Exercise 8-26 Cost Classification

Requirement

Match each cost to the appropriate cost behavior pattern shown in the graphs (a) through (
fit two or more patterns.

Solution

1. b
2. f
3. e
4. i
5. e
6. h
7. l
8. j
9. k
10. g
n in the graphs (a) through (l). Any graph can
Exercise 8-27 Cost Relationships

Comptech hired Erwin & Associates to design a new computer-aided manufacturing

capacity to produce 250 computers per day. The variable costs for each computer
fixed costs total $62,250 per month.

Required What is the average cost per unit if the facility normally expects to opera
capacity?

Solution
puter-aided manufacturing facility that has the
e costs for each computer are $150 and the

normally expects to operate at 80 percent of

Exercise 8-28 Cost Relationships
Background

The following costs are for Optical View Inc., a contact lens manufacturer:

Output in Units Fixed Costs Variable Costs

250 $4,750 $7,500
300 $4,750 $9,000
350 $4,750 $10,500
400 $4,750 $12,000

8-28 Requirements

1. Graph total cost, total variable costs, and total fixed costs.
2. Graph the per-unit total cost, per-unit variable cost, and per-unit fixed cost.
3. Discuss the behavior of the fixed, variable, and total costs.

Solution
COST

Fixed Costs
Variable Costs
Total Costs

OUTPUT

Output in Units Fixed Costs/unit Variable Costs/unit

250 $19 $30
 300 $16 $30
350 $14 $30
400 $12 $30

COST
Fixed Costs/unit
Variable Costs/unit
Total Costs/unit

OUTPUT

3. The important point of these graphs is that total fixed costs are constant, while unit fixed
changes. In contrast, unit variable costs are constant and total variable costs changes a
ufacturer:

Total Costs
$12,250
$13,750
$15,250
$16,750

Fixed Costs
Variable Costs
Total Costs

Total Costs/unit
$49
 $46
$44
$42

Fixed Costs/unit
Variable Costs/unit
Total Costs/unit

stant, while unit fixed costs change as output

able costs changes as output changes.
Exercise 8-29 Cost Estimation, Average Cost

Maribeth’s Cafe bakes croissants that it sells to local restaurants and grocery store
North Carolina, area. The average costs to bake the croissants are $0.55 for 500 and

Required If the total cost function for croissants is linear, what will be the average cos

Units Avg. Cost

500 $0.55
600 $0.50
560 ?

Solution
staurants and grocery stores in the Raleigh,
ants are $0.55 for 500 and $0.50 for 600.

what will be the average cost to bake 560?

Exercise 8-30 Cost Estimation Using Graphs, Service
Background

Lawson Advertising Agency is trying to persuade Kansas City Sailboards Company to s

money on advertising. The agency’s argument is that a positive linear relationship exists betw
advertising and sales in the sailboard industry. Sue Lawson presents these data taken from ind
for stores similar in size and market share to Kansas City Sailboards:

Advertising Expense Annual Sales

$2,500 $95,000
$3,000 $110,000
$3,500 $124,000
$4,000 $138,000
$4,500 $143,000
$5,000 $147,000
$5,500 $150,000

8-30 Requirements

1. Graph annual sales and advertising expense.

2. Do the data prove Sue's point?

Solution

1. Graph annual sales and advertising expense.

Annual Sales

1 2 3

Advertising Expense
 1 2 3

Advertising Expense

2. Do the data prove Sue's point?

There seems to be a positive linear relationship for the data between $2,500
$4,000 of advertising expense. Lawson's analysis is correct within this relevant ra
but not outside of it. Notice that the relationship between advertising expense
sales changes at $4,000 of expense.
City Sailboards Company to spend more
linear relationship exists between
ents these data taken from industry data
ards:

1 2 3
 1 2 3

r the data between $2,500 and

correct within this relevant range
etween advertising expense and
Exercise 8-31 Analysis of Regression Results

Wang Manufacturing uses regression analysis to predict manufacturing overhead

labor hours and/or machine hours, and has developed the three following regression

Regression 1 Regression 2 Regression 3

SE 33,844 45,383 31,044

R-squared 0.55 0.35 0.58
t-values:
Labor Hours 2.3 1.9
Machine Hours 1.1 0.8

Required: Which regression would you choose and why?

Solution
t manufacturing overhead costs based on
e three following regression equations.

y?
Exercise 8-32 Cost Estimation: High-Low Method; MAPE
Background
Horton Manufacturing Inc. produces blinds and other window treatments for residenti
The owner is concerned about the maintenance costs for the production machinery, as maint
previous fiscal year were higher than he expected. The owner has asked you to assist in esti
maintenance costs to better predict the firm’s profitability. Together, you have determined t
for maintenance costs is machine-hours. The data from the previous fiscal year for maintena
machine-hours are as follows:
Problem Information

Month Expense Hours

1 $2,625 1499
2 $2,670 1590
3 $2,720 1605
4 $2,820 1655
5 $2,855 1775
6 $3,005 1880
7 $2,865 1785
8 $2,905 1805
9 $2,780 1695
10 $2,570 1410
11 $2,590 1550
12 $2,890 1405

8-32 Requirements

1. What is the cost equation for maintenance cost using the high-low method?
2. Calculate MAPE for the equation you developed in part 1.

Solution

Expense Hours
High month $3,005 1,880
Low month $2,570 1,410
 Diffference $435 470

slope equals $0.9255 per machine hour

Maintenance Cost equation $1,265.00 +

Model to fit: Maintenance Expense = a + b x M (machine hours)

The graph below shows that the selected high and low points
representative of the data, but there is one significant outlier, the point
month 12

3100
3000
2900
2800
2700
2600
2500
0 500 1000 1500 2000

2. The MAPE results are shown below. Note that the outlier, point 12, has
a large MAPE. Overall, the MAPE is relatively low, due to the good fit of
the model to a set of data that is relatively linear.

HiLo
Hours Expense Estimate
1 1499 2625 2,652
2 1590 2670 2,737
3 1605 2720 2,751
4 1655 2822 2,797
5 1775 2855 2,908
 6 1880 3005 3,005
7 1785 2865 2,917
8 1805 2905 2,936
9 1695 2780 2,834
10 1410 2570 2,570
11 1550 2590 2,700
12 1405 2890 2,565

While not required for the exercise, a regression analysis on the data
produces the following results (regression 1). Note the significant
difference between the regression and the HiLo results. Also note the
relatively poor R-squared. This might be due to the outlier in month 12.
When month 12 is removed the results are shown in regression 2, a much
improved regression.

Regression 1

Regression Statistics
Multiple R 0.699085436936659
R Square 0.488720448136919
Adjusted R Square 0.437592492950611
Standard Error 116.860000802718
Observations 12

ANOVA
df SS MS
Regression 1 130537.068790555 130537.0687906
Residual 10 136562.597876112 13656.25978761
Total 11 267099.666666667

Coefficients Standard Error t Stat

Intercept -536.628580636132 704.125003357368 -0.76212118314
Hours 0.783660478951064 0.25347019680166 3.091726320646

Regression 2
Regression Statistics
Multiple R 0.964982997830891
R Square 0.931192186102694
Adjusted R Square 0.923546873447438
Standard Error 39.8737911388807
Observations 11
ANOVA
df SS MS
Regression 1 193650.727021916 193650.7270219
Residual 9 14309.2729780838 1589.919219787
Total 10 207960

Coefficients Standard Error t Stat

Intercept -1074.2735417097 247.954593469987 -4.33254140073
Hours 0.988785771658061 0.08959425699174 11.03626286838
w treatments for residential homes and offices.
ion machinery, as maintenance costs for the
sked you to assist in estimating future
r, you have determined that the best cost driver
s fiscal year for maintenance expense and

gh-low method?

Constant
$1,265.00
$1,265.00
machine hour

$0.9255 times the number of machine hours

M (machine hours)

gh and low points are

cant outlier, the point for

0 2000

, point 12, has

good fit of

MAPE
1.0%
2.5%
1.1%
0.9%
1.9%
 0.0%
1.8%
1.1%
1.9%
0.0%
4.2%
11.2%
2.3%
alysis on the data
he significant
ults. Also note the
outlier in month 12.
regression 2, a much

F Significance F
9.55877164178 0.011410355501

P-value Lower 95%

0.46358391938 -2105.51685054
0.0114103555 0.21889368806
 F Significance F
121.7990981 1.56585476E-06

P-value Lower 95%

0.00189803773 -1635.18580023
1.56585476E-06 0.786109481873
Exercise 8-33 Cost Estimation: High-Low Method; MAPE
Background

Ethan Manufacturing Inc. produces floor mats for automobiles. The owner,
Joseph Ethan, has asked you to assist in estimating maintenance costs. Together,
you and Joseph determine that the single best cost driver for maintenance costs is
machine-hours. These data are from the previous fiscal year for maintenance
expense and machine-hours:

Month Expense Machine-Hours

1 $2,600 1,690
2 $2,760 1,770
3 $2,910 1,850
4 $3,020 1,870
5 $3,100 1,900
6 $3,070 1,880
7 $3,010 1,860
8 $2,850 1,840
9 $2,620 1,700
10 $2,220 1,100
11 $2,230 1,300
12 $2,450 1,590

8-33 Requirements
iles. The owner,
e costs. Together,
aintenance costs is
r for maintenance
Exercise 8-34 The Gompertz Equation; Learning Curves
The concept of learning curves has broad application in business, medicine, and many other f
For example, the Gompertz Equation is a mathematical model used to predict the number of d
a certain age. The Gompertz Equation is very similar in form to that of the learning curve; the
differences are the use of e, the base of the natural logarithm, and there is a positive rather tha
negative exponent in the equation. The Gompertz Equation is as follows:
M(x) = AeGx
Where: M(x) is the number of deaths in a population of 100,000 of those at age x; M(x) is
often called the mortality rate
A is the initial mortality rate at age 0
G is the exponential rate of increase in mortality for an increase in age, x
e is a mathematical constant, the base of the natural logarithm, which equals
approximately 2.718281828,

The Gompertz equation is used to estimate the number of deaths at a given age. The equatio
estimated using nonlinear regression based on 2002 U.S. census data and the following estim
equation was derived (for ages 25-90). The regression had a very good fit, R-squared of 0.97

M(x) = 8.84e.08x

8-34 Requirements

1. Use the exponential function on your calculator or the EXP function in Excel to determine
mortality rate of any age you choose between 25 and 90.
2. Think of another way or two that the exponential equation like the Gompertz Equation cou
used in cost estimation.

Solution

1. Mortality
Age Rate
25 62
26 68 60 1,026
27 73 61 1,111
28 79 62 1,203
29 86 63 1,304
30 93 64 1,412
31 101 65 1,530
32 109 66 1,657
33 118 67 1,795
 34 128 68 1,945
35 139 69 2,107
36 150 70 2,282
37 163 71 2,472
38 176 72 2,678
39 191 73 2,901
40 207 74 3,143
41 224 75 3,405
42 243 76 3,689
43 263 77 3,996
44 285 78 4,329
45 309 79 4,689
46 335 80 5,080
47 362 81 5,503
48 393 82 5,961
49 425 83 6,457
50 461 84 6,995
51 499 85 7,578
52 541 86 8,209
53 586 87 8,893
54 635 88 9,633
55 687 89 10,436
56 745 90 11,305
57 807
58 874
59 947

2. An exponential equation like the Gompertz equation could be used to estimate the effe
employees working overtime on the production defect rate, or it could be used to estimate
increase in maintenance cost as a machine ages. These are examples of costs increasin
exponential rate as usage increases.
es
ss, medicine, and many other fields.
used to predict the number of deaths at
o that of the learning curve; the
and there is a positive rather than a
as follows:

00 of those at age x; M(x) is

n increase in age, x
garithm, which equals

hs at a given age. The equation was

us data and the following estimated
ery good fit, R-squared of 0.97.

P function in Excel to determine the

like the Gompertz Equation could be

 be used to estimate the effect of
or it could be used to estimate the
examples of costs increasing at an
Exercise 8-35 Regression and Utility Rates; Sustainability

For several years many utilities have been able to use regression analysis to forecast monthly ut
usage by residential customers using weather forecasts, the number of holidays, the number of d
the month and other factors. For example, the Connecticut Department of Public Utility Control (
has determined that regression, properly used, can accurately predict natural gas usage. Most g
utilities serving Connecticut have reported levels of accuracy of from 4% to 10% using regression
company, Dominion Natural Gas Company of Ohio, uses this approach, not to forecast, but to ex
customers why their natural gas bills have gone up or down compared to the prior month, and als
compared to the same month of the prior year. The bill shows total MCF (thousand cubic feet of
gas) used by the customer for that month and why the total MCF usage has changed, based on t
factors:
o Change in temperature: each degree increase in temperature causes an increase in
number of MCFs consumed. The relationship between the change in temperature an
usage of MCF is not linear, but the monthly bill shows the average change in tempera
the month and the increase or decrease in MCF related to that change
o Number of billing days in the period
o The residual , the change in usage by the customer that is not attributable to tempera
the number of days in the billing period.

