dedrick a critical review of empirical evidence

7/31/2019 Dedrick a Critical Review of Empirical Evidence

1/28

Information Technology and Economic Performance: A CriticalReview of the Empirical Evidence

JASON DEDRICK, VIJAY GURBAXANI, AND KENNETH L. KRAEMER

University of California, Irvine

For many years, there has been considerable debate about whether the IT revolutionwas paying off in higher productivity. Studies in the 1980s found no connection betweenIT investment and productivity in the U.S. economy, a situation referred to as theproductivity paradox. Since then, a decade of studies at the firm and country level hasconsistently shown that the impact of IT investment on labor productivity and economicgrowth is significant and positive. This article critically reviews the published research,more than 50 articles, on computers and productivity. It develops a general frameworkfor classifying the research, which facilitates identifying what we know, how well weknow it, and what we do not know. The framework enables us to systematicallyorganize, synthesize, and evaluate the empirical evidence and to identify both

limitations in existing research and data and substantive areas for future research.The review concludes that the productivity paradox as first formulated

has been effectively refuted. At both the firm and the country level, greater investmentin IT is associated with greater productivity growth. At the firm level, the reviewfurther concludes that the wide range of performance of IT investments among differentorganizations can be explained by complementary investments in organizational capitalsuch as decentralized decision-making systems, job training, and business processredesign. IT is notsimplya tool forautomating existing processes,but is moreimportantlyan enabler of organizational changes that can lead to additional productivity gains.

In mid-2000, IT capital investment began to fall sharply due to slowing economicgrowth, the collapse of many Internet-related firms, and reductions in IT spending byother firms facing fewer competitive pressures from Internet firms. This reduction in ITinvestmenthas haddevastatingeffects on the IT-producing sector, andmay leadto slowereconomicand productivitygrowth in theU.S.economy. While theturmoil in thetechnologysector has been unsettling to investors and executives alike, this review shows that itshould not overshadow the fundamental changes that have occurred as a result of firmsinvestments in IT. Notwithstanding the demise of many Internet-related companies, thereturns to IT investment are real, and innovative companies continue to lead the way.

Categories and Subject Descriptors: K.4.1 [Computers and Society]: Public PolicyIssues; K.4.2 [Computers and Society]: Social Issues; K.4.3 [Computers andSociety]: Organizational Impacts; K.6.0 [Management of Computing andInformation Systems]: GeneralEconomics

General Terms: Economics, Performance, Management

Additional K ey Words and Phrases: Information technology, productivity, economicperformance, firm, industry, country, management practices

Thisreview hasbeen supported by grantsfrom the U.S. National Science Foundation (Computer, Informationand Engineering Directorate, Information and Intelligent Systems Division, Program on Digital Society and

Technologies), and IBM Global Services.Authors addresses: Center for Research on Information Technology and Organizations, GraduateSchool of Management, University of California, Irvine, 3200 Berkeley Place, Irvine, CA 92697; email:{jdedrick,vgurbaxa,[email protected]}.

Permission to make digital/hard copy of part or all of this work for personal or classroom use is grantedwithout fee provided that the copies are not made or distributed for profit or commercial advantage, thecopyright notice, the title of the publication, and its date appear, and notice is given that copying is bypermission of ACM, Inc. To copy otherwise, to republish, to post on servers, or to redistribute to lists requiresprior specific permission and/or a fee.c2003 ACM 0360-0300/03/0300-0001 $5.00

ACM Computing Surveys, Vol. 35, No. 1, March 2003, pp. 128.


2/28

2 Dedrick et al.

1. INTRODUCTION

You can see the computer age everywhere butin the productivity statistics.

[Solow 1987].

Despite differences in methodology and datasources, a consensus is building that theremark-able behavior of IT prices provides the key to thesurge in economic growth.

[Jorgenson 2001].

1.1. Background

There has been a long-running debate inthe business press and the informationsystems and economics literature overwhether the information technology (IT)revolution is paying off in higher produc-

tivity. The first studies, conducted in the1980s, found no connection between ITinvestment and productivity at the levelof firms, industries, or the economy as awhole [Loveman 1994; Roach 1987, 1989,1991; Strassmann 1990]. Skeptics pointedout that heavy IT investment had occurredconcurrently with the productivity slow-down that began in 1973 in the U.S.

This so-called productivity paradoxstimulated economists, managementscientists, and information systemsresearchers to conduct more rigorousscientific analyses of the relationship be-

tween IT and productivity [Brynjolfsson1993, 1996; Bresnahan 1999; Brynjolfssonand Hitt 1995, 1996, 1998; Oliner andSichel 2000; Jorgenson 2001; Jorgensonand Stiroh 2000; Bosworth and Triplett2000; Council of Economic Advisers(CEA) 2001]. These studies, which usedlarger datasets and more refined researchmethods, revealed positive and significantimpacts from IT investments at the firmand country level. Moreover, some of thesestudies showed that the economic boomand surge in productivity of the late 1990swas largely due to heavy investment in

IT and the growth of the Internet.The debate over IT and productivity

then shifted to whether theIT-led economywould lead to permanent improvementin the prospects for economic growth, orwhether it was a temporary phenomenon,with much of the acceleration in produc-tivity driven by the business cycle and con-

centrated in just a few sectors of the econ-omy, a point of view espoused by Gordon[2000].

Given the continuing debate aboutwhether IT investments pay off, thisresearch review critically evaluates thelarge body of evidence-based research onthe subject. The purpose is to criticallysurvey the published research on IT andeconomic performance to determine whatwe know and what we do not know aboutthe returns to IT investments. The goal isto help direct future research into poten-tially productive channels so that it cancontribute to knowledge about whether ornot, as well as how, IT investments canbe effectively introduced and managed for

greater payoffs.Specifically, the aim of this article is to

(1) organize and integrate the researchon returns to IT investment in a way thatadds understanding to work in the area,(2) provide an unbiased and objective

view of the documented returns (or lack ofreturns) to IT investment at three levelsof analysisfirm, industry, and country,(3) identify the factors that contribute topayoffs from these investments, (4) eval-uate issues in current research, and(5) identify opportunities for future re-search. It is intended that this review will

help readers to understand this importantarea of research, stimulate experts to dealwith unresolved issues in ongoing andfuture research, and assist senior execu-tives in future decision-making about ITinvestments in their organizations.

We begin with a discussion of the scopeof the literature reviewed and the organi-zation of this article.

1.2. Scope of Literature Review

This review examines more than 50 em-pirical studies based on economic anal-ysis that have appeared between 1985and 2002.1 Early studies were based on

1 There is other writing on the subjectthatis not partof this review. For example, Laudon and Marr [1994]brought political perspectives to bear on understand-ing the productivity paradox when they argued thatproductivity returns from IT investment might not

ACM Computing Surveys, Vol. 35, No. 1, March 2003.


3/28

Information Technology and Economic Performance 3

Fig. 1. IT and economic performanceframework for literature review.

small samples and limited data, whereasrecent studies have been able to take ad-

vantage of both better data and largersamples, including time-series data. Thereview concentrates mainly on studieswhose results have been published inpeer-reviewed scholarly journals such as

American Economic Review, Communica-tions of the ACM, Information Systems

Research, Journal of Economic Perspec-tives, Journal of Management Informa-

tion Systems, Management Science, MISQuarterly, Organization Science, Quar-terly Journal of Economics, The Infor-mation Society, The Brookings Papers,and World Development. These journalsare known for having high standardsfor review and acceptance and there-fore are most appropriate for a criticalreview that seeks to achieve an objec-tive and balanced perspective on the re-search. The review also includes a fewother works because of their significanceor because they help to round out the re-search in an area. Finally, Brynjolfsson

[1993] and Brynjolfsson and Yang [1996]have provided excellent reviews of earlierresearch.

even be an objective in the macroculture of some or-ganizations. Similarly David [1990] and King [1996]brought historical perspectives to the IT and produc-tivity debate.

1.3. Organization of this Review

The review is organized as follows.Section 2 introduces a conceptual frame-work for thinking about IT and economicperformance. Sections 3 through 5 assessthe literature at three levels of analysis:firm, industry, and country. Section 6presents both substantive and method-ological issues that need to be addressedin future research, and Section 7 presentsspecific, high-priority recommendations

for future research based upon keydeficits in the current body of knowledge.Finally, Section 8 discusses limitationsof the survey and highlights some majorconclusions.

