Multiangle Spectropolarimetric Imager (MSPI) sensitivity to static and time-varying polarization errors is examined. For a system without noise, static polarization errors are accurately represented by the calibration coefficients, and therefore do not impede correct mapping of measured to input Stokes vectors. But noise is invariably introduced during the detection process, and static polarization errors reduce the system's signal-to-noise ratio (SNR) by increasing noise sensitivity. Noise sensitivity is minimized by minimizing the condition number of the system data reduction matrix [Appl. Opt.41, 619 (2002)]. The sensitivity of condition numbers to static polarization errors is presented. The condition number of the nominal MSPI data reduction matrix is approximately 1.1 or less for all fields. The increase in the condition number above 1 results primarily from a quarter wave plate and mirror coating retardance magnitude errors. Sensitivity of the degree of linear polarization (DoLP) error with respect to time-varying diattenuation and retardance error was used to set a time-varying diattenuation magnitude tolerance of 0.005 and a time-varying retardance magnitude tolerance of ±0.2°. A Monte Carlo simulation of the calibration and measurements using anticipated static and time-varying errors indicates that MSPI has a probability of 0.9 of meeting its 0.005 DoLP uncertainty requirement.