A customer of Dominion has used13.7 MCF in December and is charged $12.50 per MCF
total bill that month of $171.25. The following data is available to compare the current mon
weather and billing period to the prior month and to the same month last year

Usage Current Month vs Last Current Month to

Factors Month Last December

Weather 3 degrees cooler; 8 degrees

+2.5MCF warmer; - 3.5MCF

Number of 5 more days; +.05 MCF 1 less day; -0.1

Billing Days +.9 MCF
Customer MCF -         
controlled 1.8 MCF
usage8-38 Requirements

1. Determine the amount of difference in the customer’s bill from the prior month and from th
current month last year.
2. How does the Dominion Company’s billing system affect environmental sustainability?

Solution
bility

n analysis to forecast monthly utility

mber of holidays, the number of days in
artment of Public Utility Control (CDPUC)
redict natural gas usage. Most gas public
from 4% to 10% using regression. One
pproach, not to forecast, but to explain to
mpared to the prior month, and also
otal MCF (thousand cubic feet of natural
F usage has changed, based on three

mperature causes an increase in the

een the change in temperature and the
ws the average change in temperature for
ted to that change

that is not attributable to temperature or

and is charged $12.50 per MCF for a

lable to compare the current month’s
me month last year

l from the prior month and from the

environmental sustainability?
Exercise 8-36 Interpreting Regression Results
Interpreting Regression Results Recent research into the cost of various medical proced
shown the impact of certain complications encountered in surgery on the total cost of patient’
the hospital. The researchers used regression analysis and found the following results:

Total Cost for Patient = Constant, plus

a x length of stay (measured in days), plus

b x presence of one or more complications (1 if true, 0 if false), plus
c x use of a laparoscope (1 if true, 0 if false)

Where:

• a, b, c are coefficients of the regression model, and

• The laparoscope is an instrument somewhat like a miniature telescope with a fiber optic
which brings light into the abdomen. It is about as big around as a fountain pen and twice as lo
The research, based on 57 patients, showed the following regression
r-squared: 53%
constant term: $3,719

Coefficients and t-values for independent variables:

Length of Stay Complication Laproscope

Coefficient $861 $1,986 $908
t-value 10.76 4.89 2.54

Required

1. What is the estimated cost for a patient who has complications and stays in the hospital t
days, and whose surgery requires a laparoscope?
2. Which, if any, dummy variables are used in this regression?
3. Comment on the statistical measures for the model.

R Squared 53%
Constant $ 3,719
Coefficients:
Length of Stay $ 861
Complications $ 1,986
Laproscope $ 908

Solution
1. The estimated cost is:

$3,719 + 2 x $861 + 1 x $1,986 + 1 x $908 = $8,335

2. There are two dummy variables in this regression:

 presence of one or more complications.

 use of a laparoscope (or not)

3. The model has a relatively low R-squared of only 53%, but all three independent va
have good t-values (>2.0). Looking at the t-values, it appears that the strongest indepe
variable is the length of stay, and the weakest is the use of laparoscope.

The exercise is based on information from: “Hospital Costs of Uterine Artery Emboliza
M Beinfeld, J. Bosch, and G Gazette, Academic Radiology, Nov 9, No. 11, November
1300-1304.
 various medical procedures has
n the total cost of patient’s stay in
e following results:

escope with a fiber optic system

untain pen and twice as long.
following regression results:

proscope
$908
2.54

d stays in the hospital two

 $ 8,335

all three independent variables

hat the strongest independent
aroscope.

Uterine Artery Embolization…” by

ov 9, No. 11, November 2002, pp.
Exercise 8-37 Interpreting Regression Results; Appendix (Continuation of 8-36)
The Following table shows additional regression results presented by the researc
described in Exercise 8-36. There are two regressions. The right hand column shows
57 patients. The left hand column shows the results for the sub-sample of patients
without the laparoscopic surgery.

Not Laparoscopic Laparoscopic

Coefficients for Independent Variables
Regression Intercept $ 8,043 $ 3,719
Length of Stay
Coefficient* Not significant 861
Standard error for the coefficient Not applicable 80
Number of Complications
Coefficient 3,393 1,986
Standard error for the coefficient 1,239 406
Laparascopic
Coefficient Not applicable 908
Standard error for the coefficient Not applicable 358

R-squared 0.11 0.53

*Note: All independent variables are significant at the level of p=.05 (and t-value
length of stay variable in the Non-laparoscopic condition.

Required

1. Which of the two regressions has the better reliability and precision in estimating co
2. Interpret the values of each coefficient and the standard error for each coefficient.
3. What are the t-values for each of the independent variables for each regression?

Solution
ndix (Continuation of 8-36)
lts presented by the researchers in the study
The right hand column shows the results for all
r the sub-sample of patients who were treated

c Laparoscopic

$ 3,719

861
80

1,986
406

908
358

0.53
level of p=.05 (and t-value >2) except for the

and precision in estimating cost, and why?

rd error for each coefficient.
iables for each regression?
Exercise 8-38 Cost Estimation; High Low Method
Background
Albedo Inc. manufactures high-end replacement telescope lenses for amateur and professi
are seeking to upgrade the performance of their telescopes. You have just become
accountant at Albedo and Jordan Coleman, the controller, has asked you to help w
estimation for the lens manufacturing process. You review the manufacturing process an
cost driver for maintenance costs is machine-hours. These data are from the prev
maintenance expense and machine-hours:

Problem Information

Month Maint. Cost Hours

1 3210 2750
2 4650 3900
3 5175 4050
4 3350 2690 High
5 3100 2500
6 2950 2580
7 2900 2300
8 2900 2500 Low
9 4120 3160
10 4350 3325
11 3500 2780
12 3775 3000

8-38 Requirements

What is the cost equation for maintenance cost using the high-low method?

Solution Slope (b) = 1.30 =(5175-2900)/(4050-2300)

Constant (a) = -90 =5175 - 1.30x4050

Cost equation = $-90 + $1.30 x Hours

 for amateur and professional astronomers that
You have just become employed as a staff
as asked you to help with maintenance cost
manufacturing process and decide that the best
data are from the previous fiscal year for

-low method?
Problem 8-39 Cost Estimation: High-Low Method
Background
Jay Bauer Company specializes in the purchase, renovation, and resale of older homes. Jay B
and painters to do the work for him. It is essential for him to have accurate cost estimates so h
costs before he purchases a piece of property. If estimated renovation costs plus the purchase pr
estimated resale value, the house is not a worthwhile investment.

Jay has been using the home’s interior square feet for his exterior paint cost estimations. Rec
number of openings-the total number of doors and windows in a house-as a cost driver. Their
require time consuming preparatory work and careful brushwork. The rest of the house usual
spray guns, which are relatively efficient ways to apply paint to a large area. Jay has kept care
his last 12 jobs:

House Square Feet Openings

1 2,600 13
2 3,010 15
3 2,800 12
4 2,850 12
5 4,050 19
6 2,700 13
7 2,375 11
8 2,450 11
9 2,600 10
10 3,700 16
11 2,650 13
12 3,550 16

8-39 Requirements

1. Using the high-low cost-estimation technique, determine the cost of painting a 3,300-squ
Also, determine the cost for a 2,400-square-foot house with 8 openings.
2. Plot the cost data against square feet and against openings. Which variable is a better c

Solution
 older homes. Jay Bauer employs several carpenters
e cost estimates so he can determine total renovation
plus the purchase price of a house are higher than its

cost estimations. Recently he decided to include the

s a cost driver. Their cost is significant because they
t of the house usually is painted either by rollers or
ea. Jay has kept careful records of these expenses on

Cost
$3,300
$3,750
$3,100
$3,150
$4,700
$3,250
$2,800
$2,800
$2,875
$4,100
$3,200
$3,950

painting a 3,300-square-foot house with 14 openings.

ariable is a better cost driver? Why?

Problem 8-40 Cost Estimation, Machine Replacement, Ethics
Background

SpectroGlass Company manufactures glass for office buildings in Arizona and Southern Cali
result of age and wear, a critical machine in the production process has begun to produce qua
SpectroGlass is considering replacing the old machine with a new machine, either brand A
The manufacturer of each machine has provided SpectroGlass these data on the cost of ope
machine at various levels of output:

Machine A
Output (square yards) Estimated Total Costs

4,000 $54,600
7,000 $78,800
9,000 $90,300
14,000 $114,900
16,000 $132,400
24,000 $210,000

8-40 Requirements

1. If SpectroGlass's output is expected to be 22,000 square yards, which machine should i

2. As a cost analyst at SpectroGlass, you have been assigned to complete Requirement 1.
you to say that the nature of the defect is really very difficult to detect and that most cust
replacing it. He suggests that you modify your calculations to justify keeping the present
and save the company some money. What do you say?
3. Assume that brand A is manufactured in Germany and brand B is manufactured in Cana
considerations are important to SpectroGlass, in addition to those already mentioned in

Solution

1. A graph of the data shows no significant outliers and no nonlinar trend.

 E stim a te d T o ta l C o s ts

Output

Using the High-Low Method:

Machine A:
slope coefficient = $7.77
Fixed cost estimate = $23,520

Cost estimation:
x= 22,000
y= $194,460

x= 15,000
y= $140,070

Machine B:
slope coefficient = $6.10
fixed cost estimate = $45,600

x= 22,000
y= $179,800
 x= 15,000
y= $137,100

Therefore, costs are lower at both the 22,000 level and the 15,000 level for Machin

The calculations show that the costs are lower at both the 22,000 level and
the 15,000 level for Machine B, which suggests that Machine B is
preferred for production levels above at least 15,000. The answer is
wrong. Note by inspection of the chart in the text and inspection of the
graph above, that Machine A is preferred to Machine B up to 16,000 units
or more. The error in the analysis is the error in using the High-Low
method for cost estimation in this particular case. The High-Low method
derives a linear relationship from the least to the largest level of output,
while in fact the relationship between output and cost in this case is not
linear, as illustrated in the graph above. The exercise serves as a good
reminder that the application of a cost estimation method should be used
with caution, and that the review of the related graph can be a key step in
identifying potential mistakes.
2. The ethical issue presented in this case should be addressed using the
approach described in chapter 1. Here it seems important to consider the
nature and extent of the effect of the defect on customers and also
SpectroGlass. Since the glass is used in office buildings, and defects are
likely to affect the safety of those using the buildings, the cost analyst has
a responsibility to make sure that management has a clear picture of the
costs of each machine. The cost analysis should be presented to top
management in a way that makes the ethical choice apparent and
appropriate. For this reason, the calculations should not be modified.

3. In addition to the costs of the machine, SpectroGlass should be aware

of any import duties or restrictions for the purchase of the machines from
Germany or Canada. How will these restrictions and duties, if any, affect
the cost and availability of the machine? What are the effects on the
purchase decision of changes in the currencies of Canda and Germany
relative to the U.S. dollar. Also, will the purchase in either country lead to
potentially beneficial business relationships in that country. For example,
the purchase in a given country might open up new markets for
SpectroGlass.
relative to the U.S. dollar. Also, will the purchase in either country lead to
potentially beneficial business relationships in that country. For example,
the purchase in a given country might open up new markets for
SpectroGlass.
nt, Ethics

ngs in Arizona and Southern California. As a

process has begun to produce quality defects.
h a new machine, either brand A or brand B.
ass these data on the cost of operation of its

Machine B
Estimated Total Costs

$70,000
$100,000
$115,000
$137,000
$146,000
$192,000

yards, which machine should it purchase? At 15,000 square yards?

ned to complete Requirement 1. A production supervisor comes to
cult to detect and that most customers will not notice it, so he questions
ns to justify keeping the present machine to keep things the way they are

rand B is manufactured in Canada. As a U.S.-based firm, what

to those already mentioned in your answer to requirement 1?

nonlinar trend.
 Machine A
Machine B

per sq. yard

yards

per sq. yard

yards
 yards

and the 15,000 level for Machine B.

the 22,000 level and

achine B is
The answer is
inspection of the
B up to 16,000 units
the High-Low
e High-Low method
st level of output,
in this case is not
serves as a good
hod should be used
can be a key step in

addressed using the

ortant to consider the
customers and also
ngs, and defects are
the cost analyst has
a clear picture of the
be presented to top
hoice apparent and
not be modified.

ass should be aware

of the machines from
duties, if any, affect
e the effects on the
Canda and Germany
either country lead to
untry. For example,
p new markets for
either country lead to
untry. For example,
p new markets for
Problem 8-41 Cost Estimation: High-Low Method
Background
Antelope Park Amoco (APA) in Antelope Park, Alaska, has noticed that utility bills are su
the average monthly temperature is. The only thing in the shop that uses natural gas
prevailing low temperatures, the furnace is used every month of the year (though less in t
little in August). Everything else in the shop runs on electricity, and electricity use is fairly c

For a year, APA has been recording the average daily temperature and the cost of its month
and electricity.

