Airport Delays and Metropolitan Employment · 2017-01-04 · Airport Delays and Metropolitan Employment Motivation Cost of airline delays in 2007 (in $ billion): •Peterson et al.

Airport Delays and Metropolitan Employment

I-TED 2014 – International Transportation Economic Development Conference

April 10, 2014 | Dallas, Texas

Paulos Ashebir Lakew, University of California, Irvine

Volodymyr Bilotkach, Newcastle University

1

Airport Delays and Metropolitan Employment Motivation

Cost of airline delays in 2007 (in $ billion):

• Peterson et al. (2013): a 10% (30%) reduction in delays increases net welfare by $17.6 ($38.5) billion

Relevant Literature Themes1. Airport traffic and urban growth2. Determinants of airline delays 3. Cost of airline delays

Airport traffic and urban growth:

• Airport cities (Berg et al., 1996; Button & Lall, 1999)

• Airport traffic associated with higher service-sector employment and lower manufacturing employment (Brueckner, 2003; Sheard, 2014)

• Effect of air cargo traffic on urban development (Green, 2007; Button & Yuan, 2013)

DIRECTCOSTS

INDIRECTCOSTS

TOTAL COST

Carriers Pass. Economy(Demand)

JEC(2008)

19.1 12.1 9.6 (NA) 40.7

NEXTOR(2010)

8.3 16.7 4.0 (3.9) 32.9

2

Airport Delays and Metropolitan Employment Highlights

What we do:

Quantify impact of delays on employment

• Collect data on airline delays, traffic, airports,

and metro-level employment

• Construct quarterly panel

• Airports aggregated to Metropolitan

Statistical Area (MSA) cross-sections

• 40 Periods (2003Q1-2012Q4)

• OLS (2SLS) estimation with MSA fixed-effects

(IVs for endogenous variables)

• Control for exogenous city features

What we find:

Cross-sectional results:

• Frequency and length of delays increase

Total and Service-related Employment

• Extreme-weather delays have a positive

effect on Total Employment

MSA Fixed-effects results:

• Significant downward pressure by delays

on Total, Service, and Goods Employment

• Increase in the share of carrier-controlled

delays reduces Total Employment

• Results hold in both cross-sectional

and fixed-effect specifications3

Airport Delays and Metropolitan Employment Empirical Framework

• Reduced-form relationship invoked between an MSA i’s employment E, outbound (inbound) traffic T, departure (arrival) delays D, and exogenous city features X, in quarter t:

𝐸𝑖𝑡 = 𝛽𝑇𝑖𝑡 + 𝛿𝐷𝑖𝑡 + 𝛾𝑋𝑖𝑡 + 𝜃𝑡𝑄𝑡 + 𝑢𝑖 + 휀𝑖𝑡 , (1)

• where 𝑄𝑡 , 𝑢𝑖, and 휀𝑖𝑡 denote time dummies, MSA-specific intercept, and error term

• 𝑋𝑖𝑡 includes MSA population, young and old population shares, wages, and temperature

• Equation (1) treats relationship between traffic, delays, and economic development as a contemporaneous one (see Brueckner, 2003)

• Potential endogeneity of airline traffic and delays addressed by a 2SLS estimation, using a set of instrumental variables included in 𝑋𝑖𝑡

4

Airport Delays and Metropolitan Employment Socioeconomic Variables

Data source: U.S. Bureau of Labor Statistics (QCEW)

2-digit NAICS industry classifications used from BLS data (MSA level):

Dependent Variables (𝐸𝑖𝑡):

• Total employment (TOTEMP)

• Service employment (SERV)• Selected Subsectors

• Leisure & Hospitality employment (LEISHOSP)• Trade, transport and Utilities (TTU)• Professional-Business-Finance-Info. employment (PBIF)• Health, Education and Government (HEG)

• Goods employment (GOODS)• Selected Subsectors

• Manufacturing employment (MANUF)

5

Airport Delays and Metropolitan Employment Traffic and Delay Variables

TRAFFIC

• Passengers and freight/mail tons departed (landed) at U.S. airports

• Airport location data from NTAD used to link US airports to corresponding MSAs

• Office of Management and Budget’s (OMB) 2009 CBSA county delineations used to complete crosswalk

DELAYS

• Minute-level schedule delays, gate-to-gate airtime, and flight-level measures for non-stop domestic operations of U.S. major carriers

