NIH data collection approaches presents some challenges for the purposes of this study. In principle, NIH collects data by award year—which means that each year of an award is separately identified within the database.

There are occasionally difficulties in connecting every award year of each award, as grant ID numbers can change, as do other possible connecting fields such as company name and award title.

As a result, we have for the purposes on this study developed approximations for award numbers and sizes. Numbers of Phase I awards are estimated by using the first year of SBIR support. Numbers of Phase II awards are estimated by using the second year of support where the award ID indicates that this is a Phase II award. Year two of Phase II support is estimated using year three of support where the award is Phase II, and year three of Phase II support is estimated using year four of support where the award is Phase II.

These estimates are undoubtedly not completely accurate. Where companies have received a second year of support during Phase I, the third year of support could be only the first year of Phase II support.

In a similar vein, we have generated estimates for average award size by adding the average for different years of Phase II support. It would facilitate future assessments if NIH would find ways to address these data difficulties as the agency refines its ongoing evaluation and assessment program.