LMI. m8210& The AS AC Air Carrier Investment Model. Logistics Management Institute. (Revised) rff^^^mrsm^0^

Transcription

1 Lgistics Management Institute The AS AC Air Carrier Investment Mdel (Revised) NS301RD2 rff^^^mrsm^0^ Earl R. Wingrve III Peter F. Kstiuk Rbin C. Sickles David H. Gd Apprved fr public Distributin Unlimited m8210& LMI

2 THIS DOCUMENT IS BEST QUALITY AVAILABLE. THE PY FURNISHED TO DTIC NTAINED A SIGNIFICANT NUMBER OF PAGES WHICH DO NOT REPRODUCE LEGIBLY.

3 June 1996 The AS AC Air Carrier Investment Mdel (Revised) NS301RD2 Earl R. Wingrve III Peter F. Kstiuk Rbin Q Sickles David H. Gd Prepared pursuant t Department f Defense Cntract DASW01-95-C-0019 and Cntract MDA C The views expressed here are thse f the Lgistics Management Institute at the time f issue but nt necessarily thse f the Department f Defense. Permissin t qute r reprduce any part except fr gvernment purpses must be btained frm the Lgistics Management Institute. Lgistics Management Institute 2000 Crprate Ridge McLean, Virginia

4 Cntents The ASAC Air Carrier Investment Mdel 1 Backgrund 1 NASA's Rle in Prmting Aviatin Technlgy 1 NASA's Research Objective 2 Genesis f the Aviatin System Analysis Capability 2 Gals f the ASAC Prject: Identifying and Evaluating Prmising Technlgies 3 Apprach t Analyzing the Integrated Aviatin System 3 Airline Ecnmics and Investment Behavir Drive the ASAC 4 Ecnmic and Statistical Derivatin f the ASAC Air Carrier Investment Mdel 4 Intrductin 4 Overview f the Basic Air Carrier Investment Mdel 5 Air Travel Demand 6 Air Travel Supply 7 Using the Mdel 11 General Apprach 11 Frecasting Changes in Travel Demand, Airline Csts, and Aircraft Fleets 11 Travel Demand 11 Airline Csts 12 Aircraft Fleets 13 Scenaris and Frecasts 14 Baseline Scenari 14 Frecast f Operating Prfit Margins and Fare Yields in the Baseline Scenari 15 Sensitivity Analysis with Operating Prfit Cnstraint in the Baseline Scenari 18 Other Scenaris: Cmparisns 19 Cnclusin 20 Bibligraphy Bibli. 1 Appendix A. Data Descriptin in

5 Cntents (Cntinued) Appendix B. User's Guide Appendix C. Baseline Scenari Appendix D. Details f Alternative Scenaris IV

6 Illustratins 1. The ASAC Prcess 2 2. Cmpnents f the Integrated Aviatin System 4 3. Schematic f the ASAC Air Carrier Investment Mdel 6 4. Operating Prfit Margins and Aircraft Fleet Grwth fr Nine Majr Airlines 17 Tables 1. Demand Variables 7 2. Supply Variables Revenue Passenger Miles and Fleet Grwth with Different Operating Prfit Margins in the Baseline Scenari Effects f Fuel Price Changes in the Baseline Scenari Baseline and Other Scenari Frecasts 20 v

7 The AS AC Air Carrier Investment Mdel BACKGROUND NASA's Rle in Prmting Aviatin Technlgy The United States has lng been the wrld's leader in aviatin technlgy fr civil and military aircraft. During the past several decades, U.S. firms have transfrmed this psitin f technlgical leadership int a thriving industry with large dmestic and internatinal sales f aircraft and related prducts. In 1992, sales f civil aircraft peaked at $39.9 billin, with exprts f $24.3 billin. Exprts f engines/ parts, and related prducts ttaled $12.4 billin in the same year. The cmparable figures fr 1994 were $26.3 billin, $18.8 billin, and $11.8 billin, respectively. Despite its histric recrd f success, the difficult business envirnment f the past several years has stimulated cncerns abut whether the U.S. aernautics industry will maintain its wrldwide leadership psitin. Increased cmpetitin, bth technlgical and financial, frm Eurpean and ther nn-u.s. aircraft manufacturers has reduced the glbal market share f U.S. prducers f large civil transprt aircraft and cut the number f U.S. airframe manufacturers t nly tw. Order cancellatins and stretch-uts f deliveries by airlines, frthcming nise abatement requirements, and envirnmental cncerns create additinal challenges faced by U.S. prducers and purchasers f aircraft. The primary rle f the Natinal Aernautics and Space Administratin (NASA) in supprting civil aviatin is t develp technlgies that imprve the verall perfrmance f the integrated air transprtatin system, making air travel safer and mre efficient, while cntributing t the ecnmic welfare f the United States. NASA cnducts much f the basic and early applied research that creates the advanced technlgy intrduced int the air transprtatin system. Thrugh its technlgy research prgram, NASA aims t maintain and imprve the leadership rle in aviatin technlgy and air transprtatin held by the United States fr the past half century. The principal NASA prgram supprting subsnic transprtatin is the Advanced Subsnic Technlgy (AST) prgram, managed by the Subsnic Transprtatin Divisin, Office f Aernautics, NASA Headquarters. In cperatin with the Federal Aviatin Administratin and the U.S. aernautics industry, the gal f the AST prgram is t develp high-payff technlgies that supprt the develpment f a safe, envirnmentally acceptable, and highly prductive glbal air transprtatin system. NASA measures the lng-term success f its AST prgram by hw well it cntributes t an increased market share fr U.S.

