DEMAND FORECASTING

PART-2
Roll no. Name

4007 Chauhan Dhruvi Hemantbhai

4009 Chauhan Kiran Hareshbhai

4011 Darji Kajal Chandubhai

4018 Gajjar Diya Virendrakumar

4046 Panchal Dixa Yogeshbhai

4048 Panchal Riddhi Mayurkumar

4078 Shah Shruti Mayurkumar

4090 Trivedi Shreya Gajendrabhai

 MECHANICAL EXTRAPOLATION
price
100
• Assumption of this method is that future
90 90
events are continuation of past events , so
80 80 historical data can be use to predict future
70

60 60 • This method also called time series

50 50 analysis/trend projection techniques.
40

30
• This method is very popular because it is
20 20
inexpensive, simple and quick.
10

0
y ea r 1 y ea r 2 y ea r 3 y ea r 4 y ea r 5 y ea r 6
 120 250
100 200
100 200
80
80 70 150
150
60 50 100 100
Price

Price
100
40 30 50
70

20 10 50
0 0
year year year year year year year year year year year year
1 2 3 4 5 6 1 2 3 4 5 6
Year Year

• These method yield reasonable accurate result if trend of the data has persistent
tendency to move in the same direction

• If the data show significance turns then the time series analysis may not give
acceptable results
With the help of time series analysis we can get the following kind of information:
• Finding a trend value for specific year.
• Finding seasonal fluctuation in the variable.
• Predicting turning points in the future movement of the variable.

 Reasons for fluctuations:

1. Trend (T)
2. Cyclical changes (C)
3. Seasonal changes (S)
4. Irregular movement (I)
 1. Trend (T):
16 • Trend component of time series is overall direction
14 of the movement of the variable over a long period
12
10 of the time.
8 • E.X. trend in operating revenue of indian railway
6
from 1977 to 2024.
4
2 • Long term trend generally exist because some of
0
year 1 year 2 year 3 year 4 year 5 year 6 year 7
the important underlying factor like population,
national income etc. move steadily and therefore
produce only a gradual change over a time.
2. Cyclical changes (C):

• Cycle component of a time series is a wavelike, repetitive movement fluctuating about the trend of the series.
• For example depression, recovery, prosperity, and recession
• Cyclical changes are occur due to change in stockpiling or shortage of commodities, weather, government
policy etc.
• identification of cycle help in:
• Locating turning point of the movement of the variable
• Making intermediate range forecast
 3. Seasonal movement:
• The seasonal movement of time series are those repetitive, fluctuating that always occur at a particular time of a year.
• E.X. sales of woollen clothes are higher during October to January because of the
winter season.
• These help manager to make short term forecast
4. Irregular movement
• These are those variable that left over after isolating the other component.
• For example suddenly arrive medical emergency in whole country may increase the sale of pharmaceutical company.
• because these variables are random these variable cannot be measure and predict .
Decomposition of time series:
• four component may be related to each other in addictive or multiplicative form:
O= T+S+C+I (Additive model)
O = T*S*C*I (Multiplicative model) where O is original data
• Most of the series relating to business and economics are of multiplicative nature.
• Multiplicative model can be converted into addictive model by taking logarithms as:
log O = log T + log S + log C + log I
METHOD OF FINDING
TREND

Trend
projection
method

Graphic(fitti
ng trend Algebraic
Smoothing ARIMA
line by Statistical (Least
techniques technique
observation squares)
)

Logarithmic Moving
Semi- Straight Exponential
Parabolic or average
averages line smoothing
Exponential method
 1. FITTING TREND LINE BY OBSERVATION
• Easy and quick method.
• It involve merely plotting of annual sales on a graph and then estimating just by observation where the
trend line lies.
• The line can be extended to future period and corresponding year’s sale can be forecast.

Graph paper on which data can

be plotted

Arithmetic scale on both

axes Semi-logarithmic scale
• Here exponential data plotted on both linear and
Use: constant
change in Horizontal axis = arithmetic scale
logarithmic scale
absolute term Vertical axis = logarithmic • On the left (linear scale), curve grows rapidly due to
conversion of number
exponential data, by just observation it mislead the data.
• On the right (logarithmic scale), as we take log on y axis
Use: exponential so that exponential growth appear as a straight line,
data
making it much easier to analyse trend
2. Time Series Analysis employing least squares Method :-

• This technique is most widely used in a practice. It is statistical method.

