Ready for Take-off? The Economic Effects of Regional

RUHR
ECONOMIC PAPERS
Philipp Breidenbach
Ready for Take-off? The Economic
Effects of Regional Airport Expansion
#549
Imprint
Ruhr Economic Papers
Published by
Ruhr-Universität Bochum (RUB), Department of Economics
Universitätsstr. 150, 44801 Bochum, Germany
Technische Universität Dortmund, Department of Economic and Social Sciences
Vogelpothsweg 87, 44227 Dortmund, Germany
Universität Duisburg-Essen, Department of Economics
Universitätsstr. 12, 45117 Essen, Germany
Rheinisch-Westfälisches Institut für Wirtschaftsforschung (RWI)
Hohenzollernstr. 1-3, 45128 Essen, Germany
Editors
Prof. Dr. Thomas K. Bauer
RUB, Department of Economics, Empirical Economics
Phone: +49 (0) 234/3 22 83 41, e-mail: [email protected]
Prof. Dr. Wolfgang Leininger
Technische Universität Dortmund, Department of Economic and Social Sciences
Economics – Microeconomics
Phone: +49 (0) 231/7 55-3297, e-mail: [email protected]
Prof. Dr. Volker Clausen
University of Duisburg-Essen, Department of Economics
International Economics
Phone: +49 (0) 201/1 83-3655, e-mail: [email protected]
Prof. Dr. Roland Döhrn, Prof. Dr. Manuel Frondel, Prof. Dr. Jochen Kluve
RWI, Phone: +49 (0) 201/81 49 -213, e-mail: [email protected]
Editorial Office
Sabine Weiler
RWI, Phone: +49 (0) 201/81 49 -213, e-mail: [email protected]
Ruhr Economic Papers #549
Responsible Editor: Roland Döhrn
All rights reserved. Bochum, Dortmund, Duisburg, Essen, Germany, 2015
ISSN 1864-4872 (online) – ISBN 978-3-86788-627-7
The working papers published in the Series constitute work in progress circulated to
stimulate discussion and critical comments. Views expressed represent exclusively the
authors’ own opinions and do not necessarily reflect those of the editors.
Ruhr Economic Papers #549
Philipp Breidenbach
Ready for Take-off? The Economic Effects
of Regional Airport Expansion
Bibliografische Informationen
der Deutschen Nationalbibliothek
Die Deutsche Bibliothek verzeichnet diese Publikation in der deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über:
http://dnb.d-nb.de abrufbar.
Das RWI wird vom Bund und vom Land Nordrhein-Westfalen gefördert.
http://dx.doi.org/10.4419/86788627
ISSN 1864-4872 (online)
ISBN 978-3-86788-627-7
Philipp Breidenbach1
Ready for Take-off? The Economic Effects
of Regional Airport Expansion
Abstract
This paper analyzes whether the expansion of regional airports in Germany caused
positive spillover effects on the surrounding economies, exploiting the deregulation
of the European aviation market as a quasi-experiment. Such potential spillovers are
often used as an argument for the substantial annual subsidies to airports. Previous
evaluations often suffer from the problem of reverse causality, since investment
decisions are based on the economic conditions of the region. By contrast, the aviation
deregulation under the Single European Market-initiative provides an exogenous
incentive for investing in the expansion of existing regional airports. A differencein-differences approach is used to estimate the causal effects of this expansion on
regional growth. The results are sobering, though, as there is no evidence for any
positive spillover effects.
JEL Classification: R51, R42, H54
Keywords: Infrastructure investment; regional growth; airport effects
April 2015
1 RWI and RUB – I am grateful to Thomas K. Bauer, Roland Döhrn, Corinna Hentschker, Michael Kind, Alfredo
Paloyo, Christoph M. Schmidt, Torsten Schmidt and Lina Zwick for valuable comments that helped to improve this
manuscript. – All correspondence to: Philipp Breidenbach, RWI, Hohenzollernstr. 1-3, 45128 Essen, Germany, e-mail:
[email protected]
1. Introduction
Duringthepastdecades,Germanregionalairportshaveexpandedtoaremarkableextent.
RegionalpolicymakersinvestedmillionsofEurosinairportfacilities,aimingtoascertainthat
thosefulfilltherequirementsofmodernairportsinaneraofarapidlygrowingaviationmarͲ
ket. Contrary to those ambitious expectations, today nearly all German regional airports deͲ
pendonsubstantialsubsidiestocovertheirannuallosses.SincetheEuropeanUniondecided
thatthesesubsidiesviolateEuropeancompetitionlaw,itwillbeprohibitedafter2024touse
themtocoveroperationallosses.Theselegalrequirementswillcauseexistentialproblemsfor
anumberoftheseregionalairports.
Opponents of the subsidies feel vindicated by this decision, since in their assessment reͲ
gionalairportswillneverfindtheirnichebetweentheestablishedlargeairportsandwillnever
reachprofitability.Proponentsoftheairportsarguethatthenarrowfocusonthedirectlosses
failstorecognizetheirimportanceforregionaldevelopment.Theyemphasizepositivespillover
effects for the surrounding industry, alleging that service industries and highͲtech branches
particularlybenefitfromairportproximity(Sheard,2014,Brueckner,2003andButton/Taylor,
2000).Theargumentofstrongemploymenteffectsisusedtojustifythecontinuedoperation
oftheseairports(Robertson,1995).
Itisdifficulttoanalyzetheseairporteffectsempirically,sincetypicallytheexpansionofreͲ
gionalairportsistheoutcomeofeconomicandpoliticaldeliberationsinarealͲworldcontext,
and not the result of an analytical experiment. Therefore, the question what counterfactual
development an airport region would have realized without the expansion cannot be anͲ
sweredwithease.Onthecontrary,asthemanyattemptsattheeconometricevaluationofthe
(regional) growth effects of infrastructure investments demonstrates (e.g. Aschauer,1989),
theproblemsofreversecausalityandunobserved(regional)heterogeneityarealmostubiquiͲ
tous(e.g.Mukkala/Tervo,2013,Buttonetal.2010andGreen,2007).
ThispaperexploitsanabruptchangeintheregulationofEuropeanaviationthatcanbeinͲ
terpreted as a quasiͲexperiment to overcome this difficulty. Specifically, the deregulation of
theEuropeanaviationmarketin1997ledtoasubstantiallyredesignedaviationmarket(GraͲ
ham,1995),providingaparticularlystrongincentivetoexpandGermanregionalairports.DeͲ
signedtostrengthencompetitionontheairlinemarket,thisreformcausedanincreasingdeͲ
mandfortakeͲoffandlandingslotsatairports.Establishedinternationalairportswerenotable
not serve this increasing demand and, consequently, more airlines turned to operate from
regionalairports.
Contrarytoinvestmentsdrivenbypositiveregionaldevelopments,thisreformthusledto
investments which were set by exogenous changes in the structure of the aviation market
(Graham, 2010, Barrett, 2000) and can therefore be regarded as a (quasiͲ)experiment. The
paper documents that several airports indeed invested extensively to prepare their regional
airportsfortheneedsofmodernandinternationalairlinesandpassengers.Furthermore,the
location of German airports is closely linked to German military history since most of them
Ͷ
wereconvertedfrommilitarytocivilianuseinthepast(Behnen,2004).Theiroriginallocation
followedmilitarystrategiesinsteadofeconomicreasoning(Cidell,2003).Thesecircumstances
facilitate an analysis of exogenous airport expansions on the basis of a differenceͲinͲ
differencesͲapproach(DiD)whichovercomesregionalheterogeneityproblems.
Tothebestofmyknowledge,thisisthefirstpaperscrutinizingtherecentlydiscussedefͲ
fectofregionalairportsexpansionsusinganexogenouseventasidentificationstrategy.2The
resultsaresobering.Theestimationsstartwiththeapplicationofabasicmodel,ignoringany
potentialendogeneityproblems,whichcomparesprosperitylevelsinregionswithandwithout
airports. Such a preliminary approach indeed suggest airport induced regional prosperity.
However,whenendogeneityproblemsaretakenintoaccountbyexploitingtheEuropeandeͲ
regulationof1997,thiseffectvanishes.Thisresultisconfirmedbyvariousrobustnesschecks.
There is simply no evidence that spillovers spreading out from such expansions of regional
airportscouldjustifytheiroverwhelmingsubsidization.
