ECP6705

Chapter 3
Measuring and Using Demand

Regression analysis:

Elasticity:

3.1 Regression: Estimating Demand

The Capital Grille is a chain of upscale steak restaurants. Suppose that you are an upper-
level executive working for the chain. To make better decisions, you want to know the
demand for your company’s dinners. Assume that the demand function for steak dinners
at your restaurants can be written as:

Qd =a−( b∗P ) +η

d
Q =¿

P=¿

η=¿

Why don’t the points fall on a

precisely straight line?

Page 1 of 16
ECP6705
Chapter 3

Regression analysis selects the line that best fits the data.

Do we want the distance between the line and the dots to be big or small?

Residuals: the differences between the actual Qd and predicted Qd .

Ordinary Least Squares (OLS) Regression: the estimated line that minimizes the
sum of the squared residuals between actual quantities and the predicted quantities

Page 2 of 16
ECP6705
Chapter 3

Using Excel for Regression

Step 1: Enable Excel’s Analysis Tool Pack.

Step 2: Ensure data is entered correctly.

Page 3 of 16
ECP6705
Chapter 3

Using Excel for Regression (continued)

Step 3: Tell Excel to estimate the equation.

Excel Output

Page 4 of 16
ECP6705
Chapter 3
Recall, we were trying to estimate a and b in the following univariate equation.

Qd =a−( b∗P ) +η

We will NOT know a and b for sure. The numbers we get from Excel are estimates. We’ll
call the estimates a^ and b^ .

a^ =¿

^
b=¿

d
Q =¿

3.2 Interpreting the Results of Regression Analysis

Page 5 of 16
ECP6705
Chapter 3
A. Estimated Coefficients
The estimated demand curve might not be the same as the actual demand curve.

Standard Error of a coefficient is a measure of how much the estimated coefficient is

likely to change if another set of data is collected and another regression estimated.

Do we want this to be big or small?

• A small standard error means the estimated coefficient is likely

___________________________________________ to the true value.

• A large standard error means the estimated coefficient is likely

___________________________________________to the true value.

B. Confidence Intervals

Confidence interval: a range that for repeated samples has a pre-selected probability
of containing the true value of the coefficient.

Confidence Intervals Depend on:

1.

2.

What confidence interval is reported in Excel?

How is this interpreted?

We are 95% certain that the actual value of a is between the upper and lower
bounds.

In this case,

____________ confidence intervals yield less certainty about the quality of the
coefficient estimate.

C. Hypothesis Testing

Page 6 of 16
ECP6705
Chapter 3
Will demand for dinners at the steak restaurant you manage, Capital Grille, be affected
by the price of a dinner at Sea Harvest, a competing fish restaurant?

Let’s suppose that we’ve been keeping track of the price of a dinner at Sea Harvest
as well as the prices and quantity demanded for dinners at Capital Grille. We’ll add
a column to our Excel table for the price of dinner at Sea Harvest. Then, we’ll
estimate the regression again, but this time we’ll include both our dinner price (P)
and the price of a dinner at Sea Harvest (S). We use a multivariate equation, like
the following.
d
Q =a−( b∗P ) +( c∗S)+ η

In Excel, we get the following estimate for c.

The coefficient estimate for c is ____________, but how confident/certain are we that 8.04
is the true effect? Is our estimate statistically significant?

Determining Statistical Significance via a Hypothesis Test

1. Create the null hypothesis (that the coefficient is zero) and the alternative hypothesis
(that the coefficient is not zero).

2. Determine the significance level. How confident do you want to be?

Note that the higher the confidence level, the ____________ the estimated range (i.e.
confidence interval).

Page 7 of 16
ECP6705
Chapter 3
We are trying to figure out if our estimate for c is statistically significant at the 95% level.
Can we say, “We are 95% confident/certain that Sea Harvest’s prices affect demand for
Capital Grille”?

Three Ways to Test our Hypothesis

1. If the t-statistic is _________________ than the critical value, then the estimate is
statistically significant.

• t-statistic The estimated coefficient divided by its estimated standard error.

