39
Short-termism in business: causes, mechanisms, consequences
Frame 2
Why do we use econometrics?
Themainreasonsforusingeconometrics
inthisstudyisitsabilitytoassigneconomic
meaningtothecorrelationsfoundinthe
data,thepossibilitytodistinguishshort-term
andlong-termeffects,aswellasthegood
treatmentofunobservablevariablesand
one-offs.Mostoftheseadvantagesarerelated
to the use of panel data ineconometric
modelling–panel data econometrics. The
termpaneldatareferstodatasetsthathave
morethanonedimension,usuallytwo:the
entities(countries,individualsorcompanies)
andthetimeoverwhichtheseentitiesare
observed.Everyvariable(e.g.CEOtenure)
isconsideredinthesedimensions,i.e.we
knowthevalueofsuchavariablebothfor
aparticularcompanyandaparticularyear.
Wediscusstheadvantagesofpaneldata
econometricsbelow.
Seeking economic meaning in correlations
Descriptivestatisticalanalysis,asusedin
thisstudy,saysalotaboutthedata,butit
isusuallyimpropertoinferabouteconomic
relationshipsonthatbasisalone,duetosuch
thingsassectoraldifferencesbetweenrms,
whichalsohavetobetakenintoaccount.
Forinstance,somesectorstendtobemore
capitalintensivethanothers.Therefore,
armfromaheavyindustrialsectorwill
haveadifferent‘natural’levelofinvestment
activitytoacompanyfromtheretailtrade
sector,makingthesetwormsincomparable.
However,ifweanalysebothrmsinrelation
totheirsectoralbenchmarks,wearethen
abletocomparethemwithrespecttotheir
investmentactivity.Tofullyaccountforsuch
kindsofdifferences,itisnecessarytodivide
thermsintoaverylargenumberofgroups,
andthencomparetheirperformance,making
itdifculttodrawgeneralconclusionsfrom
thedataset.Contrarytodescriptivestatistical
analysis,suchproblemscaneasilybesolved
witheconometrics.
Unobservable variables
Somecompanycharacteristicsareeither
unavailableinourdatasetorareimpossibleto
quantifyatall.Ineconometrics,werefer
tosuchvariablesasunobservablevariables.
Inthecaseofthisstudy,agoodexampleis
thecompanycorporateculture,whichcan
bediscussedonlyinqualitativeterms.Aswe
havemorethan1000rmsinoursample,we
cannottellthestoryofthecorporateculture
ofeverysinglecompanywithoutasubstantial
lossofimportantdetail.Thisiswhy,from
thepointofviewofquantitativemethods,
it is sometimes reasonable to say that the
corporatecultureisunobservable.Paneldata
econometricseasilytacklesthisissue.
Toseethemeritsofpaneldatahere,imagine
thatoursamplewaslimitedtojustonesingle
year.Inthatcase,manyoftherelations
apparentinthedatawouldbespurious.For
instance,wecouldconfusethebenetsfrom
aparticularcorporateculture(whichisroughly
constantovertime)withtheimpactofthe
decisionsoftheparticularCEO.However,ifwe
considerthesamecompanyattwodifferent
pointsoftime,wecanobservetheincrement
ofthetenurewiththecorporateculture
roughlyunchanged.Becauseinthepaneldata
wehavebothtimeandindividualdimensions,
wecanstudythedifferencesbetween
companies,aswellasdevelopmentsinasingle
companyovertime.
Short-term vs. long-term effects
Short-termism is essentially the problem of
animbalancebetweenachievinglong-term
andshort-termgoals.Withapaneldataset
givingbothtimeandcompanydimensions,
econometricmethodshelpmakethese
differencesmoreformal–identifyingtheshort-
termandlong-termaspectsofthephenomena
discussedabove.
39
One-offs
Selectingonesingleyearforourstudy
couldbringusincidentalresultsduetosome
random,one-offevents.Aswestudyour
sampleover16years,suchaccidentalfactors
areaveragedoutinthepaneleconometric
models.Thispropertygivesusmoreprecise
estimatesoftheeconomicprocessesapparent
withinoursample.
39
Fortheshort-termeffects,weusepanelregressions,
whereastocapturelong-termrelationsweruncross-
sectionregressions(whichinthiscasearealsopartof
thepaneldataeconometrics).
Theconsequencesofshort-termism