Beruflich Dokumente
Kultur Dokumente
1. Introduction
In this paper, we provide empirical documentation of the effect of
activity-based costing (ABC) information on product- and customerrelated decisions made by managers in a company.
Cooper and Kaplan (1998) argue that when an entity implements
ABC, it reaps at least two important benets. First, its entire operation
is scrutinized in great detail and its performance analyzed and benchmarked against best practices. Employees are encouraged to be critical
of the status quo and to suggest improvements. This critical analysis,
they argue, can result in process improvements that promote more
efcient use of resources and hence reduce costs. Cooper and Kaplan
A previous version of this paper was titled, ABC at Insteel Industries. We would
like to thank the Insteel Industries management for generously providing us with their
time and complete access, and Dave Conrad for the data. We also thank Stan Abraham,
Sarah Eriksen, and Gregg Friedman of the Harvard Business School for their research
assistance; and Eric Noreen, Naomi Soderstrom, the seminar participants at the Harvard
Business School, Pennsylvania State University, and the April 1999 NBER conference
on Organizational Change and Performance Improvement, Santa Rosa, California, and
Ed Lazear, the discussant at the conference, for their comments.
2002 Massachusetts Institute of Technology.
Journal of Economics & Management Strategy, Volume 11, Number 2, Summer 2002, 257288
258
call this operational activity-based management. Second, ABC information may enable a rm to change the mix of products produced and
customers served, allowing it to focus on making protable products
and serving protable customers. Cooper and Kaplan call this strategic
activity-based management.
There are a number of teaching cases and other such anecdotal evidence of ABC improving managerial decisions. Likewise, there
are many other cases where ABC has no inuence on managerial
decisions, presumably because of organizational resistance to change.
To our knowledge, there has been no systematic, statistical investigation of whether ABC really inuences managerial decisions. An ABC
analysis may fail to have any impact on a rm for two reasons:
1. It may not reveal any new information to the managers, who intuitively know already what an ABC system formally captures.
2. It is conceivable that outside consultants are hired to do an ABC
analysis, but decisionmakers in the rm do not accept the ABC
numbers, which often differ signicantly from traditional cost numbers. Effective follow-up to the ABC analysis may require managerial decisions and actions signicantly different from the status
quo, resulting in organizational change and upheaval, to which
managers may be resistant.
In this study, we interview top and middle managers at Insteel
and conduct statistical analysis of rm-level data in order to shed light
on whether ABC provides new information to managers and whether
activity-based management (ABM) signicantly inuences productand customer-related decisions.
We nd that after the ABC analysis, Insteel undertook a number of process improvement actions that resulted in signicant cost
savings and displayed a higher propensity to:
1. Discontinue products that were found comparatively unprotable
in the ABC study products.
2. Increase price of products that were found comparatively unprofitable in the ABC study.
3. Discontinue customers that were found comparatively unprotable
in the ABC study.
The changes to the portfolio of customers served were similar
to but not as striking as the product-mix and pricing decisions. This
nding is consistent with senior managers intuition that product-level
decisions can be made faster than customer-level decisions.
The next section reviews the literature. Section 3 provides the
companys background and cost structure and discusses its decision
259
2. Literature Review
Surveys of company practices in the US and abroad show that for at
least several decades overhead costs have been allocated to products.
Likewise, surveys of company pricing practices document that companies typically use fully allocated product costs, rather than short-term
variable costs, to set product prices. For example, Govindarajan and
Anthony (1983) surveyed pricing practices of Fortune 1000 industrial
companies and found that 83% of them use full cost, rather than variable costs, for their pricing decisions. See Horngren et al. (2000) for a
summary of a number of surveys on cost allocation practices in US,
Australia, Ireland, Japan, and the UK as well summaries of surveys
that ask managers to rank the importance of full cost relative to variable cost in their pricing decisions. The pervasive use of full costs,
as opposed to short-term variable costs or marginal costs, in pricing
decisions runs counter to the prescriptions of microeconomics that
rms should set prices in a way that equates marginal revenues with
marginal costs plus opportunity costs (shadow prices) of constrained
resources.
Many theoretical papers try to explain this paradox and identify
conditions under which using full costs for product pricing will
maximize rm value or at least serve as a good heuristic. Hansen and
Magee (1993) show that under specied circumstances, the sunk cost
of capacity approximates the optimal opportunity cost of capacity.