2. CONCEPTUAL FRAMEWORK

In order to organize the prior researchand to identify gaps for future efforts, wedeveloped a conceptual framework (seeFigure 1) that allows us to map and as-sess the research findings. The frameworkhelps to define the key variables and re-

lationships addressed in the different re-search studies reviewed herein. Movingfrom left to right in Figure 1, the frame-work identifies the various inputs (laborand capital) to the production process andcomplementary factors of production thatinfluence the production process, and en-ables an assessment of the contribution of



4/28

4 Dedrick et al.

those inputs to outputs (value added, grossdomestic product) and to various outcomes(economic growth, labor productivity, prof-

itability, and consumer surplus). It furtherdistinguishes between firm, industry, andcountry levels of analysis. We begin ourpresentation of the framework by definingthe key terms in the title of this reviewIT investment and economic performance.

2.1. Definition of IT Investment

Traditional economic studies of productiv-ity focused on capital such as plants andequipment using aggregate measures ofcapital that include all its component cat-egories. In studies of IT and productiv-

ity, it becomes necessary to disaggregatecapital into the component categories ofinvestmentIT and the traditional formsof capital, labeled non-IT. IT investment,broadly defined, includes investments inboth computers and telecommunications,and in related hardware, software, andservices. However, as operationally de-fined in nearly all of the research includedin this review, IT investment is limitedmainly to computer hardware. In moststudies, investment is defined as an an-nualized value of the stock of computer in-

vestments including the depreciated value

of previous investments that are still inservice, or as annual spending.

2.2. Definition of Economic Performance

Economic performance can be interpretedin a variety of ways at each level of anal-ysis. At the country level, where much ofthe debate has occurred, it usually refersto economic growth, labor productivitygrowth, and consumer welfare (Figure 1).

Economic growth is the rate of changein real output, or GDP, and is measuredat the country level. Labor productivity

growth, or growth in output per worker, isa measure of the efficient use of resourcesto create value. It allows the economy toprovide lower-cost goods and services rela-tive to the income of domestic consumers,and to compete for customers in interna-tional markets [McKinsey Global Insti-tute 2001, p. 1]. Corresponding measures

focusing on the output of an industry sec-tor are utilized at the industry level.

Clearly, labor productivity growth is

also an indicator of the economic perfor-mance of firms. A firm that is more pro-ductive than its competitors will gener-ally enjoy higher profitability, which is ofcourse, also an important measure of eco-nomic performance for firms. A more pro-ductive firm will either produce the sameoutput with fewer inputs and thus experi-ence a cost advantage, or produce higher-quality output with the same inputs, en-abling a price premium. However, as willbe discussed later in the review in re-gard to firm-level research, competitioninduces other firms to catch up in produc-

tivity. Sustaining higher profits throughproductivity gains requires a firm to main-tain productivity levels higher than itscompetitors. Therefore, over time, prof-its might be competed away with the re-sult that consumers benefit. This bene-fit is measured as consumer surplus andrefers to the aggregate value realized byconsumers from their purchase of a goodless the price paid.

2.3. Modeling the Production Process

In order to better understand the IT and

productivity debate, it is useful to be-gin with a discussion of the productionprocess by which inputs are transformedinto outputs in firms and economies, andthe specific role of IT as a factor of pro-duction. Economists use two related ap-proaches to modeling the production pro-cess by which inputs are transformedinto outputs. One approach to understand-ing the output of an economic systemis production economics, which uses spe-cific functional forms, called production

functions, to model the production pro-cess [Bresnahan 1999; Brynjolfsson and

Hitt 2000]. This approach uses economet-ric techniques to relate the output of afirm, industry, or economy to the inputsbased on estimation models derived fromthe production function. Inputs typicallyaccounted for in this approach includelabor and capital, including both IT andnon-IT capital. Most of the studies at the



5/28


firm level use this approach. The primaryapproach used to model the productionprocess inherent in an economy (or in-

dustry) is growth accounting [Oliner andSichel 2000; Jorgenson and Stiroh 1999,2000; Council of Economic Advisors 2001].This method also assumes specific proper-ties of the production process and, basedon these assumptions, allocates shares ofoutput to the various inputs to production.

Output growth in firms, industries, andthe economy may arise from increases ininput levels, improvement in the qualityof inputs, and growth in the productiv-ity of inputs. The focus of the literaturereviewed here has been on understand-ing how increased levels of investment in

IT impacts labor productivity. First, laborproductivity can increase when workersare provided with more capital, a phe-nomenon called capital deepening. Sec-ond, technical progress in the productionprocess or in the quality of output can in-crease the level of output without addi-tional investment in input, a phenomenonlabeled multifactor productivity (MFP).

An increase in MFP means that for a fixedlevel and quality of inputs, a firm, indus-try or economy is achieving higher levels ofoutput. This form of productivity improve-ment is of great importance because it re-

flects structural gains that arepermanent.The framework also posits that there

are complementary factors that influ-ence the payoff from IT investments(Figure 1). At the firm level, these includeorganization and management practices[Brynjolfsson 1996; Bresnahan et al.2002; Brynjolfsson et al. 2000]; at theindustry level, they include industryorganization [Melville 2001]; and at thenational level, they might include eco-nomic structure, government policy, andinvestment in human capital [Dewan andKraemer 2000]. The following discussions

elaborate the framework for each level ofanalysis. Indeed, as we shall see below, itis these complementary factors that havebeen suggested as helping to explain someof the extraordinary estimates of returnsto IT investments.

The analysis that results from these ap-proaches indicates the relative contribu-

tion of labor and capital to output, andthe relative contribution of the differentdrivers of labor productivity growth, as

illustrated in Figure 1 above. As will beseen later, the assumptions that variousresearchers make in doing the analysiscan have substantial impacts on the re-sults and implications.

2.4. Role of IT in the Production Process

Much of the debate among economists hasaddressed these economic performance is-sues in the aggregate, at the country andindustry levels. Yet, the decision makerswho choose to invest in IT are managerswho deploy IT for use in their organiza-

tions and who use investment criteria thatare related to the outcomes at the level ofthe firm. While labor productivity is cer-tainly one often-used criterion, managersalso use measures such as profitability,market share, margins, and product vari-ety and quality as justifications for invest-ment in IT systems.

In order to understand the overall im-pact of IT at the firm level, it is use-ful to begin by thinking about the qual-itative impacts of introducing IT intoa firms production processes. Past re-search has distinguished between using

IT to automate processes, to provide bet-ter information, and to transform entireprocesses [Zuboff 1988]. The impact ofautomation is primarily the direct sub-stitution of capital for labor, consistentwith capital deepening. For example, acashier at a retail chain store using acomputer-based information system suchas a scanner can process a transactionin less time. The impact of improved in-formation is that it allows workers andmanagers to make decisions more effec-tively. For example, information providedby the store-based system allows the man-

agers to better manage inventory. Trans-formation impacts occur when a firm re-designs a process to achieve significantlyhigher levels of productivity. In our ex-ample, the firm may redesign its supplychain using a supply chain managementsystem, of which the store system is a keyelement.



6/28

6 Dedrick et al.

One key difference between IT capitaland other forms of capital is the dual rolesthat IT can play in a firm. First, like other

types of capital, IT can be used directlyas a production technology to improve la-bor productivity, as in the case of a bankstransaction processing system. However,research suggests IT has its greatest im-pact in its second role as a technology forcoordination [Bresnahan 1997; Gurbaxaniand Whang 1991; Malone et al. 1989]. Inthis literature, IT is viewed as an espe-cially potent technology that has a signif-icant impact on the costs of coordinatingeconomic activity both within and betweenorganizations. Research in this arena sug-gests that the unique value of IT is that

it enables fundamental changes in busi-ness processes and organizational struc-tures that can enhance MFP.

2.5. The Aggregate Impact of IT

While IT is deployed at the level of thefirm, the analysis of the impact of IT atthe level of an industry or the economyallows researchers to answer a relatedbut different set of questions that are ofcritical importance. At the level of theeconomy, it furthers an understanding ofthe role of IT and the IT sector in foster-

ing economic growth and, ultimately, thewealth of a country. It also facilitates adiscussion of whether steady-state growthrates are higher in an IT-intensive econ-omy. It enables the documentation andunderstanding of industry differences,and an examination of the role of industrycharacteristics such as structure andregulation in moderating the returns toIT investment.

While IT can increase productivity viacapital deepening and via MFP growth,the results might be substantially differ-ent in different industry sectors. In par-

ticular, an important distinction has tobe made between the IT-producing andIT-using sectors. The IT-producing sectorsare those which manufacture semicon-ductor, computer, or telecommunicationshardware or provide software and servicesthat enable these technologies to be usedeffectively in organizations. The IT-using

sectors are all the other sectors of theeconomy that apply IT as part of theiroperations in order to achieve greater effi-

ciency and effectiveness. They include sec-tors such as manufacturing (durable2 andnondurable), wholesale and retail trade,finance, insurance and real estate, busi-ness and professional services, and so on.

As we shall see below, there is no ques-tion that there have been very rapid im-provements in productivity and in MFP inthe IT-producing industries, particularlyin computer hardware and components asa result of research and development onproduct and process technologies.

