You are on page 1of 24

TITLE

PREDICTION OF DEMAND FROM NEED, DESIRE


AND PURCHASING POWER FOR FMCG
PRODUCTS-A NEW ALTERNATIVE.

BY

Prof. PRABAL DAS


INTERNATIONAL INSTITUTE OF MANAGEMENT SCIENCES.
AE-720, SECTOR-1, SALT LAKE, KOLKATA: 700064
EMAIL: prabaldas2010@gmail.com
MOBILE: +919836276024
ABSTRACT

In marketing we understand that need coupled with desire and purchasing power creates

demand for a product. This is a marketing myth. The problem is establishing a

mathematical relationship between them to predict demand. This is because measuring

desire or need is a tough job. There are no standard scales to measure need or desire to

perfection. Need or desire ordinarily is measured to some extent with ordinal/interval

scales like Likert scale. In such a case it becomes impossible to have a regression between

desire or need with demand, which is in ratio scale. In this paper we try to tackle this

problem. We try to establish mathematical relationship between (1) need and desire (2)

demand, need and desire and (3) demand, desire and purchasing power.

In order to establish mathematical relationship between the variables mentioned above we

depend heavily on regression techniques like Full Cubic regression and Rational regression.

The reason for avoiding linear regression is because of its poor goodness of fit between the

variables. We also use ratio scale to measure desire and need by translating respondents

desire or need into standard response by asking them how many they need (necessity) or

how many they desire (wish) to consume a particular product. In this way the psychological

aspect of desire or need gets converted into a response which is in ratio scale thus avoiding

ordinal/interval scales.

We use approximately thirty MBA students as our test sample. We also use two products

like soft drink and sachet shampoo to establish relationship between their demands and

desire, need for this test sample.


INTRODUCTION

In marketing parlance its a myth that if need for a product is coupled with desire and

buying power then demand for that product is created. The problem lies in estimating

demand from desire and need. Desire or need being a psychological variable (affective

component) it is most often measured with ordinal scales like Likert scale. Since demand is

measured in ratio scale, it is almost impossible to establish a relationship with desire or

need by using regression techniques. This is what is attempted in this research work.

In order to measure desire or need in ratio scale instead of ordinal scale, what is adopted is

a questionnaire in which simple questions like how many you need (necessity) for

consumption of a product or how many you desire to consume (wish) a product is used.

These types of questions automatically translate the psychological aspects of desire or need

into concrete response like quantity desired or needed, which are in ratio scales. Once this

is done it is easier to apply a regression technique between desire or need and demand.

In order to go ahead we take two products, soft drinks and sachet shampoo for which we

try to predict demand from desire and purchasing power. The reason for choosing two

products is to see if the mathematical relationship can be generalized. We chose

approximately thirty MBA students as test sample.

RESEARCH OBJECTIVE

We have three research objectives. They are

1) To establish a mathematical relationship between Need and Desire.


2) To establish a mathematical relationship between Demand, Need and Desire.

3) To establish a mathematical relationship between Demand, Desire and Purchasing

power.

RESEARCH METHODOLOGY

For each research objective two studies were conducted with two different products like

sachet shampoo and soft drinks so as to check if there exits any generalization. The

studies were conducted on same sample.

Survey instruments were two questionnaires, each for two products. The questionnaires

revealed information like quantity demanded by each respondent, quantity needed by

each respondent, quantity desired by each respondent and purchasing power of each

respondent for the two different products.

Sample size was approximately thirty MBA students.

Curve fitting software CURVE EXPERT PROFESSIONAL 1.6 (trial version) was used

to find out the best fit between the variables for all the three research objectives. For

each study this software checked 62 non linear mathematical forms from its library to

find out the best fitting curve. The software calculated both selection criterias like R 2

(coefficient of determination) and AICC (Akaike Information Criterion). Since our

sample size is low we chose R2 as our selection criteria. This is because AICC criteria

works better with large sample size and distorts its value in case of small sample size.

DATA ANALYSIS
RESEARCH OBJECTIVE-1

STUDY-I

In order to establish a relationship between Need and Desire in case of soft drinks we

analyzed the data, from a sample size of 26, with the software and observed the

following:

EQUATION REVEALED (BEST FITTING): Y= (A+BX)/ (1+CX+DX2)

Where Y=Need and X=Desire and A, B, C, D are coefficients.

