Chapter 4

Estimating Demand
Solutions to Exercises
1. a. The coefficient 0.582 is the price elasticity of demand. The coefficient 0.401 is the income
elasticity of demand. The coefficient 0.211 is the cross price elasticity of demand between
gasoline and cars, complementary goods.
b. The t value for the "own" price coefficient is 5.54, which is significant at the less than 1%
level. The t value for income elasticity is 2.06, which is significant at better than the 5% level.
The cross price elasticity t-value is 1.35, which is significant only at the 20% level.
2. a. 2.174
b. 0.461
c. 1.909
d. The demand is very price elastic and heavily influenced by the price of competitive
goodsmeat and poultry.
e. The quantity of haddock demanded would increase by .461(5%) = 2.31%
3. a. Dependent variable: Sales
Variable

DF

Parameter
Standard
T-ratio
Estimate
Error
_______________________________________________________________________
Intercept
1
2.4365
1.105
2.205
Price
1
.3750
.239
1.570
Income
1
5.9492
2.140
2.780
Advertising
1
6.2500
1.950
3.205

25

This edition is intended for use outside of the U.S. only, with content that may be different from the U.S. Edition. This may not be resold,
copied, or distributed without the prior consent of the publisher.

26

Chapter 4/Estimating Demand

Source of Variation
Regression
Residual
Total

DF
3
22
25

Sum of Squares
1187.343
409.750
1597.093

Mean Squares
395.781
18.625

R-Square: 0.743
F-Ratio: 21.250
b. H0: j = 0;

Ha: j 0

Reject H0 at the .05 level of significance if t (i.e., T-ratio) < -t.025,22 = -2.074 or t > t.025,22 =
+2.074. Since the calculated t-value is greater than the t-value from the table for the income and
advertising variables, one rejects the hypothesis at the .05 significance level that there is no
relationship between these variables and sales.
c. H0: 1 = 2 = 3 = 0
Ha : At least one j 0
Reject H0 at the .05 level of significance if F (i.e., F-Ratio) > F.05,3,22 = 3.05. Since the calculated
F-value is greater than the F-value from the table, one rejects at the .05 significance level the
hypothesis that there is no relationship between any of the explanatory variables and sales.
4. a. Dependent variable: SALES N: 10
Multiple R: 0.874
Squared multiple R: 0.764
Adjusted squared multiple R: 0.697
Standard error of the estimate: 17.062
Variable
Coefficient Std error Std coef
CONSTANT
344.585
84.245
0.000
PROMEXP
0.106
0.164
0.177
SELLPR
12.112
4.487 0.736

Source
Regression
Residual

Tolerance
0.4531851
0.4531851

Analysis of Variance
Sum-of-squares DF Mean-square
6612.203
2
3306.102
2037.797
7
291.114

F-ratio
11.357

T
P(2 tail)
4.090 0.005
0.648 0.537
2.699 0.031

P
0.006

Y = 344.585 + .106X1 12.112X2

This edition is intended for use outside of the U.S. only, with content that may be different from the U.S. Edition. This may not be resold,
copied, or distributed without the prior consent of the publisher.

Chapter 4/Estimating of Demand

27

b. b1 = .106
A one-unit (i.e., $1,000) increase in promotional expenditures increases expected sales by .106
( 1,000) = 106 gallons, all other things being equal.
A one-unit (i.e., $1.00) increase in the selling price decreases expected sales by 12.112 ( 1,000)
= 12,112 gallons, all other things being equal.
c. The computer output indicates that only selling price (X2) is statistically significant at the 0.05
level. The t-value for significance is t.025,7 = 2.365.
d. R2 = 0.764

The model explains about 76 percent of the variation in paint sales.

e. The F-ratio from the computer output is 11.357. The F-value for statistical significance at the
.05 level is F.05,2,7 = 4.74. Therefore we reject the null hypothesis and conclude that the
independent variables are useful in explaining paint sales.
f. Y' = 344.585 + 0.106(80) 12.112(12.50)
= 201.665
or 201,665 gallons.
g. (i) EA= (Y/X1)(X1/Y)
= .106 (80/201.665)
= .0420
(ii) EL = (Y/X2)(X2/Y)
= 12.112 (12.50/201.665)
= .751
5. a. Variable
X1
X2

Coefficient
.24
.27

Standard Error t-statistic Decision*

.032
7.50
Significant
.070

3.86

Significant

*Variable Xi is significant at the .05 level if t > t.025,30 = +2.042 or t < 2.042.
b. R2 = 0.64
Approximately 64% of the variation in sales is "explained" by the regression equation.