A critical question is whether there havebeen similar gains in productivity and

MFP outside the IT-producing industryand, if so, whether those gains can be at-tributed to investment in IT capital. Hasthe use of IT allowed industries to achievesuperior production methods than werepreviously unavailable? Put differently,are there spillovers from IT-producing in-dustries to IT-using industries?

2.6. Implications for this Review

As the foregoing suggests, a comprehen-sive review of the payoffs from IT invest-ment must examine the returns to this in-

vestment at the disaggregate level of thefirm as well as the aggregate level of in-dustries and the economy since the na-ture of the payoffs at these levels may bequite different. If IT investments are in-creasing productivity at the firm level, itmight be the case that in the aggregatethey will also increase the productivityof entire industries and countries. How-ever, it is possible that gains will show upat one level and not the other, dependingon whether individual firms capture thereturns on their investment, or whethersome or all of the gains are competed away

and flow to consumers, creating socialbenefits but not providing a measurablereturn to the firms making the invest-ment. Furthermore, in addition to firm-specific factors, industry characteristics

2 Ordinarily, durable goods manufacturing includesthe IT-producing sector.



7/28


also affect the payoffs that firms withinan industry receive from their IT invest-ments and how these payoffs are shared

between firms. It is therefore important toexamine the results at both disaggregateand aggregate levels.

In order to understand whether IT in-vestment results in greater productivity,we next look systematically at the re-search on the returns from IT investmentsfor each level of analysisfirm, industryand countryin that order. We examinethe research as a basis for understand-ing the nature, extent, and limitations ofpayoffs from IT investments. We reviewthe evidence provided by numerous sys-tematic, empirical studies. We summarize

some of the major studies in the Appendixand Tables I and II.

3. FIRM-LEVEL RESEARCH

While the productivity paradox as orig-inally framed focused on aggregatecountry-level productivity statistics, ac-tual IT investments are made by organiza-tions, mostly firms, that are interested intheir own return on investment, not thatof the country as a whole. Knowing that ITinvestment improves aggregate produc-tivity does not imply that individual firmsenjoy similar benefits. In fact, there maybe significant social benefits from IT in-

vestments that increase consumer welfarebut are not captured by the firms makingthose investments. Therefore, it is ofgreat concern to business and technologyexecutives whether their IT investmentsare paying off at the level of the firm.

3.1. IT and Firm-Level Productivity

Motivated by the productivity paradox,many firm-level studies were launched inthe 1980s and 1990s. Early studies were

unable to show that IT investments ledto payoffs, in most cases because of inad-equate data on IT investments and smallsample sizes [Brynjolfsson and Hitt 1996,2000; Brynjolfsson and Yang 1996]. Mostdiscouraging were several studies of ser-

vice firms, such as banks and insurancefirms, which showed weak or nonexis-

tent links between IT and productivity,but where output measurement is notori-ously difficult [Franke 1987; Strassmann

1990; Alpar and Kim 1991; Harris andKatz 1991]. Some studies of manufactur-ing firms did show positive returns on ITinvestment, partly because it is easier tomeasure the output of manufacturing andadjust for improvements in quality [Weill1992; Barua et al. 1995]. These studiesbegan to highlight the importance of theaccurate measurement of outputs, partic-ularly in the technology-intensive serviceindustries where the largest investmentsin IT capital were being made.

Starting around 1993, more rigorousstudies with larger samples were being re-

ported by researchers [Brynjolfsson 1993,1996; Bresnahan 1999; Brynjolfsson andHitt 1995, 1996, 1998; Lichtenberg 1995].These studies involved large U.S. corpo-rations, using data on IT capital invest-ment from marketresearch firms and fromsurveys of chief information officers andother executives, coupled with financialdata from reliable sources. The researchused econometric techniques based in pro-duction economics that relate firm output(measured as value added by a firm) to aset of inputs including labor hours, non-IT capital stock, and IT capital stock, and

estimated the marginal product or outputelasticity of IT capital.3

Each of these studies found that IT in-vestments contribute to firm productivity,and show higher gross marginal returnsthan non-IT investments. The fact thatthese researchers found a strong relation-ship between IT capital and productivitythat was not evident in earlier studies maypartly reflect the fact that the data wasmore recent, that levels of IT investmenthad increased, making it easier to distin-guish its contribution, and that over timefirms were learning to apply IT capital

more productively. They may also simplyreflect better data sets and analytical toolsthat make it possible to isolate and mea-sure the true impacts of IT investment.

3 The output elasticity of IT is the increase invalue added associated with a 1% increase in ITinvestment.



8/28

8 Dedrick et al.

More recently, Brynjolfsson and Hitt[2000] have found that payoffs to IT in-

vestment occur not just in labor productiv-

ity increases but also in MFP growth, andthat the impact on MFP growth is maxi-mized after a lag of 4 to 7 years. Gilchristet al. [2001], using the same dataset, focuson the manufacturing companies in thesample and show that IT has a substan-tial and contemporaneous impact on laborproductivity growth and on MFP growthin the durable goods sector, which exceedsthe impact that would be predicted by itsfactor share. They find that, in the non-durable goods sector, the returns to laborproductivity accrue primarily via capitaldeepening, and are consistent with IT fac-

tor share. Moreover, these returns are cor-related with decentralized computing ar-chitectures, suggesting that the diffusionand networking of computing throughoutthe organization contributes substantiallyto the payoff.

In addition to these U.S. studies, afew other studies have been conductedon firms in other countries. Greenanet al. [2001] analyzed data on Frenchfirms IT investment and productivity andcame to results consistent with the find-ings of Brynjolfsson and Hitt [1996] andLichtenberg [1995] for U.S.-based firms.

By contrast, Lal [2001] did not find a rela-tionship between IT investment and pro-ductivity in Indian garment makers. Thisis consistent with the cross-country stud-ies, which are discussed later [Dewan andKraemer 2000; Pohjola 2001], that havefound a strong relationship between ITand productivity in developed countries,but not in developing countries. With lowunit costs of labor and higher capital costs,it is not surprising that there are fewer op-portunities for capital-labor substitutionin developing countries. Also, Lals sampleincluded many small and medium-sized

firms, a group not included in most U.S.studies.

Most of the studies found that IT in-vestments were associated with highermarginal product than other capital in-

vestments. These are translated intoexcess returns by some authors, whopointed out that, in theory, all invest-

ments should pay the same risk-adjustedreturn at the margin. These returns doneed to be adjusted to account for the

high rate of obsolescence of IT capital,so that the net returns are much lower.However, Brynjolfsson and Hitt [1996] andLichtenberg [1995] found that after sub-tracting standard estimates of the cost as-sociated with the obsolescence of IT cap-ital of up to 42% per year from the grossreturns, the net returns from IT were stillhigher than those of non-IT investments.These results of firm-level studies havesometimes been taken to imply that firmsare systematically underinvesting in IT,given the high marginal returns to suchinvestments.

Some answers have been proposed tothis question in the literature and otherswill be suggested here, but we would warnthat claims of systematic underinvest-ment in IT should be viewed cautiously.First, as Brynjolfsson and Hitt [2000]pointed out, the true cost of such invest-ments may be underestimated. All studiesinclude the direct investment in computerhardware; others attempt to include labor,software, and services, but it is difficult toestimate these with a high degree of preci-sion. Importantly, they do not include thecosts of complementary investments such

as training and process reengineering thatcan be much larger than the actual di-rect investment in IT. If these costs are in-cluded on the investment side of the equa-tion, the returns might look more modest.Moreover, taking into account the largestandard deviations in the payoffs docu-mented by many studies, it is possible thatthe net returns to IT investments are con-sistent with non-IT investments.

Given these caveats, it is still possiblethat IT investment does show higher thannormal returns. There are several reasonswhy this could be so. IT investment might

be riskier than other investment. Firmsinvest when the net return is sufficient tocover the risk-adjusted cost of capital. Thiswould argue that returns need to be higherto compensate for theadditional risk. Moststudies do not assess the impact of therisk of these investments. Moreover, theremight be adjustment costs. It is difficult



9/28


and costly for firms to introduce new ITinnovations. With decreasing prices for IT,the optimal level of IT investment and cap-

ital stock increases in steady state. How-ever, firms face real costs and delays due tothe duration of software development, re-tirement of older systems, and changes inpractices that suggest that firms might notachieve these optimal levels in the shortrun. It is therefore difficult to concludethat the excess returns found in firm-level studies imply that firms are system-atically underinvesting in IT, or that man-agers are acting irrationally.

