高盛銀行業(yè)不可能三角報(bào)告英文版

4July2023|11:21PMHKTChinaFinancialServicesTesting

the‘ImpossibleTrinity’

PartI:Ninequestionsaroundlocalgov.

debtWe

introducethe?rstreportinour3-partTesting

the‘ImpossibleTrinity’

series,which

looksattheideathatbankscannotmaintainagoodbalanceofprovisions,capitalanddividendsatthesametimeowingtosqueezedearnings.Inthisreport,weassumelocalgov.

debtdefaultriskwillbelimitedaslongasdebtrolloverispermittedandnetbalancescontinuetoincrease,andassessthepotentialmulti-yearmarginlossofbanksontheback

oflocalgov.

debtrolloverduetoloweringrates.ShuoYang,

Ph.D.+852-2978-0701|shuo.yang@GoldmanSachs

(Asia)L.L.C.We

lookatninequestionsthatwethinkwillbetopofmindfor

investorsandgroupthesequestionsunderthreebroadthemes.The?rstthemeconcernsthelocalgov.

debtexposureonbankbalancesheets,intermsofsize,mixanddistribution.We

concludenon-coveredbankswith48%localgov.

debttototal

assets(vs.18%

forcoveredbanks)would

facemorechallengeswithpotentialtailrisk.Inferringfromthis,weexpectsixlargebankswithlargerbalancesheetstostepupandtakeonmorelocalgov.

debt.Thesecondthemeisbankearningsrisk,duetomarginlossonlocalgov.

debt.

WeusealossassumptionofRmb~30tn(USD4.5tn)grossadditionoflocalgov.

debtanda~30bpseffective

ratedecreaseeach

year,

withbankingsystemROEdecreasingby~100bps,

andnon-coveredbanksby~150bps

each

yearin2023E-25E.Basedonthis,we

stress-testthata~60bpsratecutperyearonlocalgov.

debtwouldtriggernon-coveredbankstoface

recapitalizationriskwithROEdecliningto1.7%.Ourthirdthemeassessesriskfactorsoflocalgov.

debtwideningthedivergenceofindividualbanks.

We

highlighttwosetsofbankstoseehowthesefactorsaffectdivergence:1)ABC(downgradetoSell,fromNeutral)andPSBC(upgradetoBuy,fromSell)onthesizeoflocalgov.

debt,and2)ICBC(downgradetoSellfromBuy)andCCB(maintainBuy)onthe%oflocalgov.

bonds.For

ourinvestmentviewsoneach

bank,pleaseseethethirdreportinthisseries,Testing

the‘ImpossibleTrinity’III:Increasingdividendrisk;PSBCuptoBuy,ICBC/ABC/IndustrialdowntoSell.GoldmanSachs

doesandseekstodobusinesswithcompaniescoveredinitsresearch

reports.Asaresult,investorsshouldbeaware

thatthe?rmmayhave

acon?ictofinterestthatcouldaffect

theobjectivityofthisreport.Investorsshouldconsiderthisreportasonlyasinglefactorinmakingtheirinvestmentdecision.For

RegACcerti?cationandotherimportantdisclosures,seetheDisclosureAppendix,orgoto/research/hedge.html.

Analystsemployedby

non-USaf?liatesarenotregistered/quali?edasresearchanalystswithFINRAintheU.S.GoldmanSachsChinaFinancialServicesNineQuestionsaroundlocalgov.

debtonbankbalancesheets1.What’s

thesizeandmixoflocalgov.

debtonbankbalancesheets?Basedonoureconomists’numbers,weestimateRmb34tn(USD4.9tn)oflocalgov.debtsitsonthebalancesheetsofcoveredbanks,which

accountfor

61%ofbankingassets.There

aretwoelementstolocalgov.

debt,namelyloans(includingshadowcredit)ofRmb18tn

(USD2.6tn)andbondsofRmb16tn

(USD2.3tn),onourestimates.We

arriveatthisbysumminguploanswithatenorofmorethan5years

(excludingmortgages),andusetheaverageportfolioshareof~48%for

localgov.

bondsheldbybanks.

