FRL-助力金融发展：生成式人工智能在跨国融资中的关键作用-Fueling financial development The crucial role of generative AI financing across nations

FinanceResearchLetters72(2025)106519

Contentslistsavailableat

ScienceDirect

FinanceResearchLetters

journalhomepage:

/locate/frl

Fuelingfinancialdevelopment:ThecrucialroleofgenerativeAIfinancingacrossnations

AbuBakkarSiddik

a

,YongLi

a

,AnnaMinDu

b

,*

,MilenaMigliavacca

c

aSchoolofManagement,UniversityofScienceandTechnologyofChina,Hefei,PRChinabTheBusinessSchool,EdinburghNapierUniversity,Edinburgh,UnitedKingdom

cDepartmentofEconomicsandBusinessAdministration,Universita、CattolicadelSacroCuore,Italy

ARTICLEINFO

ABSTRACT

JELClassifications:

O33G21C33L86D83

Keywords:

GAIfinancing

FinancialdevelopmentPublicityexposure

Cross-sectionalanalysis

ThisstudyexaminestheimpactofGenerativeAI(GAI)financingonfinancialdevelopment(FD)across21countriesusingcross-sectionaldatafrom2020to2022.Employingbothsimplelinearregressionandtwo-stageleastsquares(2SLS)toaddressendogeneity,wefindthatGAIfinancingsignificantlycontributestofinancialdevelopment,withstrongereffectsobservedinAsianandnon-Europeanregions.Regionalheterogeneityisevident,highlightingvaryingimpactsacrossdifferentsubcontinents.PolicyimplicationssuggestpromotingGAIecosystems,attractingforeigninvestment,andenhancingpublicityforGAIstartups.ThestudyhighlightstheneedforfutureresearchontheethicalimplicationsanddynamiceffectsofGAIfinancing.

1.Introduction

TheriseofGenerativeArtificialIntelligence(GAI)hasrevolutionizedvariousindustries(

DowlingandLucey,2023

;

Dwivedietal.,

2023

;

Siddiketal.,2024

),transformingthewaybusinessesoperateandimpactingbroadereconomicsystems.AsasubsetofAI,GAIusesmachinelearningmodelstogeneratecontent,includingtext,images,andvideos,mimickinghumancreativityandintelligence(

Chakrabortyetal.,2024

;

Dwivedietal.,2023

;

HermannandPuntoni,2024

).Inrecentyears,GAIfinancing,thefundingallocatedtoGAIstartupsandprojects,hasrapidlyexpandedasinvestorsrecognizethedisruptivepotentialofthesetechnologies.Duringarapidemergence,GAIstartupshaveattractedhugefundingfrominvestors,withover$25Binfunding

1in2023alone.However,thelink

betweenGAIfinancingandfinancialdevelopment(FD),particularlyatamacroeconomiclevel,remainsunderexplored.ThisstudyseekstobridgethisgapbyanalyzinghowGAIfinancinginfluencesfinancialdevelopmentacrossdifferentcountriesandregions.

Financialdevelopmentreferstothegrowthandmaturityoffinancialinstitutionsandmarkets,characterizedbyimprovedaccess,efficiency,anddepth(

AsteriouandSpanos,2019

;

MeniagoandAsongu,2018

;

Nasreenetal.,2020

).Theliteratureonfinancialdevelopmenttypicallyfocusesontraditionalsectors,includingbanking,microfinance,andequitymarkets(

Ashraf,2018

;

Bannaetal.,

2022

;

Mhadhbietal.,2021

;

WuandBowe,2010

).However,theadventofdisruptivetechnologieslikeGAIisreshapingthefinanciallandscape(

Ardekanietal.,2024

;

DowlingandLucey,2023

;

Dwivedietal.,2023

),necessitatingacloserexaminationofhowsuch

*Correspondingauthor.

E-mailaddresses:

ls190309@

,

absiddik.nub@

(A.B.Siddik),

yonglee@

(Y.Li),

a.du@napier.ac.uk

(A.M.Du),

milena.migliavacca@unicatt.it

(M.Migliavacca).