This recent research also highlightsother interesting questions that remainunresolved regarding the payoffs from in-

vestments in IT capital. First, it is notwell understood why firms in differentindustries accrue different payoffs. Forexample, it would be valuable to iden-tify the specific characteristics of durablegoods manufacturing firms that enablethem to achieve higher returns relativeto nondurable goods manufacturing firms.Second, a better understanding of the tim-ing of the payoff from investments in ITcapital is also needed. Clearly, a firmsmany individual investments in specificsystems will have different periods overwhich the payoffs will be realized. Some

systems will realize immediate payoffs,while others willrealize payoffs after a lag.The duration over which the payoffs willbe realized will also vary. Some will haveshort-term impacts and others will havelonger-term impacts. This understandingwill go a long way toward resolving the de-bate on whether the impacts of these in-

vestments are contemporaneous with theinvestiments or occur in the future.

3.2. Variance Among Individual Firms

The preponderance of evidence points to

positive and significant returns to IT in-vestment among firms. Clearly, higher lev-els of IT investment are associated withhigher levels of productivity across a largesample of companies, and this has beentrue since the mid-1980s at least. How-ever, looking at a scatter plot of IT invest-ment and productivity, as Brynjolfsson

and Hitt [1995, 1996] have presentedin several of their papers, one is struckby how widely scattered the actual data

points are around the trend line. Thisleads to the next major finding in the firm-level data.

The productivity impacts of IT in-vestments vary widely among differentcompanies.4 In other words, some firmsuse IT much more productively thanothers. Brynjolfsson and Hitt [1995] es-timated that these firm effects mayaccount for as much as half of the produc-tivity benefits attributed to IT investmentin their earlier work, but stated that theelasticity of IT remains positive and sig-nificant even after firm effects are taken

into account. Still, this raises the questionof what causes these firm effects.

Two factors stand out. First, there areidiosyncratic firm characteristics such asmarket position, rigidities in cost struc-tures (e.g., labor contracts), brand recog-nition, or the vision and leadership abili-ties of key executives, which affect a firmsstrategic options and therefore its poten-tial to derive benefits from IT investment.These can change over time, but are noteasily manipulated by management in theshort run.

Second, there are specific features of

organizational structure, strategy, andmanagement practices that can be com-pared systematically across companies.The management of a firm, throughrestructuring, new management controlsystems, the redesign of processes, or byupgrading employee training, can directlyinfluence these features.

3.3. Impact of Business Practices on Value

of IT Investments

Management practices and complemen-tary investments explain part of the varia-

tion in IT payoffs. Lovemans [1994] earlyanalysis of manufacturing firms, whichfound evidence of net marginal benefitfor non-IT investments but not for ITinvestments, highlighted complementary

4 The variance of returns to IT capital is larger thanthe variance of returns to non-IT capital.



10/28

10 Dedrick et al.

organizational factors as a possible expla-nation for the research results. He arguedthat the evidence could be interpreted as

management failure to effectively inte-grate IT with the firms business strategy,human resource management strategy,and efficient resource allocation. That is,management did not implement the orga-nizational changes that should accompanyIT investment in order to create value.

Subsequent studies at the firm levelexplicitly show that the value of IT in-

vestments is substantially impacted bythe structure and business practices ofthe firms making the investment. Forinstance, Weill [1992] showed that thequality of a firms management and its

commitment to IT enhances the contri-bution of IT investments to firm per-formance. Francalanci and Galal [1998]showed that firms with a higher propor-tion of information workers gain morefrom their IT investments than those witha lower proportion. Tallon et al. [2000]found that aligning IT with businessstrategy increased the payoffs from IT in-

vestments. In addition, firms with higherlevels of investment gained greater pay-offs from alignment. Devaraj and Kohli[2000] found that business process reengi-neering enhanced the payoffs in firms

that also made greater IT investments.Ramirez et al. [2001] found that organi-zations which invested more in IT and im-plemented management practices such asemployee involvement and total qualitymanagement received higher IT returns.

Black and Lynch [1997] studied the im-pacts of workplace practices, IT capital,and human capital development on pro-ductivity. They found that what affectedproductivity was less the presence or ab-sence of a particular management prac-tice, such as total quality management,than the way in which the practice was

implemented. Particularly important wasemployee involvementfor instance, theproportion of workers involved in regulardecision-making in a plant.

In addition to these studies of individ-ual management practices, Brynjolfssonand Hitt [2000] and Bresnahan et al.[2002] showed that firms with a cluster

of management practices, including de-centralized decision-making (which theycalled organizational capital) along with

high levels of IT investments, outper-form all others. Interestingly, firms withtraditional centralized organizations andhigh IT investments actually do worsethan similar organizations that invest lessin IT.

While the evidence shows the benefits ofcertain classes of management practices,these can be difficult to translate intospecific actions for individual companies.It is logical that executives and man-agers can improve the performance oftheir IT investments by combining theseinvestments with proven complementary

managerial practices. However, the re-search evidence is limited as to specificlinks between management practices andproductivity. In particular, understandingthe relationship between firm-specific fac-tors and management practices is criticaland by definition cannot be addressedin large-sample studies. For instance,the fact that decentralized firms earnhigher returns to their IT investmentsthan centralized firms on average is notsufficient to advise a particular firm toswitch from a centralized structure to adecentralized one. Given the firms id-

iosyncratic characteristics, a centralizedstructure might be more appropriate.

3.4. IT and Firm Financial Performance

There is mixed evidence at the firmlevel as to the impacts of IT capital onfinancial performance measures suchas profitability or market value, partlybecause the linkage is less direct. WhileIT investments can directly affect a firmsoutput and many operational indicators(e.g., inventory turnover, plant produc-tivity, product quality), a firms financial

performance is determined by a widerrange of strategic and competitive factorsthat go beyond productivity.

Several studies show a relationshipbetween IT investment and intermedi-ate measures of operational performance.Barua et al. [1995] found that IT in-

vestment affects intermediate measures



11/28


such as inventory turnover but found noevidence that the benefits extended tofirm performance as measured by return

on assets. Srinivasan et al. [1994] foundthat EDI technology at Chrysler bene-fited suppliers and buyers due to im-proved accuracy and precision of materialsmanagement information, thereby reduc-ing shipment discrepancies by over 50%within 2,700 shipments across 193 suppli-ers. Similarly, Mukhopadhyay et al. [1995]found that a materials management sys-tem at Chrysler reduced the costs associ-ated with inventory holding, obsolescenceof inventory, and transportation. Finally,Banker et al. [1990] found that a point-of-sale and order coordination system in the

Hardees restaurant chain enabled man-agers to reduce materials waste and im-prove store sales.

However, early efforts to relate IT in-vestment to financial performance havehad mixed results. Brynjolfsson and Yang[1997] found that a dollar of computercapital was associated with between$5 and $20 (depending on assumptionsin the models) in additional market cap-italization for public companies, pointingto a link between IT and financial valua-tion. The authors interpreted this findingas evidence of important but unmeasured

complementary organizational practices,or intangible assets, that are not includedin the accounting of firm-level investment,and not as evidence of IT investment re-sulting in an increase in market capital-ization. Brynjolfsson et al. [2000] foundthat when organizational capital is in-cluded in the analysis (i.e., the cluster ofcomplementary practices mentioned pre-

viously), it increases market valuationand decreases the amount attributable toIT. They also found that market valua-tion effects are greatest for firms that havehigh levels of investment in both IT and or-

ganizational capital, pointing again to thecomplementarity of the two factors.

So far, studies have failed to identify arelationship between IT investment andfirm profitability. Hitt and Brynjolfsson[1996] showed that while IT investmentaffects productivity and contributes toconsumer welfare (through lower prices

or better service, for example), it doesnot necessarily improve profitability. Theyproposed that the productivity benefits as-

sociated with IT use may be passed onto consumers through lower prices andnot lead to greater profitability. On theother hand, in our assessment, it is pos-sible and even likely that IT investmentsdo actually affect profitability, but that themodeling techniques and datasets used inthese studies are unable to measure theimpacts. As models are developed that areable to control for more of the additionalfactors that affect profitability, they mayreveal a relationship between IT invest-ment and financial performance.

3.5. IT and Labor

Firm-lavel studies have shown that ITcapital has been a net substitute for labor,as the use of IT allows firms to reduceheadcounts or to grow output faster thanemployment [Dewan and Min 1997]. Inaddition, IT use is associated with ashift toward workers with higher skilllevels, a process referred to as skill-biasedtechnical change, and these workers earnhigher wages on average. Comparingindustry sectors, Autor et al. [1998]found that the rate of skill upgrading has

been most rapid in industries that arethe most intensive users of computers.Looking at the U.S. labor force, Krueger[1993] found that workers who usedcomputers earned 10 to 15% more thannonusers. Similar results have been foundin studies of other developed countries[Chennells and Van Reenen [1999].5

Dinardo and Pischke [1997] offered acompeting perspective, finding not justa strong correlation between wages andcomputer use in German data, but equallyrobust correlations for workers who usepencils, pens, calculators, or telephones.

