We investigate the possibility that a statistical detection of the galaxy parallax shifts due to the Earth’s motion with respect to the cosmic microwave background frame (cosmic secular parallax) could be made by the Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST) or by the Nancy Grace Roman Space Telescope (NGRST), and used to measure the Hubble constant. We make mock galaxy surveys that extend to redshift z = 0.06 from a large N-body simulation, and include astrometric errors from the LSST and NGRST science requirements, redshift errors, and peculiar velocities. We include spectroscopic redshifts for the brightest galaxies (r < 18) in the fiducial case. We use these catalogues to make measurements of parallax versus redshift for various assumed survey parameters and analysis techniques. We find that in order to make a competitive measurement it will be necessary to model and correct for the peculiar velocity component of galaxy proper motions. It will also be necessary to push astrometry of extended sources into a new regime, and combine information from the different elements of resolved galaxies. In the Appendix, we describe some simple tests of galaxy image registration that yield relatively promising results. For our fiducial survey parameters, we predict an rms error on the direct geometrical measurement of H0 of $2.8{{\ \rm per\ cent}}$ for LSST and $0.8{{\ \rm per\ cent}}$ for NGRST.