Johnson and Kaplan (1987), Cooper and Kaplan (1987), Zimmerman
(1979), and Demski and Feltham (1976) suggest that sunk costs could
be a crude proxy or surrogate for the externality costs and opportunity cost of resources. Balachandran and Srinidhi (1987) use queuing
theory to model uncertainty and show that a xed charge (over and
above marginal costs) needs to be applied to achieve an optimum use
of service centers in the presence of uncertainty in demand. Fixed
charges help relieve congestion of service centers and reduce delay
costs to the rm. Whang (1989) considers information, incentive, and
congestion effects of cost allocations and shows that cost allocation,
in contrast to marginal-cost pricing, induces users to reveal their
true values and achieves ex post efciency in allocating purchased
capacity. Cooper and Kaplan (1987) argue that managers rely on full
costs, rather than marginal costs, because they view product portfolio
and pricing decisions as long-term decisions. Under that viewpoint
260
all long run costs are variable and constraints are less likely to be
binding, and full costs will serve as a proxy for long-run marginal
costs. Govindarajan and Anthony (1983) reject the notion that managers are trying to maximize prots and appeal to Herbert A. Simons
satiscing model to explain the use of full costs as opposed to variable
costs in pricing decisions.
Traditionally, all overhead was allocated to products based on
volume drivers such as labor hours, machine hours, or number of
units produced. Over the last decade, a new way to allocate overhead
costs to products, called activity-based costing, has emerged. Recent surveys indicate that 15% to 20% of companies have adopted ABC and
an equal number are considering adopting it (see Horngren et al.,
2000). The term activity-based costing was rst used in John Deere
Component Works (see March and Kaplan, 1987).
Under both traditional costing and ABC, overhead costs are assigned to cost objects such as products and customers in a two-step
process. Cost of resources such as rent, utilities, depreciation, and
indirect labor are rst assigned to cost pools using resource drivers.
For example, rent is assigned using oor space, while depreciation
is assigned using book value of assets. Under traditional costing,
cost pools were often departments (e.g., machining, assembly, and
nishing), whereas under ABC, cost pools include not only departments but also activities such as inspecting incoming materials, moving materials, and performing setups. Costs are allocated from cost
pools to cost objects using cost drivers. Under traditional costing,
cost drivers were typically volume drivers such as machine hours
and labor hours. Under ABC they include not only volume drivers,
but also activity drivers such as number of inspections, moves, and
setups. ABC recognizes that it is cheaper to produce products in larger
batches than smaller batches. For example, under ABC setup costs
are assigned to a batch whenever a setup is performed, independent
of the batch size. Likewise, product design and engineering costs are
incurred independent of the number of batches and units of a product
produced. Thus under ABC, managers must track separately expenses
that are required to produce individual units, to process batches, to
design or maintain a product, and to keep a manufacturing facility up
and running. See Figure 1 for a graphical representation of an ABC
system.
Robin Cooper and Robert Kaplan have written several cases and
articles about the design, implementation, and benets of ABC systems (see Cooper and Kaplan, 1988, 1991). The theory of a hierarchy of
costsunit-, batch-, product-, and facility-level costswas proposed
in Cooper (1990).
Product 1
$/Receipt
# Receipts
Inspect
Incoming
Materials
$/Move
# Moves
Move
Materials
Product 2
$/Setup Hour
Setup Hours
Setup
Machines
Indirect Labor
$/Setup
# Setups
Prepare
Tooling
Product 3
$/Direct Labor
Hour
Assembly
Product 4
$/Machine Hour
Machine Hours
Machining
Utilities
FIGURE 1. HOW EXPENSES FLOW FROM RESOURCES TO ACTIVITIES TO PRODUCTS UNDER ABC
Adapted from Cooper and Kaplan (1988)
Product
Cost Driver
Rate
Cost Driver
Activity
Activity
Resource
262
The theory of ABC (that overhead costs were related to nonvolume drivers, such as batch- and product-level drivers) was rst
empirically tested by Foster and Gupta (1990). They analyzed crosssectional data from 37 facilities of an electronics company and did not
nd evidence that overhead costs were related to nonvolume drivers.
Since then, however, a number of papers have found evidence of overhead costs being positively correlated with nonvolume drivers (see
Anderson, 1995; Datar et al., 1993; Banker and Johnston, 1993; and
Banker et al., 1995).
Another stream of literature addresses the question, Do ABC
implementations succeed in providing better information to managers and improving decisions? Shields (1995), Roberts and Silvester
(1996), Anderson et al. (1997), Anderson and Young (1997), and Foster
and Swenson (1997) have used surveys of companies implementing
ABC to understand what factors explain the success or failure of ABC
implementation efforts. These studies have identied the level of
team commitment, management support, and quality of other information systems as some of the factors that affect the success of ABC
efforts. Our interviews and surveys of company management support
the ndings of these studies and suggest that other factors such as
market forces and compensation structures also play a signicant
role in the ability and willingness of an organization to implement
ABM. These issues present themselves as leads for future work in the
empirical investigation of ABC implementations.
