NSF portfolio analysis
Funding vs. Research:
Chicken and Egg problem?
Questions
- Does funding lead/lag science?
- Is funding leading (published) science?
- Is science leading funding?
Has consequences for
- Career trajectories of junior faculty
- Amount of "preliminary" work needed before getting funded
- Understanding science that was never funded
Methodology
- Compare NSF funded grant abstracts (200k grant abstracts between 1990-2010) and Abstract of papers in ISI Web of Knowledge (250 million documents, 133k journals, 256 fields)
- Perform topic analysis on the combined corpora to identify scientific subareas
- Topic dynamics in ISI vs. NSF over time
- For each topic find the optimal alignment
- Correlate topics and lags with funding
Results
- Presentation [slides] at CISE/SBE AC Subcommittee on Research Portfolio Analysis. Sept 2010.
- Topic loadings for 100 topics across all disciplines: We took all funded NSF grant abstracts and research paper abstracts (indexed by ISI). The plots show topic loadings for ISI and NSF and the amount of NSF funding for that topic.
- Leads-lags for 100 Computer science topics: We took only Computer Science NSF grant abstracts and research paper abstracts (indexed by ISI). We then identified topics and aligned ISI and NSF topic time series. Plots show the lead/lag of ISI (scientific output) with regard to NSF (proposed research). Ideally ISI would lag (NSF would lead) on all topics.