• With its help a trend line fitted to the data. This line is known as ‘the line of best fit’.
• By extrapolating the trend line for future we get the corresponding figures of forecasted sales.
In this analysis there is only one independent variable time.
• This system of forecasting is considered “naive”, because it does not explain the reason for
the changes.
 Linear trend: -
 The method of least squares may be used either to fit a straight line trend or a
non-linear trend. The straight line trend may be represented by the equation:
Sales = a + b (year number)
OR, S = a + b.t
Where a and b are the constant representing the intercept and slope
respectively of the estimated straight line. In order to determine the value of
a and b, the following two normal equations need to be solved:
∑S = Na + b∑t
∑ST = a∑T +b∑t2
Where N represent number of years (month, etc.) for which data is available,
Let us see with the help of illustration as to how demand forecasting is done
with the help of least squares method.
 Illustration 1: -
The following data refers to sales, in thousands of rupees (X), of a certain product during five
years. year 1993 1994 1995 1996 1997
Sales(S) 605 715 830 790 835

Assuming the present trend continues, in which year will you expect 1994 sales to be doubled?
SOLUTION: -
Assuming that trend is linear, we may compute the trend equation as
follows:
year Fitting simple S linear trend byt least squares method
t2 St
1993 605 1 1 605
1994 715 2 4 1430
1995 830 3 9 830
1996 790 4 16 790
1997 835 5 25 835
N=5 ∑S = 3775 ∑t = 15 ∑t2= 55 ∑St=
11860
• The set of normal equations are : or,
∑S =Na +b∑t substituting the value of b in eqn. (1),
we get
∑St = a∑t +b∑t2
5a =3775 – 802.5
• Substituting the respective values in the normal a =594.5
equations, we get so,
3775 =5a +15b ……….(1) The trend equation is: S =594.5 +53.5t
Given trend value of S for 1994 (t =2) is
11860 = 15a +55b ………..(2)
715, whose double would be 1430.
• Multiplying eqn. (1) by 3 and subtracting from 1430 =594.5 +53.5t
eqn. (2), we get t = 15617
535 = 10b i.e., 16th year from 1993, viz, the year
2009.
b = 53.5
 Non-linear Trends :-
 The non-linear trends can be parabolic, exponential, geometric, etc. some of the non-
linear trend equations suitable for business forecasting are following:
i. Second and higher degree polynomial trends:

Second degree or quadratic trend S =a – bT +cT2

Third degree or cubic trend S = a – bT +cT2 –dT3
 Second degree polynomials gives us a quadratic trend, while with the help of third
degree polynomial we get cubic trend.
i. Exponential and semi log trend :
S = aebt
Log S = log a + bt
 This trend equation assumes a constant growth rate in variable S over time.
i. Double-log trend :
S = aTb
Log S = log a +b log T
 A double log function assumes that the value of elasticity of S remains constant over
time.
 However, selection of the best trend line out of the various linear and non-linear trend
equations , depends upon the theoretical considerations and empirical suitability.
 The empirical appropriateness of linear and non-linear trend can be assessed with the help
of the method of first and second
First differences, as explained in table below:
and second differences
years Sales First differences Second difference
1989-90 10 - -
1990-91 15 15-10 = 5 -
1991-92 22 22-15 = 7 7-5 = 2
1992-93 31 31-22 = 9 9-7 = 2
1993-94 43 43-31 = 12 12-9 = 3
1994-95 57 57-43 = 14 14-12 = 2
1995-96 74 74-57 =17 17-14 = 3
 In case the first differences of successive observations are nearly constant then the
straight line trend curve is most appropriate. But it is not common to find constancy
in first differences (mainly in a long run cyclical effects occur). In that case we
should not fit a linear trend to the data.
 Alternatively, a second degree of quadratic trend curve may be considered. The
appropriateness of fitting a quadratic trend curve to the given data can be tested by
the method of second differences. If second difference are almost constant, a second
degree or quadratic trend curve is most appropriate.
 when a given time series is increasing or decreasing at a constant rate it could be
more appropriate to fit exponential trend curve.
 Illustration 2 :-
Fit a trend curve having the property of parabola of second degree to the data given below
and forecast the sales for 2000:
year 1993 1994 1995 1996 199
Sales (Rs. 25 30 38 35 20
Crore)

