Research Portfolio Analysis and Topic Prominence

TL;DR

This paper introduces a new model and indicator for analyzing research portfolios by measuring topic prominence, which helps stakeholders make informed funding and hiring decisions based on visibility and demand.

Contribution

It presents a comprehensive model of science with 91,726 topics, introduces a novel prominence indicator, and links funding data to these topics for better portfolio analysis.

Findings

Topic prominence explains over one-third of funding variance.

Highly prominent topics receive more funding per researcher.

The model enables consistent and transparent research portfolio assessment.

Abstract

Stakeholders in the science system need to decide where to place their bets. Example questions include: Which areas of research should get more funding? Who should we hire? Which projects should we abandon and which new projects should we start? Making informed choices requires knowledge about these research options. Unfortunately, to date research portfolio options have not been defined in a consistent, transparent and relevant manner. Furthermore, we don't know how to define demand for these options. In this article, we address the issues of consistency, transparency, relevance and demand by using a model of science consisting of 91,726 topics (or research options) that contain over 58 million documents. We present a new indicator of topic prominence - a measure of visibility, momentum and, ultimately, demand. We assign over $203 billion of project-level funding data from STAR METRICS to individual topics in science, and show that the indicator of topic prominence, explains over one-third of the variance in current (or future) funding by topic. We also show that highly prominent topics receive far more funding per researcher than topics that are not prominent. Implications of these results for research planning and portfolio analysis by institutions and researchers are emphasized.