Student Solutions Manual To Accompany An Introduction To Econometrics A Self Contained Approach 1nbsped 9780262317184 9780262525404 - Compress

Student Solutions Manual to Accompany
An Introduction to Econometrics: A Self-Contained Approach

Student Solutions Manual to Accompany
An Introduction to Econometrics: A Self-Contained Approach
Frank Westhoff
The MIT Press

Cambridge, Massachusetts
London, England
© 2013 Massachusetts Institute of Technology
All rights reserved. No part of this book may be reproduced in any form by any electronic or
mechanical means (including photocopying, recording, or information storage and retrieval)
without permission in writing from the publisher.
This book was printed and bound in the United States of America.
ISBN: 978-0-262-52540-4
Contents
Chapter 1: Descriptive Statistics 1

Chapter 2: Essentials of Probability and Estimation Procedures 7
Chapter 3: Interval Estimates and the Central Limit Theorem 13
Chapter 4: Estimation Procedures, Estimates, and Hypothesis Testing 21
Chapter 5: Ordinary Least Squares Estimation Procedure—The Mechanics 25
Chapter 6: Ordinary Least Squares Estimation Procedure—The Properties 31
Chapter 7: Estimating the Variance of an Estimate’s Probability Distribution 37
Chapter 8: Interval Estimates and Hypothesis Testing 41
Chapter 9: One-Tailed Tests, Two-Tailed Tests, and Logarithms 49
Chapter 10: Multiple Regression Analysis—Introduction 59
Chapter 11: Hypothesis Testing and the Wald Test 63
Chapter 12: Model Specification and Development 67
Chapter 13: Dummy and Interaction Variables 73
Chapter 14: Omitted Explanatory Variables, Multicollinearity, and Irrelevant
Explanatory Variables 77
Chapter 15: Other Regression Statistics and Pitfalls 83
Chapter 16: Heteroskedasticity 89
Chapter 17: Autocorrelation (Serial Correlation) 95
Chapter 18: Explanatory Variable/Error Term Independence Premise,
Consistency, and Instrumental Variables 99
Chapter 19: Measurement Error and the Instrumental Variables Estimation
Procedure 101
Chapter 20: Omitted Variables and the Instrumental Variable Estimation Procedure 105
Chapter 21: Panel Data and Omitted Variables 109
Chapter 22: Simultaneous Equations Models—Introduction 113
Chapter 23: Simultaneous Equations Models—Identification 117
Chapter 24: Binary and Truncated Dependent Variables 123
Chapter 25: Descriptive Statistics, Probability, and Random Variables—
A Closer Look 129
Chapter 26: Estimating the Mean of a Population 133
Chapter 1: Descriptive Statistics
Solutions to Chapter 1 Prep Questions
1. One reasonable way would be to calculate the average precipitation for each month and then
choose the month with the highest average.
Solutions to Chapter 1 Review Questions
1. a. The mean is the average of the values. It describes the center of the distribution.
T
¦x
t 1
t
b. Mean[ x ]
T
3. A histogram illustrates how the values of a single data variable are distributed.
5. a. Yes.
b. No.
7. a. Cov[x, y] > 0 0 < CorrCoef[x, y] d 1

b. Cov[x, y] < 0 1 d CorrCoef[x, y] < 0
c. Cov[x, y] | 0 CorrCoef[x, y] | 0
Solutions to Chapter 1 Exercises
1. a.
Inches of
Precipitation 1964 1975
0–1 2 0
1–2 3 0
2–3 3 2
3–4 3 3
4–5 0 3
5–6 1 1
6–7 0 1
7–8 0 0
8–9 0 1
9–10 0 0
10–11 0 1
1
b.
3. a. 1975 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Precip 4.39 3.04 3.97 2.87 2.10 4.68 10.56 6.13 8.63 4.90 5.08 3.90
TwoPlusPrecip 6.39 5.04 5.97 4.87 4.10 6.68 12.56 8.13 10.63 6.90 7.08 5.90
b.
2
i. The histograms have identical shape. The TwoPlusPrecip histogram is just “slided” 2
inches to the right.
ii. The center has risen (shifted to the right) by 2 inches.
iii. The spread is unchanged.
c. i. Mean[TwoPlusPrecip] = Mean[2 + Precip]
= 2 + Mean[Precip]
= 2 + 5.02
= 7.02
ii. Var[TwoPlusPrecip] = Var[2 + Precip]
= Var[Precip]
= 5.41
d. i. 7.02
ii. 64.91
iii. 5.41
e. Yes, the answers are consistent. The histogram reveals that the
x center of the TwoPlusPrecip distribution is farther to the right; this is
consistent with its higher mean.
x spread of the two distributions is the same; this is consistent with their
equal variances.
The appropriate equation and statistical software provide the same answers.
3
5. a.
b. Yes, SatMath and SatVerbal appear to be correlated. As SatVerbal increases, SatMath

tends to increase also.
c. SatMath SatVerbal
Mean 737.0 719.0
Variance 2,774.3 3,402.3
Covariance 2,103.7
Correlation Coefficient .6847
d. i. Mean[SatSum]= Mean[SatMath] + Mean[SatVerbal]
= 737.0 + 719.0
= 1,456.0
ii. Var[SatSum] = Var[SatMath] + 2Cov[SatMath, SatVerbal] + Var[SatVerbal]
= 2,774.3 + 2u2,103.7 + 3,402.3
= 10,384.0
e. i. 1,456.0
ii. 311,520
iii. 10,384.0
f. Yes
x1 x2 ... x30
7. a. Mean[SatMath] =
30
x1 x2 ... x10 x11 x12 ... x30
=
30
x1 x2 ... x10 x x ... x30
= + 11 12
30 30
4
10 x1 x2 ... x10 20 x11 x12 ... x30
= u + u
30 10 30 20
1 2
= u Mean[SatMathFemale] + u Mean[SatMathMale]
3 3
= WgtFemale Mean[SatMathFemale] + WgtMaleMean[SatMathMale]
1 2
b. WgtFemale + WgtMale = + =1
3 3
c. Mean[SatMath] = 737
Mean[SatMathFemale] = 725
Mean[SatMathMale] = 743
Yes, the results are consistent.
Mean[SatMath] = WgtFemale Mean[SatMathFemale] + WgtMaleMean[SatMathMale]
10 20
30 30
1 2
3 3
1 2
= u 725 + u 743
3 3
725 1, 486 2, 211
= = = 737
3 3
5
Chapter 2: Essentials of Probability and Estimation
Procedures
1. a. There are 13 chances out of 52 that the card drawn is a heart; that is, the probability that
13 1
the card drawn will be a heart equals .
52 4
b. There are 4 chances out of 52 that the card drawn is an ace; that is, the probability that the
4 1
card drawn will be an ace equals .
52 13
c. There are 26 chances out of 52 that the card drawn is a red card; that is, the probability
26 1
that the card drawn will be a red card equals .
52 2
3. a. Mean[cx] = c Mean[x]
b. Mean[x + y] = Mean[x] + Mean[y]
c. Var[cx] = Var[cx] = c2Var[x]
d. Var[x + y] = Var[x + y] = Var[x] + 2Cov[x, y] + Var[y]
e. When x and y are independent: Var[x + y] = Var[x + y] = Var[x] + Var[y]
1. A random process is a process whose outcome is uncertain; that is, we cannot determine the
outcome with certainty until the process is completed.
3. A random variable is a variable that is associated with a random process. The value of a
random variable cannot be determined with certainty before the experiment is conducted.
5. a. Beforehand, we cannot determine the numerical value of the random variable with
certainty.
b. Beforehand, we can often describe the probability distribution of the random variable.
7. a. Center of the random variable’s probability distribution.

b. Spread of the random variable’s probability distribution.
7
1. Suppose that you have a deck composed of the following 10 cards:

2k 2j 2i 2h 3k
3j 3i 4k 4j 5i
Consider the following experiment:

x Thoroughly shuffle the deck of 10 cards.
x Draw one card.
x Replace the card drawn.
a. A random variable describes the outcome of a random process. The value of a random
variable cannot be determined beforehand, before the random process is completed. v is a
random variable because we cannot determine if it will equal 2, 3, 4, or 5 before the
experiment is conducted.
b. Since there are a total of 10 cards in the deck, the probability of drawing any one card is
1
:
10
Card Probability Card Probability
1 1
k
2k 3j
10 10
1 1
2j 3i
10 10
1 1
2i 4k
10 10
1 1
2h 4j
10 10
1 1
3k 5i
10 10
Now it is easy to calculate the probability distribution of v:
v Prob[v]
1 1 1 1 4
2 2k 2j 2i 2h + + + = = .4
10 10 10 10 10
1 1 1 3
3 3k 3j 3i + + = = .3
10 10 10 10
1 1 2
4 4k 4j + = = .2
10 10 10
1
5 5i = .1
10
8
v equals 2 whenever the 2k, 2j, 2i, or 2h is drawn. Therefore, the probability that v
1 1 1 1 1
equals 2 is + + + = = .4. Using analogous logic, the probability that v
10 10 10 10 10
equals 3 is .3, 4 is .2, and 5 is .1.
c. Mean[v]= ¦ vProb[v]
All v
= 2u.4 + 3u.3 + 4u.2 + 5u.1

= .8 + .9 + .8 + .5 = 3
Var[v] = ¦ (v Mean[v])2 Prob[v]
All v
v Mean[v] v Mean[v] (v Mean[v])2 Prob(v)

2 3 2 3 = 1 1 .4
3 3 33=0 0 .3
4 3 43=1 1 .2
5 3 53=2 4 .1
Var[v] = 1u.4 + 0u.3 + 1u.2 + 4u.1

= .4 + 0 + .2 + .4 = 1
d. v Portion of the repetitions
2 | 40% or .4
3 | 30% or .3
4 | 20% or .2
5 | 10% or .1
The simulation confirms the relative frequency interpretation of probability. After many,
many repetitions, the distribution of the numerical values mirrors the probability
distribution.
e. Mean of the Numerical Values of v = 3
Variance of the Numerical Values of v = 1
The mean of the probability distribution equals the mean of the numerical values. The
variance of the probability distribution equals the variance of the numerical values.
3. a. i. If a 2 was drawn on the first draw, only three 2’s remain in the deck of 9 cards.
Therefore, there are 3 chances in 9 of drawing a 2 on the second draw:
3 1
Prob[v2 = 2 IF v1 = 2] = = .3333…
9 3
ii. If a 2 was not drawn on the first draw, four 2’s remain in the deck of 9 cards.
Therefore, there are 4 chances in 9 of drawing a 2 on the second draw:
4
Prob[v2 = 2 IF v1 z 2] = = .4444…
9
b. No. The random variables v1 and v2 are not independent. Knowing whether v1 equals 2 or
not helps us predict the value of v2 because the value of v1 affects the probability of v2.
c. i. If a 2 was drawn on the first draw, 3,999 2’s remain in the deck of 9,999 cards.
Therefore, there are 3,999 chances in 9,999 of drawing a 2 on the second draw:
9
3,999
Prob[v2 = 2 IF v1 = 2] = = .39993999…
9,999
ii. If a 2 was not drawn on the first draw, 4,000 2’s remain in the deck of 9,999 cards.
Therefore, there are 4,000 chances in 9,999 of drawing a 2 on the second draw:
4, 000
Prob[v2 = 2 IF v1 z 2] = = .40004000…
9,999
d. Size of Population:
Number of Cards in Original Deck
10 10,000
Prob[v2 = 2 IF v1 z 2] .4444… .40004000…
Prob[v2 = 2 IF v1 = 2] .3333… .39993999…
Difference .1111… .00010001…
As the size of the population increases, the probability of drawing a 2 on the 2nd draw is
affected less by the card drawn on the 1st draw. When there are 10,000 cards in the
original deck, the probability of drawing a 2 on the 2nd draw is .400 to the nearest
thousandths regardless of what card was drawn on the 1st draw.
e. As the size of the population increases, the independence assumption becomes a better
approximation of reality.
f. A professional pollster need not be concerned.
5. a. Since Archie’s arrows always land within 20 centimeters of the center, the probability
that v is less than or equal to 20 must equal 1.
1
b. Prob[Points t 6] = u.10u20
2
= 1.0
1
c. Prob[Points = 10] = u (.10 .08) u 4 + .08 u 4
2
= .04 + .32
= .36
10
1
d. Prob[Points = 9] = u (.08.06) u 4 + .06 u 4
2
= .04 + .24
= .28
e. Prob[Points = 7 or Points = 9]
1
= u (.06.02) u 4 + .02 u 8
2
= .08 + .16
= .24
30 1
7. a. i. =
60 2
5 1
ii. =
60 12
25 5
iii. =
60 12
1 1 7
b. i. + =
2 12 12
5
ii.
12
5 5 5 5 5 3,125
c. u u u u = = .0126
12 12 12 12 12 248,832
11
Chapter 3: Interval Estimates and the Central Limit Theorem
1. a. A random variable describes the outcome of a random process (a draw, a coin toss, etc.).
The value of a random variable cannot be determined beforehand, before the random
process is completed.
b. i. The value of the random variable beforehand.
ii. The probability distribution of the random variable.
3. Var[T1 v1 v2 vT ]
Since Var[cx] = c2Var[x]
1
=
T2
[
Var v1 v2 vT ]
Since Var[x + y] = Var[x] + Var[y] when x and y
are independent; hence, the covariances are all 0.
1
= (Var[v1 ] Var[v2 ] Var[vT ] )
T2
Since Var[v1] = Var[v2] = … = Var[vT] = p(1 p)
1
= ( p(1 p) p(1 p) p(1
(1 p) )
T2
How many p(1 p) terms are there? A total of T.
1
= (T u p(1 p))
T2
Simplifying.
p(1 p)
T
5.
x The mean of the probability distribution describes the distribution’s center.
x The variance of the probability distribution describes the distribution’s spread.
1. a. When the mean of the estimate’s probability distribution equals the actual value, the
estimation procedure is unbiased, the estimation procedure does not systematically
overestimate or underestimate the actual value.
b. When an estimation procedure is unbiased, the variance of the estimate’s probability
distribution indicates how reliable an estimate is. On the one hand, if the variance is
small, there is a high probability that the estimate is close to the actual value. On the
other hand, if the variance is large, the probability that the estimate lies close to the actual
value is small.
13
3. As the sample size increases, the probability distribution of the random variable approaches
the normal distribution.
1. a. Mean[SatSum] = Mean[SatMath + SatVerbal]

= Mean[SatMath] + Mean[SatVerbal]
= 515 + 460
= 975
Mean[SatDiff] = Mean[SatMath SatVerbal]

= Mean[SatMath + (1)SatVerbal)]
= Mean[SatMath] Mean[SatVerbal]
= 515 460
= 55
b. CorrCoef[SatMath, SatVerbal] = .0 Since SatMath and SatVerbal are independent.
Cov[ SatMath, SatVerbal ]
CorrCoef[SatMath, SatVerbal] =
Var[ SatMath] Var[ SatVerbal ]
.0 =
100 u100
Cov[SatMath SatVerbal] = .0
Var[SatSum] = Var[SatMath + SatVerbal]
= Var[SatMath] + 2Cov[SatMath, SatVerbal] + Var[SatVerbal]
= 10,000 + 0 + 10,000
= 20,000
Var[SatDiff] = Var[SatMath SatVerbal]
= Var[SatMath + (1)SatVerbal)]
= Var[SatMath] + 2Cov[SatMath, (1)SatVerbal] + Var[(1)SatVerbal]
= Var[SatMath] 2Cov[SatMath, SatVerbal] + Var[(1)SatVerbal]
= 10,000 + 0 + 10,000
= 20,000
c. CorrCoef[SatMath, SatVerbal] = 1.0 Perfect positive correlation.
1.0 =
100 u100
Cov[SatMath SatVerbal] = 10,000
= 10,000 + 2u10,000 + 10,000
= 10,000 + 20,000 + 10,000
= 40,000
14
Var[SatDiff]= Var[SatMath SatVerbal]
= 10,000 2u10,000 + 10,000
= 10,000 20,000 + 10,000
= 0
d. CorrCoef[SatMath, SatVerbal] = .5
.5 =
100 u100
= 10,000 + 2u5,000 + 10,000
= 10,000 + 10,000 + 10,000
= 30,000
= 10,000 2u5,000 + 10,000
= 10,000 10,000 + 10,000
= 10,000
e. CorrCoef[SatMath, SatVerbal] = .5
.5 =
100 u100
= 10,000 + 2u5,000 + 10,000
= 10,000 10,000 + 10,000
= 10,000
= 10,000 2u5,000 + 10,000
= 10,000 + 10,000 + 10,000
= 30,000
f. 0.0_____ 1.0_____ between 0.0 and 1.0 X less than 0.0 _____
Typically, students who do well on one part of the SAT do well on the other.
15
3. a. No
b. Yes
c. StateDiff = SatMichigan SatAlaska
Mean[StateDiff] = Mean[SatMichigan SatAlaska]
= Mean[SatMichigan + 1uSatAlaska)]
= Mean[SatMichigan] + Mean[1uSatAlaska]
= Mean[SatMichigan] Mean[SatAlaska]
= 549 489
= 60
SD[SatMichigan] = SD[SatAlaska] = 100
Var[SatMichigan] = Var[SatAlaska] = 10,000
Cov[SatMichigan, SatAlaska] = 0 SatMichigan and SatAlaska are independent
Var[StateDiff] = Var[SatMichigan SatAlaska]
= Var[SatMichigan + 1uSatAlaska]
= Var[SatMichigan] + 2Cov[SatMichigan, SatAlaska] + Var[SatAlaska]
= 10,000 + 0 + 10,000
= 20,000
SD[ StateDiff ] Var[ StateDiff ] 20,000 141.4
0 60 45
z .42
141.4 141.4
Prob[SatMichigan > SatAlaska] = Prob[SatMichigan SatAlaska > 0]
= Prob[StateDiff > 0]
= .66
z 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
0.2 0.4207 0.4168 0.4129 0.4090 0.4052 0.4013 0.3974 0.3936 0.3897 0.3859
0.3 0.3821 0.3783 0.3745 0.3707 0.3669 0.3632 0.3594 0.3557 0.3520 0.3483
0.4 0.3446 0.3409 0.3372 0.3336 0.3300 0.3264 0.3228 0.3192 0.3156 0.3121
0.5 0.3085 0.3050 0.3015 0.2981 0.2946 0.2912 0.2877 0.2843 0.2810 0.2776
5. a. Mean[MPG] = 32
SD[MPG] = 4
16
30 32 2
z .5
4 4
Prob[MPH > 110] = .69 = 69 percent
z 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
0.4 0.3446 0.3409 0.3372 0.3336 0.3300 0.3264 0.3228 0.3192 0.3156 0.3121
0.5 0.3085 0.3050 0.3015 0.2981 0.2946 0.2912 0.2877 0.2843 0.2810 0.2776
0.6 0.2743 0.2709 0.2676 0.2643 0.2611 0.2578 0.2546 0.2514 0.2483 0.2451
b.
z 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
1.1 0.1357 0.1335 0.1314 0.1292 0.1271 0.1251 0.1230 0.1210 0.1190 0.1170
1.2 0.1151 0.1131 0.1112 0.1093 0.1075 0.1056 0.1038 0.1020 0.1003 0.0985
1.3 0.0968 0.0951 0.0934 0.0918 0.0901 0.0885 0.0869 0.0853 0.0838 0.0823
To achieve that objective, z must equal 1.29.