SOLUTION :-
The trend equation of parabola of second degree will be of the form:
S = a + b T + cTt2
Where S = sales and T = time
Its normal equation would be :
∑S = na +b∑T +c∑T2
∑ST = a∑T + b∑T2 +c∑T3
∑ST2 = a∑T2 +b∑T3 +c∑T4
 Computation of second degree parabolic
trend
Year(t) T= Sales(S) ST T2 ST2 T3 T4
(t-1995) (RS.
Crore)

1993 -2 25 -50 4 100 -8 16

1994 -1 30 -30 1 30 -1 1
1995 o 38 0 0 0 0 0
1996 1 35 35 1 35 1 1
1997 2 20 40 4 80 8 16
∑T = 0 ∑S = ∑ST = 5 ∑T2 =10 ∑ST2 = ∑T3 = 0 ∑T4 =34
148 245
Substituting the above value in the normal equation, we get
148 = 5a +10c …………..(1)
-5 = 10b …………..(2)
245 =10a + 34c ……………(3)
From equation (2), we get b = -0.5
Multiplying equation (1) by 2 and subtracting from equation (3), we get
51 = 14c
c = -3.64
substituting the value of c in equation (3), we get
10a = 245 – 34(-3.64)
10a = 245 +123.76
a = 36.88
so the trend equation in second degree parabola is
S = 36.88 – 0.5 T – 3.64T2
The estimated trend values of sales (Se) listed in col. Of the table is computed as
follows:
1993 = T = -2, so Se = 36.88 – 0.5 (-2) - 3.64(-2)2 = RS. 23.32 CR.
1994 = T = -1, so Se = 36.88 – 0.5 (-1) - 3.64(-1)2 = RS. 34.34 CR.
1995 = T = 0, so Se = 36.88 – 0.5 (0) - 3.64(0)2 = RS. 36.88 CR.
1996 = T = 1, so Se = 36.88 – 0.5 (1) – 3.64 (1)2 = RS. 32.24 CR.
1997 = T =2, so Se = 36.88 – 0.5 (2) – 3.64 (2)2 =RS. 21.32 CR.
Estimated trend value for 2000 (where T = 5) would be
Se = 36.88 – 0.5 (5 ) – 3.64 (5)2 = RS. – 56.62 CR.
Since, it gives a negative trend value, the trend equation under consideration is not
suitable.
3. Forecasting Through Decomposing a Time Series:
We know that a time series is composed of trend, seasonal fluctuation, cyclical movements
and irregular variations. If the available data is quarter wise, month wise, it is possible to
identify the seasonal effect. And, if this data is available for a sufficiently long period of time,
the trend and cyclical effect can also be found out.

4. Smoothing Methods:
Smoothing method attempt to cancel out the effect of random Variations on the values of the
time series. Once this effect is removed, a clearer indication of the underlying movement in
the series is a revealed. The two main smoothing methods are:
1) Moving average
2) Exponential smoothing
Exponential smoothing
Exponential smoothing is a very popular approach fir short term
forecasting. The weight assigned to each value reflect their degree of
importance of that value. The weights so assigned that w lies between
zero and unity (0 ≤ w ≤1).
The Smoothened value Ft is found with help of the equation:
Ft = wAt + (1 - w) Ft-1

(Smoothened Value) = w. (Current observed value) + (1 – w) .