• Critical value The value of the t-statistic above which the null hypothesis is
rejected.

• For a large number of observations, the 95 percent confidence level

(the 5 percent significance level) has a critical value of 1.96

EX) In Table 3.1, is the coefficient estimate for Competitor Price (c) statistically significant
at the 5% level?

t-statistic =

critical value =

t-statistic ____ critical value 

2. P-value: the significance level of testing the null hypothesis.

EX) In Table 3.1, is the coefficient estimate for Competitor Price (c) statistically significant
at the 5% level?

P-value =

Is this enough statistical significance?

Page 8 of 16
ECP6705
Chapter 3

3. Confidence Interval

If the 95% confidence interval includes zero:

If the 95% confidence does not include zero:

EX) In Table 3.1, is the coefficient estimate for Competitor Price (c) statistically significant
at the 5% level?

95% Confidence Interval:

Does it include zero?

D. Fit of the Regression

R2 statistic: A measure of how much of the variation in the observed values of the
dependent variable is captured by the predicted values of the regression.

EX) On page 4 of this note outline, what is R 2?

R2 =

Which regression has a lower R2?

Page 9 of 16
ECP6705
Chapter 3
3.3 Limitations of Regression Analysis
Regression Analysis is only as good as the model that created it.

If you estimate the wrong model, then regression analysis will lead an incorrect
conclusion.

Page 10 of 16
ECP6705
Chapter 3

3.4 Elasticity

Price Elasticity of Demand: a measure that indicates the degree of consumer response to
a price change.

ε=| percentage change ∈quantity demanded

percentage change∈ price |
How to interpret the elasticity coefficient:

1.

2.

3.

Examples of actual price elasticity of demand coefficients:

Gasoline:

Electricity:

Cocaine:

Health services:

Foreign travel (LR):

Page 11 of 16
ECP6705
Chapter 3
Demand is elastic on the upper region of the demand curve because, numerically, %
changes in Q are relatively large while % changes in P are relatively small.

Elasticity and Slope are not the same thing, but they are related.

Which is more elastic, the demand for cigarettes or the demand for potato chips?

Extreme Cases of Demand

Page 12 of 16
ECP6705
Chapter 3

Determinants of Elasticity

1. Number of Substitutes

more substitutes 

few substitutes 

2. Share of Budget

big share of budget 

small share of budget 

Page 13 of 16
ECP6705
Chapter 3

Elasticity and Total Revenue (TR)

Relationship Between Elasticity (ED) and Total Revenue

If ½Ed½> 1 (elastic)

If ½Ed½< 1 (inelastic)

If ½Ed½= 1 (unitary elasticity)

How Total Revenue Changes Along a Demand Curve

When the New York City Opera was faced with a growing deficit, it cuts its ticket prices by 20%
hoping to attract more customers.

At the same time, the New York City Transit Authority raised subway fares to reduce its growing
deficit.

Was one of these two opposite approaches to reducing the deficit necessarily wrong? Explain
what each is assuming about the elasticity of demand for their product.

Page 14 of 16
ECP6705
Chapter 3
Income elasticity of demand A measure of how strongly the quantity demanded
responds to a change in consumers’ income

( percentage change∈quantity demanded )

ε INC =
( percentagechange ∈income)

ε INC > 0 

ε INC < 0 

Cross-price elasticity of demand A measure of how strongly the quantity demanded

changes when the price of a related product changes

( percentage change∈quantity demanded)

ε CROSS=
( percentage change∈ price of the related good )

ε CROSS> 0 

ε CROSS< 0 

3.5. Regression Analysis and Elasticity

At Capital Grille, assume that demand for steak changes and looks like the following.

Does this look linear?

Would it be a good idea to estimate the

values for a linear

demand curve?

Page 15 of 16
ECP6705
Chapter 3

A model that looks like the graph above is log-log.

d
ln Q =a+ ( b∗ln P ) +η

In the model above, the elasticity of demand at any point is b.

Page 16 of 16