30 x
z 1.29
4
30 x = 4u1.29
30 x = 5.16
30 + 5.16 = x
x = 35.16
The mean must be increased by 35.16 32 =
3.16.
30 32
c. z 1.29
x
2
1.29
x
2
x 1.55
1.29
The standard deviation must be decreased by 4
1.55 = 2.45.
17
d. Approach #1 Cost = 3.16u100,000 =
$316,000
Approach #2 Cost = 2.45u200,000 =
$490,000
Approach #1 is less costly.
7. a. i. Recall that when betting on 1st set of twelve numbers, the mean of vi’s
probability distribution equals .027 and the variance equals 1.972.
Mean[Your TNW] = Mean[v1 + v2 + … + v50]

= Mean[v1] + Mean[v2] + … + Mean[v50]
= .027 .027 … .027
= 1.35
Var[Your TNW] = Var[v1 + v2 + … + v50]
Since the v’s are independent, all the
covariances equal 0.
= Var[v1] + Var[v2] + … + Var[v50]
= 1.972 + 1.972 + … + 1.972
= 98.60
ii. Mean[TNW] = 1.35

SD[TNW] = Var[TNW ] =
98.60 9.93
10 (1.35) 11.35
z 1.14
9.93 9.93
Prob[TNW t 10] = .1271
z 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
1.0 0.1587 0.1562 0.1539 0.1515 0.1492 0.1469 0.1446 0.1423 0.1401 0.1379
1.1 0.1357 0.1335 0.1314 0.1292 0.1271 0.1251 0.1230 0.1210 0.1190 0.1170
1.2 0.1151 0.1131 0.1112 0.1093 0.1075 0.1056 0.1038 0.1020 0.1003 0.0985
iii. Mean[TNW] = 1.35
98.60 9.93
10 (1.35) 8.65
z .87
9.93 9.93
Prob[TNW d 10] = .1922
18
z 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
0.7 0.2420 0.2389 0.2358 0.2327 0.2296 0.2266 0.2236 0.2206 0.2177 0.2148
0.8 0.2119 0.2090 0.2061 0.2033 0.2005 0.1977 0.1949 0.1922 0.1894 0.1867
0.9 0.1841 0.1814 0.1788 0.1762 0.1736 0.1711 0.1685 0.1660 0.1635 0.1611
b. i. Recall that when betting on red, the mean of vi’s probability distribution equals .027
and the variance equals .999.
Mean[Friend’s TNW] = Mean[v1 + v2 + … + v50]
= Mean[v1] + Mean[v2] + … + Mean[v50]
= .027 .027 … .027
= 1.35
Var[Friend’s TNW] = Var[v1 + v2 + … + v50]
Since the v’s are independent, all the
covariances equal 0.
= Var[v1] + Var[v2] + … + Var[v50]
= .999 + .999 + … + .999
= 49.50
ii. Mean[TNW] = 1.35
49.50 7.04
10 (1.35) 11.35
z 1.61
7.04 7.04
Prob[TNW t 10] = .0537
z 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
1.5 0.0668 0.0655 0.0643 0.0630 0.0618 0.0606 0.0594 0.0582 0.0571 0.0559
1.6 0.0548 0.0537 0.0526 0.0516 0.0505 0.0495 0.0485 0.0475 0.0465 0.0455
1.7 0.0446 0.0436 0.0427 0.0418 0.0409 0.0401 0.0392 0.0384 0.0375 0.0367
iii. Mean[TNW] = 1.35
97.23 9.86
10 (1.35) 8.65
z 1.23
7.04 7.04
Prob[TNW d 10] = .1093
z 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
1.1 0.1357 0.1335 0.1314 0.1292 0.1271 0.1251 0.1230 0.1210 0.1190 0.1170
1.2 0.1151 0.1131 0.1112 0.1093 0.1075 0.1056 0.1038 0.1020 0.1003 0.0985
1.3 0.0968 0.0951 0.0934 0.0918 0.0901 0.0885 0.0869 0.0853 0.0838 0.0823
c. Your friend appears to be more risk averse than you.
19
Chapter 4: Estimation Procedures, Estimates, and Hypothesis
Testing
other hand, if the variance is large, the probability that the estimate lies close the actual
value is small.
3. a.
Jury Finds Jury Finds
Defendant Defendant
Guilty Innocent
Defendant Actually Jury is Jury is

Innocent correct__ incorrect X correct X incorrect__
Defendant Actually Jury is Jury is

Guilty correct X incorrect__ correct__ incorrect X
For each scenario, indicate whether the jury would be correct or incorrect.
b. Defendant Actually Innocent and Jury Finds Defendant Guilty

Costs: Imprisoning an innocent young man.
Defendant Actually Guilty and Jury Finds Defendant Innocent
Costs: Allowing a guilty man to go free who then can continue to commit
crimes.
1. a.
x The mean of the estimate’s probability distribution.
x The variance of the estimate’s probability distribution.
b.
x The mean of the estimate’s probability distribution determines whether or not the
estimation procedure is biased or unbiased. If the mean equals the actual value, the
estimation procedure does not systematically underestimates or overestimates the
actual value.
21
x When the estimation procedure is unbiased, the variance of the estimate’s probability
distribution indicates the reliability of the estimation procedure. As the variance
becomes smaller, the estimate becomes more reliable.
3. a. Type I error occurs when we reject the null hypothesis when it is actually true.
b. Type II error occurs when we do not reject the null hypothesis when it is actually false.
c. As the significance level becomes smaller, it becomes more difficult to reject the null
hypothesis thereby decreasing the likelihood of Type I error and increasing the likelihood
of Type II error.
1. a. This statement is false. Even when an estimation procedure is unbiased we can never
expect the estimate to equal the actual value.
b. This statement is true whenever the estimate’s probability distribution is symmetric.
c. This is always true; it is the definition of an unbiased estimation procedure.
d. This statement is always true. The relative frequency interpretation of probability tells us
that after many, many repetitions the distribution of the actual values mirrors the
probability distribution. Accordingly, after many, many repetitions the average of the
estimates would equal the mean of the estimate’s probability distribution; the estimation
procedure would be unbiased.
1,885,178 1,885,178
3. a. DemVoterFrac .45144 | .451
1,885,178 2,290,723 4,175,901
17 17
b. .53125 | .531
17 15 32
c. Yes.
d. Cynic’s View: Despite the election results, there is no unfair districting; there is no
gerrymandering.
“Sure, the election results suggest that there may be gerrymandering, but the results
were just a coincidence. The results were just a random occurrence, a consequence of
the luck of the draw.”
e. H0: No gerrymandering Cynic is correct.
H1: Gerrymandering present Cynic is incorrect.
Prob[Results IF H0 True] = Probability that DemCongressFrac would be .531 or
more when DemVoterFrac, the fraction of voters
statewide voted for the Democratic candidate, equals
.451
f. Mean[DemCongressFrac]: .451
Var[DemDemocratFrac]: .007737
SD[DemCongressFrac]: .088
Mean[DemCongressFrac] = p = .451 where p = fraction of population statewide

voting Democratic
22
p(1 p) .451u (1 .451) .451u .549
Var[ DemCongressFrac] .007737
T 32 32
SD[ DemCongressFrac] Var[ DemCongressFrac] .007737 .08796 | .088
g. Prob[Results IF H0 True] = Probability that DemCongressFrac would be

.531 or more when DemVoterFrac, the
fraction of voters statewide voted for the
Democratic candidate, equals .451
.531 .451 .080
z .91
.088 .088
z 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
0.8 0.2119 0.2090 0.2061 0.2033 0.2005 0.1977 0.1949 0.1922 0.1894 0.1867
0.9 0.1841 0.1814 0.1788 0.1762 0.1736 0.1711 0.1685 0.1660 0.1635 0.1611
1.0 0.1587 0.1562 0.1539 0.1515 0.1492 0.1469 0.1446 0.1423 0.1401 0.1379
Prob[Results IF H0 True] = .1814
There is about one chance in five

that the election results were not a
consequence of gerrymandering.
5. a.
69, 498, 215 69, 498, 215
DemVoterFrac .5369
69, 498, 215 59,948, 240 129, 446, 455
365 365
b. .6784
365 173 538
c. Yes. Democratic.
d. Cynic’s View: Despite the election results, the Electoral College is not unfair, it does not
favor Democrats.
“Sure, the election results suggest that the Electoral College favors
Democrats, but the results were just a coincidence. The results were just a
random occurrence, a consequence of the luck of the draw.”
e. H0: No Electoral College Unfairness Cynic is correct.
H1: Electoral College Does Not Favor Democrats Cynic is incorrect.
Prob[Results IF H0 True] = Probability that DemElectColFrac would be .6784 or
nationwide voted for the Democratic candidate, equals
.5369
23
f. Mean[DemElectColFrac]: .5369
Var[DemElectColFrac]: .0004622
SD[DemElectColFrac]: .0215
Mean[DemElectColFrac] = p = .5369 where p = fraction of Democratic voters
nationwide
p(1 p) .5369 u (1 .5369) .5369 u .4631
Var[ DemElectColFrac] .0004622
T 538 538
SD[ DemElectColFrac] Var[ DemElectColFrac] .0004622 .0215
g. Prob[Results IF H0 True] = Probability that DemElectColFrac would be .6784 or
nationwide voted for the Democratic candidate, equals
.5369
.6784 .5369 .1415
z 6.55
.0215 .0215
Prob[Results IF H0 True] < .001
7. a. Yes. On average, you would expect her to win a little less than one third of the time. In
fact, she won more than two-thirds of the time.
b. Cynic’s View: Despite the results, everything was on the “up and up.”
“Sure, the results suggest that something was not on the “up and up,” but the results
were just a coincidence. The results were just a random occurrence, a consequence of
the luck of the draw.”
c. H0: Everything Was on the “Up and Up” Cynic is correct.
H1: Roulette Wheel Was Rigged to Favor the Girlfriend Cynic is incorrect.
Prob[Results IF H0 True] = Probability that the casino manager’s girlfriend would
win $20 or more if everything was on the “up and up.”
= Prob[TNW t 20 IF everything were on the “up and up”]
d. Mean[TNW] = 1.35
SD[TNW] = Var[TNW ] = 98.60 9.93
20 (1.35) 21.35
z 2.15
9.93 9.93
Prob[TNW t 20 IF everything on the “up and

up”] = .0158
z 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
2.0 0.0228 0.0222 0.0217 0.0212 0.0207 0.0202 0.0197 0.0192 0.0188 0.0183
2.1 0.0179 0.0174 0.0170 0.0166 0.0162 0.0158 0.0154 0.0150 0.0146 0.0143
2.2 0.0139 0.0136 0.0132 0.0129 0.0125 0.0122 0.0119 0.0116 0.0113 0.0110
24
Chapter 5: Ordinary Least Squares Estimation Procedure—
The Mechanics
1. a.
b. Yes. For the most part, when income increases, savings increase also.
c.
25
By choosing two points on this line, we can solve for the equation of the best fitting line.
It looks like the points (200, 15) and (1200, 155) are more or less on the line. We can use
these two points to estimate the slope:
155 15 140
.14
1200 200 1000
Now, a little algebra:
y 15
.14
x 200
y 15 = .14x 28
y = .14x 13
This equation suggests that if Americans earn an additional $1,000 of income, savings
will rise by $140. This is consistent with our theory.
d. $140 million since the slope is .14: .14u$1,000,000,000 = $140,000,000
3. a. Before the experiment is conducted, we cannot determine the numerical value of the
random variable with certainty.
b. Before the experiment is conducted, we can often calculate the random variable’s
probability distribution telling us how likely it is for the random variable to equal each of
its possible numerical values.
5. Begin with the equation for SSR:

SSR ( y1 bConst bx x1 )2 ( y2 bConst bx x2 )2 ( y3 bConst bx x3 )2
Differentiate SSR with respect to bConst and set the derivative equal to 0:
dSSR
2( y1 bConst bx x1 ) 2( y2 bConst bx x2 ) 2( y3 bConst bx x3 ) 0
dbConst
Dividing by 2.
( y1 bConst bx x1 ) ( y2 bConst bx x2 ) ( y3 bConst bx x3 ) 0
Collecting like terms.

( y1 y2 y3 ) (bConst bConst bConst ) (bx x1 bx x2 bx x3 ) 0
Simplifying.
( y1 y2 y3 ) 3bConst bx ( x1 x2 x3 ) 0
Dividing by 3.
y1 y2 y3 x x x
bConst bx 1 2 3 0
3 x
y1 y2 y3 x x2 x3
Since equals the mean of y, y , and 1 equals the mean of x, x :
3 3
y bConst bx x 0
26
Therefore,
bConst y bx x .
1. The ordinary least squares (OLS) estimation procedure minimizes the sum of squared
residuals to determine the best fitting line.
3. The parameter estimates are random variables.
1. a.
b. First, calculate the means:
5 10 30 45
x 15
3 3
14 44 80 138
y 46
3 3
Second, for each student calculate the deviation of x from its mean and the deviation of y
from its mean:
Student yt y yt y xt x xt x
1 14 46 32 5 15 10
2 44 46 2 10 15 5
3 80 46 34 30 15 15
Third, calculate the products of the y and x deviations and squared x deviations for each
student; then calculate the sums:
27
Student ( yt y )( xt x ) ( xt x )2
2
1 32u10 = 320 10 = 100
2
2 2u5 = 10 5 = 25
2
3 34u15 = 510 15 = 225
Sum = 840 Sum = 350
Now, apply the equations for bx and bConst:
T
¦ ( y y )( x x )
t 1
t t
840
bx T
2.4
350
¦ (x x )
t 1
t
2
bConst y bx x 46 2.4 u15 46 36 10

c. Calculate the sum of squared residuals by filling in the following blanks:
Student xt yt Estyt = bConst + bxxt Rest = yt Estyt Rest2
1 5 14 14 22 = 8
10 + 2.4u5 = 22 64
2 10 44 44 34 = 10
10 + 2.4u10 = 34 100
3 30 80 80 82 = 2
10 + 2.4u30 = 82 4
SSR = 168
3. a. Economic theory teaches that the supply curve is upward sloping. Hence, an increase in
the price should increase oil production.
b.
Ordinary Least Squares (OLS)
Dependent Variable: OilProdBarrels
Explanatory Variable(s): Estimate SE t-Statistic Prob
Price 92.50159 28.12319 3.289157 0.0028
Const 5944.958 478.7771 12.41696 0.0000
Number of Observations 29
Estimated Equation: EstOilProd = 5,945 + 92.5Price

c. Yes. The coefficient estimate for the price is positive suggesting that an increase in price
increases oil production.
d. i. Increase oil production by 92.5 thousand barrels.
ii. Increase oil production by 185.0 thousand barrels.
iii. Decrease oil production by 462.5 thousand barrels.
5. a. Economic theory teaches that the demand curve is downward sloping. Hence, an increase
in the price should decrease gasoline consumption.
b.
Dependent Variable: GasCons
PriceDollars 151.6556 47.57295 3.187853 0.0128
Const 516.7801 60.60223 8.527410 0.0000
28
Sum Squared Residuals 54.00000
SE of Regression 7.348469
Estimated Equation: EstGasCons = 517 151.7PriceDollars

c. Yes. The coefficient estimate for the price is negative suggesting that an increase in price
decreases gasoline consumption.
d. i. Decrease gasoline consumption by 151.7 million gallons.
ii. Decrease gasoline consumption by 303.3 million gallons.
iii. Decrease gasoline consumption by 75.8 million gallons.
iv. Increase gasoline consumption by 151.7 million gallons.
v. Increase gasoline consumption by 303.3 million gallons.
7. a. Earmarks are projects initiated by members of Congress that circumvent merit-based

criteria.
b.
Dependent Variable: Number
PartyDem1 2.361883 0.577102 4.092661 0.0001
Const 5.613527 0.424482 13.22441 0.0000
Estimated Equation: EstNumber = 5.6 + 2.3PartyDem1
c. Yes. The coefficient estimate for the Democratic party explanatory variable is positive
suggesting that Democratic members of Congress receive more earmarks than their
Republican colleagues.
29
Chapter 6: Ordinary Least Squares Estimation Procedure—
The Properties
1. NB: Because random influences are actually random, the numerical values that you reported
will no doubt differ from those below. Therefore, your means and variances which are based
on your numerical values will also differ.
Numerical
Value of
Coefficient Calculations Simulation’s Calculations
Repetition Estimate Your Mean Variance Mean Variance
1 3.40
2 4.70 4.05 .4225 4.1 .4
3 1.44 3.18 1.7955 3.2 1.8
Repetition 2 Calculations:
3.40 4.70 8.10
Mean 4.05
2 2
Deviation Squared
Value Mean From Mean Deviation
3.40 4.05 .65 .4225
4.70 4.05 .65 .4225
Sum = .8450
Sum of Squared Deviations .8450
Variance .4225
2 2
Repetition 3 Calculations:
3.40 4.70 1.44 9.54
Mean 3.18
3 3
Deviation Squared
Value Mean From Mean Deviation
3.40 3.18 .22 .0484
4.70 3.18 1.52 2.3104
1.44 3.18 1.74 3.0276
Sum = 5.3864
Sum of Squared Deviations 5.3864
Variance 1.7955
3 3
3. a. Variance of a constant times a variable: Var[cx] = c2Var[x]

b. Variance of the sum of a variable and a constant: Var[c + x] = Var[x]
31
c. Variance of the sum of two variables: Var[x + y] = Var[x] + 2Cov[x, y] + Var[y]
d. Variance of the sum of two independent variables: Var[x + y] = Var[x] + Var[y]
distribution indicates how reliable an estimate is. On the one hand, the variance is small,
there is a high probability that the estimate is close to the actual value. On the other hand,
if the variance is large, the probability that the estimate lies close the actual value is
small.
3. a. The mean of the coefficient estimate’s probability distribution is unaffected.

b. The variance of the coefficient estimate’s probability distribution decreases.
c. The coefficient estimate becomes more reliable. When the sample size increases more
information is available.
5. When the ordinary least squares (OLS) standard premises are satisfied, the ordinary least
squares (OLS) estimation procedure is the best linear unbiased estimation procedure (BLUE).
y j yi
1. a. bSlope
x j xi
b.
y j yi
bSlope
x j xi
E Const E x x j e j ( EConst E x xi ei )
x j xi
E Const E x x j e j E Const E x xi ei )
x j xi
E x x j E x xi e j ei
x j xi
E x ( x j xi ) (e j ei )
x j xi
e j ei
Ex
x j xi
1
Ex (e j ei )
x j xi
32
c.
Mean[bSlope ] [
Mean E x
1
x j xi
(e j ei ) ]
E x Mean [ x 1 x (e e )] j i
j i
Ex
1
x j xi
Mean e j ei )[ ]
Ex
1
x j xi
[
Mean[e j ] Mean[ei ] ]
Ex
d. What does the variance of bSlope‘s probability distribution, Var[bSlope], equal?
Var[bSlope ] [
Var E x
1
x j xi
(e j ei ) ]
Var [ x 1 x (e e )] j i
j i
1
( x j xi ) 2
[
Var e j ei ) ]
1
( x j xi ) 2
[
Var e j (ei )] ]
1
( x j xi ) 2
[
Var[e j ] Var[ei ] ]
1
( x j xi ) 2
[
Var[e j ] Var[ei ] ]
1
( x j xi ) 2
[
Var[e] Var[e] ]
2
Var[e]
( x j xi ) 2
3. a. There are 10 possible pair wise combinations of the 5 observations:
Obs 1 and 2 Obs 1 and 3 Obs 1 and 4 Obs 1 and 5 Obs 2 and 3
Obs 2 and 4 Obs 2 and 5 Obs 3 and 4 Obs 3 and 5 Obs 4 and 5
Since the two observations are chosen randomly the probability of each of these
1
combinations is .
10
1 1
Mean[bx] = uMean[bSlope for obs 1 and 2] + … + uMean[bSlope for obs 4 and 5]
10 10
1 1
= uEx + … + uE x
10 10
= Ex
33
b. The slopes of the 10 pair wise combinations of observations are independent so we need
not worry about the covariances.
Var[bx]
1 1
u Var[bSlope for Obs 1 and 2] u Var[bSlope for Obs 1 and 3]
10 10
1
+...+ u Var[bSlope for Obs 4 and 5]
10
1 2 1 2
u u Var[e] u u Var[e]
10 ( x1 x2 ) 2
10 ( x1 x3 ) 2
1 2
+...+ u u Var[e]
10 ( x4 x5 ) 2
1 2 1 2
u u 500 u u 500
10 ( x1 x2 ) 2
10 ( x1 x3 ) 2
1 2
+...+ u u 500
10 ( x4 x5 ) 2
1 1000 1 1000
u u
10 ( x1 x2 ) 10 ( x1 x3 ) 2
2
1 1000
+...+ u
10 ( x4 x5 ) 2
1 1000 1 1000 1 1000 1 1000
u u 2 u 2 u 2
10 62 10 12 10 18 10 24
1 1000 1 1000 1 1000
+ u 2 u 2 u 2
10 6 10 12 10 18
1 1000 1 1000
+ u 2 u 2
10 6 10 12
1 1000
+ u 2
10 6
100 100 100 100

