The design effect reported by the DEFF option on the MODEL statement in PROC SURVEYREG may be incorrectly reported as zero for some parameters.

There is no circumvention for this problem. The correct design effect can be computed manually as the ratio S^2_SurveyReg / S^2_Reg, where S^2_SurveyReg is the variance of the parameter estimate from PROC SURVEYREG and S^2_Reg is the variance of the parameter estimate obtained from running the same model in PROC REG.