CRA Logo

About CRA
CRA for Students
CRA for Faculty
Computing Community Consortium (CCC)
Government Affairs
Computing Research Policy Blog
Data & Resources
CRA Bulletin
What's New

<< Back to September 2003 CRN Table of Contents

[Published originally in the September 2003 edition of Computing Research News, Vol. 15/No. 4, pp. 1, 6.]

IT Responsible For Most Productivity Gains

by John L. King

The "productivity paradox" of missing organizational payoffs from investments in information technology has finally been put to rest. Recent research has demonstrated a major surge in U.S. productivity between 1995 and 2000 due almost entirely to IT. While investment in IT is essential to this improvement, the key to achieving payoffs from IT investments lies in changing the nature of work processes to exploit what IT offers.

The productivity paradox began in early 1986 when economist Stephen Roach demonstrated that the huge increase in organizational expenditures on IT (computers, peripheral devices, software, and related services) between 1975 and 1985 was accompanied by virtually no gains in organizational productivity. Within weeks, Fortune magazine's cover story was about "The Puny Payoff" from computers, and the rest of the business trade press soon followed. Nobel Prize winning economist Robert Solow quipped, "We see computers everywhere except in the productivity statistics."

This news did not make sense to people in the IT fields. Computers could do many things far faster and far better than people could. Their application had dramatically improved performance in all kinds of tasks, from payroll processing to air traffic control. It did not seem possible that such task-level performance would fail to show up in the productivity statistics. Yet the analyses were grounded in the best data available at that time, and the story of the productivity paradox was established as fact.

Several objections to the story arose immediately. Most were aimed at the problems with the data used in Roach's analysis. As good as they were, Roach's data came from the National Income and Product Accounts (NIPA) data maintained by the U.S. Bureau of Economic Analysis. The NIPA system was installed in the 1930s, long before the era of modern IT application. It was not clear whether the NIPA data measured the right things. In addition, the NIPA data measured effects at the level of whole industries, not at the level of individual firms where the vaunted task-level performance of computers would appear. The NIPA data could not be used to account for differences in the quality of IT implementation efforts among the organizations measured. The successes and the failures might balance each other out, with zero productivity gain as a result.

A different objection was raised by economic historians. They pointed out that flat or even declining productivity was a common feature during transitions from old to new regimes for doing complicated things. Paul David of Stanford University compared the replacement of steam engines by electric power in U.S. factories in the late 19th and early 20th centuries to the replacement of older information management practices by IT during the late 20th century. Productivity remained flat for several decades as electricity replaced steam, but then accelerated rapidly in the late 1920s as the last of the steam factories closed and the full benefit of electric power took hold. Electricity proved to be far superior to steam, but the transition took a long time. The IT revolution would play out in a similar way, David argued, with productivity rising as the full effect of IT use was felt.

More than fifteen years have passed since Roach's work appeared. IT application never slowed down. In fact, the mid and late 1990s brought the dot-com era, an extraordinary boom in IT application that coincided with one of the longest periods of economic expansion in U.S. history. If productivity were to change, the late 1990s was the time to see it. In addition, and partly in response to the productivity paradox, economic researchers were developing new measures focusing on the firm level and incorporating an improved suite of productivity measures. The results of new research on the late 1990s are coming out, and they tell a remarkable story.

Productivity grew from 1.33 percent to 2.07 percent between the periods 1975-1993 and 1995-2000, according to Dale Jorgensen of Harvard University, Mun Ho of Resources for the Future, and Kevin Stiroh of the Federal Reserve Bank of New York. This is the largest gain in many years. The gain was due mainly to "capital deepening," which means providing more effective capital investment to leverage the efforts of workers. When IT-capital deepening was separated from other capital deepening, the results are even more impressive. IT capital deepening jumped from .37 percent to .87 percent, while other capital deepening actually dropped from 0.43 percent to 0.37 percent in the 1995-2000 period. In other words, IT capital input accelerated while other capital input decelerated.

Capital deepening is not the only factor that can affect productivity: the quality of the workforce also plays a major role. During these two time periods, labor quality actually declined due to the heated economy and the easy availability of jobs as the less productive workers entered the work force. Thus, the major productivity increase occurred in spite of a decline in labor quality during that period. When productivity gains are separated into IT-related productivity and other-related productivity, the conclusions become obvious. IT-related productivity rose from 0.21 percent to 0.45 percent, while other-related productivity over the same periods grew from only 0.05 percent to 0.17 percent. In short, there was a doubling of U.S. productivity in the period 1995-2000, which effectively means that the wealth of the country was building twice as quickly. Nearly all of that improvement was due to IT.

The dot-com crash and the sluggish economy following 2000 put a damper on the celebration. But this slowdown also provided a needed respite to step back from the "irrational exuberance" of that era to reflect on the changes underway. This is particularly important when articles with titles like "IT Doesn't Matter" appear in Harvard Business Review, and academic leaders cast nervous glances at noticeable declines in applications and enrollments in undergraduate computer science and graduate information systems programs. Paul David's hopeful story suggests that IT will continue to bring major improvements in productivity over the coming years, especially given the continued pace of improvement in the underlying technology itself. The challenge now lies in better understanding the ways in which IT affects organizational performance for the better.

The effort to demonstrate IT's role in productivity has been matched by research into the "hidden assets" that complement IT investments in the quest for productivity improvement. Erik Brynjolffson of MIT and Lorin Hitt of the University of Pennsylvania define hidden assets as those that are not counted by standard economic measurement systems. Many of these hidden assets have grown up around IT implementation, including specialized software and utilities, revised work practices, new control systems, and improved analytical capacity to aid in management decision making. The ratio of hidden assets is highest in the most productive firms, reaching as much as 10:1. The value created by IT assets and hidden assets working together goes beyond standard economic measurements, affecting things such as customer convenience and service quality. Previous research not only missed the productivity impacts of IT, but failed to recognize the complexity of the mechanisms involved in improving productivity.

The story of IT and productivity has been odd. The productivity payoffs of IT have been elusive for years, yet organizations kept spending on IT in spite of this news, and dramatically accelerated IT investments in the 1990s. When econometricians finally teased out the payoffs of IT, the dot-com boom collapsed and the economy turned down. It seems ironic that it was impossible to see the evidence of productivity when the economy was roaring, but it is easy now that things have slowed down. In fact, this is to be expected because understanding almost always lags the things to be understood. It takes time to understand change of this magnitude.

The revolution launched by IT is far too large and complex to play out in just a few decades. The revolution involves economic, political, social and cultural systems that often change slowly and in subtle and sometimes ambiguous ways. The advances in IT that gave the revolution its power were accomplished by expertise in narrow areas of specialization in computer engineering, computer science, and information systems. An important role for such specialization remains, but it is increasingly important to look across and not just within these specializations, and to join with economists, psychologists, operations management specialists, library and information specialists, and others to grasp the full magnitude of the changes underway. Without this broad effort, many of those who helped to launch the revolution will see it default to those with far less sense of how it all happened, and only a distant notion of what is at stake.

Dr. John L. King (jlking [at] is Dean and Professor in the School of Information at the University of Michigan. He was recently elected to a three-year term on the CRA Board of Directors.

Search WWW Search

Copyright © 2007 Computing Research Association. All Rights Reserved. Questions? E-mail: