Research paper

Viraj_storm

Hendricksetal2007TheimpactofenterprisesystemsoncorporateperformanceAstudyofERPSCMandCRMsystemimplementations.pdf

Home >Business & Finance homework help >Research paper

www.elsevier.com/locate/jom

Journal of Operations Management 25 (2007) 65–82

The impact of enterprise systems on corporate performance:

A study of ERP, SCM, and CRM system implementations

Kevin B. Hendricks a,1

, Vinod R. Singhal b,*, Jeff K. Stratman

b,2

a Richard Ivey School of Business, The University of Western Ontario, London, Ont., Canada N6A-3K7

b College of Management, Georgia Institute of Technology, 800 West Peachtree St., NW, Atlanta, GA 30332-0520, United States

Available online 23 March 2006

Abstract

This paper documents the effect of investments in Enterprise Resource Planning (ERP), Supply Chain Management (SCM), and

Customer Relationship Management (CRM) systems on a firm’s long-term stock price performance and profitability measures such

as return on assets and return on sales. The results are based on a sample of 186 announcements of ERP implementations, 140 SCM

implementations, and 80 CRM implementations. Our analysis of the financial benefits of these implementations yields mixed

results. In the case of ERP systems, we observe some evidence of improvements in profitability but not in stock returns. The results

for improvements in profitability are stronger in the case of early adopters of ERP systems. On average, adopters of SCM system

experience positive stock returns as well as improvements in profitability. There is no evidence of improvements in stock returns or

profitability for firms that have invested in CRM. Although our results are not uniformly positive across the different enterprise

systems (ES), they are encouraging in the sense that despite the high implementation costs, we do not find persistent evidence of

negative performance associated with ES investments. This should help alleviate the concerns that some have expressed about the

viability of ES given the highly publicized implementation problems at some firms.

Keywords: Enterprise Resource Planning (ERP); Supply Chain Management (SCM); Customer Relationship Management (CRM)

1. Introduction

Enterprise systems (ES) represent an important

technology investment option for operations managers,

and have been acclaimed in the practitioner and academic

literature for their potential to improve business

performance (Akkermans et al., 1999; Davenport,

1998). For the purposes of this research, ES include

* Corresponding author. Tel.: +1 404 894 4908;

fax: +1 404 894 6030.

E-mail addresses: [email protected] (K.B. Hendricks),

[email protected] (V.R. Singhal),

[email protected] (J.K. Stratman). 1

Tel.: +1 519 661 3874; fax: +1 519 661 3959. 2

Tel.: +1 404 894 4928; fax: +1 404 894 6030.

doi:10.1016/j.jom.2006.02.002

one or more of the following applications: Enterprise

Resource Planning (ERP), Supply Chain Management

(SCM), and/or Customer Relationship Management

(CRM) systems. This paper documents the effect of

investments in ERP, SCM, and CRM systems on long-run

stock price and profitability performance. The results are

based on an analysis of a sample of 186 announcements

of ERP implementations, 140 SCM implementations,

and 80 CRM implementations at publicly traded firms.

Performance effects are examined over a five-year time

period for ERP implementations and a four-year time

period for SCM and CRM implementations. Perfor-

mance effects are also examined for both the imple-

mentation and post-implementation periods.

Firms have invested heavily in ES. AMR Research

estimates that investment in ES amounted to more than

mailto:[email protected]

http://dx.doi.org/10.1016/j.jom.2006.02.002

K.B. Hendricks et al. / Journal of Operations Management 25 (2007) 65–8266

US$ 38 billion in 2001 (Kraus and O’Brian, 2002).

Forecasters predict continued high growth in the level of

investments in ES (AMR, 2004), which makes

quantifying the financial returns of these investments

an important research issue. By estimating the long-run

financial effects of investments in ES, we shed light on

the value of these systems.

Given the level of ES investment, there is relatively

little empirical research that links investments in ES to

financial performance using objective financial perfor-

mance data. While some researchers have examined the

effect of investments in ERP systems on financial

performance, research on the effect of SCM and CRM

systems on financial performance is very limited or non-

existent. Furthermore, as we discuss in our literature

review (see Section 2), existing research on the effect of

ES systems on financial performance is not as

comprehensive and thorough as it could be in terms

of the metrics used, the methodology used to estimate

the performance effects, and the time periods covered.

Our analysis provides more rigorous and complete

evidence on the effect of ES system on performance.

The evidence in this paper also contributes to the

literature on the effect of information technology (IT)

investments on financial performance (see Dehning and

Richardson, 2002 for a recent review of this literature).

Most of these studies have primarily focused on the total

level of IT spending by firms over several years and the

impact of this spending on different financial perfor-

mance metrics. Very few studies have attempted to

examine the effects of specific type of IT investments on

performance (Chatterjee et al., 2002). Therefore, even

though we may know that at an aggregate level IT

spending positively affects performance, our knowledge

about how specific IT investments affect performance is

limited. Such knowledge can be useful in capital

budgeting and allocating decisions, and targeting

investments to those applications that give the highest

returns. Investments in ES systems require major

commitments of capital and managerial resources,

and it makes sense to carefully estimate the returns from

these investments.

Section 2 critically reviews the previous research on

ES systems and motivates the need for this study.

Section 3 briefly reviews the rationale behind the belief

that investments in ES will improve financial perfor-

mance. The contribution of this paper is a rigorous

validation of this premise. Section 4 describes the

sample collection. Section 5 describes the methods used

to estimate the long-run stock price and profitability

effects of the sample firms. The empirical results are

presented in Section 6. Section 7 discusses these results

in the context of theories on how firms develop

competitive capabilities and develops suggestions for

future research on the operational mechanisms by

which ES systems can improve performance.

2. Review of existing literature on the

relationship between ES systems and financial

performance

In documenting the effect of ES, researchers have

used objective performance data on stock returns and

accounting metrics as well as performance data

collected through surveys and experiments. With

respect to stock returns, researchers have used event

study methods to analyze the short-term stock market

reaction to announcements of ERP implementation.

Hayes et al. (2001) and Ranganathan and Samarah

(2003) estimate the stock market reaction to ERP

implementation announcements based on samples of 91

and 136, respectively. Although not directly related to

ERP systems, Chatterjee et al. (2002) examine the stock

market reaction to 112 infrastructural IT investment

announcements about technologies that provide a

platform for future business applications. These studies

find statistically significant abnormal stock market

returns ranging anywhere from 0.5% to 0.84%,

indicating that the market reacts positively to IT

investment announcements.

Using the efficient market theory one could argue

that the stock market reaction documented by these

studies is an unbiased estimate of the value of such

investments. However, abnormal returns over short

windows may not provide a complete assessment of the

value of investment. Recent research has shown that the

stock market partially anticipates many corporate

announcements, and in other cases abnormal stock

price performance is also observed subsequent to the

announcement (see Fama, 1998 for a review of some of

these studies). This suggests that to get a better idea of

the value of ERP investments, one should estimate

abnormal performance over a longer time period. This is

particularly important for ES announcements given the

complex nature of these investments, their relative

uniqueness and newness, as well as the uncertainty

associated with how the adoption and benefits of these

systems will evolve over time.

There are a couple of academic papers that use

publicly available data to examine the effect of ERP

systems on accounting metrics. Based on a sample of 50

publicly traded firms that announced ERP adoption

during 1993 and 1997 Poston and Grabski (2001)

investigate the effect of ERP adoption on profitability.

K.B. Hendricks et al. / Journal of Operations Management 25 (2007) 65–82 67

They use paired t-tests to compare the profitability in

the year before the implementation with the profitability

one, two, and three years after ERP implementation.

Although they do not find evidence of improvement in

profitability in the three years after implementation,

their results are questionable because of their metho-

dology. In estimating changes in year-to-year perfor-

mance they do not use any benchmarks to control for

changes in performance that may be related to the

sample firm’s prior performance, industry, or economy.

Therefore, it is not correct to equate changes in

performance as estimated by Poston and Grabski (2001)

to the effect of ERP investments on profitability, and

conclude that ERP systems have no effect on profit-

ability. Using benchmarks to control for normal changes

in performance is a basic and minimum requirement in

estimating performance effects of any corporate

investment or decision.

Hitt et al. (2002) analyze a sample of SAP’s ERP

implementations using accounting and stock market-

based performance measures. They find evidence of

improved financial performance during implementation,

but are unable to estimate the long-run impact of ERP

systems due to a lack of post-implementation data at the

time they conducted their study. Furthermore, they

estimate the effects using pooled regressions of data from

SAP’s ERP implementers and the remaining publicly

traded firms (control firms) without matching their

sample and control firms on prior performance. Barber

and Lyon (1996) show that in studies that use accounting

metrics matching on prior performance is critical to get

well-specified and powerful test statistics. Matching on

prior performance adjusts for mean reversion in

accounting metrics that reflects a transitory component

of performance that may have nothing to do with the

event under consideration. Without matching on prior

performance, results can be confounding as it is unclear

whether the observed abnormal performance is due to

mean reversion or due to the event under consideration.

Other academic studies have examined the effect of

ERP investments on performance using in-depth case

studies, data collected from surveys, or experiments.

McAfee’s (2002) is an example of an in-depth case

study of an ERP implementation and its effect on

performance at a single firm. A survey by Mabert et al.

(2003) found some improvements in managers’

perceptions of performance (mainly in financial close

cycles and order management) but found that few firms

had reduced direct operational costs. Stratman’s (2001)

survey found that manufacturing firms saw little change

in operational metrics. Hunton et al. (2002) experi-

mentally test the relation between ERP and perfor-

mance by presenting 63 certified financial analysts at a

financial services firm with a hypothetical case of a

company and comparing these analysts’ initial earning

forecasts with the forecasts after they are told that the

hypothetical firm has committed to investing in an ERP

system. The results of the experiment indicate that the

revisions in earnings is positive, thereby providing

support for the hypothesis that implementation of ERP

systems have a positive effect on performance. The

results from survey-based and experimental research

could be further supported by triangulation with

findings based on objective performance data.

We are aware of only one academic study that

attempts to examine the performance effects of SCM

systems using objective data. Dehning et al. (2004)

investigate the financial benefits of SCM systems by

analyzing a set of 123 manufacturing firms (SIC Codes

2011-3999) who have chosen to implement or have

implemented an SCM application. They find that that

these systems generally are associated with improved

financial performance. Unlike Poston and Grabski’s

(2001) analyses of ERP adoptions, Dehning et al. (2004)

control for industry and economy-wide effects by using

the median industry performance as benchmark. While

this is certainly better than not using any controls, they

do not control for prior performance as advocated by

Barber and Lyon (1996). As mentioned earlier, this can

confound the estimation and interpretation of abnormal

performance. To the best of our knowledge we are not

aware of any study that has rigorously analyzed the

effect of CRM systems on performance.

The above critique of the existing literature identifies

some significant gaps in our understanding of the effect

of ES system on financial performance. It also

highlights major differences between our study and

the existing literature. First, previous studies have

focused on a specific type of ES such as ERP or SCM. In

contrast, we analyze the impacts of ERP, SCM, and

CRM (three types of commonly used ES) and thereby

provide a more comprehensive assessment of the effect

of ES on performance. Second, previous studies either

use stock returns or accounting measures (but not both)

to assess the impact of ES on firm performance.

Furthermore, previous stock return studies have mainly

focused on estimating the stock market reaction over

short windows, typically over one to three days. As

mentioned earlier, given the nature of the ES investment

it may be more relevant to estimate the long-term stock

price effects of ES investments. We examine the effect

of ES on both the long-term stock returns and

profitability to examine the consistency between

different categories of performance measures. In the

K.B. Hendricks et al. / Journal of Operations Management 25 (2007) 65–8268

long run both the stock price changes and profitability

changes should point to the same conclusion about the

effect of ES on performance. Third, unlike other studies,

we examine performance effects for both the imple-

mentation and the post-implementation periods. Finally,

to isolate the effect of ES on performance, we use

methodologies that address some of the estimation and

statistical concerns and drawbacks of previous studies.

In particular, we control for prior performance, which

has been shown to result in well-specified and powerful

statistical tests (Lyon et al., 1999; Barber and Lyon,

1997, 1996; Kothari and Warner, 1997).

3. Benefits of enterprise systems

Information integration is a key benefit of ES. This

integration can replace functionally oriented and often

poorly connected legacy software, resulting in savings

in infrastructure support costs. Furthermore, improve-

ments in operational integration enabled by ES can

affect the entire organization and therefore can

positively impact firm performance. As discussed

below, ERP systems also provide benefits in the area

of transaction automation, SCM systems provide more

sophisticated planning capabilities, and CRM systems

facilitate customer relationship management.

3.1. ERP systems

ERP systems replace complex and sometimes manual

interfaces between different systems with standardized,

cross-functional transaction automation. Order cycle

times (the time from when an order is placed until the

product or service is delivered) can be reduced, resulting

in improved throughput, customer response times, and

delivery speeds (Cotteleer and Bendoly, 2006; McAfee,

2002). Similarly, automated financial transactions can

reduce cash-to-cash cycle times and the time needed to

reconcile financial data at the end of the quarter or year

(Mabert et al., 2000, 2003; McAfee, 1999; Stratman,

2001). The result is a reduction in operating capital and

the headcount of the financial area.

