Redistribution, Polarization, and Medians: Bringing Data to Downsian Puzzles