models
Forecasting

Definition
Seasonal variations are movements in the time series that
reoccur each year about the same time due to the change in
the seasons.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Time series components
Finding the trend
Time series models
Forecasting

Definition
Seasonal variations are movements in the time series that
reoccur each year about the same time due to the change in
the seasons.
Example:
Each year the unemployment rate tends to goes up in
May when students enter the summer job market, and
goes down in November when retail stores hire temporary
help to handle the Chrismas rush.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Time series components
Finding the trend
Time series models
Forecasting

Definition
Seasonal variations are movements in the time series that
reoccur each year about the same time due to the change in
the seasons.
Example:
Each year the unemployment rate tends to goes up in
May when students enter the summer job market, and
goes down in November when retail stores hire temporary
help to handle the Chrismas rush.
Sales might be higher on Friday and Saturday than on
Monday.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Time series components
Finding the trend
Time series models
Forecasting

Definition
Seasonal variations are movements in the time series that
reoccur each year about the same time due to the change in
the seasons.
Example:
Each year the unemployment rate tends to goes up in
May when students enter the summer job market, and
goes down in November when retail stores hire temporary
help to handle the Chrismas rush.
Sales might be higher on Friday and Saturday than on
Monday.
Sales of ice cream will be higher in summer than in winter,
and sales of overcoats will be higher in fall than in spring.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Time series components
Finding the trend
Time series models
Forecasting

An annual repeating seasonal variation, with low timber

production in January, increasing to a high production in April.
There is then a slight decrease in July followed by a slight
increase in October. There is also a basic upward trend.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Time series components
Finding the trend
Time series models
Forecasting

Cyclical variations
Many variables often exhibit a tendency to fluctuate above
and below the long-term trend over a long period of time.
These fluctuations are called cyclical variations. They cover
much longer time periods than do seasonal variations.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Time series components
Finding the trend
Time series models
Forecasting

Figure: NYSE margin debt indicates a cyclical variation and a

parabolic increase

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Time series components
Finding the trend
Time series models
Forecasting

Random variations
Time series also contain random variations caused by unusual
occurrences producing movements which have no discernible
pattern. These movements are unique and unlikely to reoccur
in similar fashion. They can be caused by events such as wars,
floods, earthquakes, political elections, or oil embargoes.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Time series components
Finding the trend
Time series models
Forecasting

Time series models

A time series model can be expressed as some combination of
these four components. Two types of models are commonly
associated with time series: additive model and multiplicative
model. The additve model is expressed as

Y =T +S +C +R

where
Y is the actual time series
T is the trend series
S is the seasonal component
C is the cyclical component
R is the random component.
Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting
 The components of a time series
Time series components
Finding the trend
Time series models
Forecasting

Example: If we were to develop a time series model for sales in

dollars for a local retail store, we might find that

T = $500, S = $100, C = −$25, R = −$10.

Sales would be

Y = 500 + 100 − 25 − 10 = $565.

Notice that S > 0 indicates that existing seasonal influences

had a positive impact on sales. C < 0 suggests that the
business cycle is currently in a downswing. There was some
random event which had a negative impact on sales.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Time series components
Finding the trend
Time series models
Forecasting

To the extent of our course, the additive model excludes the

cyclical component C , which means

Y = T + S + R.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Moving averages
Finding the trend
Finding the seasonal variations
Forecasting

Finding the trend

Three principal methods to find a trend.

1 Inspection: The trend line can be drawn by eye on a graph
in such a way that it appears to lie evenly between the
recorded points, that is, a line of best fit drawn by eye.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Moving averages
Finding the trend
Finding the seasonal variations
Forecasting

Finding the trend

Three principal methods to find a trend.

1 Inspection: The trend line can be drawn by eye on a graph
in such a way that it appears to lie evenly between the
recorded points, that is, a line of best fit drawn by eye.
2 Regression analysis: This method makes the assumption
that the trend line is a straight line. Periods of time are
numbered, and the regression line of the data on those
periods is found. That line is taken to be the trend.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Moving averages
Finding the trend
Finding the seasonal variations
Forecasting

Finding the trend

Three principal methods to find a trend.