This edition is intended for use outside of the U.S. only, with content that may be different from the U.S. Edition. This may not be resold,
copied, or distributed without the prior consent of the publisher.

28

Chapter 4/Estimating Demand

c. Ho: All j = 0; Ha: At least one j 0.
Reject Ho at the .05 level if the calculated F-value is greater than F.05,2,30 = 3.32. Since F =
31.402 > 3.32, one rejects Ho and concludes that the regression model "explains" a significant
proportion of the variation in quarterly sales of the cold remedy.
d. Ho: No positive or negative autocorrelation
Ha: Positive or negative autocorrelation
Reject Ho at the .05 significance level if the calculated Durbin-Watson d-statistic is either less
than dL = 1.16 from Table 6 (m = 3 and n = 33) or greater than (4 dU) = (4 1.55) = 2.45. Since
d = .4995, one concludes that significant (positive) autocorrelation is present.
e. The presence of autocorrelation will cause the least-squares procedure to underestimate the
sampling variance of the estimated regression coefficients. As a result the t-test will no longer
yield reliable conclusions concerning the statistical significance of the individual explanatory
variables. Also, the overall measures of goodness-of-fit and explanatory power of the regression
equation, namely the coefficient of determination and F-test, will no longer provide a reliable
indication of the significance of the economic relationship obtained. The correlation coefficient
between X1 and X2 would be useful in checking for multicollinearity.

6. a. ED = -.48 (exponent of P)
EY = 1.08 (exponent of Y)
b. The percentage change in furniture expenditures resulting from a one-percent change in the
value of private residential construction per household.
c. Depends on how results are to be used:
(1) physical unitsproduction planning
(2) actual dollar salesfinancial planning
If F is expressed in constant dollar terms, then Y should be also.

This edition is intended for use outside of the U.S. only, with content that may be different from the U.S. Edition. This may not be resold,
copied, or distributed without the prior consent of the publisher.

Chapter 4/Estimating of Demand

29

7. a.

b. Y = 6.279 + 3.956X
The estimated slope coefficient indicates that the selling price increases by $3,956 for each 100
square feet increase in the size of a house.
c. se = 10.132
sb = .396
t = (3.956 0)/.396 = 9.990
Since the calculated t-value is greater than the t-value from the table (t.025,13 = 2.160 or + 2.160),
one rejects the hypothesis at the .05 significance level that there is no relationship between the
selling price and the size of a house.
d. R2 = .885
e. Source of
Variation
Regression
Residual
Total

Sum of
Squares
10,244
1,335
11,579

Degrees of
Freedom
1
13
14

Mean
Squares
10,244
102.7

F = MSR/MSE = 10,244/102.7 = 99.75

This edition is intended for use outside of the U.S. only, with content that may be different from the U.S. Edition. This may not be resold,
copied, or distributed without the prior consent of the publisher.

30

Chapter 4/Estimating Demand

Since the calculated F-value is greater than the F-value from the table (F.05,1,13 = 4.67), one rejects
at the .05 significance level the hypothesis that there is no relationship between the selling price
and the size of a house.
f.

y' = 6.279 + 3.956(15) = 65.619 or $65,619

Approximate 95% prediction interval:

y' = 65.619 2(10.132) = 45.355 to 85.883,
or $45,355 to $85,883.
8. a. 0.95 = price elasticity of demand
1.40 = income elasticity of demand
0.3 = advertising elasticity of demand
0.2 = cross price elasticity of demand
0.6 = population elasticity of demand
b. Yes, all of the signs are correct. Demand for gasoline appears to be slightly price inelastic.
c. Quantity demanded would appear to fall to zero. This highlights the importance of using
demand functions to make inferences within a relevant (and reasonable) range of values.
9. a. Y = 28.915 19.105X
(6.095) (Price input in decimal form)
(t statistic in parentheses)
b. R2 = .74 (Model explains 74% of variation in quantity demanded.)

Ho: = 0; Ha: 0; Reject Ho at 0.05 significance level if

t>t.025,13 = +2.160 or t<2.160. Since t = -6.095<-2.160, reject Ho (i.e., price is
statistically significant in explaining demand).
c. Y' = 28.915 19.105(.50) = 19.363 (thousand pens)
ED = 19.105(.50/19.363) = 0.49
10. a. Dependent variable: Price
Source
Model
Error
C Total

DF
4
10
14

Sum of Squares
10339.040
1239.798
11578.837

Mean Square
2584.760
123.980

F Value
20.848

Prob > F
0.0001

This edition is intended for use outside of the U.S. only, with content that may be different from the U.S. Edition. This may not be resold,
copied, or distributed without the prior consent of the publisher.