We

alsoincludeshadowcreditinlocalgov.

debtandsummarizeitaspartofloans.Exhibit1:WeassessRmb34tn(USD4.9tn)localgov.

debtonExhibit2:BoCom/Huaxiahavelargest/smallestlocalgov.

debtexposure(as%ofIEA)at29%/9%,with16%avg.ofcoveredbanks.Asof2022,GSebalancesheetsofcoveredbanks,accoutingfor61%ofbankingassets,withloans(includingshadowcredit)ofRmb18tn(USD2.6tn)andbondsofRmb16tn(USD2.3tn).Asof2022,GSeLocalgov.debtexposureLocalgov.debtas%ofIEA(2022)9.51029%30%924%

24%Coveredbanks25%846.83Rmb34tnintotalRmb16tnbonds20%7654321020%16%Avg.16%Rmb18tnloans

&

shadowcredits13%

13%4.6315%12%

11%11%10%3.619%3.110%5%62.21.8214111.010.8220.40.30.220%Localgov.loan

as%

ofIEA

Localgov.bond

as%

ofIEA

Shadowbankingas%

ofIEALocalgov.loansLocalgov.bondsShadowbankingSource:GoldmanSachsGlobalInvestmentResearch,CompanydataSource:GoldmanSachsGlobalInvestmentResearch,Companydata2.Whataboutthelocalgov.

debtexposureofnon-coveredbanks?We

estimatenon-coveredbanks,which

accountfor

39%ofbankingassets,have

Rmb59tn(USD8.5tn)oflocalgov.

debt,makingup63%ofthetotal

basedonGSeconomists’estimatesoftotallocalgov.

debt.

We

alsoestimatethatnon-coveredbanksaccountfor48%oflocalgov.

debtasa%oftotalIEA,vs.18%for

coveredbanks,suggestingmoreearningsriskonpotentialmarginlossonlocalgov.

debtgiventhelargersize.4July20232GoldmanSachsChinaFinancialServicesExhibit3:CoveredbankshaveRmb34tn(USD4.9tn)localgov.

debt,18%oftotalassets,vs.Rmb59tn(USD8.5tn)and48%oftotalassetsfornon-coveredbanks.Exhibit4:Systemlocalgov.

debtexposurebreakdownAsof2022,GSeAsof2022,GSeLocalgov.debtexposureSystemlocalgov.debtexposurebreakdown(2022)12010080604020060%50%40%30%20%10%0%9448Coveredbanks:Rmb34tn,37%oftotalexposure48%Rmb16tn,17%Rmb32tn,35%593229%Rmb18tn,19%34161818%Non-coveredbanks:Rmb59tn,63%oftotalexposureRmb27tn,29%4527SystemCoveredbanksNon-covered

banksCoveredbanks-bondsNon-coveredbanks-loansCoveredbanks-loansNon-coveredbanks-bondsLocalgov.loans+

shadowbankingLocalgov.bondsAs%

oftotalassets(RHS)Source:GoldmanSachsGlobalInvestmentResearch,CompanydataSource:GoldmanSachsGlobalInvestmentResearch,Companydata3.Dolargebankshavelargerexposuretolocalgov.

debt?Thesixlargebanks(ICBC,

CCB,

ABC,

BOC,BoCom,

PSBC)accountfor

Rmb29tn(USD4.2tn)oflocalgov.

debt,31%oftheestimatedsystemtotal,basedonGSeconomists’estimates.

Andweassumelargerbankscansupportmorelocalgov.

debtgrowthduetotheirsizablebalancesheets,withlocalgovdebtasa%ofIEAat18%,

comparedwith29%forthesystem.

Witha1%increaseinexposure,webelievethesesixbankscouldsupportadditionallocalgov.

debtofRmb1.6tn

(USD0.2tn),potentiallyoffsetting

theslowdowningrowthoflocalgov.

debtonnon-coveredbankbalancesheets,asweassesstheexposurewould

be48%.Exhibit5:SixlargebankstotalRmb29tn(USD4.2tn)localgovdebt,31%ofestimatedsystemtotal.Asof2022,GSeExhibit6:Weexpectlargebankscansupportmorelocalgov.

debtgrowthontheirlargebalancesheet,withlocalgov.

debtas18%ofIEA,vs.29%ofsystem.Asof2022,GSeSystemlocalgov.debtexposurebreakdown(2022)Localgov.debtas%oftotalassets60%Smallerbanks,

Rmb5tn,5%48%50%40%Largebanks,Rmb29tn,31%29%30%Non-coveredbanks,Rmb59tn,64%18%20%10%0%16%Non-coveredbanksSystemSixlargebanks(covered)Smallerbanks(covered)Non-covered

banksLargebanksSmallerbanksSource:GoldmanSachsGlobalInvestmentResearch,CompanydataSource:GoldmanSachsGlobalInvestmentResearch,Companydata4.What’s

thegrossandnetadditionoflocalgov.

