Complementary constraints: Separation of powers, rational voting, and constitutional design

Jide Okechuku Nzelibe, Matthew C. Stephenson

Research output: Contribution to journalArticlepeer-review

29 Scopus citations


This Article explores how the separation of powers affects voters' electoral strategies, and how this interaction influences the performance of different institutional arrangements. We show that when one political agent, such as the President, acts unilaterally, voters are likely to respond asymmetrically to policy successes and failures in order to offset the risk that the President may be biased or "captured" by special interest groups. When political agents act in concert - such as when the President seeks congressional authorization for a policy initiative - voters prefer a more refined strategy, with less acute as ymmetries between political rewards and punishments. Our analysis has positive and normative implications. First, it suggests that presidents do not always prefer to operate with as little congressional interference as possible. Second, it provides a rationalist account for "responsibility shifting" by elected officials - behavior that is usually thought to derive from voter confusion or irrationality. Third, it suggests that separation of powers does not necessarily induce "gridlock" or otherwise reduce the likelihood of policy change. Fourth, it suggests that although separation of powers enhances the efficacy of the electoral constraint on politicians, voter welfare is higher when separation of powers is "optional" rather than mandatory, as when the President may seek congressional authorization for policy initiatives but is not required to do so.

Original languageEnglish (US)
Pages (from-to)617-654
Number of pages38
JournalHarvard Law Review
Issue number3
StatePublished - Jan 2010

ASJC Scopus subject areas

  • Law


Dive into the research topics of 'Complementary constraints: Separation of powers, rational voting, and constitutional design'. Together they form a unique fingerprint.

Cite this