Theremainderofthepaperisorganizedasfollows.Section2summarizestheexistingliterͲ
ature,emphasizingendogeneityissuesandtheinstitutionalsettingsofthederegulation.SecͲ
tion3describesthedatasetandoutlinestheDiD.Theresultsandvariousrobustnesschecks
arepresentedinsection4,and,finally,section5concludes.
2. EUaviationmarketreformassourceofexogenousvariation
2.1. Existingliteratureandconceptualchallenges
ProponentsofregionalairportsarguethatairportsactasadrivingforceofregionaldevelͲ
opment, because airports tend to increase income and employment in the local economy
(ELFAA,2004).Theytypicallyfocusonthreetransmissionchannels(seeButton,2010forfurͲ
thersubdivisions):(i)directeffects,realizedthroughemploymentandinvestmentsattheairͲ
port,(ii)indirecteffects,inthechainofsuppliersofgoodsandservicesrelatedtotheairport,
and (iii)induced effects which comprise the surplus of employment through spending of diͲ
rectlyandindirectlyemployedindividuals.Besidethesethreechannels,airportsaresupposed
tohaveacatalyticeffectbyimprovingproductivityandattractingeconomicandtouristicactivͲ
ities(e.g.EUCommitteeoftheRegions,2004,Cezanne/Mayer,2003).
However,obtainingempiricalevidenceforsuchairporteffectsisachallengingtask.ArguaͲ
bly,thesimplestmethodforexaminingwhethertheexistenceofanairportaffectsgrowthin
the surrounding region – a direct comparison of growth between airportͲ and nonͲairportͲ
regions – will lead to biased estimates, due to omitted regional heterogeneity. Even though
onemightbeabletocontrolforawiderangeofobservablevariables,thisidentificationstratͲ
egywouldstillbepronetosufferingfromunobservableheterogeneity,inparticularembodied
insuchregionalpreͲconditionswhicharelikelytobecorrelatedwiththeexistenceofanairͲ
2
TheUSaviationderegulationact(1978)hasbeenexploitedasexogenousgivenchangesofairservices(BloninͲ
gen/Cristea(2012)butasshownbelow,thisreformdidnotmarkcomparablechangesforregionalairports.
ͷ
port.Sincethelocationofanairportisnotrandom,theremightbehigherprobabilityforthe
erection of an airport in a prospering region. In this context, the application of fixed effects
estimations in panel data frameworks (Islam,1995) is of little help since the existence of an
airportisafixedeffectitself.
Focusingonairportactivities,suchasflights,passengersorcargodevelopment(Floridaet
al., 2012) helps to introduce further variation in the dataset. However, as long as increased
activities do not have an experimental exogenous character, it is not possible to distinguish
whetherimprovingregionalconditionsinfluenceairportactivityasfoundbyGoetz(1992)and
Dobruszkesetal.(2011)orviceversaassuggestedbytheairportproponents.Specifically,this
problemoccursiffutureregionaldevelopmentisanticipatedandairportexpansionsarebased
upon it. In this case, time series analyzes, e.g., provided by Green (2007), Mukkala/Tervo
(2013)orButton/Yuan(2013)cannothelptoidentifyaleadingandafollowingprocess.3
Morepromisingidentificationstrategiescanbederivedfromexogenousevents.Brueckner
(2003) and Sheard (2014)4, for instance, apply the concept of instrumental variables in their
analyses.However,findingexogenouseventsorproperinstrumentswhichareabletopredict
airport size but do not correlate with regional circumstances is a challenging task. BloningͲ
en/Cristea(2012)considertheUS“AviationDeregulationact”5in1978whichwasendorsedto
promotecompetitionintheaviationmarketasanexogenousevent.ActingundermarketpresͲ
sureafterthereform,airlinesfocusedtheiractivitiesonthecentralairportssincesubsidiesfor
peripheralconnectionswerecut.Theseshiftsinitiatedremarkableincreasesanddecreasesin
the various airports’ activities (Burghouwt/Hakfoort, 2001), exogenous from the respective
regional development. Based on the deregulation act, Bloningen/Cristea(2012) observe that
increasingairportactivitiesaffectpopulationgrowth,percapitaincomeandemploymentposiͲ
tively.
Although, their paper offers a promising methodological approach for the evaluation of
generalairporteffectsthesetupislessrelevantfortheparticularexaminationofregionalairͲ
porteffects.Theirstudyisrestrictedtotheexaminationofexogenouslyinducedgrowthofbig
airportsincentralmetropolitanareas.Bydefinition,regionalairports(inthescopeofthispaͲ
per) are rather small and they are located in less central areas. Therefore, effects obtained
3
Beside this, the applied Granger causality tests do only give a hint for potential economic causality (MukkaͲ
la/Tervo,2013).
4
Sheard(2014)exploitstheUS1944NationalAirportPlan,whichmarksanexogenouscomponentofthesizeof
today’sairportswithoutbeingdirectlyinfluencedbythelaterdevelopmentoftheUScities.
5
SincethepreͲreformaviationmarketintheUSdidnotreallyholdasanexampleforanopenmarketbutrather
offeredahighdegreeofgovernmentalregulations,thederegulationactledtosubstantialshiftsinthemarketstrucͲ
ture. The market situation was described by obstacles such as peripheral connections with higher governmental
subsidiesthanattainedticketturnaroundsontheonehandandother'hotconnections'whereonlyalimitednumͲ
ber of flights were allowed although there was a demand for much more flights on the other hand (BloningͲ
en/Cristea,2012).Sincethissetupwasnotsustainablefortherapiddevelopmentoftheaviation,USgovernment
passedaradicalreformofthesystemfor1978.SeeBloningen/Cristea(2012)foradetaileddescriptionofthisreͲ
form.
͸
from the US deregulation act cannot be translated to the expansion of regional airports and
thegrowtheffectsofthesurroundingregions.
2.2. EUaviationmarketreform
EncouragedbytheapparentsuccessofmarketderegulationsintheUS,in1983theEuroͲ
peanCommissionstartedaderegulationinitiative,theSingleEuropeanAviationMarket(GraͲ
ham,1997).ThepostͲreformperiodintheUSdemonstratedhowenhancedcompetitioncould
leadtoanincreaseofsuppliersandflightsand,decreasingprices.ThederegulationoftheEuͲ
ropean market was split into three separate steps. The first two steps were implemented in
1988 and 1990 (Graham,1995) and were characterized by rather small changes such as the
permissionofbilateralintraͲEUagreements,thevalidityofcompetitionrulesfortheaviation
andimplementationofthreeboundedfarezoneswhichallowedticketstobesuppliedbelow
thestandardminimumfares(Schenk2004:95ff).6
Substantialchangesinthestructurewereinitiatedbythethirdstep(Graham,1997).This
included the harmonization of the airline licensing processes, entire liberalization of ticket
faresandtheabolitionofcapacityregulationsbetweenmemberstates.TheinvolvedsuspenͲ
sionofallcabotageͲrestrictions7openedthemarketforarangeofnewairlines(Schenk2004:
98).ThesetrendschangedthemarketdramaticallysinceasubstantialnumberofairlinesenͲ
tered the German market as further competitors, leading to an increased number of flights
(see Thompson, 2002 for France). Nevertheless, slots (for departures and arrivals), ground
operationservicesandbookingsystemsrepresentedabottleneckfortheoperationofincreasͲ
ing traffic. National carriers enjoyed grandfathering rights for the slots and services without
convincing“useitorloseit”rules(Schenk,2004).Thus,theappropriateaccessofnewcompetͲ
itorstotheestablishedairportswashinderedandtheyhadtodiverttheirbusinesstoregional
airports.
TheincreasednumberofcompetitorsandtheshortageofslotsatestablishedairportsproͲ
vided strong incentives to regional policy makers to engage into airport expansion. Graham
(1997) documents substantial benefits for regional airports and for the newly emerging reͲ
gional airlines.8 Furthermore, the reform incentivized regional policy makers to expand the
airportsrightin1997whennewairlinesenteredthemarketandincreasedthedemandforthe
services of regional airports to conduct their operations. Thus, the reform forms a (quasiͲ
)experiment,sincethetimingoftheexpansionswasdeterminedbylegislation,notbyeconomͲ
icconsiderations.Germanyimplementedtheregulationaslateaspossibleinthebeginningof
1997,foryearsafteritpassedEuropeanCouncilin1993.ThislongperiodensuresthatplanͲ
nershadenoughtimetopreparetheairportsfortheincreasingdemandforservices.MoreoͲ
6
SeeGraham(1995/1997)fordetaileddescriptionofthefirsttworeformstepsandSchenk(2004)fortheirconͲ
creteimplementationintheGermancase.