debteachyear?We

estimateRmb~30tn(USD4.5tn)ofgrossadditionoflocalgov.

debteach

yearto2025E,withRmb20tn(USD3.0tn)rolloverbalanceandRmb10tn

(USD1.5tn)

netaddition.4July202339XdU9YnW8VyWaQcM8OnPrRpNnOjMrRqPeRoOrP8OnMsOwMnOpMwMmNqRGoldmanSachsChinaFinancialServicesFor

thepurposesofthisanalysis,we

assumeanaverage5yeardurationoflocalgov.debt,andunchangedweightsoflocalgov.

debtonbankbalancesheets.

This

wouldmean20%oflocalgov.

debtrolloverand~10%

balancegrowtheach

year.Exhibit7:WeestimateRmb30/31/34tn(USD4.2/4.4/4.8tn)grosslocalgov.

debtincrease,withRmb19/21/23tn(USD2.7/3.0/3.3tn)ofrolloverandRmb11/10/11tn(USD1.5/1.4/1.5tn)ofnetincreasein2023E/24E/25E...Exhibit8:...comparedwithnew

systemcreditgrowthadditionofRmb25/23/25tn(USD3.5/3.3/3.5tn),assumingannualbalancegrowthof~10%.Grossincreaseoflocalgov.debt(system)Gross/netincreaseoflocalgov.debtvs.systemcredit401403435302531120100801123301021111920343112615105306011525251052394401110112002023E2024E2025E020222023E2024E2025ENetincreaseofsystemcredit

Grossincreaseoflocalgov.debt

Netincreaseoflocal

gov.debtLocalgov.debtRolloverNewincreaseGrossincreaseSource:GoldmanSachsGlobalInvestmentResearch,CompanydataSource:GoldmanSachsGlobalInvestmentResearch,Companydata5.What’s

thepotentialmarginlossonlocalgov.

debt?We

estimatesystemROE

willdecreaseby~100bps

each

year

in2023E-25E,duetomarginlossonlocalgov.

debt,elsebeingequal.Lookingatcoveredbanks,wearriveata~60bpsROEdecrease,andnon-coveredbanksat~150bps.We

quantifythemarginriskofbothstock

rolloverandincreased?ows

oflocalgov.

debt,assumingdefaultriskislimitedaslongasrolloverisallowedandnetbalancesincrease.We

expectloweringratesonbothstock

rolloverandincreasesinlocalgov.

debt?owsgiventheincreasingsize

anddif?cultyofpreventingdefaultrisk.

We

assumea1%ratedecreaseonrolloverdebteach

year

and~150

bpsspreadlossbetweenlocalgov.

debtandtheoverallbankrateasopportunitycosttomodelthemarginlossoflocalgov.

debt.Thisimpliestheeffective

interestrateonlocalgov.

debtwoulddecreaseby

~30bpseach

year

onaveragein2023E-25E.This

would

translateintoaNIMdecreaseofcoveredbanksbyanaverage3-10bps

each

yearin2023E-25E.

We

believeICBC/ABC/Industrialbankwouldhave

alargerNIMcontractionof7.3/6.6/9.4bps

vs.the6.0bpsaverageforcoveredbankseach

year

2023-25E,whileCMBandBOCwouldhave

asmallerNIMdecreaseof4.0/3.3bps,mainlyduetothesizeoflocalgov.

debt.

Themorelocalgov.debtheldbybanks,themoreNIMdilution.Basedontheseassumptions,andwiththemulti-yearmargincontractionledbyloweringratesofincreasinglocalgov.

debt,bankingsystemCET1ratiocouldbe9.9%,withthatofnon-coveredbanksremainingabove

theminimumrequirementat7.9%

byend-2025E,althoughtheCET1

ratiowoulddecreaseby~80bpscumulativelyin2023E-25E.4July20234GoldmanSachsChinaFinancialServicesExhibit9:WeestimatesystemROEdecreaseby~100bpseachyearin2023E-25E,duetomarginlossonlocalgov.