• >15 minutes departure delays (by origin)

• >15 minutes arrival delays (by destination)

• Starting June 2003, cause-of-delay data: • Carrier controlled: Carrier and Late Aircraft

• Exogenous to carrier: Extreme Weather, National Air System (NAS), and Security

• Cancellations (by origin) and Diversions (by destination)

Data source: U.S. Department of Transportation (Bureau of Transportation Statistics) and NTAD (2012)Form 41 Traffic (T-100 Segment Tables) and On-time Performance databanks

6

Airport Delays and Metropolitan Employment Instruments

• HUB indicates hub cities for passenger (cargo) carriers• HUB = 1 if a carrier at an airport serves at least 25 (20)

destinations/quarter (focus cities dropped), HUB = 0 otherwise• If hub city has multiple airports, HUB is equal to fraction of city’s

airports (to discount the hub airport’s share of traffic)

• Example: For Los Angeles-Long Beach-Santa Ana MSA, HUB = ¼

• SLOT denotes slot-controlled airports operating at capacity• DCA, EWR, JFK, LGA, and ORD

• LEISURE dummy for Las Vegas, NV and Orlando, FL• Hub cities for FedEx and UPS central-sorting airports

(Memphis, TN and Louisville, KY) captured by a SORT indicator

7

Airport Delays and Metropolitan Employment Instruments

• PROXIMITY dummy• Captures traffic-diversion from small-to-large cities (for better services, network

connection, lower fares, facilities, etc.)

• Indicator of a small passenger (cargo) MSA that is within 150 miles of a large one• Small and Large MSAs identified based on their annual traffic output (k-means clustering)

• Smallest and largest airports in the airport identified within the small and large MSAs, respectively

• Dummy constructed to equal 1 if distance between these airports is <= 150 miles

• Airport-to-airport Great Circle distances calculated using NTAD airport coordinates

Small PASS. (CARGO) MSA: < 300K pass. (15K US tons of freight) per year

Large PASS. (CARGO) MSA: > 5 million pass. (175K US tons of freight) per year

Smallest airport in MSA

Largest airport in MSA

8

Airport Delays and Metropolitan Employment Instruments

WEATHER

Data source: National Oceanic and Atmospheric Administration’s (NOAA) Global Historical Climatology Network (GHCN) stations

From selected weather stations located at (in vicinity of) airports in our sample –

• PRCP: Precipitation (rain and melted snow in mm), MSA average

• SNOW: and Snowfall (in mm), MSA average

9

Airport Delays and Metropolitan Employment Controls

WEATHER

• Maximum January Temperature (TMAXJAN)• MSA averages of highest January temperatures recorded at the corresponding

airport GHCN stations (converted to degrees Celsius)

DEMOGRAPHIC

Data source: U.S. Census Bureau’s Intercensal Estimates

• Population (POP)

• YOUNG POP share: 14 and younger

• OLD POP share: 65 and older

10

Airport Delays and Metropolitan Employment Preliminary results (by origin MSA)

• Impact of traffic on employment:• Cross-sectional results: consistent with Brueckner (2003) and Sheard (2014)

• Fixed-effects results: coefficient on TRAFFIC higher for Total and Goods Employment (0.11 and 0.42, respectively), lower for Service-sector Employment (0.06)

• Impact of departure delays and traffic on employment:• Cross-sectional results:

• Frequency (count of delays), length of delays (mean and median), and Cancellations all associated with higher levels of Total Employment and Service Employment

• Results hold for OLS and 2SLS estimations

• Goods employment unaffected by delays

• Fixed-effects results:

• Frequency and length of delays (and total sum of delayed minutes) put significant downward pressure on Total, Service, and Goods Employment

11

Airport Delays and Metropolitan Employment Preliminary results (by destination MSA)

• Impact of arrival delays and traffic on employment:• Follow patterns similar to departure delays (by origin MSA)

• Cross-sectional results: • Frequency and length of delays increase Total Employment and Service Employment

• Fixed-effects results:• Delays put significant downward pressure on Total, Service, and Goods Employment

• Extreme Weather delays consistently have a positive effect on Total Employment (as their share increases, while reducing share of carrier-controlled delays)

• Results mostly hold in cross-sectional analysis

• Increase in share of carrier-controlled delays reduces Total Employment• Results hold in both cross-sectional and fixed-effect specifications

12