joseph emmens

Econ Best Job Market Paper Award 2024 UniCredit (Runner-up)

[latest version]

Abstract

In this paper, I study the impact of an expanding scientific and technological frontier on team innovations. To do so, I present a novel framework that integrates inventor teams and their patent texts. I model collaboration directly through a Bayesian model of Natural Language Processing. Applied to patent text data, this model builds a map of inventors, teams, and research fields, referred to as the knowledge space. Applied to over 2.2 million U.S. patents from the USPTO PatentsView database, this framework allows me to tackle unanswered questions on how teams create new knowledge. Specifically, I investigate the effect of prior work on a team’s ability to produce a breakthrough–an innovation that sparks a new and successful research field. Leveraging high-dimensional patent text data, I back out two new measures: breakthrough patents and a team’s knowledge field, the set of research fields accessible to the team. I combine this with data on premature inventor deaths as a quasi-natural experiment. This identifies how team innovations change as they pivot to more or less advanced research fields. The framework unifies key elements of collaboration. Teams build on existing knowledge, and prior work both supports and obstructs innovation. I show that teams generate more breakthroughs when building on enough prior work to incorporate valuable knowledge, but not so much as to stifle novelty.

Keywords: Teams, Innovation, Patents, Topic Modelling

With Mihai Codreanu, Alex Imas, Juan Mateos-Garcia, Evalyne Muiruri, Arthur Turrell, Julian Jacobs, Atoosa Kasirzadeh, Ana Trišović, Yiyuan Chen, Tanya Rodchenko, Catherine Pollard, Scott Strand, Daniel Rock, Zanna Iscenko, Fabien Curto Millet, Neil Thompson and James Manyika

[Paper Link]

Abstract
Scientific progress is a key driver of economic growth and prosperity. There is great excitement - but also concerns - about the impacts of AI on science, but so far little data. We provide early insights on this from three data sources: a sample of 15 million Gemini interactions, an inventory of over 2,600 specialized AI models across disciplines, and a survey of over 600 scientists. We map these data to a new taxonomy of scientific tasks to study how scientists are using AI. Four main findings emerge. First, we find broad adoption and coverage: scientists use AI more than most other occupations. Specialized AI models have broad disciplinary coverage and are highly cited. Nearly half of the scientists surveyed report using some form of AI every day. Second, we document evidence that LLMs (proxied through Gemini usage) and specialized models act as complements–LLMs are used for general analysis, coding, and manuscript preparation, while specialized models provide domain-specific predictions, data generation and classification. Third, scientists report large productivity gains from using AI: a saving of nearly 7 hours per week, time which is primarily re-invested in more research. Finally, we show that AI is already changing the scientific process. As some stages of scientific research become easier, bottlenecks shift downstream. Scientists report an increased backlog of untested hypotheses and substantial demand for output verification. Our findings suggest that AI holds significant potential to increase scientific productivity. However, as with other sectors, its ultimate impact will be governed by complex task interdependencies and investment into the elimination of emerging bottlenecks.

Keywords: Artificial intelligence, Economics of Science, Productivity, Economic Growth

With Dennis C. Hutschenreiter, Felix Noth, Stefano Manfredonia and Tommaso Santini

[RFBerlin Discussion Paper Series]

Abstract

We study whether common ownership affects the direction of technological change. We develop a task-based model with multiple local labor markets in which commonly owned firms internalize wage externalities from portfolio rivals when hiring from the same labor pool, increasing incentives to automate. We establish causality by exploiting institutional investor mergers in a dynamic DiD design, using U.S. data on institutional ownership, establishment level employment, and text-classified automation patents. Increases in common ownership among local labor-market rivals raise firms’ automation propensity by 22.7 percentage points and reduce employment growth. The effect disappears when firms do not compete within labor markets.

Keywords: Common Ownership, Automation, Local Labor Markets, Market Power.

With Jingnan Liu, Ana Trišović and Neil Thompson

[Paper and Data Link]

Abstract

Scientific discovery is a fundamental engine of human progress and economic growth, yet we lack detailed evidence on the nature of scientific work. This paper introduces the Science Task Taxonomy, a new framework that maps scientific work across 232 subfields, including both commercial and non-commercial sectors. In particular, we introduce three levels of task categories, from Level-1 tasks shared across the sciences to Level-3 tasks specific to individual subfields and sectors, with Level-2 providing an intermediate level of granularity. The taxonomy runs from 12 Level-1 task areas, such as "Analyze and model quantitative research data'', through 114 Level-2 and 2,433 Level-3 areas, down to a pool of 208,202 representative tasks, such as "Develop cell-based assays to measure pharmacodynamic biomarkers''. Using this taxonomy, we provide a systematic account of how the tasks performed by scientists differ from the tasks in the wider economy. We show that scientific work is more cognitive, requires fewer interpersonal interactions, is less manual across each of the physical, psychomotor and sensory dimensions. Science is also less codifiable, and the median scientific task takes over two hours longer to complete. We also find significant variation between scientific occupations along every dimension we measure. This paper moves the measurement of the idea production function to the task level, with important implications for our understanding of the effect of AI on science.

Keywords: Science, Research and Development, Task-based Modelling