Airline Fuel Efficiency Ranking Bo Zou University of Illinois at Chicago Matthew Elke, Mark Hansen University of California at Berkeley 06/10/2013 1 1
Outline Introduction Airline selection Mainline efficiency ranking Mainline-regional affiliation Summary 2 2
Introduction Airlines nowadays are more intent to improve fuel efficiency Rising fuel prices Concerns about climate change More environmentally conscious travelers Capability to evaluate airline fuel efficiency becomes important 3 3
Introduction Quantify fuel efficiency of large jet operators in the US airline industry in 2010 Mainline airline efficiency assessment based on multiple approaches Consider regional-mainline affiliation Consider routing structure heterogeneity and passenger O-D trip oriented efficiency Provide a generic and transparent airline fuel efficiency scheme 4
Mainline airline selection 250 Aircraft Gauge Demarcation 200 Avg seats/aircraft 150 100 50 0 5 5
Mainline airline selection Importance of the 15 carriers in system wide fuel consumption Cumulative Fuel Consumptions 1.2E+10 1E+10 8E+09 6E+09 4E+09 2E+09 0 These 15 carriers made up over 80% of all fuel consumed in 2010 Regional airlines made up the other approximately 20% of fuel consumed (excluding Cargo) 6 6
Regional Airline selection Set a threshold of 500,000 enplaned passengers (22) Regionals significant domestic contribution: 20% of fuel consumed 30% of Revenue Aircraft Miles 25% of enplaned passengers 7 7
Contribution by Study group 1.20E+10 Gallons of Fuel 1.00E+10 8.00E+09 6.00E+09 4.00E+09 2.00E+09 0.00E+00 37 airlines studied (15 mainlines + 22 Regionals) ~99% of fuel used >99% of Revenue Aircraft Miles >99% of enplaned passengers 8 8
Mainline airline rankings Ratio approach Frontier based approaches 9 9
Ratio approach Primary Metric used: Fuel / RPM How many gallons of fuel needed on average to move one passenger, one mile Fuel/RPM Fuel/ASM X ASM/RPM Fuel/PM PM/ASM 10 10
Mainline Ranking (Fuel/RPM) Fuel (gal) / RPM 0.0220 0.0200 0.0180 0.0160 0.0140 0.0120 0.0100 0.0080 0.0060 0.0040 0.0020 0.0000 11 11
Fuel inefficiency scores 1.2 Inefficiency Score 1 0.8 0.6 0.4 0.2 0 12 12
Efficiency frontier approach Consider two outputs RPM: measuring mobility Departures (DEP): measuring accessibility and capturing effects of takeoffs/landings on fuel use Fuel consumption efficiency: measured against some best achieved level (fuel frontier): Fuel frontier: how much minimum fuel needed in order to produce a certain amount of output 13 13
Concept of frontier Fuel consumption B D Frontier A C u B : inefficiency of observation B Output 14 14
Specification I: Deterministic frontier 15 15
Specification I: Deterministic frontier Deterministic frontier 16 16
Specification I: Deterministic frontier Deterministic frontier Inefficiency 17 17
Deterministic frontier: Computing inefficiency Deterministic frontier Inefficiency 1. Estimate the frontier parameters using Ordinary Least Square method 2. Calculate residuals 3. Calculate 4. Average by airline to obtain airline inefficiency scores 18 18
Limitations of deterministic frontier Deterministic frontier Inefficiency 1. Random noise (e.g. luck and bad weather) 2. Measurement errors Frontier itself can be stochastic 19 19
Specification II: Stochastic frontier 20 20
Specification II: Stochastic frontier 21 21
Specification II: Stochastic frontier Random noise 22 22
Specification II: Stochastic frontier Random noise Inefficiency 23 23
Specification II: Stochastic frontier Random noise Inefficiency 24 24
Specification II: Stochastic frontier Stochastic frontier Random noise Inefficiency 25 25
Specification II: Stochastic frontier Stochastic frontier Inefficiency 1. We assume inefficiency follows a truncated normal distribution ( always non-negative) 26 26
Specification II: Stochastic frontier Stochastic frontier Inefficiency 1. We assume inefficiency follows a truncated normal distribution ( always non-negative) 2. depends on environmental factors not identically distributed 27 27
Specification II: Stochastic frontier Stochastic frontier Inefficiency 1. We assume inefficiency follows a truncated normal distribution ( always non-negative) 2. depends on environmental factors not identically distributed 28 28
Stochastic frontier: Computing inefficiency Stochastic frontier Inefficiency Estimate the frontier parameters ( ) using the Maximum Likelihood method Calculate Average by airline to obtain airline inefficiency scores 29 29
Frontier estimation results Deterministic Frontier Stochastic Frontier Est P-value Est P-value Ln(RPM) 0.869 0.000 0.823 0.000 Ln(DEP) 0.150 0.000 0.200 0.000 Constant -2.726 0.000-2.344 0.000 Ln(stage length) 0.147 0.037 Ln(ac size) -0.189 0.058 R 2 0.996 Likelihood 79.59 RTS 0.996 0.977 30 30
Frontier estimation results Deterministic Frontier Stochastic Frontier Est P-value Est P-value Ln(RPM) 0.869 0.000 0.823 0.000 Ln(DEP) 0.150 0.000 0.200 0.000 Constant More -2.726 departures 0.000 consume -2.344 more fuel 0.000 Ln(stage length) 0.147 0.037 Ln(ac size) -0.189 0.058 R 2 0.996 Likelihood 79.59 RTS 0.996 0.977 31 31