1Retrievedfrom

https://dealroom.co/guides/generative-ai

/10.1016/j.frl.2024.106519

Received6October2024;Receivedinrevisedform14November2024;Accepted24November2024Availableonline26November2024

1544-6123/©2024TheAuthor(s).PublishedbyElsevierInc.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense

(/licenses/by-nc-nd/4.0/

).

A.B.Siddiketal.FinanceResearchLetters72(2025)106519

2

innovationsdrivefinancialgrowth.GAIhasthepotentialtorevolutionizefinancialservices,improvedecision-makingprocesses,andenhancemarketefficiency.Yet,despiteitsgrowingprominence,researchontherelationshipbetweenGAIfinancingandfinancialdevelopmentremainslimited.

Intherapidlyevolvingfinancesector,digitaltechnologiesarereshapingtraditionalmodelsanddrivingimpactfulinnovation.Frommobilebankinganddigitalwalletstofintechandrobo-advisors,digitalfinancehastransformedconsumerbehaviorandredefinedfinancialsystemsglobally(

Congetal.,2024

).AgrowingbodyofresearchemphasizesAI’stransformativepotential:

Ardekanietal.

(2024)

introducedFinSentGPT,asentimentanalysismodelsurpassingtraditionalmethods,while

Panetal.(2024)

demonstratedAI’sroleinstrengtheningregulatoryoversight.

Daníelssonetal.(2022)

alsohighlightAI’spotentialrisks,suchasamplifyingsystemicvulnerabilitiesdespiteitsefficiencies.(

Siddiketal.,2024

)foundthatinvestorinfluencesignificantlyenhancesfundingforGAIstartups,whiletechnologicalinfluenceislimited,underscoringtheimportanceofinvestornetworksinsupportingGAIexpansion.AlthoughthesestudiesillustrateAI’sbroadimpact,theydonotspecificallyaddressGAIfinancing’sroleinadvancingfinancialdevelopment.

DespiteextensiveresearchonAI’simpactonfinancialsystems(

Ardekanietal.,2024

;

Panetal.,2024

;

Sachanetal.,2024

),agapremainsinunderstandinghowGAIfinancingcontributestomacro-levelfinancialdevelopment.MoststudiesfocusonAI’stechnicalapplications,butfewexaminehowthefundingofGAIstartupsaffectsbroaderfinancialgrowth,particularlyindevelopingregions.AsGAItechnologiesbecomeincreasinglyintegratedintofinancialservices(

AlmeidaandGonçalves,2024

;

DowlingandLucey,2023

;

Dwivedietal.,2023

),itisessentialtoassesstheireconomicimpact,especiallyregardingfinancialsystemgrowthandmaturity.Toaddressthisgap,ourstudyinvestigatestheimpactofGAIfinancingonfinancialdevelopmentusingcross-sectionaldatafrom2020to2022.Itexploresthreekeyquestions:(1)HowdoesGAIfinancinginfluencefinancialdevelopmentacrosscountries?(2)WhatroledoespublicityexposureplayinsecuringGAIfinancing?(3)AretheresignificantregionaldifferencesinGAIfinancing’simpactonfinancialdevelopment?ThisstudyaimstoprovideacomprehensiveunderstandingofGAIfinancing’seffectsonfinancialdevelopmentatbothcountryandregionallevels.

ThenoveltyofthisstudyliesinitsfocusonGAIfinancingasadriveroffinancialdevelopment,arelationshiplargelyoverlookedintheliterature.WhilepreviousstudieshaveexaminedAI’sroleinfinancialsystems,theyhavenotaddressedGAIfinancing’smacro-economicimplications.Byanalyzingcross-sectionaldataacrossmultiplecountries,thisstudyoffersauniqueviewonhowGAIstartupscontributetofinancialdevelopment.Additionally,introducingpublicityexposureasaninstrumentalvariableprovidesinsightsintohowmediavisibilityaffectsinvestorconfidenceandfunding.Thestudyalsoexploresregionalheterogeneity,offeringvaluableinsightsforpolicymakerstopromotefinancialgrowthacrossdifferentsubcontinents.