7
CabotagerestrictionsinterdicttheprovisionofanationalroutebynonͲdomesticairlines.Oninternationalroutes
onlythedomesticairlinesfromoneoftheconnectedcountriesisallowedtoprovidetheroute.
8
Section3providesevidenceforactualinvestmentsatGermanregionalairportsinthistime.
͹
ver, preͲreform adjustment of the airͲservices (Ashenfelter’s Dip problem, Ashenfelter 1978)
couldnotoccursincetherestrictiveregulationswerestillintact.
Thescopeofthereformstrengthensitsinterpretationasanexogenouseventwithrespect
to regional development. The European Commission intended to intensify competition beͲ
tweenairlines(Graham,1998),whiletheincentivesforregionalairportexpansionswereonly
sideeffectsofthereform.Thiscontrastswithregionalpolicymeasureswhicharespecifically
designed to compensate for disadvantages of the targeted regions. As Behnen (2004) points
out,thereformdidnotonlycausearevolutionintheskybut,especiallyinGermany,alsoon
theground.
Moreover, the initial location of German regional airports is also quite unrelated to the
economic circumstances surrounding them, since most of the today’s regional airports have
served as military bases in some stages of their existence, before they were converted into
civilianuse(Behnen,2004).Therefore,theirlocationwasnotdrivenbytheeconomicneedsof
aregionorfavorableeconomicpreͲconditions(Cidell,2003),butwasratherbasedonmilitary
strategiesandtheassociateddistributionoftheairforce.
Takentogether,themilitarybackgroundoftheinitiallocationsofregionalairports,theenͲ
compassingscopeofthereformwhichdidnotintendtosupporttheregionalairports,andthe
exogenousimpetusleadingtotheinvestmentsintoregionalairportexpansionformthreearͲ
gumentswhichfacilitatetheexaminationofthecausaleffectsofairportexpansionsonregionͲ
alprosperity.
3.DataandIdentificationStrategy
TheDiDapproachprovidesasoundidentificationstrategyfortheevaluationofthepotenͲ
tialregionalgrowtheffectswhichmighthavebeeninducedbythederegulationoftheaviation
market.Sincethedataset9containsannualinformationonGermancounties(NUTS3level)for
theperiodfrom1991to2008,thepreͲ1997yearsserveaspreͲtreatmentperiodandthelatter
astreatmentperiod.Thedemarcationintotreatmentandcontrolregionsismorechallenging.
Obviously,thoseregionswitharegionalairportareregardedasthetreatmentgroup.Asthe
researchquestionfocussesontheeffectsforregionalairportsanddetailedreactionsofestabͲ
lished(international)airportsonthederegulationremainunclear,suchregionswithinternaͲ
tionalairportsareomittedintheevaluation.10
Forthispurpose,thedistinctionbetweeninternationalandregionalairportshastobedeͲ
fined.ThemaindefinitionofregionalairportsinthispaperreliesonthepreͲreformpassenger
figuresprovidedbytheGermanAirportAssociation(ADV).Allthoseairportswithlessthanone
millionpassengersin1996areincludedinthetreatmentgroupasregionalairports.ThisarbiͲ
9
EconomicvariablesaretakenfromtheFederalInstituteforResearchonBuilding,UrbanAffairsandSpatialDeͲ
velopmentBBR(2009)andBBR(2011).
10
Possiblytheywerealsopositivelyaffectedbythereformandgainedhigherefficiencyoftheirtraffic.
ͺ
trarydefinitionisextensivelytestedintherobustnesscheckswhichcontainvaryingthresholds
from0.5millionto5millionpassengersperyear.Lesspromisingdistinctionsaretheairports’
legal permissions (which are subdivided into international and regional permissions)11 or the
categorization of its members provided by the ADV itself (which suffers from selfͲselection
problems).12GermancountieswithoutaregionaloraninternationalairportserveastheconͲ
trolgroup.
ThegrowthofGDPperlaborforce(GDPpl)isthemostpromisingindicatorwhichcancapͲ
turetherangeofassumedspilloversfromregionalairportstoregionalprosperity.Therefore,
thegrowthofnominalGDPperlaborforceisusedasthemainoutcomevariable.13Sincethe
GDPperlaborforcemayalsohavesomeminorshortcomings14furtherestimationsintheroͲ
bustnesschecksareappliedwiththegrowthoftotalGDP,GDPpercapitaandemploymentas
dependentvariables.
In contrast to the majority of airport evaluations which exploit terms of air services
(Allroggen/Malina,2014),thispaperfocussesonthepureexistenceofinfrastructurecaptured
inacrosssectionaltreatmentdummy(ai)turning1iftheregionhasaregionalairportand0
otherwise.Thetimedummy(tt)indicatesthepostͲtreatmenteraandtheirinteraction(DiDit)
marksthevariableofinterest,theDiDindicator.StartingwithapureDiDasfirstestimation,a
varyingsetofcontrols(xkit)issubsequentlyincludedtoprovideanindicationoftherobustness
oftheestimates.
An increasing set of control variables can account for regional heterogeneity will tend to
improvetheprecisionoftheestimates.However,especiallyinthecontextofregionaleconomͲ
ics, the inclusion of controls is problematic as they might themselves be an outcome of the
treatment (Angrist/Pischke (2008)15, Becker etal.2014). In the context of airport expansions
this is relevant for, e.g., regional investments which might be increased by the expansions.
Developments of these “bad controls” which are driven by the treatment may bias the estiͲ
mated effect of the treatment on the outcome variable, due to the correlation between the
badcontrolandtheoutcomevariable.
The lagged level of population, population density and employment are included as conͲ
trols.Furthermore,thelaggedlevelofthelefthandsidevariable,thelevelofGDPperlabor
force,isincluded.Thisisstandardformodelsbasedonneoclassicaltheorybutposesproblems
11
However, the offered flight destinations do not justify this distinction. International flights are also provided
from those airports which are legally defined as regional airports. See destinations of, e.g., DortmundͲWickede
(http://www.dortmundͲairport.com/f2a0c5cf806929ea/passengersͲvisitors)
12
Therobustnesschecksalsoincludeestimationsdefiningthethelegallydefinedregionalairportsastreatment
group.
13
GDPcanonlybeprovidedoncurrentpricelevelsincethereisnoinformationonpriceindicesforthedeflation
onthechosenlevelofregionalentities(Destatis,2015)
14
Underthestrictassumptionofafixedcapitalstock,theGDPperlaborforcemaydecreasebytheincreaseof
employment.
15
Angrist/Pischke(2008)denotethosecontrolvariablesas“badcontrols”.
ͻ
regardingitsimplicitdynamiccomponent.Sincetheoutcome(gyi,t=ln(yi,t)Ͳln(yi,tͲ1))correlates
withthemodel’serrorterm(ߝit),theregressor(yi,tͲ1) alsocorrelateswiththeerrorterm,leadͲ
ing to biased estimates (see Nickell 1981 or Baltagi2008 for an overview). As Bruno (2005)
shows,thedynamiccorrectedfixedeffectestimatorbasedonaninitialBlundell/Bond(1998)
estimationprovidessatisfactoryresultsforrathershortsamplesperiods.Resultsbasedonthis
estimationmethodarepresentedintheappendix.
Theestimationsarebasedonthefollowingmodel.
K
gyi ,t
G1ai G 2tt G 3 DiDit E1 ln( yi ,t 1 ) ¦ E k ln( xi ,t 1,k ) H i ,t (1)
k 2
wherei=1,…,NisthecrossͲsectionalandt=1,…,Tisthetimedimension,ɴkandɷ1,ɷ2andɷ3are
regression coefficients to be estimated, and ɸi,t is an independent and identically distributed
errorterm(i.i.d.).AlleconomicvariablesaretakenfromBBR(2011).Estimationisbasedonthe
modelbeingtransformedintofirstdifferences,therebyexcludingtheindividualai
'gyi ,t
K
G 2 'tt G 3 'DiDit E1' ln( yi ,t 1 ) ¦ E k ' ln( xi ,t 1, k ) ui ,t (2)
k 2
withuit=ɸi,tͲɸi,tͲ1.