1 Inspection: The trend line can be drawn by eye on a graph
in such a way that it appears to lie evenly between the
recorded points, that is, a line of best fit drawn by eye.
2 Regression analysis: This method makes the assumption
that the trend line is a straight line. Periods of time are
numbered, and the regression line of the data on those
periods is found. That line is taken to be the trend.
3 Moving average: This method attempts to remove
seasonal variations by a process of averaging.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Moving averages
Finding the trend
Finding the seasonal variations
Forecasting

Moving averages
Moving average (MA) is a series of arithmetic averages over a
given number of time periods. It is the estimate of the long
run average of the variable.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Moving averages
Finding the trend
Finding the seasonal variations
Forecasting

Moving averages
Moving average (MA) is a series of arithmetic averages over a
given number of time periods. It is the estimate of the long
run average of the variable.

A moving average will have the effect of smoothing out the

data, producing a movement with fewer peaks and valleys, and
hence can be used to find the trend.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Moving averages
Finding the trend
Finding the seasonal variations
Forecasting

Moving averages
Moving average (MA) is a series of arithmetic averages over a
given number of time periods. It is the estimate of the long
run average of the variable.

A moving average will have the effect of smoothing out the

data, producing a movement with fewer peaks and valleys, and
hence can be used to find the trend.
MA is computed by averaging the values in the time series
over a set number of time periods. The same number of time
periods is retained for each average by dropping the oldest
observation and picking up the newest.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Moving averages
Finding the trend
Finding the seasonal variations
Forecasting

Moving averages
Moving average (MA) is a series of arithmetic averages over a
given number of time periods. It is the estimate of the long
run average of the variable.

A moving average will have the effect of smoothing out the

data, producing a movement with fewer peaks and valleys, and
hence can be used to find the trend.
MA is computed by averaging the values in the time series
over a set number of time periods. The same number of time
periods is retained for each average by dropping the oldest
observation and picking up the newest.
The value of MA is related to the midpoint of the overall
period.
Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting
 The components of a time series
Moving averages
Finding the trend
Finding the seasonal variations
Forecasting

Example (1)
The following table gives the Snowmobile sales for Arthur
Monitor Inc, with both 3-month and 5-month MA.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Moving averages
Finding the trend
Finding the seasonal variations
Forecasting

Example (1)
Month Sales ($100) 3-month MA 5-month MA
Jan 52
Feb 81 60.00
Mar 47 64.33 59.00
Apr 65 54.00 63.20
May 50 62.67 56.00
Jun 73 56.00 58.60
Jul 45 59.33 55.60
Aug 60 51.76 61.40
Sep 50 63.00 55.80
Oct 79 58.00 59.20
Nov 45 62.00
Dec 62
Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting
 The components of a time series
Moving averages
Finding the trend
Finding the seasonal variations
Forecasting

90
80
70
60
50 Sales

40 3-month MA

30 5-month MA

20
10
0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Moving averages
Finding the trend
Finding the seasonal variations
Forecasting

When an odd number of time periods is used in the moving

average, the results are centered at the middle time period.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Moving averages
Finding the trend
Finding the seasonal variations
Forecasting

When an odd number of time periods is used in the moving

average, the results are centered at the middle time period.
However, if there is an even number of time periods in the
MA, an adjustment must be made. As in an even case, there
is no middle observation at which the value can be centered.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Moving averages
Finding the trend
Finding the seasonal variations
Forecasting

Example (2)
Consider the quarterly sales data for Sun Shine Greeting Cards
in the following table. The data run from the first quarter of
2009 (2009-1) to the last quarter of 2011 (2011-4).
A four-period (quarter) MA is calculated. The first entry of
42.5 is actually located between 2009-2 and 2009-3. The
remaining entries are similarly off-centered.
To center the MA, we take the mean of each successive pairs
of moving averages. The average of the first and second
values 44.13 is then centered at 2009-3. The next entry of 45
is centered at 2009-4.
The remaining values are likewise centered at their respective
time periods.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Moving averages
Finding the trend
Finding the seasonal variations
Forecasting

