Election polls are designed to do exactly that, whereas this sample isn't anything even approaching a random subset that could be used for inference about the w

Election polls are designed to do exactly that, whereas this sample isn't anything even approaching a random subset that could be used for inference about the whole population. It may be the case that in state courtrooms judges are the same, worse, or better, but making that inference on the basis of federal-level data is probably unjustified.