Avg. Temp F Utility Cost

January 31 760
February 41 629
March 43 543
April 44 510
May 46 275
June 50 233
July 53 220
August 60 210
September 50 305
October 40 530
November 30 750
December 20 870

8-41 Requirements

Use the high-low method to estimate utility cost for the upcoming months of January and F
near-record average temperature of 10 degrees Fahrenheit; temperatures in February are
Fahrenheit.

Solution
d that utility bills are substantially higher the colder
that uses natural gas is the furnace. Because of
e year (though less in the summer months and very
electricity use is fairly constant throughout the year.

and the cost of its monthly utility bills for nature gas

onths of January and February. The forecast for January is a

atures in February are expected to average 40 degrees
Problem 8-42 Regression Analysis
Problems 8–42 through 8–46 are based on Armer Company, which is accumulating data to us
its annual profit plan for the coming year. The cost behavior pattern of the maintenance
determined. The accounting staff has suggested the use of linear regression to derive an
maintenance hours and costs. Data regarding the maintenance hours and costs for the last year
of the regression analysis follow:

Hours of Activity
January 480
February 320
March 400
April 300
May 500
June 310
July 320
August 520
September 490
October 470
November 350
December 340
Sum 4,800
Average 400

Average cost per hour ($43,200/4,800) =

a (intercept)
b coefficient
Standard error of the estimate
R-squared
t-value for b
 8-42 Requirements

If Armer Company uses the High-Low method of analysis, the equation for the relationship
maintenance cost is?

Solution

Cost
High Point $4,470
Low Point $2,820
Difference $1,650

Estimated slope coefficient = $7.50

Estimated intercept = $570.00
which is accumulating data to use in preparing
or pattern of the maintenance costs must be
linear regression to derive an equation for
hours and costs for the last year and the results

Maintenance
Costs
$4,200
$3,000
$3,600
$2,820
$4,350
$2,960
$3,030
$4,470
$4,260
$4,050
$3,300
$3,160
$43,200
$3,600

,200/4,800) = $9.00

684.65
7.2884
34.469
0.99724
60.105
e equation for the relationship between hours of activity and

Activity
520
300
220

per hour
per month
Problem 8-43 Regression Analysis

Problems 8–42 through 8–46 are based on Armer Company, which is accumulating data to
its annual profit plan for the coming year. The cost behavior pattern of the maintenan
determined. The accounting staff has suggested the use of linear regression to derive
maintenance hours and costs. Data regarding the maintenance hours and costs for the last ye
of the regression analysis follow:

Hours of Activity
January 480
February 320
March 400
April 300
May 500
June 310
July 320
August 520
September 490
October 470
November 350
December 340
Sum 4,800
Average 400

Average cost per hour ($43,200/4,800) =

a (intercept)
b coefficient
Standard error of the estimate
R-squared
t-value for b
 8-43 Requirements

Based on the data derived from the regression analysis, 420 maintenance hours in a month
should be budgeted at:

Solution
ich is accumulating data to use in preparing
pattern of the maintenance costs must be
near regression to derive an equation for
urs and costs for the last year and the results

Maintenance
Costs
$4,200
$3,000
$3,600
$2,820
$4,350
$2,960
$3,030
$4,470
$4,260
$4,050
$3,300
$3,160
$43,200
$3,600

43,200/4,800) = $9.00

684.65
7.2884
34.469
0.99724
60.105
enance hours in a month mean that maintenance costs
Problem 8-44 Regression Analysis
Problems 8–42 through 8–46 are based on Armer Company, which is accumulating data to
its annual profit plan for the coming year. The cost behavior pattern of the maintenance
determined. The accounting staff has suggested the use of linear regression to derive
maintenance hours and costs. Data regarding the maintenance hours and costs for the last yea
of the regression analysis follow:

Hours of Activity
January 480
February 320
March 400
April 300
May 500
June 310
July 320
August 520
September 490
October 470
November 350
December 340
Sum 4,800
Average 400

Average cost per hour ($43,200/4,800) =

a (intercept)
b coefficient
Standard error of the estimate
R-squared
t-value for b
 8-44 Requirements

The coefficient of determination for Armer's regression equation for maintenance activities i

Solution

The coefficient of determination is also referred to as R-squared, which in this case =

hich is accumulating data to use in preparing
r pattern of the maintenance costs must be
inear regression to derive an equation for
ours and costs for the last year and the results

Maintenance
Costs
$4,200
$3,000
$3,600
$2,820
$4,350
$2,960
$3,030
$4,470
$4,260
$4,050
$3,300
$3,160
$43,200
$3,600

$43,200/4,800) = $9.00

684.65
7.2884
34.469
0.99724
60.105
 for maintenance activities is?

d, which in this case = 99.724%

Problem 8-45 Regression Analysis
Problems 8–42 through 8–46 are based on Armer Company, which is accumulating data to u
its annual profit plan for the coming year. The cost behavior pattern of the maintenance
determined. The accounting staff has suggested the use of linear regression to derive a
maintenance hours and costs. Data regarding the maintenance hours and costs for the last year
of the regression analysis follow:

Hours of
Activity
January 480
February 320
March 400
April 300
May 500
June 310
July 320
August 520
September 490
October 470
November 350
December 340
Sum 4,800
Average 400

Average cost per hour ($43,200/4,800) =

a (intercept)
b coefficient
Standard error of the estimate
R-squared
t-value for b
 8-45 Requirements

The percent of the total variance that can be explained by the regression equation is:

Solution

The percentage of the variation in the dependent variable, cost, that is explained by change
is called the "coefficient of determination" or "R-Squared."

In the present regression equation, the R-squared statistic =

 which is accumulating data to use in preparing
vior pattern of the maintenance costs must be
of linear regression to derive an equation for
hours and costs for the last year and the results

Maintenance
Costs
$4,200
$3,000
$3,600
$2,820
$4,350
$2,960
$3,030
$4,470
$4,260
$4,050
$3,300
$3,160
$43,200
$3,600

ur ($43,200/4,800) = $9.00

684.65
7.2884
34.469
0.99724
60.105
e regression equation is:

ost, that is explained by changes in the independent variable, hours,

99.724%
Problem 8-46 Regression Analysis

Problems 8–42 through 8–46 are based on Armer Company, which is accumulating data t
its annual profit plan for the coming year. The cost behavior pattern of the maintenan
determined. The accounting staff has suggested the use of linear regression to derive
maintenance hours and costs. Data regarding the maintenance hours and costs for the last y
of the regression analysis follow:

Hours of Activity
January 480
February 320
March 400
April 300
May 500
June 310
July 320
August 520
September 490
October 470
November 350
December 340
Sum 4,800
Average 400

Average cost per hour ($43,200/4,800) =

a (intercept)
b coefficient
Standard error of the estimate
R-squared
t-value for b
 8-46 Requirements

At 400 hours of activity, Armer management can be approximately two-thirds confident that
be in the range of?

Solution

Std error of the regression = $34.47

CI (no. of std deviations) = 1
x= 400
estimated y = $3,600 for the month

CI (67%):
Upper Limit = $3,634.48
Lower Limit = $3,565.54
 is accumulating data to use in preparing
ttern of the maintenance costs must be
ar regression to derive an equation for
and costs for the last year and the results

Maintenance
Costs
$4,200
$3,000
$3,600
$2,820
$4,350
$2,960
$3,030
$4,470
$4,260
$4,050
$3,300
$3,160
$43,200
$3,600

43,200/4,800) = $9.00

684.65
7.2884
34.469
0.99724
60.105
o-thirds confident that the monthly maintenance costs will
Problem 8-47 Regression Analysis
Pilot Shop is a catalog business providing a wide variety of aviation products to pilots through
the world. Maynard Shephard, the recently hired assistant controller, has been asked to develo
cost function to forecast shipping costs. The previous assistant controller had forecast ship
department costs each year by plotting cost data against direct labor-hours for the most recen
months and visually fitting a straight line through the points. The results were not satisfactory.

After discussions with the shipping department personnel, Maynard decided that shipping c
could be more closely related to the number of orders filled. He based his conclusion on the
that 10 months ago the shipping department added some automated equipment. Furthermore
believes that using linear regression analysis will improve the forecasts of shipping costs. C
data for the shipping department have been accumulated for the last 25 weeks. He ran
regression analyses of the data, one using direct labor-hours, and one using the number of car
shipped. The information from the two linear regressions follows:

Regression 1 Regression 2
Equation SC = 804.3 + 15.68DL SC = 642.9 + 3.92NR
R-squared 0.3650 0.7290
Standard error of the
estimate 2.6520 1.8840
t-value 1.8900 3.4600

where: SC = total shipping department costs

DL = total direct labor-hours
NR = number of cartons shipped

Required

1. Identify which cost function (regression 1 or regression 2) that Pilot Shop should adap
forecasting total shipping department costs and explain why.
2. If Pilot Shop projects that 600 orders will be filled the coming week, calculate the t
shipping department costs using the regression you selected in requirement 1.
3. Explain two or three important limitations of the regression you selected in requiremen
and identify one or two ways to address the limitations. Specifically include in your discus
the effect, if any, of the global nature of Pilot Shop’s business.

(CMA Adapted)

Regression 1 Regression 2
R-squared 0.365 0.729
Standard Error 2.652 2
Constant 804.3 642.9
Coefficient 15.68 3.92
t-value 1.89 3.46
Number of orders to be filled 600

Solution
 products to pilots throughout
has been asked to develop a
ntroller had forecast shipping
-hours for the most recent 12
s were not satisfactory.

decided that shipping costs

ed his conclusion on the fact
equipment. Furthermore, he
asts of shipping costs. Cost
last 25 weeks. He ran two
e using the number of cartons

ession 2
.9 + 3.92NR
7290

8840
4600

Pilot Shop should adapt for

ng week, calculate the total

requirement 1.
u selected in requirement 1,
ally include in your discussion
Problem 8-48 Analysis of Regression Results
Rock n’ Roll Heaven is an outdoor pavilion that presents musical performers throug
season, from late spring to early fall. Rock n’ Roll presents a diverse venue of ar
approximately 40 events each season. In order to better project its costs and expec
Rock n’ Roll uses regression analysis to project expected ticket sales for upcoming
performer. The regression results shown below are derived from the three most recen
dependent variable for Rock n’ Roll is the number of paying tickets holders for each
independent variables are:

1. whether or not this particular performer appeared at Rock n’ Roll previously (a dum
no and 1 if yes)
2. the spending on advertising targeted to the performer’s appearance
3. the performer’s local sales of CDs in the most recent year prior to their appearance
4. the number of television appearances for the performer in the most recent year
5. the number of public appearances in the U.S. by the performer in the recent year

Independent Variables Results

Regression intercept 1,224

Attendance at prior concert

Coefficient 3,445
t-value 4.11

Spending on advertising
Coefficient 0.113
t-value 1.88

Performer's CD sales
Coefficient 0.00044
t-value 1.22

Television appearances
Coefficient 898
t-value 2.4

Other Public performances

Coefficient 1,233
t-value 3.7
 Coefficient 898
t-value 2.4

Other Public performances

Coefficient 1,233
t-value 3.7

R-squared 0.88
Standard error of the estimate 2,447

Required:
1. Using the above regression, what attendance would be predicted for a perf
appeared at Rock n’ Roll previously, had 6 other public appearances but no TV ap
sales of $10 million, and Rock n’ Roll planned to spend $35,000 on advertising?
2. Evaluate the precision and reliability of the regression results shown above. Wha
do you proposed for the regression? Which variables should be deleted, and whi
should be added, and why?

