Non-reporting of R&D and Software

Research finds that R&D and software expenditures are highly skewed: large firms spend much more relative to their sizes than do small firms (e.g., see James and Xiupeng Wang. 2025. “The Intangible Divide: Why do so few firms invest in innovation?†Center for Economic Studies Working Paper, CES25-15). To what extent is this the result of firms not reporting these survey items and to what extent are many firms reporting zero values for these items? Moreover, how accurate is firm reporting relative to external sources? One concern is that firms might over- or under-report their actual investments. Perhaps small firms under-report because they have a harder time tracking intangible investments. On the other hand, perhaps large firms re-categorize expenses to exaggerate R&D spending in order to earn bigger R&D tax credits; this is not a concern with software. This note first establishes the extent of non-reporting and zero reporting. Then, using a linear probability model, it looks at the correlates of non- and zero reporting. Finally, it checks the external validity of survey estimates by comparing R&D spending on personnel reported in the BRDIS survey to labor compensation costs of scientists and engineers estimated with data from the Current Population Survey.