SIG DSA - Data Science and Analytics for Decision Support
Loading...
Paper Type
Complete
Paper Number
1726
Description
Conventional demand models (e.g., gravity model) in air transport literature tend to rely heavily on the mainstream econometric variables (e.g., distance, population, and GDP), which cannot be dynamically measured or used for short-term predictions. This study seeks to complement the short-term predictability of such conventional approaches by introducing dynamic predictors while alleviating the endogeneity by implementing panel data modeling analysis. Utilizing 40,072 air passenger data stacked in 3,344 city pairs over twelve months in 2020, we demonstrate that a large variability in demand can be explained by a handful of non-conventional variables such as internet search volume and geometric mobility indicators. The performance of our fixed effect model was dramatically improved by adding the regional intensity of google search for “airport” and “flight” and by adding the measure of people’s time spent at residential areas in the origin and destination state (Adj. R2 to .74).
Recommended Citation
Hopfe, David H.; Lee, Kiljae; and Yu, Chunyan, "Modeling US Air Passenger Traffic Demand: Dynamic Data" (2022). AMCIS 2022 Proceedings. 21.
https://aisel.aisnet.org/amcis2022/sig_dsa/sig_dsa/21
Modeling US Air Passenger Traffic Demand: Dynamic Data
Conventional demand models (e.g., gravity model) in air transport literature tend to rely heavily on the mainstream econometric variables (e.g., distance, population, and GDP), which cannot be dynamically measured or used for short-term predictions. This study seeks to complement the short-term predictability of such conventional approaches by introducing dynamic predictors while alleviating the endogeneity by implementing panel data modeling analysis. Utilizing 40,072 air passenger data stacked in 3,344 city pairs over twelve months in 2020, we demonstrate that a large variability in demand can be explained by a handful of non-conventional variables such as internet search volume and geometric mobility indicators. The performance of our fixed effect model was dramatically improved by adding the regional intensity of google search for “airport” and “flight” and by adding the measure of people’s time spent at residential areas in the origin and destination state (Adj. R2 to .74).
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.
Comments
SIG DSA