8 civil aircraft and aircraft cmpnent prducers and t the increased effectiveness and capacity f the natinal air transprtatin system. NASA's Research Objective T meet its bjective f assisting the U.S. aviatin industry with the technlgical challenges f the future, NASA must identify research areas that have the greatest ptential fr imprving the peratin f the air transprtatin system. Therefre, NASA seeks t develp the ability t evaluate the ptential impact f varius advanced technlgies. By thrughly understanding the ecnmic impact f advanced aviatin technlgies and by evaluating hw thse new technlgies wuld be used within the integrated aviatin system, NASA aims t balance its aernautical research prgram and help speed the intrductin f high-leverage technlgies. Genesis f the Aviatin System Analysis Capability Technlgy Integratin is the element f the AST prgram designed t ensure that the technlgies NASA develps are timely and cnsistent with ther develpments in the aviatin system. One f the bjectives f the Technlgy Integratin element is t develp an Aviatin System Analysis Capability (ASAC). This analytical capability will give NASA and ther rganizatins in the aviatin cmmunity greater ability t evaluate the ptential ecnmic impacts f advanced technlgies. ASAC is envisined primarily as a prcess fr understanding and evaluating the impact f advanced aviatin technlgies n the U.S. ecnmy. ASAC cnsists f a diverse cllectin f mdels, databases, analysts, and individuals frm the public and private sectrs brught tgether t wrk n issues f cmmn interest t rganizatins within the aviatin cmmunity. ASAC als will be a resurce available t thse same rganizatins t perfrm analyses; prvide infrmatin; and assist scientists, engineers, analysts, and prgram managers in their daily wrk. ASAC will prvide this assistance thrugh infrmatin system resurces, mdels, and analytical expertise, as well as thrugh its rle as a cnductr and rganizer f large-scale studies f the aviatin system and advanced technlgies. Figure 1 displays this cncept. inputs Outputs Databases Tls and mdels ASAC Prcess Plicy studies Cst-benefit analyses Knwledge and Cmmunicatins and analytical methds...,. cnsensus-building Figure 1. The ASAC Prcess

9 Gals f the ASAC Prject: Identifying and Evaluating Prmising Technlgies The principal bjective f ASAC is t develp credible evaluatins f the ecnmic and technlgical impact f advanced aviatin technlgies n the integrated aviatin system. These evaluatins wuld then be used t assist NASA prgram managers t select the mst beneficial mix f technlgies fr NASA t invest in, bth in brad areas, such as prpulsin r navigatin systems, and in mre specific prjects within the brader categries. Generally, engineering analyses f this kind require multidisciplinary expertise, pssibly using several mdels f different cmpnents and technlgies, giving cnsideratin t multiple alternatives and utcmes. These types f analyses will be mst effective if they use infrmatin and inputs frm rganizatins and analysts frm different parts f the aviatin cmmunity. In this way, the studies will use the expertise f peple arund the United States and build acceptance frm the start f the research effrt. In additin t the need fr identifying brad directins fr investments in technlgy, there is als a need t prvide researchers at NASA and elsewhere with the ability t quickly evaluate the ecnmic ptential f alternative technlgies and systems. By prviding engineers better infrmatin n ptential markets fr technlgies and data n hw the current system wrks, ASAC will help NASA engineers incrprate the needs f their custmers mre easily int their rutine wrk. These types f prblems are mst likely t invlve investigatins int specific technical designs f aircraft r subsystems that wuld be readily substituted fr existing equipment nw used by peratrs, withut requiring significant changes t ther aviatin cmpnents. With such infrmatin, researchers culd mre easily evaluate the utility f alternative designs and btain quick estimates f the value f their design cncepts. Others using ASAC in this way wuld be analysts frm industry, gvernment, and universities. Apprach t Analyzing the Integrated Aviatin System The aviatin technlgies that are mst likely t be useful are nt necessarily the mst technically advanced. Rather, it is critical that NASA and industry invest in the technlgies that have the mst prmising payffs. High-payff technlgies are thse that clearly demnstrate a capacity fr ecnmically viable perfrmance enhancements frm the perspective f thse rganizatins that will purchase and perate the technlgies. Because new aviatin technlgies will be intrduced int a cmplex system, it is critical that the ptential impact f any prpsed technlgy be analyzed frm a systemwide perspective. Otherwise, the ptential impact may be ver- r underestimated due t the unexamined interdependencies with ther elements f the aviatin system. Figure 2 shws the cmpnents f the integrated aviatin system.

10 Air traffic management and airprts Figure 2. Cmpnents f the Integrated Aviatin System Airline Ecnmics and Investment Behavir Drive the ASAC The ASAC differs frm previus NASA mdeling effrts in that the ecnmic behavir f buyers and sellers in the air transprtatin and aviatin industries is central t its cnceptin. T link the ecnmics f flight with the technlgy f flight, ASAC requires a parametrically based mdel that links airline peratins and investments in aircraft with aircraft characteristics. That mdel als must prvide a mechanism fr incrprating air travel demand and prfitability factrs int the airlines' investment decisins. Finally, the mdel must be flexible and capable f being incrprated int a wide-ranging suite f ecnmic and technical mdels that are envisined fr ASAC. The remainder f this reprt describes a prttype air carrier investment mdel, develped by LMI, that meets these requirements. ENOMIC AND STATISTICAL DERIVATION OF THE ASAC AIR CARRIER INVESTMENT MODEL Intrductin In creating the ASAC Air Carrier Investment Mdel, we had sme specific gals in mind. A primary bjective was t generate high-level estimates frm

11 brad industry-wide supply and demand factrs. We envisined being able t frecast the demand fr air travel under a variety f user-defined scenaris. Frm these air travel demand frecasts, we then culd estimate the derived demand fr the factrs f prductin; mst imprtant, the number f aircraft in the fleets f U.S. passenger air carriers. We culd als gauge the financial health f the airline industry as expressed in its perating prfit margins. T create the mdel, we first identified 85 key U.S. airprts frm which flights riginate, and then develped airprt-level demand mdels fr passenger service prvided by majr air carriers. Furthermre, we linked the air carrierspecific demand schedules t an analysis f the carriers' technlgies via their cst functins expressed in terms f the prices f the majr inputs labr, fuel, materials, and flight equipment. Flight equipment was mdeled in an especially detailed way by incrprating sme key perating characteristics f aircraft. 1 Frm the cst functins, we generated derived demand schedules fr the factrs f prductin, in particular aircraft fleets. The derived demand schedules are functins f the price f the factr f prductin, prices f ther factrs, parameters that describe the aircraft and the netwrk used by a carrier, and the level f passenger service supplied. Because it is s capital-intensive, the airline industry must earn an perating prfit margin f between 4 and 6 percent if it is ging t maintain and expand its aircraft fleet. Accrdingly, we added an perating prfit margin cnstraint t the mdel. When this ptin is activated, passenger fare yields are adjusted up r dwn t ensure that the target perating prfit margins are met. Overview f the Basic Air Carrier Investment Mdel As shwn in Figure 3, the ASAC Air Carrier Investment Mdel starts with the factrs affecting the demand fr air passenger travel at the airline and airprt levels. It then examines the determinants f airline cst functins and the resulting industry supply curve. The bjective f bth analyses is t btain parametric estimates fr the air travel demand and airline cst functins. These parametric estimates can then be cmbined with user-specified values f key supply and demand variables t generate industry-level frecasts f revenue passengermiles (RPMs) flwn, 2 number f aircraft in the fleets f U.S. passenger air carriers, and perating margins under varius scenaris. 1 Acting under subcntract t LMI, Prfessr Rbin Sickles f Rice University and Prfessr David Gd f Indiana University generated the data sets and perfrmed an ecnmetric study f majr U.S. airlines. They were assisted by Anthny Pstert, a Ph.D. student at Rice University. See the bibligraphy fr a listing f previusly published studies by Sickles and Gd. 2 One revenue passenger (persn receiving air transprtatin frm the air carrier fr which remuneratin is received by the air carrier) transprted ne statute mile.