Chapter 4/Estimating of Demand

Root MSE
Dep Mean
C.V.
Variable
Intercep
Size
Rooms
Age
Garage

11.134621
100.587
11.06968
DF
1
1
1
1
1

R-Square
ADJ R-Sq

Parameter Est.
14.735136
3.921432
3.585118
0.118145
2.831695

Std. Error
27.483494
0.755262
4.470708
0.640698
9.716504

31
0.8929
0.8501

t-ratio
0.536
5.192
0.802
0.184
0.291

Prob > |t|

0.6036
.0004
0.4412
0.8574
0.7767

Y' = 14.7351 + 3.9214X1 + 3.5851X2 0.1181X3 2.8317X4

b. a = 14.7351
No significant economic meaning.
b1 = 3.9214
An increase of 100 sq. ft. in size increases the expected selling price by $3921, all other things
remaining constant.
b2 = 3.5851
An increase of 1 room increases the expected selling price by $3,585, all other things remaining
constant.
b3 = .1181
An increase of 1 year in age decreases the expected selling price by $118, all other things
remaining constant.
b4 = 2.8317
An attached garage decreases the expected selling price by $2,832, all other things remaining
constant.
c. From the computer output, only X1(size) is significant in explaining the selling price at the 5%
significance level or better.
d. R-Square = 0.8929. The regression model explains about 89% of the variation in selling price.
e. From the computer output, F is significant at the .0001 level.

This edition is intended for use outside of the U.S. only, with content that may be different from the U.S. Edition. This may not be resold,
copied, or distributed without the prior consent of the publisher.

32

Chapter 4/Estimating Demand

f. y' = 14.7351 + 3.9214(18) + 3.5851(7) .1181(15) 2.8317(1)
= 76.3426
se = 11.1346 (Root MSE from computer output)
y' 2 se = 76.3426 2(11.1346) = 54.0734 to 98.6118, or $54,073 to $98,612

11. a. Dependent variable (Y)Amount of Life Insurance

Independent variable XAnnual Income
b.

c. Y = 11.148 + 1.492X
The estimated slope coefficient (b = 1.492) indicates that the amount of life insurance held by
executives increases by 1.492 x $1000 = $1,492 for each $1000 increase in annual income.
d. se = 43.34
sb = 0.565
t = (1.492 - 0)/.565 = 2.641
Since the calculated t-value is greater than the t-value from the table (t.025,10 = 2.228 or +2.228),
one rejects the hypothesis at the .05 significance level that there is no relationship between the
amount of life insurance held and annual income.

This edition is intended for use outside of the U.S. only, with content that may be different from the U.S. Edition. This may not be resold,
copied, or distributed without the prior consent of the publisher.

Chapter 4/Estimating of Demand

33

e. R2 = .41
f. Source of
Variation
Regression
Residual
Total

Sum of
Squares
13,088
18,787
31,875

Degrees of
Freedom
1
10
11

Mean
Squares
13,088
1,879

F.05,1,10 = 4.96
F = MSR/MSE = 13,087/1879 = 6.966
Since the calculated F-value is greater than the F-value from the table, one rejects at the .05
significance level the hypothesis that there is no relationship between the amount of life insurance
held and annual income.
g. y' = 11.148 + 1.492(80) = 130.508 or $130,508
Approximate 95% prediction interval:
y' 2(43.344) = 130.508 86,688 = 43.82 to 217.196, or
$43,820 to $217,196.
12. a. ED = 2.15 (the exponent of P)
b. EA = 1.05 (the exponent of A)
c. The exponent of N (3.70) represents the elasticity of the quantity demanded with respect to the
proportion of the population under12 years of age. It indicates that the quantity demanded will
increase (decrease) by 3.70 percent for each one percent increase (decrease) in the proportion of
the population under 12 years of age.
13. a. Y=390.376 14.263X2
b. The estimated intercept value of 390.376 indicates that sales (Y) will be equal to 390.376
(X1000) = 390,376 gallons when the selling price (X2) is equal to zero. This value of X2 lies far
outside the range over which the regression line was estimated (recall that the lowest selling price
in the sample was $12.00 in sales region 9) and the sales estimate has no practical economic
significance. The estimated slope coefficient of 14.263 indicates that expected sales (Y) will
decrease by 14.263 (X1000) = 14,263 gallons for each additional $1.00 increase in the selling
price (X2).

This edition is intended for use outside of the U.S. only, with content that may be different from the U.S. Edition. This may not be resold,
copied, or distributed without the prior consent of the publisher.