2.Materialsandmethods

2.1.Dataandvariables

WeexaminetheimpactofGAIfinancingonthefinancialdevelopmentof21selectedcountriesusingacross-sectionaldataset.OurmeasureofGAIfinancingisbasedonthetotalfunding(inUSDmillion)receivedbyGAIstartupsestablishedbetween2010and2022,withdatasourcedfromCrunchbase,acomprehensiveplatformprovidinginformationonstartups,investments,andfundingactivities

.

(

LeeandGeum,2023

;

Uddinetal.,2024

;

ZbikowskiandAntosiuk,2021

).WeselectedcountriesthathadatleasttwoGAIstartupswithrecordedfunding,resultinginafinalsampleof384startupsacross21countries.

ToassessGAIfinancing’sinfluenceonfinancialdevelopment,wecollectedfinancialdevelopmentindexdatafromtheInternationalMonetaryFund(IMF)fortheyears2020,2021,and2022.Althoughouranalysisfocusesonthesethreeyears,reflectingthemostrecentandreliabledataavailable,weacknowledgethatthislimitedtimeframemayrestrictinsightsintolonger-termdynamiceffects.However,ourchoiceofacross-sectionalapproach,withdatafrom2020to2022,allowsforafocusedanalysisofshort-termimpactsinarapidlyevolvingtechnologicalcontext.Additionally,weincorporatedcontrolvariablesatbothcompanyandcountrylevels,detailedin

Table1

,toensuretherobustnessofourfindings.Tofurtheraddresspotentialendogeneityconcerns,weemployedPublicity

Table1

Variabledescriptions.

VariableType

VariableName

Symbol

Description

Source

Dependent

Financial

FD

Acompositeindexmeasuringthedevelopmentoffinancialinstitutionsandmarkets,

IMF

Variable

Development

basedondepth,access,andefficiency,fortheyears2020,2021,and2022.

Independent

GenerativeAI

GAIF

Thenaturallogarithmoftotalfunding(inmillionUSD)receivedbyindividualGenerative

Crunchbase

Variable

Financing

AIstartups.

ControlVariables

NumberofEmployeesStartupAge

StartupITSpending

ForeignDirect

Investment

NationalIncome

NOE

AGE

ITspendFDI

NI

AcategoricalvariablerepresentingtherangeofemployeesinGAIstartups(e.g.,1–10,11–50,51–100).

ThenumberofyearssincetheGAIstartupwasfounded.

ThenaturallogarithmoftotalannualITspendingbythestartup(inUSD).

NetinflowsofforeigndirectinvestmentasapercentageofGDPfortheyears2020,2021,and2022.

Thenaturallogarithmofadjustednetnationalincome(currentUSD)fortheyears2020,2021,and2022.

Crunchbase

CrunchbaseCrunchbaseWDI

WDI

Instrumental

PublicityExposure

PE

Thenaturallogarithmoftotalmediacoverage,includingpressmentions,articles,and

Crunchbase

variable

newsfeatures.

A.B.Siddiketal.FinanceResearchLetters72(2025)106519

3

Exposureasaninstrumentalvariable.

ThestatusandtrendsofGAIstartupsacross21countriesrevealsignificantvariationinboththenumberofstartupsandthefundingtheyhavereceived,asshownin

Fig.1

.TheUnitedStatesdominatesthelandscape,with229startupssecuringover$34.9billioninfunding,farsurpassinganyothercountry.TheUnitedKingdomfollows,with31startupsreceiving$642.75million.OthercountriessuchasGermanyandCanadaalsostandout,with17and16startupsreceiving$1.27billionand$1.07billion,respectively.Incontrast,smallermarketslikeArgentina,Estonia,andSouthKoreahavefewerstartups,typicallyreceivinglessthan$10millioninfunding.CountrieslikeAustralia,Brazil,Israel,andFrancehavemoderatenumbersofstartups,eachsecuringbetween$67millionand$487million.ThedataindicatesthatwhiletheU.S.istheclearleader,EuropeanandAsiancountriesarealsoseeingmeaningfulde-velopmentsinGAIfunding,albeitonasmallerscale.ThisdiversedistributionhighlightsthegrowingglobalinterestinGAI,thoughinvestmentishighlyconcentratedinafewkeymarkets.