12 1. Estimate airline and airprt-level demand Own fare yield Cmpetitr fan: yield Per capita incme Ppulatin Unemplyment rate >>Ä»M*>>>j«*X 2. Estimate airline cst functins Airline utputs Input quantities Input prices Stage length Lad factr Seats per aircraft Aircraft age Percentage jets Percentage wide-bdied aircraft >»»ÄO»KWJ»>x«'i»»»>:wx«'»w»>: Fare yield Incme grwth Ppulatin grwth Unemplyment rate Revenue passenger-miles Predicted number f aircraft in fleet Operating margins 5. Outputs frm Labr csts Energy prices Materials prices Capital prices Average stage lengths Lad factr Average seats per aircraft Average age f aircraft Percentage jets Percentage wide-bdied aircraft Supply Travel demand Airline cst functins 3. Parametric estimates 4. Scenari variables Figure 3. Schematic f the ASAC Air Carrier Investment Mdel Air Travel Demand Our first analytical task was t develp a mdel f demand fr an airline's passenger service. Frm a particular airprt at rigin i, carrier j will generate a certain level f passenger traffic. The U.S. Department f Transprtatin's (DOTs) Origin and Destinatin data recrd a 1 in 10 sample f all tickets; frm these, the RPM service riginating at a particular airprt fr a particular carrier was cnstructed. Demand fr a carrier's service is driven by the carrier's passenger fare yield (measured by the average ticket price fr flights riginating at airprt i divided by the average number f RPMs flwn), its cmpetitrs' yields, and the size and ecnmic prsperity f the market. We mdeled the ecnmic characteristics f the Standard Metrplitan Statistical Area (SMSA) surrunding the 85 airprts in the study in terms f the area's ppulatin, per capita incme, and unemplyment rate. The perid under cnsideratin was frm the first calendar quarter f 1979 thrugh the last calendar quarter f The demand functin, in equatin frm, is tftu Dt f i f j(pt,i,j, Pt,i r X i,i) f [Eq. 1] where q tij is the scheduled demand (in RPMs) riginating at time t frm airprt i fr carrier;'; p tij is the average yield fr service riginating at time t frm airprt i

13 fr carrier ;'; and p tic is the average yield fr the ther carriers generating traffic at time t frm airprt i. The x t,. are the ther demand characteristics at time t fr airprt i. Cnventinal treatments fr firm and airprt fixed effects were used. These effects capture thse imprtant characteristics f a particular city that are nt easily measured, such as turism effects. We used a lg-lg specificatin fr Equatin 1, s that the regressin cefficients may be interpreted as elasticities. Ttal demand fr an air carrier's passenger service was then cnstructed by summing the airprt-specific demand equatins. In terms f Equatin 1, the ttal demand fr a carrier's service is given by ap Qt,j = X Qt,i,ir [Eq.2] where ap is the number f airprts (85). Table 1 shws the demand variables that were incrprated int the mdel. All f the explanatry variables were fund t be statistically significant at the 95 percent level f cnfidence. 3 Table 1. Demand Variables Variable Name Cefficient T-rati Own fares LNAVEOWN Cmpetitrs' fares LNAVEOTHER Per capita incme LNPCI Ppulatin LNPOP Unemplyment rate LNUNRATE Nte: Estimates f firm and airprt variables are nt reprted. Air Travel Supply The secnd majr cmpnent f ur ecnmetric study explains ttal carrier csts in terms f utput quantities, factr prices, aircraft attributes, and netwrk traits. The cst analysis was based mainly n bservatins frm the DOT Frm 41 data (discussed in mre detail in Appendix A). The cst data fllw 13 U.S. passenger air carriers with quarterly bservatins between the beginning f 1979 and the end f These firms are the set f frmer certificated 3 The partial regressin cefficients shw the effects f changes in the independent variables (e.g., wn fares, and cmpetitrs' fares) n the dependent variable (i.e., ttal demand fr an air carrier's passenger service). The T-ratis shw the degree t which the partial regressin cefficients are statistically different frm zer. Fr degrees f freedm ver 30, a T-rati f 1.96 prvides 95 percent cnfidence that the partial regressin cefficient is nt zer.

