Technological Progress & Productivity Measurement

December 4, 2009
Ajay K. Agrawal and Benjamin Jones, Organizers

Pierre Azoulay, MIT and NBER; Joshua S. Graff Zivin, UC, San Diego and NBER; and Gustavo Manso, MIT
Incentives and Creativity: Evidence from the Academic Life Sciences

Despite its presumed role as an engine of economic growth, we know surprisingly little about the drivers of scientific creativity. Azoulay, Graff Zivin, and Manso exploit key differences across funding streams within the academic life sciences to estimate the impact of incentives on the rate and direction of scientific exploration. Specifically, they study the careers of investigators of the Howard Hughes Medical Institute (HHMI), which tolerates early failure, rewards long-term success, and gives its appointees great freedom to experiment; and grantees from the National Institute of Health, who are subject to short review cycles, pre-defined deliverables, and renewal policies unforgiving of failure. Using a combination of propensity-score weighting and difference-in-differences estimation strategies, the authors find that HHMI investigators produce high-impact papers at a much higher rate than two control groups of similarly-accomplished NIH-funded scientists. Moreover, the direction of their research changes in ways that suggest the program induces them to explore novel lines of inquiry.


Heidi Williams, Harvard University
Intellectual Property Rights and Innovation: Evidence from the Human Genome

Williams shows how intellectual property (IP) on a given technology affects subsequent downstream innovation. She analyzes the sequencing of the human genome by the public Human Genome Project and the private firm Celera, and estimates the impact of Celera's gene-level IP on subsequent scientific research and product development outcomes. Celera's IP applied to genes sequenced first by Celera, and was removed when the public effort re-sequenced those genes. Williams tests whether as of 2009 genes that ever had Celera's IP differ in subsequent downstream innovation from genes sequenced by the public effort over the same time period -- a comparison group that appears balanced on ex ante gene-level observables. A complementary panel analysis traces the effects of removal of Celera's IP on within-gene flow measures of innovation. Both analyses suggest that Celera's IP led to reductions in subsequent scientific research and product development outcomes by about 30 percent. Celera's short-term IP thus appears to have had persistent negative effects on subsequent innovation relative to a counterfactual of Celera genes having always been in the public domain.


Alex Oettl, University of Toronto
Productivity and Helpfulness: Implications of a New Taxonomy for Star Scientists

It is surprising that the prevailing performance taxonomy for scientists (Star versus Non-Star) focuses only on individual output and ignores social behavior, because innovation is often characterized as a communal process. To address this deficiency, Oettl expands the traditional taxonomy that focuses solely on productivity and adds a second, social dimension to the taxonomy of scientists: helpfulness to others. Using academic paper citations to capture scientist productivity and the receipt of academic paper acknowledgements to measure helpfulness, he classifies a group of 415 immunologists into four distinct categories of human capital quality: All-Stars who have both high productivity and helpfulness; Lone Wolves who have high productivity but average helpfulness; Mavens who have average productivity but high helpfulness; and Non-Stars who have both average productivity and helpfulness. Looking at the change in quality-adjusted publishing output of an immunologist's coauthors after the immunologist's death, he finds that the productivity of coauthors of All-Stars decreases by 35 percent on average, coauthors of Mavens by 30 percent on average, and the coauthors of Lone Wolves by 19 percent, all relative to the decrease in productivity of coauthors of Non-Stars. These findings suggest that our current conceptualization of star scientists, which focuses solely on individual productivity, is both incomplete and potentially misleading because Lone Wolves may be systematically overvalued and Mavens undervalued.

Timothy Simcoe, Boston University and NBER
What's in a (Missing) Name? Status Signals in Open Standards Development

How much are we influenced by authors' identity when evaluating their work? If names matter, are we responding to status, reputation, or a signal of underlying quality? Simcoe addresses these questions in the context of open standards development. He exploits a natural experiment, whereby author names were occasionally replaced by et al in a series of email messages used to announce new submissions to the Internet Engineering Task Force (IETF). He compares the effect of obscuring high- versus low-status author names. His results suggest that name-based signals can explain up to two-thirds of the difference in publication outcomes across status cohorts. However, this signaling effect disappears for a set of pre-screened proposals that receive more attention than a typical submission. He also shows that high-status authors receive more forward citations from other IETF participants than do low-status authors, while cites from outside the focal community (from U.S. patents and academic journal articles) exhibit no difference. These findings suggest that status signals are important for drawing attention to new ideas, which is important for developing these ideas and bringing them forward to publication.


Yuriy Gorodnichenko, UC, Berkeley and NBER; and Monika Schnitzer, University of Munich
Financial Constraints and Innovation: Why Poor Countries Don't Catch up

Gorodnichenko and Schnitzer examine micro-level channels of how financial development can affect macroeconomic outcomes such as the level of income and export intensity. Specifically, the paper investigates theoretically and empirically how financial constraints affect a firm's innovation and export activities. Theoretical predictions are tested using unique firm survey data which provides direct measures for innovations and firm-specific financial constraints and information on shocks to firms' internal funds that can serve as firm-level instruments for financial constraints. There is unambiguous evidence that financial constraints strongly adversely affect the ability of domestically owned firms to innovate and to export and hence to catch up to the technological frontiers. Furthermore, the negative effect of financial constraints on productivity is amplified as these constraints force export and innovation activities to become substitutes even when these activities are natural complements. Findings reported in the paper can help explain why poor countries don't catch up, despite increasing globalization.


Lynn Wu, MIT Sloan; and Erik Brynjolfsson, MIT and NBER
The Future of Prediction: How Google Searches Foreshadow Housing Prices and Sales

To make effective decisions, consumers, executives, and policymakers must make predictions. However, most data sources, whether from the government or businesses, are typically available only after a substantial lag, at a high level of aggregation, and for variables that were specified and collected in advance. This hampers the effectiveness of real-time predictions. A critical advance in IT research has been the development of powerful search engines and the underlying Internet infrastructure. Wu and Brynjolfsson demonstrate how data from such search engines provides a highly accurate but simple way to predict future business activities. Applying their methodology to predict housing market trends, they find that their housing search index is strongly predictive of the future housing market sales and prices. Specifically, each percentage point increase in their housing search index is correlated with additional sales of 67,700 houses in the next quarter. The use of search data produces out-of-sample predictions with a mean absolute error of just 0.102, a substantial improvement over the 0.441 mean absolute error of the baseline model which uses conventional data but does not include any search data. They also demonstrate how these data can be used in other markets, such as laptop sales. In the near future, this approach can transform prediction in numerous markets, and thus business and consumer decisionmaking.