Thehostcountyofanairportmightnotbetherightdelineationforcapturingtheairport’s
economiccontribution.Typically,airportsarenotlocatedinthecenterofcounties;thisespeͲ
cially holds for airports located in bigger cities. They are rather situated in the periphery of
citiesorinadjacentcounties.Thus,simplyconsideringthehostcountiesofairportsasrelevant
regional unit is not appropriate, since their economic effects spread out to adjacent regions
(i.e.,spilloverscrosscountyborders).ThisproblemistackledbydefiningimputedairportreͲ
gions which deliberately construct buffers around the airport’s reference point (see Paloyo
etal.2010fordetails).
Thechosenradiusof15kilometersroughlyrepresentsthemeanradiusofGermancounͲ
ties.Thus,thebuffersizeisrathersmall.16Thepaper concentratesontheregionaleffectsin
the immediate proximity of an airport. Furthermore, since many local municipalities provide
largesharesofthesubsidiesandtheyjustifythisbythepositivespillovers,suchsmallbuffers
are the right demarcation for the objective of this paper. Thus, hinterland effects are intenͲ
tionallydisregardedin thisapproach. WhetherairtransportsupplyhasapositiveoverallimͲ
pactonGermanseconomyisaseparatequestion.
Theeconomiccharacteristicsoftheseairportbuffersaredefinedbythevariablesofthose
administrativecountieswhicharelocatedinthebuffer.Precisely,thebuffervariablesaredeͲ
fined by the withinͲbuffer weighted means, with the respective spatial shares serving as
16
Thefollowingresultsarenotsensitivetoavariationofbuffersizesfrom10to30kilometer(withexceptionof
the30kmͲdefinition,notshowninthepaper,butavailableuponrequest).
ͳͲ
weights.Sincenearlyallairportbuffersconsistofmorethanonecounty(theonlyexceptionis
HannoverͲLangenhagen), the number of observations is much smaller than the number of
regions.17 Although DüsseldorfͲWeeze and Memmingen are regional airports today, they are
ignoredintheempiricalanalyses,sincetheydidnotserveasregionalairportsin1997.Weeze
was opened in 2003 while civilian use in Memmingen started in 2004. All these limitations
reduce the original sample size of 413 German counties to 271 observed regional units. An
overviewontheregionalandinternationalairportsaswellasamapwiththebuffersisprovidͲ
edinFigureA.1intheappendix.
DescriptivestatisticsofthesampleareprovidedinTable1.ItappearsthatregionswithareͲ
gionalairportandthosewithoutthesefacilitiesdonotdiffersubstantially.ThemeanGDPper
laborforcegrowthis0.002percentagepointshigherintheairportregions.18Theairportand
nonͲairportregionsdonotdiffersubstantiallyinalmostallthevariablemeans,asthetͲtestin
thelastcolumndocuments.Onlythelogofpopulationandemploymentdisplayhighermeans
intheairportregions.
Table1:DescriptiveStatisticsforairportandnonͲairportregions
GDP(growth)
GDPpl(growth)
GDPpc(growth)
GDPpl
(x103)
GDPpc
(x103)
ln(employment)
lagged
ln(Population)
Density
(x103)
Airportregions(24)
Mean
Min/
(Std.Dev.) Max
0.030
Ͳ0.041/
(0.039)
0.218
0.027
Ͳ0.036/
(0.037)
0.211
0.033
Ͳ0.040/
(0.039)
0.222
48.725
21.502/
(8.333)
68.455
23.431
7.429/
(6.399)
37.273
4.258
1.807/
(0.884)
5.739
5.007
2.821/
(0.793)
6.525
0.006
0.000/
(0.009)
0.066
NonͲAirportregions
(247)
Mean
Min/
(Std.Dev.) Max
0.028
Ͳ0.260/
(0.039)
0.252
0.025
Ͳ0.262/
(0.039)
0.294
0.029
Ͳ0.241/
(0.040)
0.261
49.804
17.702/
(8.650)
90.099
23.608
6.230/
(9.071)
76.558
4.100
2.915/
(0.540)
6.861
4.893
3.649/
(0.546)
7.184
0.010
0.001/
(0.033)
1.027
tͲstatistic
Ͳ0.505
Ͳ0.621
Ͳ1.065
1.808*
Ͳ0.102
Ͳ9.198***
Ͳ7.904***
2.061**
Note:AllvariablesaretakenfromBBR(2011).***,**,*denotesignificantdifferences
atthe1%Ͳ,5%Ͳand10%Ͳlevel.
17
The buffers are also constructed for the international airports which are ignored in the further course of the
analyses.
18
CloserexaminationoftheminimalgrowthratesinnonͲairportregionsshowthattheseareoutlierswhichare
supposedtobeartificiallyconstructedbycorrectionsoftheofficialstatistics.
ͳͳ
TheinvestmentactivitiesatGermanregionalairportsreflecttheincentiveswhichthedereguͲ
lationoftheaviationmarketprovidedtoregionalpolicymakers.Varioustypesofinvestments
were undertaken; airports expanded their terminals or runways and better connections to
publictransportwereestablished.Inaddition,someairportsprovidedregularairͲserviceswith
scheduledflightsforthefirsttimeafterthereform.Onlyforthreeofthe24regionalairports,
therearenoexplicitinvestmentsfoundfor1997andtheadjacentyears.19
4. Results
Toprovideafirstimpression,resultsfromabasicestimationsetupdemonstratewhether
airportregionsdisplayahigherGDPperlaborforcethannonͲairportͲregions.20Notethatthe
motivationfortheseestimationsismerelytoillustratetheendogeneitybiasofsuchasetupin
comparisontothesubsequentidentificationstrategies,basedonthe(quasiͲ)experiment.FoͲ
cusingonthispreferableidentificationapproach,aDiDmodelisestimatedafterthecommon
trend assumption is tested. These baseline estimations are followed by various robustness
checkstoconsiderthreecrucialissuesoftheidentification:First,theAshenfelter’sDipprobͲ
lem, second the sensitivity to changes of the definition and the spatial demarcation of the
treatmentgroupandthirdchangingdefinitionsofthecontrolgroup.Anumberofsecondary
testsarepresentedintheappendix.
4.1. Estimationresults
ThefirstimpressioninTable2showsthattheexistenceofanairport–unconstrainedon
internationalorregional(column(i)and(ii))–clearlycorrelateswiththelevelofregionalGDP
perlaborforce.Thespecificationincludingcontrolvariables(column(ii))suggeststhattheGDP
per labor force is 3.5% higher compared to regions without an airport. Omitting all internaͲ
tionalairportsfromthesamplereducesthiseffectto1.9%incolumn(iv),althoughitremains
significant.However,thesefindingscanonlydrawanincompletepicture,sincetheestimation
approachdisregardsanyendogeneityproblems.
19
Excludingthesethreeregionsfromtheestimationsdoesnotchangethefollowingresults.
20
IncontrasttosubsequentDiDͲestimationsbasedonthegrowthoftheGDPperlaborforce,thisestimationis
based on the its level values since potential airport induced growth effect faded in after their opening and they
shouldshowupinhigherlevelvaluestoday.
ͳʹ
Table2:PooledͲOLSwithoutDiDforallandregionalairports
DependentVar:
LogofGDPpl
Airport
WestGermany
ln(Population)
ln(PopulationͲ
Density)
StateͲDummies
CountyͲtype
TimeDummies
No.ofRegions
No.ofObs.
Allairports
(i)
(ii)
0.040***
0.035***
(0.009)
(0.004)
0.347***
(0.011)
0.005***
(0.001)
0.036***
(0.003)
N
Y
N
Y
N
Y
287
287
4845
4495
RegionalAirports
(iii)
(iv)
Ͳ0.015
0.019***
(0.009)
(0.004)
0.158***
(0.012)
0.002*
(0.001)
0.032***
(0.003)
N
Y
N
Y
N
Y
271
271
4624
4288
Note: ***,**,* denote significance at the 1%Ͳ, 5%Ͳ and 10%Ͳlevel. Robust standard
errorsclusteredoncountyͲlevelinparentheses.
