Research needs to be policy-oriented. But at times, researchers need to take the liberty to undertake 'policy irrelevant research' (A. Zaidi) - i.e. to use a research space within which policies themselves and their underlying assumptions can be critically questioned and debated. The present paper argues for such intellectual space taking the example of decentralization and power devolution. At the level of policy-orientation, researchers for example can study the performance of related ventures, in order to recommend potential improvements to the policy-makers. At a more basic level though, researchers need to ask for the pre-conditions that make decentralization and power devolution such a powerful discourse today. Among others, this leads to the question of state-subject relations.