The effect of channel parameter mismatch on the performance of iterative multiuser detection using the belief propagation algorithm is investigated. The mismatch of two parameters are considered: the channel coefficients and the channel noise variance. The mismatch of channel noise variance corresponds to a mismatch of SNR. The following results are observed using simulations: Overestimation of SNR has very little effect on the performance, whereas underestimation may cause some degradation, but only for very large values of mismatch. The effect of the mismatch of channel coefficients grows with the MSE and may be significant for practical values of channel estimation errors.