The productivity paradox of information technology

TL;DR

Brynjolfsson reviews the “IT productivity paradox”—large increases in computing power and IT investment without clear aggregate productivity gains—and argues the shortfall is at least partly due to measurement and methodological deficiencies rather than solely mismanagement. He organizes explanations into measurement errors, lags, redistribution, and mismanagement, and notes that some firm-level evidence finds high returns to IT.

Key findings

• U.S. delivered computing power increased by more than two orders of magnitude since 1970 while service-sector productivity stagnated; Solow’s summary is quoted: “we see computers everywhere except in the productivity statistics.”
• Roach’s cited statistics indicate output per production worker grew 16.9% between the mid-1970s and 1986, while output per information worker decreased 6.6%.
• Four explanations for the paradox are assessed: measurement errors, lags, redistribution, and mismanagement. The paper argues measurement difficulties may account for “the lion’s share” of the gap; Baily and Gordon’s revisions add 2.3% per year to productivity between 1973 and 1987 in finance, insurance, and real estate.
• Brynjolfsson and Hitt found service firms’ return on investment averaged over 60% per year, and the study with the largest and most detailed data set found no productivity shortfall. Lags are also noted: executives expected up to five years for IT payoffs, and an econometric study found two-to-three-year lags.

Methods

• Review and synthesis, not new primary data collection.
• Based on a computerized literature search of 30 leading journals in information systems and economics, plus discussions with leading researchers.
• Assesses principal empirical studies of IT and productivity across manufacturing and services and evaluates four hypotheses for the paradox.

Limitations

• The paper states there is still no definitive answer to whether IT’s productivity impact has actually been unusually low; research tools remain “blunt,” and the measurement problem is becoming more severe. It also cautions that “a shortfall of evidence is not necessarily evidence of a shortfall.”
• The paper states that studies usually spell out data and method limitations, but media may report only surprising conclusions, and managers tend to overemphasize productivity studies.
• Evident from the text: as a literature review, its conclusions depend on cited studies and their measurement and methodological limitations; it provides no new empirical estimates.