CHAPTER FIVE

DEMAND ESTIMATION

— Estimating demand for the firm’s product is an essential and continuing

process. After all, decisions to enter new market, decisions concerning
production, planning production capacity, and investment in fixed assets
inventory plans as well as pricing and investment strategies are all
depends on demand estimation.
— The estimated demand function provides managers with an accurate
way to predict future demand for the firm’s product as well as set of
elasticities that allow managers to know in advance the consequences
of planned changes in prices, competitors’ prices, variations in
consumers’ income, or the expected changes in any of the other factors
affecting demand.
— This chapter will provide you with a simplified version of the simple and
multiple regression analyses and techniques that belong to a field called
“Econometrics”, which focuses on the use of statistical techniques and
economic theories in dealing with economic problems.
— Managers may not need to estimate demand by themselves, especially
in big firms. They may assign such technical tasks to their research
department or hire outsider consulting firms (outsourcing) to do the job.
— However, a manager does need at least to have some basic knowledge
of econometrics, to be able to read and understand reports.
— By the end of this chapter, you will be able to do simple demand
estimation, or at least to be able to read and understand the computer
printouts and reports presented to you.
— In the following pages, we will study regression analysis and how it can
be used in demand estimation and how to find the coefficients of
demand equation. The question is how these coefficients are estimated,
or generally how demand is estimated.

Page 1 of 22
 Regression Analysis
— Regression analysis is a statistical technique for finding the best
relationship between dependent variable and selected independent
variable(s).
— Dependent variable: depends on the value of other variables. It is the
primary interest to researchers.
— Independent (explanatory) variable: used to explain the variation in
the dependent variable.
— Regression analysis is commonly used by economists to estimate
demand for a good or service.
— There are two types of statistical analysis:
1. Simple Regression Analysis:
The use of one independent variable
Y = a + bX +µ
Where:
Y: dependent variable, amount to be determined
a: constant value; y-intercept
b: slope (regression coefficients), or parameters to be estimated (it
measures the impact of independent variable)
X: independent (or explanatory) variable, used to explain the variation in
the dependent variable
µ: random error

2. Multiple Regression Analysis:

The use of more than one independent variable
Y = a + b1X1 + b2X2 + ….+ bkXk + µ
(k= number of independent variable in the regression model)
— The well-known method of ordinary least squares (OLS) is used in our
regression analysis. Some of the assumptions of OLS include.
1. Independent (explanatory) variables are independent from the
dependent variable and independent from each other.

Page 2 of 22
 2. The error terms (µ) are independent and identically distributed
normal random variables, with mean equal to zero.

How regression analysis is done?

There are certain steps to conduct reg. analysis.
1. Identify the relevant variables
2. Obtain data on the variables
3. Specify the regression model
4. Estimate the parameters (coefficients)
5. Interpret the results
6. Statistical evaluation of the results (testing statistical significance of
model)
7. Use the results in decision making (forecasting using reg. results).

1. Identification of variables and data collection:

— Here, we try to answer the question of what variables to be included in
regression analysis; what variables are important
— The role of economic theory, the availability of data and other
constraints help in determining which variables to be included
— The role of economic theory:
o In estimating the demand for a particular good or service, we have
to determine all the factors that might influence this demand.
o Economic theory helps by considering the right set of variables to
be considered when estimating demand for the good. It also helps
in determining the relationship between Qd and these variables;
e.g.; we expect a negative sign for the coefficient of P because of
the negative relation between P and Qd. if the good under
consideration is a normal good, we expect a positive sign the for
the coefficient of income because of the positive relationship
between income and the demand for normal good, etc…

Page 3 of 22
— In reality, however, the availability of data and the cost of generating
new data may determine what to include.
o Some variables are easy to find and to measure and quantify, like
prices, number of consumers and may be income …
o Sometimes it is difficult to get data for the original variables Ö use
proxy
o Some variables are hard to quantify such as location (urban,
suburban, rural) or tastes and preferences (like, dislike, indifferent,
…) ⇒ use dummy (binary) variables (1 if the event occurs and
zero otherwise. E.g., urban, 0 otherwise) or (1 if like, 0 otherwise).
o The main types of data used in regression are:
1. Cross sectional: provide information about the variables
for a given time period (different individuals, goods, firms,
countries …)
2. Time series: give information about variables over a
number of periods of time (years, months, daily,…)
3. Pooled (Panel): Combinations of cross section and time
series data
o Data for studies pertaining to countries, regions, or industries are
readily available and reliable.
o Data for analysis of specific product categories may be more
difficult to obtain. The solution is to buy the data from data
providers, perform a consumer survey, focus groups, etc.

2. Specification of the model:

— This is where the relation between the dependent variable (say, Qd) and
the factors affecting it (the independent or explanatory variables) are
expressed in regression equation.
— The estimation of regression equation involves searching for the best
linear relationship between variables.

Page 4 of 22
— The commonly used specification is to express the regression equation
in the additive liner function.
— If equation is non-liner such as Multiplicative such as Q = APbYc,
transform nonlinear into linear using logarithm
— It is double log (log is the natural log, also written as ln)
Log Q = a + bLog P + cLog Y
— For the purpose of illustration, let us assume that we have obtained
cross-sectional data on college students of 30 randomly selected
college campuses during a particular month, with the following equation.
Qd = a + b1P + b2T - b3Pc + b4L + µ
Where:
Qd: Quantity demanded of pizza (average number of slices per capita
per month)
P: Average price of slice of pizza (in cents)
T: Annual tuition as proxy for income (in thousands of $s)
Pc: Price of cans of soft drinks (in cents)
L: Location of campuses (1 if urban area, 0 otherwise)
a: Constant value or Y intercept
bi: Coefficient of independent variables to be estimated (slope)
µ: Random error term standing for all other omitted variables.
— The effect of each variable (the marginal impact) is the coefficient of that
variable in the regression equation. The impact of P is b1 (dQ/dP), the
impact of T is b2 (dQ/dT), etc…..
— The elasticity of each variable is calculated as usual:
dQ P P
o Ed = × = b1 ×
dP Q Q
dQ T T
o ET = × = b2 ×
dT Q Q
dQ Pc P
o E Pc = × = b3 × c
dPc Q Q

dQ L L
o EL = × = b4 ×
dL Q Q

Page 5 of 22
3. Estimation of the regression coefficients:
— Given this particular set up of regression equation we can now estimate
the values of coefficients of the independent variable as well as the
intercept term. Using ordinary least squares (OLS) method
— Usually statistical and econometrics packages are used to estimate
regression equation using excel and many other statistical packages
such as SPSS, SAS, EViews, LimDep, TSP…
— Results are usually reported in regression equation or table format,
containing certain information such
Qd = 26.67 – 0.088P + 0.138T – 0.076Pc – 0.544L
(0.018) (0.0087) (0.020) (0.884)
R2 = 0.717 (The coefficient of determination)
2
R = 0.67 (Adjusted R2
SE of Q estimate (SEE) = 1.64
F = 15.8 (F-Statistics)
Standard errors of the coefficient are listed in parentheses.

4. Interpretation of the regression coefficients:

— Analyzing regression results involves the following steps
o Checking the signs and magnitudes
o Computing elasticity coefficients
o Determining statistical significance
— It also involves two tasks:
o Interpretation of coefficients
o Statistical evaluation of coefficients
— What are the expected magnitude and signs of the estimated
coefficient?
— Check signs of the coefficient according to economic theory and see if
they are as expected:
P: when price increases, Qd for Pizza decreases (negative sign)

Page 6 of 22
 T: Sign for proxy of income depends on whether pizza is a normal or
inferior good. (+,-)
Pc: expected sign for Pc is (-) because of complementary relation (Pc
increases, demand for pizza decreases)
L: Expected sign is (-) because in urban areas students have varieties
of restaurants (more substitutes), ⇒ they will consume less pizza
than their counterparts in other areas will.
— Check the effect of each independent variable on the dependent
variable according to economic theory.
— With regard to magnitude, we can see that each estimated coefficient
tells us how much the demand for pizza will change relative to a unit
change in each of the independent variables.
b1: a unit change in P changes Qd by 0.088 units in the opposite
direction.
b2: for a $1000 change in tuition, demand changes by 0.138 units.
b3: for a unit change in Pc, demand changes by 0.076 in opposite
direction
b4: students in urban areas will buy about half (0.544) less than those in
other areas.
— Magnitude of regression coefficients is measured by elasticity of each
variable.
If P=100 (cents), T=14($000), Pc=110 (cents), L= 1
Qd = 26.67 – 0.088(100) + 0.138(14) – 0.076(110) – 0.544(1) = 10.898
100
E d = −0.088 × = −0.807 is somewhat inelastic
10.898
14
E T = 0.138 × = 0.177 no great impact
10.898
110
E pc = −0.076 × = −0.767 is inelastic
10.898
1
E L = −0.544 × = −0.05 dose not really matter
10.898

Page 7 of 22
 5. Statistical evaluation of the regression results
— Regression results are based on a sample.
— How confident are we that these results are truly reflective of
population?
— The basic test of the statistical significance using each of the estimated
regression coefficients is done separately using t-test.