2.2.Econometricmodels

ToexaminetheimpactsofGAIfinancingonfinancialdevelopment,weutilizesimplelinearregressionmodelsfortheyears2020,2021,and2022.Themodelsforeachyeararespecifiedseparately,asshowninthefollowingequations:

Fortheyear2020:

FDc=β0+β1GAIFc+β2NOEc+β3Agec+β4ITspendc+β5FDIc+β6NIc+ϵc……………..(M1)

Fortheyear2021:

FDc=β0+β1GAIFc+β2NOEc+β3Agec+β4ITspendc+β5FDIc+β6NIc+ϵc……………..(M2)

Fortheyear2022:

FDc=β0+β1GAIFc+β2NOEc+β3Agec+β4ITspendc+β5FDIc+β6NIc+ϵc……………..(M3)

Where:

FDcrepresentsfinancialdevelopmentforcountrycfortheyears(2020,2021,and2022).GAIFcisthetotalfundingreceivedbyGAI

Fig.1.ThestatusofGAIstartupswiththeirrespectivefundinginthesamplecountries.

4

A.B.Siddiketal.FinanceResearchLetters72(2025)106519

startupsincountryc.β2toβ6arecontrolvariablesforcountryc.εcistheerrortermforcountrycfortheyears(2020,2021,and2022).Toovercomepotentialendogeneityissuesandenhancetherobustnessofourbasemodel,weintroducethetwo-stageleastsquares

(2SLS)method.ThisapproachhelpsmitigateanybiasthatmayarisefromendogeneityintherelationshipbetweenGAIfinancingandfinancialdevelopment.Byusing2SLS,weaddresspotentialreversecausalityoromittedvariablebiasthatcouldaffecttheaccuracyoftheestimatedcoefficients.Moreover,weconductaheterogeneityanalysisbasedonsubcontinent-wisedivisions.ThisallowsustoexaminewhethertheimpactsofGAIfinancingonfinancialdevelopmentvaryacrossdifferentregions.Thesubcontinent-wiseanalysishelpsidentifywhetherregionalfactors,suchaseconomicconditionsorinstitutionaldifferences,leadtovariationsintheeffectsofGAIfinancingonfinancialdevelopment.Thisanalysisiscrucialtounderstandingthebroaderapplicabilityofourfindingsacrossdiversegeographiccontexts.

3.Resultsanddiscussion

3.1.Descriptivestatistics

Thedescriptivestatisticsprovideinsightsintothevariablesusedintheanalysisacrosstheyears2020,2021,and2022,asshownin

Table2

.Financialdevelopmentshowsrelativelystablemeansacrossthethreeyears,withslightvariationinstandarddeviation.GAIFhasameanof6.59(logarithmicform),withaconsiderablerangefrom3.65to10.05.ControlvariablessuchasNOE,startupage,andITspendingshowmoderatevariationacrossobservations.

Thecorrelationmatricesindicatetherelationshipsbetweenthevariables,asshownin

TableA1(a

-c

inAppendix).Forallyears,GAIFhasapositiveandsignificantcorrelationwithFD(around0.22),suggestingamoderateassociation.ITspendingalsoshowsapositivecorrelationwithFD,whileFDIexhibitsastrongnegativerelationshipwithFD,particularlyin2020and2022.NIconsistentlyshowsastrongpositivecorrelationwithFD,reinforcingitsimportanceasacontrolvariable.Additionally,thecorrelationbetweenGAIFandNOE(0.639),indicatingthatasthefinancingincreases,thenumberofemployeestendstoincreaseaswell.Similarly,GAIFandITspendingarepositivelycorrelated(around0.202),meaningthatstartupswithhigherfinancingalsotendtospendmoreonITinfrastructure.

3.2.Benchmarkregression

Thebenchmarkregressionresults(

Table3

)showthatGAIFconsistentlyhasapositiveandsignificantimpactonFDacrossallthreeyears(2020,2021,and2022).ThecoefficientsforGAIFrangebetween0.018and0.031,indicatingthathigherGAIfinancingcon-tributestoincreasedfinancialdevelopment.ThissuggeststhatGAIstartupsplayacrucialroleinenhancingthefinancialdevelopmentofthecountriesinthesample.ThemodelsalsoshowthatNOEhasanegativeandsignificanteffectonFD,implyingthatlargerGAIstartups(intermsofemployees)maynotnecessarilyboostfinancialdevelopment.Ontheotherhand,AGEandITspendingshownosignificantinfluenceonfinancialdevelopment.