(Previous smoothened value)
The weight ‘w' is also called smoothing constant. The contribution of
remote values to St becomes less at each successive time period.
While Selecting the best value of ‘w' rule of the thumb may be prescribed.
1. When the magnitude of the random variations is large, give lower value to w.
2. When the magnitude of the random variation is moderate, a large value be
assigned.
The following example illustrates three different series of smoothened
values: simple moving averages and exponential smoothing with w = 0.1 and
with w = 0.4 .
ARIMA Method (Box-Jenkin technique) :
Box and Jenkin developed a method of forecasting using auto regressive integrated moving
average (ARIMA). This This method is therefore known as ARIMA Method. It can be used
primarily for short term forecasting.
There are five stages of analysis in the method:
1) Removal of the trend:
This method is useful, in case of stationary time series. Those time series that don’t have a
long-term trend component. To remove underlying trend if any in the time series, method of
differencing is applied.
2) Model identification:
Box-Jenkin model involves the following combination:
The order of involvement of auto regressive terms
Number of differences of the original series necessary to remove any inherent trend
The order of moving average terms.
Out of the various possible combination, those combination are selected that provide
an adequate fit to the underlying time series. This is done by matching computed auto
correlations with the auto-correlation function given by the combination.
3) Parameter estimation:
Once a particular combination of the three elements is identified, the method of least square
is used to fit this model to the time series.
4) Verification:
The “goodness” of the estimated model is checked by analysing the residuals it generates. If
these residuals do not show any specific pattern, it is a “good” fit. If it is not “good” fit, we
need to repeat the process by starting from stage 2 and try to develop new combination.
5) Forecasting:
Once we arrive at a model which is good fit, we can use its coefficient to generate forecasts of
future values.
Since box-Jenkin model is complex and is based on search method, it is possible to handle
forecasting with the help of computers. This creates three kind of problems:
a) Computer programmes designed to handle this model are very expensive. Cost of
computational become higher because through this method we can only get short
term forecasts.

b) The difficulties in the operation of this model reduces it’s popularity with the
prepares and users of the forecast.

c) Another difficulty lies with the nature of the model. Since it is primarily a search
model, it requires a highly qualified and experienced forecaster.
BAROMETRIC TECHNIQUE (LEADING INDICATOR TECHNIQUE) :
Barometric technique is based on the presumption that relationship can exist among various
economic time series. There are three kinds of relationships among economic time series:
a) Leading series:
It consists of the data that move ahead of the series being compared. Ex. Birth rate of the
children is the leading series for demand of seats in school
b) Coincident series:
When data in series move up and down along with some other series, it is known as
coincident series.
Ex. A series of data on national income is often coincident with the series of
employment in economy over a short period.
c) Lagging series:
Where data moves up and down behind the series being compared.
Ex. Data on industrial wages over time is a lagging series when compared with series of
price index for industrial workers.
The leading, coincident and lagging series can be used to forecast change in an economic
variable.
 COINCIDENT INDICATORS AND LAGGING
INDICATORS

• These are those variables whose movement coincides with or falls behind
general economic activity or market trend.
• Common examples of coincident indicators are gross national product,
index of industrial production, retail sales, no. of industrial employees, etc.
• Common lagging indicators are manufacturer's stock level and consumer
credit out standing.
 Leading Indicators
Those variables whose movement precedes the movement of some other
related variable are known as leading indicators.

Ex:-
• Loan application with financial institution
• Birth rate
• Enrolment in school
• Price of food grains
STAGES:

1. Locate the leading indicator

2. Estimate a mathematical/statistical relationship

3. Find out Forecasted values of the variable

4. If possible, verify the validity of the forecast

Limitations:

 Difficult to Identify
 Proper Indicators may not exist
 Changes in fashion and tastes
 Lead time issues
 Mainly useful for deriving information regarding direction of change, but not of
much use of to find the magnitude change in variable.
 INDEX NUMBER

 To overcome some of the problems of leading indicators method, the

following methods were developed:

(i)Diffusion indices
(ii)Composite indicators
Diffusion indices:

• In this method a group of indicators is initially chosen.

• Then the percentage of the group of chosen indicators which have fallen (or,
risen) over the last period is plotted against time to get the diffusion index.
• The diffusion index is typically calculated as the proportion of indicators that
are increasing, usually expressed as a percentage.
For example:- if there are say, 9 proper leading indicators for forecasting the
construction activity of dwelling units and if by plotting we find that say, 6
indices show a rise. Calculate diffusion index.
Ans: 66.7%
Limitations:

 When used to forecast business cycles, the lead time between change in
chosen indicators and the turning points in business cycle is very small.
(lead to difficulties in interpretation)

 It is not possible to predict the magnitude of change.

COMPOSITE INDICATORS:

 Composite indicators or indices consist of a weighted average of the chosen

leading indicators.

