I would guess that in an industrial environment, with an
ungrounded or poorly grounded (i.e. "earthed") case, you
are adding 50/60Hz hum to the local "ground" and some
rectification / charge pumping (like with input ESD diode
networks) is adding to the real signal.
You might repeat the experiment with the case tied hard
to a known good earth ground, and in a location without
much electrical equipment, fluorescent lighting and so on.
You could put an oscilloscope to the internal input point
and eyeball the waveform as you touch / un-touch the
case, to get quantitative. If you've ever just touched the probe
tip on a 'scope you have an idea of how large human body
line frequency "pickup" can be against a high impedance
sense-point.