The most realistic model for current-to-voltage measurements of electrical impedance tomography is the complete electrode model which takes into account electrode shapes and contact impedances at the electrode/object interfaces. When contact impedances are small, numerical instability can be avoided by replacing the complete model with the shunt model in which perfect contacts, that is zero contact impedances, are assumed. In the present work we show that using the shunt model causes only a (almost) linear error with respect to the contact impedances in modelling absolute current-to-voltage measurements. Moreover, we note that the electric potentials predicted by the two models exhibit genuinely different Sobolev regularity properties. This, in particular, causes different convergence rates for finite element approximation of the potentials. The theoretical results are backed up by two dimensional numerical experiments. arXiv:1312.4202v2 [math.AP] 29 Jan 2015 2 *. Interestingly, the subspace projection property of the SM yields a better -of order O(z s ) for any s < 1 -rate for the electrode voltage. In other words, if the CEM is replaced by the SM, the error in the practical electrode measurement data exhibits almost linear * In [6] the exponent s = 1 2 is shown to be optimal for a generic two dimensional problem for Laplacian with fixed boundary data (1.2).