62 122 182 242
100 100 100
+ 2 2 2
6 12 18
100 100
+ 2 2
6 12
100
+ 2
6
400 300 200 100

62 122 182 242
11.1 2.08 .62 .17 13.98
c. Yes
34
5. a. i.
Price 92.50159 28.12319 3.289157 0.0028
Const 5944.958 478.7771 12.41696 0.0000
Estimated Equation: EstOilProdBarrels = 5,945 + 92.5Price
ii.
Dependent Variable: OilProdBarrelsPlus1000
Price 92.50159 28.12319 3.289157 0.0028
Const 6944.958 478.7771 12.41696 0.0000
Estimated Equation: EstOilProdBarrelsPlus1000 = 6,945 + 92.5Price
Best fitting line: EstOilProdBarrelsPlus1000 = 6,945 + 92.5Price
b. The best fitting line
shifts up by 1,000.
c. All the points in the
scatter diagram have
shifted up by 1,000.
Consequently, the
best fitting line, the
line that minimizes
the sum of squared
residuals has also
shifted up by 1,000.
7. a. i.
PriceDollars 151.6556 47.57295 3.187853 0.0128
Const 516.7801 60.60223 8.527410 0.0000
Estimated Equation: EstGasCons = 516.8 151.7PriceDollars
35
ii.
PriceDollarsPlus2 151.6556 47.57295 3.187853 0.0128
Const 820.0912 155.6644 5.268328 0.0008
Estimated Equation: EstGasCons = 820.1 151.7PriceDollarsPlus2
b. The estimated constants

differ.
c. The estimated coefficients
are identical.
d. All the points in the scatter
diagram have shifted to the
right by 2.00.
Consequently, the best
fitting line, the line that
minimizes the sum of
squared residuals has also
shifted to the right by 2.00.
36
Chapter 7: Estimating the Variance of an Estimate’s
Probability Distribution
other variance, if the variance is large, the probability that the estimate lies close the
actual value is small.
3. a.
Rise 90 66 24 6
b. Slope =
Run 25 5 20 5
y 66 6
x 5 5
5 y 330 6 x 30
5 y 6 x 300
6
y x 60 1.2 x 60
5
c. 0
Rise 90 70 20
d. i. Slope = 1
Run 25 5 20
y 70
1
x5
y 70 x 5
y x 65
ii. 0
Rise 90 86 4 1
e. i. Slope =
Run 25 5 20 5
y 86 1
x 5 5
5 y 430 x 5
5 y x 425
1
y x 85 1.2 x 85
5
ii. 0
f. 0
37
Var[e ]
1. Var[bx ] T
¦ (x x )
t 1
t
2
3. While this is an unbiased estimation procedure for the variance of the error term’s probability
distribution, it does not help us. We can never compute the values of the error terms because
the error terms are unobservable.
5. We divide by the degrees of freedom because they equal the number of pieces of information
that are available to estimate the variance of the error term’s probability distribution.
1. First, calculate the means:

5 10 30 45
x 15
3 3
14 44 80 138
y 46
3 3
Second, for each student calculate the deviation of x from its mean and the deviation of y
from its mean:
Student yt y yt y xt x xt x
1 14 46 32 5 15 10
2 44 46 2 10 15 5
3 80 46 34 30 15 15
Third, calculate the products of the y and x deviations and squared x deviations for each
student; then calculate the sums:
Student ( yt y )( xt x ) ( xt x )2
32u10 = 320
1 102 = 100
2u5 = 10
2 52 = 25
3
34u15 = 510 152 = 225
Sum = 840 Sum = 350
Now, apply the formulas:
T
¦ ( y y )( x x )
t 1
t t
840
bx T
2.4
350
¦ (x x )
t 1
t
2
bConst y bx x 46 2.4 u15 46 36 10
38
3. Finally, use the quiz data to estimate the variance and standard deviation of the coefficient
estimate’s probability distribution, the coefficient estimate’s standard error.
Estimated Variance of the Coefficient Estimate’s Probability Distribution:
EstVar[e ] 168
EstVar[bx ] T
.48
350
¦ (x x )
t 1
t
2
Standard Error of the Coefficient Estimate:

SE[bx ] EstVar[bx ] .48 .69
5. a. Yes X No ____
SE of Regression EstVar[e] 78.09
EstVar[e ] 78.092 6, 098
b. Yes X No ____
SE[bx ] EstVar[bx ] 4.49
EstVar[bx ] 4.492 20.16
c. Yes X No ____
EstVar[e ]
EstVar[bx ] T
¦ (x x )
t 1
t
2
T
EstVar[e ] 6, 098
¦ (x x )
t 1
t
2
EstVar[bx ] 20.16
302.48
d. Yes X No ____
T
¦(y
t 1
t y )( xt x )
bx T
¦ (x x )
t 1
t
2
T T
¦ ( yt y )( xt x ) bx u ¦ ( xt x )2 15.26 u 302.48 4, 615.8

t 1 t 1
e. Yes X No ____
SSR
EstVar[e ]
Degrees of Freedom
SSR EstVar[e ] u Degrees of Freedom 6, 098 u (10 2) 6, 098 u 8 48, 784
f. Yes ____ No X
39
Chapter 8: Interval Estimates and Hypothesis Testing
5 10 30 45
1. x 15
3 3
T
¦ (x x )
t 1
t
2
(5 15)2 (15 15) 2 (25 15) 2 102 02 102 100 0 100 200
Var[e ] 50
Var[bx ] T
.25
200
¦ (x x )
t 1
t
2
|68.5%
|95.5%
|99.7%
Actual Values From To Repetitions Between
Ex Var[e] Value Value From and to Values
2 50 1.5 2.5 |68.5%
2 50 1.0 3.0 |95.5%
2 50 .5 3.5 |99.7%
3. z equals the number of standard deviations the value lies from the mean:
Value of Random Variable Distribution Mean
z
Distribution Standard Deviation
Number of Stadard Deviations from the Mean
1. Interval Estimate Question: What is the probability that the estimate, _____, lies within
____ of the actual value____?
3.
x Step 0: Formulate a model reflecting the theory to be tested.
x Step 1: Collect data, run the regression, and interpret the estimates.
x Step 2: Play the cynic, challenge the evidence, and construct the null and alternative
hypotheses.
x Step 3: Formulate the question to assess the null hypothesis and the cynic’s view.
x Step 4: Use the general properties of the estimation procedure, the probability distribution
of the estimates, to calculate Prob[Results IF H0 True].
x Step 5: Decide on the standard of proof, a significance level.
41
5.
x Each estimation procedure is unbiased; each estimation procedure does not systematically
underestimate or overestimate the actual value.
x The ordinary least squares (OLS) estimation procedure for the coefficient value is the
best linear unbiased estimation procedure (BLUE).
1. a. PetroConsPC = PetroCons*1000/Pop
PriceReal = PriceNom*100/CPI
b.
Dependent Variable: PetroConsPC
PriceReal 540.5367 203.4389 2.656999 0.0289
Const 1418.403 144.7593 9.798354 0.0000
Estimated Equation: EstPetroConsPC = 1,418 540.5PriceReal

Interpretation of Estimates:
bPriceReal = 540.5: A $1 increase in the real price of gasoline reduces per capita
Nebraska petroleum consumption by 540.5 gallons.
c. The Student t-distribution must be used since we must estimate the variance of the
coefficient estimate’s probability distribution.
d. First, note that the degrees of freedom equal 8:
Degrees of Freedom = Sample Size – Number of Estimated Parameters
= 10 – 2
=8
i. Convert 400 into standard errors:
400
400 equals = 1.97 standard errors.
203.4
The probability of the
coefficient estimate
being within 400 of the
actual coefficient value
equals the probability of
the coefficient estimate
being within 1.97
standard errors of the
actual coefficient value,
that is, between t’s of
1.97 and +1.97.
42
Using the Econometrics Lab:
Left tail: Right tail:
Degrees of Freedom: 8 Degrees of Freedom: 8
t: 1.97 t: 1.97
Left tail probability = .04 Right tail probability = .04
The probability of the coefficient estimate being within 200 of the actual
coefficient value equals .92.
1.00 (.04 + .04) = 1.00 .08 = .92
ii. 300 of the actual coefficient value? .82
Convert 300 into standard errors:
300
300 equals = 1.47 standard errors.
203.4
The probability of the coefficient estimate being within 300 of the actual coefficient value
equals the probability of the coefficient estimate being within 1.47 standard errors of the
actual coefficient value, that is, between t’s of 1.47 and +1.47.
t: 1.47 t: 1.47
equals .82.
1.00 (.09 + .09) = 1.00 .18 = .82
iii. 200 of the actual coefficient value? .64.
Convert 200 into standard errors:
200
200 equals = .98 standard errors.
203.4
equals the probability of the coefficient estimate being within .98 standard errors of the actual
coefficient value, that is, between t’s of .98 and +.98.
t: .98 t: .98
equals .64
1.00 (.18 + .18) = 1.00 .36 = .64
3. a. Economic theory teaches us that the demand curve is downward sloping. Accordingly,
theory postulates that an increase in the price should decrease consumption.
b. Step 0: Formulate a model reflecting the theory to be tested.
Model: GasConst = EConst + EPPriceDollarst + et
The theory suggests that EP should be negative.
43
Step 1:
PriceDollars 151.6556 47.57295 3.187853 0.0128
Const 516.7801 60.60223 8.527410 0.0000
Estimated Equation: EstGasCons = 516.8 151.7PriceDollars

bPriceDollars = 151.7: A $1 increase in the gasoline price decreases gasoline consumption
by 151.7 million gallons per day.
Step 2: Play the cynic, challenge the evidence, and construct the null and alternative
hypotheses.
Cynic’s view: Despite the results, the price of gasoline has no impact on gasoline
consumption. The results were just “the luck of the draw.”
Now, we construct the null and alternative hypotheses:
H0: EP = 0 Cynic is correct: Price has no impact on gasoline consumption
H1: EP < 0 Cynic is incorrect: Higher prices reduce gasoline consumption
Step 3: Formulate the question to assess the null hypothesis and the cynic’s view.
Question for the Cynic:
x Generic Question: What is the probability that the results would be like those we
actually obtained (or even stronger), if the cynic is correct and price actually has no
impact?
x Specific Question: The regression’s coefficient estimate was 151.7: What is the
probability that the coefficient estimate in one regression would be 151.7 or less if
H0 were actually true (if the actual coefficient, EP, equals 0)?
Answer: Prob[Results IF Cynic Correct] or Prob[Results IF H0 True]
Step 4: Use the general properties of the estimation procedure, the probability distribution
OLS estimation If H0were Number of Number of
procedure unbiased true SE column observations parameters
Mean[bP] = EP = 0 SE[bP] = 47.57 DF = 10 2 = 8
44
Recall that the Prob. column
reports the “tails probability:”
Prob. Column (Tails

Probability): The probability
that the coefficient estimate,
bP, resulting from one
regression would be 151.7
or less, if the actual
coefficient, EP, equals 0.
.0128
Prob[Results IF H0 True] = | .006
2
The probability that the coefficient estimate in one regression would be 151.7 or
less if H0 were actually true (if the actual coefficient, EP, equals 0) is .006.
c. Step 5: Decide on the standard of proof, a significance level
If we adopt a 1 or 5 percent significance level, we conclude that this probability is
small and reject the null hypothesis thereby supporting the theory.
5. a. Step 0: Formulate a model reflecting the theory to be tested.

Model: Numbert = EConst + EDemPartyDem1t + et
The allegation suggests that EDem should be positive.
Step 1: Collect data, run the regression, and interpret the estimates.
PartyDem1 2.361883 0.577102 4.092661 0.0001
Const 5.613527 0.424482 13.22441 0.0000
Estimated Equation: EstNumber = 5.6 + 2.36PartyDem1

bDem = 2.36: A Democratic member of Congress receives 2.36 more earmarks than
his/her non-Democratic colleagues.
hypotheses.
Cynic’s view: Despite the results, party membership has no impact on earmarks. The
results were just “the luck of the draw.”
H0: EDem = 0 Cynic is correct: Party affiliation has no impact on earmarks.
H1: EDem > 0 Cynic is incorrect: Democratic party affiliation increases
earmarks.
45
actually obtained (or even stronger), if the cynic is correct and party affliliation
actually has no effect on earmarks?
probability that the coefficient estimate in one regression would be 2.36 or more if
H0 were actually true (if the actual coefficient, EDem, equals 0)?
OLS estimation If H0were Number of Number of
Mean[bDem] = EDem = 0 SE[bDem] = .577 DF = 451 2 = 449
Recall that the Prob.

column reports the “tails
probability:”
Prob. Column (Tails

Probability): The
probability that the
coefficient estimate, bDem,
resulting from one regression would be 2.36 or more, if the actual coefficient, EDem,
equals 0.
.0001
Prob[Results IF H0 True] = < .0001
2
The probability that the coefficient estimate in one regression would be 2.36 or more
if H0 were actually true (if the actual coefficient, EDem, equals 0) is less than .0001.
b. Step 5: Decide on the standard of proof, a significance level
If we adopt a 1 or 5 percent significance level, we conclude that this probability is small
and reject the null hypothesis thereby supporting the allegation.
7. a. Economic theory teaches us that the supply curve is upward sloping. Accordingly, theory
postulates that an increase in the price should increase production.
b. Step 0: Formulate a model reflecting the theory to be tested.
Model: OilProdBarrelst = EConst + E PPricet + et
Theory suggests that EP should be positive.
46
Price 92.50159 28.12319 3.289157 0.0028
Const 5944.958 478.7771 12.41696 0.0000
Estimated Equation: EstOilProdBarrels = 5,945 + 92.5Price

bP = 92.5: A 1 dollar increase in the price increases oil production by 92.5 thousand
barrels.
hypotheses.
Cynic’s view: Despite the results, price has no impact on oil production. The results were
just “the luck of the draw.”
H0: EP = 0 Cynic is correct: Price has no impact on oil production.
H1: EP > 0 Cynic is incorrect: A higher price increases oil production.
x Generic Question: What is the probability that the results would be like those
we actually obtained (or even stronger), if the cynic is correct and price
actually has no impact on oil production?
probability that the coefficient estimate in one regression would be 92.5 or
more if H0 were actually true (if the actual coefficient, EP, equals 0)?
OLS estimation If H0 were Number of Number of
Mean[bP] = EP = 0 SE[bP] = 28.1 DF = 29 2 = 27
Recall that the Prob. column reports

the “tails probability:”
Prob. Column (Tails Probability):

The probability that the coefficient
estimate, bP, resulting from one
regression would be 92.5 or more, if
the actual coefficient, EP, equals 0.
47
.0028
Prob[Results IF H0 True] = = .0014
2
The probability that the coefficient estimate in one regression would be 92.5 or more
if H0 were actually true (if the actual coefficient, E P, equals 0) is .0014.
c. Step 5: Decide on the standard of proof, a significance level
The Prob[Results IF H0 True] equals .0014, less than .01. So, even with a 1 percent
significance level, we would reject the null hypothesis and thereby support the
theory.
48
Chapter 9: One-Tailed Tests, Two-Tailed Tests, and Logarithms
1. a. Q EConst P E P
dQ
EConst E P P E P 1
dP
P P 1
b.
Q EConst P EP EConst P EP 1
dQ P 1
c. EConst E P P E P 1
EP
dP Q EConst P EP 1
3. a. PuQ = 1,000
b. i. EConst = 1,000
ii. EP = 1
PuQ = 1,000
Q = 1,000P 1
d log( z ) 1
5.
dz z
1. a. Theory: E>c or E<c

H0: E = c H0: E = c
H1: E > c H1: E < c
b. Theory: E=c
H0: E = c
H1: E z c
1. a. Necessity.
b. Yes, the demand should be inelastic because the demand for a necessity is inelastic.
c. Demand is inelastic whenever the absolute value of the price elasticity of demand is less
than 1. But now remember that the numerical value of the price elasticity of demand will
be negative since the demand curve is downward sloping. Accordingly, the numerical
value of the price elasticity of demand should be greater than 1.
d. Step 0: Formulate a model reflecting the theory to be tested.
Constant price elasticity model:
PetroConsPC = EConst PriceRealEP where EP = Price elasticity of demand
49
Taking natural logarithms of both sides.
log(PetroConsPC) = log(EConst) + EPlog(PriceReal)
LogPetroConsPC = c + EPLogPriceReal
where LogPetroConsPC = log(PetroConsPC)
c = log(EConst)
LogPriceReal = log(PriceReal)
Theory suggests that the absolute value of EP should be less than 1; since EP is
negative, theory suggests that EP should be greater than 1.
Generate two new variables:
x LogPetroConsPC = log(PetroConsPC)
x LogPriceReal = log(PriceReal)
Dependent Variable: LogPetroConsPC
LogPriceReal 0.365454 0.131853 2.771670 0.0242
Const 6.813011 0.048429 140.6798 0.0000
Estimated Equation: EstLogPetroConsPC = 6.81 .365LogPriceReal

bP = .365: 1 percent increase in the price decreases petroleum consumption by .365
percent. That is, the estimate of the price elasticity of demand equals .365.
Step 2: Play the cynic, challenge the evidence, and construct the null and alternative hypotheses.
Cynic’s view: Sure, the coefficient estimate from regression suggests that the demand
is inelastic, but this is just “the luck of the draw.” In fact, the price elasticity of
demand is not greater than 1, it equals 1.
H0: EP = 1 Cynic’s view is correct: Actual price elasticity of demand equals

1
H1: EP > 1 Cynic’s view is incorrect: Demand is inelastic; the elasticity of
demand is greater than 1
Step 3: Formulate the question to assess the cynic’s view.
x Generic Question: What is the probability that the results would be like those
we actually obtained (or even stronger), if the cynic is correct and actual
price elasticity of demand equals 1?
x Specific Question: The regression’s coefficient estimate was .365: What is the
probability that the coefficient estimate, bP, in one regression would be .365 or
more, if H0 were actually true (if the actual coefficient, EP, equals 1)?
50
If the null hypothesis
were true, the actual
coefficient would equal
1. Since ordinary least
squares (OLS) estimation
procedure for the
coefficient value is
unbiased, the mean of the
probability distribution of
coefficient estimates would be 1. The SE column of the printout provides us with the
standard error of the coefficient estimates, .132. The degrees of freedom equal the
number of observations, 10, less 2, since we are estimating two parameters, the constant
and the coefficient; consequently the degrees of freedom equal 8.
OLS estimation Assume H0 Number of Number of

procedure unbiased is true SE column observations parameters
Mean[bP] = E P = 1 SE[bP] = .132 DF = 10 2 = 8
Econometrics Lab: Prob[Results IF H0 True] = .0007

Clever Algebraic Manipulations to Calculate Prob[Results IF H0 True].
Alternatively, we can use a clever algebraic manipulation approach by defining
EClever so that it equals 0 when E P equals 1:
EClever = E P + 1
and
EP = EClever 1
Next, consider the model and perform some algebra:
LogPetroConsPC = c + EPLogPriceReal
LogPetroConsPC = c + (EClever 1)LogPriceReal
LogPetroConsPC + LogPriceReal = c + ECleverLogPriceReal
yClever = c + ECleverLogPriceReal
where yClever = LogPetroConsPC + LogPriceReal
Expressing the null and alternative hypotheses in terms of EClever:
H0: EClever = 0 EP = 1 Cynic’s view is correct: Actual price elasticity
of demand equals 1
H1: EClever > 0 EP > 1 Cynic’s view is incorrect: Demand is inelastic
51
Dependent Variable: yClever
LogPriceReal 0.634546 0.131853 4.812516 0.0013
Const 6.813011 0.048429 140.6798 0.0000
.0013
Prob[Results IF H0 True] = | .0007
2
We have obtained the same value for Prob[Results IF H0 True].
e. Step 5: Decide on the standard of proof, a significance level
Since Prob[Results IF H0 True] equals .0007, we reject the null hypothesis that
demand is unit elastic even at the 1 percent significance level. This supports the
theory that demand is inelastic.
3. a. Step 0: Formulate a model reflecting the theory to be tested.