TosupporttheapplicationoftheDiDapproach,Figure1displaysthegrowthoftheGDP
per labor force separately for treated and nonͲtreated regions. The basic assumption of the
DiDimpliesthatbothgroupswouldhavedevelopedequallyintheabsenceofthetreatment.
Sincethisisanassumptiononthecounterfactualitcannotbestatisticallytested.AvisualinͲ
spection of Figure1 shows only minor level differences in the preͲtreatment development
which can be captured by regional fixed effects (Angrist/Pischke, 2008), therefore the DiD
seemstobeasuitablemethod.
This impression is corroborated by “placebo regressions” with altered definitions of the
treatment in the robustness tests. Note further, that a simple “before and after” analysis of
airportregionsinsteadofaDiDwouldprovidemisleadingresultssinceitonlyfocusesonthe
highergrowthofGDPperlaborforceamongairportregionsafterthereformwithoutaccountͲ
ingforthequitesimilardevelopmentofthecontrolregionsovertime.
ͳ͵
Figure1:CTA,TreatmentandnonͲTreatmentgrowthrate
(AnnualGDPperlaborforcegrowthrate)
ThemainresultsinTable3donotshowastatisticalsignificantexpansioneffectofthereͲ
gionalairportsongrowth.Thevariableofinterest(DiDͲestimator)whichaccountsforreformͲ
inducedgrowthremainsinsignificant.Thegrowingnumberofcontrolvariablesdoesnothave
anyinfluenceonthestatisticalsignificanceoftheDiDcoefficient.Theseresultsareconfirmed
byrathersimilarresultsinTableA.1whichconsiderthedynamiccorrectionofthefixedeffect
modelasproposedbyBruno(2005).SincethedynamiccorrectioninitiallyatartsfromaBlunͲ
dellͲBond (1998) estimation which also faces some shortcomings, e.g., potentially imprecise
estimators in a rather small crossͲsectional (Bruno 2005), and since the dynamic component
seemsnottobiasstandardfixedeffectestimations,thefurtherestimationsarebasedonthe
standardfixedeffectmodel.
ͳͶ
Table3:DiDͲFixedeffectsforregionalairports
Dep.
Variable:
GDPplgrowth
DiDͲestimator
PostͲTreat
ln(GDPpl)
lagged
ln(employment)
lagged
ln(Population)
lagged
ln(Density)
lagged
Constant
TimeDummies
No.ofRegions
No.ofObs.
(i)
Ͳ0.004
(0.008)
Ͳ0.023***
(0.002)
(ii)
Ͳ0.004
(0.004)
0.004***
(0.001)
Ͳ0.160***
(0.008)
0.092***
(0.017)
(iii)
Ͳ0.003
(0.005)
0.002
(0.002)
Ͳ0.149***
(0.009)
0.054**
(0.022)
0.084***
(0.026)
Ͳ0.000
(0.001)
(iv)
Ͳ0.002
(0.003)
0.003
(0.002)
Ͳ0.295***
(0.012)
Ͳ0.079***
(0.022)
0.126***
(0.032)
Ͳ0.001**
(0.001)
0.042***
(0.002)
N
271
4352
0.261***
(0.076)
N
271
4352
Ͳ0.035
(0.105)
N
271
4288
0.844***
(0.134)
Y
271
4288
Note:***,**,*denotesignificanceatthe1%Ͳ,5%Ͳand10%Ͳlevel.RobuststandarderͲ
rorsclusteredoncountyͲlevelinparentheses.
4.2. RobustnessTests
Potential weaknesses of the main results are tested in the following robustness checks.
First, the estimations of a DiD might be biased since reform effects are anticipated and preͲ
reformadjustmentstothenewscenerytakeplace(Ashenfelter’sDip).AlthoughairportplanͲ
nersdidanticipatethereformtheycouldnotbenefitfrompreͲreformadjustmentssincethe
market situation could not change notably before the deregulation. Therefore, an earlier inͲ
vestment did not lead to advantages in the preͲreform period. Nevertheless, construction
measures which were required before the deregulation to provide adequate infrastructure
maybiastheresults.Thisbiasmaybetwofold,ontheonehandreformͲinducedconstructions
mayhavecausedpreͲreformgrowthandthereforebiastheexpansioneffectdownwards.On
theotherhand,airportoperationsmayhavebeenreducedduetosuchconstructionsinpreͲ
reformyearsleadingtoanupwardbiasoftheestimatedexpansioneffect.
TheAshenfelter’sdipproblemconcernsthepreͲreformyears.Sincetheexpansioneffects
may have taken time to spread out and to attract airport activities (and since some investͲ
mentswerenotcompletedintime)thefirsttwopostͲtreatmentyearsareexcludedinfurther
estimations.TheresultsofthetwostrategiesarepresentedinTable4,column(i)and(ii)focus
ontheAshenfelter’sdipandexcludetheyears1996and1995,thelattercolumnsexcludethe
years 1997 and 1998. Table 4 shows that the estimated DiD coefficients remain insignificant
whenexcludingtheseyears.
ͳͷ
Table4:DiDͲFixedeffectswithexcludedyears
(DependentVariable:GrowthofGDPperlaborforce)
Exclusionof…
DiDͲestimator
PostͲTreat
ln(GDPpl)
lagged
ln(employment)
lagged
ln(Population)
lagged
ln(Density)
lagged
Constant
TimeDummies
No.ofGroups
No.ofObs.
1996
Ͳ0.005
(0.004)
0.009***
(0.003)
Ͳ0.296***
(0.013)
Ͳ0.082***
(0.024)
0.132***
(0.033)
Ͳ0.001**
(0.001)
0.824***
(0.137)
Y
271
4024
1995&1996
Ͳ0.007
(0.005)
0.019***
(0.004)
Ͳ0.305***
(0.013)
Ͳ0.077***
(0.026)
0.133***
(0.035)
Ͳ0.001*
(0.001)
0.830***
(0.140)
Y
271
3760
1997
Ͳ0.002
(0.003)
0.016***
(0.003)
Ͳ0.293***
(0.012)
Ͳ0.080***
(0.024)
0.128***
(0.033)
Ͳ0.002***
(0.001)
Ͳ0.293***
(0.012)
Y
271
4021
1997&1998
Ͳ0.002
(0.003)
0.010***
(0.003)
Ͳ0.291***
(0.012)
Ͳ0.075***
(0.024)
0.134***
(0.032)
Ͳ0.002***
(0.001)
Ͳ0.291***
(0.012)
Y
271
3754
Note:***,**,*denotesignificanceatthe1%Ͳ,5%Ͳand10%Ͳlevel.Robuststandarderrorsclusteredon
countyͲlevelinparentheses.
Assecondrobustnesscheck,thetreatmentgroupisvariedbythedefinitionofthethreshͲ
old between regional and international airports. Since the assumption to treat all airports
(listedbytheADV)withatmostonemillionpassengersperyearasregionalairportsmightbe
crucial, this threshold is varied from 0.5 million to 5 million passengers in columns (i)Ͳ(v) in
Table5.Sincethoseairportregionswithpassengernumbersabovethethresholdareexcluded
from the estimations, the number of included regions and observations increase with an inͲ
creasing threshold. The last column restricts the treatment group on the legally defined reͲ
gional airports independent of the respective passenger numbers.21 The results in Table 5
showthattheDiDcoefficientisrobusttothedifferentdefinitionsofthetreatmentgroup.
21
TheestimationsinTable4coverthesamecontrolsascolumn(iv)inTable3.Thus,resultsofcolumn(ii)(1milͲ
lionpassengers)equalscolumn(iv)inTable3.
ͳ͸
Table5:DiDͲFixedeffectsforregionalairports
Dep.Variable:
GDPplgrowth
DiDͲestimator
PostͲTreat
ln(GDPpl)
lagged
ln(employment)
lagged
ln(Population)
lagged
ln(Density)
lagged
Constant
TimeDummies
No.ofRegions
No.ofObs.