Data Input

Independent Variables Results

Regression intercept 1,224

Attendance at prior concert

Coefficient 3,445
t-value 4.11

Spending on advertising
Coefficient 0.113
t-value 1.88

Performer's CD sales
Coefficient 0.00044
t-value 1.22

Television appearances
 Coefficient 898
t-value 2.4

Other Public performances

Coefficient 1,233
t-value 3.7

R-squared 0.88
Standard error of the estimate 2,447

Dummary variable:
"no" = 0
"yes" = 1

Requirements

1. Using the above regression, what attendance would be predicted for a performer who ha
previously, had six other public appearances but no TV appearances, and had CD sales
planned to spend $35,000 on advertising?
2. Evaluate the precision and reliability of the regression results shown above. What chang
regression? Which variables should be deleted, and which do you think should be added

Solution

1. Estimated attendance:

Dummary variable = 1
No. of other public appearances = 6
No. of TV appearances = 0
CD sales = $10
Planned advertising = $35,000

Estimated attendance =

2. The overall reliability of the regression, as measured by R-squared is very good,

error of the estimate, at 2,447 is reasonably small, considering the level of predicted
the other hand, two of the five independent variables have unsatisfactory t-valu
2. The overall reliability of the regression, as measured by R-squared is very good,
error of the estimate, at 2,447 is reasonably small, considering the level of predicted
the other hand, two of the five independent variables have unsatisfactory t-valu
advertising variable and the CD sales variables are less than 2.0, indicating thes
significant relationship to number of ticket holders. Rock n’ Roll should consider remov
Other potentially useful variables include dummy variables for the timing of the perform
holiday weekend, early or late in the season, the prior appearance was on a rainy day
related to the performer’s popularity, such recent appearances in the print media,
single, etc.
usical performers throughout a 6 month
s a diverse venue of artists in a set of
ject its costs and expected attendance,
ket sales for upcoming events for each
om the three most recent seasons. The
tickets holders for each event, and the

n’ Roll previously (a dummy variable, 0 if

pearance
prior to their appearance
he most recent year
mer in the recent year
be predicted for a performer who had
pearances but no TV appearances, CD
00 on advertising?
ults shown above. What changes if any
ld be deleted, and which do you think
ed for a performer who had appeared at Rock n' Roll
ances, and had CD sales of $10 million, and Rock n' Roll

hown above. What changes, if any, do you propose for the

ou think should be added, and why?

million

20,422

R-squared is very good, at 88% and the standard

ng the level of predicted attendance, 20,422. On
ve unsatisfactory t-values. The t-values for the
R-squared is very good, at 88% and the standard
ng the level of predicted attendance, 20,422. On
ve unsatisfactory t-values. The t-values for the
han 2.0, indicating these variables have a non-
l should consider removing them from the model.
the timing of the performer’s appearance (near a
ance was on a rainy day, etc), and other variables
ces in the print media, release of a new CD or
Problem 8-49 Correlation Analysis
PolyChem is a large manufacturer of packaging materials for supermarkets a
applications; the packages are used by customers to carry away their purchases.
succeeded for many years by providing a high quality product and superior cu
Recently, additional competitors have entered the market, both local and from othe
PolyChem is finding that it must increasingly compete on price. PolyChem’s stra
with the increased competition is to market its products to smaller retailers that w
the firm’s quality and service, as well as the firm’s ability to customize the product –
designs and colors to the packaging material. Until recently, the firm determine
based on simple averages of materials purchases, plant labor and overhe
management is now interested in improving, if possible, the accuracy of its cost in
start, Cheryl Greenberg, the management accountant, obtains the following sample
1) from the plant manager, showing the machine number, the order size (quantity
the machine setup time (in hours per unit; setup time also includes clean up time a
run), runtime (the operating time for the machine to produce the order), and a
complexity of the order based on a subjective rating where 1= less complex and 2 i
(complexity relates to the number and type of images and colors printed on the pack

Cheryl wants to run some regression analyses to better understand this data and
obtains a correlation analysis which shows the simple correlation between each of
Table 1. The results are shown in Table 2. Cheryl understands that each o
numbers in Table 2 is equivalent to the R-squared for a simple linear regressi
variable, as follows: (correlation between two variables) 2 = the R-squared for si
analysis between these two variables. To illustrate, note that the correlation be
number and order quantity = -0.33919. The R-squared for the regression betw
variables (with either as the dependent variable) is (-0.33919)2 = 0.1151. Cheryl al
negative correlation means that that the two variables are inversely related – when
the other decreases.

Table 1 Plant Data for PolyChem

Machine Order Order Per Unit

Number Size Complexity Setup time Runtime
2 480 1 0.002 0.042
2 489 1 0.000 0.043
2 480 2 0.005 0.042
4 180 1 0.004 0.040
4 2160 1 0.002 0.035
4 1377 1 0.002 0.040
4 120 2 0.004 0.040
4 540 1 0.003 0.041
4 360 2 0.014 0.041
 2 489 1 0.000 0.043
2 480 2 0.005 0.042
4 180 1 0.004 0.040
4 2160 1 0.002 0.035
4 1377 1 0.002 0.040
4 120 2 0.004 0.040
4 540 1 0.003 0.041
4 360 2 0.014 0.041
4 1080 2 0.011 0.038
4 300 1 0.004 0.043
4 2400 2 0.005 0.035
4 81 2 0.046 0.041
8 360 1 0.002 0.043
8 120 1 0.002 0.043
8 120 2 0.007 0.042
8 60 2 0.008 0.042
8 240 1 0.008 0.043
8 60 2 0.005 0.047

Table 2: Correlation Results for PolyChem’s Plant Data

Setup
Number Order Size Complexity time Runtime
Number 1
Order Size -0.33919 1
Complexity 0.071001 -0.07095 1
Setup time -0.03805 -0.20952 0.4521388 1
Runtime 0.346651 -0.80882 -0.140537 -0.06534 1

Note to Table 2: correlations with absolute value > 0.4 are statistical
0.10; correlations with absolute value > 0.5 are statistically significant at

Requirements
1. Analyze the findings in Table 2 and asses how, if at all, order size and complexi
and runtime. What other findings in Table 2 are of particular interest?
2. How can your analysis in 1 above help PolyChem become more competitive?

Solution
1. The correlation analysis shows that only one of the correlations is significant at the
size and runtime, and the relationship is negative, or inverse. That is, the larger t
smaller the runtime per unit. Based on an actual company, this result is due to the fac
operators slowed the machine time at the start of each order to ensure that the o
properly before getting the machine up to the normal runtime speed. The effect of t
larger orders saved the company In two ways. First, it should reduce the average p
(there is evidence of this in the data, but the correlation of -0.209 is not statistically
setup time varies per batch (order) and not by units in the order. Second, the larger
machine operators to operate the machines at a higher than average speed relativ
speeds for the smaller orders, thus saving runtime on the larger orders.

Another informative aspect of the correlation analysis is to show the positive (0.4
significant (p=0.08) relationship between complexity and setup time per unit. This
complexity tends to increase setup time, and intuitive result.

2. The information above is particularly useful to PolyChem as it begin to focus on sm

order to find profitable alternatives to the low-cost competition it now faces. The key po
smaller and more customized orders will increase setup and runtime costs, as il
correlation analysis. Smaller orders lead to slower runtime, and more complex orders le
time, for higher overall unit costs for these smaller orders. The company should consid
and other cost control measures to ensure the success of this new strategy.

This problem is adapted from the Laurent Company case, which is included in
accompanies this text.
als for supermarkets and other retail
away their purchases. PolyChem has
roduct and superior customer service.
both local and from other countries, and
price. PolyChem’s strategy for dealing
smaller retailers that would appreciate
ustomize the product – adding different
tly, the firm determined product costs
ant labor and overhead. The firm’s
e accuracy of its cost information. As a
ins the following sample of data (Table
the order size (quantity, in thousands),
ncludes clean up time after the order is
uce the order), and a measure of the
1= less complex and 2 is more complex
lors printed on the packaging material).

nderstand this data and as a first step

elation between each of the variables in
nderstands that each of the correlation
simple linear regression between the
= the R-squared for simple regression
that the correlation between machine
for the regression between these two
9)2 = 0.1151. Cheryl also recalls that a
nversely related – when one increases,
 Setup
time Runtime

1
-0.06534 1

lue > 0.4 are statistically significant at p <

statistically significant at p < 0.05

order size and complexity affect setup time

ar interest?
e more competitive?
ations is significant at the 0.05 level – order
se. That is, the larger the order size, the
his result is due to the fact that the machine
der to ensure that the order was running
me speed. The effect of this practice is that
uld reduce the average per unit setup time
-0.209 is not statistically significant) since
der. Second, the larger orders allowed the
n average speed relative to the operating
rger orders.

show the positive (0.452) and marginally

up time per unit. This means that greater

as it begin to focus on smaller customers in

it now faces. The key point is that selling in
nd runtime costs, as illustrated from the
d more complex orders lead to longer setup
e company should consider pricing policies
s new strategy.

e, which is included in the casebook that

Problem 8-50 Regression Analysis
United States Motors Inc. (USMI) manufactures automobiles and light trucks and distributes
sale to consumers through franchised retail outlets. As part of the franchise agreement, de
must provide monthly financial statements following the USMI accounting procedures manu
has developed the following financial profile of an average dealership that sells 1,500 new
annually.
AVERAGE DEALERSHIP FINANCIAL PROFILE
Composite Income Statement

Sales $30,000,000
Cost of goods sold 24,750,000
Gross profit $5,250,000
Operating costs
Variable expenses 862,500
Mixed expenses 2,300,000
Fixed expenses 1,854,000
Operating income $233,500

USMI is considering a major expansion of its dealership network. The vice president of m
has asked Jack Snyder, corporate controller, to develop some measure of the risk associa
the addition of these franchises. Jack estimates that 90 percent of the mixed expenses sh
variable for purposes of this analysis. He also suggested performing regression analyse
various components of the mixed expenses to more definitively determine their variability.

Required

1. Calculate the composite dealership profit if 2,000 units are sold.

2. Assume that regression analyses were performed on the separate components of the mix
expenses and that a coefficient of determination value of 0.60 was determined as applica
aggregate mixed expenses over the relevant range.
a. Define the term relevant range.
b. Explain the significance of an R-squared value of 0.60 to USMI’s analysis.
c. Describe the limitations that may exist in applying the composite-based relation
specific new dealerships that have been proposed.
d. Define the standard error of the estimate.
3. The regression equation that Jack Snyder developed to project annual sales of a dealer
an R-squared of 60 percent and a standard error of the estimate of $4,500,000. If the p
annual sales for a dealership total $28,500,000, determine the approximate 95
confidence range for Jack’s prediction of sales.
4. What is the strategic role of regression analysis for USMI?

(CMA Adapted)
Input Data

Annual sales volume, average dealership =

Sales Revenue
CGS
Gross Profit
Operating expenses:
Variable $862,500
Mixed $2,300,000
Fixed $1,854,000
Operating income

% of mixed costs that are variable =

estimated sales volume (part 1) =

Std error of the regression (part 3) =

Projected sales dollars (part 3) =
No. std deviations (for CI, part 3) =

Solution

1. Assuming that all purchases of autos for resale (cost of good sold) represent variable costs:

Selling price = $20,000

Variable cost per unit:
CGS = $16,500
Operating costs = $575
Mixed operating costs = $1,380 $18,455
Total Fixed Costs $2,084,000

Operating profit when x = 2,000 units per year:

Sales $40,000,000
VC $36,910,000
FC $2,084,000
Operating profit = $1,006,000

2. a. The relevant range is the band or range of activity within which specified cost relationships (beh
assumptions) remain valid and fixed costs remain fixed.

b. The R-squared value is a measure of the goodness of fit between the independent and dep
variable, the extent to which the independent variable accounts for the variability in the dep
variable. An R-squared value of 0.60 indicates that 60% of the total variation in mixed expe
explained by the regression equation.
2. a. The relevant range is the band or range of activity within which specified cost relationships (beh
assumptions) remain valid and fixed costs remain fixed.

b. The R-squared value is a measure of the goodness of fit between the independent and dep
variable, the extent to which the independent variable accounts for the variability in the dep
variable. An R-squared value of 0.60 indicates that 60% of the total variation in mixed expe
explained by the regression equation.

c. The composite-based relationships may not be realistic and could result in incorrect pred
Application of these relationships to specific new dealerships may not allow for regional varia
items such as wages and rents and may not include factors that are peculiar to the start-
dealership.

d. The standard error of the estimate is the measure of precision of the regression. The standa
of the estimate helps to determine the range of the accuracy of the estimate with a given de
confidence.

3. Using the regression that Jack Synder developed, the approximate range of sales that could occur
calculated below:

Range of sales = Estimated (mean) sales +/- (std error x std dev for specified CI)
=$28,500,000 +/- ($4,500,000 x 2)

Upper Limit = $37,500,000

Lower Limit = $19,500,000

Note that this is a relatively wide range for the prediction. Consider that the ratio o
amount predicted is $4,500/$28,500 = 15.8%, an indication of a relatively poor SE
regression. The R-squared value of 60% is further indication of the weakness of t
of the lack of precision of the predictions generated by using the model.