t- Test
— t-test is conducted by computing t-value or t-statistic for each of the
estimated coefficient, to test the impact of each variable separately.
— t = (estimated coefficient – population value of the coefficient) / standard
error of the coefficient
b̂ i − b i
t=
Sb i

b̂ i
bi is assumed equal to zero in the null hypothesis ⇒ t=
Sb i

b̂ i
— We usually compare the estimated (observed) t-value ( t = ) to the
Sb i

critical value from t-table, t α, n-k-1

where:
α = level of significance (it is an error rate of unusual samples with their
false inference from a sample to a population)
n = number of observations,
k = number of independent/ explanatory variables.
n-k-1 = degrees of freedom: the number of free or linearly independent
sample observations used in the calculation of statistic.

To compare estimated t-value to critical t-value.

Page 8 of 22
First: form the hypotheses:
— Null hypothesis, H0: bi = 0
The null hypothesis means that there is no relationship between
independent variable and dependent variable. i.e. the variable in
question has no effect on dependent variable when other factors are
held constant.
— Alternative hypothesis, Ha: bi ≠ 0
The alternative hypothesis means that there is linear relationship
between independent variable and the dependent variable.
— Since there are two hypotheses, rejecting one implies the other is
automatically accepted (not rejected)

Second: Calculate t-value (observed t-value) of all independent

variables:
— In the pizza example:
− 0.088
tp = = −4.89
0.018
0.138
tT = = 1.58
0.087
− 0.076
tp = = −3.80
c
0.020
− 0.544
tL = = −0.615
0.884

Third: Determine your level of significance (say 5%).

— Using the rule of two, we can say that estimated coefficient is
statistically significant (has an impact on the dependent variable) if t-
value is greater than or equal to 2.
— In the pizza example above
o P & Pc are greater than 2 ⇒ statically significant ⇒ the whole
population has an effect on demand.

Page 9 of 22
 o T& L are Less than 2 ⇒ statistically insignificant ⇒ the population
as no effect on demand.
— α = 0.05, n = 30, k= 4
t α, n-k-1 = t 0.05, 30-4-1 = t 05, 25 = 2.060

Reject H0
Accept H0 Reject H0

-2.060 0 2.060

Fourth: Conclusion
— Compare absolute t-value with the critical t-value:
— If absolute t-value > critical t-value, reject H0 and conclude that
estimated coefficient is statistically significant, otherwise accept H0.
var. t-value critical Decision Conclusion
P 4.889 > 2.060 reject significant
T 1.683 < 2.060 don’t reject not significant
Pc 3.800 > 2.060 reject significant
L 0.615 < 2.060 do not reject not significant
— Significant means there is linear relationship between the independent
and dependent variables. The independent variable has a true impact
on the dependent variable or it is important in explaining variation in the
dependent variable (Qd in our example),
— Not significant means there is no linear relationship between the
independent and dependent variables

Page 10 of 22
 Testing the performance of the Regression Model – R2
— The overall results are tested using the coefficient of determinations, R2.
— R2 is to evaluate the deterministic power of the regression model.
— R2 is used to test whether the regression model is good, i.e., to test the
goodness of fit of the regression line to actual data.
— R2 measures the percentage of total variation in the dependent variable
that is explained by the variation in all of the independent variables in
the regression model.
RSS ESS
R2 = = 1−
TSS TSS
Where,
TSS: total sum of squares; it is the sum of squared total variation in the
dependent variables around its mean (explained & unexplained)
RSS: regression sum of square (explained variation)
ESS: Error of square (unexplained variation)
0 < R2 < 1
— R2 = 0 ⇒ variation in the dependent variable cannot be explained at all
by the variation in the independent variable.
— R2 = 1 ⇒ all of the variation in the dependent variable can be explained
by the independent variables
— For statistical analysis, the closer R2 is to one, the better the regression
equation; i.e., the greater the explanatory power of the regression
equation.
— Low value of R2 indicates the absence of some important variables from
the model.
— In our example, R2 = 0.717. This means that about 72% of the variation
in the demand for pizza by college students can be explained by the
variation in the independent variables.
— The value of R2 is affected by:
o The value of independent variables: the way R2 is calculated
causes its value to increase as more independent variables are

Page 11 of 22
 added to the regression model even if these variables do not have
any effect on the dependent variable.
o Types of data used: other factors held constant, time series data
generally produce a higher R2 than cross-sectional data. This is
because of series data has built-in trend over time to keep
dependent and independent variables moving closely together. A
good example of this is a time-series analysis of aggregate
consumption regressed on aggregate disposable income.
Regression analysis of this consumption function commonly
produces R2 of .95 and above.