debt,elsebeingequal.Comparedwithcoveredbankswhichata~60bpsROEdecrease,non-coveredbankswoulddecreasec.~150bps.Exhibit10:Theeffectiveratecuteachyearwouldbe~30bps,with1%ratecutfor20%balancerolloverand150bpsopportunitycostfor~10%newgrowth,whichwouldtranslatetoaNIMdecreaseofcoveredbanksby3-10bpseachyearin2023E-25E.MarginlossimpacttoROELocalgov.debteffectiveratetrajectory(basecase)14360120.720.610.55340320300280260240220200-1810-161.10864200.967.60.897.4-1811.311.1-1210.81.573.67.81.423.41.383.1350-18-113162862023E2024E2025E2023E2024E2025E2023E2024E2025E256Covered

banksSystemNon-covered

banks2022

Balance

Newrollover

growth2023E

Balance

Newrollover

growth2024E

Balance

New

2025Erollover

growthAdj.ROEvs.beforeadj.Source:GoldmanSachsGlobalInvestmentResearch,CompanydataSource:GoldmanSachsGlobalInvestmentResearch,CompanydataExhibit11:WehighlightICBC/ABC/IndustrialbankwouldhavelargerNIMcontractionof7.3/6.6/9.4bpsvs.6.0bpsaverageofcoveredbankseachyearin2023-25E,whileCMBandBOCwouldhaveasmallerNIMdecreaseof4.0/3.3bps,mainlyduetothesizeoflocalgov.

debt.Exhibit12:Onourassumptions,withthemulti-yearmargincontractionledbyloweringratesofincreasinglocalgov.

debts,bankingsystemCET1ratiocouldbe9.9%,withthatofnon-coveredbanksremainingabovethemin.requirementat7.9%byend-2025E,thoughtheCET1ratiowoulddecreaseby~80bpscumulativelyin2023E-25E.NIMdilution(2023-25Eavg.)MarginlossimpacttoCET-1ratio0.0-1.0-2.014120.0611.60.1111.50.1511.30.1010.40.1810.11080.259.9-3.0-2.90.158.50.288.20.397.9-4.0-5.0-6.0-7.0-8.0-9.0-10.0-3.3-4.06-4.2-4.84-5.7Avg.-6.0bps2-6.6-7.1-7.3

-7.402023E2024E2025E2023E2024E2025E2023E2024E2025ECoveredbanksSystemNon-coveredbanks-9.4-9.5Adj.CET-1ratiovs.

beforeadj.Source:GoldmanSachsGlobalInvestmentResearch,CompanydataSource:GoldmanSachsGlobalInvestmentResearch,Companydata6.Howlowcouldtheratesonlocalgov.

debtgo?Ourstress-testshowsthat,iftheeffective

interestrateonlocalgov.

debtdecreasesby183bpscumulativelyin2023E-25E,or~60bpseach

year,

non-coveredbankswould

facerecapitalizationriskwiththeCET1ratiodippingbelow7.5%,

themin.requirement,byend-2025E.

AndthesystemROEwould

be6.6%vs.7.4%

inourbasecase,withcovered/non-coveredbanksROEat10.3%/1.7%

vs.10.8%/3.1%ofourbasecasein2025E.Thissuggeststhat,beforereachingthepointofnon-coveredbanks(39%ofbankingassets)requiringrecapitalization:1)theeffective

interestrateonlocalgov.

debtcouldgolower,

asourstresstestsuggests30bpsmoreratecutseach

year

or~100bpslower

intotalfor2023E-25E;or2)moretimecouldbeallowedfordebtrollover,

assuminga?xedsetupofratecutsonlocalgov.

debtinourbasecase,elsebeingequal.4July20235GoldmanSachsChinaFinancialServicesExhibit13:Westress-testthatwitheffectiveinterestrateonlocalgov.

debtdecreasingby~60bpseachyear,

non-coveredbankswouldfacerecapitalizationriskwiththeCET1ratiodippingbelow7.5%,themin.requirementbyend-2025E.Exhibit14:ThesystemROEwouldbe6.6%vs.7.4%basecase,andcovered/non-coveredbanksROEwouldarriveat10.3%/1.7%vs.10.8%/3.1%ofbasecasein2025E.Adj.CET-1ratio(stresscase)Adj.ROEinbase/stresscase(2025E)2.85%3.0%2.5%12.010.810.314.02.25%12.010.01.68%2.0%1.5%1.0%0.5%0.0%-0.5%-1.0%10.08.06.04.02.00.08.06.04.02.00.07.4Requiredmin.7.56.611.511.411.210.310.07.99.78.47.53.11.72023E2024E2025ENon-coveredbanksCoveredbanksRequiredmin.SystemEffectiverate(RHS)BasecaseStresscaseBasecaseStress

caseBasecaseStresscaseCoveredbanksSystemNon-coveredbanksSource:GoldmanSachsGlobalInvestmentResearch,CompanydataSource:GoldmanSachsGlobalInvestmentResearch,Companydata7.Whatwouldhappentocoveredbanksifthelocalgov.