They argued that these findings castdoubt on the interpretation that the wagedifferential reflects returns to computeruse, but reflect, in fact, the nature of thework and the implied skill sets of the

5 These authors have provided a broad survey of re-search in this area.



12/28

12 Dedrick et al.

workers. Moreover, as Chennells and VanReenen [1999] pointed out, there is muchevidence that workers with the best skills

are given the best technology to use.It is also important to identify what

mechanism might account for the rela-tionship between computer use and skilllevel. Bresnahan [1999] argued that theprocess of skill upgrading is due to orga-nizational changes related to computer-ization rather than to individual use ofcomputers. In particular, computer sys-tems enable work to be shared between aworker and the system, with many stan-dard and repetitive tasks now conductedby the system but many of the higher cog-nitive tasks still conducted by the worker.

Correspondingly, much clerical work isconducted by automated systems today,changing the nature of clerical work tofocus on more complex situations andthose that require human intervention.In the case of highly educated workers,work is supported rather than automatedby computers. In this view, organizationalcomputing systems have been a substi-tute for low- and middle-skill white collarworkers while creating more demand forhigh-skill workers. This process could ex-plain the higher skill levels and wage ratesassociated with IT use.

In an empirical study, Bresnahan et al.[2002] tested the relationship between ITuse, organizational change, and skill lev-els at the firm level. They found that theuse of IT, along with complementary work-place reorganization and a higher rate ofintroduction of new products and services,all tend to result in greater use of high-skilled labor. They also found that orga-nizational changes accompanied by tech-nology change may have a greater impacton skill levels than technology changealone. These findings are consistent withthe view that IT-enabled organizational

changes are responsible for the shift to-ward higher-skilled workers.

3.6. Summary of Firm-Level Studies

While earlier studies showed mixed re-sults, nearly all major studies since themid-1990s have shown positive and sig-

nificant returns to IT investments, andin most cases higher gross returns thanfor other investments. An important point

is that the data utilized in the studiesrun from the late 1980s to the mid-1990s,before the Internet boom and before theadvent of the so-called New Economy.

As such, the research shows that the is-sue of whether firms benefit from theirIT investments can be separated from thequestion of whether the late 1990s produc-tivity surge at the country level was a tem-porary development or the beginning of along-term structural shift in the economy.

While average returns have been high,there is a great deal of variance amongfirms in returns to IT investments. Com-

plementary management practices suchas decentralization of decision-making,business process redesign, and total qual-ity management are found to be criticalto the level of returns to IT investmentachieved by firms.

Studies also have shown that IT capi-tal can be substituted for other types ofcapital and labor, and that IT investmentis associated with a shift to higher-skilledworkers. One explanation is that orga-nizational computing systems have beena substitute for low- and middle-skilledwhite collar workers while creating more

demand for high-skilled workers.Firm-level studies have so far failed

to show a clear link from IT investmentto profitability. The failure to documentthese results most likely has stemmedfrom the inability to quantify and incor-porate the various unobservable factorsthat determine a firms competitive posi-tion and outcomes.

While firm-level studies show that ITinvestments have higher gross marginalproducts than non-IT investments, thereare reasons to be skeptical of claimsthat firms are systematically underinvest-

ing in IT. Once factors such as incom-plete accounting of complementary invest-ments, high rates of obsolescence, and riskadjustments are taken into account, thereturns to IT investments are likely to lookmore normal.

Finally, it should be noted that most ofthe firm-level research has involved data



13/28


on large firms, as it is more difficult toget good financial data on smaller firms,which are usually privately held. There-

fore, the results of firm-level studies can-not be taken to represent the entire uni-

verse of firms. It may be that a morerepresentative sample would show eitherhigher or lower returns to IT investment.This is an area ripe for future research ifbetter datasets can be developed on smalland mid-sized firms.

4. INDUSTRY-LEVEL RESEARCH

While some of the firm-level studies havefocused on a particular industry, or havecompared results from different sectors

(e.g., services vs. manufacturing), there isa dearth of studies using aggregate dataat the industry level. Attempts at suchstudies have suffered from inconsisten-cies in U.S. industry-level data as identi-fied in Baily and Gordon [1988] and else-where, which are discussed in Section 6.

As one might expect from such limited re-search with serious data problems, therehas been considerable divergence in theresults.Indeed, as we shall see,someof themost important open questions pertain tothe breadth of the IT payoff among indus-try sectors. That is, do the payoffs from ITinvestment occur across a large number ofindustries, or are they confined to a few?

4.1. Average Labor Productivity Growth

A number of recent studies of the produc-tivity revival of the late 1990s have shownthat labor productivity growth acceleratedin many industry sectors in the 19951999period over earlier periods [Jorgenson andStiroh 2000; Council of Economic Advi-sors 2001; Stiroh 2001a, 2001b; Bailyand Lawrence 2001; Nordhaus 2001]. Onestudy [Gordon 2000] found that acceler-

ation of labor productivity growth wasconcentrated in the durable goods manu-facturing sector, and most of that in theIT-producing industries. However, Gordon[2001] has since updated his study to in-clude data for 2000 and now finds accel-eration in labor productivity outside thedurable goods sectors as well.

The research has also shown that thereis variation across industries. Studies bythe Council of Economic Advisors [2001]

have shown that the overall pattern is oneof positive and in some cases very sub-stantial change (Table I). The CEA studieshave also shown that this positive changein labor productivity is associated withgreater IT investment. Those industriesthat have made greater investments in ITalso experienced greater change in laborproductivity. For example, as shown at thebottom of Table I from the 2001 CEA study,average productivity growth from 19951999 was four times greater (4.18% vs.1.05%) in industries with intense IT in-

vestment than in those with less intense

investment. Moreover, the increase in av-erage productivity for the industries withintense IT investment was also greater be-tween the 19891995 and 19951999 peri-ods than it was for those with less intenseinvestment (Table I).

The CEA findings are reinforced byStiroh [2001a, 2001b], who compared pro-ductivity gains during the 1990s in 61 in-dustry sectors and founds that two-thirdsshowed a positive shift in labor produc-tivity after 1995. Moreover, he found thatIT-intensive industries (those with higherthan average levels of IT capital as a share

of total capital) showed a 1.3% higher la-bor productivity acceleration than otherindustries from the early to late 1990s,and had higher productivity growth inboth periods. This provides further evi-dence that IT use was strongly identifiedwith the acceleration of labor productivityin the late 1990s.

A study by the McKinsey Global Insti-tute [2001] found that 38 industry sec-tors, accounting for 70% of GDP, expe-rienced productivity increases post-1995.Six sectors accounted for 74% of the totalproductivity increase, and another 26 sec-

tors accounted for 26%. Thus, a reason-able conclusion from the McKinsey study,which is consistent with the Council ofEconomic Advisors [2001] and Stiroh[2001a, 2001b] analyses, is that the pro-ductivity revival was broad-based as morethan half of the industry sectors experi-enced increased productivity.



14/28

14 Dedrick et al.

Table I. Labor Productivity Growth by Industry, 19891999

Industry 19891995 19951999 Change

Private industries .88 2.31 1.43

Agriculture, forestry, fisheries .34 1.18 0.84

Mining 4.56 4.06 .50

Construction .10 .89 .79

Manufacturing 3.18 4.34 1.16

Durable goods 4.34 6.84 2.51

Nondurable goods 1.65 1.07 .59

Transportation 2.48 1.72 .76

Trucking and warehousing 2.09 .78 2.82

Transportation by air 4.52 4.52 .00

Other transportation 1.51 2.14 .63

Communications 5.07 2.66 2.41

Electric, gas and sanitary services 2.51 2.42 .09

Wholesale trade 2.84 7.84 4.99

Retail trade .68 4.93 4.25

Finance, insurance and real estate 1.70 2.67 .97

Finance 3.18 6.76 3.58

Insurance .28 .44 .72

Real estate 1.38 2.87 1.49

Services 1.12 .19 .93

Personal services 1.47 1.09 2.55

Business services .16 1.69 1.85

Health services 2.31 1.06 1.26

Other services .72 .71 .01

Industries by intensity of IT use

Intense IT use 2.43 4.18 1.75

Less intense IT use .10 1.05 1.15

Value added per full-time equivalent employee; average annual percent change.Source: Council of Economic Advisors [2001].