Another benefit of ERP systems is that all enterprise

data are collected once during the initial transaction,

stored centrally, and updated in real time. This ensures

that all levels of planning are based on the same data and

that the resulting plans realistically reflect the prevailing

operating conditions of the firm. For example, a single,

centrally developed forecast ensures that operational

processes remain synchronized and allows the firm to

provide consistent order information to customers

(Bancroft et al., 1998).

Taken together, the standardized firm-wide transac-

tions and centrally stored enterprise data greatly

facilitate the governance of the firm (Scott and Vessey,

2000; McAfee and Upton, 1996). ERP reports provide

managers with a clear view of the relative performance

of the various parts of the enterprise, which can be used

to identify needed improvements and take advantage of

market opportunities (AT Kearney, 2000; Boston

Consulting Group, 2000).

3.2. SCM systems

The primary benefit of SCM systems is better

operational and business planning. The MRP II and

ERP systems of the nineties usually included only

rough-cut capacity planning logic, with basic finite-

capacity planning functionality limited to key work

centers (Vollmann et al., 2005). SCM systems use finite-

capacity planning algorithms that do not require

iterative adjustments to the master schedule (Raman

and Singh, 1998), and real-time planning capabilities

allow firms to react quickly to supply and demand

changes. Coordinated planning and flow of materials

and information among supply chain partners can

mitigate the ‘‘bullwhip effect’’ (Lee et al., 1997). There

is a rich literature in OM on the benefits of better supply

chain planning and coordination (Cachon and Fisher,

2000; Cheung and Lee, 2002; Milner and Kouvelis,

2002). Recent empirical research has demonstrated that

reducing forecasting and planning errors that result in

supply chain disruptions avoids value destruction

(Hendricks and Singhal, 2003). Increased revenue,

increased productivity, operational cost savings, lower

inventory, and reduced order-to-fulfillment cycle time

are some of the benefits from SCM system implementa-

tions (Nucleus Research, 2003a).

3.3. CRM systems

CRM is a synthesis of many existing principles from

relationship marketing (Jancic and Zabkar, 2002; Sheth

et al., 2000; Morgan and Hunt, 1994), and the broader

issue of customer-focused management. CRM systems

provide the infrastructure that facilitates long-term

relationship building with customers. Some examples

of the functionality of CRM systems are sales force

automation, data warehousing, data mining, decision

support, and reporting tools (Katz, 2002; Suresh, 2004).

CRM systems also reduce duplication in data entry and

maintenance by providing a centralized firm-wide

database of customer information. This database

replaces systems maintained by individual sales people,

K.B. Hendricks et al. / Journal of Operations Management 25 (2007) 65–82 69

institutionalizes customer relationships, and prevents

the loss of organizational customer knowledge when

sales people leave the firm. Centralized customer data

are also valuable to firms managing multiple product

lines. In many cases customers will overlap across

different lines of business, providing an opportunity for

increasing revenues through cross-selling.

Based on the above review and discussion of the

literature we hypothesize:

H1 (2 and 3). Investments in ERP (SCM and CRM) systems lead to improvements in financial performance

as measured by stock returns and profitability.

4. Sample selection procedure and data

description

Business Wire, Dow Jones News Service, PR News-

wire, and the Wall Street Journal are our primary sources

for collecting the sample of firms that have invested in

ES. We started with the set of all announcements during

1991–1999 that mention ES providers by name.

Although more than 25 providers are used in our search,

the major ones are SAP, Oracle, JD Edwards, and

Peoplesoft for ERP systems; i2 Technologies, Manugis-

tics, Aspen Technology, American Software, and

Logility Inc. for SCM systems; and Siebel and SCT

Corp. for CRM systems. These announcements men-

tioned firms that have invested in ES. To be included in

the final sample a firm must have stock price information

on the Center for Research on Security Prices (CRSP)

database and accounting information on the COMPU-

Table 1

Description of the sample of firms that have made announcements of inves

Measure Mean Med

(Panel A) Descriptive statistics for the sample of ERP investment announc

Sales (US$ million) 6958.6 1678

Total assets (US$ million) 10329.9 1461

Equity market value (US$ million) 10092.9 2002

Employment (thousands) 31.2 9

(Panel B) Descriptive statistics for the sample of SCM investment announc

Sales (US$ million) 10020.8 4089

Total assets (US$ million) 13735.9 3253

Equity market value (US$ million) 18882.2 3265

Employment (thousands) 38.5 16

(Panel C) Descriptive statistics for the sample of CRM investment announ

Sales (US$ million) 9609.5 2876

Total assets (US$ million) 25199.7 3404

Equity market value (US$ million) 27377.2 5216

Employment (thousands) 32.2 14

STAT database. Nearly 98% of the announcements were

made during 1995–1999.

Our sample consists of 406 firms with the following

breakdown: 186 announcements of investments in ERP

systems, 140 in SCM systems, and 80 in CRM systems.

Panel A of Table 1 presents statistics on the sample based

on the most recent fiscal year completed before the date

of the announcement of investing in ES. The mean

(median) ERP observation represents a firm with annual

sales of nearly US$ 6956 million (US$ 1679 million) and

total assets of US$ 10,330 million (US$ 1461 million).

SCM observations represents firms with mean (median)

annual sales of nearly US$ 10,020 million (US$ 4089

million) and total assets of US$ 13,735 million (US$

3253 million) whereas CRM observations represent firms

with mean (median) annual sales of nearly US$ 9609

million (US$ 2879 million) and total assets of US$

25,199 million (US$ 3404 million). Overall, it appears

that larger firms have invested in ES. Furthermore, the

size of firms that invest in SCM and CRM applications is

larger than those that invest in ERP applications. This

may be because SCM and CRM applications are

relatively new compared to ERP applications.

Our sample has two distinct sets of announcements.

One set indicates that the firm has started or plans to

start the implementation of an ES application. The other

set indicates that the firm has completed the imple-

mentation of an ES application. Of the 186 ERP

announcements, 35 are on completed implementations;

of the 140 SCM announcements, 12 are on completed

implementations; and of the 80 CRM announcements, 5

are on completed implementations.

ting in ES

ian S.D. Maximum Minimum

ements

.7 16651.8 168919.0 0.53

.6 36612.7 355935.0 4.00

.3 2983.9 333672.0 6.01

.2 64.3 608.0 0.01

ements

.0 18931.1 120279.0 0.53

.6 46632.2 405200.0 23.54

.0 55572.1 508329.5 11.15

.0 59.2 340.0 0.01

cements

.8 15431.6 99820.0 4.12

.1 6050.5 355935.0 5.58

.8 5517.8 333672.1 5.66

.0 49.9 293.0 0.10

K.B. Hendricks et al. / Journal of Operations Management 25 (2007) 65–8270

5. Methodology for estimating performance

effects of investments in ES

This section describes our methodology including

the period over which the performance effects are

measured and the approach used to estimate the long-

term stock price and profitability effects of investments

in ES.