Frontier estimation results Deterministic Frontier Stochastic Frontier Est P-value Est P-value Ln(RPM) 0.869 0.000 0.823 0.000 Ln(DEP) 0.150 0.000 0.200 0.000 Constant -2.726 0.000-2.344 0.000 Ln(stage length) 0.147 0.037 Ln(ac size) Longer -0.189 stage length 0.058 => R 2 0.996 lower fuel efficiency Likelihood 79.59 RTS 0.996 0.977 32 32
Frontier estimation results Deterministic Frontier Stochastic Frontier Est P-value Est P-value Ln(RPM) 0.869 0.000 0.823 0.000 Ln(DEP) 0.150 0.000 0.200 0.000 Constant -2.726 0.000-2.344 0.000 Ln(stage length) 0.147 0.037 Ln(ac size) -0.189 0.058 R 2 0.996 Likelihood 79.59 RTS 0.996 0.977 Constant RTS: 1% increase in RPM 33 and DEP => ~1% fuel increase 33
1.2 Inefficiency scores under different approaches (ordered by Fuel/RPM) Inefficiency Score 1 0.8 0.6 0.4 0.2 0 Fuel/RPM Stochastic frontier Deterministic frontier 34 34
Inefficiency score correlation Fuel/RPM Fuel/RPM 1 Deterministic Frontier Stochastic Frontier Deterministic Frontier Stochastic Frontier 0.83 1 0.71 0.98 1 35 35
Inefficiency scores (ordered by Fuel/RPM) 1.2 Inefficiency Score 1 0.8 0.6 0.4 0.2 0 Fuel/RPM Stochastic frontier Deterministic frontier 36 36
Inefficiency scores (ordered by Fuel/RPM) 1.2 Inefficiency Score 1 0.8 0.6 Frontier approaches penalize airlines with fewer departures everything else being equal 0.4 0.2 0 Fuel/RPM Stochastic frontier Deterministic frontier 37 37
Mainline-Regional Affiliation Assigning regional RPM s to mainlines Constructing adjusted fuel efficiency measures 38 38
Mainline-Regional Affiliation Domestic air traffic based on: Hub-and-spoke system Mainlines contract out service on lower demand and marginal segments to affiliated regionals Regional airline traffic share steadily increasing 39 39
Types of Affiliations Mainline-regional relationships take two forms Regional is a wholly owned subsidiary of the mainline carrier (i.e. American Eagle) Regional is an Independent company contracting its services out to the mainline airline(s) (i.e. SkyWest) Mainline controls ticketing and scheduling for the regional, and aircraft are under mainline brand Type of relationship determines how RPMs can be accurately apportioned 40 40
Apportioning Regional RPMs Apportion a regional s entire RPMs to one mainline carrier If the regional is a wholly owned subsidiary of that mainline If the regional is independent but has an exclusive agreement with a mainline carrier Apportion regional RPMs for each segment it served If a regional is independent and flies for multiple mainline carriers out of an airport If a regional flies for more than one mainline on a segment, use the ratio of mainline traffic to split RPMs 41 41
Mainline + Regional RPMs 1.20E+11 1.00E+11 8.00E+10 RPM (main) RPM (main+reg) RPMs 6.00E+10 4.00E+10 2.00E+10 0.00E+00 42 42
Mainline + Regional RPMs Regional Carrier Apportioned RPM % RPM app. PSA (new) 1,677,034,927 100% Piedmont Airlines 518,216,513 100% Pinnacle Airlines 4,210,577,910 100% Colgan Air 573,433,520 97% Trans States Airlines 741,021,563 98% CommutAir (champlain ent) 145,073,561 100% Compass Airlines 2,210,100,086 100% Atlantic Southeast Airl. 5,384,997,748 98% Freedom Airlines 315,123,971 100% GoJet Airlines 1,530,592,216 100% American Eagle 7,386,172,780 100% Comair 2,919,863,879 100% Skywest 10,971,400,000 93% Executive Airlines (New) 264,017,675 100% Horizon Air 2,224,661,874 100% Chautauqua Airlines 1,690,870,678 79% Shuttle America 2,609,768,611 96% ExpressJet/Cont. Exp. 7,808,116,996 95% Mesaba Airlines Inc. 3,381,681,196 99% Mesa Airlines 3,538,361,387 91% Republic Airlines/Hughes 5,569,788,120 94% Air Wisconsin 1,820,269,811 100% 43 43
Mainline-Regional composite fuel Ratio approach inefficiency Frontier approach 44 44
Considering regional affiliations reduce the fuel efficiency of the mainline airlines under the ratio approach Regional carriers provide services with high accessibility, and therefore can boost efficiency, leading to smaller overall efficiency change 45 45
Passenger O-D trip oriented fuel efficiency Regional carriers operations support hugand-spoke services provided by mainlines Hub-and-spoke carriers should be penalized if considering efficiency of moving passengers from their origins to destinations 46 46
Passenger O-D trip oriented fuel Ratio approach efficiency where Frontier approach 47 47
Ratio approach penalizes hub-andspoke carriers Frontiers are reshaped, but difference in inefficiency scores w/ and w/o considering circuity is small 48 48
Summary This research examines fuel efficiency of large jet operators in the US Multiple approaches are employed, providing different perspectives on efficiency measurement Regional carriers are important to the overall fuel efficiency of mainline-regional composites Considering circuity changes the traditional airline efficiency picture. However, the effect seems not significant using frontier methods Future research may consider fuel efficiency vs. overall technical efficiency 49 49
Thank you! Questions? 50 50
Calculated circuity 51 51