OilProdBarrels = EConst PriceEP where EP = Price elasticity of demand
log(OilProdBarrels) = log(EConst) + EPlog(Price)
LogQ = c +EPLogP
where LogQ = log(QilProdBarrels)
c = log(EConst)
LogP = log(Price)
Theory asserts that EP equals .10.
We must generate the two variables: the logarithm of quantity and the logarithm of
price:
x LogQ = log(OilProdBarrels)
x LogP = log(Price)
52
Dependent Variable: LogQ
LogP 0.213620 0.069544 3.071739 0.0048
Const 8.324825 0.187512 44.39626 0.0000
Estimated Equation: EstLogQ = 8.32 + .2136LogP

bP = .2136: A 1 percent increase in the price increases oil production by .2136 percent.
That is, the estimate of the price elasticity of supply equals .2136. The evidence
suggests that the elasticity of supply does not equal .10. More specifically, the
estimate is .1136 from .10.
hypotheses.
Cynic’s view: Sure the coefficient estimate from regression suggests that the price
elasticity of supply does not equal .10, but this is just “the luck of the draw.” In fact,
the actual price elasticity of supply equals .10.
H0: EP = .10 Cynic’s view is correct: Actual price elasticity of supply equals .10
H1 EP z .10 Cynic’s view is incorrect: Actual price elasticity of supply does not equal .10
x Generic Question: What is the probability of obtaining a result like the one
obtained from the regression (or even stronger), if the cynic is correct and
actual price elasticity of supply equals .10?
x Specific Question: The regression’s coefficient estimate was .2136: What is
the probability that the coefficient estimate, bP, in one regression would be at
least .1136 from .10, if H0 were actually true (if the actual coefficient, EP,
equals .10)?
53
If the null hypothesis were true, the
actual coefficient would equal .10.
Since ordinary least squares (OLS)
estimation procedure for the
coefficient value is unbiased, the
mean of the probability distribution
of coefficient estimates would be
.10. The SE column of the printout
provides us with the standard error
of the coefficient estimates. The
degrees of freedom equal the
number of observations less 2, since we are estimating two parameters, the constant and
the coefficient.
Mean[bP] = E P = .10 SE[bP] = .0695 DF = 29 2 = 27
i. We can use the Econometrics Lab to compute Prob[Results IF H0 True]:

We shall calculate this probability in two steps:
x First, calculate the probability of the estimated lying in the right hand
tail. Calculate the probability that the estimate lies .1136 or more above
.10; that is, the probability that the estimate lies at or above .2136.
x Second, calculate the probability of the estimated lying in the left hand
tail. Calculate the probability that the estimate lies .1136 or more below
.10; that is, the probability that the estimate lies at or below .0136.
Left Right
Tail Tail
p p
Prob[Results IF H0 True] = .0569 + .0569 | .114.
ii. Alternatively, we can use a clever algebraic manipulation approach by defining EClever
so that it equals 0 when E P equals .10:
EClever = E P .10
and
EP = EClever + .10
Next, consider the model and perform some algebra:
LogQ = c + EPLogP
LogQ = c + (EClever + .10)LogP
LogQ .10LogP = c + ECleverLogP
yClever = c + ECleverLogPriceReal
where yClever = LogQ .10LogP
Expressing the null and alternative hypotheses in terms of EClever:
54
H0: EClever = 0 EP = .10 Cynic’s view is correct: Actual price elasticity
of supply equals .10
H1: EClever z 0 EP z .10 Cynic’s view is incorrect: Actual price
elasticity of supply does not equal .10
Dependent Variable: yClever
LogP 0.113620 0.069544 1.633795 0.1139
Const 8.324825 0.187512 44.39626 0.0000
Next, calculate Prob[Results IF H0 True] focusing on EClever:
Question: What is the

probability of
obtaining a result like
the one calculated
from the regression (a
coefficient estimate,
bClever, of .1136, .1136
from 0), if the cynic’s
view and the null
hypothesis were correct (that is, if the actual coefficient, EClever, equals 0)?
.1139 .1139
Answer: Prob[Results IF H0 True] = + | .114.
2 2
This is the same answer as before. By a clever algebraic manipulation, we can get
the statistical software to perform the calculations.
b. Step 5: Decide upon the standard of proof, what constitutes proof beyond a reasonable
doubt. Decide on the significance level, the dividing line between small and large
probability:
Even at a 10 percent significance level, we do not reject the null hypothesis that the
elasticity of supply equals .10. That is, at a 10 percent significance level, the estimate
of .2136 is not statistically different from .10.
5. a. Most regard cigarettes as addictive. Accordingly, the price elasticity of demand for
cigarettes should be inelastic.
b. Step 0: Formulate a model reflecting the theory to be tested:
CigConsPC = EConst PriceConsumerEP where EP = Price elasticity of demand
55
Taking natural logarithms of both sides..
log(CigConsPC) = log(EConst) + EPlog(PriceConsumer)

LogCigConsPC = c + EPLogPriceConsumer
where LogCigConsPC = log(CigConsPC)
c = log(EConst)
LogCigConsPC = log(PriceConsumer)
Theory suggests that the absolute value of EP should be less than 1; since EP is
negative, theory suggests that EP should be greater than 1.
x LogCigConsPC = log(CigConsPC)
x LogPriceConsumer = log(PriceConsumer)
Dependent Variable: LogCigConsPC
LogPriceConsumer 1.443914 0.237416 6.081787 0.0000
Const 6.379758 0.381990 16.70135 0.0000
Estimated Equation: EstLogCigConsPC = 6.38 1.44 LogPriceConsumer

bP = 1.44: A 1 percent increase in the price decreases gasoline consumption by 1.44
percent. That is, the estimate of the price elasticity of demand equals 1.44.
c. No. The estimate for the price elasticity of demand suggests that demand is elastic, not
inelastic.
7. a. Step 0: Formulate a model reflecting the theory to be tested:

HoursPerWeek =EConst WageEW where EW = Wage elasticity of supply
log(HoursPerWeek) = log(EConst) + EWlog(Wage)
LogHoursPerWeek = c + EWLogWage
where LogHoursPerWeek = log(HoursPerWeek)
c = log(EConst)
LogWage = log(Wage)
Theory suggests that the absolute value of EP should be less than 1.

x LogHoursPerWeek = log(HoursPerWeek)
x LogWage = log(Wage)
56
Dependent Variable: LogHoursPerWeek
LogWage 0.768171 0.294727 2.606380 0.0107
Const 0.037095 1.093466 0.033925 0.9730
Estimated Equation: LogHoursPerWeek = .037 + .768 LogWage

bW = .768: A 1 percent increase in the wage increases hours worked by .768 percent.
That is, the estimate of the wage elasticity of supply equals .768.
Step 2: Play the cynic, challenge the evidence, and construct the null and alternative hypotheses.
Cynic’s view: Sure, the coefficient estimate from regression suggests that the supply is
inelastic, but this is just “the luck of the draw.” In fact, the wage elasticity of supply is
not less than 1, it equals 1.
H0: EP = 1 Cynic’s view is correct: Actual wage elasticity of supply equals 1
H1: EP < 1 Cynic’s view is incorrect: Supply is inelastic; the wage elasticity of
supply is less than 1
actually obtained (or even stronger), if the cynic is correct and actual wage
elasticity of supply equals 1?
x Specific Question: The regression’s coefficient estimate was .768: What is the
probability that the coefficient estimate, bP, in one regression would be .768 or less,
if H0 were actually true (if the actual coefficient, EP, equals 1)?
of the estimates, to
calculate Prob[Results
IF H0 True].
If the null hypothesis
were true, the actual
coefficient would
equal 1. Since ordinary
least squares (OLS)
estimation procedure
for the coefficient
value is unbiased, the
mean of the probability distribution of coefficient estimates would be 1. The SE column
of the printout provides us with the standard error of the coefficient estimates, .295. The
degrees of freedom equal the number of observations, 92, less 2, since we are estimating
two parameters, the constant and the coefficient; consequently the degrees of freedom
equal 90.
57
Mean[bWage] = E Wage = 1 SE[bP] = .295 DF = 92 2 = 90
Econometrics Lab: Prob[Results IF H0 True] = .2168

b. Step 5: Decide on the standard of proof, a significance level.
Even at the 10 percent significance level we cannot reject the null hypothesis casting
suspicion on the theory.
58
Chapter 10: Multiple Regression Analysis—Introduction
1. a.
Q EConst P E I E ChickP E
P I CP
EConst P E I E ChickP E
P I P EI
EConst P E I E ChickP E ChickP E

P I P I
EP EI
EConst (ChickP
P
)(ChickP
I
)
EP EI
EP EI
P I
EConst ( ChickP )( ChickP )
b.
P
is unchanged
ChickP
I
is unchanged
ChickP
Q is unchanged
c. If EP + EI + ECP = 0, then ECP = EP EI.
In part b, we showed that if ECP = EP EI, then Q is unchanged when the good’s own
price, income, and the price of other goods all double.
3. log(Q) = log(EConst) + E Plog(P) + EIlog(I) + ECPlog(ChickP)

Since EClever = EP + EI + ECP, ECP = EClever EP EI; now, substitute for ECP
= log(EConst) + EPlog(P) + EIlog(I) + (EClever EP EI)log(ChickP)

Multiplying out the last term
= log(EConst) + EPlog(P) + EIlog(I) + ECleverlog(ChickP) EPlog(ChickP) EIlog(ChickP)
Rearranging terms
= log(EConst) + EPlog(P) EPlog(ChickP) + EIlog(I) EIlog(ChickP)+ ECleverlog(ChickP)
Factoring EP from the second and third terms and EI from the fourth and fifth terms
= log(EConst) + EP[log(P) log(ChickP)] + EI[log(I) log(ChickP)] + ECleverlog(ChickP)
59
1. Multiple regression analysis includes more than one explanatory variables; simple regression
analysis includes only a single explanatory variable.
3. No.
5. Yes, if the exponents sum to 0.
1.
a.
x Additional labor increases value added.
x Additional land under cultivation increases value added.
x Additional machinery increases value added.
b.
x ELabor > 0.
x ELand > 0.
x EMachinery > 0.
c.
x H0: ELabor = 0
H1: ELabor > 0
x H0: ELand = 0
H1: ELand > 0
x H0: EMachinery = 0
H1: EMachinery > 0
d.
Dependent Variable: ValueAdded
Labor 267.6604 11.53321 23.20780 0.0000
Land 2013.179 794.9466 2.532471 0.0125
Machinery 21917.88 1012.361 21.65027 0.0000
Const 5.28E+08 5.20E+08 1.015321 0.3119
Estimated Equation: EstValueAdded = 528,000,000 + 268Labor

+ 2,013Land + 21,918Machinery
bLabor = 268: 1 more farm worker, while the amount of cultivated land and machinery
remain constant, increases value added by $268.
bLand = 2,013: 1 more square kilometer of cultivated land, while the amount of labor
60
and machinery remain constant, increases value added by $2,013.
bMachinery = 21,918: We estimate that 1 more tractor, while the amount of labor and
cultivated land remain constant, increases value added by $21,918.
e. Note that one-tail tests are appropriate. The Prob column indicates that we would not
reject any of our null hypotheses. Consequently all our theories are supported.
3. a. Taking natural logarithms and adding an error term:

log(ValueAddedt) = log(EConst) + ELaborlog(Labort) + ELandlog(Landt)
+ EMachinerylog(Machineryt) + et
b. Estimate the parameters of the model.
Dependent Variable: LogValueAdded
LogLabor 0.405212 0.051607 7.851944 0.0000
LogLand 0.028024 0.062265 0.450084 0.6534
LogMachinery 0.421532 0.029000 14.53552 0.0000
Const 11.27838 0.680996 16.56159 0.0000
c. H0: ELabor + ELand + EMachinery = 1
H1: ELabor + ELand + EMachinery z 1
5. a. i. Downward Sloping Demand Theory: As the price of cigarettes rise, per capita
cigarette consumption decreases.
ii. Conventional Wisdom: Cigarettes are an inferior good; as per capita income rises, per
capita cigarette consumption decreases.
b. CigConsPCt = EConst + EPPriceConsumert + EIIncPCt + et
c. Theory: EP < 0. Theory: EI < 0.
d. H0: EP = 0 H0: EI = 0
H1: EP < 0 H1: EI < 0
e.
Dependent Variable: CigConsPC
PriceConsumer 16.04703 3.668014 4.374856 0.0001
IncPC 0.318916 0.588471 0.541939 0.5904
Const 155.0632 20.12211 7.706110 0.0000
Estimated Equation: EstCigConsPC = 155 16.0PriceConsumer .32IncPC

bP = 16.0: A $1 increase in the price of a pack of cigarettes, while per capita income
remains constant, decreases per capita cigarette consumption by 16 packs per year.
61
bI = .32: A $1,000 increase in per capita income, while the price of cigarettes
remains constant, decreases per capita cigarette consumption by .32 packs per year.
f. One-tail tests are appropriate. Both theories suggest that the actual coefficient is negative.
x The regression results lend support to the downward sloping demand theory. As
the Prob column indicates we easily reject the null hypothesis at the 1 percent
significance level.
x The conventional wisdom asserting that cigarettes are an inferior is shaky. While
the coefficient is negative we cannot come close to rejecting the null hypothesis
at the 1, 5, or even 10 percent significance level.
7. a. i. As the price of cigarettes rise, the youth smoking rate should decrease.
ii. As per capita income rises, the youth smoking rate should decrease.
b. SmokeRateYoutht = EConst + EPPriceConsumert + EIIncPCt + et
c. Theory: EP < 0. Theory: EI < 0.
d. H0: EP = 0 H0: EI = 0
H1: EP < 0 H1: EI < 0
e.
Dependent Variable: SmokeRateYouth
PriceConsumer 1.455381 0.665547 2.186743 0.0337
IncPC 0.202018 0.106776 1.891985 0.0645
Const 33.45011 3.651081 9.161700 0.0000
Estimated Equation: EstSmokeRateYouth = 33.5 1.46PriceConsumer

.20IncPC
bP = 1.46: A $1 increase in the price of a pack of cigarettes, while per capita income
remains constant, decreases the youth smoking rate by 1.46 percentage points.
bI = .20: A $1,000 increase in per capita income, while the price of cigarettes
remains constant, decreases the youth smoking rate by .20 percentage points.
f. One-tail tests are appropriate. Both theories suggest that the actual coefficient is negative.
In both cases, we reject the null hypothesis at the 5 percent significance level, but not at
the 1 percent significance level.
62
Chapter 11: Hypothesis Testing and the Wald Test
1. log(Q) = log(EConst) + EPlog(P) + EIlog(I) + ECPlog(ChickP) + et

Since EP + EI + ECP = 0, ECP = (EP + EI)
= log(EConst) + EPlog(P) + EIlog(I) (EP + EI)log(ChickP) + et
Multiplying through
= log(EConst) + EPlog(P) + EIlog(I) EPlog(ChickP) EIlog(ChickP) + et
Rearranging terms
= log(EConst) + EPlog(P) EPlog(ChickP) + EIlog(I) EIlog(ChickP) + et
Factoring EP from the second and third terms and EI from the fourth and fifth terms
= log(EConst) + EP[log(P) log(ChickP)] + EI[log(I) log(ChickP)] + et
3. a. Before the experiment is conducted, we cannot determine the numerical value of the
random variable with certainty.
b. Before the experiment is conducted, we can often calculate the random variable’s
probability distribution telling us how likely it is for the random variable to equal each of
its possible numerical values.
1. No, the restricted sum of squared residuals cannot be less than the unrestricted sum. This is
true because the ordinary least squares (OLS) estimation procedure “chooses” the parameter
estimates that minimize the sum of squared residuals. Placing a restriction on one or more of
the parameters can only increase, not decrease, the sum of squared residuals.
3. A two-tailed t-test is a special case of the Wald test in which the value of a single coefficient
is restricted to equal 0.
1. a H0: E Labor + E Land + EMachinery = 1

H1: E Labor + E Land + EMachinery z 1
b. See the solutions to exercises 1 and 3 from Chapter 10.
3. Assess the constant returns to scale theory using the Wald approach the “easy way” using
statistical software.
63
Dependent Variable: LogValueAdded
LogLabor 0.405212 0.051607 7.851944 0.0000
LogLand 0.028024 0.062265 0.450084 0.6534
LogMachinery 0.421532 0.029000 14.53552 0.0000
Const 11.27838 0.680996 16.56159 0.0000
Degrees of Freedom
Value Num Dem Prob
F-statistic 7.223722 1 129 0.0081
5. a.
PriceSupplier 34.16907 10.66315 3.204408 0.0024
Tax 12.69334 4.040626 3.141428 0.0029
Const 207.6718 38.53283 5.389477 0.0000
Estimated Equation: EstCigConsPC = 207.7 34.2PriceSupplier 12.7Tax

b. Interpret the coefficient estimates.
x A $1 increase in the price of cigarettes received by suppliers decreases per
capita cigarette consumption by 34.2 packs.
x A $1 increase in the cigarette tax decreases per capita cigarette consumption
by 12.7 packs.
7. a.
PriceSupplier 34.16907 10.66315 3.204408 0.0024
Tax 12.69334 4.040626 3.141428 0.0029
Const 207.6718 38.53283 5.389477 0.0000
SSRU = 19,665.41
DFU = 51 3 = 48
64
b. CigConsPCt = EConst + EPriceSupplierPriceSuppliert + ETaxTaxt + et
H0: EPriceSupplier = ETax
= EConst + ETaxPriceSuppliert + ETaxTaxt + et
= EConst + ETax(PriceSuppliert + Taxt) + et
Since PriceConsumert = PriceSuppliert + Taxt
= EConst + ETaxPriceConsumert + et
c.
PriceConsumer 17.02657 3.168679 5.373398 0.0000
Const 148.6762 16.19163 9.182288 0.0000
SSRR = 20,822.81
DFR = 51 2 = 49
d. SSRR = 20,822.81 DFR = 49
SSRU = 19.665.41 DFU = 48
SSRR SSRU = 1,157.40 DFR DFU = 1
( SSRR SSRU ) / ( DFR DFU ) 1,157.40 /1 1,157.40

