The aim of the present paper is to quantify the dependence of the estimates of luminosities and stellar mass content of early-type galaxies on the different models and model parameters which can be used to analyse the observational data. The paper is organized in two parts. The first one analyses the dependence of the ratios and of the k-corrections in different bands on model parameters (initial mass function, metallicity, star formation history, age), assuming some among the most popular spectrophotometric codes usually adopted to study the evolutionary status of galaxies: Bruzual & Charlot (BC03), Charlot & Bruzual (CB08), Maraston (Ma05), Fioc & Rocca-Volmerange (PEGASE), Silva et al. (GRASIL). The second part of our work is dedicated to quantify the reliability and systematics affecting the mass and luminosity estimates obtained by means of the best-fitting technique applied to the photometric spectral energy distributions (SEDs) of early-type galaxies at 1 < z < 2. To this end, we apply the best-fitting technique to some mock catalogues built on the basis of a wide set of models of early-type galaxies. We then compare the luminosity and the stellar mass estimated from the SED fitting with the true known input values. The goodness of the mass estimate is found to be dependent on the mass estimator adopted to derive it, but masses cannot anyhow be retrieved better than within a factor of 2–3, depending on the quality of the available photometric data and/or on the distance of the galaxies since more distant galaxies are fainter on average and thus affected by larger photometric errors. Finally, we present a new empirical mass estimator based on the K-band apparent magnitude and on the observed(V−K) colour. We show that the reliability of the stellar mass content derived with this new estimator for early-type galaxies and its stability are even higher than those achievable with the best classic estimators, with the not negligible advantage that it does not need any multiwavelength data fitting.