TIME SERIES

Time series is used to understand, evaluate, and interpret changes in economic phenomena over
time with the hope of more correctly anticipating the course of future events.

Components of time series

The main components of time series are as follows:

1. Secular trend:

The general tendency of a data to increase or decrease over a long period of time is known as secular
trend or trend. Eg. National income shows upward tendency, birth or death or illiteracy shows downward
tendency.

2. Seasonal variation

Seasonal variations are those variation which occur with some degree of regularity with in a specific
period of one year or shorter. Eg, Climatic conditions, social customers, religious functions etc. The
prices of price will grow up in the sowing season and will come down in the harvest season.

3. Cyclic variation

Cyclic variations are periodic movements which occurs at intervals (or periods) of more than one year.
Eg. Business cycles, trade cycles

4. Irregular fluctuations

Irregular fluctuations are those caused by unusual, unexpected and accidental events such as
earthquake, strike, flood etc.

Additive model
O =T+S+C+I

Multiplicative model
O  T  S C  I

Where O = original data

S= seasonal variation
C= cyclical variation
I =irregular fluctuation
Short term fluctuations = actual values –trend values.

1. Secular trend
The general tendency of a data to increase or decrease over a long period of time is known as
trend or secular trend.
 Methods of measuring trend
a) Free hand curve method.
b) Method of semi averages.
c) Method of moving averages.
d) Method of least squares.

a) SEMI AVERAGE METHOD

According to this method, the original data is divided into two equal parts. When the number of
years is even, there is no problem. But if the number of years is odd, the total number of years is divided
into two groups, omitting the middle year. For example, when the data given is for1998 to 2007 then the
two parts are1998 to 2002 and 2003 to 2007. If the data is for 1998 to 2006 the two parts are 1998 to 2001
and 2003 to 2006 (omitting 2002). The figures from the two parts are averaged. The average of each part
is plotted on the graph against the middle year of each part. A straight line shall then be drawn to pass
through these two points plotted on the graph. The straight line shows the trend. Trend values can be
measured from this line.

Ex1: Apply the method of semi averages for determining trend.

year 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
values 45 58 62 50 70 72 68 70 78 75
Solution:

year values Semi average

1999 45
2000 58
2001 62 (285/5)=57
2002 50
2003 70
2004 72
2005 68
2006 70 (363/5)=73.6
2007 78
2008 75

Ex2.Apply the method of semi averages for determining trend.

year 2000 2001 2002 2003 2004 2005 2006 2007 2008
values 10 12 15 20 18 25 24 28 34
 Solution:

year values Semi average

2000 10
2001 12 57/4=14.25
2002 15
2003 20
2004 18
2005 25
2006 24
2007 28 111/4=27.75
2008 34

b) MOVING AVERAGE METHOD

Moving average of a time series is a new series obtained by finding out successive items chosen
on the basis of periodicity of fluctuations, dropping off one item and adding the next item at each stage.
The moving average may be for three, four, five, six years and so on, according to the size and periodicity
of fluctuations of the data. Suppose moving average is to be calculated for five years, we will take the
average for first five years and will place it against the middle year of the five. Now leaving the first year
and adding the sixth year, take the average of the five years. Place this average against the middle of those
five years. We will go on in this way taking the average after leaving the first year and taking one next
year. For example, if the original values are a, b, c, d, e, f… then three yearly moving averages are,
abc bcd cd e abcd e
, , etc. Five yearly moving averages are ,
3 3 3 3
bcd e f
, etc.
3

If the moving average is to be calculated for even number of years, say, four six, etc., then the
procedure will be different. For four years moving average, we will calculate the average of first four
years and will calculate the average. This average will be placed in between second and third years, i.e.,
in the middle of four years. Leaving the first year, calculate the average of next four years and place it in
the middle of these four years and so on. Then we find the average of the two moving averages already
calculated, taking first and second, second and third etc. This is called moving average centered. We will
place the first average centered against the middle of the two moving averages. This average will be
against 3rd year middle of the two moving averages. This average will be against 3rd year of the original
data. In this way we calculate averages centered for the other years. The centered moving averages will be
the trend values.

Ex.1. Calculate 7 yearly moving average for the following data on number of commercial and industrial
failures in a country during 1992-2007.

Year 1992 93 94 95 96 97 98 99 200 01 02 03 04 05 06 07

No of 23 26 28 32 20 12 12 10 9 13 11 14 12 9 3 1
families
year values 7 yearly moving 7 yearly moving
total average
1992 23 …… …….
1993 26 ……. …….
1994 28 ……. ……..
1995 32 153 21.19
1996 20 140 20.0
1997 12 123 17.6
1998 12 108 15.4
1999 10 87 12.4
2000 9 81 11.6
2001 13 81 11.6
2002 11 78 11.1
2003 14 71 10.1
2004 12 63 9.0
2005 9 ……. …..
2006 3 ……. …..
2007 1 ……. …..