4. A key issue for USMI is the risk of expanding its dealership network. Regression analysis allows fi
predictions about the effect of the proposed expansion on sales and profits. While Jack Snyder's m
or precise (part 3 above), the approach is certainly worthwhile, especially if Jack can determine a w
improve its reliability and precision, perhaps by incuding better independent variables. Strategicall
predict sales, using regression and/or other methods, will be in a stronger competitive position--the
focused and effective.
ight trucks and distributes them for
franchise agreement, dealerships
ounting procedures manual. USMI
ship that sells 1,500 new vehicles

AL PROFILE
ment

k. The vice president of marketing

easure of the risk associated with
of the mixed expenses shown are
rming regression analyses on the
termine their variability.

ate components of the mixed

was determined as applicable to

USMI’s analysis.
composite-based relationships to

t annual sales of a dealership has

ate of $4,500,000. If the projected
ne the approximate 95 percent
 1,500

$30,000,000
$24,750,000
$5,250,000

$5,016,500
$233,500

90.00%
2,000

$4,500,000
$28,500,000
2

esent variable costs:

cified cost relationships (behavioral

n the independent and dependent

or the variability in the dependent
tal variation in mixed expenses is
cified cost relationships (behavioral

n the independent and dependent

or the variability in the dependent
tal variation in mixed expenses is

uld result in incorrect predictions.

not allow for regional variations in
t are peculiar to the start-up of a

he regression. The standard error

e estimate with a given degree of

ge of sales that could occur during the year is

or specified CI)

on. Consider that the ratio of the standard error to the

ation of a relatively poor SE and a relatively poor
ication of the weakness of the model, and therefore
y using the model.

egression analysis allows financial managers to make

ofits. While Jack Snyder's model is not particularly reliable
lly if Jack can determine a way to modify his model to
dent variables. Strategically, firms that are better able to
ger competitive position--the firm's planning will be more
Problem 8-51 Cost Estimation, High-Low Method, Regression Analysis
Problem Information

DVD Express is a large manufacturer of affordable DVD players. Management recently became aware
costs resulting from returns of malfunctioning products. As a starting point for further analysis, Bridget
the controller, wants to test different forecasting methods and then use the best one to forecast quarte
for 2010. The relevant data for the previous three years is as follows:

2007 Return 2008 Return 2009 Return

Quarter Expenses Quarter Expenses Quarter Expenses
1 $15,000 1 $16,200 1 $16,600
2 $17,500 2 $17,800 2 $18,100
3 $18,500 3 $18,800 3 $19,000
4 $18,600 4 $17,700 4 $19,200

The result of a simple regression analysis using all 12 data points yielded the following:

Intercept term $16,559

Coefficient estimate $183.22
R-squared 0.27
t-statistic 1.94
SE 1,128

8-51 Requirements

1. Calculate the quarterly forecast for 2010 using the high-low method and regression analysis. Recommend which m
Bridget should use.
2. How does your analysis in requirement 1 change if DVD Express manufactures its products in multiple global prod
facilities to serve the global market?

Solution
 recently became aware of rising
urther analysis, Bridget Forrester,
st one to forecast quarterly expenses

is. Recommend which method

ts in multiple global production

Problem 8-52 Regression Analysis; Use of the Internet
Economists and business planners often need to make projections of interest rates in orde
plan for the purchase of equipment and other assets. The timing of these investments is a
firm’s financial management and can have a dramatic effect on the firm’s profitability. This
important during times of liquidity problems as we have had in the financial markets. As a
retail firm that requires significant amounts of seasonal borrowing, you are interested in for
rate for the next several months. You think you can get a good prediction for the consume
(CPI) and the unemployment rate months ahead, based on other studies and economic fo
you have access. You have decided to use regression analysis to develop a model to pre
from the CPI and the unemployment rate. The data for monthly interest rate data (prime r
Federal Reserve Board can be found at the FED website (http://www.stls.frb.org/fred/data/
and the employment and CPI information is provided by the U.S. Bureau of Labor Statistic
(http://www.bls.gov/data/home.htm). The data for 24 months from October 2006 through
provided below.

Prime Rate CPI

10/1/2006 8.25 604.6
11/1/2006 8.25 603.6
12/1/2006 8.25 604.5
1/1/2007 8.25 606.3
2/1/2007 8.25 609.6
3/1/2007 8.25 615.4
4/1/2007 8.25 619.1
5/1/2007 8.25 622.9
6/1/2007 8.25 624.1
7/1/2007 8.25 624
8/1/2007 8.25 622.8
9/1/2007 8.03 624.5
10/1/2007 7.74 625.9
11/1/2007 7.50 629.6
12/1/2007 7.33 629.2
1/1/2008 6.98 632.3
2/1/2008 6.00 634.1
 3/1/2008 5.66 639.6
4/1/2008 5.24 643.5
5/1/2008 5.00 648.9
6/1/2008 5.00 655.5
7/1/2008 5.00 658.9
8/1/2008 5.00 656.3
9/1/2008 5.00 655.4

Required

1. Develop a regression model to predict interest rates using the Bureau of Labor statis
Reserve Board data above. Evaluate the results of the regression.

2. Use the model you developed to predict the prime rate for December 2008 if the CP
be 670 and the unemployment rate is expected to be 6.6 in December.

3. What other economic data can you find on the web that would help in predicting the
rate? Include the web links in your answer, if appropriate.

4.To follow up on the regression analysis in part (1), go to the web sites indicated above
most recent data, and re-run the regressions. Compare your results to the finding for th
part (1). Why the difference, if any?

Solution
Note: This problem was written in September 2008, and the solution below reflects
the period through September 2008. The solution also includes a presentation of th
through February 2009, the latest data at the time the solutions manual was prepare

There are a variety of possible answers to this question. In the example solution ab
taken the consumer price index and the unemployment rate as plausible predictors
prime loan rate. We hypothesize that an increase in unemployment is likely a sign
downturn which will signal FED officials to consider reducing the FED rate, leading t
in the prime loan rate. We also hypothesize a positive relationship between the con
index (CPI) and the loan rate; an increase in inflation will prompt the FED to raise in
The results are shown in the regression report below.

Regression: the regression equation (from Excel) is as follows:

 Regression Statistics
Multiple R 0.92778207
R Square 0.86077957
Adjusted R Square 0.84752048
Standard Error 0.54523148
Observations 24

ANOVA
df SS
Regression 2 38.5984712281231
Residual 21 6.24282460521019
Total 23 44.8412958333333

Coefficients Standard Error

Intercept 43.6186304 7.55113239545229
CPI -0.05165308 0.015644105628544
Unemployment Rate -0.82532909 0.530750123206441

The results show that the CPI variable is significant with a p-value less than
.01. The relationship between the CPI and the loan rate is negative,
indicating that increases in the CPI tend to be associated with decreases in
the loan rate, not predicted in our hypothesis.

The relationship between loan rate and the unemployment rate is negative
as expected, indicating that a rise in the unemployment rate is associated
with a fall in the prime loan rate, but the relationship is not significant
(p=.123).
Data for the problem are contained in the imbedded Excel
spreadsheet below; double-click and copy the data for your own use.
Note that the data includes an update for the months through
February 2009, as explained later in this solution note.

Prime Rate Unemployment Rate

10/1/2006 8.25 4.4
11/1/2006 8.25 4.5
12/1/2006 8.25 4.4
1/1/2007 8.25 4.6
2/1/2007 8.25 4.5
 3/1/2007 8.25 4.4
4/1/2007 8.25 4.5
5/1/2007 8.25 4.5
6/1/2007 8.25 4.6
7/1/2007 8.25 4.7
8/1/2007 8.25 4.7
9/1/2007 8.03 4.7
10/1/2007 7.74 4.8
11/1/2007 7.50 4.7
12/1/2007 7.33 5.0
1/1/2008 6.98 4.9
2/1/2008 6.00 4.8
3/1/2008 5.66 5.1
4/1/2008 5.24 5.0
5/1/2008 5.00 5.5
6/1/2008 5.00 5.6
7/1/2008 5.00 5.8
8/1/2008 5.00 6.2
9/1/2008 5.00 6.2
10/1/2008 4.56 6.6
11/1/2008 4.00 6.8
12/1/2008 3.61 7.2
1/1/2009 3.25 7.6
2/1/2009 3.25 8.1

The overall results for the regression are very significant with an R-squared
of .86 and a p-value less than .01. To further study the regression results
we obtained the correlation matrix for the three variables, as follows:

Prime Rate CPI

Prime Rate 1
CPI -0.9191021 1
Unemployment Rate -0.88797906 0.908601093570959
Note the high correlation between each of the independent variables and the de
variable. Also, note the high correlation, .9086, between the two independent
variables. This indicates that the two variables are both strong predictors of the
dependent variable, and that one of the independent variables will tend to domin
other in the multiple regression. The results are likely to be improved by using s
 Note the high correlation between each of the independent variables and the de
variable. Also, note the high correlation, .9086, between the two independent
variables. This indicates that the two variables are both strong predictors of the
dependent variable, and that one of the independent variables will tend to domin
other in the multiple regression. The results are likely to be improved by using s
regressions for each variable. A simple regression for each variable is shown b

Simple Regression for CPI only:

Regression Statistics
Multiple R 0.9191021
R Square 0.84474867
Adjusted R Square 0.83769179
Standard Error 0.56252962
Observations 24

ANOVA
df SS
Regression 1 37.8796251056907
Residual 22 6.96167072764263
Total 23 44.8412958333333

Coefficients Standard Error

Intercept 53.469207 4.24031247977241
CPI -0.07375657 0.006741294543187

Simple Regression for Unemployment rate only

Regression Statistics
Multiple R 0.88797906
R Square 0.78850682
Adjusted R Square 0.77889349
Standard Error 0.65656227
Observations 24

ANOVA
df SS
Regression 1 35.3576675531488
Residual 22 9.48362828018458
 Total 23 44.8412958333333

Coefficients Standard Error

Intercept 18.9490833 1.31595908967553
Unemployment Rate -2.41757008 0.26693971309145

Looking at the three regressions together, it is clear they are very much the
same, but the regression with both variables is slightly better for R-squared
and SE. Since our focus is on predicting the dependent variable, and not
on interpreting the coefficients of the independent variables, we choose to
use the combined equation.

At March 31, 2009, the time of this writing, the available prime rate, CPI,
and unemployment rate data were as follows (note that the unemployment
rates for June through September 2008 were adjusted by the Bureau of
Labor Statistics; the CPI and prime rate data were unchanged):

Month Prime Rate CPI

6/2008 5.00 655.5
7/2008 5.00 658.9
8/2008 5.00 656.3
9/2008 5.00 655.4
10/2008 4.56 648.8
11/2008 4.00 636.3
12/2008 3.61 629.8
1/2009 3.25 632.5
2/2009 3.25 635.6
3/3009 3.25 not available

Using the regression above with both the CPI and Unemployment
variables, we would predict the December 2008’s prime rate to be:

43.6186 - .05165 x 629.8 - .8253 x 7.2 = 5.15%

The actual rate in December 2008 was 3.61%, quite a difference from the
model’s prediction. The reason for the difference is the sharp decrease in
the CPI relative to the pattern of the prior months. For comparison
purposes, the CPI-only model predicted the December prime rate at 7.02%
and the Unemployment-rate-only model predicted the December prime rate
at 1.54%. Note how the CPI-only model predicts too high because of the
steep fall in the CPI during these months, while the Unemployment-only
model predicts too low because of the steep increase in unemployment
during these months. The model with both predictors turned out to be the
 the CPI relative to the pattern of the prior months. For comparison
purposes, the CPI-only model predicted the December prime rate at 7.02%
and the Unemployment-rate-only model predicted the December prime rate
at 1.54%. Note how the CPI-only model predicts too high because of the
steep fall in the CPI during these months, while the Unemployment-only
model predicts too low because of the steep increase in unemployment
during these months. The model with both predictors turned out to be the
best overall model.

A second regression for both CPI and Unemployment rate was run using all
data through February 2009

Regression Statistics
Multiple R 0.958279708537546
R Square 0.918299999794804
Adjusted R S0.912015384394404
Standard Err0.539900022810993
Observations 29

ANOVA
df SS MS
Regression 2 85.1848899 42.5924449291033
Residual 26 7.5787929 0.29149203463131
Total 28 92.7636828

Coefficients Standard Error t Stat

Intercept 36.4082857590012 4.41617073 8.24431118668505
Unemploymen
-1.27761219742752 0.11616344 -10.9984022262341
CPI -0.0366472377489 0.00749795 -4.88763428387757

Using the revised regression above, the predicted prime rate for Decembe
2008 is 4.13%, a much more accurate prediction than any of the
regressions based only on data through September 2008. The revised
regression appears to capture some of the changes in the latter part of
2008 which are critical for accurate predictions.
Overall, the results show the sensitivity of the regression model to sharp
increases in the underlying data used in the model. This indicates how
important it is to validate the model with other information, to use as
complete a set of independent variables as possible, and to continuously
update the model as the underlying data change, especially in a time of
rapid change as was true in the latter part of 2008 through early 2009.

For comparison to a more stable period, see the results (shown below) of
this analysis for the years prior to July 2006.
complete a set of independent variables as possible, and to continuously
update the model as the underlying data change, especially in a time of
rapid change as was true in the latter part of 2008 through early 2009.

For comparison to a more stable period, see the results (shown below) of
this analysis for the years prior to July 2006.

Loan Rate
Jan 2004 4.00
Feb 2004 4.00
Mar 2004 4.00
Apr 2004 4.00
May 2004 4.00
Jun 2004 4.01
Jul 2004 4.25
Aug 2004 4.00
Sep 2004 4.00
Oct 2004 4.43
Nov 2004 4.93
Dec 2004 5.15
Jan 2005 5.25
Feb 2005 5.49
Mar 2005 5.58
Apr 2005 5.76
May 2005 5.98
Jun 2005 6.01
Jul 2005 6.25
Aug 2005 6.44
Sep 2005 6.59
Oct 2005 6.75
Nov 2005 7.00
Dec 2005 7.15
Jan 2006 7.26
Feb 2006 7.50
Mar 2006 7.53
Apr 2006 7.76
May 2006 7.93
Jun 2006 8.02
Jul 2006 8.25

Regression Statistics
Multiple R 0.97682931
R Square 0.9541955
Adjusted R Square 0.95092375
Standard Error 0.32504064
Observations 31

ANOVA
df SS
Regression 2 61.6257989279576
Residual 28 2.95823978171978
Total 30 64.5840387096774

Coefficients Standard Error

Intercept -20.00957 9.88314432296763
CPI 0.05827348 0.012248468210166
Unemp Rate -1.56943283 0.555312101376202

The results for 2004-2006 show that both variables are significant, with p-
values less than .01 for each. The relationship between loan rate and the
unemployment rate is negative (as it was in 2006-2008), indicating that a
rise in the unemployment rate is associated with a fall in the prime loan
rate. In contrast to 2006-2008, the relationship between the CPI and the
loan rate is positive, indicating that increases in the CPI tend to be
associated with increases in the loan rate, as we expected in our original
hypothesis.