F = = = 2.825
SSRU / DFU 19, 665.41/ 48 409.69
e. Prob[Results IF H0 True] = .0993.
65
Chapter 12: Model Specification and Development
1. 0.
3. a. EUnemCurrent < 0
b. H0: EUnemCurrent = 0
H1: EUnemCurrent < 0
5. a. 1912 election stands out because third parties received more than a third of the vote.
b. Theodore Roosevelt, a former Republican President, split from his party and formed a
third party called the Bull Moose Party. One of his opponents was his former vice
president.
1. An artificial model is associated with an “original” model. The original model is designed to
assess a theory. The artificial model is not designed to assess the theory; it assesses the
properties of the original model.
3. a. i. No.
ii. No.
b. i. Yes.
ii. Yes.
1. a.
x Higher real GDP growth rates should increase the vote for the incumbent’s party.
x Higher inflation rates should decrease the vote for the incumbent’s party.
x More consecutive terms should decrease the vote for the incumbent’s party (throw
the ****s out).
b. VotePresPartyTwot = EConst + EGdpGrowthRealGdpGrowtht
+ EInflGdpCurrentInflGdpCurrentt + E PresTermsPresPartyTermst + et
The theories imply:
x EGdpGrowth > 0.
x EInflGdpCurrent < 0.
x EPresTerms < 0.
67
c.
Dependent Variable: VotePresPartyTwo
RealGdpGrowth 0.435612 0.221348 1.967995 0.0598
InflGdpCurrent –0.656971 0.293145 –2.241111 0.0338
PresPartyTerms –3.144539 1.078933 –2.914488 0.0072
Const 58.38144 2.595120 22.49663 0.0000
Estimated Equation: EstVotePresPartyTwo = 58.38 + .436RealGdpGrowth

.657InflGdpCurrent 3.145PresPartyTerms
bGdpGrowth = .436: A 1 percentage point increase in the GDP growth rate increases the
popular vote of the incumbent President’s party by .436 percentage points.
bInflGdpCurrent = .657: A 1 percentage point increase in inflation as measure by the
GDP price deflator decreases the popular vote of the incumbent President’s party by
.657 percentage points.
bPresTerms = 3.145: One additional term the incumbent President’s party occupies the
White House decreases the popular vote of the incumbent President’s party by 3.145
percentage points.
d. GDP Growth Inflation Presidential Terms
H0: EGdpGrowth = 0 H0: EInflGdpCurrent = 0 H0: EPresTerms = 0
H1: EGdpGrowth > 0 H1: EInflGdpCurrent < 0 H1: EPresTerms < 0
e. GDP Growth Inflation Presidential Terms
.0598 .0338 .0072
| .030 | .017 | .004
2 2 2
We reject the null hypotheses for the GDP growth and inflation theories at the 5
percent significance level, but do not reject the null hypotheses at the 1 percent
significance level. We reject the presidential terms null hypothesis at the 1
percent significance level.
3. a. Artificial Model: Tax = EConst + ETobProdPCTobProdPCt + JEstTax2 EstTaxt2 + et

Ramsey RESET Test
Dependent Variable: Tax
TobProdPC 0.076634 0.050998 1.502701 0.1395
Const 3.387954 2.322631 1.458670 0.1512
Fitted^2 2.693214 1.324261 2.033749 0.0475
Critical Result: The Fitted^2 coefficient is 2.69. The estimate does not equal 0; the
estimate is 2.69 from 0. This evidence suggests that the new form of
the information adds explanatory power.
68
H0: JEstTax2 = 0 New form of the information adds NO explanatory power
H1: JEstTax2 z 0 New form of the information adds explanatory power
Prob[Results IF H0 True] =.0475
We do not reject the null hypothesis at the 1 percent significance level, but we do reject
the null hypothesis at the 5 percent significance level. So, perhaps we could do better by
using the information in a different form.
b.
SqrtTobProdPC 0.187790 0.068316 2.748863 0.0083
Const 1.389676 0.110588 12.56625 0.0000
H0: ESqrtTobProdPC = 0
H1: ESqrtTobProdPC < 0
.0083
Prob[Results IF H0 True] = = .0042
2
We do reject the null hypothesis at the 1 percent significance level.
c. Artificial Model:
Taxt = E Const + E TobProdPCSqrtTobProdPCt + JEstTax2 EstTaxt2 + et
Ramsey RESET Test
SqrtTobProdPC 0.139110 0.302547 0.459796 0.6477
Const 0.549077 1.751680 0.313457 0.7553
Fitted^2 1.025663 0.924855 1.108999 0.2730
Critical Result: The Fitted^2 coefficient is 1.03. The estimate does not equal 0; the
estimate is 1.03 from 0. This evidence suggests that the new form of
H0: JEstTax2 = 0 New form of the information adds NO explanatory power
H1: JEstTax2 z 0 New form of the information adds explanatory power
Prob[Results IF H0 True] =.2730
We do not reject the null hypothesis at the 10 percent significance level. That is,
we do not reject the notion that the new form of the information adds no
explanatory power. Hence, there is no compelling need to consider a different
specification of the model.
69
5. a.
x The downward sloping theory of demand suggests that higher prices reduce the youth
smoking rate.
x Conventional wisdom suggests that higher incomes reduce the youth smoking rate.
b. SmokeRateYoutht = EConst + EPPriceConsumert + EIIncPCt + et
x EP < 0.
x EI < 0.
c.
PriceConsumer 1.455381 0.665547 2.186743 0.0337
IncPC 0.202018 0.106776 1.891985 0.0645
Const 33.45011 3.651081 9.161700 0.0000
Estimated Equation: EstSmokeRateYouth = 33.45 1.46PriceConsumer

.20IncPC
bP = 1.46: A $1 per pack increase in cigarette prices decreases the youth smoking
rate by 1.46 percentage points.
bI = .20: A 1 thousand dollar increase in per capita income decreases the youth
smoking rate by .20 percentage points.
d. Price Income
H0: EP = 0 H0: EI = 0
H1: E P < 0 H1: EI < 0
e. Prob[Results IF H0 True]:
Price Income
.0337 .0645
| .017 | .032
2 2
We reject both the price and income null hypotheses at the 5 percent significance level,
but not at the 1 percent significance level.
f. Artificial Model: SmokeRateYoutht = EConst + E PPriceConsumert + EIIncPCt
+ JEstYouthRate EstYouthRatet2 + et
Ramsey RESET Test
PriceConsumer 4.059803 5.996945 0.676979 0.5017
IncPC 0.540266 0.809220 0.667638 0.5076
Const 58.23491 99.14360 0.587379 0.5598
Fitted^2 0.103279 0.111604 0.925399 0.3595
70
Critical Result: The Fitted^2 coefficient is .103. The estimate does not equal 0; the
estimate is .103 from 0. This evidence suggests that the new form of
H0: JEstYouthRate = 0 New form of the information adds NO explanatory
power
H1: JEstYouthRate z 0 New form of the information adds explanatory power
We do not reject the null hypothesis at the 1, 5, or 10 percent significance level. That is
we do not reject the notion that the new form of the information adds no explanatory
power. Hence, there is no compelling need to consider a different specification of the
model.
71
Chapter 13: Dummy and Interaction Variables
1.
dSSR
2( y1 bConst ) 2( y2 bConst ) 2( y3 bConst ) 0
dbConst
y1 bConst y2 bConst y3 bConst 0
y1 y2 y3 3bConst
y1 y2 y3
bConst
3
NB: bConst equals the mean of y.
3. a. The annual increase in internet use expressed in percentage terms.

b.
x Year: The internet is an emerging technology; consequently, we would expect users to
increase over time. E Year > 0.
x CapitalHuman: Citizens in nations with more human capital will be able to take better
advantage of the internet; hence, more human capital would increase internet use.
ECapHum > 0.
x CapitalPhysical: Nations with more human capital are better able to operate the
internet; hence, more physical capital would increase internet use. ECapPhy > 0.
x GdpPC: Citizens in nations with higher per capital GDP will be able to afford the
internet; hence, higher per capital GDP would increase internet use. EGDP > 0.
x Auth: Authoritarian nations will discourage internet use because it is difficult for
governments to control its content; hence, more authoritarian nations will decrease
internet use. EAuth < 0.
1. A dummy variable takes on only two values: 0 and 1. The dummy variable divides all the
observations into two disjoint groups.
3. When we use averages to draw conclusions we implicitly assume that the only a single factor
affects the dependent variable. Typically, this is not a realistic assumption. For example,
when we use average salaries for men and women to investigate gender discrimination, we
implicitly assume that an individual’s gender is the only factor affecting his/her salary.
73
1. a. Additional articles should increase a faculty member’s salary: EArt > 0.

b.
Dependent Variable: Salary
SexF1 10584.60 5492.306 1.927168 0.0554
Experience 2435.660 211.0136 11.54267 0.0000
Exper_SexF1 1086.915 397.0372 2.737564 0.0068
Articles 1783.056 839.0052 2.125202 0.0348
Const 37992.42 3862.495 9.836240 0.0000
Estimated Equation: EstSalary = 37,992 + 10,585SexF1 + 2,436Experience

1,087Exper_SexF1 + 1,783Articles
bArt = 1,783: 1 additional article increases a faculty member’s salary by $1,783.
c. H0: EArt = 0
H1: EArt > 0
.0348
d. Prob[Results IF H0 True] = | .017
2
We reject both the published articles null hypotheses at the 5 percent significance
level, but not at the 1 percent significance level.
3. a.
x The seniority system provides long time members of Congress more influence than
newly elected members; hence additional terms should increase the number of
earmarks. E Terms > 0.
x Conventional wisdom suggests that liberals are more supportive of government
projects than conservatives; hence, liberal members of Congress should be more
active in acquiring earmarks. ELiberal > 0.
x Equity suggests that low income states should receive more Federal government
assistance than high income states; hence, high income states should receive fewer
earmarks. E Income < 0.
b. Use the ordinary least squares (OLS) estimation procedure to estimate the coefficients.
Interpret the coefficient estimates.
74
Terms 0.260593 0.068817 3.786763 0.0002
ScoreLiberal 0.052310 0.011216 4.663870 0.0000
IncPC 0.083213 0.047239 1.761533 0.0789
Const 6.445098 2.010177 3.206235 0.0014
Estimated Equation: EstNumber = 6.45 + .261Terms

+ .052ScoreLiberal .083IncPC
bTerms = .261: We estimate that an additional term in Congress increases a member’s
earmarks by .261.
bLiberal = .052: We estimate that a 1 point (on a 0-100 scale) increase in a member
liberal score increases the member’s earmarks by .052.
bIncPC = .083: We estimate that a 1 thousand dollar increase in a state’s per capita
income reduces a member earmarks by .083.
c. Seniority Liberal Equity
H0: ETerms = 0 H0: ELiberal = 0 H0: EIncome = 0
H1: ETerms > 0 H1: ELiberal > 0 H1: EIncome < 0
d. Seniority Liberal Equity
.0002 .0001 .0789
| .0001 | <.0001 | .039
2 2 2
We reject the seniority, liberal, and equity null hypotheses at the 5 percent
significance level. At the 1 percent significance level, we reject the seniority and
liberal null hypotheses, but not equity null hypothesis.
5. a. Allegation: Members of Congress from the Northeast receive more earmarks than their
colleagues from other parts of the country.
ENE > 0.
b.
Terms 0.266513 0.068549 3.887917 0.0001
ScoreLiberal 0.048719 0.011280 4.319022 0.0000
IncPC 0.153651 0.056707 2.709577 0.0070
RegionNorthEast 2.061917 0.927902 2.222127 0.0268
Const 9.300573 2.377948 3.911176 0.0001
Estimated Equation: EstNumber = 9.30 + .267Terms + .049ScoreLiberal

.154IncPC + 2.06RegionNorthEast
75
bNE = 2.06: Members of Congress from the Northeast receive 2.06 more earmarks
than their colleagues from other regions of the nation.
c. H0: ENE = 0
H1: ENE > 0
.0268
d. Prob[Results IF H0 True] = | .013
2
We reject the allegation’s null hypothesis at a 5 percent significance level, but not at the 1
percent significance level.
76
Chapter 14: Omitted Explanatory Variables, Multicollinearity,
and Irrelevant Explanatory Variables
1. Goal of Multiple Regression Analysis: Multiple regression analysis attempts to sort out the
individual effect of each explanatory variable. An explanatory variable’s coefficient estimate
allows us to estimate the change in the dependent variable resulting from a change in that
particular explanatory variable while all other explanatory variables remain constant.
value is small.
5. a. The value of one variable does not allow us to predict the value of the other.
b. We use a scatter diagram plotting each variable’s deviation from its mean illustrates
independence. Loosely speaking, when the variables are independent their scatter
diagram points are spread “evenly” across all four quadrants.
c. It equals 0.
7. a. 1.00
b. Yes, something unusual occurs: EViews error message: “Near singular matrix.”
1. a. The ordinary least squares (OLS) estimation procedure for an included explanatory
variable’s value can be biased.
b. An omitted explanatory variable bias problem arises whenever the omitted explanatory
variable:
x influences the dependent variable;
x is correlated with an included explanatory variable.
3. a. The ordinary least squares (OLS) estimated procedure for the included explanatory
variables’ values will be unbiased.
b. The variance of the probability distributions for the values of the relevant explanatory
variables will be larger when the irrelevant explanatory variable is included than when
the irrelevant variable is excluded.
77
1. a.
x The downward sloping demand curve theory suggests that higher prices decrease per
capita cigarette consumption. E Price < 0.
x Conventional wisdom suggests that college educated individuals smoke less. EEduColl
< 0.
x Tobacco is important to the economy of those states that produce large quantities of
tobacco; hence, tobacco use will be more acceptable in those states and residents will
be more receptive to smoking. ETobProd > 0.
b.
PriceConsumer 11.18450 3.464492 3.228322 0.0023
EducCollege 1.235767 0.580736 2.127934 0.0386
TobProdPC 0.725932 0.310685 2.336554 0.0238
Const 150.4259 17.75788 8.470934 0.0000
Estimated Equation: EstCigConsPC = 150.4 11.2PriceConsumer

1.24EducCollege + .726TobProdPC
bPrice = 11.2: A $1 rise in the price of a pack of cigarettes decreases per capita
cigarette consumption by 11.2 packs per year.
bEduColl = 1.24: A 1 percentage point increase in college educated individuals
decreases per capita cigarette consumption by 1.24 packs per year.
bTobProd = .726: A 1 pound rise in a state’s per capita tobacco production increases per
capita cigarette consumption by .736 packs per year.
c. Demand Education Production
H0: EPrice = 0 H0: EEduColl = 0 H0: ETobProd = 0
H1: EPrice < 0 H1: EEduColl < 0 H1: ETobProd > 0
d. Demand Education Production
.0023 .0386 .0238
| .001 | .019 | .012
2 2 2
x At the 5 percent significance level we reject the demand, education, and
production null hypotheses.
x At the 1 percent significance level, we reject the demand null hypothesis, but not
education and production null hypothesis.
78
3. a.
x The downward sloping demand curve theory suggests that higher prices decrease per
capita cigarette consumption. E Price < 0.
x Conventional wisdom suggests that college educated individuals smoke less. EEduColl
< 0.
x Conventional wisdom suggests that higher income individuals smoke less.
EI < 0.
x Tobacco is important to the economy of those states that produce large quantities of
tobacco; hence, tobacco use will be more acceptable in those states and residents will
be more receptive to smoking. ETobProd > 0.
b.
PriceConsumer 12.30477 3.417315 3.600714 0.0008
EducCollege 2.543609 0.881137 2.886736 0.0059
IncPC 1.582120 0.818423 1.933132 0.0594
TobProdPC 0.737257 0.302072 2.440662 0.0186
Const 134.8625 19.04746 7.080337 0.0000
Estimated Equation: EstCigConsPC = 134.9 12.3PriceConsumer

2.54EducCollege + 1.58IncPC
+ .737TobProdPC
bPrice = 12.3: A $1 rise in the price of a pack of cigarettes decreases per capita
bEduColl = 2.54: A 1 percentage point increase in college educated individuals
decreases per capita cigarette consumption by 2.54 packs per year.
bI = 1.58: A 1 thousand dollar increase in per capita income increases per capita
bTobProd = .737: A 1 pound rise in a state’s per capita tobacco production increases
per capita cigarette consumption by .737 packs per year.
The estimate of the income coefficient is surprising. It is positive rather than negative
thereby contradicting the conventional wisdom.
c. Since the income results refute the theory we only formulate null and alternative
hypotheses for the three other theories.
Demand Education Production
H0: EPrice = 0 H0: EEduColl = 0 H0: ETobProd = 0
H1: EPrice < 0 H1: EEduColl < 0 H1: ETobProd > 0
d. Demand Education Production
.0008 .0059 .0186
| .0004 | .003 | .009
2 2 2
79
x At the 1 percent significance level, we reject the demand, education, and production
null hypotheses.
x Since the income results refute the theory perhaps we should rethink the role of
income.
5. a.
x As a member of Congress has been in the House longer, he/she gains influence.
Consequently, he/she is more able to garner earmarks. ETerms > 0.
x Liberal members of Congress have a more activist view of government’s role than
conservative members. Consequently, members of Congress who have high liberal
ratings should be more active in their pursuit of earmarks. E Liberal > 0.
b.
Terms 0.256432 0.068947 3.719247 0.0002
ScoreLiberal 0.045667 0.010589 4.312564 0.0000
Const 3.122135 0.696198 4.484549 0.0000
Estimated Equation: Number = 3.12 + .256Terms + .046ScoreLiberal

bTerms = .256: An additional term in Congress increases a member’s earmarks by .256.
bLiberal = .046: A 1 point (on a 0-100 scale) increase in a member liberal score
increases the member’s earmarks by .046.
c. Terms Liberal
H0: ETerms = 0 H0: ELiberal = 0
H1: ETerms > 0 H1: ELiberal > 0
d. Terms Liberal
.0002 .0001
| .0001 | <.0001
2 2
x At the 1 percent significance level we reject the terms and liberal null hypotheses.
80
7. a.
x As a member of Congress has been in the House longer, he/she gains influence.
Consequently, he/she is more able to garner earmarks. ETerms > 0.
x Liberal members of Congress have a more activist view of government’s role than
conservative members. Consequently, members of Congress who have high liberal
ratings should be more active in their pursuit of earmarks. E Liberal > 0.
x Democratic members of Congress have a more activist view of government’s role
than their non-Democratic colleagues. Consequently, Democratic members of
Congress should be more active in their pursuit of earmarks. EDem > 0.
b.
Terms 0.259824 0.070022 3.710597 0.0002
ScoreLiberal 0.039769 0.023079 1.723173 0.0856
PartyDem1 0.367293 1.276776 0.287672 0.7737
Const 3.198703 0.746051 4.287513 0.0000
Estimated Equation: EstNumber = 3.20 + .260Terms + .040ScoreLiberal

+ .367PartyDem1
bTerms = .260: An additional term in Congress increases a member’s earmarks by .260.
bLiberal = .040: A 1 point (on a 0–100 scale) increase in a member liberal score
increases the member’s earmarks by .040.
bDem = .367: A Democratic member of Congress increases the number of earmarks by
.367.
c. Terms Liberal Democrat
H0: ETerms = 0 H0: ELiberal = 0 H0: EDem = 0
H1: ETerms > 0 H1: ELiberal > 0 H1: EDem > 0
d. Terms Liberal Democrat
.0002 .0856 .7737
| .0001 | .043 | .387
2 2 2
x At the 1 percent significance level we reject the terms null hypothesis, but we do not
reject the liberal or Democrat null hypotheses.
x At the 5 percent significance level we reject the terms and liberal null hypotheses, but
we do not reject the Democrat null hypothesis.
81
Chapter 15: Other Regression Statistics and Pitfalls
1. a.
Dependent Variable: LogUsersTV
Year 0.022989 0.015903 1.445595 0.1487
CapitalHuman 0.036302 0.001915 18.95567 0.0000
CapitalPhysical 0.001931 0.000510 3.789394 0.0002
GdpPC 0.058877 0.012338 4.772051 0.0000
Auth 0.063345 0.012825 4.939278 0.0000
Const 44.95755 31.77155 1.415025 0.1575
Estimated Equation: EstLogUsersTV = 45.0 + .023Year