14 carriers that existed thrughut the study perid and accunt fr well ver 95 percent f the dmestic air traffic. Frm the Frm 41 data, we generated a separate set f demand equatins fr each f the carrier's factrs f prductin based n standard ecnmic assumptins cncerning the cst-rrunimizing behavir f a carrier. In turn, these demand equatins permit examinatins f the impact f factr price and factr prductivity changes, fleet and netwrk cnfiguratins, and aircraft perating characteristics. Scheduled RPM traffic fr carrier j at time t was cnstructed as the sum f riginating traffic supplied by the carrier fr all airprts frm which it ffered flights. This was the first f the tw utputs cnsidered in the cst functin belw. The secnd was the level f nnscheduled RPM service. The tw generic utput categries at time t fr carrier; are designated y 3 and y t., fr scheduled and nnscheduled RPM demand, respectively. The factrs f prductin are labr, energy, materials, and capital. Factr prices are labeled w. In the mdel, capital refers t aircraft fleets nly. Capital ther than aircraft, such as grund structures and grund equipment, is included in the materials categry. Omitting the time and firm subscripts, the transcendental lgarithmic (translg) cst functin is given by In TC = cx + X cilny, + X X a,)lnyilny ; + ß,lniü, IZ ßijlnWiinWj + X p; aircraft attributesi\nw cap i ta, [Eq. 3] 2 + X A-, netwrk traitsi. Cst shares fr labr, energy, and materials are given by M, = ß f + Xß lnw,. [Eq.4] The cst share fr capital is 4 4 M-cajntd = Pcapitai + X ficapitaim Wj + X p, aircraft attributes^. [Eq. 5] The translg cst equatin can be viewed rughly as a secnd-rder apprximatin f the cst functin dual t a generic prductin functin. Symmetry and linear hmgeneity in input prices are impsed n the cst functin by the restrictins: a*, = a n, \/i f j; ß (/ = ß ;! -, Vz',;'; X ; ß«= 1; X, ßi, = 0; and X, p, = 0.

15 Summary statistics based n the translg cst equatin and its assciated share equatins are prvided by the Mrishima and Allen-Uzawa substitutin elasticities. 4 Several measures f returns t scale can als be btained frm the parameter estimates. Aircraft attributes are mdeled frm varius characteristics f the aircraft fleet. A majr cmpnent f airline prductivity grwth is measured by changes in these attributes ver time. Fr example, all ther things being equal, newer aircraft types are expected t be mre prductive than lder types. The mst significant cntributin t prductivity grwth in the 1960s was the intrductin f jet equipment. While this innvatin was widely adpted, it was nt universal fr carriers thrughut the data sample. Newer wing designs, imprved avinics, and mre fuel-efficient prpulsin technlgies als make flight equipment mre prductive. Once an aircraft design is certified, a large prtin f the technlgical innvatin becmes fixed fr its prductive life. In an engineering sense, transprtatin industries tend t be characterized by increasing returns t equipment size. Fixed csts fr fuel, pilts, terminal facilities, and even landing slts can be spread ver mre passengers. Hwever, large aircraft size is nt withut ptential disecnmies. As equipment size increases, it becmes mre difficult t fine-tune air traffic scheduled capacity n a particular rute. Because airline capacity (reflected by available seat-miles) is cncentrated int fewer and fewer departures, quality f service als declines (the prbability decreases that a flight is ffered at the time a passenger desires it mst). This raises particular difficulties in cmpetitive markets where an airline's capacity must be adjusted in respnse t the behavir f rival carriers. Deregulatin has accentuated this liability by virtually eliminating mnplies in dmestic high-density air travel markets. On the ther hand, deregulatin has increased the ttal vlume f traffic thrugh mre vigrus fare cmpetitin, smewhat attenuating this liability. In any event, the perating ecnmies f increased equipment size must be traded ff against limited flexibility. Tw attributes f the carrier's netwrk are als included in the mdel: average stage length and passenger lad factr. Stage length allws us t accunt fr different ratis f csts due t grund-based resurces cmpared with csts attributable t the actual stage length flwn. Shrter flights use a higher prprtin f grund-based systems per passenger-mile f utput than d lnger flights. Als, shrter flights tend t be mre circuitusly ruted by air traffic cntrl and spend a lwer fractin f time at an efficient altitude than lnger flights. Passenger lad factr can be viewed as a cntrl fr capacity utilizatin and macrecnmic demand shcks. Many transprtatin studies als interpret it as a prxy fr service quality. As lad factrs increase and the netwrk becmes less resilient, the number and length f passenger flight delays generally increase as d the number f lst bags and ticketed passengers wh are bumped. Inflight service levels als decline since the number f flight attendants is nt generally adjusted upward as passenger lad factr increases. 4 The Mrishima and Allen-Uzawa substitutin elasticities are measures f the degree t which the varius factrs f prductin may substitute fr ne anther, hlding factr prices and the level f prductin cnstant.

16 Estimates f the lng-run cst functin and summary statistics fr varius elasticities are prvided in Table 2. Table 2. Supply Variables Variable Name Cefficient T-rati Labr price LNLP N/A Labr price squared LNLP A Labr x energy LNLPEP Labr x materials LNLPMP Labr x capital LNLPKP Energy price LNEP N/A Energy price squared LNEP A Energy x materials LNEPMP Energy x capital LNEPKP Materials price LNMP N/A Materials price squared LNMP A Materials x capital LNMPKP Capital price LNKP N/A Capital price squared LNKP A Scheduled demand LNSQ Scheduled demand squared LNSQ A Nnscheduled demand LNNQ Nnscheduled demand squared LNNQ A Scheduled x nnscheduled demand LNSQNQ Stage length LNSL Lad factr LNLF Average seats XLNAS Average age XLNAA Percentage jets a XXPJ Percentage wide-bdied aircraft 3 XXPWB Nte: Estimates f firm and quarterly dummy variables are nt reprted. * All ther variables are expressed as natural lgarithms. 10

17 USING THE MODEL General Apprach The jint mdel f supply and demand fr cmmercial passenger air service specified in ur study and the inferences abut the demand fr airplanes that are imbedded in ur ecnmetric results allw us t simulate the effects f emerging airframe and engine technlgies by mdifying characteristics f the planes in service. We can als simulate the grwth in ttal system demand fr passenger service and fr factr inputs such as the number f aircraft in the fleet. We incrprated the estimated regressin cefficients int a spreadsheet mdel. Details f the mdel and user instructins are fund at Appendix B. We fllw several general steps when evaluating scenaris: First, we predict the change in RPMs based upn ecnmic frecasts and the demand equatin estimates. Next, we estimate airline revenues based upn frecast RPM grwth and hypthesized changes in ticket prices. Then we estimate airline perating csts n the basis f frecasted RPM grwth, changes in input prices, and changes in aircraft and netwrk characteristics. We predict the aircraft inventry frm airline perating csts, the capital share equatin, and hypthesized changes in aircraft price and aircraft size. Finally, we cmpare frecasts frm the ASAC Air Carrier Investment Mdel with predicted changes in RPMs, aircraft fleet, and perating margins frm ther published frecasts. Frecasting Changes in Travel Demand, Airline Csts, and Aircraft Fleets TRAVEL DEMAND T predict changes in travel demand, the mdel starts with actual airline utput fr the last tw quarters f 1993 and the first tw quarters f 1994 and changes it ver time based n the estimated demand functin cefficients and predicted changes in the explanatry variables. The equatin fr predicting annual changes in demand is % ARPM = Z ß, % AX ;, [Eq. 6] where the ß ; are the cefficients estimated frm the ecnmetric mdel and the X; are the explanatry variables. Due t the lgarithmic structure f the statistical mdel, the cefficients are interpreted as elasticities. Fr example, the cefficient f n per capita incme means that a 1 percent increase in per capita incme raises the demand fr air travel by percent. The annual percentage change in per capita incme, ppulatin, and unemplyment are parameters entered by the user. The baseline mdel uses 11