In this paper, we complement the earlier surveys on ABC implementations by focusing on decisions made within one company after
an ABC study. Using statistical analysis and interviews, we document
whether the ABC system affects product- and customer-related decisions at that company.
263
Insteel Industries, Inc., manufactures and markets concrete reinforcing products, industrial wire, bulk nails, collated fasteners, PC strand,
and tire bead wire. The companys primary markets are the construction, home furnishings, appliance, and tire manufacturing industries.
Headquartered in Mt. Airy, North Carolina, the company currently
operates eight manufacturing facilities serving markets nationwide.
Annual sales revenues are about $300 million.
This paper studies the managerial and organizational issues
related to ABC at the Andrews, South Carolina, plant of Insteel
Industries. Four product lines are produced at the Andrews plant:
steel wire of different gauges (called bright wire or industrial
wire in the trade), galvanized wire of different gauges, wire mesh,
and nails. In 1996, about 477 individual products were spread across
these four product lines. However, 20% of the products accounted for
85% of Andrewss annual revenues of about $60 million. Likewise,
20% of Andrewss customers accounted for 95% of all sales.
The steel and galvanized-wire product lines are produced to
order; the other two product lines, nails and mesh products, are produced to stock (about 20 days inventory). While other Insteel plants
also produce wire products, all of Insteels nail products are manufactured at the Andrews plant, in Nail Mill A (commodity nails and
sinkers, a large-volume, nondifferentiated product line) and Nail Mill
B (pallet nails, manufactured to tight tolerances on diameter, heading,
and threading specications).
The primary raw material is hot-rolled steel rod, purchased in
large rolls of about 4000 pounds from both domestic and foreign suppliers. Wire drawing, which is the rst production stage, is common
to all of Andrewss products and generally operates at capacity. See
Figure 2 for a graphical representation of Insteels production process.
Prior to 1996, Insteel did not have a sophisticated cost system.
The price per ton of steel rod, the basic raw material, was closely
Galv
Galv
Wire
Wire
Nail Machine # 2
Mesh
Mesh
to
Nail Machine # 55
Wire Galvanizer
Nail Machine # 1
to
Wire Drawer # 16
Wire Drawer # 3
Wire Drawer # 2
Wire Drawer # 1
Cleaning House
Steel Rod
Wire
Wire
Mesh
Mesh
Bright
Bright
Nails
Nails
Nail
Galv
Galv
Galv
Nails
Nails
264
monitored. This price was used to estimate the material costs of all
of Insteels productsby multiplying the weight of the product by
the price of steel rod. Although freight was billed as a separate cost
on each invoice, customers were charged an average freight charge on
each shipment, independent of the actual destination. Price quotations
were based only on the weight of the product and estimated freight
costs.
In 1996, after years of product and customer proliferation, Insteel
decided to implement an ABC system. Insteels management recognized that the consumption of overhead resources had increased
substantially and varied signicantly over its product and customer
portfolio. Wishing to limit the upheaval caused by a change in the
status quo, and also to maximize the possibility of enthusiastic adoption, Insteels management selected the Andrews plant as its rst ABC
site. The plant had had a history of good communications between
plant management and workers and supervisors, it had minimal
operational overlap with other plants, and its manager, Bill Sronce,
was an enthusiastic champion of the process. He communicated
the objective and the process of the ABC study to all employees of
the plant. Dave Conrad, the director of cost management, was also
assigned full-time to the project. He worked in close coordination
with the consultants to implement the technology, analyze the data,
and train Insteels staff in the ABC methodology.
At Insteel, the ABC team took their rst ABC snapshot of operations at the Andrews plant in the summer of 1996 with the assistance of a big-six accounting rm. The ABC team analyzed Andrewss
operations and identied 12 business processes. Within each business
process, a number of activities were identieda total of 146 activities. Next, 426 employees were surveyed to estimate how they allocated their time to different activities. All overhead resources were
then collected in 80 cost pools and assigned to cost objects such as
products and customers (some cost pools included multiple activities).
This assignment was done by selecting cost drivers that link the performance of activities to demands made by individual products. For
example, the cost of the material-handling activity was assigned to
products based on the number of moves for that product, on the
ground that it is the number of times raw materials and work-inprocess is moved that causes material handling resources to be consumed, rather than the weight of materials moved. See Table I for
examples of unit-, batch-, product-, and customer-level activities and
the cost driver associated with each activity.