2.
The model for predicting the prime rate is:
Prime Rate = 43.6186 – 0.05165 x CPI – 0.8254 x Unemployment Rate
For the data given for December:
Prime Rate = 43.6186 – 0.05165 x 670 – 0.8254 x 6.6 = 3.565%

As noted in part above, while these predictions for the CPI and
unemployment rate might have been reasonable in September 2008, they
were quite off the mark in December 2008.
For the data given for December:
Prime Rate = 43.6186 – 0.05165 x 670 – 0.8254 x 6.6 = 3.565%

As noted in part above, while these predictions for the CPI and
unemployment rate might have been reasonable in September 2008, they
were quite off the mark in December 2008.

The instructor may choose to require a prediction for any month

(other than December 2008), a month which might be more applicable
to the date of the assignment.

3. Other sources of data include for example data from the US Statistical
abstract, and the Bureau of Economic Statistics, though these data are onl
annual or quarterly, respectively.

US Statistical abstract (annual data)

http://www.census.gov/statab/www/

US Bureau of Econ Stats (quarterly data)

http://www.bea.doc.gov/bea/dn/nipaweb/SelectedTables.asp?Selected=Y

4. Answers will vary depending on the time periods chosen. The solution
for data through February 2009 is shown in part 1 above.
 ions of interest rates in order to effectively
ing of these investments is a critical part of the
n the firm’s profitability. This is particularly
the financial markets. As a CFO of a large
ng, you are interested in forecasting the prime
d prediction for the consumer price index
er studies and economic forecasts to which
is to develop a model to predict the prime rate
ly interest rate data (prime rate) from the
//www.stls.frb.org/fred/data/irates/mprime)
S. Bureau of Labor Statistics
rom October 2006 through September 2008 is

CPI for all urban

areas
Unemployment
Rate
4.4
4.5
4.4
4.6
4.5
4.4
4.5
4.5
4.6
4.7
4.7
4.7
4.8
4.7
5.0
4.9
4.8
 5.1
5.0
5.5
5.5
5.7
6.1
6.1

g the Bureau of Labor statistics and Fed eral

ession.

or December 2008 if the CPI is expected to

December.

would help in predicting the prime interest

e web sites indicated above, and include the

r results to the finding for the regression in

he solution below reflects the data for

cludes a presentation of the results
utions manual was prepared.

In the example solution above, We have

ate as plausible predictors of the monthly
employment is likely a sign of economic
ing the FED rate, leading to a reduction
elationship between the consumer price
prompt the FED to raise interest rates.
 MS F Significance F
19.2992356141 64.92 1.02066043E-09
0.29727736215

t Stat P-value
5.7764356478 9.83E-06
-3.30175954482 0.0034
-1.55502382679 0.13488

a p-value less than

s negative,
with decreases in

nt rate is negative
ate is associated
t significant

Excel
or your own use.
through
e.

CPI
604.6
603.6
604.5
606.3
609.6
 615.4
619.1
622.9
624.1
624
622.8
624.5
625.9
629.6
629.2
632.3
634.1
639.6
643.5
648.9
655.5
658.9
656.3
655.4
648.8
636.3
629.8
632.5
635.6

nificant with an R-squared

dy the regression results
ariables, as follows:

Unemployment Rate

1
ndent variables and the dependent
ween the two independent
oth strong predictors of the
variables will tend to dominate the
to be improved by using simple
ndent variables and the dependent
ween the two independent
oth strong predictors of the
variables will tend to dominate the
to be improved by using simple
r each variable is shown below.

MS F Significance F
37.8796251057 119.706 2.2937728E-10
0.31643957853

t Stat P-value
12.6097327143 1.53E-11
-10.9410105643 2.29E-10

MS F Significance F
35.3576675531 82.0223 7.09441659E-09
0.43107401274
 t Stat P-value
14.3994470792 1.11E-12
-9.05661451639 7.09E-09

ar they are very much the

ghtly better for R-squared
ndent variable, and not
variables, we choose to

ilable prime rate, CPI,

e that the unemployment
sted by the Bureau of
unchanged):

Unemp. Rate
5.6
5.8
6.2
6.2
6.6
6.8
7.2
7.6
8.1
8.5

Unemployment
prime rate to be:

te a difference from the

s the sharp decrease in
For comparison
mber prime rate at 7.02%
the December prime rate
oo high because of the
e Unemployment-only
ase in unemployment
ors turned out to be the
 For comparison
mber prime rate at 7.02%
the December prime rate
oo high because of the
e Unemployment-only
ase in unemployment
ors turned out to be the

ent rate was run using all

F Significance F
146.118726651 7.23E-15

P-value Lower 95%

1.00168805E-08 27.3307
2.8194029E-11 -1.51639
4.52063871E-05 -0.05206

ed prime rate for December

n than any of the
ber 2008. The revised
ges in the latter part of

gression model to sharp

del. This indicates how
formation, to use as
sible, and to continuously
e, especially in a time of
08 through early 2009.

results (shown below) of

sible, and to continuously
e, especially in a time of
08 through early 2009.

results (shown below) of

CPI Unemp Rate

554.9 5.7
557.9 5.6
561.5 5.7
563.2 5.5
566.4 5.6
568.2 5.6
567.5 5.5
567.6 5.4
568.7 5.4
571.9 5.4
572.2 5.4
570.1 5.4
571.2 5.2
574.5 5.4
579 5.1
582.9 5.1
582.4 5.1
582.6 5.0
585.2 5.0
588.2 4.9
595.4 5.1
596.7 4.9
592 5.0
589.4 4.9
593.9 4.7
595.2 4.8
598.6 4.7
603.5 4.7
606.5 4.6
 607.8 4.6
609.6 4.8

MS F Significance F
30.812899464 291.647 1.78942881E-19
0.10565142078

t Stat P-value Lower 95%

-2.02461578001 0.05254 -40.25427309876
4.75761386956 5.38E-05 0.033183632812
-2.82621759806 0.00859 -2.706938092642

les are significant, with p-

between loan rate and the
6-2008), indicating that a
a fall in the prime loan
between the CPI and the
the CPI tend to be
e expected in our original

254 x Unemployment Rate

254 x 6.6 = 3.565%

or the CPI and

e in September 2008, they
254 x 6.6 = 3.565%

or the CPI and

e in September 2008, they

ction for any month

might be more applicable

ta from the US Statistical

though these data are only

dTables.asp?Selected=Y

ods chosen. The solution

1 above.
Problem 8-53 Learning Curves
Business schools have commonly observed that when job opportunities are down, those
business seek to enter MBA programs . To test this hypothesis, we decided to look at th
applicants for the Graduate Management Admissions Test (GMAT) which is required for
M.B.A. programs and to compare that to the unemployment rate. The GMAT data is onl
annual basis, so we also use annual unemployment data. The GMAT data is from the G
Management Admissions Council (http://www.gmac.com/gmac) while the unemploymen
the website of the U.S. Bureau of Labor Statistics (http://www.bls.gov/data/home.htm). W
both the total unemployment rate and the rate for college graduates 25 years of age or o
2001 through September 2008 is shown below. The unemployment rate is the median v
months in the year (or the 9 months in 2008). The percent change (from prior year) in G
from the GMAC’s publication “2008 Application Trends Survey.”

Unemployment Rate % Change in

Unemployment Rate(%) for College Grads GMAT Takers
2000 4.0 1.6 4
2001 4.6 2.2 3
2002 5.8 2.9 12
2003 6.0 3.1 (7)
2004 5.5 2.7 (9)
2005 5.1 2.3 (4)
2006 4.6 2 6
2007 4.7 2.1 5
2008 5.5 2.3 10

Required

1. Use regression on the above data to determine whether there is a relationship betwe
unemployment rate and the number of GMAT exam takers. Use both measures of u
see if there is a difference in your results. State whether you expect the relationship
negative (inverse).
2. Assume that the relationship between the unemployment rate and the number of GM
lagged relationship, and a change in the unemployment rate in one year leads to a c
takers in the following year. Use regression again, and compare your results to thos

Solution
s are down, those interested in
cided to look at the number of
ch is required for application to most
GMAT data is only available on an
data is from the Graduate
he unemployment data is taken from
ata/home.htm). We have data for
years of age or older. The data for
e is the median value for the 12
m prior year) in GMAT exam takers is

elationship between the

oth measures of unemployment and
ct the relationship to be positive or

he number of GMAT takers is a

year leads to a change in GMAT
our results to those in part (1).
Problem 8-54 Learning Curves
The Air Force Museum Foundation has commissioned the purchase of 16 Four F Six
aircraft. They will be built completely from scratch to the exact specifications used
further authentication, the aircraft will be made using the technology and manufacturin
when the originals were built. Each of the 16 will be flown to Air Force and aviation mu
country for exhibition. Aviation enthusiasts can also visit the production facility to s
aircraft were built in 1938.
Soren Industries wants to bid on the aircraft contract and asked for and received ce
about the Four F Sixes from the Air Force. The information includes some of the o
builders of the original aircraft. The available information is for the total accumulated tim
and thirty-second aircraft, respectively, were completed.

Output Total Hours

1 250
8 1,458
32 4,724

Required

1. If Soren Industries expects that the manufacturing time will be the same as it wa
hours will it take to build the 16 aircraft for the Air Force Museum Foundation?
2. What is the role of learning curves in Soren Industries’ business for contracts such a

Input Data

Output Total Hours

1 250
8 1,458
32 4,724

Requirements

1. If Soren Industries expects that the time spent per unit will be the same as it was in 1938
the 16 aircraft for the Air Force Museum Foundation?
2. What is the role of learning curves in Soren Industries' business for contracts such as th

Solution

The average production hours per unit obviously decreased as the output increased. This d
that of a 90 learning curve.

1. An estimate of the hours required to build 16 aircraft is 2,624 hours, as follows:

Estimated learning curve = 0.90

Output Avg. Time Total Time

1 250 250
2 225 450
4 203 810
8 182 1,458
16 164 2,624

2. The role of learning curves is to help predict future costs when

significant learning takes place in the work. When learning is present,
unit costs increase at a nonlinear, decreasing rate, so that linear
estimation methods such as regression and the high-low method are
not as appropriate. The learning curve method takes into account the
nonlinear learning behavior in the situation.
e purchase of 16 Four F Sixes, pre–World War II
e exact specifications used for the originals. As
echnology and manufacturing processes available
o Air Force and aviation museums throughout the
the production facility to see exactly how such

d asked for and received certain cost information

ion includes some of the old cost data from the
for the total accumulated time as the first, eighth,

e will be the same as it was in 1938, how many

seum Foundation?
usiness for contracts such as this?

e the same as it was in 1938, how many hours will it take to build
ness for contracts such as this?

s the output increased. This decrease corresponds very closely to

4 hours, as follows:

ct future costs when

n learning is present,
rate, so that linear
high-low method are
akes into account the
Problem 8-55 Learning Curves
Ben Matthews and David Everhart work for a landscaping company in Twin Citie
Their principal job is to lay railroad ties to line the sidewalks around apartment com
install flower boxes. The first time Ben and David undertook one of these projects, t
hours. Their goal by the end of the summer was to be able to finish an apartment
hours, one working day. They performed eight of these jobs and had an 80 percent le
Assume that all apartment complexes are approximately the same size.

Required Did they reach their goal? If not, what would the learning rate have to h
them to have accomplished their goal?

Data Inputs

Project # Time
1 17

Learning curve rate = 80%

Solution
g company in Twin Cities, Oklahoma.
s around apartment complexes and to
k one of these projects, they spent 17
e to finish an apartment complex in 8
and had an 80 percent learning curve.
same size.

e learning rate have to have been for

Problem 8-56 Learning Curves
Problem Information

Emotional Headdress (EH) is a Des Moines, Iowa, manufacturer of avant garde hat
11, 2004, the company purchased a new machine to aid in producing various
Production efficiency on the new machine increases with the workforce experience
cumulative output on the new machine increases, average labor time per unit decr
of at least 3,200 units. As EH’s cumulative output doubles from a base of 100 un
labor time per unit declines by 15 percent. EH’s production varies little from month t
hats per month.