Ex2. Work out the trend values by “centered 4 yearly moving average nmethod” for the following data
and plot the given values and trend values on a graph.

Year 1997 98 99 2000 01 02 03 04 05 06 07 08

Tonnage 1102 1250 1180 1340 1212 1317 1452 1549 1586 1476 1614 1586
of cargo
cleared
Solution:

year values Total of 4 years 4 yearly moving 4 yearly

average moving average
centered
1997 1102 …
1998 1250 …
1999 1180 4872(between 2nd 1218 1231.75
and 3rd line)
2000 1340 4982 1245.5 1253.881296.25
2001 1212 5049 1262.25 1356.37
2002 1317 5321 1330.35 1429.25
2003 1452 5530 1382.5 1495.88
2004 1549 5904 1476 1536
2005 1586 6063 1515.75 1560.88
2006 1476 6225 1556.25 ……
2007 1614 6262 1565.5 ……..
2008 1586 ….
 c) Method of least squares.
The principle of least squares provides us with a mathematical or analytical device to obtain a
mathematical curve which will be fit to the given series. The technique of obtaining this mathematical
curve by the principle of least squares is known as curve fitting.

The principle of least squares states that the sum of the squares of the deviations between the
observed values and trend values is least. The technique can be used to fit linear as well as non-linear
trends.

Trend may be linear or no-linear. Linear trend is one which gives the straight line when plotted on
a graph paper while non-linear trend is one which gives non-linear curves like parabola, exponential curve,
logistic curve etc.

LINEAR TREND (LINE OF BEST FIT)

A straight line trend can be fitted to the data by the method of curve fitting based on the most
popular principle called principle of least squares. Such a straight line is also known as line of best fit. It
gives us a straight line from which the sum of the deviations on either side will be equal to zero. That is
the vertical distances at various points on one side from this line will be equal to the total of vertical
distances at various points on other side of the straight line. When sum of deviations on either side equal
to zero, the sum of the squares of these deviations will be least. Let the line fit be described by an
equation of the type Y = a + b x, where ‘y’ is the value for dependent variable, a and b are two unknown
constants whose values are to be determined.

To find a and b, we apply the method of least squares. Let ‘E’ be the sum of squares of the
deviations of all the original values from their respective values derived from the equations, so that
E  [ y  (a  bx )] 2

E E
By calculus method, for minimum  0 and  0 .Thus, we get the two equations known
a b
as normal equations. They are,

 y  na  b x
 xy  a x  b x 2

Solving these two normal equations, we get a and b. Substituting these values in the equations y = a +bx,
we get the trend equation.
Note: When  x  0 , the normal equation becomes  y  na and  xy  b x 2
so that a 
y
n

and b 
 xy .
x 2

Problem: The following are the annual profits in thousands of Rupees in a certain business.

Year 2002 2003 2004 2005 2006 2007 2008

Profit in 60 72 75 65 80 85 95
(‘000Rs)
Using the method of least squares fit a straight line to the above data. Also make an estimate of profit in
2009. Estimate the trend values for all years. Compute short term fluctuations. Plot the given values and
trend values on a graph.

Solution:

Year Profit in Deviation xy X2

(t) (‘000Rs) from midyear
2005 = t-2005
2002 60 -3 -180 9
2003 72 -2 -144 4
2004 75 -1 -75 1
2005 65 0 0 0
2006 80 1 80 1
2007 85 2 170 4
2008 95 3 285 9
Total 532 0 136 28

The equation of straight line trend is y = a+bx where a 

y and b 
 xy .
n x 2

532 136
i.e., a   76 and b   4.86
7 28

 The equation is y = 76 + 4.86x

For the year 2009, x=4 then y = 76 + (4.86  4) = 95.44
 Profit in 1998= 95.44 thousand rupees.
Trend values
When x = -3, y = 76 + (4.86  -3) = 61.42
When x = -2, y = 76 + (4.86  -2) = 66.28
When x = -1, y = 76 + (4.86  -1) = 71.14
When x = 0, y = 76 + (4.86  0) = 76
When x = 1, y = 76 + (4.86  1) =80.86
When x = 2, y = 76 + (4.86  2) = 85.72
When x = 3, y = 76 + (4.86  3) =90.58
We can also plot the actual values and trend values and draw graph for them.
Ex.2 using 2003 as origin obtain a straight line equation by the method of least squares

year 1999 2001 2002 2003 2004 2005 2008

values 140 144 160 152 168 176 180
Find the trend values of the missing years 2000,2006, and 2007.

Year Value X=t-2003 xy X2

t (y)
1999 140 -4 -560 16
2001 144 -2 -288 4
2002 160 -1 -160 1
2003 152 0 0 0
2004 168 1 168 1
2005 176 2 352 4
2008 180 5 900 25
1120 1 412 51
Here

 y  na  b x
 xy  a x  b x 2