5.1. Choosing the period over which to measure

performance impacts

In deciding the period over which to measure

performance changes, we focus on longer periods to

fully capture the performance effects of ES. We also

examine the performance during the implementation

period as well as the post-implementation period. Based

on the evidence in the literature, we use a two-year

implementation period for ERP systems. O’Leary

(2000) and McAfee (1999) report that ERP installation

takes between 1 and 3 years, with an average of 21

months. Stratman (2001) collected detailed timeline

information from 43 respondents, and estimates that the

average time from start of installation to live is 17.1

months. The weighted average implementation time

from the survey of 479 US manufacturing firms by

Mabert et al. (2000) is about 17.5 months. Given the

above evidence, we choose to use a two-year

implementation period.

The evidence on the time required to implement

SCM and CRM applications is limited. Compared to

ERP systems, SCM and CRM systems are less complex

and easier to implement. Raman and Singh’s (1998)

case study on i2 Technologies indicated that imple-

mentation of SCM systems can take about 6–12 months.

Nucleus Research (2003a,b) reports a 1.5-year imple-

mentation period for SCM systems. Our discussion with

an SCM expert at SAP suggests that a 12-month

implementation period seems reasonable. We are

unable to find much information about implementation

time for CRM. A summary of white papers found at

crmguru.com indicated that 50% of the projects get a

payback within 18 months, suggesting that implemen-

tation time might be short. Based on the above evidence,

we choose to use a one-year implementation period for

SCM and CRM systems.

The existing literature provides little guidance on

what should be the appropriate length of the post-

implementation period over which one should measure

the benefit of investments in ES. There does seem to be

an agreement that the benefits will be felt over a long

time period. Given this, we choose a three-year post-

implementation period for ERP, SCM, and CRM

applications. Overall, we examine the changes in

financial performance over a five-year period for

ERP systems (a two-year implementation period and

a three-year post-implementation period), and a four-

year period for SCM and CRM systems (a one-year

implementation period and a three-year post-imple-

mentation period).

For sample firms that plan to implement ES, the five-

year period in the case of ERP systems (four-year period

in the case of SCM and CRM systems) starts on the

announcement date. For sample firms that have

announced they have completed the implementation

of an ES application, we set the starting date back by

two years (one year) relative to the announcement date

of completion of the implementation of ERP system

(SCM and CRM systems), and use these reset dates to

measure the five-year and four-year periods. This

adjustment of dates allows us to align the starting time

for both the planned and completed implementations.

5.2. Methodology for estimating the long-term

stock price effects

We estimate the long-run buy-and-hold abnormal

returns using daily returns. An abnormal return is the

difference between the return on a stock and the return

on an appropriate benchmark. There has been con-

siderable discussion in the literature on the appropriate

methodology for computing long-run abnormal returns

(Lyon et al., 1999; Fama, 1998; Barber and Lyon, 1997;

Kothari and Warner, 1997). This discussion has focused

on two issues. The first issue is the appropriate factors

that should be controlled for in computing long-run

abnormal returns. The current consensus seems to be

that abnormal return computations should control for

size, market-to-book ratio of equity, and prior perfor-

mance. The second issue is the interpretation of the

statistical significance of long-run abnormal returns.

Barber and Lyon (1997) and Kothari and Warner (1997)

report that test statistics from many commonly used

methods are misspecified because these methods do not

adjust for cross-sectional dependency. Barber and Lyon

(1997) and Lyon et al. (1999) find that abnormal returns

using matching portfolios of similar firms give well-

specified tests. We implement this approach as follows:

Step 1 I

n each month, all eligible NYSE firms are

sorted into deciles according to their market

value of equity. Next all AMEX and NASDAQ

firms are placed into the appropriate size

portfolio. The smallest size decile portfolio is

K.B. Hendricks et al. / Journal of Operations Management 25 (2007) 65–82 71

further divided into quintiles, resulting in 14 size

portfolios. Each portfolio is further divided into

quintiles according to their market-to-book ratio

of equity, resulting in 70 portfolios. Each

portfolio is further divided into 3 portfolios

based on the stock price performance of firms in

that portfolio over the previous year, resulting in

210 portfolios for each month where firms in

each portfolio are similar in terms of size,

market-to-book ratio, and prior performance.

Over 300 months, the mean (median) number of

firms over 63,000 portfolios is about 22 (21).

Step 2 W

e identify the portfolio that a sample firm is

assigned to in the first month of the start of the

sample firm’s measurement period. Since all

other firms in this portfolio are similar to the

sample firm on size, market-to-book ratio, and

prior performance, all these firms can be

considered as benchmarks for the sample firm.

For example, Allied Signal announced on

October 7, 1998, that they plan to implement

an ERP system. In this case we search the 210

portfolios for August 1998 and identify the

portfolio that includes Allied Signal. Suppose

that Allied Signal is in portfolio #20, which also

includes 25 other firms. These 25 firms then

become the benchmark against which Allied

Signal’s performance is evaluated.

Step 3 A

sample firm’s abnormal return is the

difference between its buy-and-hold return

and the average of the buy-and-hold returns of

all other firms that belong to the sample firm’s

portfolio (see Hendricks and Singhal, 2005 for

more details). For example, in the case of Allied

Signal, we compute the buy-and-hold return of

Allied Signal and the 25 other firms that belong

to portfolio #20. If Allied Signal’s return over a

particular period is say 25%, and the average

return of the 25 other firms in portfolio #20 is

15%, then Allied Signal’s abnormal return is

10%.

Step 4 S

tatistical inference is based on a simulation

approach (see Lyon et al., 1999). The idea is to

compute an empirical distribution of abnormal

returns for a portfolio that has similar

characteristics to those of the sample portfolio,

and compare where the abnormal return of the

sample portfolio falls on this distribution. To

achieve this we create a pseudo-sample where

for each sample firm we randomly select, with

replacement, a firm that belongs to the

portfolio assigned to the sample firm. This

randomly selected firm is assigned the same

announcement date as that of the sample firm.

Thus, in the case of Allied Signal, we would

select one of the 25 firms that belong to

portfolio #20 and assign it the date of October

7, 1998. Once this is done for all sample firms,

the mean abnormal performance for this

pseudo-sample is computed using the portfolio

approach discussed in Step 3. This results in

one observation of the mean abnormal perfor-

mance from a pseudo-sample. We repeat this

process 1000 times to obtain 1000 mean

abnormal return observations.