A recent study by Triplett and Bosworth[2002] focused on productivity in 27 in-dustries in the services sector. It is thefirst study to look at this sector in suchdetail and identify the impacts of IT andother factors on productivity growth. Thestudy found that, post-1995, the mostIT-intensive industries in the U.S. econ-omy are overwhelmingly services indus-tries and that labor productivity growthin the services industries has proceededat about the economy-wide rate. More-over, labor productivity growth is broadlybasedit is not limited to just a few large

services industries. These findings are es-pecially important because they show thatthe productivity improvement post-1995was broad, and they dispel the belief thatthe inherent nature of services makes pro-ductivity improvements less likely thanin the manufacturing sectors of the econ-omy as a whole. Moreover, they show that

IT, through capital deepening, played animportant role in labor productivity bothpre- and post-1995: It was often not newIT, or new IT investment, that was as-sociated with rapid productivity change,but instead IT capital technology that hadbeen around for a decade or two [Triplettand Bosworth 2002, p. 18].

4.2. Multifactor Productivity Growth in

IT-Producing Industries

There is considerable agreement amongeconomists that multifactor productiv-

ity has increased in the IT-producingindustries [Gordon 2000; Jorgenson andStiroh 2000; Oliner and Sichel 2002;Council of Economic Advisors 2001]. Asan example, Jorgenson [2001] attributestwo-thirds of the growth in MFP in the19951999 period to the IT-producing sec-tor (Figure 2). Specifically, he attributed



15/28


Fig. 2. Contribution of IT-producing industries to MFP. (Source: Based ondata from Jorgenson [2001].)

the growth in MFP to continuing technicalinnovation (R&D) in the semiconductorand computer industries. More rapid pricedeclines occurred from 19941999, withthe result that computer and telecommu-nications equipment prices have declinedby 90 percent over the period. This in-crease in productivity in the IT-producingsector has naturally contributed to MFPgrowth in the whole economy given itsincreasing share of the national economy.

4.3. Multifactor Productivity Growth in theIT-Using Industries

While there is agreement about MFPgrowth in the IT-producing industries,there is some debate about whether MFPgrowth has also increased in the IT-usingindustries. Most studies have attributedsome multifactor productivity growth inthe recent 5 years to the IT-using indus-tries as well as the IT-producing industries[Whelan 2000; Jorgenson and Stiroh 2000;Oliner and Sichel 2002; Council of Eco-nomic Advisors 2001; Baily and Lawrence2001]. However, Gordon [2000] found no

evidence of MFP acceleration outside ofthe IT-producing and durable goods man-ufacturing industries.

This controversy about the contribu-tion of MFP in IT-using industries wasa motivation for the recent Triplett andBosworth [2002] study of the services sec-tor. They found that the sources of la-

bor productivity improvement in the ser-vices industries were growth in MFP,IT capital deepening, and increased useof outsourcing.6 However, they concludedthat MFP was the dominant factor con-tributing to the productivity accelerationin the services industries post-1995, con-tributing well over half of the acceleration.

4.4. Summary of Industry Studies

The overall pattern of industry-level stud-ies consistently has shown positive re-

turns in the form of labor productivity in-creases from IT investments. They alsohave shown that labor productivity hasincreased more in industries that use ITmore intensively. Most studies showedthat productivity increases in the man-ufacturing industries were higher thanthose in services. However, a recent studyhas shown that most services indus-tries have experienced productivity in-creases comparable to those in manufac-turing, and that these increases have beenpresent throughout the 1990s. Thus, thepattern of productivity increases is broad,

encompassing many industries in both themanufacturing and services sectors.

There is agreement that MFP growthhas increased in the IT-producing sector

6 Labor productivity improvement from outsourcingoccurs because of the use of more specialized, andhence more productive, producers.



16/28

16 Dedrick et al.

due to technical innovation in the semi-conductor and computer industries, butthere has been some disagreement about

whether MFP growth has also increasedin the IT-using sector. The most recent ev-idence suggests that MFP growth has in-creased in the IT-using industries, and,most significantly, that MFP has increasedin the services industries, which havehistorically posed difficult measurementproblems.

5. COUNTRY-LEVEL RESEARCH

As discussed earlier, economists mainlyuse growth accounting to estimate the con-tribution of inputs to productivity and out-

put. As discussed in prior sections, eco-nomic growth can result from a greaterlevel of inputs (labor and capital), im-proved quality of the inputs, and greateroverall efficiency in the combination ofinputs in production. The efficiency withwhich these factors of production are com-bined can increase as a result of im-provements in production methods, suchas managerial practices, organizationalchanges, and innovative ways of produc-ing goods and services.

The research shows that all these fac-tors explain some of the trends in na-

tional economic growth. The key questionin our review is the specific contributionof IT capital to this growth, both in termsof labor productivity increases via capitaldeepening and from multifactor produc-tivity growth.7

5.1. Average Labor Productivity Growth

The first studies conducted at the countrylevel in the late 1980s and early 1990s con-cluded that the contribution of IT to pro-ductivity and economic growth was nonex-

7 While there is an important conceptual distinc-tion between the sources of productivity growth, thedistinction between the sources of growth in laborproductivity in empirical analyses is often a roughand ready practical one, as measurement issues anddata strengths and weaknesses limit the researcher.It should be pointed out that it is therefore difficultto precisely interpret the allocation of productivityimprovements to capital deepening and to MFP interms of the qualitative impacts of IT systems.

istent or slight [Roach 1987, 1989, 1991;Oliner and Sichel 1994; Jorgenson andStiroh 1995]. One explanation advanced

for this conclusion was that IT investmentwas too small a portion of the capital stockin the economy to have substantial eco-nomic effects [Sichel 1997]. For example,IT capital as a share of total capital in-

vestment in nominal dollars in the U.S.was 3.5% in 1980 and 9% in 1990. Duringthe 1990s, IT capital investment increaseddramatically, reaching 22% of total U.S.capital investment. One major factor inthis increased rate of adoption of IT wasan acceleration of the decline in computerprices from an average of 17% annuallyfrom 19591995 to roughly 32% for the pe-

riod 19951999 [Jorgenson 2001]. Clearly,the decreasing prices of IT have resultedin a significant increase in its demand, en-couraging organizations to substitute ITfor labor and for other forms of capital suchas plant, machinery, or equipment.

With this increased investment, recentstudies have found that IT investmentshave had a major impact on labor produc-tivity and economic growth at the countrylevel. U.S. labor productivity, which grewat 1.5% per year in the 19731995 period,grew at the rate of 3.1% per year in 19952000. Similarly, gross domestic product

(GDP) grew at 3% per year in the earlierperiod and accelerated to 4.8% per yearduring the later period [Council of Eco-nomic Advisors 2001]. This accelerationin recent productivity and GDP growthhas been attributed in significant part byseveral macroeconomic studies to the im-pact of IT investment [e.g., Oliner andSichel 2000; Jorgenson 2001; Jorgensonand Stiroh 2000; Council of Economic Ad-

visors 2001], some of which were authoredby researchers who had previously ex-pressed a contrary opinion. Thus, thereis considerable agreement that IT invest-

ments have had a major impact on laborproductivity and economic growth at thecountry level.

Although much of the current focusis on the IT-led productivity surge ofthe late 1990s, it is important to pointout that these contributions are not new.Recent studies have argued that there



17/28


Fig.3. Contribution ofIT andnon-IT capital investmentto GDPgrowth.(Source: Based on data from Jorgenson [2001].)

Fig. 4. Contribution of IT and non-IT capital investment to labor pro-ductivity growth. (Source: Based on data from Jorgenson [2001].)

have been significant and positive im-pacts from IT investments for countriesfor decades [Jorgenson 2001; Oliner andSichel 2000; Bosworth and Triplett 2000].While the impacts of IT capital invest-ment were lower because of its lower shareof capital stock, IT investment has con-tributed to U.S. economic and productiv-ity growth for decades, even when thegrowth rate in labor productivity was low

(Figures 3 and 4). While there continues tobe a debate over the magnitude of the ITcontribution to productivity, there is con-

vincing evidence of significant and pos-itive long-term impacts from IT invest-ments on national productivity.

Early studies by Jorgenson and Stiroh[1995] reported a modest contribution

of IT to productivity growthabout 6%of yearly productivity growth of 2.94%(Table II) for the period 19591973. Insubsequent periods, these researchersfound the contribution of IT to be consid-erably greater. For example, during theperiod 19731995, Jorgenson and Stiroh[2000] found that IT contributed about13% of the 3.04% economic growth and27% of the 1.4% labor productivity growth

in the U.S. Oliner and Sichel [2000] foundslightly higher contributions (Table II).