18 estimates f ppulatin grwth published by the Bureau f Labr Statistics. Per capita incme grwth is nt directly input int the mdel. Instead, the user prvides estimates f the lng-run annual grwth rates in grss dmestic prduct and ppulatin. The mdel then calculates the annual change in per capita incme and uses it t generate the demand frecast. Fare variables are treated in ne f tw pssible ways. User-defined rates f change in fare yields can be input directly int the mdel, and their effects will be estimated immediately. The secnd mde f peratin, as described later in the reprt, allws the user t set a series f prfit rate cnstraints fr each f the fur, five-year intervals in the frecast perid. The user then instructs the mdel t vary the fare yields until the prfit rate cnstraints are met. The ecnmetric estimates f the demand functin are based n quarterly traffic vlume fr each airline and airprt in the sample. While it is pssible t build the demand frecasts up frm this highly detailed level, it wuld be timecnsuming and prbably add mre inaccuracy t the final estimate. Instead, we use the actual scheduled and nnscheduled RPM data fr the dmestic and internatinal rutes f U.S. passenger airlines as the starting pint, and grw demand at the rate indicated by Equatin 6. This impses the cnstraint that utput grws at the same rate fr each airline. While bviusly inaccurate, this is nt a significant bias in the mdel since ur gal at this time is t frecast industry-wide demand, csts, and aircraft fleets. Fr lng-run frecasts such as thse generated by the mdel, it is immaterial whether the aggregate demand fr air travel is satisfied by a particular carrier such as United Airlines r Delta Airlines. Fr purpses f frecasting fares and fr calculating industry travel demand, the wn-fare and ther-fare changes are assumed t be identical. Therefre, the verall price effect is the sum f the tw cefficients. The net effect shws that air passenger travel is sensitive t price changes, but nt unusually s. The mdel predicts that a 10 percent reductin in fares will increase RPMs by 10.7 percent. This implies that after hlding ther factrs cnstant - such as ppulatin and incme changes in air fares will have virtually n effect n ttal revenues cllected bv the industry. AIRLINE STS Equatin 3 describes the airline cst equatin estimated fr the mdel. As shwn, ttal csts are a functin f airline utputs, input prices, and aircraft and airline netwrk attributes. Using the supply parameter estimates shwn in Table 2, Equatin 3 can easily be used t prduce a time series f predicted 12

19 changes in airline csts. Using the lg-lg structure f the equatin t ur advantage, the fllwing frecast equatin is derived: %ATC = i a,%ay«+ Ü cxi/%ay,-%ay/ + 2 ß;%Aw, i=i <<n=i i=i X X ßi,%Azüi%Aw, + X p; %Aaircraft attributesi %Aw capita i [Eq. 7] iö'r=l ' ;=l 2 + X A,,- % Anetwrk traits ir 1=1 where %A means annual percentage change in the variable. In Equatin 7, factr prices 10, aircraft attributes, and netwrk traits are userdefined variables in the ASAC Air Carrier Investment Mdel. Scheduled and nnscheduled utput changes are estimated directly in the demand mdel frecasting cmpnent and then input int the cst functins. Therefre, changes in utput cannt be made directly by the user. As with the demand frecasts, ttal csts are prjected frward frm the baseline defined by the reprted data. The mdel increases the csts at the rates predicted by the mdel, given utput frecasts, input price changes, and changes in aircraft and netwrk characteristics. AIRCRAFT FLEETS Estimating the aircraft fleet required t meet the frecasted travel demand is a smewhat mre invlved prcess. Fur factrs enter int the frecast f aircraft fleets: The changes in ttal airline csts The estimated share f aircraft csts in ttal csts The frecasted change in aircraft capital csts The grwth in average aircraft size. Changes in ttal airline csts were discussed in the previus sectin. Referring t Equatin 5, the aircraft share f ttal csts is a functin f input prices and aircraft attributes. As with the cst and demand frecasts, we update the capital share equatin thrugh the frecast perid as a functin f the rates f 13

20 change in the factr price and aircraft attribute parameters. The equatin fr changes in the capital cst share is 4 4 A capital cst share = X ß cap it fl /,,%Atü,- + X pj%aaircraft attributes^. [Eq. 8] The resulting capital share time-series predicts the fractin f ttal csts that will be spent n aircraft investments. Frm Equatin 8, the capital share varies with changes in the price f aircraft and with changes in aircraft characteristics. By multiplying this share estimate by ttal csts, we btain a time-series f capital investments in aircraft. The final pieces f infrmatin needed t calculate the number f planes in the aircraft fleet are the predicted level f aircraft prices and the average aircraft size. Bth variables are defined by users f the mdel. The aircraft price variable can include mre than simply the implicit rental price. As it reflects a mre cmprehensive measure f aircraft wnership csts, it can als be used t reflect the impact f changes in the prductivity f aircraft. The rate f grwth in aircraft size is als selected by the user, with size measured by the average number f seats. The prduct f the aircraft price index and the average size are divided int the aircraft investment t get the estimated number f planes in each airline's fleet. In equatin frm the frmula is., capital share x ttal cst number f planes =. [Eq. 9] aircraft price x average size The required fleets fr all the airlines are then summed t get the industry estimate. SCENARIOS AND FORECASTS Baseline Scenari Using the baseline values specified in Appendix C fr the supply and demand variables, the ASAC Air Carrier Investment Mdel prjects annual grwth in travel demand f 4.55 percent fr the perid f 1994 thrugh This predictin cmpares quite favrably with annual grwth frecasts f 4.74 percent and 4.69 percent frm the Being Cmpany (Being) and the Federal Aviatin Administratin (FAA), respectively. In terms f the number f planes required t satisfy this grwth in travel demand, the ASAC mdel prjects annual grwth in the U.S. cmmercial airline fleet f 2.36 percent fr the perid f 1994 thrugh This predictin is between Being's frecast f a 2.13 percent annual grwth and the FAA's frecast f a 3.28 percent annual grwth. Other details fr the baseline scenari are fund in Appendix C. 14