Step 5 T

he empirical distribution of the mean abnormal

returns from 1000 pseudo-portfolios is used to

test whether the mean abnormal return for the

sample portfolio is significantly different from

zero. We compute the p-value as the fraction of

the 1000 pseudo-samples with mean abnormal

returns less than the mean abnormal return of the

sample portfolio. For example, if 600 pseudo-

samples have mean abnormal returns above the

mean abnormal return of the sample, then the p-

value is 0.400. Using the empirical distribution

to compute p-values explicitly accounts for

cross-sectional dependencies, which has been a

major source of concern about the validity of p-

values from conventional test statistics (Lyon

et al., 1999).

To pool observations across time, for each sample

firm, we translate calendar time to event time as follows.

The announcement date is day 0 in event time, the next

trading date is day 1, and trading day after that is day 2,

and so on. Since a year typically has 250 trading days,

the implementation period for ERP systems (SCM and

CRM systems) span event days 0–500 (0–250) and the

post-implementation period spans event days 501–1250

(251–1000).

5.3. Methodology for estimating the long-term

operating performance effects

To estimate the profitability effects of investments in

ES, we analyze changes in operating return on assets

(ROA) and operating return on sales (ROS). ROA

(ROS) is the ratio of operating income to book value of

total assets (sales), where operating income is defined as

sales less cost of goods sold, and selling, general, and

administrative expenses. We focus on operating income

over other income measures (for example, net income or

earnings per share) because it is a cleaner measure of

K.B. Hendricks et al. / Journal of Operations Management 25 (2007) 65–8272

performance as it is not obscured by special items, tax

considerations, or capital structure changes.

To control for various factors unrelated to invest-

ments in ES that could affect the performance, we

compare the performance of each sample firm against

an appropriately chosen comparison group. We estimate

abnormal performance as the change in the sample

firm’s performance minus the change in the median

performance of the comparison group. More formally,

let PIt1 and PIt2 be the performance level in year t1 and t2 (where t2 > t1), respectively, for the sample firm I. Let PCt1 and PCt2 be the median performance level in year

t1 and t2, respectively, for the comparison group for

sample firm I. Then API, the abnormal performance of

sample firm I is

API ¼ðPIt2 � PIt1Þ�ðPCt2 � PCt1Þ

In earlier studies of operating performance the

choice of the comparison group was generally based on

industry and size. Furthermore, some researchers used a

single firm as the basis for comparison while others used

a portfolio of firms as their comparison group. More

recently, Barber and Lyon (1996) develop robust

guidelines on selecting comparison groups that give

well-specified and powerful test statistics. They

emphasize the importance of selecting comparison

groups that have similar prior performance as that of the

sample firms as well as using a portfolio of firms as the

comparison group. We implement the findings of

Barber and Lyon (1996) using a four-step procedure.

Step 1 F

or each sample firm we identify all firms that

have the same two-digit SIC code as that of the

sample firms and whose ROA (ROS) in the

starting year of the measurement period is

within 90–110% of the sample firm. All firms

that meet these criteria are considered part of the

comparison group for the sample firm. The 90–

110% filter on performance is used because this

range yields well-specified test statistics (Barber

and Lyon, 1996).

Step 2 I

f we do not find any firms in Step 1, then we

attempt to match performance within the 90–

110% filter using all firms in the same one-digit

SIC code.

Step 3 I

f we do not find any firms in Step 2, then we

attempt to match performance within the 90–

110% filter without regard to SIC code.

Step 4 I

f we do not find any firms in step 3, then we

chose the firm that is closest in performance

without regard to SIC code.

The mean (median) number of firms in the comparison

groups is 20 (14) for the ERP sample, 17 (12) for the SCM

sample, and 28 (18) for the CRM sample. To test the

sensitivity of our results from the above approach, we

create a second comparison group, which takes the firms

identified in the first comparison group and includes only

those firms whose total assets are within a factor of 10 of

the total assets of the sample firm. Since these results are

very similar to the first comparison group, we only report

the results from the first comparison group.

Abnormal performance can be reported as the

change in the level of performance or as the percent

change in the level of performance. To see the

difference between these two methods, consider a

sample firm where the ROA has changed from 3% to

5%. In this case the change in the level of ROA is 2%,

whereas the percent change in ROA is 66.66%. If the

starting ROA is negative, then percent change calcula-

tions are nonsensical. Therefore, basing results on

percent change require that we exclude firms (sample or

comparison group firms) that have negative starting

ROA. This not only diminishes the power of statistical

tests, but can also lead to biases in test statistics.

Because of these issues with the percent change

method, we report abnormal performance based on

the change in the level of performance.

To pool observations across time, for each firm in our

sample, we translate calendar year to event years as

follows. The year of the announcement date is year 0 in

event year, the next year is year 1, and year after that is year

2, and so on. For ERP systems (SCM and CRM systems)

the implementation period spans years 0–2 (0–1) and the

post-implementation period spans years 2–5 (1–4).

6. Empirical results

Outliers can influence the mean values of long-term

performance effects. To control for outliers, all

abnormal stock return results are reported after capping

the data at the 1.0% level in each tail. Outliers are a

more serious problem with accounting data. To control

for outliers all profitability results are reported after

capping the data at the 2.5% level in each tail. Even with

capping, outliers can still influence the mean values of

long-term performance effects. Therefore, we will put

more emphasis on non-parametric statistics such as

median and percent of sample firms with positive

abnormal performance. In addition to reporting the

parametric tests on changes in the mean, we report two

non-parametric tests. The Wilcoxon signed-rank test is

used to test whether the median of the changes is

significantly different from zero, and the binomial sign

K.B. Hendricks et al. / Journal of Operations Management 25 (2007) 65–82 73

test is used to test whether the percent of sample firms

experiencing positive abnormal performance is sig-

nificantly different from 50%. Consistent with our

hypotheses, we measure statistical significance using

one-tailed tests.

6.1. Results for investments in ERP systems

Table 2 presents results for the sample of firms that

invested in ERP systems. During the two-year imple-

mentation period, the stock price performance of the

sample firms fared poorly relative to the benchmark

portfolios. The mean (median) abnormal return during

this time period is �11.96% (�23.04%). A p-value of 0.064 indicates that the mean abnormal return of the

sample firm is lower than the mean abnormal returns of

936 out of the 1000 pseudo-portfolios abnormal returns.

Of the 186 sample firms, only 40% of the sample firms do

better than the median return of the firms that belong to

their assigned portfolios, significantly lower than 50%

( p-value � 0.01). The abnormal stock price performance during the implementation period is negative and

statistically significant.