 By combining several individual series this method reduces the possibility

of existence of random fluctuations, thereby saving from the possibility of
false prediction of downturn or upturn in the variable under forecast.
The steps to get composite indices are the following:

1. Select a set of proper leading indicators for forecasting.

2. Remove trends from the series and smooth the data using the moving average method.
3. Assign weights to indicators, based on following characteristics:
 Economic significance of the variable
 Extent to which the movement in the variable conform to given cyclical pattern.
 Stability in the timing of movement of the indicator
 Smoothness of the movement of the chosen indicator
 Currency of indicator
4.Calculate a composite index by multiplying the values of the indicators by their
respective weights and summing them up.
STATISTICAL METHODS:
We take time as explanatory variable in forecasting exercise, we may assume that factors
influencing sales in the past will behave exactly in same manner in future too.
It not considers factors like competitor’s advertising, competitor’s price, own price and
advertisement. These factors can change, therefore time will not able to explain the movement
of sale
So, in such a case we use statistical methods. These methods are also called economic
methods.
1. Naive models:
Naïve models are often effective as more sophisticated models, cheaper and easier to use.
It is generally useful when situation are stable or experiencing only gradual changes.
In this model the current value is used as a forecast for the next period.
2. Correlation and regression method:
• It is most popular method of forecasting.
• Unlike time series the correlation and regression analysis does not limit itself to” time” as
the independent variable.
• It considers other factors that can be affected to sales.
Correlation Analysis:
Correlation in fact means covariance.
Positive correlation: when two series vary in the same direction . (r>0)
Ex: The relationship between family ‘s income and expenditure
Negative correlation: when two series vary in the opposite direction . ( r<0)
Ex: fuel prices and vehicle sales
It may be noted that two sets of variables may be related at same date or with a time lag. Ex acreage and
output of various crops depend upon the prices of the crops at the time of sowing period.
i.e. the price of wheat in period t determines acreage shown under wheat in next period(t+1) and
therefore output of wheat in period ( t+1)

 correlation analysis can be of two type:

1. Simple or partial correlation: When only one independent variable is taken to explain variations in
dependent variable
2. Multiple correlation: when there are more than one independent variable taken to explain variations
in dependent variable .

The closeness of independent and dependent variable can be found with the help of coefficient of
correlation.
The value of r is greater than 0.5 it implies a strong tendency for variation in production
and in fuel consumption. But since r is negative it means when production increases fuel
consumption decrease, and vice versa.
Regression equation method:
Export of good: X=a (national income) + b( domestic price of x)+ c( international price of
x) + d( weather)
If trend of the dependent variable is other than linear we fit a non – linear regression
equation for forecasting.
It can be any forms parabolic, logarithmic, exponential etc. depending on the way the trend
of the dependent variable behaves.
Fitting simple linear regression: In this case a straight line is fitted to the data containing
one dependent variable and only one independent variable
Ex sales = a+b(price)
Fitting of straight line regression equation can be done either Graphical method Or by least
squares method.
1) Graphical method:
Least Square method:
- It is a method to predict the behavior of dependent variable due to change in independent variable. In least square
method of estimating regression line S = a + bp, we need to find the values of the constants a and b with the help of
the normal equations:

………..
 By simplifying (1) and (2) we get the regression coefficient of S on P and the intercept of the regression
equation (a) :

 By substituting the calculated values of constants a and b in the regression equation S = a + bp, we get the
required regression equation to forecast S.
 We can also find regression equation by either of the following two formulas (which can be used according to
the type of data available):
- Note:
Fitting Non-linear Regression equation:
1. Logarithmic Model:
• Let us plot the data given in illustration 6 into (a) the usual arithmetic scale and (b) the
semi-logarithmic scale graph papers. if we find that the first graph shows a curved
freehand trend line, while the second graph shows a linear trend in the price series over
time, then a log-linear model would be more appropriate to fit. The log-linear model
would be of the form:
log S = a+b*P
where, log S = logarithm of sales, and
P = Price
Functions of the form S = aPb and b=Price elasticity of P
2. Parabolic Regression Model:
• Sometimes we need to fit a curved trend line which, by a change in variable ,could not
be reduced to a linear form.
Let us assume that it is a second-degree polynomial given by the equation:
S= a+bP+cP²
The statistics a, b, c, etc. can be calculated from the set of normal equations
∑S=Na+b∑P+c∑P²
∑SP=a∑P+b∑P²+c∑P³
∑SP²=a∑P²+b∑P³+c∑P⁴
3. Multiple Regression Analysis:
- When more than one independent variable is taken in the regression model, we use multiple
regression Co-efficient and equation.
A multiple regression model, for sale;