STANDARD ERROR = 2.572998

CORRELATION COEFFICIENT (R) = 0.830963

COEFFICIENT OF DETERMINATION (R2) = 0.690499

DOF = 22

AICC = 51.89

A= - 1.466 with standard error = 39.73. Confidence Interval (95%) = - 83.86 to 80.93

B= 3.66 with standard error = 45.37. Confidence Interval (95%) = - 90.44 to 97.77

C= 1.0831 with standard error = 14.2745. Confidence Interval (95%) = - 28.52 to 30.68

D= -.0260 with standard error = .3371. Confidence Interval (95%) = - .726 to 0.64
From the above graphical relationship we observe that as desire increases need

increases but up to a point after which it decreases even if desire increases. This is

because the second order derivative of the equation is found to be negative

For testing hypothesis of individual regression coefficients we postulate that each

regression coefficient is of zero value in case of null hypothesis and not zero value for

alternative hypothesis. We take confidence interval estimation for hypothesis testing

and observe that all of the regression coefficients confidence interval includes the null-

hypothesized value of zero, thus accepting null hypothesis with 95% confidence. In

other words the test is insignificant.


STUDY-II

In order to establish a relationship between Need and Desire in case of sachet shampoos

we analyzed the data, from a sample size of 33, with the software and observed the

following:

BEST FITTING EQUATION: Y= (A+BX)/ (1+CX+DX2)

Where Y=Need and X=Desire and A, B, C, D are coefficients.

STANDARD ERROR = 0.9952

CORRELATION COEFFICIENT (R) = 0.819

COEFFICIENT OF DETERMINATION (R2) = 0.6717

DOF=29

AICC = 2.2478

A= 1.584 with standard error = 0.2675. Confidence Interval (95%) = 1.0377 to 2.13

B= - 0.0744 with standard error = 0.0635. Confidence Interval (95%) = - 0.20 to 0.055

C= - 0.1665 with standard error = 0.0097. Confidence Interval (95%) = -0.186 to -0.146

D= 0.0077 with standard error = 0.006915. Confidence Interval (95%) = .0058 to 0.0096
In case of shampoo also we observe that need increases as desire increases but up to a point

after that it decreases as desire increases.

In this study we observe that regression coefficients a, c and d is statistically significant

because their 95% confidence interval does not include the null-hypothesized value zero

thus rejecting the null hypothesis.

OBSERVATION ON RESEARCH OBJECTIVE-I

It is interesting to note that study-2 reveals a better result than study-1 but it is also

worthwhile to note that both the study expresses same mathematical relationship between

Need and Desire. Though it is too early to state that this is the generalized expression

between Need and Desire but the possibility cannot be ruled out. The correlation coefficient
and the coefficient of determination are also near similar between the two studies but mild.

In study-I the co variances between the parameters or coefficients seem to be very high

signifying dependency between the coefficients. However in study-II the covariance matrix

suggests no dependencies among the parameters or coefficients. The covariance matrix is

not provided here for both the studies.

DATA ANALYSIS ON RESEARCH OBJECTIVE-II

STUDY-I

In this study we try to establish a relationship between Demand, Need and Desire for soft

drinks. The data collected from the questionnaire for a sample size of 26 are analyzed with

the help of curve fitting software and the following observations were made:

Equation Revealed (Best Fitting):

Y=a+ bX1+cX2+dX 12+eX22+fX13+gX23+hX1X2+iX12X2+jX1X22

Where Y=Demand, X1=Desire, X2=Need

And a, b, c, d, e, f, g, h, i and j are coefficients.