debtratewentlowertowardstheneedforrecapitilization?Ifwestresstestratesonlocalgov.

debtgoinglower

at~60bpseach

year

in2023E-25E,wecouldseecoveredbankshave

aNIMdecreaseof12bps

each

year,

drivingROEandCET1ratiodecreaseby

~120/10bps

each

year,

assumingcreditcostdoesnotclimbupandelsebeingequal.Bytheend-2025E,thestresstestedROEandCET1

ratiowould

be10.3%/11.2%

onaverage,or52/14bps

lower

thanthebasecase.Exhibit15:Withthestresstesttosetratesonlocalgov.

debtistoExhibit16:...andCET-1

ratiowouldbe11.2%,or~14bpslowerthanbasecase,bytheend-2025E.lowerat~60bpseachyearin2023E-25Evs.basecaseof~30bps,ROEofcoveredbankswouldbe10.3%,or52bpslowerthanofbasecase...Adj.ROEinbase/stresscase(2025E)Adj.CET-1ratioinbase/stresscase(2025E)17.015.013.011.09.00.315.00.0914.00.80.10

0.1913.00.712.011.010.09.00.0811.50.80.81.00.40.169.916.40.39.80.069.60.79.313.90.19

0.200.79.10.139.114.512.9

12.90.198.80.38.50.98.40.158.713.00.208.311.7

11.610.6

10.28.09.59.47.07.05.06.0Stresscasevs.basecaseStresscasevs.basecaseRequiredmin.Source:GoldmanSachsGlobalInvestmentResearch,CompanydataSource:GoldmanSachsGlobalInvestmentResearch,Companydata8.Doesthelocalgov.

debtmixmatterasanoffsettobankearnings?Withmorebondsinlocalgov.

debtportfolios,andmoretax(0%taxratevs.~15%

ofloans)andcapitalsavings(10%

(LGGB)or20%(LGSB)riskweightingvs.100%

ofloans),itmeansbankscanbetteroffset

themarginlossonlocalgov.

debt.Therefore,webelievethe%ofbondsinlocaldebtportfoliosisanimportantfactorin4July20236GoldmanSachsChinaFinancialServicesdifferentiatingbanks’bottom-linegrowth.Mostbanksguidetheywould

liketotakemorelocalgov.

bondsontotheirbalancesheets;despitenominalyieldsonlocalgov.

bondnotbeinghigh,thetax

andcapitalsavingscanprovidemoreupsidetoadj.return.Banks

alsoguidethatifthecentralbankallowsmorelocalgov.

bondsasquali?edcollateralforliquidity,localgov.

bondscanbemoreattractiveassetstoenhancedebtportfolioreturns,amidloweringrates.We

estimatethesystemadjustedlocalgov.

debtyieldcouldbe0.8%higheronaveragein2023E-25E,thankstothetaxandcapitalbene?ts

oflocalgov.

bonds.We

estimatetheweightedcapitalchargesoflocalgov.

bondscouldbe39%,blendedby29%ofLGGB(localgovernmentgeneralbonds)witha10%

riskweighting,43%ofLGSB(localgovernmentgeneralbonds)witha20%riskweighting,and28%ofimplicitLGFV(localgovernment?nancingvehiclebonds)witha100%

riskweighting.This

couldhelpbankssave

capitalby61%,comparedwithloanswhich

require100%

riskweighting.

We

believethesaved

capitalcoulddriveadditionalreturnsonnormalizedROEof11%

for

coveredbanks,assumingthereturnoncapitalforeach

coveredbankremainsunchanged.Thereby,

withinclusionofthetaxandcapitalbene?tsoflocalgov.

bonds,wecouldseethedecreaseinROEduetomarginlossoflocalgov.

debtnarrowby1.1ppts

forcoveredbanksand1.9ppts

fornon-coveredbanks.Withinourcoverage,we

have

bankswithlargerlocalgov.

bondexposures,i.e.PSBC/CMB(83%/68%bond%localgov.

debt),andbankswithsmallerexposures,i.e.IndustrialandBoCom(24%/31%),comparedwiththe50%averageofcoveredbanks.Exhibit17:Morebondsinthelocalgov.