TreatmentͲThreshold:Max.passengersperyear
0.5million 1million
2million
2.5million 5million
Ͳ0.002
Ͳ0.003
Ͳ0.003
Ͳ0.004
Ͳ0.004
(0.003)
(0.003)
(0.003)
(0.003)
(0.003)
0.009***
0.003
0.003
0.003
0.009***
(0.003)
(0.002)
(0.002)
(0.002)
(0.003)
Ͳ0.296*** Ͳ0.296*** Ͳ0.294*** Ͳ0.296*** Ͳ0.296***
(0.012)
(0.012)
(0.012)
(0.012)
(0.012)
Ͳ0.082*** Ͳ0.082*** Ͳ0.080*** Ͳ0.082*** Ͳ0.081***
(0.022)
(0.022)
(0.022)
(0.022)
(0.022)
0.131***
0.131***
0.130***
0.132***
0.132***
(0.031)
(0.031)
(0.031)
(0.031)
(0.031)
Ͳ0.001**
Ͳ0.001**
Ͳ0.001**
Ͳ0.001**
Ͳ0.001**
(0.001)
(0.001)
(0.001)
(0.001)
(0.001)
0.834***
0.837***
0.830***
0.830***
0.824***
(0.133)
(0.133)
(0.132)
(0.132)
(0.131)
Y
Y
Y
Y
Y
270
271
276
278
279
4288
4320
4368
4400
4416
Legally
defined
Ͳ0.005
(0.004)
0.009***
(0.003)
Ͳ0.294***
(0.012)
Ͳ0.079***
(0.023)
0.131***
(0.031)
Ͳ0.001**
(0.001)
0.810***
(0.131)
Y
264
4160
Note:***,**,*denotesignificanceatthe1%Ͳ,5%Ͳand10%Ͳlevel.Robuststandarderrorsclusteredon
countyͲlevelinparentheses.
Furthermore,thesizeofthebufferswhichmarkthetreatedregionsaroundtheairportsis
varied.TheradiusofthetreatmentͲbufferisdoubled(30km)toobservepotentialeffectsina
larger group of regions. Based on this enlarged buffer size, the DiD coefficient turns slightly
negative (presented in column (i) of Table6). In addition, one might be concerned that the
stable unit treatment value assumption (SUTVA) is violated in the estimations setup. If the
regions within treatment buffer affect the outcome of the adjacent control regions by some
negativeorpositivespillovers,thecoefficientofthetreatmentestimatorisbiased.
To overcome this problem, again the 15 km buffers are considered as treatment group.
But,topreventadirecttransitionfromtreatmenttocontrolgroup,allregionswithinthe30
kmbufferbutoutsidethe15kmbufferareignoredandneitherconsideredasmembersofthe
treatment nor the control group. Thus, spillovers from treated regions cannot directly influͲ
ence the control group (column (ii) of Table6). This estimation does not show a significant
result.
ͳ͹
Table6:DiDͲFixedeffectsvariedbuffersize
(DependentVariable:GrowthofGDPperlaborforce)
(i)
(ii)
Excludedring
30kmbuffer from15Ͳ30km
Ͳ0.005*
Ͳ0.005
(0.003)
(0.003)
0.005*
0.012***
(0.003)
(0.003)
Ͳ0.295***
Ͳ0.004
(0.012)
(0.005)
Ͳ0.080***
Ͳ0.150***
(0.022)
(0.013)
0.127***
0.054
(0.031)
(0.033)
Ͳ0.001**
0.087**
(0.001)
(0.035)
0.849***
0.814***
(0.135)
(0.200)
Y
Y
271
154
4288
2426
DiDͲestimator
PostͲTreat
ln(GDPpl)
lagged
ln(employment)
lagged
ln(Population)
lagged
ln(Density)
lagged
Constant
TimeDummies
No.ofGroups
No.ofObs.
Note:***,**,*denotesignificanceatthe1%Ͳ,5%Ͳand
10%Ͳlevel.RobuststandarderrorsclusteredoncountyͲ
levelinparentheses.
Third,onemightbeconcernedthatthebroaddelineationofthecontrolgroupwhichcomͲ
prises all German regions without an airport is inappropriate. As discussed in section 2, airͲ
portsmightbelocatedinregionswithspecificeconomiccharacteristics.Hence,ifthesecharͲ
acteristics determine the probability for having an airport on the one hand and the growth
expectations over the treatment period on the other hand, the empirical strategy, may lead
astray.Ineconometricterms,thecharacteristicsofthecontrolvariablesmaynotoverlapbeͲ
tweenthetreatmentregionsandacriticalmassofcontrolregionsinthesample.Accordingto
thisconcern,controlregionswhichdiffersubstantiallyfromtheairportregionsintheirregionͲ
alcharacteristicswouldhavetobeexcludedfromtheestimation.
In the spirit of a matching approach, a propensity score is estimated based on a probit
model.
K
pi
¦E
k
ln( zi ) vi
(3)
k 1
Onthelefthandside,thedummypiindicatesifaregionhasanairport(pi=1)ornot(pi=0).On
therighthandside,ziindicatesaverybroadsetofregionalcontrols, ߚkaretherelatedcoeffiͲ
ͳͺ
cientstobeestimatedand,ʆiisani.i.d.errorterm.Sincevariationovertimedoesnotreally
offerfurtherinformation,themodelisestimatedwithobservationsoftheyear1996,thelast
preͲreformobservation.Theestimatedoutcomes( p̂ )indicatearegionalprobabilityforhostͲ
inganairport.Therangeoftheestimatedairportprobabilities( p̂ )oftheairportregionsdeͲ
cides which regions enter the control group.22 Only those nonͲairport regions with a
( p̂ ) within the range of ( p̂ ) of airport regions are considered in the control group. Regions
withanexceptionallylowprobabilityforhavinganairportareexcluded.
Table7:ProbitfortheAirportprobability
(DependentVariable:AirportDummy)
DependentVar.:
AirportDummy
ln(GDP)
ln(GDPpw)
ln(GDPpl)
Positivemigration
ln(Population)
ln(Density)
WestGermany
Constant
CountyTypes
FederalStateDummies
Lowest p̂ ofapt.region
Highest p̂ ofapt.region
No.ofObs
(i)
20.465**
(8.762)
Ͳ1.935
(1.981)
Ͳ16.391*
(8.529)
1.046**
(0.458)
Ͳ0.564**
(0.257)
Ͳ19.591**
(8.785)
Ͳ0.205
(1.121)
Ͳ16.706***
(5.971)
Y
Y
0.015
0.975
287
Note: ***,**,* denote significance at the 1%Ͳ,
5%Ͳ and 10%Ͳlevel. Robust standard errors in
parentheses.
AsTable7shows,therangeofpredictedairportprobabilities( p̂ )forairportregionscover
aspectrumfrom1.5%to99.8%.Thisresultsuggeststhatthereisbasicallynoregionwhichhas
22
Togiveanexample,ifthelowestpredictedairportprobabilityofanairportregionis50%,allthosecontrolreͲ
gionswithaprobabilitybelow50%areexcludedfromthefollowingDiDestimation.
ͳͻ
norealisticprobabilityforhostinganairport.Thus,thebroadselectionofthecontrolgroups
posesnoproblems.23Thisfindingsupportstheargumentthatthemilitarybackgroundofmost
airportsismuchmoreimportantforairportlocationthananyeconomicconsiderations.
Furtherrobustnesschecksareprovidedintheappendix.TableA.2presentstwo“placebo
regressions”whichrefertoahypotheticalderegulationeventduringthepreͲtreatmentperiod
ignoring all postͲtreatment observations and another hypothetical deregulation event during
the postͲtreatment period ignoring all preͲtreatment observations. While an artificial treatͲ
ment during the preͲtreatment period strengthens the previously discussed common trend
assumption, the latter artificial treatment can rather indicate effects of the treatment which
fadeinafteracertaintime.Theyears1994and2000arechosenforthehypotheticalreform.
Inbothcases,thecoefficientoftheDiDremainsinsignificant.
Toensurethattheseresultsarenotbasedonthelowerproductivityofnewjobscausedby
theairportexpansion,regressionswiththegrowthoftotalGDPgrowth,GDPpercapitagrowth
andgrowthoftheemploymentasdependentvariablesareappliedandreportedinTableA.3
(column(i)Ͳ(iii)).Noneofthesechangesleadstodivergingfindings.Toavoidthedeterioration
oftheaviationmarketafter9/11,thesampleperiodends2001incolumn(iv)whichdoesnot
changetheobservedresults.