+ .036CapitalHuman + .002CapitalPhysical
+ .059GdpPC + .063Auth
b. H0: EYear
TV
= .010
H1: EYear
TV
z .010
A two-tailed test is appropriate.
c. Prob[Results IF H0 True] = Probability that the coefficient estimate would be at least
.033 from .010, if H0 were true (if the actual coefficient
equals, EYear
TV
, .010)
Left Tail Right Tail

p p
Prob[Results IF H0 True] = .0191 + .0191 | .038.
83
3. a. i. Error message: “Near singular matrix.”
ii. 1
iii. No
b. i. Error message: “Near singular matrix.”
ii. No
c. i. Error message: “Near singular matrix.”
ii. No
d.
Dependent Variable: TotalSalary
HomeSalary 1.000000 8.58E-17 1.17E+16 0.0000
VisitSalary 1.000000 8.61E-17 1.16E+16 0.0000
Const 0.000000 4.24E-15 0.000000 1.0000
The standard errors are very small, approximately 0.
e.
Dependent Variable: Attendance
PriceTicket 590.7836 184.7231 3.198211 0.0015
HomeSalary 783.0394 45.23955 17.30874 0.0000
Const 9246.429 1529.658 6.044767 0.0000
Estimated Equation: EstAttendance = 9,246 591PriceTicket

+ 783HomeSalary
Interpretation: bPriceTicket = 591. We estimate that a $1.00 increase in the price of
tickets decreases attendance by 591 per game.
bHomeSalary = 783. We estimate that a $1 million increase in the home
team salary increases attendance by 783 per game.
i. 591.
ii. 20.232.
iii. 20232.
No.
Dependent Variable: Attendance
PriceTicket 1896.379 142.8479 13.27552 0.0000
HomeSalary 0.088467 0.484536 0.182580 0.8552
Const 3697.786 1841.286 2.008263 0.0451
84
Estimated Equation: EstAttendance = 3,698 + 1,896PriceTicket .088HomeSalary
iv.
x The estimate of the ticket price coefficient changes dramatically, from 591 to
1,896.
x The estimate of the home salary coefficient changes dramatically, from 783.0 to
.088.
5. No.
1. Our approach thus far has followed two steps:

x First, we present the theory.
x Second, we analyze the data and determine whether the data are consistent with the
theory.
We have started with the theory and then decide whether the data is compatible with the
theory.
The confidence interval approach reverses this process. Confidence intervals indicate the
range of theories with which the data are compatible.
x First, we analyze the data.
x Second, we consider the theories and determine which theories are consistent
with the data.
We start with the data and then decide what theories are compatible with the data.
3. Goal of Multiple Regression Analysis: Multiple regression analysis attempts to sort out the
individual effect of each explanatory variable. An explanatory variable’s coefficient estimate
allows us to estimate the change in the dependent variable resulting from a change in that
particular explanatory variable while all other explanatory variables remain constant.
1. a.
Dependent Variable: LogUsersInternet
Year 0.449654 0.017078 26.32965 0.0000
CapitalHuman 0.023725 0.002470 9.606597 0.0000
GdpPC 0.118177 0.011461 10.31146 0.0000
Auth 0.095836 0.013999 –6.845761 0.0000
Const 899.3201 34.17432 –26.31567 0.0000
85
Estimated Equation: EstLogUsersInternet = 899.3 + .450Year +
.024CapitalHuman + .002CapitalPhysical + .118GdpPC
.096Auth
b. EYear
LB
.4161
EUB
Year .4832
The 95 percent confidence interval is from .4161 to .4832.
3. a.
PriceReal 124.7236 77.09510 1.617789 0.1444
Const 609.9538 58.12769 10.49334 0.0000
Number of Observations 10 Mass1 = 1
Estimated Equation: EstPetroConsPCMass = 610.0 124.7PriceReal

b. 610.0
c. 124.7
5. a. i. EstPetroConsPCMass = bMass + bPMassPriceReal

ii. bMass
iii. bPMass
b. i. EstPetroConsPCNeb = bNeb + bPNebPriceReal
ii. bNeb
iii. bPNeb
c. bMass = 610.0 bPMass = 124.7 bNeb = 1,418.4 bPNeb = 540.5
d.
Mass1 609.9538 143.2589 4.257704 0.0006
Neb 1418.403 106.8526 13.27438 0.0000
PriceReal_Mass 124.7236 190.0051 0.656422 0.5209
PriceReal_Neb 540.5367 150.1664 3.599585 0.0024
Estimated Equation: EstPetroConsPC = 610.0Mass1 + 1,418.4Neb

124.7PriceReal_Mass 540.5PriceReal_Neb
e. Yes.
7. a. i. EstPetroConsPCMass = bConst + bMass + (bP + bPMass)PriceReal

ii. bMass
iii. bP + bPMass
86
b. i. EstPetroConsPCNeb = bConst + bNeb + (bP + bPNeb)PriceReal
ii. bNeb
iii. bP + bPNeb
c. No, we have 5 unknowns, but only 4 equations:
bMass = 610.0 bP + bPMass = 124.7
bNeb = 1,418.4 bP + bPNeb = 540.5
d. Error message: Near singular matrix
e. Yes
87
Chapter 16: Heteroskedasticity
1.
x Error Term Equal Variance Premise: The variance of the error term’s
probability distribution for each observation is the same; all the variances equal
Var[e]:
Var[e1] = Var[e2] = … = Var[eT] = Var[e]
x Error Term/Error Term Independence Premise: The error terms are
independent: Cov[ei, ej] = 0.
Knowing the value of the error term from one observation does not help you
predict the value of the error term for any other observation.
x Explanatory Variable/Error Term Independence Premise: The explanatory
variables, the xt’s, and the error terms, the et’s, are not correlated.
Knowing the value of an observation’s explanatory variable does not help
you predict the value of that observation’s error term.
3. The error term equal variance premise played an important role. If this premise were violated
the variance would differ from observation to observation. Var[e] would not exist. The
expression for Var[bx] would be much more complicated.
5. Model: LogUsersInternett = EConst + EGDPGdpPCt + et

Divide both sides of the equation by
GdpPCt .
LogUsersInternett EConst GdpPCt et
Tweaked Model : EGDP
GdpPCt GdpPCt GdpPCt GdpPCt
EConst
EGDP GdpPCt Ht
GdpPCt
et
where H t
GdpPCt
89
et
Var[εt ] Var[ ]
GdpPCt
p Arithmetic of variances: Var[cx ] c 2 Var[x ]
1
Var[et ]
GdpPCt
p Var[et ] =V u GpcPCt where V equals a constant
1
V u GpcPCt
GdpPCt
V
1. Heteroskedasticity occurs whenever the variance of the error terms’ probability distributions
is not identical; that is, whenever the variance of the error terms’ probability distributions
differ from observation to observation.
3. a. The ordinary least squares (OLS) estimation procedure for the coefficient value is
unbiased.
b.
x The ordinary least squares (OLS) estimation procedure for the variance of the error
term’s probability distribution is flawed because it is based on a false premise; hence,
the standard errors and call calculations based on the standard errors are flawed.
x The ordinary least squares (OLS) estimation procedure is not the best linear unbiased
estimation procedure (BLUE).
1. a.
x A state with a higher crime rate should have more court cases and hence would spend
more on its judicial system: ECrimes > 0.
x A state with a higher per capita GDP has a larger tax base and hence, can afford to
spend more on its judicial system: EGDP > 0.
b.
Dependent Variable: JudExp
CrimesAll 0.321427 0.408012 0.787787 0.4348
GdpPC 0.298469 0.063159 4.725673 0.0000
Const –2425.185 2705.168 –0.896501 0.3746
Estimated Equation: EstJudExp = 2,425 + .321CrimesAll + .298GdpPC
90
bCrimesAll = .321: A 1 per 100,000 persons rise in the crime rate increases judicial
expenditures by $.321 per 100,000 persons.
bGDP = .298: A $1 rise in per capita state income increases judicial expenditures by
.298 per 100,000 persons.
Critical Result: CrimesAll coefficient estimate equals .321. This evidence, the positive
sign of the coefficient estimate, suggests that higher crime rates
increase judicial expenditures thereby supporting the theory.
GdpPC coefficient estimate equals .298. This evidence, the positive
sign of the coefficient estimate, suggests that higher per capita GDP
increases judicial expenditures thereby supporting the theory.
c. Crime Rate GDP
H0: ECrimes = 0 H0: EGDP = 0
H1: ECrimes > 0 H1: EGDP > 0
d. Crime Rate GDP
.4348 .0001
| .217 | < .0001
2 2
x At the 1 percent significance level we reject the GDP null hypothesis.
x Even at a 10 percent significance level we cannot reject the crime rate null
hypothesis.
3. a.
Original Model: JudExpt = EConst + ECrimesCrimesAllt + EGDPGdpPCt + et
Divide both sides of the equation by GdpPCt
.
JudExpt
Tweaked Model :
GdpPCt
EConst CrimesAllt GdpPCt et
ECrimes EGDP
GdpPCt GdpPCt GdpPCt GdpPCt
EConst CrimesAllt
ECrimes EGDP GdpPCt Ht
GdpPCt GdpPCt
et
where H t
GdpPCt
91
et
Var[εt ] Var[ ]
GdpPCt
p Arithmetic of variances: Var[cx] c 2 Var[x]
1
Var[et ]
GdpPCt
p Var[et ] =V u GpcPCt where V equals a constant
1
V u GpcPCt
GdpPCt
V
b.
Dependent Variable: AdjJudExp
AdjCrimesAll 0.308650 0.375938 0.821013 0.4158
AdjGdpPC 0.299664 0.062667 4.781868 0.0000
AdjConst 2413.696 2517.482 0.958774 0.3426
5. a.
x A state with a higher unemployment rate should increase its poverty rate: EUnem > 0.
x A state with a higher burglary rate should depress economic conditions increasing the
poverty rate: EBurg > 0.
b.
Dependent Variable: PovRate
UnemRate 0.446262 0.380817 1.171854 0.2472
Burglaries 0.006463 0.001780 3.630155 0.0007
Const 4.949594 2.049562 2.414952 0.0197
Estimated Equation: EstPovRate = 4.95 + .446UnemRate + .00646Burglaries

bUnem = .446: A 1 percentage point rise in the unemployment rate increases the
poverty rate by .446 percentage points.
bBurglaries = .00646: A 1 per 100,000 persons rise in the burglary rate increases the
poverty rate by .00646 percentage points. A 100 per 100,000 persons rise in the
burglary rate increases the poverty rate by .646 percentage points.
92
Critical Result: Unem coefficient estimate equals .446. This evidence, the positive sign
of the coefficient estimate, suggests that higher unemployment rates
increase the poverty rate thereby supporting the theory.
Burglaries coefficient estimate equals .00646. This evidence, the
positive sign of the coefficient estimate, suggests that higher a higher
burglary rate increases the poverty rate thereby supporting the theory.
c. UnemRate Burglaries
H0: EUnem = 0 H0: EBurg = 0
H1: EUnem > 0 H1: EBurg > 0
d. UnemRate Burglaries
.2472 .0007
| .124 | .0004
2 2
x At the 1 percent significance level we reject the Burglaries null hypothesis.
x Even at a 10 percent significance level we cannot reject the unemployment rate null
hypothesis.
7. a.
Original Model: PovRatet = EConst + EUnemUnemRatet + E BurgBurglariest + et
Multiply both sides of the equation by Popt .
Tweaked Model : PovRatet Popt
EConst Popt EUnemUnemRatet Popt E Burg Burglariest Popt et Popt
EConst Popt EUnemUnemRatet Popt E Burg Burglariest Popt Ht
where H t et Popt
Var[εt ] Var ª¬et Popt º¼

p Arithmetic of variances: Var[cx] c 2 Var[x]
Popt Var[et ]
V
p Var[et ] = where V equals a constant
Popt
V
Popt
Popt
V
93
b.
Dependent Variable: AdjPovRate
AdjUnemRate 0.814434 0.331685 2.455442 0.0178
AdjBurglaries 0.005746 0.001291 4.451748 0.0001
AdjConst 3.059615 2.097694 1.458561 0.1513
94
Chapter 17: Autocorrelation (Serial Correlation)
1.
x Error Term Equal Variance Premise: The variance of the error term’s
probability distribution for each observation is the same; all the variances equal
Var[e]:
Var[e1] = Var[e2] = … = Var[eT] = Var[e]
x Error Term/Error Term Independence Premise: The error terms are
independent: Cov[ei, ej] = 0.
Knowing the value of the error term from one observation does not help you
predict the value of the error term for any other observation.
x Explanatory Variable/Error Term Independence Premise: The explanatory
variables, the xt’s, and the error terms, the et’s, are not correlated.
Knowing the value of an observation’s explanatory variable does not help
you predict the value of that observation’s error term.
3. The error term/error term independence premise played an important role. If this premise
were violated, the error term covariance terms would not equal 0. The expression for Var[bx]
would be much more complicated.
5. a. The value of one variable does not allow us to predict the value of the other.
b. We use a scatter diagram plotting each variable’s deviation from its mean illustrates
independence. Loosely speaking, when the variables are independent that scatter diagram
points are spread “evenly” across all four quadrants.
c. Both are approximately 0.
7. Rest = yt Estyt
Substituting for yt
yt = EConst + Exxt + et
= EConst + Exxt + et Estyt

Substituting for et
et = Uet1 + vt
= EConst + Exxt + Uet1 + vt Estyt
Substituting for Estyt
Estyt = bConst + bxxt
= EConst + Exxt + Uet1 + vt (bConst + bxxt)
Rearranging terms
= (EConstbConst) + (Exbx)xt + Uet1 + vt
95
1. Autocorrelation exists whenever one observation’s error term is correlated with the previous
observation’s error term.
3. a. The ordinary least squares (OLS) estimation procedure for the coefficient value is
unbiased.
b.
x The ordinary least squares (OLS) estimation procedure for the variance of the error
term’s probability distribution is flawed because it is based on a false premise; hence,
the standard errors and call calculations based on the standard errors are flawed.
x The ordinary least squares (OLS) estimation procedure is not the best linear unbiased
estimation procedure (BLUE).
1. a. The downward sloping theory of demand postulates that a rise in the real price of
petroleum should decrease petroleum consumption. E P < 0.
b.
PriceReal 53.63777 18.00446 2.979138 0.0054
Const 563.4323 16.26989 34.63038 0.0000
Estimated Equation: EstPetroConsPC = 563.4 53.6PriceReal

bP = 53.6: A $1 increase in the real price of petroleum decreases per capital
petroleum consumption by 53.6 gallons.
Critical Result: Price coefficient estimate equals 53.6. This evidence, the negative
sign of the coefficient estimate, suggests that higher petroleum prices
reduce petroleum consumption.
c. H0: E P = 0
H1: E P < 0
.0054
d. | .003
2
At the 1 percent significance level we reject the null hypothesis.
96
3. a. Autocorrelation model:
et = Uet1 + vt vt’s are independent
or
vt = et Uet1
Start with the original model:
PetroConsPCt = EConst + E I PriceRealt + et
Lag it one year and multiply by U:
U PetroConsPCt–1 = UEConst + EIU PriceRealt–1 + Uet–1
Subtract
PetroConsPCt U PetroConsPCt–1 = (1U)EConst + EI(PriceRealt UPriceRealt–1)
+ et Uet–1
Substituting vt for et Uet–1:
PetroConsPCt U PetroConsPCt–1 = (1U)EConst + EI(PriceRealt UPriceRealt–1) + vt
vt’s are independent
b. Use the estimated value of U to generate two new variables:
AdjPetroConsPC = PetroConsPC .77PetroConsPC(1)
AdjPriceReal = PriceReal .77PriceReal(1)

Dependent Variable: AdjPetroConsPC
AdjPriceReal 47.58822 21.48296 2.215161 0.0340
Const 131.0316 4.954810 26.44533 0.0000
5. a.
x An increase in per capita income should reduce the crime rate. EI < 0.
x An increase in the unemployment rate should increase the crime rate. EUn > 0.
b.
Dependent Variable: CrimesAll
IncPC 0.693721 0.116784 5.940206 0.0000
UnemRate 114.1572 111.4697 1.024109 0.3201
Const 14897.77 2369.554 6.287163 0.0000
Estimated Equation: CrimesAll = 14,898 .694IncPC + 114.2UnemRate

bI = .694: A $1 increase in per capita income decreases the crime rate by .694 per
100,000.
bUn = 114.2: A 1 percentage point increase in the unemployment rate increases the
crime rate by 114.2 per 100,000.
97
Critical Result: IncPC coefficient estimate equals .694. This evidence, the negative
sign of the coefficient estimate, suggests that higher per capita income
reduces crime thereby supporting the theory.
UnemRate coefficient estimate equals 114.2. This evidence, the
positive sign of the coefficient estimate, suggests that higher
unemployment rates increase crime thereby supporting the theory.
c. Income Unemployment
H0: E I = 0 H0: EUn = 0
H1: E I < 0 H1: EUn > 0
.0001 .3201
d. | < .0001 | .160
2 2
x At the 1 percent significance level we reject the income null hypothesis.
x Even at the 10 percent significance level we do not reject the unemployment
null hypothesis.
7. a. Autocorrelation model:
et = Uet1 + vt vt’s are independent
or
vt = et Uet1
Start with the original model:
CrimesAllt = EConst + EIIncPCt + EUnUnemRatet + et
Lag it one year and multiply by U:
UCrimesAllt–1 = UEConst + EIUIncPCt–1 + EUnUUnemRatet–1 + Uet–1
Subtract
CrimesAllt UCrimesAllt–1 = (1U)EConst + E I(IncPCt UIncPCt–1) + EUn(UnemRatet
UUnemRatet–1) + et Uet–1
Substituting vt for et Uet–1:
CrimesAllt UCrimesAllt–1 = (1U)EConst + E I(IncPCt UIncPCt–1)
+ EUn(UnemRatet UUnemRatet–1) + vt
vt’s are independent
b. Use the estimated value of U to generate three new variables:
AdjCrimesAll = CrimesAll .79CrimesAll(1)
AdjIncPC = IncPC .79IncPC(1)
AdjUnemRate = UnemRate .79UnemRate (1)
Dependent Variable: AdjCrimesAll
AdjIncPC 0.324473 0.122196 2.655344 0.0173
AdjUnemRate 144.7625 55.71745 2.598153 0.0194
AdjConst 1747.274 467.3090 3.739012 0.0018
98
Chapter 18: Explanatory Variable/Error Term Independence
Premise, Consistency, and Instrumental Variables
1. a. i.
ii. Decrease.
iii. Negatively.
b. i.
ii. Observationxt et yt
1 2 4 11
2 6 2 11
3 10 3 14
4 14 2 11
5 18 1 14
6 22 4 13
iii. Plot the x and y values for each of the 6
observations.
c. See above.
d. The best fitting line is less steeply sloped than the actual line.
1. a. Upward bias.
b. Downward bias.
c. Unbiased.
3. Only the mean of the estimate’s probability distribution is important. An estimation

procedure is unbiased if and only if the mean of the estimate’s probability distribution equals
the actual value:
Mean[Est] = Actual Value
5. No. An unbiased estimation procedure will not be consistent whenever the variance of the
estimate’s probability distribution remains constant or increases as the sample size increases.
7. The instrument and the

x “problem” explanatory variable must be correlated.
x error term must be independent.
99
1. a. .50
b. .0156
c. 32 percent
3. Random Sampling Procedure
5. Comparing the answers:

x The unbiased procedure results in 32 percent of the repetitions being “close to” the
actual value when the sample size is 16.
x The biased procedure results in 36 percent of the repetitions being “close to” the
actual value when the sample size is 25.
An unbiased procedure is not always better than a biased procedure. When
x the magnitude of the bias for the biased procedure is sufficiently small
and
x the variance of the biased procedure is smaller than the variance of the unbiased
procedure
The biased procedure can produce more reliable estimates in the sense that a greater
proportion of the estimates will be close to the actual value.
100
Chapter 19: Measurement Error and the Instrumental
Variables Estimation Procedure
1. a. Sometimes you will be a little premature in clicking the stop watch button. Other times
you will be a little late. It is humanly impossible to measure the actual elapsed time
perfectly. No matter how careful you are, sometimes the measured value will be a little
low and other times a little high.
b. i. Half
ii. Half
iii. 0
1. a. The ordinary least squares (OLS) estimation procedure for the coefficient value will still
be unbiased.
b. The variance of the coefficient estimate’s probability distribution will be greater as a
consequence of dependent variable measurement error. This occurs because measurement
error introduces more “more uncertainty” into the mix.
3. a. No. The ordinary least squares (OLS) estimated procedure for the coefficient value is
biased when measurement error is present (and the actual value of the coefficient is
nonzero).
b. While the instrumental variable (IV) estimation procedure for the coefficient value is still
biased, it is consistent.
1. Explanatory variable measurement error would be introduced. The ordinary least squares
(OLS) estimation procedure for the value of the SATScores coefficient would be biased
toward 0 if SAT scores do have an impact on exam scores.
3. Dependent variable measurement error would be introduced. The ordinary least squares
(OLS) estimation procedure
x for the value of the ProbScores and SATScores coefficients would not be biased.
But
x the variance of the estimates’ probability distributions would be greater as a
consequence of the randomization.
101
5. a.
Dependent Variable: IncAnnPC
Coll 0.783234 0.077323 10.12934 0.0000
Const 12.76024 2.140549 5.961198 0.0000
Estimated Equation: EstIncAnnPC = 12.76 + .783Coll

b.
Dependent Variable: Covered
EstIncAnnPC 0.319072 0.125334 2.545781 0.0141
Const 75.42874 4.296441 17.55610 0.0000
Estimated Equation: EstCovered = 75.43 + .32EstIncAnnPC

bEstIncAnnPC = .32: A $1,000 increase in permanent per capita disposable income
increases the state’s health coverage by .32 percentage points.
Critical Result: The EstIncAnnPC coefficient estimate equals .32. This evidence, the
positive sign of the coefficient estimate, suggests that increases in
permanent disposable income increase health insurance coverage
thereby supporting the theory.
7. a.
x A state with a higher crime rate should have more court case and hence would spend
more on its judicial system: ECrimes > 0.
x A state with a higher per capita GDP has a larger tax base and hence, can afford to
spend more on its judicial system: EGDP > 0.
b.