Example (2)
Period Sales ($’000) 4-quarter MA Centered 4-q MA
2009-1 40
2009-2 45
2009-3 38 42.50 43.13
2009-4 47 45.75 45.00
2010-1 53 44.25 45.38
2010-2 39 46.50 44.63
2010-3 47 42.75 42.50
2010-4 32 42.25 43.00
2011-1 51 43.75 42.50
2011-2 45 41.25 44.00
2011-3 37 46.75
2011-4 54
Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting
 The components of a time series
Moving averages
Finding the trend
Finding the seasonal variations
Forecasting

60

50

40

30 Sales
Centered 4-quarter MA
20

10

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Moving averages
Finding the trend
Finding the seasonal variations
Forecasting

Finding the seasonal variations

The additive model of time series analysis is

Y = T + S + R.

If we assume that the random component R is negligible, the

seasonal component can be found as

S =Y −T

called the de-trended series.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Moving averages
Finding the trend
Finding the seasonal variations
Forecasting

Example (3)
Output at a factory appears to vary with the day of the week.
Output (in ’000 units) over the last three weeks has been as
follows
Week 1 Week 2 Week 3
Mon 92 94 96
Tue 115 121 127
Wed 104 107 110
Thu 131 135 140
Fri 74 76 78

Find the seasonal variation for each of the fifteen days, and
the average seasonal variation for each day of the week using
the moving average method.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Moving averages
Finding the trend
Finding the seasonal variations
Forecasting

The multiplicative model of time series analysis can be

expressed as
Y = T × S × R.
In this model, only T is expressed in the original units, S and
R are stated in terms of percentages.
If the random component R is negligible, which means R ≈ 1
in this case, the seasonal percentage is calculated by

S = Y /T .

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Moving averages
Finding the trend
Finding the seasonal variations
Forecasting

Example (4)
From the data in Example 3, using multiplicative model, find
the seasonal variation for each of the fifteen days, and the
average seasonal variation for each day of the week.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Forecasting by extrapolation
Finding the trend
Deseasonalization
Forecasting

Forecasting

Forecasting by extrapolation
Extrapolation is the process of estimating, beyond the original
observation range, the value of a variable on the basis of its
relationship with another variable.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Forecasting by extrapolation
Finding the trend
Deseasonalization
Forecasting

Forecasting

Forecasting by extrapolation
Extrapolation is the process of estimating, beyond the original
observation range, the value of a variable on the basis of its
relationship with another variable.

Forecasts of future values should be made as follows.

1 Calculate a trend line using moving averages or regression.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Forecasting by extrapolation
Finding the trend
Deseasonalization
Forecasting

Forecasting

Forecasting by extrapolation
Extrapolation is the process of estimating, beyond the original
observation range, the value of a variable on the basis of its
relationship with another variable.

Forecasts of future values should be made as follows.

1 Calculate a trend line using moving averages or regression.
2 Use the trend line to forecast future trend line values.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Forecasting by extrapolation
Finding the trend
Deseasonalization
Forecasting

Forecasting

Forecasting by extrapolation
Extrapolation is the process of estimating, beyond the original
observation range, the value of a variable on the basis of its
relationship with another variable.

Forecasts of future values should be made as follows.