This cost allocation exercise resulted in Insteels rst ABC analysis and its rst ABC database (essentially a set of linked Excel spreadsheets and Access databases), maintained by the ABC team. Cost of
265
TABLE I.
Cost (MM)
Cost Driver
Cleaning house
Production
Tons cleaned
Pounds produced
14
Material handling
Changeovers
Number of moves
Number of changeovers
Product
02
Number of products
Customer
89
Invoicing
Selling products
Number of invoices
Sales calls
Batch
155
Activities/Cost Pools
Material costs: Cost of steel rods and other materials used in production. These costs were assigned to products and customers based on
pounds produced and sold, respectively.
Conversion costs: Overhead resources consumed in converting steel
rod to nished wire or nail products. This category includes
unit-level costs (such as plant labor), batch-level costs (such as
wire-drawing setup resources), plant-level costs (such as plant
managers salary), and product-level costs (such as the costs of
special wire-drawing dies for particular products). Each component
of conversion costs was allocated to products using an appropriate
cost driver.
Customer costs: Overhead resources consumed in serving special
customer needs such as the packing and loading costs, orderprocessing and invoicing costs, cost of postsales services, and
carrying cost of receivables. Customer costs were traced to each
customer and allocated to products based on quantity of each
product purchased by that customer.
Capital costs: Cost of capital related to product- and customerspecic assets. Assets were rst allocated to products and customers. A weighted average cost-of-capital charge was then applied
to these assets and then charged to the corresponding product or
customer. For example, accounts receivable were identied with
266
4.1
267
268
matched the billed freight charge with the actual freight cost, revealed
this pattern of discrepancy.
When Insteel realized after the 1996 ABC study how much they
were actually incurring in quality costs, they put in place a team
responsible for probing deeper and understanding better what was
causing the quality costs to be incurred in the rst place and for suggesting steps to reduce them. Insteel realized that certain foreign suppliers of rods were lower in price but supplied poorer-quality rods
that caused breakdowns in Insteels manufacturing process. The lower
price of those suppliers did not compensate for the quality costs.
Insteel switched to higher-quality rod suppliers. Insteel also realized
that smaller-diameter wire products are more likely to break and disrupt the manufacturing process. Insteel migrated its product mix to
more large-diameter wire products. Such initiatives led to reduction
in quality costs from $6.7 million in 1996 to $4.9 million in 1997.
It is hard to estimate how much of these savings would have
been realized had Insteel not conducted an ABC analysis. From interviews with Insteel managers and sitting in on senior management
meetings, it appears that the activity analysis gave them an appreciation of the scope and quantied the magnitude of the improvement
potential, thereby allowing them to prioritize among various process
improvement possibilities. In the opinion of the senior management,
the ABC system also helped them keep track of the savings and ensure
that the promised payback was actually realized.
The plant manager reported that no other initiatives, such as
total quality management (TQM) and just-in-time (JIT), were being
introduced or practiced in the plant simultaneously with the adoption of ABC. Applying our past experience with at least ten ABC
implementations and our visits to more than a dozen JIT/TQM plants,
we observed certain factors at Andrews that lead us to believe that
the process improvements were indeed the result of the ABC study.
Under JIT or TQM, process improvements are undertaken by trained,
motivated, and empowered employees. Data about the manufacturing
process is collected, displayed, and shared with and by the workers,
either through charts and diagrams or via computers on the plant
oor. Workers are cross-trained and wage contracts are redesigned
to encourage teamwork and creative problem solving while working
with reduced buffer inventories, smaller lot sizes, and tightly coupled
workstations. Continuous process improvement is then a byproduct of
worker-level initiatives.
In contrast, in an ABC study, potential process improvements
emerge from the cost data analysis when certain activities show up
as being disproportionately expensive. Teams of managers usually
269
270
271
TABLE II.
76 (38%)
59 (27%)
65 (30%)
112 (58%)
52 (42%)
81 (51%)
83 (53%)
81 (42%)
71 (58%)
77 (49%)
75 (47%)
(C) By 1996 Protability and 1997 Price Changes or Availability Relative to 1996
No. of products sold in 1996 and also
112 (46%)
in 1997 for which prices were decreased
in 1997 relative to 1996
52 (26%)
81 (37%)
83 (38%)
Total
A: number of products found protable in the 1996 ABC study; B: number of products found unprotable in the
1996 ABC study; C: number of products above or equal to the median 1996 gross margin per pound; D: number of
products below the median 1996 gross margin per pound.