Table 2

Performance results for the sample of firms investing in ERP systems

Panel A

Performance measures Implementation period

(days 0–500)

Number of observations 186

Mean abnormal return (%) �11.96 (0.064) Median abnormal return (%) �23.04 Percent of sample firms

with returns greater than

its portfolio median

39.78 (�2.79)a

Panel B

Performance measures Implementation

period from years 0 to 2

Post-imp

period f

Observation Mean Median Percent

positive

Observa

Abnormal change in the

level of return on assets

186 1.03

(1.74) c

0.56

(1.76) c

56.45

(1.76) c

167

Abnormal change in the

level of return on sales

186 0.58

(1.28)

0.36

(1.42)

53.22

(0.88)

167

Panel A: Results on the mean abnormal stock return ( p-value from the empir

parenthesis), the median abnormal stock return, and the percent of sample firm

their assigned benchmark portfolios (the binomial sign test Z-statistic in pa

Panel B: Results on abnormal return on assets and return on sales. T-statistics

binomial sign test Z-statistic for the percent positive are reported in parent a

Significantly different from zero (50% in the case of percent positive) b

Significantly different from zero (50% in the case of percent positive) c

Significantly different from zero (50% in the case of percent positive)

The results for the three-year post-implementation

period are mixed. The mean abnormal return of 10.97%

is statistically significant ( p-value = 0.043) and similar

in magnitude to the loss experienced during the

implementation period. However, the median abnormal

return is �1.03%. Only 51.07% of the firms do better than the median return of the firms that belong to their

assigned portfolio, insignificantly different from 50%.

Overall, only one of the three statistics suggests positive

abnormal performance.

When the performance is examined over the full five-

year period (the combined implementation and post-

implementation periods), there is no evidence of

abnormal performance. The mean abnormal return is

�5.06%, insignificantly different from zero ( p- value = 0.41). The median abnormal return is

�11.39%. Nearly 52% of the sample firms do better than the median return of the firms that belong to their

assigned portfolio, insignificantly different from 50%.

The evidence suggests that over the five-year period, the

stock price performance of firms that invest in ERP

systems is no different from that of their benchmark

portfolios.

Post-implementation

period (days 501–1250)

Implementation and

post-implementation period

(days 0–1250)

186 186

10.97 (0.043) �5.06 (0.41) �1.03 �11.39 51.07 (0.29) 51.61 (0.44)

lementation

rom years 2 to 5

Implementation

and post-implementation

period from years 0 to 5

tion Mean Median Percent

positive

Observation Mean Median Percent

positive

0.37

(0.52)

0.60

(0.81)

53.89

(1.01)

186 1.09

(1.53)

1.03

(2.49) a

57.53

(2.05) b

0.31

(0.38)

1.04

(1.93) c

57.48

(1.90) c

186 �0.15 (�0.18)

0.67

(1.42)

56.98

(1.90) c

ical distribution created from 1000 replications of pseudo-portfolios in

with returns greater than the median return of the firms that belong to

rentheses).

for the mean, Wilcoxon signed-rank test Z-statistic for the median, and

heses.

at the 1% level for one-tailed test.

at the 2.5% level for one-tailed test.

at the 5% level for one-tailed test.

K.B. Hendricks et al. / Journal of Operations Management 25 (2007) 65–8274

The results on changes in profitability (Panel B of

Table 2) provide some evidence of improvements in

profitability. The mean and median changes in ROA are

positive for the implementation, post-implementation,

and the combined implementation and post-implemen-

tation periods. The positive changes in ROA during the

implementation period are statistically significant at the

5% level. Although the changes in ROA during the post-

implementation period are positive, none of the changes

are statistically significant. However, during the

combined implementation and post-implementation

periods the median change of 1.03% in ROA is

significantly different from zero ( p-value � 0.01), and nearly 58% of the sample firms experienced positive

abnormal change in ROA, significantly different from

50% ( p-value � 0.025). The evidence suggests that ERP adopters show an improvement in ROA.

When ROS is used as the performance metrics, eight

of the nine performance metrics (three for each time

period) are positive but only three changes are

statistically significant at the 5% level. These are the

median change during the post-implementation period,

the percent of sample firms that experience positive

abnormal change during the post-implementation period,

and the percent of sample firms that experience positive

abnormal change during the combined implementation

and post-implementation periods. Although there is some

evidence of positive abnormal changes in ROS, the

results are not as strong as that of the changes in ROA.

Overall the evidence suggests that although firms

that invest in ERP systems do not experience a

statistically significant increase in stock returns, there

is some evidence to suggest that profitability improves

over the combined implementation and post-imple-

mentation periods. To examine the ERP results in more

detail, we segment the sample into four different

subsamples. We briefly summarize these findings. The

detailed results are available from the authors.

To estimate the effect of ERP systems during the

post-implementation period more precisely we examine

the results only for those announcements that indicated

that the firm has completed the implementation of an

ERP system. In this case we know when the

implementation was completed, and hence the time

period after the completion would more accurately

represent the performance effects during the post-

implementation period. The results for this subsample

are very similar to the overall sample.

We also examine the results for those announce-

ments that indicated that the firm has started or planned

to start the implementation of an ERP system. In this

case we know when the implementation was started but

not necessarily when it was completed. By examining

the performance of these firms over the five-year period,

we shed some light on the payback from ERP systems

over a five-year period. Again the results for these

subsamples are very similar to the overall sample.

We also segment our sample into manufacturing and

service firms to see if the benefits from investments in

ERP systems are more or less for manufacturing or

service firms. We did not find any evidence to suggest

that the benefits of ERP implementation are different for

manufacturing or service firms. Basically, the results for

these two subsamples are very similar to those for the

full sample.

6.2. Results for investments in SCM systems

Table 3 presents results for the sample of firms that

invested in SCM systems. During the one-year

implementation period, the mean abnormal return is

�1.46%, insignificantly different from zero ( p- value = 0.624). The median abnormal return is

�8.11%. Of the 140 sample firms, about 45% of the sample firms did better than the median return of the

firms that belong to their assigned portfolios, insignif-

icantly different from 50%. Basically, the abnormal

stock price performance during the implementation

period is not statistically significant.

During the post-implementation period the mean

abnormal return of 18.06% is statistically significant ( p-

value = 0.025). However, the median abnormal return is

�4.91%. Nearly 51% of the firms do better than the median return of the firms that belong to their assigned

portfolio, insignificantly different from 50%. Over the

full four-year period, the mean abnormal return of

18.75% is statistically significant at the 7% level. The

median is �9.24%. Half the sample firms do better than the median return of the firms that belong to their

assigned portfolio. Overall, there is some evidence of

positive abnormal stock price performance during the

four-year period.

The results for the accounting metrics (see Panel B of

Table 3) provide strong support that firms that invest in

SCM systems show improvements in ROA and ROS.

