The Council of Science Editors is not a body I've come across before, but I chanced upon their website while looking for something else, and given this site's recent interest in the subject, I wondered what their position is on data sharing. It's here:
Thorough peer review may require access to data and analyses that are not provided in a submitted manuscript, and sometimes such access is needed after publication as well. Editors should establish policies on access that address the following issues.
Organizations that sponsor research should encourage the publication of the results and should provide access to data if requested by journals for the purpose of peer review. Sponsoring organizations may limit access to data by others both during the research and after it is concluded, but should have no right to control the dissemination or interpretation of the results of the research and should provide access to any data needed for peer review.
Submission of an original article to a journal should carry with it the implied consent to provide access to data if needed for editorial evaluation and peer review. Journals should also have the right to review data on which manuscripts are based after publication, should questions arise regarding the validity of the work or of errors in it. This right of journals of access to data should be expressly stated by editors as part of their published editorial policies and in their guidelines for authors.
Editors should request access to data for the sole purpose of evaluating a manuscript for publication or in the case of a challenge to the validity of a work after publication. Editors, reviewers, and journal staff have a responsibility to keep the data confidential and not to use it for their own purposes in any way, or otherwise directly benefit from their access to the data that results from their role in the peer review process.
This is a bit of a disappointment as it seems to adopt the "weak" approach to data, assuming that authors will comply with requests for access after publication. Given that we know this doesn't happen in practice, it looks to me as if the society needs to revisit its policy and adopt the strong approach of the econometrics journals.
What was also interesting was that they seem to have examined the issue in more detail at their 2008 Annual Meeting. The schedule for the session reads as follows:
Most journals require that their authors share the data to support the conclusions presented in published papers. However, the devil is in the details when it comes to enforcing this requirement. Journal editors are being asked to arbitrate disagreements between authors and researchers with regard to just how much data must be shared. And the problem becomes even more complicated when a political agenda is behind the request for data. Come to this session to learn more about the issues surrounding data sharing.
Political agenda? Whoever could they mean? Surely they are not suggesting that someone's motives are relevant to whether they get the data or not? This aside, the statement here gets to the nub of the problem. If you are relying on authors complying with requests after the event, you have an enforcement issue. You just have to hope that your author is going to toe the line, and if they don't then you have a problem. Why do the journals do this to themselves? If they demand the data up front, the whole issue goes away. It's not rocket science.