Adjusted R2, R 2
(The adjusted coefficient of determination),
— As more and more variables are added, R2 usually increases.
— Therefore, we use R 2 to account for this “inflation" in R2 so that
equation with different numbers of independent variables can be more
fairly compared
— In our example, R 2 = 0.67 which indicates that about 67% of the
variation in Qd of pizza is explained by the variations in the independent
variables while 33% of these variations are unexplained by the model.
— R 2 is calculated as

R 2 = R2 −
k
n − k −1
(1 − R2 )
4
R 2 = 0.72 − (1 − 0.72) = 0.67
25

Page 12 of 22
 F- test
— F-test is used to test the impact of overall explanatory power of the
whole model, or the joint effect of all explanatory variables as a group.
(i.e., testing the overall performance of the regression coefficients)
— F-test measures the statistical significance of the entire regression
equation rather than of each individual coefficient as the t-test is
designed to do.
— If it is used in simple regression (i.e., for a regression equation with only
one independent variable), then in effect it provides the same test as the
t-test for this particular variable.
— The F-test is much more useful when two or more independent variables
are used.
— It can then test whether all of these variables taken together are
statistically significant from zero, leaving the t-test to determine whether
each variable taken separately is statistically significant.
— As in the t-test, we have to set our hypotheses.

First: form the hypotheses:

— H0: All bi = 0 ( b1= b2 = b3 =…= bk = 0)
(k= number of independent variable in the regression model)
There is no relation between the dependent variable and independent
variables. The model cannot explain any of the variation in the
dependent variable
— Ha: at least one bi ≠ 0
A linear relation exists between dependent variable and at least one of
the independent variables.

Page 13 of 22
Second: Calculate F-value
— F = (explained variation/k) / (unexplained variation/n-k-1)

∑ (Ŷ − Y )
2
/k RSS / k
— F= =
∑ (Y − Ŷ ) ESS / n − k − 1
2
/n − k −1
RSS: regression some of squares
ESS: error sum of squares
n: number of observation
k: number of explanatory variables
— But F maybe re-written in term of R2 as follows
R2 / k
— F=
( )
1 − R2 / n − k − 1
— In our example: F=15.8
— The greater the value of F-statistics, the more confident the researcher
would be that variables included in the model have together a significant
effect on the dependent variable, and the model has a high explanatory
power.
— Thus, the F test examines the significance of R2

Third: Determine your level of significance (say 5%)

— F α, k, n-k-1

α: level of significance
k: number of independent variables
n: number of observations or sample size
k, n-k-1: degrees of freedom
— in our example: F.05, 4, 30-4-1 = F.05, 4, 25 = 2.76

Fourth: Compare F-value (observed F) with critical F-value

— If F > Critical F value Ö reject H0 and conclude that a linear relation
exists between the dependent variable and at least one of the
independent variable

Page 14 of 22
 F= 15.8 > F.05, 4, 25 = 2.76
Reject H0, there is a linear relationship between the dependent variable
and at least one of the independent variables. The entire regression
model accounts for a statistically significant proportion of the variation in
the demand for pizza.

6. Forecasting:
— Future values of demand can easily be predicted or forecasted by
plugging values of independent variables in the demand equation.
— Only we have to be confident at a given level that the true Y is close to
the estimated Y.
— Since we do not know the true Y, we can only say that it lies between a
given confidence interval.
— The interval is Ŷ+ t α, n-k-1 X SEE
— Confidence interval tells that we are, say, 95% confident that the
predicted value of Qd lies approximately between the two limits.

Implications of Regression Analysis for Decision Making

— Regression analysis can show which are the important factors, judging
from whether the variable is significant or not (what variables passed the
t-test and what did not pass.)
— The magnitude and the level of coefficient indicate the importance of the
variables.
— Computing elasticity will help in determining what may happen to total
revenue.

Page 15 of 22
 Correlation
— A measure of association is the correlation coefficient, r.
— Correlation coefficient, r, indicates the strength and direction of a
linear relationship between two random variables
— The correlation is defined only if both of the standard deviations are
finite and both of them are nonzero
— If r = 0 the variables are independent
— If r = 1, the correlation is perfect and positive. This is the case of an
increasing linear relationship
— If r = -1, the correlation is perfect and negative. This is the case of an
decreasing linear relationship
— If the value is in between, it indicates the degree of linear dependence
between the variables
— The closer the coefficient is to either -1 or 1, the stronger the correlation
between the variable
— The correlation coefficient is defined in terms of the covariance:
cov( X, Z ) σ
— corr ( X, Z ) = = XZ
var( X) var( z ) σ X σ Z

–1 ≤ corr(X,Z) ≤ 1
corr(X,Z) = 1 mean perfect positive linear association
corr(X,Z) = –1 means perfect negative linear association
corr(X,Z) = 0 means no linear association

Association and Causation

— Regressions indicate association, but beware of jumping to the
conclusion of causation
— Suppose you collect data on the number of swimmers at a beach and
the temperature and find:
Temperature = 61 + .04 Swimmers,
and R2 = .88.