Emotional Headdress has developed a new style of men’s hat, the Morrisey, to
machine. One hundred Morrisey hats can be produced in a total of 25 labor-hou
produce each Morrisey hat are $16.25, excluding direct labor cost. EH’s direct l
Fixed costs are $8,000 per month, and EH has the capacity to produce

Decrase in Avg DLHs/Unit when output doubles 15%

Average Hats/Month: 800
Fixed cost per month $8,000
Capacity (hats per month) 3,200

New hat (the Morrisey):

Other Direct Costs/Hat: $16.25
Markup on Cost: 125%
Batch size 100
Hours per batch 25
Direct labor cost per hour $15.00

8-56 Requirements

1. Emotional Headdress wishes to set the selling price for a Morrisey hat at 125% of the ha
level of 100 units, what is the estimated selling price?

2. The company has received an order for 1,600 Morrisey hats from Smiths, Inc. Smiths is
Should the company accept Smiths' order and produce the 1,600 hats? Explain.

Solution
1. Emotional Headdress wishes to set the selling price for a Morrisey hat at 125% of the ha
level of 100 units, what is the estimated selling price?

Cumulative Average Total Labor Labor Costs/ Total Labor Total Direct
Output Labor Time Time Unit Costs Costs/Unit
100 0.25000 25 $3.75 $375.00 $20.00
200 0.21250 42.5 $3.19 $637.50 $19.44
400 0.18063 72.25 $2.71 $1,083.75 $18.96
800 0.15353 123 $2.30 $1,842.38 $18.55
1600 0.13050 209 $1.96 $3,132.04 $18.21
3200 0.11093 355 $1.66 $5,324.46 $17.91

Calculation of the first row above (100 units):

25 hours/100 hats = 0.25 hrs/hat
0.25 hours x $15/hr =$3.75/hat
100 units x $3.75/unit = $375
$3.75/unit (labor) + $16.25/unit (all else) = $20.00/unit
$8,000 fixed costs/800 units per month = $10.00 fixed cost/unit
The selling price is $ 25.00 ($20 x 1.25, from table and calculations above)
of avant garde hats and headwear. On March
roducing various established product lines.
kforce experience. It has been shown that as
time per unit decreases up to the production
a base of 100 units produced, the average
little from month to month and averages 800

, the Morrisey, to be produced on the new

l of 25 labor-hours. All other direct costs to
ost. EH’s direct labor cost per hour is $15.
acity to produce 3,200 hats per month.

t at 125% of the hat-production cost. At the production

ths, Inc. Smiths is offering $20 for each hat.

s? Explain.
t at 125% of the hat-production cost. At the production

tal Direct Total Costs/

Unit
$30.00
$29.44
$28.96
$28.55
$28.21
$27.91

lculations above)
Problem 8-57 Learning Curves
Hauser Company, a family-owned business, engineers and manufactures a line of
bikes under the trade-name Trailite. The company has been in business for almo
has maintained a profitable share of the recreational vehicle market due to its rep
quality products. In addition, Hauser’s engineering department has kept the compan
by incorporating the latest technology in the Trailite bikes. Most subassembly work
subcontracted to reliable vendors. However, the final assembly and inspection o
performed at Hauser’s plant. Hauser recently developed a new braking system
Model-500 dirt bike. Because of the company’s current availability of productio
Walsh, production manager, recommended that the first lot of the new brak
manufactured in-house rather than by subcontractors. This 80-unit production run
completed. The cumulative average labor-hours per unit for the braking system
Hauser’s experience with similar products indicates that a learning curve of 80 perc
and that the learning factor can be expected to extend only through the fourth produ
batch). Hauser’s direct labor cost is $14.50 per direct labor-hour. Its manageme
whether to continue producing the braking system in its own plant or to subcontract
Lane, Hauser’s purchasing agent, has received a proposal from MACQ, a compan
component assembly. MACQ has done work in the past for Hauser and has prove
and reliable. The terms of MACQ’s proposal are negotiable, and before beginning
them, Joyce has decided to conduct some relevant financial analysis.

8-57 Requirements

Assumed learning curve = 80% Therefore, the produ

the following schedule:

Direct labor cost/hour = $14.50

Batch size (units) = 80

80% Learning Curve

Cumul Average Total Increase Time Per
Output Time Time in Time Unit
80 60.00 4,800 4,800 60.00
160 48.00 7,680 2,880 36.00
320 38.40 12,288 4,608 28.80
 Note: the total time of 12,288 can also be derived by using the powerfunction
functions). Use the formula Y = ax-b , where Y = average time for the first unit (
learning rate of 80%. Set x = 4 and b = -0.322 in the Excel function to find 0.63
38.396 hours. Since Y = average time per unit, then 320 x 38.396 = 12,287 ho

Required

1. Hauser Company has an immediate requirement for a total of 1,000 units of the braking
direct labor costs to produce the required braking system if it manufactures the units in-h
2. A consultant has advised Joyce that the learning rate for this application might be closer
projected costs of having a 75% learning curve compared to an 80% learning curve?
3. What conditions in a manufacturing plant, if present, would offest the potential benefits o
strategic role of learning curve analysis for Hauser Company?

Solution
manufactures a line of mopeds and dirt
n in business for almost 20 years and
e market due to its reputation for high-
nt has kept the company in the forefront
Most subassembly work for the bikes is
mbly and inspection of all products is
a new braking system for the Trailite
availability of production capacity, Jim
t lot of the new braking system be
80-unit production run has now been
or the braking system was 60 hours.
arning curve of 80 percent is applicable
hrough the fourth production run (80 per
or-hour. Its management must decide
plant or to subcontract this work. Joyce
from MACQ, a company specializing in
Hauser and has proved to be efficient
and before beginning discussions with
analysis.

Therefore, the production time will likely follow

using the powerfunction in Excel (one of the "Math and Trig"
ge time for the first unit (or batch) and b = 0.322 for a
Excel function to find 0.63993; 60 hours x 0.63993 =
20 x 38.396 = 12,287 hours.

000 units of the braking system. Determine Hauser's future

anufactures the units in-house.
plication might be closer to 75%. What is the effect on
80% learning curve?
st the potential benefits of the learning curve? What is the
Problem 8-58 Cost Estimation: Regression Analysis
Background

Plantcity is a large nursery and retail store specializing in house and garden plants and supplies. Jean Raouth, the assistant
manager, is in the process of budgeting monthly supplies expense for 2010. She assumes that in some way supplies expense
is related to sales, either in units or in dollars. She has collected these data for sales and supplies expenses for June 2007
through December 2009, and has estimated sales for 2010.

Problem Information

Date Supplies Expense Sales Units Sales Dollars

June-07 $2,745 354 $2,009
July-07 $3,200 436 $2,190
August-07 $3,232 525 $2,878
September-07 $2,199 145 $1,856
October-07 $2,321 199 $2,168
November-07 $3,432 543 $2,152
December-07 $4,278 1,189 $2,463
January-08 $2,310 212 $1,999
February-08 $2,573 284 $2,190
March-08 $2,487 246 $1,894
April-08 $2,484 278 $2,134
May-08 $3,384 498 $3,210
June-08 $2,945 424 $2,850
July-08 $2,758 312 $2,265
August-08 $3,394 485 $2,435
September-08 $2,254 188 $1,893
October-08 $2,763 276 $2,232
November-08 $3,245 489 $3,004
December-08 $4,576 1,045 $3,309
January-09 $2,103 104 $2,195
February-09 $2,056 167 $2,045
March-09 $3,874 298 $2,301
April-09 $2,784 398 $2,345
May-09 $2,345 187 $1,815
June-09 $2,912 334 $2,094
July-09 $2,093 264 $1,934
August-09 $2,873 333 $2,054
September-09 $2,563 143 $1,977
October-09 $2,384 245 $1,857
November-09 $2,476 232 $2,189
December-09 $3,364 1,122 $3,433
January-10 180 $1,600
Feb-10 230 $2,000
March-10 190 $1,900
April-10 450 $2,400
May-10 350 $2,300
June-10 350 $2,300
July-10 450 $2,500
Aug-10 550 $3,000
Sept-10 300 $2,500
Oct-10 300 $2,500
Nov-10 450 $3,200
Dec-10 950 $3,900

8-58 Requirements

1. Develop the regression that Jean should use based on the above data and using the regression procedures in Excel (or an
equivalent regression software program). Evaluate the reliability and precision of the regression you have chosen.
2. What are the predicted monthly figures for supplies expense for 2010?

Solution

1. The spreadsheet regression output for Plantcity is shown in

Exhibits 8-58A, B and C. Exhibit 8-58A shows the regression which
includes both predictors, sales dollars and sales units, while Exhibit
8-58B shows sales dollars only, and Exhibit 8-58C shows sales units
only.
Exhibit 8-58A
Regression Statistics
Multiple R 0.836460729157
R Square 0.699666551421
Adjusted R Squ 0.678214162237
Standard Error 356.8016908504
Observations 31

ANOVA
df SS MS F Significance F
Regression 2 8304227.68892371 4152113.844462 32.614854476754 4.857937E-08
Residual 28 3564608.50462468 127307.4465937
Total 30 11868836.1935484

Coefficients Standard Error t Stat P-value Lower 95% Upper 95%Lower 95.0%
Upper 95.0%
Intercept 1720.993362588 410.348175379628 4.193983221678 0.0002492767569 880.43324065 2561.553 880.43324 2561.5535
Units 1.66307944346 0.351697453451073 4.728721880528 5.821969113E-05 0.9426598776 2.383499 0.9426599 2.383499
Dollars 0.21261161571 0.214591376743227 0.99077427498 0.330281101402 -0.2269588872 0.652182 -0.226959 0.6521821

Exhibit 8-58B
SUMMARY OUTPUT (Sales Dollars versus Supplies Expense)

Regression Statistics
Multiple R 0.678100394807
R Square 0.459820145438
Adjusted R Squ 0.441193253901
Standard Error 470.1909446631
Observations 31

ANOVA
df SS MS F Significance F
Regression 1 5457529.98469503 5457529.984695 24.685822888572 2.76888E-05
Residual 29 6411306.20885336 221079.5244432
Total 30 11868836.1935484

Coefficients Standard Error t Stat P-value Lower 95% Upper 95%Lower 95.0%
Upper 95.0%
Intercept 650.5468078563 451.027588894694 1.442365885977 0.1599132876272 -271.90817238 1573.002 -271.9082 1573.0018
X Variable 1 0.956144724064 0.192441985298979 4.968482956454 2.768879672E-05 0.5625566773 1.349733 0.5625567 1.3497328

Exhibit 8-58C
SUMMARY OUTPUT(Sales Units versus Supplies Expense)

Regression Statistics
Multiple R 0.830142974266
R Square 0.689137357723
Adjusted R Squ 0.678417956266
Standard Error 356.6886877876
Observations 31

ANOVA
df SS MS F Significance F
Regression 1 8179258.4136743 8179258.413674 64.288790790758 7.655438E-09
Residual 29 3689577.77987409 127226.8199957
Total 30 11868836.1935484

Coefficients Standard Error t Stat P-value Lower 95% Upper 95%Lower 95.0%
Upper 95.0%
Intercept 2112.016479739 112.329041572044 18.80205199102 8.695777825E-18 1882.2777977 2341.755 1882.2778 2341.7552
X Variable 1 1.918401432713 0.239260971326455 8.018029108874 7.655437922E-09 1.4290578094 2.407745 1.4290578 2.4077451

The
Theprecision
precisionof theof regression
the regression
shown inshown8-58A isingood,
8-58Awithisa good, with a
standard
standard error
error
of theofestimate
the estimate
of 356 relative
of 356 to relative
a dependentto avariable
dependent variable
Dollars

with
withvalues
valuesaveraging
averaging
about 3,000.
aboutAlso,
3,000.
the reliability
Also, the of reliability
the model isof the model is
Units

quite
quitegood,
good,with with
an R-squared
an R-squared
of 68%, ofan F68%,
valueanof 32.6
F value
and aoft- 32.6 and a t-
value
valueon on
sales
sales
units units
of 4.7.ofHowever,
4.7. However,
the t-valuetheon the
t-value
sales on
dollars
the sales dollars Supplies Expens e
variable
variableis poor,
is poor,
as shown
as shown
by the low
by t-value
the low (.99).
t-value (.99). Sales Dollars

The
Theregression
regression usingusing
sales dollars
sales only
dollars
(Exhibit
only8-58B)
(Exhibit
is somewhat
8-58B) is somewhat
worse
worse while
whilethe regression
the regression
on salesonunits
sales
(Exhibit
units 8-58C)
(Exhibit
gives 8-58C) gives
almost
almostequivalent
equivalentR-squared
R-squared
and standard
and standard
errors valueserrors
to thevalues
model to the model
with
with
both
both
unitsunits
and dollars.
and dollars.
BecauseBecause
the regression
the onregression
sales unitson sales units
only
only
is simpler
is simplerand has
anda has
lowerastandard
lower standard
error and higher
errorR-squared,
and higher R-squared,
the
themodel
model
using using
only sales
only units
salesis units
a logical
is choice
a logical for the
choice
cost for the cost
estimation
estimationmodelmodel
in thisin
case.
this case.