34

Chapter 4/Estimating Demand

Ha: 2 0
b2= 14.263
sb2 = 2.909

c. H0: 2=0

t = (14.263 0)/2.909 = 4.903

Since the calculated t-value is less than the t-value from the table (t.025,8 = 2.306), one rejects the
null hypothesis at the .05 significance level that there is no relationship (i.e., 2=0) between paint

sales and the selling price.

d. R2 = .75 The regression equation, with selling price as the independent variable, explains 75
percent of the variation in paint sales in the sample.
e. Source of
Variation
Regression
Residual
Total

Sum of
Squares
6489.812
2160.188
8650.000

Degrees of
Freedom
1
8
9

Mean
Squares
6489.812
270.024

H0: 2=0

Ha: 2 0
F = MSR/MSE = 6489.812/270.024 = 24.034
Since the calculated F-value is greater than the F-value from the table (F.05,1,8 = 5.32), one rejects
the null hypothesis at the .05 significance level that there is no relationship between selling price
And paint sales.
f. Xp = $14.50 se = 16.432
Y' = 390.376 14.263(14.50) = 183.563
or 183,563 gallons
Y' - 2se = 183.563 2(16.432) = 150.699
Y' + 2se = 183.563 + 2(16.432) = 216.427
or from 150,699 gallons to 216,427 gallons.
g. ED = (dY/dX2)(X2/Y)
= 14.263 (14.50/183.563)
= 1.13

This edition is intended for use outside of the U.S. only, with content that may be different from the U.S. Edition. This may not be resold,
copied, or distributed without the prior consent of the publisher.

Chapter 4/Estimating of Demand

14. a. QD = 10,425 2910PX + .028A + 11,100 Pop

35

QD/P = 2910

QD = 10,425 2910(5) + .028(1,000,000) + 11,100(.5)

QD = 29,425

ED = 2910(5/29,425)
= 0.49

QD = 10,425 2910(10) + .028(1,000,000) + 11,100(.5)

QD = 14,875
ED = 2910(10/14,875)
= 1.96
b. QD/A = 0.028

QD = 10,425 2910(5) + .028(2,000,000) + 11,100(.5)

QD = 57,425
EA = .028(2,000,000/57,425) = 0.98

c. Yes, all t values are well in excess of t.025,21 = 2.074.

______________________________________________________________________

Solution to Case Exercise: Soft Drinks

1. The linear demand estimation is as follows:
R2 = 0.70
QD= 514.2 + 2.93 TEMP + 1.22 INCOME 242.9 PRICE
(4.12)
(O.80)
(5.58)
SSE = 38.26
and for the log-linear model,
log QD= 1.050 + 1.12 log TEMP + 0.22 log INCOME 3.12 log PRICE R2 = 0.67
(1.72)
(4.23)
(1.19)
( 4.92)
SSE = 0.11
where the numbers in parentheses are t-scores.
2. Both temperature and price are statistically significant with expected signs while income is
insignificant in its effect on soft drink demand. For the linear model, the price elasticity of
demand is (Q/P) (Mean P/Mean Q)
242.97 x ($2.2025/158.2083) = 3.38
and for the log-linear model 3.12.

This edition is intended for use outside of the U.S. only, with content that may be different from the U.S. Edition. This may not be resold,
copied, or distributed without the prior consent of the publisher.

36

Chapter 4/Estimating Demand

This point elasticity at the mean price and quantity across the states is in the elastic range, as
expected. Note, however, that these are market-level price elasticities, so no firm behavior is
directly implied by this estimate. Nevertheless, an elastic demand at the market level does imply
elastic firm-level demand at comparable prices, comparable price sensitivity, and the smaller
quantities facing each firm.

3. Omitting price from the regression, one obtains for the log-linear model
log QD= 0.16 + 1.72 log TEMP 0.152 log INCOME
(5.96)
( 0.73)

R2 = 0.49
SSE = 0.137

4. Omitting both price and temperature yields a linear model as follows:

QD = 254.6 5.37 INCOME
R2 = 0.11
( 2.11)
SSE = 64.2
For the log-linear model, one obtains
QD = 4.47 0.552 INCOME
( 2.13)

R2 = 0.09
SSE = 0.18

No, a marketing plan should not be designed specifically to introduce canned soft drink machines
into low-income neighborhoods. And students should not offer the negative and significant
income parameter estimate above as their reason. The above regression does NOT call for
relocating canned soft drink machines away from low-income neighborhoods. The regression
coefficient on income has been biased downward by the omission of price and temperature
enough to make an insignificant factor appear negative and significant in its effect on demand.
This illustrates the critical importance of using analytical reasoning and demand theory to
correctly specify a regression model.

This edition is intended for use outside of the U.S. only, with content that may be different from the U.S. Edition. This may not be resold,
copied, or distributed without the prior consent of the publisher.