272
273
TABLE III.
Correlation with
Change in Volume
from 1996 to 1997
p-value
Sample
Size
067
028
026
011
015
015
237
622
641
847
786
795
316
316
316
316
316
316
274
275
(p = 0822), indicating that the likelihood of price increase for a profitable product is not signicantly different from that of an unprotable
product, when protability is measured by gross margin per pound.4
From Table II(A) we know that products found unprotable in
the 1996 ABC study are more likely to be discontinued than protable
products. It is conceivable that Insteel discontinued products by raising their prices to a level that would take them off the market. For
completeness, we include in discontinued products those for which
prices were increased in 1997 and run the analysis once again. The
results are shown in Table II(C). Naturally, relative to Table II(B) only
the second row of Table II(C) has increased frequencies, and the ndings are considerably stronger: From columns A and B of Table II(C),
74% of ABC-based loss-making products faced price increases or discontinuance, vs. 54% of prot-making products ( 2 = 193, signicant
at less than 0.01%). In contrast, the gross-margin-based frequency distributions in columns C and D show almost no difference between
high- and low-gross-margin products ( 2 = 004 with p-value 0.84),
further strengthening our ndings in support of Hypothesis 2. Thus,
it appears that 1997 product pricing decisions are more closely associated with ABC protability numbers than the old cost-accounting
systems gross-margin-based numbers.
4.2.1.3 Impact of ABC on Customer Portfolio. Next, we analyze
customer-level data to test whether the evidence broadly shows that
the sales force acted on these customer-level ndings of the ABC
analysis. Dening customers who appear in the 1996 and the 1997
data as continuing customers, and 1996 customers that dropped out
of Insteels 1997 customer list as discontinued customers, we formulate
the following hypothesis on customers:
Hypothesis 3: Customers found to be unprotable in the ABC study
are more likely to be discontinued in 1997 than customers found to
be protable.
Within the nail product group, we have 1996 and 1997 information about the total revenue from a customer, the total product
cost, customer cost, and freight cost, and the net protability of the
customer. The data we have on wire customers needs additional
decomposition before it can be used for a test of this hypothesis;
hence we restrict attention to customers for nail products. The results
are summarized in Table IV. Unlike the product-level data used in
4 The results are virtually identical when protability is measured as gross margin
per dollar of sales.
276
Protable
Unprotable
66 (66%)
20 (53%)
34 (34%)
18 (47%)
100 (100%)
38 (100%)
Total
Table II, we do not have the materials cost by customer and are
unable to estimate gross margin by customer. Hence we are unable
to perform gross-margin-based tests for customer continuationdiscontinuation decisions. We see in Table IV that 66% of customers
who were protable in 1996 were continuing customers in 1997, versus 53% for those that were unprotable in 1996. The 2 test statistic
of 2.1 is signicant only at the 15% level.
Compared to multivariate tests, univariate tests are simpler and
provide more economic intuition. However they are not informative of
marginal effects. To understand the marginal effect of ABC data after
controlling for volume and gross margins, we turn to multivariate
probit analysis.
4.2.2 Multivariate Analyses of Product and CustomerRelated Decisions. The results of the univariate analyses above
277
xed costs, and this would also cause Insteel to be reluctant to discontinue the product or re the customer. Since volume is also related
to protability, the hypothesis we wish to test is then
Hypothesis 4: The higher the 1996 ABC prots, the higher the 1996
gross margins, and the greater the 1996 share of volume of a product
(or customer), the higher will be the 1997 probabilities of continuing
to sell a product, of lowering product price, and of continuing to sell
to a customer.
To test this hypothesis, we conduct probit regressions of the decision to continue a product or raise prices against 1996 ABC prot per
pound (ABC PrPP96), 1996 gross margin per pound (GMPP96), and
share of volume of product sold in 1996 (LBS PCT96). Table V gives
univariate statistics and correlations for these variables: panel (A) is
based on all 440 products that were sold in 1996, panel (B) is based
on 316 continued products (those that were sold in 1996 and 1997),
and panel (C) is based on the 138 continued nail customers. As mentioned in the section on univariate tests, we do not have gross-margin
information for customers. We note that for products GMPP96 is negatively correlated with LBS PCT96, indicating that Insteel was selling a larger relative volume of lower-gross-margin products. Not surprisingly, GMPP96 is positively correlated with ABC PrPP96. At the
product and at the customer level, sales volume is not signicantly
correlated with ABC protability.