5. Conclusion
The EU Commission has recently announced that subsidization of airports which merely
survive due to substantial public support will be prohibited after 2024. Since most of the
smallerandregionalairportsinGermanyarecurrentlysubsidized,theyarefacingsevereprobͲ
lemsfortheirfutureexistence.Proponentsoftheregionalairportsemphasizetheimportance
ofpositivespilloversonemploymentandeconomicgrowththroughouttheregion.Thispaper
probes if German regional airports indeed generate a better economic performance in their
environment.
Investments in infrastructure such as airports are an outcome of economic performance
and future economic expectations and, correspondingly, evaluations suffer from the related
endogeneityproblems.ThederegulationoftheEuropeanaviationmarketmarksanexogenous
event which can be seen as a (quasi)Ͳexperiment for the expansion of regional airports. An
increasingnumberofairlinesdemandedfurtheroperationslotsinGermanyandshiftedtothe
regionalairports.Andthoseairportspreparedtheirinfrastructurewithmassiveinvestmentsto
fulfill the airlines’ requirements, quite independent from the contemporaneous economic
conditions. Furthermore,themilitary background ofmostregionalairportsmakestheir locaͲ
tionlessdependentontheeconomicconditionsintheirenvironment.
Based on the assumed exogeneity of the expansion, a DiD is applied which conducts the
period after the reform in 1997 as treatment period. For the spatial definition of the treatͲ
23
TheassociatedDiDwithouttheregionswithaprobabilitybelow1.5%isnotshownhere.
ʹͲ
ment,regionalbufferswitharadiusof15kmareconstructedaroundtheairports.Ignoringthe
possible endogeneity problems would lead to a positive estimation of airport effects on the
GDP per labor force. This appears to drive earlier positive findings such as Allroggen/Malina
(2014). However, when taking into account the preferred deregulationͲbased identification
strategy,theestimatedeffectsarenegligible.
A broad set of robustness checks strengthens these findings. The results are robust to a
change of the definition of regional airports, various demarcations of the control and treatͲ
mentgroups,avoidanceofanAshenfelter’sdipbytheexclusionofyearsaroundthederegulaͲ
tion,andchangesofthedependentvariable.Areasonforthesesoberingresultsmightbethe
overwhelmingopportunitycostsoftheairportoperation.Sincemanymunicipalitiesspenthigh
amountsintheoperationoftheairports,thiscapitalistiedupbytheairportandother–posͲ
sibly better investments – are precluded. This overall result is supported by a recently pubͲ
lishedreportoftheEuropeanCourtofAuditorsthatdetectedsevereunsuccessfulsubsidized
airportͲprojects in southern European countries with sobering costͲbenefit relations (EUCA
21/2014).
NotethattheevidencepresentedheredoesnotsuggestanyconclusionsregardingtheefͲ
fects of international airports. It may be the case that airports need to exceed a certain
thresholdtofadeoutspillovers.Furthermore,thehighdensityofairportsinGermanymaybe
areasonfortheresults,sincefurtherbenefitsofanexpandedairportmightberatherlowina
denseairportnetwork.Besidestheadvancesoftheappliedidentificationstrategy,thismight
beafurtherexplanationforthedifferingfindingsinthispapercomparedtoexistingliterature
forothercountries.
Basedontheeconomiceffects,thispaperdoesnotconfirmconcernsofregionalpoliticians
thatregionswillsufferonceairportsubsidieswillbecut.Sincetheexpansionofairportshad
nopositiveeffects,adownscalingofactivitiestowardssustainableairportswithoutsubsidiesis
not supposed to have major negative effects on the surrounding municipalities. In addition,
theprovisionofbetterregionalairtransportinfrastructuredoesnotseemtobeapromising
instrumenttostimulategrowthinlaggingregions.
ʹͳ
References
Allroggen,F.,&R.Malina,2014.Dotheregionalgrowtheffectsofairtransportdifferamong
airports?.JournalofAirTransportManagement,37,1Ͳ4.
Angrist,J.D.,&J.S.Pischke,2008.Mostlyharmlesseconometrics:Anempiricist'scompanion.
Princetonuniversitypress.
ArbeitsgemeinschaftDeutscherVerkehrsflughäfen(ADV),2015,Verkehrszahlen–Archiv.
Onlineavailable:http://www.adv.aero/verkehrszahlen/archiv/
Aschauer,D.A.,1989.Ispublicinfrastructureexpenditureproductive?.JournalofMonetary
Economics,23,pp.177–200.
Augurzky,B.,Bauer,T.K.,&Schaffner,S.(2006).CopaymentsintheGermanHealthSystemͲ
DoTheyWork?.RWIDiscussionPapers:43.
Baltagi,B.(2008).Econometricanalysisofpaneldata(Vol.1).JohnWiley&Sons.
Barrett,S.D.,2000.AirportcompetitioninthederegulatedEuropeanaviationmarket.Journal
ofAirTransportManagement,6(1),13Ͳ27.
BBR(FederalOfficeforBuildingandRegionalPlanning),2011.LaufendeBeobachtungendes
BundesinstitutsfürBauͲ,StadtͲundRaumforschung,variousissues.Bundesamtfür
BauwesenundRaumordnung,Bonn.
BBR(FederalOfficeforBuildingandRegionalPlanning),2009.Siedlungsstrukturelle
Kreistypen.Publishedonline:http://www.bbsr.bund.de.
Becker,S.O.,Boeckh,K.,Hainz,C.,&L.Woessmann,2014.TheEmpireIsDead,LongLivethe
Empire!LongͲRunPersistenceofTrustandCorruptionintheBureaucracy.TheEconomic
Journal.(onlinefirst)
Blonigen,BruceA.&Cristea,AncaD.,2012.AirportsandUrbanGrowth:Evidencefroma
QuasiͲNaturalPolicyExperiment(No.w18278).NationalBureauofEconomicResearch.
Brueckner,J.K.2003.Airlinetrafficandurbaneconomicdevelopment.UrbanStudies40(8):
1455Ͳ1469.
Bruno,G.S.,2005.EstimationandinferenceindynamicunbalancedpanelͲdatamodelswitha
smallnumberofindividuals.StataJournal,5(4),473.
Burghouwt,G.,J.Hakfoort,2001.TheevolutionoftheEuropeanaviationnetwork,1990–
1998,JournalofAirTransportManagement,Volume7,Issue5,Pages311Ͳ318
Button,K.J.,Doh,S.andYuan,J.,2010.Theroleofsmallairportsineconomicdevelopment,
JournalofAirportManagement,4,pp.125–136.
Button,K.&Taylor,S.,2000.Internationalairtransportationandeconomicdevelopment.
JournalofAirTransportManagement6:209Ͳ222.
Button,K.,&Yuan,J.,2013.Airfreighttransportandeconomicdevelopment:anexamination
ofcausality.UrbanStudies,50(2),329Ͳ340.
ʹʹ
Cezanne,W.,andA.Mayer,2003.FlugplatzalsWirtschaftsfaktorͲMöglichkeitenundFormen
derLandeplätzealsWirtschaftsfaktor,AnalysederimLandvorhandenenSituation.
CommitteeoftheRegions,2004.OwnͲinitiativeopinionoftheCommitteeoftheRegionson
‘LowͲcostairlinesandterritorialdevelopment’.OfficialJournaloftheEuropeanUnion.
(2004/C318/02).Brussels
Cidell,J.,2014.Theroleofmajorinfrastructureinsubregionaleconomicdevelopment:an
empiricalstudyofairportsandcities.JournalofEconomicGeography
Cidell,J.,2003.Theconversionofmilitarybasestocommercialairports:existingconversions
andfuturepossibilities.JournalofTransportGeography11,93–102.
Destatis,2015.RegionaldatenbankDeutschland.ErläuterungBruttoinlandsprodukt,
Bruttowertschöpfung.PublishedOnline:https://www.regionalstatistik.de/genesis/online
EuropeanLowFaresAirlineAssociation,2004.LiberalizationofEuropeanAirTransport:The
BenefitsofLowFaresAirlinestoConsumers,Airports,RegionsandtheEnvironment.
Brussels
GoetzA.R.1992.AirpassengertransportationandgrowthintheU.S.urbanSystem,1950Ͳ
1987.GrowthandChange23:218Ͳ242.