21 FORECAST OF OPERATING PROFIT MARGINS AND FARE YIELDS IN THE BASELINE SCENARIO When applied t the baseline scenari, ur mdel predicts increasing prfitability fr the airline industry during the frecast years. Fr example, the industry perating prfit margin rises t 8 percent by the year 2005 and cntinues t grw thereafter as csts fall mre rapidly than fares. This is clearly unreasnable fr the highly cmpetitive airline industry. T make the mdel reflect actual industry cnditins mre faithfully, three imprtant characteristics f the industry need t be incrprated int the mdel: Cmpetitin amng airlines that keeps perating prfits at realistic levels Links between airline csts and fare yields Interdependency between fares and prfitability. The ASAC Air Carrier Investment Mdel accmmdates these features with a straightfrward extensin. It builds an industry-wide target prfit rate int the mdel. T meet the target prfit rate, the mdel adjusts fare yields until the target is met. This apprach incrprates the impact f cmpetitin int the frecast and allws the degree f cmpetitin t be set directly thrugh the target margins. By chsing an apprpriate prfit rate, the user can als ensure that adequate capital is available t finance the purchase r lease f the aircraft needed t satisfy the grwing demand fr air travel. As implemented in the mdel, separate target prfit rates can be set fr each f the fur, five-year intervals within the frecast perid. Specifying fur distinct perids permits the user t include changes in the ecnmic envirnment during the frecast perid. Fr example, many financial analysts tday claim that airlines will nt purchase additinal aircraft until their balance sheets are "repaired." One way t implement this cncept is t set a higher prfit margin during the first five-year interval and then reset the target at a lwer, histrically reasnable level. Such a scenari will keep fares and prfits at a higher level fr five years, while reducing the derived demand fr aircraft and ther inputs. The mdel des nt impse the margin cnstraint in every single year. Instead, the mdel iterates changes in fare yield until the target margin in the final year f each interval is satisfied. Since the mdel uses a cnstant rate f fare change within each five-year interval, the perating margin des nt equal the target until the final year f the perid. In practice, the prfit margin mves in equal increments within the interval. If the target margins are the same at the beginning and end f the five-year interval, the margin will be the same in each year. This apprach explicitly lets fare changes be set by the degree f cmpetitin and the level f csts thrughut the industry. It allws fr a market-based mechanism fr translating cst changes int prfits and fare changes. One implicatin f this apprach is that cst-reducing technlgies will primarily 15

22 benefit the traveling public and nt result in higher prfits fr the airlines ver the lng run. While sme airlines may benefit fr a shrt while, cmpetitin will eventually drive fares dwn as mst airlines adpt the cst-reducing technlgy. This analysis is cnsistent with ecnmic thery and als appears t be an accurate descriptin f the airline industry. The relatively lw prfit margins reprted by the airline industry demnstrate the speed with which innvatins and new technlgies diffuse thrughut the industry. The ease f entry fr new airlines with access t cheap lder aircraft keeps prfit margins lw, and it is unlikely that this situatin will change in the near future. Several alternative prfit measures culd be used t implement this apprach in ur mdel. We chse t use the perating prfit margin, which is revenues minus perating csts, divided by revenues. The perating margin des nt reflect interest paid n debts r a return t cmmn sharehlders, bth imprtant elements f cst in a capital-intensive industry such as the airlines. Capital expenses vary significantly frm airline t airline, and in particular, will be strngly affected by whether the airline flies new r ld aircraft. An equally imprtant questin is what target perating margin shuld be used in the mdel. Being states that an perating prfit margin f abut 5 percent is prbably required fr the airline industry t remain healthy enugh t meet increasing travel demands and purchase new aircraft. An examinatin f the histrical data tends t cnfirm this cnclusin. Figure 4 shws perating margins and the percentage change in aircraft fleets fr nine majr air carriers (American, Cntinental, Delta, Eastern, Nrthwest, Trans Wrld, United, USAir, and Suthwest) frm 1978 thrugh While there is clearly a great amunt f variability in the year-t-year numbers, the years f greatest and mst cnsistent grwth in fleets was the mid-1980s. This was als the nly extended perid f prfitability fr the industry during these years. While the change in aircraft fleets may be smewhat skewed because f the effect f mergers ver this time, the numbers clearly demnstrate a strng crrelatin between prfitability and aircraft inventries. The results are reinfrced when ne cnsiders that new aircraft deliveries in the early 1990s were frequently frm rders placed much earlier. The chart demnstrates clearly the imprtance f incrprating a limit n airline prfits in the investment mdel. 16

23 Year Change in aircraft fleet Operating margin Figure 4. Operating Prfit Margins and Aircraft Fleet Grwth fr Nine Majr Airlines Table 3 shws the impact n the grwth f revenue passenger miles (RPMs) and the U.S. aircraft fleet frm varying the perating prfit margin cnstraint. Frm 1995 thrugh 2005, the mdel predicts that RPMs wuld grw 65.4 percent with a target margin f 2.5 percent. This scenari represents a lw-fare envirnment, with fare yields falling abut 1.8 percent per year frm 1995 thrugh The aircraft fleet grws by 33.2 percent during this perid under the lw-fare scenari. Unfrtunately, the lw perating prfit margins will hinder the ability f the airline industry t finance the required grwth in the fleet thrugh internal cash flw r external surces. In cntrast, the high-prfit scenari sees RPM grwth falling t 57.7 percent ver the 10-year interval. Aircraft fleet inventries grw nly 27.4 percent. Althugh the industry has high prfits with which t finance fleet expansin, there is less f a requirement t d s. The reasn fr the slwer industry grwth rates is that fares are higher. Fares fall nly abut 0.8 percent annually frm 1995 thrugh 2000, and then fall by 1.8 percent per year during the fllwing fiveyear perid. On the basis f the histrical data described by Figure 4, the lw-fare, high- RPM grwth scenari reflects the current cmpetitive envirnment. A return t fare regulatin r a wave f mergers that restrict entry by lw-cst airlines culd result in a mre prfitable industry envirnment with higher fares. The lngrun equilibrium is likely the ne envisined by Being, with an perating margin f abut 5 percent prbably required fr the airline industry t remain financially healthy enugh t meet travel demands and purchase new aircraft. 17