[Where, a, b, c, d and e are the partial regression Co-efficient which show the effect of cor-responding
variables on sales]
Here, 'u' represents the effect of all the variables which have been left out in the equation but have an
effect on sales.
How sale is forecasted?
- If the expected values of the independent variable are substituted in equation, the sales will be
forecasted.
Advantage:
- The effect of a large no. Of variables can be taken into account.
- Enables the business to experiment with what might happen under extreme to unlikely
conditions.

Disadvantaged:
- forecasts based on the past data.
- The accuracy of measurement of independent variables determines the degree to which
the confidence can be placed in the forecasted values of the dependent variables.
 ECONOMETRIC MODELS
 The main feature that distinguishes econometric models from time series models is their
perception of what influences the future value of the variable for forecast.

Time series analysis rely only on time as a Econometric models try to identify all
causing change in the variable to be those economic and demographic variable
predicted. under forecast and build up a cause-effect
relationship.

FOR EXAMPLE:

To predict the future demand of a particular brand of air-conditioner, an econometric

model would need series of data on certain independent variables like price of that brand
and its substitutes, advertisement expenditure by the brand and its rivals, weather
conditions, aggregate personal disposable income and other variables influencing demand.
A linear model relating these independent variables to the dependent variables , viz., demand of
air-conditioners, may be stated as:
 Forecasting through econometric models involve THREE STAGES:
1) Identification of Variables and the Functional Form:
a) Which variables influence the variable to the forecasted.
 You first figure out which factors (variables) affect the demand for air-conditioners.

b) Which form of mathematical relationship between the independent variable and the
forecast variables is most appropriate.
 Next, decide what kind of mathematical relationship connects the variables and the
demand. For example, should the relationship be linear (like a straight line) or more
complex?

 The choice of these variables and mathematical form may be either deducted from
economic analysis or be based on past empirical evidence, or both.
2) Estimation of Parameters:
 Using the series of data of the independent variables and the forecast variable, the
values of α's and β's are estimated from equation which “best” represents the behaviour
in the past (known as “best fit”).
 In the estimation of parameters, we aim to find the coefficients (the α's and β's) that tell
us how strongly each independent variable (like price, advertising, weather, etc.) affects
the dependent variable (like demand).

3) Finding the Forecast Values:

 It is assumed that the best fit equation would represent the future trend of the forecast
variable. The future values of independent variables are, therefore, predicted and

substituted in the ‘best fit’ equation to get the corresponding values of the forecast
variable.
After estimating the model, you can use it to predict future demand by plugging in future
values of the variables (like future prices or income levels). This is how you forecast
demand based on your model.
 Econometric models can be broadly classified into TWO categories:
a) Single equation models
b) Simultaneous models
a) Single equation models: The casual relationship is expressed in terms of one equation
only( the demand equation of air-conditioners).
b) Simultaneous models: Consist of a set of equations where a variable affects another variable
and in turn also affected by the other variable. For example, in a business, sales, prices,
income, and costs are all related. If you change one, it impacts the others. The simultaneous
equation model captures these interrelationships. When variables are both independent and
dependent at the same time we need to build a complete system of simultaneous equations,
known as simultaneous model.
 Those variable whose values are determined by the model are known as endogenous
variables. They are influenced by other variables in the system. For example, sales might
be an endogenous variable because it depends on things like price or income.
 On the other hand, exogenous variables are those whose values originate from outside the
system. and are not influenced by other variables in the system. For example, government
policy or external economic conditions could be exogenous.