Relying on our earlier denition of continued products and customers, but recognizing that discontinued products may only have
been rendered dormant by high prices, in one of our tests we combine
discontinued products and those that experienced a price increase.
Specically, we conduct probit analyses of four decisions and report
the results in Table VI:
0001
0006
ABC PrPP96
Mean
Variable
LBS PCT96
0010
ABC PrPP96
0161
0002
LBS PCT96
GMPP96
0167
Mean
GMPP96
Variable
TABLE V.
0074
0003
0112
0024
0086
Q1
0006
00003
0119
Med
0034
0002
0214
Q3
1192
0149
1683
Max
0090
(0060)
0153
(0001)
1000
GMPP96
0542
0045
Min
0016
0086
Q1
0010
00001
0118
Med
0036
00007
0215
Q3
0554
0046
0819
Max
0106
(0059)
0120
(0033)
1000
GMPP96
2333
0077
Min
Std Dev
0160
0009
0138
Std Dev
(A) Independent Variables for Probit Analyses 1 and 3 (see Table VI)
0153
(0001)
0033
(0489)
1000
ABC PrPP96
0018
(0753)
0106
(0059)
1000
LBS PCT96
0120
(0033)
0018
(0753)
1000
ABC PrPP96
Correlation Matrix
(p-values in parentheses)
0033
(0489)
0090
(0060)
1000
LBS PCT96
Correlation Matrix
(p-values in parentheses)
278
Journal of Economics & Management Strategy
Mean
0007
0013
Variable
LBS PCT96
ABC PrPP96
0199
0018
Std Dev
169
Min
0005
0001
Q1
0015
0002
Med
0047
0005
Q3
0131
0162
Max
0056
(0512)
1000
0056
(0512)
1000
ABC PrPP96
Correlation Matrix
(p-values in parentheses)
LBS PCT96
Likelihood ratio
(00001)
(3) = 404
0031
0068
0032
Marginal
Effecte
316
(04065)
(3) = 291
1416
(0161)
9529
(0678)
0307
(0639)
0002
(0989)
Coefcient
(p-value)
0029
0002
0016
Marginal
Effecte
440
(00382)
(3) = 842
1563
(0017)
4313
(0533)
0465
(0367)
0254
(0017)
Coefcient
(p-value)
0034
0002
0023
Marginal
Effecte
138
(00001)
(3) = 452
2593
(0154)
2729
(0001)
N/A
0363
(0045)
Coefcient
(p-value)
0053
0416
Marginal
Effecte
(p-value)
440
1551
(0011)
ABC PrPP96
1372
(00011)
LBS PCT96
0705
(0167)
GMPP96
0551
(0001)
Coefcient
(p-value)
Probit 1: Product
Continuation Decisiona
Intercept
Variable
TABLE VI.
280
Journal of Economics & Management Strategy
281
insignicant, negatively inuences the decision to continue selling a product. This nding lends further support to our test of
Hypothesis 1 [Table II(A)]. Probit 2 is overall insignicant, and the
individual coefcients are insignicant. Treating price increases and
product discontinuance as inseparable in the data, probit 3 (which
is overall signicant) indicates that the higher the ABC prots, the
lower is the propensity to raise price or discontinue a product. Finally,
probit 4, which is signicant, indicates that share of volume is the
only determinant of the decision to continue to serve a customer.5
The coefcient on ABC protability, although positive as predicted,
is not highly signicant (p = 154), and this is completely congruent
with the weak results we obtained for the inuence of ABC numbers
in the test of Hypothesis 3 (Table IV). The lower sample size for the
customer-related tests, compared to product related tests, could also
be responsible for the lower signicance levels.
Table VI also provides the marginal effect of each variable on the
probability of affecting activity-based management decisions. Overall,
a change within the interquartile range of ABC prots per pound
(from $0.26 per pound to $0.39 per pound in the entire sample of 440
products) reduces the probability of discontinuance or price increase
by about 3%. The most dramatic marginal effect is that of a customers
share of volume: a change within the interquartile range (from 0.06%
of total pounds to 0.50% of total pounds) raises the probability of
retaining the customer by almost 42%. This is strong evidence of the
importance of large customers to Insteels business.
Overall, our univariate tests provide some evidence of activitybased management, particularly the management of product mix and
prices. The multivariate tests provide evidence of ABM and show that
even after controlling for gross margins and volume-related signicance, ABC protability had an inuence on product mix and pricing
decisions, but not (as much) on customer portfolio decisions.