Standard Error = 1.4

Correlation Coefficient (r) = 0.963

Coefficient of Determination (r2) = 0.9286

DOF = 16

AICC = 34.31
a=0.2812 with standard error=2.023. CI (95%) = -4.009 to 4.57

b=0.8381 with standard error=0.71. CI (95%) = - 0.688 to 2.33

c=0.2028 with standard error=1.36. CI (95%) = - 2.68 to 3.09

d= - 0.091 with standard error=0.126. CI (95%) = - 0.359 to 0.1764

e= - 0.162 with standard error=0.3846. CI (95%) = - 0.97 to 0.653

f=0.0068 with standard error=0.005. CI (95%) = - 0.0051 to 0.0187

g= - 0.0136 with standard error=0.0348. CI (95%) = - 0.087 to 0.0603

h=0.1357 with standard error=0.19967. CI (95%) = - 0.287 to 0.559

i= - 0.033138 with standard error=0.011099. CI (95%) = -0.0566 to - .009609

j=0.0494 with standard error=0.0229. CI (95%) = .000709 to 0.098092


The graphical representation of the relationship (full cubic) between demand, need and

desire is shown above.

The study reveals a very authentic relationship. Though there exists a relationship between

need and desire as founded in study-I, but still for demand prediction purpose its a useful

equation. The correlation coefficient and coefficient of determination is very high. The

covariance matrix (not provided here) suggests little bit of dependency between parameter

a and c (covariance= - 0.96), and parameter a and b (covariance= - 0.48). Otherwise there

exists no dependency among other parameters.

Testing the null hypothesis that regression coefficients are equal to zero, we find that only

coefficients i and j are significant at 95% confidence interval.


The high values of R and R2 may be because of the relationship that exists between the two

independent variables.

STUDY-II

In this study the quest is to find the relationship between the same Demand, Desire and

Need but with different product. This time we chose sachet shampoo. The test sample size

was 29. The data obtained was analyzed with the same curve fitting software and the

following observation was made:

Equation Revealed (Best Fitting):

Y=a+ bX1+cX2+dX 12+eX22+fX13+gX23+hX1X2+iX12X2+jX1X22

Where Y=Demand, X1=Desire, X2=Need

And a, b, c, d, e, f, g, h, i and j are coefficients.

Standard Error = 2.929

Correlation Coefficient (r) = 0.9365

Coefficient of Determination (r2) = 0.877

DOF = 19

AICC = 77.55

a= - 3.77 with standard error=5.166. CI (95%) = - 14.517 to 7.10

b=2.359 with standard error=1.0507. CI (95%) = 0.1597 to 4.55


c= - 0.8637 with standard error=1.19. CI (95%) = - 3.35 to 1.62

d= - 0.1092 with standard error=0.061057. CI (95%) = - 0.237 to 0.0185

e= 0.042875 with standard error=0.102. CI (95%) = - 0.1708 to 0.2565

f= - 0.001023 with standard error=0.001576. CI (95%) = - 0.004322 to 0.00227

g= 0.002096 with standard error=0.006470. CI (95%) = - 0.01144 to 0.01563

h=0.029339 with standard error=0.109979. CI (95%) = - 0.20085 to 0.2595

i= 0.008235 with standard error=0.008017. CI (95%) = - 0.008545 to 0.0250

j= - 0.008781 with standard error=0.01263. CI (95%) = - 0.035 to 0.01765

The graph below shows the relationship between demand, desire and need for sachet

shampoo. The study reveals a very authentic relationship between demand, desire and

need. For the purpose of demand prediction the equation is of outmost importance in spite

of the existence of a relationship between the two independent variables. It is quite possible

that high value of R and R2 is due to this relationship.

The value of the parameters lies within 95% confidence interval. Though there exists a

little bit dependency between parameter a and b (covariance= - 0.350), and a and c

(covariance= - 0.26) but other parameters are independent of each other. There exists a

high standard error because of low sample size.

Hypothesis testing of regression coefficients reveals only coefficient b significant.


OBSERVATION ON RESEARCH OBJECTIVE-II

It is noteworthy to reveal that both the study suggests same mathematical relationship

(form) between Demand and two independent variables namely Desire and Need. The

result of study-I and study-II are very encouraging and invites further research. High value

of correlation coefficient and coefficient of determination may be due to mild relationship

between the two independent variables but for the purpose of demand prediction the

mathematical relationship holds good. A little bit high value of standard error may be

attributed to small sample size. It is of course adventurous to opine that the equation

revealed is a generalized one; however future research could be promising.


DATA ANALYSIS

RESEARCH OBJECTIVE-III

STUDY-I

Since in research objective II the two independent variables, Need and Desire have

relationship, so we drop variable Need and introduce variable Purchasing power in our

study.