Graham,B.,Guyer,C.,1999.Environmentalsustainability,airportcapacityandEuropeanair
transportliberalization:irreconcilablegoals?.JournalofTransportGeography7,165–180.
Graham,B.,1998.Liberalization,regionaleconomicdevelopmentandthegeographyof
demandforairtransportintheEuropeanUnion.JournalofTransportGeography,6(2),
87Ͳ104.
Graham,B.,1997.RegionalairlineservicesintheliberalizedEuropeanUnionsingleaviation
market.JournalofAirTransportManagement,3(4),227Ͳ238.
GrahamB.,1995.GeographyandAirTransport.Wiley,Chichester.
Green,R.2007.Airportsandeconomicdevelopment.RealEstateEconomics35:91Ͳ112.
FrédéricDobruszkes,MoritzLennert,GillesVanHamme,2011.Ananalysisofthe
determinantsofairtrafficvolumeforEuropeanmetropolitanareas,JournalofTransport
Geography,Volume19,Issue4,Pages755Ͳ762
Mawson,J.,1997.AirtransportliberalizationintheEuropeanUnion:anassessment.Regional
Studies,31(8),807Ͳ812.
MukkalaK.,TervoH.,2013.Airtransportationandregionalgrowth:whichwaydoesthe
causalityrun?.EnvironmentandPlanningA45(6)1508–1520
Nickell, S. (1981). Biases in dynamic models with fixed effects. Econometrica: Journal of the
EconometricSociety,1417Ͳ1426.
Nijkamp,P.,Geenhuizen,M.,1997.Europeantransport:challengesandopportunitiesfor
futureresearchandpolicies.JournalofTransportGeography5,4Ͳ11.
ʹ͵
Paloyo,A.R.,Vance,C.,&Vorell,M.2010.Theregionaleconomiceffectsofmilitarybase
realignmentsandclosuresinGermany.DefenceandPeaceEconomics,21(5Ͳ6),567Ͳ579.
Robertson,J.A.W.1995.Airportsandeconomicregeneration,JournalofAirTransport
Management,Volume2,Issue2,June1995,Pages81Ͳ88
Sheard,N.,2014.Airportsandurbansectoralemployment.JournalofUrbanEconomics,80,
133Ͳ152.
Thompson,I.,2002.Airtransportliberalisationandthedevelopmentofthirdlevelairportsin
France,JournalofTransportGeography,Volume10,Issue4,December2002,Pages273Ͳ
285
VanPraag,B.M.S.andB.E.Baarsma,2005.UsingHappinessSurveystoValueIntangibles:
TheCaseofAirportNoise.TheEconomicJournal,115:224–246.
Vickerman,R.,Spiekermann,K.,Wegener,M.,1999.Accessibilityandeconomicdevelopment
inEurope.RegionalStudies33:1Ͳ15.
ʹͶ
TableA.1:CorrectedDiDͲFixedeffects(Bruno,2005)
Dep.
Variable:
GDPpllevel
DiDͲestimator
PostͲTreat
ln(GDPpl)
lagged
ln(employment)
lagged
ln(Population)
lagged
ln(Density)
lagged
TimeDummies
No.ofRegions
No.ofObs.
(ii)
Ͳ0.006
(0.005)
0.008***
(0.002)
0.899***
(0.005)
0.109***
(0.01“)
N
271
4352
(iii)
Ͳ0.003
(0.005)
Ͳ0.011***
(0.002)
0.916***
(0.006)
0.023
(0.017)
0.175***
(0.025)
Ͳ0.000
(0.001)
N
271
4288
(iv)
Ͳ0.003
(0.004)
0.003
(0.003)
0.761***
(0.008)
Ͳ0.084***
(0.018)
0.196***
(0.025)
Ͳ0.001
(0.001)
Y
271
4288
Note:***,**,*denotesignificanceatthe1%Ͳ,5%Ͳand10%Ͳlevel.RobuststandarderͲ
rorsclusteredoncountyͲlevelinparentheses.
ʹͷ
TableA2:PlaceboRegressions(DiDͲFixedeffects)
(DependentVariable:GrowthofGDPperlaborforce)
DiDͲestimator
PostͲTreat
ln(GDPpl)
lagged
ln(employment)
lagged
ln(Population)
lagged
ln(Density)
lagged
(i)
Hypothetical
Treatmentin1994
(excludingt>1996)
Ͳ0.007
(0.006)
0.010***
(0.004)
Ͳ0.434***
(0.021)
Ͳ0.097
(0.060)
0.672*
(0.388)
Ͳ0.620
(0.418)
(ii)
Hypothetical
Treatmentin2000
(excludingt<1997)
0.001
(0.003)
0.058***
(0.004)
Ͳ0.327***
(0.020)
Ͳ0.056**
(0.028)
Ͳ0.004
(0.042)
Ͳ0.001
(0.001)
Constant
TimeDummies
No.ofGroups
No.ofObs.
4.779
(4.314)
Y
271
1056
1.531***
(0.166)
Y
171
3232
Note: ***,**,* denote significance at the 1%Ͳ, 5%Ͳ and 10%Ͳlevel.
RobuststandarderrorsclusteredoncountyͲlevelinparentheses.
ʹ͸
Ͳ0.005
(0.005)
0.008***
(0.003)
Ͳ0.290***
(0.014)
Ͳ0.220***
(0.025)
0.283***
(0.037)
Ͳ0.001*
(0.001)
0.627***
(0.148)
Y
271
4288
Dep.Var.GDPpc
growth
Dep.Var.GDP
growth
Ͳ0.004
(0.005)
0.008***
(0.003)
Ͳ0.293***
(0.014)
Ͳ0.202***
(0.025)
0.219***
(0.039)
Ͳ0.001*
(0.001)
0.883***
(0.152)
Y
271
4288
(ii)
(i)
(iii)
Dep.Var.
Employment
growth
Ͳ0.002
(0.003)
0.005***
(0.001)
0.002
(0.006)
Ͳ0.122***
(0.010)
0.092***
(0.016)
0.000
(0.000)
0.039
(0.067)
Y
271
4288
Ͳ0.001
(0.003)
0.020***
(0.004)
Ͳ0.377***
(0.016)
Ͳ0.103***
(0.033)
0.227***
(0.051)
Ͳ0.003***
(0.001)
0.736***
(0.237)
Y
272
2384
Skippedafter
2001
(iv)
(vi)
BiggestPassenͲ
RegionalAirports
gersurplus(10
legallydefined
Apt)
Ͳ0.005
Ͳ0.011***
(0.004)
(0.003)
0.009***
0.009***
(0.003)
(0.003)
Ͳ0.294***
Ͳ0.293***
(0.012)
(0.013)
Ͳ0.079***
Ͳ0.079***
(0.023)
(0.023)
0.131***
0.128***
(0.031)
(0.032)
Ͳ0.001**
Ͳ0.002***
(0.001)
(0.001)
0.810***
0.816***
(0.131)
(0.134)
Y
Y
264
257
4160
4072
(v)
Note:***,**,*denotesignificanceatthe1%Ͳ,5%Ͳand10%Ͳlevel.RobuststandarderrorsclusteredoncountyͲlevelinparentheses.
Dep.Variable:
GDPplgrowth(or
seecolumn
headline)
DiDͲestimator
PostͲTreat
ln(GDPpl)
lagged
ln(employment)
lagged
ln(Population)
lagged
ln(Density)
lagged
Constant
TimeDummies
No.ofGroups
No.ofObs.
TableA.3:Variousrobustnesscheck(DiDͲFE)
FigureA.1:Airportandtherespectivebufferregions
Altenburg
Augsburg
Bayreuth
BraunschweigͲWolfsburg
Dortmund
ErfurtͲWeimar
FrankfurtͲHahn
Friedrichshafen
Hof
Karlsruhe
KasselͲCalden
Kiel
Lübeck
Magdeburg
Mönchengladbach
Münster
Neubrandenburg
Paderborn
Rostock
Saarbrücken
Schwerin
Siegerland
StralsundͲBarth
Zweibrücken
•
•
•
•
•
•
•
•
•
•
•
•
•
BerlinͲSchönefeld
BerlinͲTegel
Bremen
Dresden
Düsseldorf
FrankfurtamMain
Hamburg
Hannover
Köln/Bonn
Leipzig/Halle
München
Nürnberg
Stuttgart
InternationalAirports
RegionalAirports
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•