We discussed our ndings with Insteels plant mangers, sales
managers, and CFO. There was a general consensus that armed with
the ABC information, discontinuing an unprotable product or discouraging its purchase was one of the early initiatives undertaken
with respect to Andrewss products. Next, production and sales of
hidden-prot products were ramped up. For example, in Narayanan
and Sarkar (1998) we document their aggressive move into pallet
nails. The ABC analysis had also highlighted the high cost of the dies
5 The results of probit 4 are the same whether volume is dened as total pounds
of products sold to a customer in 1996, or total dollar sales to that customer. For consistency across the four probit analyses, we use product or customer pounds (as a fraction
of total pounds sold for all of Andrews).
282
that wear out in the wire drawing process and, hence, the greater relative protability of large-diameter wire. The sales force tried to push
these large-diameter wire products more aggressively after the ABC
product cost information was made available to them.
When we asked them why they had been unable to act more
decisively with respect to customers, they pointed out that consolidations and vertical integration within the industry during the period
of interest limited Insteels ability to jettison customers shown to
be unprotable in the ABC analysis. Specically, they pointed out
that some of Insteels most protable customers had been bought by
Insteels competitors, which were integrating forward into the value
chain. The loss of that protable business in 1997 meant that Andrews
would have excess capacity if it red more customers. The plant was
better off retaining some large unprotable customers who covered
their variable cost and made a positive contribution to xed cost.
283
(sales, less all costs, including materials, conversion, and customerand facility-level costs, but excluding allocated corporate overheads)
of Insteels plants improved after the introduction of ABC. For example, the gross margins at the Andrews plant improved from 8.28% in
19951996 to 11.26% in 19981999.
To control for industry-wide factors we compared the nancial
performance of Insteel with that of other rms with four-digit
SIC code 3310. Since Insteel introduced ABC into its non-Andrews
plants only in 1998, we can think of the nancial performance
in 1999 as being post-ABC. We examined return on equity (net
income/owners equity), return on assets (operating income before
interest expense/total assets), and stock returns. Insteel outperformed
industry median numbers under all three metrics in 1999, whereas
it had lagged behind the industry median under all three metrics in
1996. For instance, the Insteels returns on assets in 1996 and 1999
were 2.91% and 5.95%, whereas the industry medians were 5.38% and
5.09%, respectively. While by no means conclusive evidence, these
upward trends in gross margins and returns are compatible with the
managements perception of ABM having a positive nancial impact
at Insteel.
5. Organizational Issues7
The adoption of ABC at Insteel developed from a project in one location in 1996 to a company-wide initiative in 1999. Interviews with
the management indicate that they consider this effort to have been
successful in improving the bottom line despite the cost of analysis,
implementation, and managerial energy invested in the effort. Judging from our own observations at Insteel and ve other companies,
readings of case histories of similar projects, and input from Insteel
managers, we can say that worker involvement, intensive communication, and commitment from the sales force and upper management
were the key determinants of the perceived success of this project.
5.1
284
5.2
Reaction of Managers
285
plant. But the senior management felt that mesh production was in
the long-term interests of Insteel. The company had its beginnings in
the agricultural and construction mesh business, and upper management was reluctant to drop mesh. The mesh-products sales manager,
who had been a long-time employee, made signicant efforts and successfully improved the protability of mesh products within one year.
She organized more efcient loading of trucks with rolls of mesh and
renegotiated prices with key customers.8
We observed the greatest enthusiasm for ABM among Insteel
middle managers, who acted quickly in gathering up the low-hanging
fruit of the rst ABC analysis, i.e., identifying and implementing process improvements with a quantiable payback. Quick wins were
shared and celebrated with all associates (a term that includes plantoor workers). This helped generate and maintain momentum for
the project when changes related to product mix inevitably caused
upheaval in the plant. John Calcagni, the sales manager for industrial
wire products pointed out that the ABC approach was the rst tool
that allowed them to model the effects of labor, product mix, and customer demandsi.e., factors other than volumeon protability and
nancial performance, and it could also be used to guide and justify
product and customer-mix changes made by the sales force.
5.3
Sales-Force Reaction
286
For many of these products, prices were raised drastically, and the
customers were thus encouraged to leave. Some products were not
offered at all in 1997, and customers were encouraged to substitute
other Insteel products.9 One of Insteels largest customers, walked
in the face of raised prices and discontinued purchases after the 1996
study, but recently, having tried other sources, has become an Insteel
customer again. This, the sales force feels, is vindication for their new
ABC-based strategy.
6. Conclusions
In this study we sought to understand the impact of ABC on managers product pricing, product portfolio, and customer portfolio
decisions.
9 Note that in our empirical analyses, both of these types of products would be
counted as discontinued products.