In this study we analyze the data to establish relationship between Demand, desire and

purchasing power for soft drinks. The sample size was 25. We analyze the data using the

software curve expert professional 1.6. The best fitting curve along with other observations

are given below:

Equation Revealed (Best Fitting):

Y=a+ bX1+cX2+dX 12+eX22+fX13+gX23+hX1X2+iX12X2+jX1X22

Where Y=Demand, X1=Desire, X2=Purchasing Power (Budget)

And a, b, c, d, e, f, g, h, i and j are coefficients.

Standard Error = 5.315

Correlation Coefficient (r) = 0.9044

Coefficient of Determination (r2) = 0.8179

DOF = 15

AICC = 100.7569
a= 14.968 with standard error=11.847. CI (95%) = - 10.28 to 40.22

b= - 1.536 with standard error=1.4. CI (95%) = - 4.52 to 1.452

c= 0.0998 with standard error=0.136. CI (95%) = - 0.19 to 0.39

d= 0.027207 with standard error=0.0406. CI (95%) = - 0.059 to 0.113

e= - 0.00225 with standard error=0.0011.CI (95%) = - .0049 to 0.000144

f= 0.000074 with standard error=0.000384. CI (95%) = - 0.000745 to 0.000892

g= 0.000004 with standard error=0.000004. CI (95%) = - 0.00004 to 0.000012

h=0.020042 with standard error=0.011402. CI (95%) = - 0.0042 to 0.044

i= - 0.000354 with standard error=0.000110. CI (95%) = - 0.00058 to 0.000120

j= 0.000005 with standard error=0.000026.CI (95%) = - 0.000051 to 0.000061


The above graph depicts the relationship between Demand, Desire and Purchasing Power.

The relationship is full cubic in nature. With practically no relationship between the

independent variables like desire and purchasing power and high correlation coefficient

(R) and coefficient of determination (R 2), automatically suggests that the equation has not

only high predictive power but also estimates parameter values accurately. The high value

of standard error is attributable to small sample size. The low standard error of the

parameter values also substantiates the accuracy level of the equation. The parameters lie

within 95% confidence interval. The parameters a and b (covariance= - 0.53) have a

dependency. However other parameters are independent of each other.


Testing the null hypothesis that regression coefficients are equal to zero, we find only

coefficient i significant.

STUDY-II

In this study we try to find out a relationship between the same variables like Demand,

Desire and Purchasing Power but for sachet shampoo. The sample size was 28. We analyze

the data with the same software and the following observations were made:

Best Fitting Equation:

Y=a+ bX1+cX2+dX 12+eX22+fX13+gX23+hX1X2+iX12X2+jX1X22

Where Y=Demand, X1=Purchasing power, X2=Desire

And a, b, c, d, e, f, g, h, i and j are coefficients.

Standard Error = 3.0359

Correlation Coefficient (r) = 0.93

Coefficient of Determination (r2) = 0.867

DOF = 18

AICC = 77.818

a= 6.76 with standard error=5.219. CI (95%) = - 4.2 to 17

b= - 0.83 with standard error=0.383. CI (95%) = - 1.64 to 0.03

c= 1.495 with standard error=1.163. CI (95%) = - 0.948 to 3.938


d= 0.0076 with standard error=0.0056. CI (95%) = - 0.004 to 0.0195

e= - 0.122 with standard error=0.08099. CI (95%) = - 0.2924 to 0.047

f= - 0.000057 with standard error=0.000072. CI (95%) = - 0.000207 to 0.000094

g= 0.003391 with standard error=0.00122. CI (95%) = 0.000828 to 0.005886

h=0.060509 with standard error=0.026358. CI (95%) = 0.0051 to 0.115886

i= - 0.000177 with standard error=0.000782. CI (95%) = - 0.001465 to 0.0018

j= - 0.001994with standard error=0.001730. CI (95%) = -0.0056 to 0.00164


The graph above describes demand with respect to desire and purchasing power for sachet

shampoo. In this case also the best fitting equation is a full cubic equation with high

correlation coefficient(R) and coefficient of determination (R2) with practically no

relationship between the independent variables like desire and purchasing power. Small

sample size contributes to high standard error. The equation revealed not only has high

predictive power but also estimates independent variables and parameters or coefficients

accurately. The parameters lie within 95% confidence interval. Though there exists mild

dependency between parameter a and c (covariance= - 0.46) but other parameters are

independent of each other. Leaving aside parameter a (standard error=5.21), other

parameters have very low standard error.


