On "Nonstationarity of AGN variability" -- selection effect

A new paper accepted by ApJL, arXiv:2001.04471 "Nonstationarity of AGN
variability: The only way to go is down!" by Caplar et al, concludes
that visible quasar luminosity is linearly decreasing due to
cosmological effects, and that this change is visible over a 15-year
timeline. It seems to me that a selection effect is a more likely
cause.

Their data sample is 5919 SDSS-DR7 quasars which were re-observed in
the recent HSC survey, with a redshift-dependent magnitude decrease
observed over that time (14.85 years). But what is omitted is that
SDSS-DR7 quasars needed to be of a certain minimum brightness for a
spectrum to be obtained by them. This means that of a specified
complete population of quasars, only the brighter ones would get the
DR7 spectrum -- therefore you'd expect a subsequent average decrease
in brightness. The fainter members of that population (too faint for
SDSS-DR7 to take a spectrum) may well brighten in that 15-year
interval, but HSC doesn't take their spectrum and so they are not
included in the sample. For the sample to be unbiased, HSC should
take spectra in the same way as SDSS did, and so classify the quasars,
and then the sample can be selected from both ends to see the unbiased
behaviour across that 15-year baseline.

The corollary to their methodology would be that if HSC did classify
quasars from their data, and you selected those HSC-found quasars and
then matched them to all SDSS-DR7 pipeline optical data (including
those which were not classified by SDSS-DR7), that you'd likely find
that most of those objects were *brightening* in the 15-year interval.
Let's see that test as a control.

The authors use, as a control sample, a group of non-variable stars
and observe that they don't vary in the same way as the quasars.
Well, why did they use non-variable stars to compare to quasars which
are known to vary? Why didn't they use variable stars?

So I think that the paper has not made its case.

Eric Flesch