10 This system change has also led to a complete overhaul of their product and
customer codes and a consolidation of their activity dictionary. This has made new,
post-1997 data hard to compare with the data used in this study.
287
Interviews with managers and our statistical analysis lend support to the hypothesis of changes in managerial decisions at Insteel
following the ABC study. The rationalizing of the product portfolio
and pricing decisions appears to have been inuenced by the ABC
study. Customers are now billed for freight in a more direct way. The
company has made several process changes attributable to the ABC
study. It has introduced new product lines. However, the rationalizing
of the customer portfolio, although initiated by the company, cannot
be statistically detected. These strategy changes will, we suspect, be
reected in the data in the future.
References
Anderson, S., 1995, Measuring the Impact of Product Mix Heterogeneity on Manufacturing Overhead Cost, Accounting Review, 70 (3), 363387.
and S. Young, 1997, Evaluation of Activity Based Costing Systems: The Impact of
Contextual and Procedural Factors, Working Paper, University of Michigan, Ann
Arbor.
, J. Hesford, and S. Young, 1997, Implementing Activity Based Costing: A Field
Study of ABC Development Teams in Two U.S. Automobile Firms, Working Paper,
University of Michigan, Ann Arbor.
Balachandran, K.R. and B.N. Srinidhi, 1987, A Rationale for Fixed Cost Allocation,
Journal of Accounting Auditing and Finance, Spring, 151169.
Banker, R. and H. Johnston, 1993, An Empirical Study of Cost Drivers in the US Airline
Industry, Accounting Review, 68 (3), 576601.
, G. Potter, and R. Schroder, 1995, An Empirical Analysis of Manufacturing Overhead Cost Drivers, Journal of Accounting and Economics, 19, 115137.
Cooper, R., 1990, Cost Classications in Unit-Based and Activity-Based Manufacturing
Cost Systems, Journal of Cost Management, Fall, 414.
and R. Kaplan, 1987, How Cost Accounting Systematically Distorts Product
Costs, in W. Bruns, Jr., and R.S. Kaplan, eds., Accounting and Management: Field Study
Perspectives, Harvard Business School Press.
and
, 1988, Measure Costs Right: Make the Right Decisions, Harvard Business Review, SeptemberOctober, 9798.
and
, 1991, Prot Priorities from Activity-Based Costing, Harvard Business
Review, MayJune.
and
, 1998, Cost and EffectUsing Integrated Cost Systems to Drive Protability
and Performance, Boston: Harvard Business School Press.
Datar, S., S. Kekre, T. Mukhopadhyay, and K. Srinivasan, 1993, Simultaneous Estimation of Cost Drivers, Accounting Review, 68 (3), 602614.
Demski, J.S. and G.A. Feltham, 1976, Cost Determination: A Conceptual Approach, Iowa
State University Press.
Foster, G. and M. Gupta, 1990. Manufacturing Overhead Cost Driver Analysis, Journal
of Accounting & Economics, 12 (13), 309337.
and D. Swenson, 1997, Measuring the Success of Activity Based Cost Management and Its Determinants, Journal of Management Accounting Research, 9, 109141.
Govindarajan V. and R.N. Anthony, 1983, How Firms Use Cost Data in Price Decisions, Management Accounting, July, 3036.
288
Hansen S. and R. Magee, 1993, Capacity Cost and Capacity Allocation, Contemporary
Accounting Research, 9 (2), 635660.
Horngren C., G. Foster, and S. Datar, 2000, Cost Accounting: A Managerial Emphasis,
Upper Saddle River, NJ: Prentice Hall.
Johnson H. and R. Kaplan, 1987, Relevance Lost, Harvard Business School Press.
March, A. and R. Kaplan, 1987, John Deere Component Works, Harvard Business
School Case 187-107.
Narayanan, V. and R. Sarkar, 1998, Insteel Wire Products: ABM at Andrews, Harvard
Business School Case 198-087.
Roberts, M.W. and K. Silvester, 1998, Why ABC Failed and How It May Yet Succeed,
Journal of Cost Management Research, Winter, 2335.
Rosenberg, N., 1976, Perspectives on Technology, Cambridge University Press.
Shields, M., 1995, An Empirical Analysis of Firms Implementation Experiences with
Activity Based Costing, Journal of Management Accounting Research, 7 (Fall), 148166.
Whang, S., 1989, Cost Allocation Revisited: An Optimality Result, Management Science,
October, 12641273.
Zimmerman, J., 1979, The Costs and Benets of Cost Allocation, The Accounting
Review, July, 504521.