Originally Posted by
xjk1
Obviously, using seat maps can be hugely problematic for this, which is acknowledged in section 2.2 by the author. He concludes, however, that "the measurement error associated with seat maps is likely to be small" upon "perform[ing] two analyses". Do you think this is correct?
The correlation would be airline-specific at best given the variety of booking patterns, charges, mix of passenger type and access to advance seat assignments. Beyond that, even within an airline, the trends will vary significantly by route. On a recent international DL trip the domestic connection was nearly full at OLCI and the plane was full at departure. On the international segment only ~50% of seats were assigned when OLCI opened and the flight went out near 80% full, maybe more. My hopes of an empty row were dashed though I was fortunate to have one of the few empty seats remaining in my row.
Not to mention airlines tweak availability based on connecting bookings, not just direct sales on a single flight. The Inventory and Revenue management systems are incredibly complex. Reducing them to seat maps and published inventory is not going to lead to great results.