Page 16 of 22
 o Surely the temperature and the number of swimmers is positively
related, but we do not believe that more swimmers caused the
temperature to rise.
o Furthermore, there may be other factors that determine the
relationship, for example the presence of rain, or whether or not it
is a weekend or weekday.
— Education may lead to more income, and also more income may lead to
more education. The direction of causation is often unclear. But the
association is very strong.

Page 17 of 22
 Regression Problems

Identification Problem:
— The identification problem refers to the difficulty of clearly identifying the
demand equation because of the effects of both supply and demand that
are often reflected in data used in the analysis.
— The estimation of demand may produce biased results due to
simultaneous shifting of supply and demand curves.
— Advanced estimation techniques, such as two-stage least squares and
indirect least squares, are used to correct this problem.

Multicollinearity
— Two or more independent variables are highly correlated, thus it is
difficult to separate the effect each has on the dependent variable.
— Passing the F-test as a whole, but failing the t-test for each coefficient is
a sign that multicollinearity exists.
— A standard remedy is to drop one of the closely related independent
variables from the regression

Autocorrelation
— Also known as serial correlation, occurs when the dependent variable
relates to the independent variable according to a certain pattern.
— Possible causes:
o Effects on dependent variable exist that are not accounted for by the
independent variables.
o The relationship may be non-linear
— The Durbin-Watson (DW) statistic is used to identify the presence of
autocorrelation.
— To correct autocorrelation consider:
o Transforming the data into a different order of magnitude
o Introducing leading or lagging data

Page 18 of 22
 APPENDIX

Using MS-Excel in Regression:

1. Open a new file in Excel program and name it example-1.
2. Label the first column on your file Q, the second P. Insert the following
data collected by the research team in 1999.
Observations 1 2 3 4 5 6 7 8 9 10
Meals 180 590 430 250 275 720 660 490 700 210
Price 475 400 450 550 575 375 375 450 400 500
Observations 11 12 13 14 15 16 17 18 19 20
Meals 150 120 500 150 600 220 200 280 160 300
Price 480 650 300 330 350 660 650 540 720 600

3. Open Tools in the menu bar, choose Data Analysis and move to number
4 bellow. If Data Analysis does not appear on the Tools menu, click
“Add-Ins….”. on the Tools menu. In the Add-Ins window, choose
“Analysis ToolPack” and press OK.

Page 19 of 22
4. Now, open tools once again and click the new title “Data Analysis”,
and click Regression in the Data Analysis window, then OK.

5. In the regression dialog box, For “Input Y range”, select the Q column
of your data including the label cell. Move the cursor to “Input X range”,
select the P column from your data. Check the square beside “labels”.
6. Click “Output Range”, and click a cell bellow your data where you like
printing of the results to start.

Page 20 of 22
 7. Click OK. The results print out will look exactly as follows:

SUMMARY OUTPUT

Regression Statistics
Multiple R 0.66119647
R Square 0.43718077
Adjusted R Square 0.40591304
Standard Error 158.38729
Observations 20

ANOVA
df SS MS F Significance F
Regression 1 350756.1452 350756.1 13.98184982 0.001501539
Residual 18 451557.6048 25086.53
Total 19 802313.75

Page 21 of 22
 Coefficients Standard Error t Stat P-value
Intercept 903.598862 149.8239232 6.031072 1.05758E-05
P -1.1075257 0.296190743 -3.73923 0.001501539

Excel Exercise

1. Use the data in your text page 169 to confirm the results presented in
the text
2. The following table contains data on the number of apartments rented
(Q), the rental price (P) in BDs, the amount spent on advertisement (AD)
in hundreds of BDs, and the distance between the apartments and the
university (Dis) in miles.

Q 28 69 43 32 42 72 66 49 70 60
P 250 400 450 550 575 375 375 450 400 375
AD 11 24 15 31 34 22 12 24 22 10
Dis 12 6 5 7 4 2 5 7 4 5

a. Use Excel program to regress Q on the three explanatory variables.

b. Write the estimated demand equation for apartments.

Page 22 of 22