Generic Tests for a Nonlinear Model of Congressional Campaign Dynamics
Mebane, Walter R.
I develop a statistical model based on a generic third-order Taylor series
approximation for differential equation systems that exhibit Hopf bifurcation
in order to use district-level cross-sectional data to test a nonlinear
dynamic formal model of campaign contributions, district service and voting
during and after a U.S. House election. The statistical model represents the
key nonlinearities of the formal model's Cournot-Nash equilibrium in a highly
robust fashion. For data from the years 1984--85 and 1986--87, non-nested
hypothesis tests (implemented using a calibrated, parametric bootstrap method)
show that under assumptions of multivariate normality, the nonlinear model is
vastly superior to the generic linear alternative defined by the sample mean
vector and covariance matrix.
non-nested hypothesis tests
Document ID Number