Improvements are observed in both the implementation

and post-implementation periods, with mean and

median changes in ROA and ROS generally positive

and most are statistically significant at the 2.5% level or

better. The results for the combined implementation and

post-implementation periods indicate that the median

change in the level of ROA is 1.78%. The median

change in the level of ROS is 1.44%. Both these changes

are statistically significant ( p-value � 0.01). More than

K.B. Hendricks et al. / Journal of Operations Management 25 (2007) 65–82 75

Table 3

Performance results for the sample of firms investing in SCM systems

Panel A

Performance measures Implementation

period (days 0–250)

Post-implementation

period (days 251–1000)

Implementation

and post-implementation

period (days 0–1000)

Number of observations 140 140 140

Mean abnormal return (%) �1.46 (0.624) 18.06 (0.025) 18.75 (0.068) Median abnormal return (%) �8.11 �4.91 �9.24 Percent of sample firms

with returns greater than its

portfolio median

45.71 (�1.01) 51.42 (0.34) 50.00 (0.00)

Panel B

Performance measures Implementation

period from years 0 to 1

Post-implementation

period from years 1 to 4

Implementation

and post-implementation

period from years 0 to 4

Observation Mean Median Percent

positive

Observation Mean Median Percent

positive

Observation Mean Median Percent

positive

Abnormal change in the

level of return on assets

141 0.95

(2.18) b

0.76

(2.61) a

60.99

(2.61) a

130 1.58

(2.35) a

01.58

(4.31) a

59.23

(2.10) b

141 2.98

(4.31) a

1.78

(3.59) a

60.99

(2.61) a

Abnormal change in the

level of return on sales

141 0.65

(1.80) c

0.58

(2.50) a

59.57

(2.27) b

131 0.43

(0.77)

0.43

(1.75) c

57.69

(1.75) c

141 1.33

(2.59) a

1.44

(2.78) a

62.41

(2.95) a

Panel A: Results on the mean abnormal stock return ( p-value from the empirical distribution created from 1000 replications of pseudo-portfolios in

parenthesis), the median abnormal stock return, and the percent of sample firm with returns greater than the median return of the firms that belong to

their assigned benchmark portfolios (the binomial sign test Z-statistic in parentheses).

Panel B: Results on abnormal return on assets and return on sales. T-statistics for the mean, Wilcoxon signed-rank test Z-statistic for the median, and

binomial sign test Z-statistic for the percent positive are reported in parentheses. a

Significantly different from zero (50% in the case of percent positive) at the 1% level for one-tailed test. b

Significantly different from zero (50% in the case of percent positive) at the 2.5% level for one-tailed test. c

Significantly different from zero (50% in the case of percent positive) at the 5% level for one-tailed test.

60% of the sample firms experience positive abnormal

changes in ROA and ROS during the combined

implementation and post-implementation periods. All

the results for the combined implementation and post-

implementation periods are statistically significant at

the 1% level or better. Overall the results indicate that

investments in SCM systems improved profitability.

Since very few of the SCM investment announce-

ments in our sample indicated that the firm had

completed the implementation of an SCM system, it is

not meaningful to analyze this subsample separately.

We also segment the SCM announcements sample into

those made by manufacturing firms (about 80% of the

sample). The results for this subsample are generally

consistent with the results for the full sample.

6.3. Results for investments in CRM systems

Table 4 presents results for the sample firms that

invested in CRM systems. During the implementation

period, the mean abnormal return is 6.84%, insignif-

icantly different from zero ( p-value = 0.15). During

the post-implementation period the mean abnormal

return of �3.07% is not statistically significant ( p- value = 0.617). The percent of sample firms that do

better than the median return of the firms that belong to

their assigned portfolio is insignificantly different from

50%. Overall, there is no evidence of positive or

negative abnormal stock price performance during the

implementation and post-implementation periods. Over

the full four-year period, the mean (median) abnormal

return is �15.22% (�12.41%), and nearly 53% of the sample firms do better than the median return of the

firms that belong to their assigned portfolio. However,

none of these performance changes are statistically

significant. Basically, investments in CRM systems

have had little effect on the stock returns of investing

firms.

The results of abnormal stock price performance are

corroborated by the results on changes in operating

performance (see Panel B of Table 4). Changes in ROA

and ROS are generally positive during the implementa-

tion, post-implementation, and the combined imple-

mentation and post-implementation period. Except for

the mean change in ROA and ROS during the combined

implementation and post-implementation periods, none

K.B. Hendricks et al. / Journal of Operations Management 25 (2007) 65–8276

Table 4

Performance results for the sample of firms investing in CRM systems

Panel A

Performance measures Implementation

period (days 0–250)

Post-implementation

period (days 251–1000)

Implementation

and post-implementation

period (days 0–1000)

Number of observations 80 80 80

Mean abnormal return (%) 6.84 (0.152) �3.07 (0.617) �15.22 (0.804) Median abnormal return (%) �4.66 �12.12 �12.41 Percent of sample firms

with returns greater than its

portfolio median

53.75 (0.67) 48.75 (�0.22) 52.50 (0.45)

Panel B

Performance measures Implementation

period from years 0 to 1

Post-implementation

period from years 1 to 4

Implementation

and post-implementation

period from years 0 to 4

Observation Mean Median Percent

positive

Observation Mean Median Percent

positive

Observation Mean Median Percent

positive

Abnormal change in the

level of return on assets

81 1.01

(0.79)

0.25

(�0.80) 56.79

(1.22)

77 0.79

(0.58)

0.02

(0.84)

50.64

(0.11)

81 2.21

(1.94) c

0.25

(1.46)

55.55

(0.99)

Abnormal change in the

level of return on sales

81 2.40

(1.57)

0.67

(1.63)

58.02

(1.44)

77 0.75

(0.57)

8.02

(0.96)

53.25

(0.57)

81 4.64

(2.25) b

1.37

(1.48)

54.32

(0.78)

Panel A: Results on the mean abnormal stock return ( p-value from the empirical distribution created from 1000 replications of pseudo-portfolios in

parenthesis), the median abnormal stock return, and the percent of sample firm with returns greater than the median return of the firms that belong to

their assigned benchmark portfolios (the binomial sign test Z-statistic in parentheses).

Panel B: Results on abnormal return on assets and return on sales. T-statistics for the mean, Wilcoxon signed-rank test Z-statistic for the median, and

binomial sign test Z-statistic for the percent positive are reported in parentheses. a Significantly different from zero (50% in the case of percent positive) at the 1% level for one-tailed test. b

Significantly different from zero (50% in the case of percent positive) at the 2.5% level for one-tailed test. c

Significantly different from zero (50% in the case of percent positive) at the 5% level for one-tailed test.

of these changes are statistically significant. Overall,

investments in CRM systems seem to have had little

impact on profitability.

Since very few of the CRM investments announce-

ments in our sample indicated that the firm has

completed the implementation of a CRM system, it is

not meaningful to analyze this subsample separately.

We also segment the CRM announcements sample into

those made by manufacturing firms (about 67% of the

sample) and service firms (about 33% of the sample). In

both these subsamples, the results are generally

consistent with the results for the full sample.