Testing for null hypothesis that regression coefficients are zero, we find that coefficients b, g

and h are statistically significant with 95% confidence interval.

OBSERVATION ON RESEARCH OBJECTIVE-III

Both the studies reveal same full cubic form of relationship between demand, desire and

purchasing power. Is it a step towards a generalized equation to estimate demand from two

independent variables like desire and purchasing power?

CONCLUSION

Research Objective-I & II met with discrepancies. In research objective-I the correlation

coefficient and coefficient of determination for both the studies were mild thus reducing the

strength in the relationship between Need and Desire. More over in study I, all the

coefficients were insignificant for the test sample. In research objective-II the strength in

the relationship of the dependent variable (Demand) and two independent variables like

Desire and Need was very strong in both the studies but this could be due to the mild

relationship between the two independent variables like Need and Desire. May be for the

purpose of demand prediction its good but for estimating parameters or coefficients it may

not be that accurate.

Research Objective-III was the best research and has promising future. The strength of the

relationship between the variables was very good for both the studies. The results could

have been better if the research was carried out on a larger sample size and many samples

as well as few more products. In that case we could have understood whether

generalization exists or not.


The residual analysis in all the studies were not unlikely and with few outliers.

The success of the research lies in ratio scale measurement of need and desire. Instead of

measuring desire with psychological likert scale we translated the psychological aspects of

desire and need into concrete quantity desired as well as needed.

REFERENCE:

David Lester, Judith Hvezda, Shannon Sullivan and Roger Plourda, Maslows Hierarachy

of Needs and Psychological Health, The Journal of General Psychology, 1983, 109, 83-85.

W. Fredman Raaij, Kassaya Wandwossen (1978), Motivation-Need Theories and

Consumer Behaviour, Advances in consumer Research, vol 05, pp: 590-595.

George Brooker (1975), An Instrument to Measure Consumer Self Actualization,

Advances in Consumer Research, vol 02, pp: 563-576

Chris Heathwood, Desire Satisfaction and Hedonism, available at www.google.co.in,

accessed on 22nd April-11, 9-45pm.

Dale o. Jorjenson, Measurement of Desire for Control of the Physical Environment

Psychological Reports, 1978, 42, 603-608.

Wikipedia, the free encyclopedia, Spline Interpolation available at www.google.co.in,

accessed on 26th July at 10pm.


Cubic Spline Interpolation, available at

www.arcetri.astro.it/irlab/library/recipes/bookcpdf/c3-3.pdf, accessed on 25th Oct 12 at 12-

30 pm.

Wikipedia,the free encyclopedia, akaike Information Criterion, available at

en.wikipedia.org/wiki/Akaike_information_criterion, accessed on 3rd May 12 at 8pm

Shibhata,R,Asymtotically Efficient Selection of the Order of the Model for estimating

parameters of a linear process, Annals of Statistics, 80, 147-164.

Shibhata,R, An Optimal Selection of Regression Variables, Biometrics, 68, 45-54.

SAMPLE QUESTIONNAIRE

SACHET SHAMPOO

NAME :

AGE :

SEX :

1.0 WHICH IS YOUR FAVOURITE BRAND OF SHAMPOO?

2.0 HOW MANY SACHETS/POUCH OF SHAMPOO YOU CONSUME IN A WEEK?

3.0 HOW MANY SACHETS/POUCH OF SHAMPOO WOULD YOU WISH OR DESIRE TO CONSUME IN A
WEEK?
4.0 HOW MANY SACHETS/POUCH OF SHAMPOO IS A NECESSARY FOR YOU IN A WEEK?

5.0 HOW MANY SACHETS/POUCH OF SHAMPOO IS A MUST FOR YOU IN A WEEK?

5.0 WHAT IS YOUR MONTHLY BUDGET FOR SACHET SHAMPOO?

You might also like