24 Table 3. Revenue Passenger Miles and Fleet Grwth with Different Operating Prfit Margins in the Baseline Scenari Operating prfit margin target Revenue passenger-miles (billins) Aircraft fleet % ,512 4,972 5,747 Grwth (%) Annual fare changes (%) % ,532 5,108 5,905 Grwth (%) Annual fare changes (%) % ,551 5,243 6,061 Grwth (%) Annual fare changes (%) SENSITIVITY ANALYSIS WITH OPERATING PROFIT NSTRAINT IN THE BASELINE SCENARIO The enhanced mdel can nw be used t calculate the direct and secndary effects f cst changes in the industry. By hlding perating prfit margins cnstant at the 5 percent target, changes in fuel prices r ther inputs can be lgically allcated between airline industry prfit and fare effects. Table 4 shws the effects f changes in fuel prices ver the frecast perid. The baseline scenari assumes a 0.9 percent annual increase in the real price f fuel. Fr the sensitivity analysis, we use annual changes f 1.0 percentage pint abve and belw the baseline value. As Table 4 shws, changes f 1.0 percentage pint per year in the grwth f energy prices measurably affect industry frecasts. 5 With energy prices rising 0.9 percent annually and perating prfit margins set at 5 percent, ttal airline industry utput rises 61.6 percent frm 1995 thrugh If energy prices fall by 0.1 percent per year, RPMs grw by 64.1 percent. Cnversely, if energy prices rise by 1.9 percent annually, airline industry utput rises nly 58.8 percent. 5 This is a ceteris paribus analysis where a single factr the price f energy is allwed t vary while all ther factrs are held cnstant. In this regard, this illustrative sensitivity analysis differs frm the "il price shck" scenari discussed later in this reprt. 18

25 The difference in the size f the industry-wide aircraft fleet in the year 2005 under cases f higher and lwer energy prices is 221 planes. At an average $66 millin per aircraft, this amunts t a swing in revenues f $14.6 billin fr the aircraft manufacturing industry. Table 4. Effects f Fuel Price Changes in the Baseline Scenari Revenue passenger-miles (billins) Aircraft fleet Energy price change % ,525 5,056 5,794 Grwth (%) Annual fare changes (%) % ,532 5,108 5,905 Grwth (%) Annual fare changes (%) % ,539 5,160 6,015 Grwth (%) Annual fare changes (%) Other Scenaris: Cmparisns T demnstrate the reasnableness and utility f ur mdel, we evaluated a set f alternative scenaris. These are summarized in Table 5. As nted earlier, the general effect f the 5 percent perating margin target is t pass sme f the benefit t the traveling public during favrable scenaris and t pass sme f the csts t the traveling public when the scenari is unfavrable. Lwer r higher airfares are the mechanism fr transmitting the benefits r csts t cnsumers. Fr example, when ecnmic grwth is high and unemplyment is lw, hlding perating prfit margins t 5 percent results in lwer airfares than wuld therwise ccur. This increases the grwth rate fr travel demand and the number f aircraft in the fleet. Cnversely, when there is an il price shck, the 5 percent perating prfit margin target means that the traveling public absrbs sme f the increase in the price f energy thrugh higher airfares. Cnsequently, travel demand grws less rapidly and the number f aircraft in the fleet shrinks mre than when the prfit targets are nt binding. Details f the scenari frecasts and illustrative printuts frm the ASAC Air Carrier Investment Mdel are in Appendix D. 19

26 Table 5. Baseline and Other Scenari Frecasts Scenari Annual change fr affected variables (%) Baseline values (%) Grwth in travel demand ( ) (%) Grwth in aircraft fleet ( ) (%) Industry perating margins in 2005 (%) Baseline High grwth with lw emplyment Incme = 3.0 Unemplyment rate= When perating margin targets are binding Oil price shck Incme = 0.0 Energy price = 2.0 When perating margin targets are binding Airline fare war Fare yield = When perating margin targets are binding NCLUSION T link the ecnmics f flight with the technlgy f flight, NASA's Aviatin System Analysis Capability (ASAC) requires a parametrically based mdel that links airline peratins and investments in aircraft with aircraft characteristics. That mdel als must prvide a mechanism fr incrprating air travel demand and prfitability factrs int the airlines' investment decisins. Finally, the mdel must be flexible and capable f being incrprated int a wideranging suite f ecnmic and technical mdels that are envisined fr ASAC. The ASAC Air Carrier Investment Mdel meets all f these requirements. The prttype mdel incrprates ecnmetric results frm the supply and demand curves faced by U.S. air carriers. It prvides analysts with the ability t prject revenue passenger-miles flwn, number f aircraft in the fleet, and perating margins under varius user-defined scenaris. The mdel als demnstrates the feasibility f incrprating sphisticated regressin results int a user-friendly spreadsheet envirnment. Future wrk will invlve refining the 20

27 ecnmetric results, and the spreadsheet may be replaced with prgramming cde. 21