ThecontrolvariablesFDIandNIhavesignificanteffectsonfinancialdevelopment.FDIconsistentlyshowsanegativeandsig-nificantrelationshipwithFD,whilenationalincomehasastrongpositiveimpactacrossallyears,highlightingtheroleofbroadereconomicfactorsinshapingfinancialdevelopment.ThehighR-squaredvalues(0.594–0.620)acrossthemodelsindicatethattheindependentvariablesexplainasubstantialportionofthevariationinfinancialdevelopment.Thesefindingsunderscoretheimpor-tanceofbothGAIfinancingandmacroeconomicvariablesindeterminingfinancialdevelopment,whilealsopointingtopotentialareasforfurtherinvestigation,suchastheroleofITspendingandfirmcharacteristics.

3.3.Robustnessanalysis

Theoutcomesofthe2SLSapproachpresentedin

Table4

indicatetherobustnessoftherelationshipbetweenGAIFandFDwhileaddressingpotentialendogeneity.Inthefirststage,PublicityExposure(PE)isusedasaninstrumentalvariableforGAIF,showinga

Table2

Descriptivestatistics.

Variable

Observation

Mean

Std.Dev.

Min

Max

FD2020

384

0.839

0.132

0.280

0.950

FD2021

384

0.840

0.136

0.252

0.939

FD2022

384

0.841

0.141

0.225

0.929

GAIF

384

6.590

1.033

3.653

10.053

NOE

383

2.117

1.234

1.000

8.000

AGE

384

4.521

2.744

1.000

14.000

Itspend

384

2.321

2.767

0.000

7.109

FDI2020

384

0.197

0.976

−1.219

3.140

FDI2021

384

0.629

0.972

−2.634

3.454

FDI2022

384

0.586

0.859

−4.048

3.884

NI2020

384

29.438

1.464

23.960

30.514

NI2021

384

29.543

1.443

24.121

30.606

NI2022

384

29.637

1.427

24.260

30.691

5

A.B.Siddiketal.FinanceResearchLetters72(2025)106519

Table3

Benchmarkregression.

Models

(M1)

(M2)

(M3)

(M4)

(M5)

(M6)

Variables

FD2020

FD2020

FD2021

FD2021

FD2022

FD2022

GAIF

0.028***

0.018***

0.029***

0.023***

0.031***

0.023***

(4.331)

(3.308)

(4.444)

(4.150)

(4.538)

(4.007)

NOE

−0.016***

−0.019***

−0.019***

(−3.311)

(−3.800)

(−3.726)

AGE

−0.001

−0.000

−0.000

(−0.333)

(−0.190)

(−0.034)

Itspend

0.003

0.002

0.002

(1.526)

(1.318)

(1.147)

FDI2020

−0.031***(−4.587)

NI2020

0.050***(10.794)

FDI2021

−0.025***(−5.481)

NI2021

0.067***(20.923)

FDI2022

−0.022***(−3.967)

NI2022

0.068***(18.870)

Constant

0.657***

−0.702***

0.647***

−1.235***

0.637***

−1.265***

(15.455)

(−5.139)

(14.753)

(−13.211)

(14.057)

(−11.995)

N

384

384

384

384

384

384

R2

0.047

0.600

0.049

0.620

0.051

0.614

adj.R2

0.044

0.594

0.047

0.613

0.049

0.608

Note:Significantat*p<0.1,**p<0.05,***p<0.01.