 Simultaneous models consist of TWO kinds of equations.

i. Identities
ii. Behavioural equations
1. identities represent relationship that are necessarily true .
For example , if we write an equation starting that the sum of final
goods and services in economy during year (Yt) equals the sum of
expenditure on consumption ( Ct) Government expenditure (Gt)
and investment(It ), it would be an identity , because it is true by
definition .
Yt = Ct +Gt It…….(1)
• This equation represents the total output or income (Yt) in an
economy. So, Yt is the sum of household consumption and a
combined factor of government spending and investment.
2. Behavioural equations, On the other hand, represent the
expected behavior of physical processes or individuals. For
example , if we state that consumption (Ct) depends upon the
level of income during t(yt) :
Ct = b0 + b1 Yt +et……(2)
• b0 is the intercept, representing the base level of
consumption when income is zero.
• b1 is the marginal propensity to consume (MPC),
showing how much consumption will increase if income
increases by one unit.
• Yt represents income during the period.
• et is the error term, accounting for random factors
affecting consumption that are not captured by the
model.
Equation (3): Behavioral Equation for Investment
It = a0−a1 it + a2 P t −1 + et…….(3)
This is a behavioral equation for investment It
It states that investment depends on:
a0​= A constant value representing basic investment.
−a1 ​it​ = This indicates that investment is negatively related to the interest rate [it​]. So, if
interest rates go up, investment tends to go down.
a2 Pt−1 ​= This shows that investment is positively related to past profits Pt−1​. If
businesses made more profits in the previous period, they are more likely to invest more.
et​ = Disturbance term, capturing other factors affecting investment that are not included
in the model.
This equation reflects how firms make investment decisions, balancing the cost of borrowing
(interest rates) against expected returns (past profits). Then the two equation 2 and 3 may
be stated as behavior equations.
Advantages of simultaneous equation model :
1) This models are able to identify the reasons for change in the values of variable.
 The model helps you understand why changes happen in the variables (like why
demand for a product increases or decreases). It can trace back to specific causes, such
as price changes, income fluctuations, or changes in other relevant factors.
2) In a simultaneous equation model it is possible to forecast the effect of policy variables by
tracing them as independent variables.
 By analyzing how different variables interact, the model allows you to predict the impact
of policy decisions (for example, changes in taxes or interest rates) on the economy or
business. You can forecast how a change in one variable might affect others over time.
3) As a predicted values are compared with the actual data, it is a possible to improve the
accuracy of prediction of this model.
 The model compares its predicted values (what it forecasts) with actual data from the
real world. By doing this, it can adjust and improve its predictions, making future
forecasts more reliable. This process helps reduce errors in predictions.
 Simultaneous Equation Method
 The Simultaneous Equation Method is also called the complete systems approach. It
looks at all variables together to figure out how they affect each other at the same time. This
method is used when we are trying to understand how different factors influence each other
in an economic or decision-making situation.
 In this method, a set of equations is created where each equation shows how one variable
depends on other variables. These dependent variables are also called endogenous variables
because they are affected by other factors within the system. The number of equations needs
to match the number of unknowns (variables we are trying to find).
 Once the system of equations is written, we solve it using methods like two-stage least
squares. This is a statistical process that helps in estimating the unknowns. In the first stage,
we look at variables that the management cannot control (called exogenous variables), such
as consumer income or lagged variables (like sales from the previous period). These
variables help in predicting the unknowns in the next period.
 The main advantage of this method is that it allows us to predict future outcomes by
considering how all the factors interact with each other at once. This method also handles
both exogenous variables (things outside the system’s control) and endogenous variables
(things within the system's control).

 However, it can be complicated because we have to estimate the future values of several
variables at the same time. Unlike simpler methods like regression, it requires more steps
to solve. Even though it provides a detailed understanding, it is not as popular because of
its complexity.
Demand Forecasting Of New Products
Life cycle segmentation analysis

• Each product has a life cycle : introduction , growth , maturity ,saturation and decline.
• The sales of a new product in any particular market segment tend to follow an S-
shaped curve.
Introduction: Focus on quality; advertising has low impact.
Growth: Early adopters bought; advertising is key to meet demand.
Maturity: Price becomes critical; rivals enter, and advertising
continues.
Saturation: Price is low; product differentiation in packaging and
quality becomes important.
Decline: New uses for the product are sought; advertising and
quality matter, but price has minimal impact.
Timing of each stage affects marketing strategy, and
segmenting the market helps adjust tactics for less-developed
segments.

THANK YOU