Context. The cosmological concordance model (ΛCDM) matches the cosmological observations exceedingly well. This model has become the standard cosmological model with the evidence for an accelerated expansion provided by the type Ia supernovae (SNIa) Hubble diagram. However, the robustness of this evidence has been addressed recently with somewhat diverging conclusions. Aims. The purpose of this paper is to assess the robustness of the conclusion that the Universe is indeed accelerating if we rely only on low-redshift (z 2) observations, that is to say with SNIa, baryonic acoustic oscillations, measurements of the Hubble parameter at different redshifts, and measurements of the growth of matter perturbations. Methods. We used the standard statistical procedure of minimizing the χ 2 function for the different probes to quantify the goodness of fit of a model for both ΛCDM and a simple nonaccelerated low-redshift power law model. In this analysis, we do not assume that supernovae intrinsic luminosity is independent of the redshift, which has been a fundamental assumption in most previous studies that cannot be tested. Results. We have found that, when SNIa intrinsic luminosity is not assumed to be redshift independent, a nonaccelerated low-redshift power law model is able to fit the low-redshift background data as well as, or even slightly better, than ΛCDM. When measurements of the growth of structures are added, a nonaccelerated low-redshift power law model still provides an excellent fit to the data for all the luminosity evolution models considered. Conclusions. Without the standard assumption that supernovae intrinsic luminosity is independent of the redshift, low-redshift probes are consistent with a nonaccelerated universe.