Yeah, that was me.
P-value for "PMCO IDBed PAX at a higher rate than PMCO in 2Q 2011" is about 2.8 x 1E-66. (Sorry -- I can't figure out how to get superscripts in FT's version of BB code. Anybody have a workaround for that?)
The P-value (for a given conclusion) is basically the probability of having gotten a false positive. More precisely, if I claim "X" with a P-value of
p, then I'm asserting that I have data that (1) support my claim of "X", and (2) would occur with probability less than
p if "X" was in fact false.
So, the simple interpretation is: if PMCO's overbooking/bumping process is actually less IDB-prone than PMUA's, then in fact we
could have gotten the data posted by
aacharya in that old thread by random chance. But the probability of getting those data --
or any equally extreme or more extreme data -- is less than 1 in 100,000,000,000,000,000,000,000,000,000,000,000,00 0,000,000,000,000,000,000,000,000,000, if in fact PMCO's overbooking/bumping process is less IDB-prone.
Which leads me to conclude that, indeed, PMCO is probably more IDB-prone.
In other words, the data are statistically significant.
That absolutely does
not mean that it's operationally significant. UA IBDed about 1 in 10,000 pax, while CO IDBed about 1.7 in 10,000 pax. For context, if
ua1flyer was at risk of IDB (I'm sure he's not), over his whole career we'd expect him to have around 1 IDB at that rate. So it's not a big risk for any given pax.
It's also probably not a major issue for the airline's bottom line, but some of the related VDB issues mentioned above (profit from overselling at high fares vs. loss from liberal voucher policies) may be.