The acceleration of labor productivitygrowth from 19951999 was due in partto rapid growth in IT investment. Themajor reason for the increased impacton productivity was simply that invest-ment in IT was increasing at a faster



18/28

18 Dedrick et al.

Table II. Contributions of IT to GDP and Productivity Growth

Jorgenson and Stiroh [2000]; Jorgenson [2001] 19591973 19731995 19951999

GDP growth (annual rate) 4.32 3.04 4.08

Capital contribution (percent of total) 33 50 71

IT contribution to GDP growth (percent of total) 4 13 28

Productivity growth (annual rate) 2.94 1.40 2.11

IT contribution to productivity growth (percent) 6 27 42

Oliner and Sichel [2000] 19731995 19951999

GDP growth (annual rate) 2.99 4.82

Capital contribution (percent of total) 42 38

IT contribution to GDP growth (percent of total) 17 23

Productivity growth (annual rate) 1.52 2.67

IT contribution to productivity growth (percent) 31 41

Sources: Original studies, plus calculations by Bosworth and Triplett [2000] and the authors.These studies were selected for special focus because they are comparable in that they include similartime periods, the same methodology, and the same definition of IT to include computer hardware andsoftware and telecommunications equipment.

rate and the accumulated investment inIT represented a substantially greatershare of the total capital stock thanin prior periods.8 Thus, IT capital con-tributed more then to the growth of theeconomy than it did in earlier periods.Jorgenson [2001] showed that during theperiod 19951999, IT capital contributedabout 28% of the 4.08%, yearly economicgrowth and about 42% of the 2.11 percentgrowth in labor productivity in the UnitedStates. Oliner and Sichel [2000] estimated

very similar contributions (Table II).

There has been some debate in the re-cent literature about the share of theseimprovements attributable to structuralchanges resulting from technical progress,or changes in the trend line, versus thosethat are due to the effects of the busi-ness cycle. Short-run growth can raiseboth measured productivity and invest-ment; short-run declinecan reduce both. It

8 The nominal share of IT investment as a percent oftotal business investment grew from 2.6% in 1970to 3.5% in 1980 to 9% in 1990 and 22% in 1999.One major factor in this increased rate of adoptionof IT was an acceleration of the decline in computer

prices, from an average rate of 17% annually from19591995, to roughly 32% annually for the period19951999 [Jorgenson 2001]. Clearly, the decreasingprices of IT have resulted in a significant increase indemand for it, encouraging organizations to substi-tute IT for labor and for other forms ofcapital suchasplant, machinery, or equipment. Although IT capitalis not a very large portionof total capital investment,there is no evidence that IT investment has reacheda point of diminishing returns.

is difficult to say with reasonable certaintywhat the precise shares of impacts due totrend and cycle are,9 though it is clear thatboth matter. Gordon [2000] attributed asignificant share of the 19952000 pro-ductivity growth acceleration to businesscycle effects, while other studies [Coun-cil of Economic Advisors 2001; Stiroh2001] have shown little or no cyclical ef-fects. These different results have sub-stantially different implications for themagnitude of the impact of IT investmenton productivity.

Regardless, the importance of thesefindings is that the broad and continuinguse of IT has made a significant differencein long-term labor productivity growth.The Internet and electronic commercemight contribute additionally. A study byLitan and Rivlin [2001] estimated thelikely productivity impact from the Inter-net across eight industry sectors, whichaccount for about 70 percent of the na-tions GDP. While admittedly speculative,the study estimates that the impact ofthe Internet over 5 years could trans-late into an annual contribution of 0.2 to

0.4% to the baseline trend of productivitygrowth.

9 There are considerable measurement issues re-lated to output. While aggregate statisticsare widelybelieved to capture business cycle effects, they arelikely to underestimate qualityimprovements in out-puts, and particularly those that are enabled by ITuse.



19/28


5.2. Average Labor Productivity Growth in

Other Developed Countries

While there has notbeen as much researchcomparable to that on the United Stateselsewhere in the world, most of the forego-ing trends have also been found in otherdeveloped countries of Europe and Asia.For example, Schreyer [1999] looked atG-7 countries and found that IT made apositive contribution to productivity andeconomic growth in all of the countriesduring the period 19901996. Another Or-ganisation for Economic Co-operation andDevelopment (OECD) study [Daveri 2000]updated and extended the analysis to 18countries. While there were differences be-

tween the two studies, the essential find-ings were the sameIT capital has con-tributed to growth, and because IT hasbeen growing faster than labor input, itcontributes to labor productivity throughcapital deepening.

These findings have been corroboratedby several larger and more robust stud-ies. Two contemporary studies of 36 (plus)countries worldwide came to the interest-ing conclusion that wealthier, industrial-ized countries showed a positive and sig-nificant relationship between IT, growthand productivity, but that there was no ev-

idence of such a relationship for develop-ing countries [Dewan and Kraemer 1998,2000; Pohjola 2001]. Dewan and Kraemerhypothesized that this gap was due to thelow levels of IT investment relative to GDPin developing countries, and to the lackof complementary assets such as the nec-essary infrastructure and knowledge-baseto support effective use of IT. Kraemerand Dedrick [2001] studied a sample of 43developed and developing countries andfound that growth in IT investment perworker was positively correlated with la-bor productivity growth, but the level of IT

investment was not.

5.3. Summary of Country Studies

The large price/performance changes in ITequipment have stimulated increases inIT capital investment in the U.S. and othercountries in the expectation of improved

economic performance. The surge in IT in-vestment during the mid- to late-1990s ledto a sharp acceleration in labor productiv-

ity growth, but it is important to recog-nize that IT investments have been payingoff in terms of labor productivity for over30 years. As IT has become a larger shareof total capital investment by firms, so hasits contribution to labor productivity andto economic growth.

6. EVALUATION OF THE STATE OF

RESEARCH

A careful assessment of the literaturebrings to the forefront a range of under-lying research issues that makes it diffi-

cult to precisely estimate the returns toIT investment.

6.1. Measurement Problems

Accurate estimation of the returns to ITinvestment requires accurate measure-ment of the inputs and outputs in the pro-duction processes of firms and industries.Measurement issues are quite dauntingin this field. In particular, measuring out-puts in the services sector, which owns themajority of IT capital, is very difficult, asis accounting for changes in the intangi-

ble attributes of products such as qual-ity and variety in the manufacturing sec-tor[Bosworth and Triplett 2000]. Accuratemeasurement of firm outputs requiresdata on quality-adjusted prices for theseoutputs, which is usually unavailable.

On the input side, it has been quitechallenging to develop quality-adjustedprice indexes for IT inputs. In the case ofhardware, government agencies like theBureau of Economic Analysis, in concertwith academic and computer industryeconomists, have made significant stridesin developing quality-adjusted price

indexes for computer equipment [Coleet al. 1986; Dulberger 1989]. On the otherhand, it has proven to be very difficultto account for investments in software.It is not only conceptually challenging todefine units of software; it is also difficultin practice to account for the large invest-ments that firms have made in custom



20/28

20 Dedrick et al.

software. While there has been consider-able progress in developing price indexesfor packaged software, the same is not

true for custom software. Indeed, it wasonly in 1999 that software was reclassifiedas an investment rather than an expensein the national accounts.

6.2. Statistical Issues

There are estimation issues as well; afew key concerns are discussed here. Inproduction function approaches, perhapsthe most significant of these is the no-tion of simultaneity in investment andgrowth due to unobservable factors. Forexample, a firm with attractive growth op-

tions may choose to invest in increasingamounts of IT to enable its growth. Sta-tistical techniques may find evidence ofa correlation between IT investment andgrowth, but not recognize that these are si-multaneously determined by an unobserv-able factorfor example, a firms growthoptionsand erroneously attribute thisgrowth to IT investment. Virtually allstudies employ advanced techniques toaddress this concern; what is uncer-tain is how successful these techniquesare in distinguishing between these twoeffects.

These same problems arise withmacroeconomic data: Is an increase ininvestment a cause of an increase in GDP,or vice versa? Aggregate labor produc-tivity tends to increase when the labormarket is tight since firms try to squeezemore output from their existing workers.The very low rates of unemployment inthe later 1990s would naturally lead toan increase in measured productivity.Researchers have attempted to adjustfor these business cycle effects, but itis still debatable how well they havesucceeded.

In spite of legitimate concerns aboutmeasurement, data, and statistical mod-els, the evidence of positive and signifi-cant productivity gains related to IT in-

vestment is still strong. The issues raisedpoint to difficulties in arriving at preciseestimates of returns on investment, andof sorting out the relative contributions of

labor productivity versus MFP. The factthat a large number of studies using differ-ent datasets and different models come to

similar conclusions makes for strong ev-idence of payoffs to IT investments rel-ative to capital deepening. On the otherhand, better measurement of output qual-ity could lead to better estimates of thereturn to IT investment. Accordingly, onepromising avenue for further research isthe development of better measures of in-

vestment in capital and labor, and of meth-ods of accounting for previously uncountedinvestments.