strongandsignificantpositiveeffectonGAIFacrossallmodels,withcoefficientsaround0.80.ThisdemonstratesthatPUisastronginstrumentforpredictingGAIfinancing.Inthesecondstage,GAIFcontinuestohaveapositiveandsignificantimpactonFDinallthreeyears(2020,2021,and2022).Thecoefficientsrangefrom0.029to0.034,confirmingthathigherGAIfinancingpositivelyinfluencesfinancialdevelopment,evenafteraccountingforendogeneity.Theunder-identificationtest(Kleibergen-Paap)showsthatthemodelsarewell-identified,andtheweakidentificationtestconfirmsthestrengthoftheinstrument.Additionally,theHansenJstatisticforoveridentificationdoesnotrejectthenullhypothesis,indicatingthattheinstrumentsarevalid.Thus,the2SLSresultsreinforcethefindingsfromthebenchmarkmodel,showingthatGAIfinancingsignificantlyenhancesfinancialdevelopment,andtheinstrumentalvariableapproacheffectivelyaddressespotentialbias.

3.4.Heterogeneityanalysis

TheheterogeneityanalysisexaminesthevaryingimpactsofGAIFonFDacrossdifferentsubcontinentalregions,asoutlinedin

Table5

.InPanelA(Asianvs.non-Asiancountries),GAIFhasasignificantandpositiveimpactonFDinAsiancountriesacrossallyears,withcoefficientsrangingfrom0.043to0.056.However,innon-Asiancountries,GAIF’simpactismuchweaker,withcoefficientsclosetozeroandstatisticalsignificanceonlyin2021(0.008,p<0.1).ThissuggeststhatGAIfinancingplaysamorecriticalroleindrivingfinancialdevelopmentinAsiancountriescomparedtonon-Asianregions.

PanelB(Europeanvs.non-Europeancountries)showsasimilardivergence.Innon-Europeancountries,GAIFhasapositiveand

Table4

TheestimationusinganIVapproachisconductedthrougha2SLSmethod.

Models

(M1)

(M2)

(M3)

(M4)

(M5)

(M6)

Variables

GAIF

FD2020

GAIF

FD2021

GAIF

FD2022

PE

0.796***

0.802***

0.799***

(13.560)

(13.770)

(13.680)

GAIF

0.029**

0.031**

0.034***

(2.690)

(2.920)

(3.350)

Constant

3.417***

−0.738**

2.597***

−1.251***

2.228***

−1.284***

(3.320)

(−2.670)

(3.730)

(−9.240)

(2.940)

(−12.110)

Controlvariables

Yes

Yes

Yes

Yes

Yes

Yes

N

384

384

384

384

384

384

Underidentificationtest

56.052***

57.043***

56.639***

Weakidentificationtest

183.901

189.503

187.238

HansenJstatistic

0.000

0.000

0.000

Note:Significantat*p<0.1,**p<0.05,***p<0.01.

6

A.B.Siddiketal.FinanceResearchLetters72(2025)106519

Table5

Heterogeneityanalysissubcontinentwise.

PanelA:Asianandnon-Asiancountries.

Models

Asian

Non-Asian

Asian

Non-Asian

Asian

Non-Asian

Variables

FD2020

FD2020

FD2021

FD2021

FD2022

FD2022

GAIF

0.056**

0.005

0.052**

0.008*

0.043*

0.008

(2.379)

(1.068)

(2.181)

(1.695)

(1.959)

(1.531)

Controlvariables

Yes

Yes

Yes

Yes

Yes

Yes

Constant

1.072

−0.610***

0.543

−1.154***

0.396

−1.123***

(1.393)

(−4.972)

(0.651)

(−13.789)

(0.423)

(−12.247)

N

40

343

40

343

40

343

R2

0.338

0.600

0.226

0.665

0.249

0.667

adj.R2

0.218

0.593

0.085

0.659

0.112

0.661

PanelB:European(EURO)andnon-European(NEURO)countries.

ModelsEURO

NEURO

EURO

NEURO

EURO

NEURO

VariablesFD2020

FD2020

FD2021

FD2021

FD2022

FD2022

GAIF0.012

0.020***

0.029*

0.016***

0.028

0.015***

(0.710)

(3.644)

(1.886)

(2.913)

(1.565)

(2.732)

ControlvariablesYes

Yes

Yes

Yes

Yes

Yes

Constant−1.407***

−0.497**

−0.696**

−1.719***

−1.675***

−2.480***

(−5.041)

(−2.154)

(−2.327)

(−13.504)

(−5.367)

(−13.873)

N85

298

85

298

85

298

R20.563

0.607

0.664

0.642

0.593

0.667

adj.R20.529

0.599

0.638

0.635

0.561

0.660

PanelC:Latin&northAmerica(L&NA)andothers.

