Pop-Up Journal on the Returns to R&D

Why the range is so wide

Published estimates of the social return to R&D span roughly an order of magnitude. The US government invests around 0.6 percent of GDP in scientific and technological R&D on the premise that this spending generates substantial social benefit, yet estimates of that benefit range from about two dollars to twenty dollars or more per dollar invested.

This is placeholder text for a demonstration. The commissioned review essay would be written by the journal's authors and imported here. The structure below shows how such an essay would be organised and how it would cite the record set held alongside it.

What the estimates measure

Not all of these numbers answer the same question. Jones and Summers (2022) work from aggregate R&D expenditure shares in the national accounts. Fieldhouse and Mertens (2025) estimate from R&D capital stocks, which also capitalise software expenditures. Bloom, Schankerman and Van Reenen (2013) identify spillovers at the firm level by separating technological proximity from product market rivalry. Reading these as competing estimates of one parameter is a mistake, and much of the apparent disagreement in the literature comes from doing so.

Where the uncertainty actually lies

Three questions account for most of the remaining spread:

• The counterfactual growth path. What would have happened without the research?
• The treatment of health gains, which dominate welfare calculations when included.
• The boundary of what counts as R&D, particularly software and process innovation.

What would narrow it

The commissioned research programme described in this issue is aimed at the second and third of these. The expert elicitation, repeated annually, tracks whether the distribution of informed belief moves as that evidence arrives.