7. OPPORTUNITIES FOR FUTURE

RESEARCH

7.1. Sources of Productivity Growth

The research evaluated in this reviewhighlights a set of fundamental issues andquestions that are critical to developingan understanding of the mechanisms bywhich IT pays off in higher productivity.In particular, IT can impact labor pro-ductivity via capital deepening, and inMFP growth through improvements inproduction methods. The first finding isconsistent with a traditional neoclassi-cal economics view which also impliesthat firms receive diminishing returnsfrom continuing investment in IT asopportunities for investment decline withincreasing levels of IT stock. In this view,technical progress originates exogenouslyin the computer industry, and ongoinginvestment by firms in the outputs of theIT industry drives productivity growth.On the other hand, a payoff in laborproductivity via capital deepening plusMFP growth may be indicative of constantor increasing returns. Such a finding re-quires the identification of a mechanismby which capital might not experiencedecreasing returns.10 In particular, oneexplanation for this structure of returns isthe possibility of spillovers in which firmsbenefit not just from a private investmentin an asset but also from a growth in the

10 This notion is central to new growth theory, whichfocuses on endogenous growth and constant or in-creasing returns; c.f. Romer [1986].



21/28


asset stock of all firms. The McKinseyGlobal Institute [2001] provided a tellingexample of this phenomenon in the retail

sector, where Wal-Marts productivitygains from innovations in IT and associ-ated management practices have spurredcompetitors to make similar investmentsleading to productivity improvements inthe industry broadly.

7.2. Spillover Effects

An understanding of whether thesespillovers exist and how they occur is cen-tral to developing a comprehensive frame-work for understanding the returns to ITinvestment and for developing guidelines

for the successful deployment of thesetechnologies. A critical feature of this de-bate is whether IT is like traditional formsof capital, such as tangible assets and hu-man capital, or whether it is more likeknowledge capital, which is significantlydifferent. In the case of traditional capi-tal investment, returns accrue primarilyto the firm making the investment andreceive diminishing returns from continu-ing investment. On the other hand, someeconomists hold that knowledge capitalcan be owned and used by many par-ties simultaneously, leading to potential

spillovers, and that the returns may bedifficult for a single firm to capture in thepresence of spillovers to other firms. Thesespillovers can lead to endogenous techni-cal progress.

Clearly, IT capital has aspects of bothforms of capital. As a production technol-ogy, it is similar to traditional forms ofcapital. In its informational and trans-formational roles, it is similar to knowl-edge capital. Best practice information re-garding the management of technology,complementary organizational practices,and techniques for better information use

does lend itself to use by many firms.Such knowledge is often diffused by enti-ties such as technology user groups, aca-demic institutions, management consult-ing firms, and, especially, labor mobility.It is often the case that competing firmsrapidly copy IT investments made by in-novative firms.

7.3. IT and Firm Financial Performance

Firm-level studies have failed to iden-

tify a relationship between IT investmentand firm profitability. Hitt and Brynjolf-sson [1996] showed that while IT invest-ment affects productivity and contributesto consumer welfare (through lower pricesor better service, for example), it doesnot necessarily improve profitability. Theyproposed that the productivity benefits as-sociated with IT use may be passed on toconsumers through lower prices and notlead to greater profitability. On the otherhand, it is possible and even likely thatIT investments do actually affect prof-itability, but the modeling techniques and

datasets used in these studies are unableto measure the impacts. It is important,therefore, to develop better datasets andalso to develop models that are able tocontrol for more of the additional factorsthat affect profitability, with the possibil-ity of revealing a relationship between ITinvestment and financial performance. Itis also important to measure the impactsof IT on intermediate outputs such as in-

ventory levels, planning cycles, asset uti-lization, and other measures of operationsperformance, which are known to have adirect link with profitability.

7.4. Excess Returns at the Firm Level

Another issue that deserves further at-tention is the high returns to IT use thatsome firms appear to have accrued. Someevidence suggests that firms in durablegoods industries have achieved substan-tially larger returns than firms in non-durable goods industries, while other ev-idence suggests that the returns to ITinvestment are broader and accrue to awider range of firms if lagged payoffs aretaken into account. As pointed out here,IT must have a high gross return to allow

for rapid depreciation and obsolescence,and it is also the case that investmentsin complementary assets such as software,training, and organizational transforma-tions have been undercounted. It is impor-tant to develop an understanding of themechanisms by which these returns ac-crue to firms. Studies should also attempt



22/28

22 Dedrick et al.

to include adjustments for the risk in-volved with IT investments.

7.5. Timing of Payoff from IT Investments

A better understanding of the timing ofthe payoff from investments in IT capitalis also needed. Clearly, a firms many in-dividual investments in specific systemswill have different periods over whichthe payoffs will be realized. Some sys-tems will realize immediate payoffs, whileothers will realize payoffs after a lag.The duration over which the payoffs willbe realized will also vary. Some will haveshort-term impacts and others will havelonger-term impacts. This understanding

will go a long way toward resolving the de-bate over whether the impact of these in-

vestments is contemporaneous or occursin the future.

7.6. Industry Differences

It is not well understood why firms in dif-ferent industries accrue different payoffs.

At aggregate levels, an explanation that isgenerally consistent with the traditional,neoclassical approach has been advanced.That is, in computer-using industries, themechanism through which IT provides

a payoff is increasing labor productivityvia capital deepening; in the computer-producing sector, and the durable goodssector, more generally, the mechanism isprimarily technical progress, measuredas growth in MFP. While the evidencefor this is compelling overall, there aresome important unanswered questions.For one, it is unclear why some IT-intensive industriessuch as consumerbanking and insurancehave not seengains until recently in labor productivityin spite of long-term, large investmentsin IT. While the difficulty in measuring

outputs in these industries is one of thelikely explanations, more research is re-quired to fully understand this result. Itis also somewhat unclear why durablegoods manufacturing achieves significantMFP gains while nondurable manufactur-ing does not. Finally, it is not understoodwhy technical progress has accelerated in

recent years in the computer-producingsectors of the economy, although somehave argued that intensified competition

in the semiconductor industry in the late1990s led to faster product cycles.

8. CONCLUSION

As the foregoing review shows, IT andeconomic performance has become a keyarea of research in the information sys-tems field with contributions being madeby information systems researchers, man-agement scientists, and economists. Thearea has grown from less than a dozenstudies in the 1980s to more than 50 stud-ies in the 1990s.The researchon IT payoffs

is complex, employing a number of analyt-ical tools to study a variety of firms, in-dustries, and countries. Different studiessometimes come to conflicting conclusions,and researchers have different interpreta-tions of what the data mean. What is par-ticularly significant about these studies isthat whereas the early studies were incon-clusive, studies that are more recent haveshown considerable convergence. Beyondall the complexity of the field, three fun-damental conclusions clearly emerge fromthis review.

First, the productivity paradox as origi-

nally stated by Robert Solow [1987], whichwas always more of a straw man than aneconomic analysis, has been put to rest. Anumber of major studies have documentedthe significant impact of IT investment onthe productivity of firms, industries, andcountries, showing that computers do, infact, show up in the productivity statistics.

Second, while the so-called New Econ-omy and the high IT returns capturedthe medias attention in the late 1990s,IT investments actually have been in-creasing productivity for more than threedecades.

Third, and most significantly for firms,although returns to IT investments arepositive on average, there is a wide rangeof performance among different compa-nies, with some doing much better thanothers. Some of these differences can beexplained by idiosyncratic firm differencesthat result in different opportunities to



23/28


employ IT productively. In addition, thereis strong evidence that investments in or-ganizational capital through management

practices such as decentralized decision-making, job training, and business processrestructuring have a major impact on re-turns to IT investments. The value of ITinvestments needs to be seen in relation-ship to investments in such organizationalcapital, as the two are complementary. ITis not simply a tool for automating exist-ing processes, but is more importantly anenabler of organizational changes that canlead to productivity gains.

While Solows paradox has been re-solved, this review and evaluation of theliterature indicates that the issue of re-

turns to IT investments is far more com-plex than the original formulation, andhence more research is needed. The eval-uation highlighted a range of underlyingissues that need to be addressed in on-going and future research. Chief amongthese are measurement issues regardingthe inputs and outputs of firms and indus-tries. Improved measurement of inputs,especially of software and intangible as-sets, such as R&D and human capital, isa first step. A much more important butdifficult step is improved measurements ofoutputs. This is particularly critical for the

services sector, where measurement prob-lems are the most severe. IT returns areleast understood and possibly underesti-mated in this sector that accounts for two-thirds of the U.S. economy.

This review has also identified priorityareas for futureresearch. Three are partic-ularly important for professional practice.The first is further analysis of the mecha-nismsby which some firms receivehigh re-turns from IT use, and in particular, there-turns from investments in complementaryassets. The second priority is explainingwhy some IT-intensive industries have not

seen gains in labor productivity in spite oflarge investments in IT.

dedrick a critical review of empirical evidence

Documents