1 Calculate a trend line using moving averages or regression.
2 Use the trend line to forecast future trend line values.
3 Adjust these values by the average seasonal variation.
With the additive model, add the variation. With the
multiplicative model, multiply the trend value by the
variation proportion.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Forecasting by extrapolation
Finding the trend
Deseasonalization
Forecasting

Example (5)
For the past seven years, business conditions for Rainbow
Enterprises have been rather bleak. The CEO has collected
quarterly totals of the number of employees who have been
laid off over the past four years in the following table.
The CEO would like to forecast the number of layoffs for each
quarter of 2015, using both additive and multiplicative models
of time series analysis.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Forecasting by extrapolation
Finding the trend
Deseasonalization
Forecasting

Example (5)
Time Layoffs Time Layoffs
2011-1 25 2013-1 35
2011-2 27 2013-2 37
2011-3 32 2013-3 37
2011-4 29 2013-4 39
2012-1 28 2014-1 38
2012-2 32 2014-2 42
2012-3 34 2014-3 44
2012-4 38 2014-4 45

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Forecasting by extrapolation
Finding the trend
Deseasonalization
Forecasting

The multiplicative model is better than the additive model for

forecasting when the trend is increasing or decreasing over
time. The additive model suffers from the somewhat
unrealistic assumption that the components are independent
of each other. In most instances, movements in one
component will have an impact on other components.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Forecasting by extrapolation
Finding the trend
Deseasonalization
Forecasting

The multiplicative model is better than the additive model for

forecasting when the trend is increasing or decreasing over
time. The additive model suffers from the somewhat
unrealistic assumption that the components are independent
of each other. In most instances, movements in one
component will have an impact on other components.
The multiplicative model is often preferred. It assumes that
the components interact with each other and do not move
independently.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Forecasting by extrapolation
Finding the trend
Deseasonalization
Forecasting

The multiplicative model is better than the additive model for

forecasting when the trend is increasing or decreasing over
time. The additive model suffers from the somewhat
unrealistic assumption that the components are independent
of each other. In most instances, movements in one
component will have an impact on other components.
The multiplicative model is often preferred. It assumes that
the components interact with each other and do not move
independently.
Since extrapolation is based on historical data alone, and does
not include effects of developments, it can be used for
short-term forecasts only for specific areas, where no untypical
developments are expected.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Forecasting by extrapolation
Finding the trend
Deseasonalization
Forecasting

Deseasonalization
Deseasonalization is the process of removing seasonal
variations from the data to leave a figure indicating the trend.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Forecasting by extrapolation
Finding the trend
Deseasonalization
Forecasting

Example (6)
Actual sales figures for four quarters, together with
appropriate seasonal adjustment factors derived from the
previous data, are as follows. In the table, Additive SA means
the additive seasonal adjustments and Multiplicative SA means
multiplicative seasonal adjustments.
Quarter Sales ($’000) Additive SA Multiplicative SA
1 150 +3 1.02
2 160 +4 1.05
3 164 -2 0.98
4 170 -5 0.95

Deseasonalize these data.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Forecasting by extrapolation
Finding the trend
Deseasonalization
Forecasting

Example (7)
The percentage of employees absent from work was recorded
over a four-week period as follows.
Week Mon Tue Wed Thu Fri
1 8.4 5.1 5.7 4.8 6.3
2 8.1 5.5 6.0 4.6 6.5
3 8.4 5.6 6.2 5.0 6.8
4 8.6 5.6 6.3 4.9 6.9

Forecast the percentage of employees absent in each day in

week 5 by both additive and multiplicative models.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting

 The components of a time series
Forecasting by extrapolation
Finding the trend
Deseasonalization
Forecasting

Example 8: Data in the given file provide the monthly retail

sales of motor vehicle and parts dealers, in millions of dollars,
of the United States, in the period of 2011-2014.
1 Calculate the retail sales in each quarter of the years
2011, 2012, 2013 and 2014. Calculate a four-quarter
moving average trend centered on actual quarters and
then find the seasonal variations by both additive and
multiplicative models. Predict the retail sales in the four
quarters of 2015 by both models.
2 Given that the actual retail sales, in millions of dollars, in
quarters 1, 2 and 3 of 2015 were respectively 267529,
275239 and 280567. Compare these numbers with your
prediction and discuss the reliability of the two models.

Dr. Nguyen Ngoc Phan Time Series Analysis and Forecasting