perishable, particularly for earlier award years. For firms that had more than two awards, 20 percent were selected, but no less than two.
Top Performers. The problem of skew was dealt with by ensuring that all Phase IIs known to meet a specific commercialization threshold (total of $10 million in the sum of sales plus additional investment) were surveyed (derived from the DoD commercialization database). Since 56 percent of all awards were in the random and firm samples described above, only 95 Phase IIs were added in this fashion.
Coding. The project database tracks the survey sample, which corresponds with each response. For example, it is possible for a randomly sampled project from a firm that had only two awards to be a top performer. Thus, the response could be analyzed as a random sample for the program, a random sample for the awarding agency, a top performer, and as part of the sample of single or double winners. In addition, the database allows examination of the responses for the array of potential explanatory or demographic variables.
Total Number of Surveys. The approach described above generated a sample of 6,410 projects and 4,085 firm surveys—an average of 1.6 award surveys per firm. Each firm receiving at least one project survey also received a firm survey. Although this approach sampled more than 57 percent of the awards, multiple award winners, on average, were asked to respond to surveys covering about 20 percent of their projects.
The questionnaire drew extensively from the one used in the 1999 National Research Council assessment of SBIR at the Department of Defense, The Small Business Innovation Research Program: An Assessment of the Department of Defense Fast Track Initiative.1 That questionnaire in turn built upon the questionnaire for the 1991 GAO SBIR study. Twenty-four of the twenty-nine questions on the earlier NRC study were incorporated. The researchers added twenty-four new questions to attempt to understand both commercial and noncommercial aspects, including knowledge base impacts, of SBIR, and to gain insight into impacts of program management. Potential questions were discussed with each agency, and their input was considered. In determining questions that should be in the survey, the research team also considered which issues and questions were best examined