Dependent Variable: JudExp
CrimesAll 0.321427 0.408012 0.787787 0.4348
GdpPC 0.298469 0.063159 4.725673 0.0000
Const 2425.185 2705.168 0.896501 0.3746
Estimated Equation: EstJudExp = 2,425 + .321CrimesAll + .298GdpPC

bCrimesAll = .321: A 1 per 100,000 persons rise in the crime rate increases judicial
102
expenditures by $.321 per 100,000 persons.
bGDP = .298: A $1 rise in per capita state income increases judicial expenditures by
$.298 per 100,000 persons.
Critical Result: CrimeAll coefficient estimate equals .321. This evidence, the positive
sign of the coefficient estimate, suggests that higher crime rates
increase judicial expenditures thereby supporting the theory.
GdpPC coefficient estimate equals .298. This evidence, the positive
sign of the coefficient estimate, suggests that higher per capita GDP
increases judicial expenditures thereby supporting the theory.
9.
x First, we used annual disposable income as an explanatory variable and applied the
ordinary least squares (OLS) estimation procedure. We estimated that a 1 per 100,000
persons rise in crime rate increases judicial expenses by $.321 per 100,000 persons. But
we believe that an explanatory variable measurement error problem is present here.
x Second, we used an instrumental variable (IV) approach which resulted in a higher
estimate for the impact of permanent income. We estimated that a 1 per 100,000 persons
rise in crime rate increases judicial expenses by $5.12 per 100,000 persons.
103
Chapter 20: Omitted Variables and the Instrumental Variable
Estimation Procedure
1. a.
yt = EConst + Ex1x1t + Ex2 x2t + et
= EConst + Ex1x1t + (Ex2x2t + et)
= EConst + Ex1x1t + Ht
Ht = Ex2x2t + et
b. Yes, the explanatory variable, x1t, and the second model’s error term, Ht, are positively
correlated.
x1t and x2t
positively correlated
Included Typically, omitted
o
variable x1t up variable x2t up
p p Ht = Ex2x2t + et
x1t up Ht up Ex2 > 0
x1t and Ht positively correlated

c. Since the explanatory variable, x1t, and the error term, Ht, are positively correlated, the
ordinary least squares (OLS) estimation procedure for the value of the coefficient
estimate, Ex1, will be biased upward.
3. .60
1. The instrument and the

x “problem” explanatory variable must be correlated.
x error term must be independent.
1. a. Since George W. Bush, a Republican, was the incumbent President, a rising

unemployment should hurt the Republicans and increase the votes received by the
Democrats.
b. EUnTrend > 0.
105
3.
x Instrument/”Problem” Explanatory Variable Correlation: The instrument,
AdvDegt, must be correlated with the “problem” explanatory variable, PopDent.
x Instrument/Error Term Independence: The instrument, AdvDegt, and the error
term, Ht, must be independent.
5.
x Instrument/”Problem” Explanatory Variable Correlation: The correlation
coefficient of AdvDegt and PopDent equals .77. This along with the results from
IV Regression 1 suggests that this condition is satisfied.
x Instrument/Error Term Independence: We cannot determine if this condition is
met.
7.
VoteDemPartyTwot = EConst + EGdpGth RealGdpGrowtht
+ EPopDenPopDent + ELibLiberalt + et
+ (ELibLiberalt + et)
+ Ht
where Ht = ELibLiberalt + et
PopDent and Liberalt

positively correlated
Included Typically, omitted
variable o variable
PopDent up Liberalt up
p Ht = E PopDen Liberalt + et
Ht up ELib > 0
PopDent and Ht positively correlated

p
Ordinary least squares (OLS) estimation
procedure for the value of the coefficient will be
biased upward.
106
9. a.
Dependent Variable: PopDen
AdvDeg 314.3409 37.54406 8.372587 0.0000
Const –2748.974 393.1890 –6.991482 0.0000
Estimated Equation: EstPopDen = 2,749.0 + 314.3AdvDeg
b.
Dependent Variable: VoteDemPartyTwo
RealGdpGrowth 0.867041 0.416006 2.084205 0.0425
EstPopDen 0.008464 0.000957 8.844707 0.0000
Const 49.19420 1.049180 46.88823 0.0000
Estimated Equation: VoteDemPartyTwo = 49.2 .87 RealGdpGrowth

+ .008EstPopDen
bGdpGth = .87: A 1 percentage point rise in the real GDP growth rate decreases the
vote of the Democrats by .87 percentage points
bEstPopDen = .008: A 1 person increase in a state’s population density increases the
state’s Democratic vote by .008 percentage points; a 100 person increase in a state’s
population density increases the state’s Democratic vote by .8 percentage points.
107
Chapter 21: Panel Data and Omitted Variables
1. Fill in each of the following blanks with a Yes or a No:

OLS Bias Question: Is the explanatory Violated:
Satisfied: Explanatory
variable/error term independence premise Explanatory Variable
Variable and Error
satisfied or violated? and Error Term
Term Independent
Correlated
Is the OLS estimation procedure ~ ~
for the value of the coefficient unbiased p p
or biased? Unbiased Biased
OLS Reliability Question: Are the error Satisfied Violated

term equal variance and error term/error ~ ~
term independence premises satisfied or ~ ~
violated?
p p
Can the OLS calculations for the
Yes No
standard error be “trusted?”
Is the OLS estimation procedure for the

Yes No
value of the coefficient BLUE?
1. For each cross section entity, the value of the omitted variable does not vary from time period
to time period.
3. For each period, the value of the omitted variable does not vary from cross section entity to
cross section entity.
5. With heteroskedasticity and autocorrelation we exploited the relationship among the error
terms to improve the estimation procedure. The random effects approach uses the same
strategy.
109
1. a.
x The downward sloping demand curve theory suggests that a rise in the price of beer
decreases per capita beer consumption. E P < 0.
x Conventional wisdom suggests that beer is an inferior good; as per capita income
rises, per capita beer consumption decreases. EI < 0.
b.
Dependent Variable: BeerPC
Price 80.51845 304.6151 0.264329 0.7916
IncPC 2.414644 1.109782 2.175782 0.0301
Const 394.2818 122.3245 3.223244 0.0014
Cross Sections 51
Periods 9
Estimated Equation: EstBeerPC = 394.3 80.5Price 2.4IncPC

bP = 80.5: A $1 increase in the real price decreases per capita consumption by 80.5
cans per year.
bI = 2.4: A $1,000 increase in real per capita disposable income decreases per capita
consumption by 2.4 cans per year.
Critical Result: The Price coefficient estimate equals 80.5. This evidence, the
negative sign of the coefficient estimate, suggests that higher beer
prices decrease per capita beer consumption thereby supporting the
theory.
The IncPC coefficient estimate equals 2.4. This evidence, the
negative sign of the coefficient estimate, suggests that higher per
capita income decreases per capita beer consumption thereby
supporting the theory.
c. Price Disposable Income
H0: E P = 0 H0: E I = 0
H1: E P < 0 H1: E I < 0
d. Price Disposable Income
.7916 .0301
| .3958 | .015
2 2
x Even at the 10 percent significance level we cannot reject the price null
hypotheses that price has no impact on beer consumption.
x At the 5 percent significance level, we reject the null hypothesis that
disposable income has no impact on beer consumption. On the other hand, at
the 1 percent significance level, we do not reject the null hypothesis.
110
3.
Dependent Variable: LogUsersInternet
Year 0.449654 0.017078 26.32965 0.0000
CapitalHuman 0.023725 0.002470 9.606597 0.0000
GdpPC 0.118177 0.011461 10.31146 0.0000
Auth 0.095836 0.013999 –6.845761 0.0000
Const 899.3201 34.17432 –26.31567 0.0000
Cross Sections 110
Periods 8
The coefficient estimate of Year is positive suggesting that per capita Internet use is
growing over time after accounting for human capital, physical capital, per capita GDP,
and the political environment. More specifically, we estimate that per capita Internet use
is growing by 45 percent annually after accounting for the other explanatory factors.
5. a. The downward sloping demand curve theory suggests that a rise in the price of gasoline
decreases per capita motor fuel consumption.
b.
Dependent Variable: MotorFuelPC
Price 50.72342 21.87900 2.318361 0.0224
Const 509.3529 21.98667 23.16644 0.0000
Cross Sections 3
Periods 35
Estimated Equation: EstMotorFuelPC = 509.4 50.7Price

bP = 50.7: A $1 increase in the real price decreases per capita consumption by 50.7
gallons per year.
Critical Result: The Price coefficient estimate equals 50.7. This evidence, the
negative sign of the coefficient estimate, suggests that higher gasoline
prices decrease per capita motor fuel consumption thereby supporting
the theory.
c. H0: E P = 0
H1: E P < 0
.0224
d. | .011
2
111
At the 5 percent significance level, we reject the null hypothesis that the real price of
gasoline has no impact on beer consumption. On the other hand, at the 1 percent
significance level, we do not reject the null hypothesis.
e.
The first 35 observations are for Arkansas, the second 35 observations for Massachusetts,
and the last 35 observations for Washington. The residuals for
x Arkansas are positive.
x Massachusetts are negative.
x Washington approximately zero.
This suggests that the inclusion of cross section random effects may be
appropriate.
112
Chapter 22: Simultaneous Equations Models—Introduction
1. See Appendix 22.1.
3. a. Multiple regression analysis attempts to sort out the individual effect that each
explanatory variable has on the dependent variable.
b. Each explanatory variable’s coefficient reveals the individual impact that the explanatory
variable has on the dependent variable; that is, each explanatory variable’s coefficient
tells us how changes in that explanatory variable affect the dependent variable while all
other explanatory variables remain constant.
c. 'y = bx1'Vbl1
d. 'y = bx2'Vbl2
e. 'y = bx1'Vbl1 + bx2'Vbl2
5. a. P = 6 + .01Inc + .4FeedP
= 6 + .01u4,000 + .4u40
= 6 + 40 + 16
= 50
Q = 160,000 + 500Inc 4,000FeedP

= 160,000 + 50u4,000 4,000u40
= 160,000 + 200,000 160,000
= 200,000
b. On a sheet of graph paper, plot the demand and supply
curves to illustrate the equilibrium.
1. Endogenous variables are variables whose values are determined “within” the model.
3. a. No.
b. No.
5. a. No.
b. Yes.
113
1. a. bID estimates how the quantity of beef demanded changes when income changes while
the price of beef remains constant.
QD
bID while P remains constant
'I
'QD = bID 'I while the price of beef remains constant
b. i. 'P = 1.0562'FeedP + .018825'Inc
ii. 0 = 1.0562'FeedP + .018825'Inc
Solving for 'FeedP
1.0562'FeedP = .018825'Inc
.018825
'FeedP = 'Inc
1.0562
= .017233'Inc
c. i. 'Q = 332.00'FeedP + 17.347'Inc
ii. From b2: 'FeedP = .017233'Inc
= 332.00u(.017233'Inc) + 17.347'Inc
= 5.917'Inc + 17.347'Inc
= 23.264'Inc
'Q
23.264 while P remains constant
'Inc
bID = 23.264
3.
log(Qt ) DConst
Q
D FP
Q
log( FeedPt ) D IQ log( Inct ) H tQ
log( Pt ) D Const
P
D FP
P
log( FeedPt ) D IP log( Inct ) H tP
5. a. Own Price Elasticity of Demand = bPD

bPD estimates the own price elasticity of demand, the percent change in quantity
demanded resulting from a 1 percent change in the price while income remains constant.
'LogQ D
bPD while LogInc remains constant
'LogP
While LogInc remains constant: 'LogInc = 0
'LogQ = .08452'LogFeedP + .3595'LogInc = .08452'LogFeedP
'LogP = .3317'LogFeedP + .7503'LogInc = .3317'LogFeedP
'LogQ D .0852'LogFeedP .0852
bPD .255
'LogP .3317'LogFeedP .3317
Interpretation: We estimate that a 1 percent increase in the price of beef decreases the
quantity of beef demanded by .225 percent.
114
b. Own Price Elasticity of Supply = bPS
bPS estimates the own price elasticity of supply, the percent change in quantity supplied
resulting from a 1 percent change in the price while the feed price remains constant.
'LogQ S
bPS while LogFeedP remains constant
'LogP
While LogFeedP remains constant: 'LogFeedP = 0
'LogQ = .08452'LogFeedP + .3595'LogInc = .3595'LogInc
'LogP = .3317'LogFeedP + .7503'LogInc = .7503'LogInc
S 'LogQ S .3595'LogInc .3595
b .479
'LogP .7503'LogInc
P
.7503
Interpretation: We estimate that a 1 percent increase in the price of beef increases the
quantity of beef supplieded by .479 percent.
7.
Dependent Variable: Q
FeedP 533.6727 139.9106 3.814384 0.0003
Inc 10.77606 1.628569 6.616890 0.0000
Const 54222.01 12057.98 4.496775 0.0000
Estimated Equation: EstQ = 54,222 533.67FeedP + 10.776Inc

Q
aFP = 533.67: A 1 cent increase in the price of cattle feed decreases the quantity of
beef by 533.67 million pounds.
aIQ = 10.776: A 1 billion dollar increase in real disposable income increases the
quantity of beef by 10.776 million pounds.

Dependent Variable: P
FeedP 1.342682 0.126652 10.60134 0.0000
Inc 0.010833 0.001474 7.347909 0.0000
Const 39.19066 10.91532 3.590428 0.0006
Estimated Equation: EstP = 39.19 + 1.3427FeedP + .010833Inc
115
P
aFP = 1.3427: A 1 cent increase in the price of cattle feed increases the price of beef
by 1.3427 cents.
aIP = .010833: A 1 billion dollar increase in real disposable income increases the price
of beef by .010833 cents.
a. 'Q = 533.67'FeedP + 10.776'Inc
b. 'P = 1.3427'FeedP + .010833'Inc
9. a. An increase in police expenditures provides the populace with more police protection
which decreases the crime rate: E PC < 0.
b. An increase in the unemployment rate increases the chances that crimes will be
committed: EUC > 0.
11.
Crimest DConst
c
DUCUnemRatet D GC GdpPCt H tC
PoliceExpt DConst
P
DUPUnemRatet D GP GdpPCt H tP
13. a. bPC estimates how the crime rate changes when police expenditures change while the
unemployment rate remains constant.
'Crimes
bPC whileUnemRate remains constant
'PoliceExp
b. While UnemRate remains constant, 'UnemRate = 0:
'Crimes = 23.814'UnemRate .025271'GdpPC = .025271'GdpPC
'PoliceExp = 12.174'UnemRate + .019163'GdpPC = .019163'GdpPC
'Crimes .025271'GdpPC .025271
bPC 1.319
'PoliceExp .019163'GdpPC .019163
116
Chapter 23: Simultaneous Equations Models—Identification
1.
Dependent Variable: P
FeedP 1.056242 0.286474 3.687044 0.0003
Inc 0.018825 0.005019 3.750636 0.0003
Const 33.02715 31.04243 1.063936 0.2895
Estimated Equation: EstP = 33.027 + 1.0562FeedP + .018825Inc
3. a.
Dependent Variable: Q
EstP 921.4783 113.2551 8.136309 0.0000
FeedP –1305.262 121.2969 –10.76089 0.0000
Const 108291.8 16739.33 6.469303 0.0000
Estimated Equation: EstQ = 108,292 921.5EstP 1,305Inc

b. They are identical.
1. A model is underidentified whenever we cannot estimate all the parameters of the original
model using the reduced form (RF) estimation procedure.
3. a. The two procedures provide identical results whenever the model is identified or
underidentified.
b. The two procedures provide different results whenever the model is overidentified. The
reduced form (RF) estimation provides multiple estimates for one or more of original
model’s parameters; two stage least squares (TSLS) provides only a single estimate.
117
1. a.
LogFeedP 0.178585 0.027126 6.583483 0.0000
Const 12.89211 0.101850 126.5791 0.0000
Q
Estimate: aFP = .1786
'LogQ = .1786'LogFeedP

Dependent Variable: LogP
LogFeedP 0.135433 0.074912 1.807907 0.0732
Const 4.586163 0.281270 16.30522 0.0000
P
Estimate: aFP = .1354
'LogP = .1354'LogFeedP
i. Yes. We can still estimate the own price elasticity of demand because the feed price
data “tells us” when the supply curve shifts allowing us to use the equilibria to
estimate the slope of the demand curve.
bPD 1.319
ii. No. Without chicken price and income data we do not know when the demand curve
shifts. We cannot use the equilibria to estimate the slope of the supply curve.
b. i. Yes. There is one estimate: 1.319.
Two Stage Least Squares (TSLS)
Instrument(s): LogFeedP
LogP 1.318616 0.669960 1.968203 0.0514
Const 18.93950 3.412885 5.549410 0.0000
Estimate bPS = 1.319

ii. No.
118
3. a.
LogFeedP 0.092794 0.028875 3.213667 0.0017
LogInc 0.333895 0.062472 5.344703 0.0000
LogChickP 0.040721 0.059085 0.689185 0.4921
Const 9.583209 0.490918 19.52098 0.0000
Estimates: Q
aFP = .0928 Q
aInc = .3339 Q
aCP = .04072
'LogQ = .0928'LogFeedP + .3339'LogInc + .04072'LogChickP