28 Bibligraphy Andersn, J., and M. Kraus. "Quality f Service and the Demand fr Air Travel." Review f Ecnmics and Statistics, Vl. 63,1981: Arntt, R. J., and J. E. Stiglitz. "Cngestin Pricing t Imprve Air Travel Safety." Transprtatin Safety in an Age f Deregulatin. New Yrk: Oxfrd University Press, 1989: Avmark, Inc. "The Cmpetitiveness f the Eurpean Cmmunity's Air Transprt Industry." Prepared fr the Cmmissin f the Eurpean Cmmunities, Bailey, E. E., and D. M. Kirstein. "Can Truth in Airline Scheduling Alleviate the Cngestin and Delay Prblem?" Transprtatin Safety in an Age f Deregulatin. New Yrk: Oxfrd University Press, 1989: Banker, R., A. Charnes, and W. W. Cper. "Sme Mdels fr Estimating Technical and Scale Efficiencies in Data Envelpment Analysis." Management Science, Vl. 30,1984: Barla, P., and S. Perelman. "Technical Efficiency in Airlines Under Regulated and Deregulated Envirnments." Annals f Public and Cperative Ecnmics, Vl. 60, N. 1,1989: Ben Akiva, M., and S. Lerman. Discrete Chice Analysis. Cambridge, Massachusetts: MIT Press, Bernstein, J. I., and M. I. Nadiri. "Rates f Return n Physical and R&D Capital and Structure f the Prductin Prcess: Crss Sectin and Time Series Evidence." Advances in Ecnmetric Mdeling. Bstn: Kluwer Academic Publishers, 1989: Brenstein, S. "Airline Mergers, Airprt Dminance, and Market Pwer." American Ecnmic Review, Vl. 80,1990: Brenstein, S. "The Evlutin f U.S. Airline Cmpetitin." Jurnal f Ecnmic Perspectives, Vl. 6,1992: Brenstein, S., and N. L. Rse. "Cmpetitin and Price Dispersin in the U.S. Airline Industry." Mime, Brander, J. A. and A. Zhang. "Market Cnduct in the Airline Industry: An Empirical Investigatin." RAND Jurnal f Ecnmics, Vl. 21,1990: Bibli. 1

29 Bresnahan, T. F. "Empirical Studies in Industries with Market Pwer." Handbk f Industrial Organizatin, Vl. II, 1990: Brwn, R. "Measuring 1992's Medium Term Dynamic Effects." NBER Wrking Paper #3166,1989. Brueckner, J. K., and P. T. Spiller. "Cmpetitin and Mergers in Airline Netwrks." Internatinal Jurnal f Industrial Organizatin, Vl. 9, 1991: Brueckner, J. K., N. J. Dyer, and P. T. Spiller. "Fare Determinatin in Airline Hub and Spke Netwrks." RAND Jurnal f Ecnmics, Vl. 23, N. 3, 1992: Brueckner, J. K., and P. T. Spiller. "Fares and Netwrk 'Feed': Estimating Ecnmies f Traffic Density in Airline Hub and Spke Systems." Mime, Buttn, K. Airline Deregulatin: Internatinal Experiences. New Yrk: New Yrk University Press, Calingaert, M. "The 1992 Challenge frm Eurpe: Develpment f the Eurpean Cmmunity's Internal Market." N.P.A. Reprt #237,1988. Captain, P. Cmpetitin and Efficiency in the Eurpean Airline Industry. Unpublished Ph.D. dissertatin, Rice University, Hustn, Carltn, D., W. Landes, and R. Psner. "Benefits and Csts f Airline Mergers: A Case Study." Bell Jurnal f Ecnmics, Vl. 11,1980: Caves, D., L. Christensen, and M. Tretheway. "Ecnmies f Density versus Ecnmies f Scale: Why Trunk and Lcal Service Airline Csts Differ." RAND Jurnal f Ecnmics, Vl. 15,1984: Caves, D., L. Christensen, and E. Diewert. "Multilateral Cmparisns f Output, Input, and Prductivity Using Superlative Index Numbers." Ecnmic Jurnal, Vl. 92,1982: Charnes, A., and W. W. Cper. "Preface t Tpics in Data Envelpment Analysis." Annals f Operatins Research, Vl. 2,1985: Charnes, A., W. W. Cper, and E. Rhdes. "Measuring Efficiency f Decisin Making Units." Eurpean Jurnal f Operatinal Research, Vl. 2, 1978: Christensen, L. R, and D. W. Caves. "Glbal Prperties f Flexible Functinal Frms." American Ecnmic Review, Vl. 70, N. 3,1980: Bibli. 2

30 Crnwell, C, P. Schmidt, and R. C. Sickles. "Prductin Frntiers with Crss Sectinal and Time Series Variatin in Efficiency Levels." jurnal f Ecnmetrics, Vl. 46,1990: Diewert, W. E. "Exact and Superlative Index Numbers." Jurnal f Ecnmetrics, Vl. 4,1976: Dreshler, L. "Weighting f Index Numbers in Multilateral Internatinal Cmparisns." Review f Incme and Wealth, Vl. 19,1973: Emersn, M., et al. The Ecnmics f New Yrk: Oxfrd University Press, Ericsn, R., and A. Pakes. "Empirical Implicatins f Alternative Mdels f Firm Dynamics." Mime, Evans, D. "Tests f Alternative Theries f Firm Grwth." Jurnal f Plitical Ecnmy, Vl. 95,1987: Evans, D. "The Relatinship Between Firm Grwth, Size, and Age: Estimates fr 100 Manufacturing Industries." Jurnal f Industrial Ecnmics, Vl. 35, 1987: Fare, R., et al. "Prductivity Grwth, Technical Prgress, and Efficiency Change in Industrialized Cuntries." Mime, Gng, B. H., and R. C. Sickles. "Finite Sample Evidence n the Perfrmance f Stchastic Frntiers and Data Envelpment Analysis Using Panel Data." Jurnal f Ecnmetrics, Vl. 51,1992: Gd, D., and E. Rhdes. "Prductive Efficiency, Technlgical Change, and the Cmpetitiveness f U.S. Airlines in the Pacific Rim." Jurnal f the Transprtatin Research Frum, Vl. 31,1991: Gd, D., M. I. Nadiri, and R. C. Sickles. "The Structure f Prductin, Technical Change, and Efficiency in a Multiprduct Industry: An Applicatin t U.S. Airlines." NBER Wrking Paper #3939,1992. Gd, D. and R. C. Sickles. "East Meets West: A Cmparisn f the Prductive Perfrmance f Eastern Eurpean and Western Eurpean Air Carriers." Mime, Gd, D., L. H. Reller, and R. C. Sickles. "E.C. Integratin and the Structure f the Franc-American Airline Industries: Implicatins fr Efficiency and Welfare." Mime, Bibli. 3