As per the SEP Syllabus for 2nd Semester B.

Com
Bangalore University

BUSINESS DATA ANALYSIS

MODULE 5: TIME SERIES

The analysis of the statistical data becomes necessary for

a businessman or an economist in order to predict or to
estimate for future demand of a product, or the future
economic movements, by studying the past behavior of
the data. Hence the ‘analysis of the time series is the study
of the past with the object of prediction for future’.

To estimate for the future the first step is to collect

information from the past. So, the data is observed,
collected and recorded at successive intervals of time.
Such data is called time series data.

Variables change with the time. For example, figures of

population, agriculture production, sales, exports. imports,
employment and electrical consumption etc, are changing
with the time. These changes may be either a year, a
month, a week etc.

Hence “A set of figures relating to a variable according to a

time is called Time Series”.​
Also, “A chronological arrangement of statistical data is
called a Time Series”.

Consider the following example of figures of production of

a firm.​

If we observe the above table, the production is increasing,

although for some years, it has decreased. The rise or fall
of production may be due to some causes or influences or
reasons, and are all called ‘factors’.
PURPOSES/USES OF TIME SERIES ANALYSIS

The analytical study of these time series will help to:

●​ Understand the past, present & future behavior of the

data
●​ Predict / plan for future
●​ Control present performance and
●​ Facilitates comparison.

COMPONENTS OF TIME SERIES

“The factors which are responsible for the fluctuations

occurring in the time series are called Components of time
series.”

The components/variations of time series are classified

into four main classes. They are-

●​ Secular Trend
●​ Seasonal Trend
●​ Cyclical Variations
●​ Irregular/ Random Variations.

1. Secular Trend: The term secular trend/Trend refers to the

tendency of the variable to increase or decrease or to
remain steady over a period of time.
For Example, The values of population, prices, sales,
literacy etc. are all increasing. As against the death rate,
illiteracy, travel by bullock carts is decreasing. The rise or
fall may be steep or gradual. But they show an increasing
trend, if we observe over a sufficiently long period of time.​
(Here the time ‘t’ > 1 year)

2. Seasonal Trend: The term seasonal variations basically

refer to the variations caused annually by the seasons of
the year. But also includes the variations of any kind which
are periodic in nature and whose period is shorter than
one year (Here the time ‘t'< 1 year). There are two
important factors which are responsible for seasonal
variations namely​
(a) climatic and weather conditions​
(b) Customs, habits and traditions of the people.​
For Example:

●​ Sales of cool drinks, ice creams are more in the

summer season
●​ Sale of stationery will be more in June, July.
●​ The commercial bank business may reach a peak
around the first week of every month.
Hence, a clever businessman will arrange for the
production or to maintain stock according to the needs of
the season and will earn maximum profit.

3. Cyclical Variations: Cyclical variations are the

fluctuations spread over a period of more than one year.
(Here time ‘t'>1 year). Most of the time series relating to
economics and business show some kind of cyclical
variation. Hence ‘cyclical variation is an oscillatory
variation which occurs in four stages, such as:

●​ Prosperity
●​ Decline/Recession
●​ Depression and
●​ Improvement/Recovery

Business cycle is:​

​
During peak/prosperity, the curve increases, the business
is in boom, the transactions are much more than expected.
After reaching the peak of activities the business declines,
the curve slopes down. This is called the period of
decline/recession slowly to depression - the business
activities are at the lowest. Then follows a period of
improvement in activities and the curve again starts to
rise. In this manner the cycle repeats. The interval of time
from one prosperity to the other is called the period of
cycle. (Here the time ‘t ‘ > OR < 1 year)

4. Irregular or Random variations: Irregular variations are

those changes of time series which are irregular in nature
and do not show any pattern. The causes of irregular
variations are due to accidental happenings such as wars,
earthquakes, floods, famine, fire, strikes, etc. These factors
are unpredictable. Generally such variations last for a
short period.

METHODS OF MEASURING TREND

Following methods are used for the measurement of trend:

●​ Graphical Method
●​ Semi-Average Method
●​ Moving Average Method
●​ Least Squares Method
Method of Moving Averages

In this method simple arithmetic means are calculated

successively by taking some specified number of values at
a time ( period of moving average ‘m’), say 3, 4, 5 years.
The aim of averaging is to remove the short term variations
that are present in the time series of a periodic type. Here
the period of moving average (m) is the period covering
the number of consecutive values taken at a time.

In 3 yearly moving averages, the first moving average is

the average of 1st, 2nd and 3 rd observations and is
written against the middle i.e. the 2nd time point. Then
dropping the 1st value and adding the next three values
and continuing this process, until all values are included.

If m is even number ie.(m = 2 or 4), first moving averages

with period ‘m’ are found, these do not belong to any of the
time points and so again ‘moving averages with period 2 of
the moving averages are calculated, and are called
centered moving averages’. Such calculated values are
called trend values Yc​
The graph of a time series is called a “Historigram” and
that of the trend values is called a “Trend Line”.

Merits:
 ●​ This is a simple mathematical method
●​ If few observations are added to the data the trend
values are not affected.
●​ This method is suitable if the time series shows an
irregular trend

Demerits:

●​ Trend values cannot be computed for all years.

●​ Forecasting values for further years is not possible.

Method of Least Squares

In this method a mathematical relation is developed

between the time (x) and the values (y). The relation may
be used to fit

●​ Linear trend
●​ Quadratic trend
●​ Exponential

“Method of least square is a method of fitting a

mathematical relation to the time series such that the sum
of squared deviations of the observed and trend values is
the least”.
Here a relation is derived such that Σ(Y – Yc)2 is least . This
process results in normal equations, {i.e. in reducing the
errors between actual and trend values} ‘The process of
minimisation of sum of squared errors results in some
equations which are called normal equations.

