I mostly agree, as services is the easiest part of GDP to inflate.
BTW, here's the proof for my previous assertion that
their intensity resolution and dynamic range is both too low.
Original paper:
When your data is too low bits, you can't separate similar intensity measurements, and when it saturates, arbitrarily high intensities are binned together with moderate ones.
How bad is the underestimation? I've tried looking for direct comparisons between urban and rural, but it is surprisingly hard to find hard measurements. Best I could do was an Indian (IK, IK) study on indoor lighting. It turns that even indoor lighting - 100% within human control - can differ by 3 orders of magnitude from 30-28500 lux. For rural areas this value would be near 0 at night. 3 orders of magnitude cannot be accurately captured by a 6 bit number.
They could've used the Chinese radiometric satellite which has 14-bit data and 3 channel RGB sensors for more accuracy.
But this is not surprising at all, since
DMSP was designed to measure solely urban vs rural, not urban light intensity. As for why this information might be useful... we all know.
They also factually get something wrong: China subscribes to SDSS, but was not listed in page A13. Not a big deal but shows that the author should be a bit more careful.