Local measurements of the Hubble expansion rate are affected by structures like galaxy clusters or voids. Here we present a first fully relativistic treatment of this effect, studying how clustering modifies the mean distance (modulus)-redshift relation and its dispersion. The best estimates of the local expansion rate stem from supernova observations at small redshifts (0.01<z<0.1). It is interesting to compare these local measurements with global fits to data from cosmic microwave background anisotropies. In particular, we argue that cosmic variance (i.e. the effects of the local structure) is of the same order of magnitude as the current observational errors and must be taken into account in all future local measurements of the Hubble expansion rate.