Normal equations are the equations which are used for

finding the coefficients (constants say a, b) of the relation
which is fitted by the method of least squares.

Merits:

●​ This method is mathematically easy, gives accurate

trend values
●​ Trend values can be computed for all time points.
●​ Future values can be predicted for the given time
points.

Demerits:

●​ Difficult to apply to an algebraic equation.

●​ Entire calculations have to be done all over again
whenever any observations are added to the time
series.

The Straight line/Linear trend equation is: y = a + bx​

Normal Equations are​
na + bΣx = Σy​
aΣx + bΣx2 = Σxy​
Where, y – actual value; n-number of years; x – deviations
for time; a and b are constants determined by solving the
normal equations. After getting the values of a and b the
equation is fitted - called the line of best fit as Yc - the
trend line.

NOTE: Using deviation method always we can get Σx = 0

and so

a = Σy/n

and

b = Σxy/Σx2

Example:1

For the following time series calculate the trend by the

method of 3 yearly moving averages.

Answer:
In the 3 yearly moving average there are no trend values in
1st and last time points.

Here Yc – trend values.

Example 2.​
For the following time series, calculate the trend by the
method of 3 yearly moving averages. Show a trend on a
graph.​

Answer:​
Computation of 3 yearly moving averages.​
Let x and y be the year and values of the time series.​
From the graph the time series shows a decreasing
tendency.

Example 3.​
For the following time series, calculate the trend by the
method of 3 yearly moving averages. Plot the trend values
on a graph.​
Answer:​
Calculation of trend values by 3 yearly moving averages.
(Let x and y be the year and values.)​
Example 4.​
Calculate the trend values by finding three yearly moving
averages.​

Answer:​
Let X and Y be the year and sales. Here the period of
moving average ‘m’ = 3​

Example 5.​
For the following time series, compute the trend by five
yearly moving averages.​

Answer:​
Computation of 5 yearly moving averages. Let x and y be
the year and values.​
There are no trend values for the first 2 times points and
last 2 times points.​
Yc {5 yearly moving average} trend values.

Example 6.​
For the following time series, compute the trend values by
the method of five yearly moving averages.​

Answer:
Let x and y be the year and Index of the time series.​

Example 7.​
For the following time series, compute the trend by 5
yearly moving averages.​

Answer:​
Computation of trend values by the method of 5 yearly
moving averages.​
Let x and y be the year and values.​

Example 8.​
Obtain the trend values for the following time series by the
method of four yearly centered moving averages.

Computation of 4 yearly centered moving averages. Let x

and y be the year and values.​
Here there are no trend values for the first 2 time points
and last 2 time points.​

Example 9.​
Obtain the trend values for the four centered averages.​

Answer:​
Calculation of trend by 4 yearly moving averages. Let x
and y be the year and values.​

From above, the time series shows a decreasing tendency.

Example 10.​
For the given time series, fit a straight line trend of the type
y = a + bx by the method of least squares.​

Answer:​
Let x and y be the year and production.​
The straight line equations y = a+bx​
Normal equations are.​
na + b Σx = Σ y ……..(1)​
a Σx + b Σx2 = Σxy……..(2)​
Here x = X – x̄( odd time points)/ deviations​
x̄ = 1996​

From (1) 5a + b. 0 = 289​

∴ a = \frac{289}{5} = 57.8​
From (2) a.0 + b.10 = 35​
∴b = \frac{35}{10} = 3.5​
Hence the fitted straight line trend equation is ŷ = 57.8 +
3.5x​
Note: The trend values for the given years can be obtained
by substituting the values of x as -2, -1, 0, 1, 2 if n is odd
and -5, -3, -1, 1, 3, 5 if n is even.

Example 11.​
Fit an equation of the type y = a + bx to the following
Time-series data.​

Answer:​
The straight line equation is y = a + bx​
The normal equations are na + b Σx = Σy​
a Σx+b Σx2= Σ xy​
Here n = 8, even,​
take x=2 (x- x̄); x̄ = 2001.5​
Let X be the year and Y be the imports.​

The fitted straight line equation is : Yc = 30.875 -0.375x

Example 12.​
Compute 4 yearly moving averages for the following data :​

Answer:​
Let X and Y be the years and values​

Example 13.​
Fit a straight line to the following time series, show the
trend line on a graph​

Answer:​
Let x and y be the year and values The Straight line
equation is​
y = a + bx​
Normal equations are –​
na + bΣx = Σy ………… (1)​
aΣx + bΣx2 = Σxy ………….(2)​
Here x̄ = 1992​

From (1) 7a + b.O = 3106​

∴ a = 443.71​
From (2) a.O + b.28 = 2045​
∴ b= 73.04​
∴ The fitted straight line trend equation is

Yc = 443.71 + 73.04x​
Trend value Yc for 1989; Here x = -3​
∴ Y1989 = 443.71 + 73.04 (-3) = 443.71 – 219.12

Y1989= 224.50​
Trend value for 1990 ; Here x = -2​
Y1990 = 443.71 + 73.04(-2) = 297.63​
Similarly all remaining trend values can be obtained by
substituting x= -1,0,1,2 and 3 Straight line trend:​

Example 14.​
For the given time series, fit a straight line trend of the type
y = a + bx by the method of least squares and find the
trend value for the year 1995.​

Answer:​
X Here X=2(x-x) \n = 6 X = 1990.5​
From (1) 6a + b.0 = 620 ∴ a = 103.33​
From (2) a.O + b.70 = 302 ∴ b = 4.31​
The fitted straight line trend equation is.​
Yc = 103.33 + 0.057x​
Estimation of production for the year 1995​
Here x = 9​
Y1995 = 103.33 + 4.31(9)= 142.09.

**********************************************************************