For
Forfurther
further
regression
regression
analysis
analysis
on this data,
on this
consider
data, theconsider
graphs the graphs
below
belowwhich
which
shows
shows
evidence
evidence
of seasonality
of seasonality
in the data. in the data.
Dollars

Units

Supplies Expense
Sales Dollars

Since the
the graphs
graphsshow
showclear
clearevidence
evidenceofofseasonality,
seasonality,another
anothertrytry
of the
of the
model
model with
with seasonality
seasonalityincluded
includedwould
wouldbebea auseful
usefulnext step.
next step.TheTheaddition
addition
of a seasonal
seasonal variable
variablefor
forthe
themonth
monthofofDecember
December improved
improvedthethemodel in in
model
Exhibit
Exhibit 8-58C
8-58Csubstantially.
substantially.
The
The seasonal
seasonal model
model for
forsales
salesdollars
dollarsisisshown
shownininExhibit
Exhibit8-58D.
8-58D.Note thethe
Note
substantial
substantial improvement
improvementininR-squared;
R-squared;alsoalso note
note that thethe
that seasonal
seasonal variable
variable
is significant.
is significant. The
The coefficient
coefficienton
onthe theseasonality
seasonalityvariable
variableis is
negative
negative
because supplies
because suppliesexpense
expensedoes
doesnot
notrise
riseasasfast
fastasasunits sold
units in in
sold December.
December.

Exhibit 8-58D
SUMMARY OUTPUT (Seasonal Regression)

Regression Statistics
Multiple R 0.859051741856
R Square 0.737969895185
Adjusted R Squ 0.719253459127
Standard Error 333.2733965545
Observations 31

ANOVA
df SS MS F Significance F
Regression 2 8758843.8017205 4379421.90086 39.428975307562 7.193109E-09
Residual 28 3109992.39182789 111071.156851
Total 30 11868836.1935484

Coefficients Standard Error t Stat P-value Lower 95% Upper 95% Lower 95.0%
Upper 95.0%
Intercept 1815.23365721 167.018364795732 10.8684674253 1.483997825E-11 1473.1116618 2157.356 1473.1117 2157.3557
Units 2.949465937682 0.503693179088494 5.855679727528 2.700318193E-06 1.9176960737 3.981236 1.9176961 3.9812358
Season -1042.0362195 456.167928402683 -2.284325912928 0.0301360468856 -1976.4549112 -107.6175 -1976.455 -107.6175

2. Predicted monthly figures for supplies expense for 2010 using regression results from Exhibit 55D

Units Seasonality Predicted Expense

Jan 180 0 2346
Feb 230 0 2494
Mar 190 0 2376
Apr 450 0 3142
May 350 0 2848
Jun 350 0 2848
Jul 450 0 3142
Aug 550 0 3437
Sep 300 0 2700
Oct 300 0 2700
Nov 450 0 3142
Dec 950 1 3575
Problem 8-59 Cross-Sectional Analysis
Background

WasteTec is a large construction company that specializes in the construction of l

plants and recycling plants. A major cost driver in either type of facility is the
example, the capacity of a recycling plant is measured the number of tons of wa
plant can process. These plants can vary in size from a few hundred TPD to as
TPD. Regression analysis is a useful method to estimate the cost of a new p
equation developed from prior plant construction projects. The dependent varia
actual construction cost of each project, while the independent variable is the TPD
sample of some recent projects and the related construction costs (in thousands).

Problem Information

Location TPD
Commerce, CA 360
Hudson Fall, NY 400
Layton, UT 420
Oxford Township, NJ 450
Savannah, GA 500
Poughkeepsie, NY 506
Panama City, FL 510
Ronkonkoma, NY 518
Okahuma,FL 528
Spokane, WA 800
Arlington, VA 975
Camden, NJ 1,050
York, PA 1,344
Bridgeport, CT 2,250
Chester, PA 2,688

8-59 Requirements

1. Develop a regression model to predict the cost of a propopsed new plant in Babylon, NY
 of 750 TPD. What is the predicted cost for the Babylon plant, using your regression?
2. Evaluate the precision and reliability of the regression you have developed. How might i

Solution
s in the construction of large waste water treatment
er type of facility is the capacity of the plant. For
e number of tons of waster per day (TPD) that the
ew hundred TPD to as many as several thousand
ate the cost of a new plant by using a regression
. The dependent variable of the regression is the
ndent variable is the TPD for the plant. Below is a
on costs (in thousands).

Cost
$59,369
$77,013
$50,405
$75,779
$87,439
$57,463
$60,730
$84,457
$88,119
$152,902
$127,021
$163,395
$139,302
$344,852
$448,073

new plant in Babylon, NY, which will have a required capacity

sing your regression?
e developed. How might it be improved?
Problem 8-60 Regression Analysis: Cross-Sectional Analysis; Calculation of a R
Background
Jim Manzano is the general partner of an investment group that owns a number of comm
including a chain of 15 convenience stores located in the greater metropolitan area of Clevelan
the recent increase in inventory theft and waste (he calls it “spoilage”) in his stores. Spoilag
percent in each of the past two years. In some stores, the main reason is theft; in others, it is d
others, merchandise actually does spoil and must be thrown out. Jim has collected data on sp
recent month and is looking for patterns of spoilage relative to store size (measured by squa
employees, and total sales) and to the location of the store (location 1 is an area where few arr
or vandalism are made, and location 3 is for areas with high arrests). Jim is not sure, but he
managing convenience stores, that a relationship exists among these factors. A colleague to
called “cross-sectional” regression would suit his needs. The cross-sectional regression is to
for the dependent variable to the predicted value as a basis for assessing the reasonableness of
often used in cases similar to Jim’s in whom the accuracy or reasonableness of the reported de
effect, the cross-sectional regression develops a model that represents the overall patterns in a
will be identified by the largest error terms in the regression. The following data are for the mo

Problem Information

Store Number of
Number Inventory Spoilage Square Footage Employees
1 $1,512 2,400 8
2 $3,005 3,900 10
3 $1,686 3,200 12
4 $1,908 3,400 12
5 $2,384 3,750 9
6 $4,806 4,800 10
7 $2,253 3,500 8
8 $1,443 3,000 10
9 $3,755 5,550 15
10 $1,023 2,250 15
11 $1,552 2,500 9
12 $2,119 3,500 16
13 $5,506 7,500 15
14 $3,034 5,700 16
15 $772 2,200 8
 TOTALS $36,758 57,150 173

Requirements

1. Use regression analysis that predicts inventory spoilage at each of the 15 stores. Use a
independent variables (or a combination) you think appropriate and explain your answer. A
reliability of the regression you select.

2. Using the regression equation you developed in requirement 1, determine which of the 1
spoilage that is out of line relative to the entire chain of stores. Explain your choice.

Solution

1. Use regression analysis that predicts inventory spoilage at each of the 15 stores.
independent variables (or a combination) you think appropriate and explain your an
precision nad reliability of the regression you select.

The first regression we try includes all four independent variables: square feet, number of e
dollars. Each of these variables has a plausible relationship to inventory spoilage. We find
below) that the R-squared value is very good (95%), the SE is relatively low at 15% (370.5
for two of the variables (location and square feet), poor for the sales variable, and margina

Variable t-value
Square Feet 2.86
Employees -1.89
Location 3.63
Sales -0.33

Regression Statistics
Multiple R 0.97305677
R Square 0.94683948
Adjusted R 0.92557527
Square
Standard Error 370.518856
Observations 15

ANOVA
df SS MS F
Regression 4 24451628 6112907 44.5274
Residual 10 1372842 137284
Total 14 25824470

Coefficients Std Error t Stat P-value

Intercept -201.784428 393.7116 -0.51252 0.61942

Footage 0.62591525 0.218101 2.86984 0.01667
Employees -73.7720711 38.88984 -1.89695 0.08706
Location 879.37668 242.0028 3.63375 0.00458
Sales -0.00074804 0.002235 -0.33469 0.74477

The negative t-value on “employees” suggests that the spoilage at some stores might
but the t-value is not strong enough for strong conclusions. The t-value on “sales”
should consider deleting the variable from the model; moreover, can we explain w
variable is negative? With this thinking, we decide to re-run the model keeping o
variables: location and square feet. The results are shown below, under “Regressi

The second regression, shown below, has comparable values for R-squared and
improved. Additionally, the F-value almost doubles, meaning a more statistically re
reasons we have chosen to rely on this second regression model to complete the ana
the regression results there is a residual report which shows the predicted and actual v
store, and the error term (“residual”). A large positive residual is unfavorable while a
favorable.

2. Using the regression equation you developed in requirement 1, determine which o

inventory spoilage that is out of line relative to the entire chain of stores. Explain yo

Stores 6 and 7 have relatively high spoilage for their given levels of square feet and locatio
relationships for all 15 stores, as captured in the regression model. Why then are these two
Jim has now a basis for beginning an investigation. Jim might also want to investigate why
so unexpectedly low at stores 12 and 14, to perhaps discover the factors (beyond the 4 va
considered) associated with these stores that have contributed to their success.
Regression Two: Square Footage and Location Only
Regression Statistics
Multiple R 0.96194073
R Square 0.92532997
Adjusted R 0.91288496
Square
Standard Error 400.865101
Observations 15
ANOVA
df SS MS F
Regression 2 23896156 1.20E+07 74.3535
Residual 12 1928314 160693
Total 14 25824470

Coefficients Std Error t Stat P-value

Intercept -777.895194 292.1796 -2.66239 0.0207

Footage 0.45825253 0.126322 3.62765 0.00347
Location 926.553993 256.4631 3.61282 0.00356

RESIDUAL
OUTPUT
Store Number Predicted Residuals Actual
Spoilage Spoilage
1 1,248 264 $1,512
2 2,862 143 3,005
3 1,615 71 1,686
4 1,707 201 1,908
5 2,794 -410 2,384
6 4,201 605 4,806
7 1,753 500 2,253
8 1,523 -80 1,443
9 3,619 136 3,755
10 1,180 -157 1,023
11 1,294 258 1,552
12 2,679 -560 2,119
13 5,439 67 5,506
14 3,687 -653 3,034
15 1,157 -385 772
 sis; Calculation of a Regression Equation

owns a number of commercial and industrial properties,

ropolitan area of Cleveland, Ohio. Jim is concerned about
e”) in his stores. Spoilage has increased by more than 20
n is theft; in others, it is damage and vandalism; and in still
m has collected data on spoilage at each of his stores in the
e size (measured by square feet of floor space, number of
1 is an area where few arrests for theft, disorderly conduct,
. Jim is not sure, but he suspects, based on his experience
se factors. A colleague told him that a type of regression
sectional regression is to compare the actual known value
ng the reasonableness of the actual value. This approach is
bleness of the reported dependent variable is a concern. In
s the overall patterns in all the data and the unusual stores
owing data are for the most recent month’s operations:

Location Sales
1 $312,389
2 $346,235
1 $376,465
1 $345,723
2 $453,983
3 $502,984
1 $325,436
1 $253,647
2 $562,534
1 $287,364
1 $198,374
2 $333,984
3 $673,345
2 $588,947
1 $225,364
 $5,786,774

of the 15 stores. Use any of the four potential

d explain your answer. Also evaluate the precision nad

determine which of the 15 stores might have inventory

plain your choice.

each of the 15 stores. Use any of the four potential

e and explain your answer. Also evaluate the

square feet, number of employees, location type, and sales

ntory spoilage. We find from the regression results (see
ively low at 15% (370.51/(36,758/15)). The t-values are good
s variable, and marginal for the number-of-employees variable.
 Sig F.
2.43E-06

Lower 95% Upper 95%

-1079.03 675.4598
0.139956 1.111874
-160.424 12.8799
340.1608 1418.593
-0.00573 0.004232

ge at some stores might be due to under-staffing,

The t-value on “sales” is weak enough that we
over, can we explain why the coefficient on this
n the model keeping only the two independent
elow, under “Regression Two.”

es for R-squared and SE, but the t-values are

g a more statistically reliable model. For these
del to complete the analysis for Jim. Note below
e predicted and actual values for spoilage at each
is unfavorable while a large negative residual is

nt 1, determine which of the 15 stores might have

n of stores. Explain your choice.

f square feet and location type, based upon the

Why then are these two stores so different?
want to investigate why the level of spoilage is
actors (beyond the 4 variables already
heir success.
 Sig F.
1.73E-07

Lower 95% Upper 95%

-1414.5 -141.291
0.18302 0.733485
367.7689 1485.339
Problem 8-61 Regression Analysis in Tax Court Cases
Since at least the late 1960s, the court systems in the United States and elsewhere have a
regression analysis as evidence in court cases. In many instances, however, because of lim
or errors in developing the regression analysis, tax courts question or deny the regression ev
A study was performed recently to determine the factors in the regression analysis that th
considered in determining whether regression evidence was admissible.

Required What factors regarding the development of a regression analysis do you suspect
courts considered in determining the acceptability of a regression analysis as evidence?

Solution
 tates and elsewhere have accepted
ces, however, because of limitations
on or deny the regression evidence.
e regression analysis that the court
missible.

ion analysis do you suspect the tax

n analysis as evidence?