ModelsL&NA

Others

L&NA

Others

L&NA

Others

VariablesFD2020

FD2020

FD2021

FD2021

FD2022

FD2022

GAIF0.002

0.061***

0.001

0.064***

0.001

0.070***

(0.500)

(4.197)

(0.402)

(4.839)

(0.340)

(4.572)

ControlvariablesYes

Yes

Yes

Yes

Yes

Yes

Constant−3.990***

−0.959***

−2.801***

−0.076

−1.466***

−0.920***

(−14.387)

(−3.137)

(−43.830)

(−0.233)

(−4.192)

(−2.685)

N253

130

253

130

253

130

R20.704

0.378

0.934

0.495

0.581

0.348

adj.R20.697

0.348

0.932

0.470

0.571

0.316

Note:Significantat*p<0.1,**p<0.05,***p<0.01.

significanteffectonFDinallthreeyears,withcoefficientsrangingfrom0.015to0.020.Incontrast,theimpactofGAIFinEuropeancountriesislesssignificant,showingmoderateeffectsonlyin2021(0.029,p<0.1).InPanelC(Latin&NorthAmericavs.othercountries),GAIFhasaconsistentlystrongandsignificantimpactonFDin"Other"regions,withcoefficientsaround0.061to0.070acrossallyears.InLatin&NorthAmerica,however,GAIFshowsnosignificanteffectonFD.Overall,theheterogeneityanalysishighlightsregionaldifferences,withGAIFshowingastrongerimpactonfinancialdevelopmentinAsianandnon-Europeancountriescomparedtootherregions.

4.Conclusion

OurstudyunderscoresthesignificantroleofGAIFinadvancingfinancialdevelopmentacross2020,2021,and2022,withrobustevidencefrom2SLSanalysisaffirmingGAIF’spositiveimpactonFD.Notably,weobservesubstantialregionalheterogeneity,asGAIFexertsastrongerinfluenceonFDinAsianandnon-EuropeancountriescomparedtoEuropeandtheAmericas.ThisregionaldisparitysuggeststhatGAIFismoreeffectiveinenhancingfinancialsystemsinsomeareas,whileotherfactorsmaydriveFDinothers.Thesefindingshavemultifacetedpolicyimplications.Policymakers,banks,financialinstitutions,andfinancialregulatorsshouldcollaboratetocreateconduciveenvironmentsforGAIstartupstoenhancefinancialdevelopment.InAsiananddevelopingregions,regulatorybodiesshouldfosterfavorableconditionsbyprovidingfinancialincentives,enhancinginfrastructure,andestablishingsupportiveregulatoryframeworks.Additionally,financialinstitutionsandbanksshouldactivelyinvestinandcollaboratewithGAIstartupstodriveinnovationinfinancialservices.ForEuropean,LatinAmerican,andNorthAmericanmarkets,financialregulatorsshouldfocusonintegratingGAIintoexistingfinancialframeworks,encouragingpartnershipsbetweenGAIfirmsandtraditionalfinancialentities.ThisapproachcouldhelpovercomelimitationsintheseregionswhereGAIFhaslessimpact.Furthermore,banksandfinancialin-stitutionsintheseareascanplayapivotalrolebyofferingtargetedfinancialproductsandservicesthatalignwithGAI-drivenin-novations.Finally,financialregulatorsacrossallregionsshouldaddressethicalconcerns—suchasbiasesinAI-drivendecision-making,privacyrisks,andpotentialjobdisplacement—bydevelopingframeworksthatpromotefairnessandtransparency.PubliccampaignsandinitiativestopromoteGAIcanalsoincreasevisibilityandattractinvestors,helpingdrivesustainablefinancialdevelopmentacrossdiversecontexts.Limitationsofthisstudyandrecommendationsforfutureresearchareoutlinedin

TableA2

(seeappendix).

A.B.Siddiketal.FinanceResearchLetters72(2025)106519

7