Economists from the USPTO (Chief Economist Stuart Graham and former visiting scholar Deepak Hegde) recently posted a new paper to SSRN: Do Inventors Value Secrecy in Patenting? Evidence from the American Inventor's Protection Act of 1999. Their answer to their title's question is essentially that inventors don't value secrecy as much as you might think.
Under the American Inventors Protection Act of 1999 (AIPA), U.S. patent applications are published 18 months after filing, but applicants who forgo foreign patent protection may choose to delay publication until the patent is granted. Graham and Hegde report that only 7.5% of applications filed from 2001 to 2005 chose pre-grant secrecy, though I think the more interesting figure is the percentage who chose secrecy out of those who could have chosen secrecy, which is about twice this (because about 50% of applicants chose to forgo foreign protection both before and after AIPA enactment). In either case, it is interesting that most inventors choose disclosure over secrecy—though I wonder how much these results depend on sticky default rules rather than an affirmative preference for disclosure.
Graham and Hegde also find that conditional on forgoing foreign protection, small inventors are not more likely to choose pre-grant secrecy over disclosure, and that small inventors are "more likely than large corporations to prefer disclosure over secrecy for their most important inventions" (where "important" patents have longer time lags to grant, more claims, more maintenance fees paid, and more citations). This raises a related question about default rules: are small inventors less likely to know that they have to file a nonpublication request at the time of filing to opt out of disclosure at 18 months (thus making the default of publication more sticky for small inventors)? My guess is that if the PTO asked each applicant who did not seek foreign protection to make an affirmative choice for either secrecy or disclosure, the percentage choosing secrecy would be higher, but this intuition might be wrong. In any case, given the benefits of early patent disclosure for making useful technical information available, a default of publication makes sense.
The authors argue that their "findings question the conventional wisdom—which seems to have shaped important policy—that the disclosure of patent applications harms U.S. invention by increasing the risk of imitation for small inventors." As I've argued before (see p. 582-83), even an 18-month delay limits the ability of researchers to use patents to learn about cutting-edge technologies. One approach might be to make the default publication date for all patents much earlier, while allowing patent applicants to affirmatively opt out of early publication (and to publish at 18 months for those seeking foreign protection or at the time of grant for those seeking only U.S. protection).
Update: Note that the authors' finding that small inventors are more likely to opt for disclosure