6.4. Are the results driven by the control group

methodology?

Before we discuss the implications of our results, we

need to address an important issue that has been raised

during the review process. The issue is whether our

results could be driven by the control sample, as it is not

clear that the controls have not invested in ES. Hence, it

is plausible that the some of the insignificant results that

we find with respect to ES systems are because the

controls have adopted ES systems. We certainly cannot

claim that all firms in our control set have not

implemented ES. However, as discussed below we

believe that the chances are low that our results are

driven by the possibility that a subset of control firms

may have implemented ES.

First, our sample has 406 ERP, SCM, and CRM

announcements. Given that at any point in time more

than 5000 firms are publicly traded, our control firms

will come from a sample of more than 4600 firms. If

most of these 4600 firms have adopted ES, then one

would be very much concerned about our results. While

we do not know what the controls have done, the

chances that most of the controls have ES are quite low.

The adoption of ERP systems is still limited among

midsize and small firms, and the adoption is even lower

for SCM and CRM systems. Furthermore, given that our

last announcement is in December 1999, the adoption

rate among controls in 1999 is likely to be much lower

than today.

Second, even if some of the 4600 control firms have

adopted ES, it should not have much of an impact on our

stock price performance results because of the method

K.B. Hendricks et al. / Journal of Operations Management 25 (2007) 65–82 77

used to create the 1000 pseudo-samples. In each

pseudo-sample we use size, prior performance, and

market to book ratio to select 406 firms from the sample

of more than 4600 firms. We then compare the results of

the sample firms against the results of these 1000

pseudo-samples. If each of these 1000 pseudo-samples

is dominated by control firms that have implemented ES

then our results would be a source of concern. While

this could happen in a few of the pseudo-samples, the

chances of this happening in most of the pseudo-

samples are very low as process of creating the pseudo-

samples is quite randomized.

Third, in analyzing the performance effect of ES on

ROA and ROS, we match each sample firm to a

comparison group that consists of firms from the same

industry, and which have similar performance char-

acteristics. On average each comparison group consists

of 20 firms. Furthermore, we estimate abnormal

performance relative to the change in the median

performance of the comparison group. The median of

the comparison group is less likely to be impacted by

non-identified ES adopters. For example, with this

approach if more than 50% of the firms in each

comparison group adopt ES, and ES adoption leads to

positive results, then it would be a cause of concern as

these adopters would drive the median, and negate any

positive effect observed in the sample firms.

Finally, we note that our sample is based on firms that

have started their implementation between 1991 and

1999. At least in the case of SCM and CRM systems our

sample is likely to have firms that are early adopters,

which minimizes the chances that the control firms may

have also implemented SCM and CRM systems.

The above discussion provides some rationale of

why the chances are low that our results are impacted by

controls that may have also implemented ES. The

purpose of controls is to control for broad economic and

industry factors. No control process will be perfect on

all dimensions. With a large sample of firms and

reasonable level of randomization in the selection

process, there should not be any systematic bias in the

selection of controls. Nonetheless, given the concerns

about whether the controls are driving the results we

report some additional analyses on early ES adopters to

test the robustness of our results.

6.5. Analysis of the performance of early adopters

of ERP

Since it is not clear that controls have not invested in

ES, another way of dealing with the issue of whether

this could drive our results is to examine the perfor-

mance of early adopters. While at a conceptual level the

issue of controls adopting ES still remains, at a practical

level the issue would be less of a concern for the early

adopters. It is less likely that most of our controls are

early adopters. We restrict our analysis only to the ERP

sample as ERP systems have been around longer than

SCM and CRM systems. We define as early adopters

those firms that have made announcements of invest-

ments in ERP systems before 1998. Table 5 reports the

abnormal stock price and operating performance results

for the early adopter.

The evidence indicates that abnormal stock price

performance results of early adopters are not that

different from the overall sample. The abnormal return

during the implementation period is negative, and is

positive during the post-implementation period. How-

ever, the mean abnormal returns over the implementa-

tion and post-implementation periods are insignificantly

different from zero.

The profitability performance (see Panel B of

Table 5) of early adopters appears to be stronger than

the results for the full ERP sample (see Panel B of

Table 2). Over the combined implementation and post-

implementation periods, the improvements in ROA and

ROS are positive and statistically significant. There is

some evidence of statistically significant positive

changes in ROA and ROS during the post-implementa-

tion period, which is stronger than what was observed

for the full ERP sample. Consistent with the results for

the full sample, there is weak evidence of improvements

in profitability during the implementation period. The

evidence of Table 5 suggests that early adopters may

have benefited more from ERP implementation when

compared to later adopters.

7. Summary and future research

Our analysis of the financial benefits of ES

implementations yields mixed results. In the case of

adopters of ERP systems, we find some evidence of

improvements in profitability but not in stock returns.

The results for improvements in profitability are

stronger in the case of early adopters of ERP systems.

On average, adopters of SCM system experience

positive abnormal returns as well as improvements in

profitability. There is no evidence of improvements in

stock returns or profitability for firms that have invested

in CRM. Although our results are not uniformly positive

across the different ES systems, they are encouraging in

the sense that despite the high implementation costs, we

do not find persistent evidence of negative performance

associated with ES adoption. This should help alleviate

K.B. Hendricks et al. / Journal of Operations Management 25 (2007) 65–8278

Table 5

Performance results for the early adopters (1997 and before) ERP systems

Panel A

Performance measures Implementation

period (days 0–500)

Post-implementation

period (days 501–1250)

Implementation

and post-implementation

period (days 0–1250)

Number of observations 104 104 104

Mean abnormal return (%) �6.48 (0.167) 14.99 (0.075) �1.82 (0.36) Median abnormal return (%) �16.46 0.23 �13.43 Percent of sample firms

with returns greater than its

portfolio median

40.8 (�2.04)c 52.88 (0.58) 52.88 (0.87)

Panel B

Performance measures Implementation

period from years 0 to 2

Post-implementation

period from years 2 to 5

Implementation

and post-implementation

period from years 0 to 5

Observation Mean Median Percent

positive

Observation Mean Median Percent

positive

Observation Mean Median Percent

positive

Abnormal change in the

level of return on assets

110 1.16

(1.60)

0.45

(1.51)

54.54

(0.95)

102 0.47

(0.51)

1.51

(1.25)

59.80

(2.27) b

110 1.55

(1.59)

1.51

(2.51) a

60.91

(2.27) b

Abnormal change in the

level of return on sales

110 1.43

(1.78) c

0.66

(1.78) c

55.45

(1.14)

102 0.80

(0.52)

1.74

(2.97) a

64.71

(2.97) a

110 �0.28 (�0.19)

1.75

(2.08) b