LogFeedP 0.320692 0.087094 3.682121 0.0004
LogInc 0.716095 0.188434 3.800250 0.0002
LogChickP 0.054309 0.178218 0.304733 0.7611
Const 2.373348 1.480750 1.602802 0.1117
Estimates: P P P
aFP = .3207 aInc = .7161 aCP = .05431
'LogP = .3207'LogFeedP + .7161'LogInc + .05431' LogChickP
i. Yes. We can estimate the own price elasticity of demand because the feed price data
“tells us” when the supply curve shifts allowing us to use the equilibria to estimate
the slope of the demand curve.
bPD .289
ii. Yes. We can estimate the own price elasticity of supply because the chicken price
and income data “tell us” when the demand curve shifts allowing us to use the
equilibria to estimate the slope of the supply curve. The reduced form (RF)
estimation procedure provides two estimates:
'LogQ S .3339'LogInc .3339
bPS .466
'LogP .7161'LogInc .7161
'LogQ S .04072'LogChickP .04072
bPS .450
'LogP .05431'LogChickP .05431
119
b. i. Yes. There is one estimate: .289.
Instrument(s): LogFeedP LogInc LogChickP
LogP 0.289355 0.044935 6.439347 0.0000
LogInc 0.541101 0.027894 19.39826 0.0000
LogChickP 0.056435 0.030582 1.845364 0.0675
Const 8.896469 0.191751 46.39605 0.0000
Estimate bPD = .289

ii. Yes. There is one estimate: .480.
Instrument(s): LogFeedP LogInc LogChickP
LogP 0.480214 0.157434 3.050255 0.0028
LogFeedP –0.243622 0.057725 –4.220402 0.0000
Const 10.68977 0.749584 14.26094 0.0000
Estimate: bPS = .480
5. a.
LogInc 0.717732 0.077586 9.250758 0.0000
LogPorkP 0.347416 0.052969 6.558847 0.0000
Const 3.458030 0.645048 5.360887 0.0000
Estimates: Q Q
aInc = .7177 aPP = .3474
'LogQ = .7177'LogInc + .3474'LogPorkP
120
LogInc 0.268894 0.138588 1.940241 0.0564
LogPorkP 0.044724 0.094616 0.472686 0.6379
Const 2.253325 1.152212 1.955651 0.0546
Estimates: P
aInc = .2689 P
aPP = .04472
'LogP = .2689'LogInc .04472' LogPorkP
i. No. Without feed price data we do not know when the supply curve shifts. We cannot
use the equilibria to estimate the slope of the demand curve.
ii. Yes. We can still estimate the own price elasticity of supply because the pork price
and income data “tell us” when the demand curve shifts allowing us to use the
equilibria to estimate the slope of the supply curve. The reduced form (RF)
estimation procedure provides two estimates:
'LogQ S .7177'LogInc .7177
bPS 2.67
'LogP .2689'LogInc .2689
'LogQ S .3474'LogPorkP .3474
bPS 7.77
'LogP .044721'LogPorkP .04472
b. i. No.
ii. Yes. There is one estimate: 2.67.
Instrument(s): LogInc LogPorkP
LogP 2.667617 1.593295 1.674277 0.0985
Const 0.180328 6.847074 0.026337 0.9791
Estimate: bPS = 2.67
7. None.
9. One.
121
Chapter 24: Binary and Truncated Dependent Variables
1. The more densely populated states tend to have more voters who are registered Democrats
than Republicans; consequently, Democrats would be more likely to win states whose
population density is high. EDen > 0.
3. a.
Dependent Variable: WinDem1
PopDen 0.001001 0.000236 4.247280 0.0001
Const 0.192315 0.074299 2.588406 0.0127
Estimated Equation: EstWinDem1 = .192 + .001PopDen

b. i. PopDen = 1.2
EstWinDem1 = .1922 + .001u1.2 = .1923 + .0012 = .1935
ii. PopDen = 321.0
EstWinDem1 = .1923 + .001u321.0 = .1923 + .3210 = .5133
iii. PopDen = 566.6
EstWinDem1 = .1923 + .001u566.6 = .1923 + .5666 = .7589
iv. PopDen = 821.3
EstWinDem1 = .1923 + .001u821.3 = .1923 + .8213 = 1.0146
v. PopDen = 1,025.0
EstWinDem1 = .1923 + .001u1,025.0 = .1923 + 1.0250 = 1.2173
vi. PopDen = 1,162.0
EstWinDem1 = .1923 + .001u1,162.0 = .1923 + 1.1620 = 1.3543
123
5.
a. The scatter diagram appears to support the theory.

b. The values of the dependent variable, Salary, have a lower bound of 414,000.
1. The dependent variable represents a probability.
3. Yes. In a probit model, the best fitting is S-shaped which can never be less than 0 or greater
than 1. Hence, estimates of the dependent variable always lie between 0 and 1.
1. a. The party winning a state’s Electoral College vote depends on the state’s population
density; as a state becomes more densely populated, the Democrats rather than the
Republicans become more likely to win. EDen < 0.
b.
Dependent Variable: WinRep1
PopDen 0.001001 0.000236 4.247280 0.0001
Const 0.807685 0.074299 10.87080 0.0000
124
Estimated Equation: EstWinRep1 = .808 .001PopDen
bDen = .001: A 1 person per square mile increase in a state’s population density
decreases the probability of a Republican victory by .001.
Critical Result: The PopDen coefficient estimate equals .001. This evidence, the
negative sign of the coefficient estimate, suggests that higher
population density decreases the probability of a Republican victory
c. H0: EDen = 0
H1: EDen < 0
.0001
d. Prob[Results IF H0 True] = < .0001
2
At the 1 percent significance level, we reject the null hypothesis that population density
has no impact of the probability of a Republican victory.
e. Estimated Equation: EstWinDem1 = .192 + .001PopDen
Yes, the results are consistent. A 1 person per square mile increase in population density
increases the probability of a Democratic victory by .001 which is equivalent to
decreasing the probability of a Republican victory by .001.
3. a.
Probit
Explanatory Variable(s): Estimate SE z-Statistic Prob
PopDen 0.009060 0.003073 2.947983 0.0032
Const –0.899796 0.342966 –2.623574 0.0087
population density increases the probability of a Democratic victory
b. H0: EDen = 0
H1: EDen > 0
.0032
c. Prob[Results IF H0 True] = | .002
2
At the 1 percent significance level, we reject the null hypothesis that population density
has no impact on the probability of a Democratic victory.
125
5. a.
Probit
PopDen 0.008500 0.003279 2.592533 0.0095
UnemTrend 0.175342 0.389554 0.450110 0.6526
Const 1.003798 0.416943 2.407520 0.0161
population density increases the probability of a Democratic victory
The UnemTrend coefficient estimate equals .175. This evidence, the
positive sign of the coefficient estimate, suggests that an upward trend
in the unemployment rate increases the probability of a Democratic
victory thereby supporting the theory.
b. Yes. The signs of the coefficient estimates lend support to the theories.
c. The precise numerical values differ, but this is hardly surprising. What is a little
troubling, however, is the Prob. column value for the UnemTrend coefficient estimate has
changed dramatically. Even at the 10 percent significant level we can no longer reject the
null hypothesis that the unemployment trend has no effect on the probability of a
Democratic victory.
7.
The scatter diagram appears to support the theory.
126
9.
Tobit
Dependent Variable: HeatDegDays
HighTemp 0.951145 0.027690 34.34919 0.0000
Const 72.71311 1.873004 38.82165 0.0000
a. The absolute value of the Tobit estimate is greater indicating that the Tobit best fitting
line is more steeply sloped than the ordinary least squares (OLS) best fitting line.
b. The Tobit estimation procedure accounts for the fact that heating degree days are
truncated at 0. Those days that require cooling rather than heating are ignored when we
just consider heating degree days.
127
Chapter 25: Descriptive Statistics, Probability, and Random
Variables—A Closer Look
3
1. a. i. = .6.
5
2
ii. = .4.
5
1
b. i. = .5.
2
1
ii. = .5.
2
3
c. i. = .75.
4
1
ii. = .25.
4
3. a. Mean[v ] ¦ v Prob[v]
All v
b. Var[v ] ¦ (v Mean[v])
All v
2
Prob[v]
1. a. The mean is the average of the values.

b. The mode is the most frequently occurring value.
c. The median is the value that is in the middle; at least half the values lie at the median or
above and at least half lie at the median or below.
3. An event tree visually displays the mutually exclusive outcomes (events) of a random
process.
5. a. The value of one random variable does not affect the probability distribution of the other.
b. 0.
129
1. a.
b. Mean = 719
Mode = 750
Median = 730
c. The left tail is longer than the right tail. Accordingly, the left tail “drags” the mean down.
3. a.
130
b. Prob[Yellow Guilty AND Yellow Reported] = .9 u .8 = .72
Prob[Yellow Guilty AND Orange Reported] = .9 u .2 = .18
Prob[Orange Guilty AND Orange Reported] = .1 u .8 = .08
Prob[Orange Guilty AND Yellow Reported] = .1 u .2 = .02
c. Prob[Yellow Reported] = .74
Prob[Orange Reported] = .26
d. Prob[Yellow Guilty IF Yellow Reported]
Prob[Yellow Guilty AND Yellow Reported] .72
.97
Prob[Yellow Reported] .74
Prob[Orange Guilty IF Yellow Reported]
Prob[Orange Guilty AND Yellow Reported] .02
.03
Prob[Yellow Reported] .74
Prob[Yellow Guilty IF Orange Reported]
Prob[Yellow Guilty AND Orange Reported] .18
.69
Prob[Orange Reported] .26
Prob[Orange Guilty IF Orange Reported]
Prob[Orange Guilty AND Orange Reported] .08
.31
Prob[Orange Reported] .26
e. No. Regardless of which cab the eyewitness reports a Yellow cab is still more likely to be
guilty.
f. There are two pieces of information:
x 90 percent of the cabs are Yellow.
x Eyewitness testimony is 80 percent accurate.
Since 90 percent of the cabs are Yellow and the eyewitness testimony is only 80 percent
accurate, relative number of cabs trumps the eyewitness testimony.
5. a.
131
b.
i. Prob[Prize behind Door 2 IF Monty opens Door 1]
Prob[Prize behind Door 2 AND Monty opens Door 1]
Prob[Monty opens Door 1]
1 1
3 3 2
1 1 1 3

3 6 2
ii. Prob[Prize behind Door 3 IF Monty opens Door 1]
1 1
6 6 1
1 1 1 3

3 6 2
iii. Prob[Prize behind Door 1 IF Monty opens Door 2]
1 1
3 3 2
1 1 1 3

3 6 2
iv. Prob[Prize behind Door 3 IF Monty opens Door 2]

1 1
3 6 1
1 1 1 3

3 6 2
c. Stay.
132
Chapter 26: Estimating the Mean of a Population
1.
Mean [ T1 ( v 1 + v 2 + … + v T) ]
Mean[cx] = cMean[x]
1
=
T
[
Mean v1 + v2 + … + vT]
Mean[x + y] = Mean[x] + Mean[y]

1
= ( Mean[v1] + Mean[v2] + … + Mean[vT])
T
Mean[v1] = Mean[v2] = … = Mean[vT] = ActMean
1
= ( ActMean + ActMean + … + ActMean)
T
There are T ActMean terms
1
= (TuActMean)
T
Simplifying
= ActMean
value is small.
value is small.
133
3. We should divide by the sample size less 1. Variance equals the average of the squared
deviations. To calculate an average we divide by the number of pieces of available
information. When estimating the population variance of a population, the number of pieces
of available information equals the sample size less 1.
5. We use the Student t-distribution because we must estimate the standard deviation.
1. a.
8
¦v
t 1
t
EstMean
8
8.9 6.4 6.2 6.7 7.5 5.3 6.4 6.2 53.6
6.7
8 8
b. Laptop vi EstMean Dev Sqr Dev
1 8.9 6.7 2.2 4.84
2 6.4 6.7 0.3 0.09
3 6.2 6.7 0.5 0.25
4 6.7 6.7 0 0
5 7.5 6.7 0.8 0.64
6 5.3 6.7 1.4 1.96
7 6.4 6.7 0.3 0.09
8 6.2 6.7 0.5 0.25
Sum = 8.12
Sum of Squared Deviations = 8.12

c.
Sum of Squared Deviations Using Estimated Population Mean
EstVarAll
Sample Size 1
8.12 8.12
1.16
8 1 7
d. Knowing whether one battery’s life was greater than or less than average would not help
us predict the life of any other battery.
e.
ActVarAll 1.16
Var[ EstMean] .145
T 8
SE[ EstMean] Var[ EstMean] .145 .3808
134
3. a.
64
¦v
t 1
t
EstMean
64
8 u 8.9 8 u 6.4 8 u 6.2 8 u 6.7 8 u 7.5 8 u 5.3 8 u 6.4 8 u 6.2
64
8 u 53.6 428.8
6.7
64 64
b. Sum of Squared Deviations = 8u8.12 = 64.96

c.
Sum of Squared Deviations Using Estimated Population Mean
EstVarAll
Sample Size 1
64.96 64.96
1.03111...
64 1 63
d.
ActVarAll 1.03
Var[ EstMean] .016111...
T 64
SE[ EstMean] Var[ EstMean] .016111... .1269
e. Mean: 7.0
Standard Error: .1269
Value: 6.7
Degrees of Freedom: 63
f. At the 5 percent significance level, we
would reject the null hypothesis, reject
the manufacturer’s claim that mean
battery life was 7.0. On the other hand, at
the 1 percent significance level we would not reject the null hypothesis.
5. Sum of Variance Variance

Squared Estimate of Estimate of
Deviations Population’s of EstMean’s
Sample from Probability Probability
Size EstMean EstMean Distribution Distribution Prob[Results IF H0 True]
8 6.7 8.12 1.16 .145 .3808
64 6.7 64.96 1.03111… .01611 .0106
120 6.7 121.80 1.0235 .008529 .0008
As the sample size increases and the estimated mean continues to equal 6.7 our confidence
that the mean of the population falls short of the 7.0 hours claimed by the manufacturer is
reinforced.
135
7. As the sample size increases, use of the Student t-distribution becomes less critical. As the
sample size increases, the difference in the probabilities declines.
136

Student Solutions Manual To Accompany An Introduction To Econometrics A Self Contained Approach 1nbsped 9780262317184 9780262525404 - Compress

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

Student Solutions Manual To Accompany An Introduction To Econometrics A Self Contained Approach 1nbsped 9780262317184 9780262525404 - Compress

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Student Solutions Manual To Accompany An Introduction To Econometrics A Self Contained Approach 1nbsped 9780262317184 9780262525404 - Compress

Uploaded by

Copyright:

Available Formats

Student Solutions Manual to Accompany

An Introduction to Econometrics: A Self-Contained Approach

An Introduction to Econometrics: A Self-Contained Approach

The MIT Press

Chapter 1: Descriptive Statistics 1

Solutions to Chapter 1 Prep Questions

Solutions to Chapter 1 Review Questions

7. a. Cov[x, y] > 0 0 < CorrCoef[x, y] d 1

Solutions to Chapter 1 Exercises

b. Yes, SatMath and SatVerbal appear to be correlated. As SatVerbal increases, SatMath

Solutions to Chapter 2 Prep Questions

Solutions to Chapter 2 Review Questions

7. a. Center of the random variable’s probability distribution.

1. Suppose that you have a deck composed of the following 10 cards:

Consider the following experiment:

= 2u.4 + 3u.3 + 4u.2 + 5u.1

v Mean[v] v  Mean[v] (v  Mean[v])2 Prob(v)

Var[v] = 1u.4 + 0u.3 + 1u.2 + 4u.1

Solutions to Chapter 3 Prep Questions

Solutions to Chapter 3 Review Questions

Solutions to Chapter 3 Exercises

1. a. Mean[SatSum] = Mean[SatMath + SatVerbal]

Mean[SatDiff] = Mean[SatMath  SatVerbal]

To achieve that objective, z must equal 1.29.

Mean[Your TNW] = Mean[v1 + v2 + … + v50]

ii. Mean[TNW] = 1.35

c. Your friend appears to be more risk averse than you.

Solutions to Chapter 4 Prep Questions

Defendant Actually Jury is Jury is

Defendant Actually Jury is Jury is

b. Defendant Actually Innocent and Jury Finds Defendant Guilty

Solutions to Chapter 4 Review Questions

Solutions to Chapter 4 Exercises

Mean[DemCongressFrac] = p = .451 where p = fraction of population statewide

g. Prob[Results IF H0 True] = Probability that DemCongressFrac would be

Prob[Results IF H0 True] = .1814

There is about one chance in five

Prob[TNW t 20 IF everything on the “up and

Solutions to Chapter 5 Prep Questions

5. Begin with the equation for SSR:

Collecting like terms.

Solutions to Chapter 5 Review Questions

3. The parameter estimates are random variables.

Solutions to Chapter 5 Exercises

b. First, calculate the means:

bConst y  bx x 46  2.4 u15 46  36 10

Estimated Equation: EstOilProd = 5,945 + 92.5Price

Estimated Equation: EstGasCons = 517  151.7PriceDollars

7. a. Earmarks are projects initiated by members of Congress that circumvent merit-based

Solutions to Chapter 6 Prep Questions

3. a. Variance of a constant times a variable: Var[cx] = c2Var[x]

Solutions to Chapter 6 Review Questions

3. a. The mean of the coefficient estimate’s probability distribution is unaffected.

Solutions to Chapter 6 Exercises

b. The estimated constants

Solutions to Chapter 7 Prep Questions

Solutions to Chapter 7 Exercises

1. First, calculate the means:

v Mean[v] v Mean[v] (v Mean[v])2 Prob(v)

Mean[SatDiff] = Mean[SatMath SatVerbal]

To achieve that objective, z must equal 1.29.

ii. Mean[TNW] = 1.35

bConst y bx x 46 2.4 u15 46 36 10

Estimated Equation: EstGasCons = 517 151.7PriceDollars

bConst y bx x 46 2.4 u15 46 36 10

¦ ( yt y )( xt x ) bx u ¦ ( xt x )2 15.26 u 302.48 4, 615.8

Estimated Equation: EstPetroConsPC = 1,418 540.5PriceReal

Estimated Equation: EstGasCons = 516.8 151.7PriceDollars

Mean[bP] = EP = 0 SE[bP] = 47.57 DF = 10 2 = 8

Mean[bDem] = EDem = 0 SE[bDem] = .577 DF = 451 2 = 449

Mean[bP] = EP = 0 SE[bP] = 28.1 DF = 29 2 = 27

Estimated Equation: EstLogPetroConsPC = 6.81 .365LogPriceReal

H0: EP = 1 Cynic’s view is correct: Actual price elasticity of demand equals

Mean[bP] = E P = 1 SE[bP] = .132 DF = 10 2 = 8

Mean[bP] = E P = .10 SE[bP] = .0695 DF = 29 2 = 27

Estimated Equation: EstLogCigConsPC = 6.38 1.44 LogPriceConsumer

Estimated Equation: LogHoursPerWeek = .037 + .768 LogWage

Mean[bWage] = E Wage = 1 SE[bP] = .295 DF = 92 2 = 90

EConst P E I E ChickP E ChickP E

= log(EConst) + EPlog(P) + EIlog(I) + (EClever EP EI)log(ChickP)

= log(EConst) + EP[log(P) log(ChickP)] + EI[log(I) log(ChickP)] + ECleverlog(ChickP)

Estimated Equation: EstCigConsPC = 155 16.0PriceConsumer .32IncPC

Estimated Equation: EstSmokeRateYouth = 33.5 1.46PriceConsumer

Estimated Equation: EstCigConsPC = 207.7 34.2PriceSupplier 12.7Tax

( SSRR SSRU ) / ( DFR DFU ) 1,157.40 /1 1,157.40

Estimated Equation: EstSmokeRateYouth = 33.45 1.46PriceConsumer