debtportfolios,moretax(0%taxratevs.~15%ofloans)andcapitalsavings(10%(LGGB)or20%(LGSB)riskweightingvs.100%ofloans)meanbankscanoffsetthemarginlossoflocalgov.

debtExhibit18:Weestimatethesystemadjustedlocalgov.

debtyieldcouldbe0.8%higheronaveragein2023E-25E,thankstothetaxandcapitalbene?tsoflocalgov.

bonds.MixTaxrate9%Riskweighting69%RmbtnBalanceMix(bond)Totalbenefitsoflocalgov.bond(2023E)Totallocalgov.debt(onshore)Localgov.bond

(onshore)Official94483514211445100%52%37%15%22%14%48%7001.0%100%72%29%43%28%4%39%1.1%0%16%6000.8%4080.9%0.7%0.5%0.3%0.1%-0.1%-0.3%-0.5%LGGB0%10%0.6%5004003002001000LGSB0%20%Implicit(LGFVbond)Localgov.

loan15%15%100%100%26214622912910017813346SystemCoveredbanksNon-coveredbanksAdditionalreturnon

capitalTaxsaving△localgov.

bondyield(RHS)Source:GoldmanSachsGlobalInvestmentResearch,CompanydataSource:GoldmanSachsGlobalInvestmentResearch,Companydata4July20237GoldmanSachsChinaFinancialServicesExhibit19:Withtheinclusionofthetaxandcapitalbene?tsoflocalgov.

bonds,wecouldseethedecreaseinROEduetomarginlossoflocalgov.

debtnarrowingby1.1pptsforcoveredbanksand1.9pptsfornon-coveredbanks.Exhibit20:Withinourcoverage,wehavebankswithlargerlocalgov.

bondexposures,PSBC/CMB(83%/68%bond%localgov.

debt),andbankswithsmallerexposures,IndustrialandBocom(24%/31%),comparedwiththe50%averageofcoveredbanks.Asof2022,GSeAdj.ROEafter/beforeincludingbenefitsLocalgov.mix(%)14.012.010.08.012.51.210012.21.1116144521711.91.190807060504030201003132386248524452489.31.59.11.58.81.4546238636869835.51.95.31.95.01.93.1686.06411.311.110.85241354.07.87.631247.42.03.63.40.02023E

2024E

2025E

2023E

2024E

2025E

2023E

2024E

2025ECoveredbanksSystemNon-coveredbanksLocalgov.bond

as%

oftotal

Local

gov.loanas%

of

totalShadowbanking

as%

oftotalAdj.ROEaftermargindilutionROEchangeonbenefitsAdj.ROEafterbenefitsSource:GoldmanSachsGlobalInvestmentResearch,CompanydataSource:GoldmanSachsGlobalInvestmentResearch,Companydata9.What’s

thepotentialearningsimpactforcoveredbanks?We

estimateanaverage6bpsPPOPROA

decreaseforourcoveredbanksin2023E-25Eduetomarginlossonlocalgov.

debt.

With

theinclusionoftaxandcapitalsavings,wecouldseealower

ROEdecreasefor

ourcoveredbanks,by~1.1ppts

each

year

to12.5%/12.2%/11.9%

in2023E-25E.We

thinkbothsizeandmixoflocalgov.

debtarefactorsthatdifferentiate

banks,asweexpectthelargerlevelsoflocaldebtheldby

banks,themoremarginlossontheback

ofloweringratesoflocalgov.

debt,whilethemorebondsinlocalgov.

debtportfolios,themoretaxandcapitalbene?ts

cannarrowthebottom-linedecrease.Thus,

wescreenourcoveredbanks,andhighlighttwogroups:1)ABC(downgradetoSell,fromNeutral)andPSBC(upgradetoBuy,

fromSell)onthesizeoflocalgov.

debt,and2)ICBC(downgradetoSell,fromBuy)andCCB(maintainBuy)onthe%oflocalgov.

bonds(seePart

IIIintheseriesforinvestmentviews

ofeach

bank).nABCandPSBConthesizeoflocalgov.

debt,giventheirsimilarlystrongdepositbases.ABChas20%localgov.

debtasa%ofIEA,comparedwith13%

forPSBC.Assuminglocalgov.

debtonbothbanksrollsover

withloweringrates,thePPOPROA

ofABCwoulddecreaseby

7bpseach

year

onaveragein2023E-25E,comparedwith5bpsofPSBC,elsebeingequal.Moreover,

whenfactoringinthetaxandcapitalsavingsoflocalgov.

bonds-PSBChas83%bondexposureintheirlocalgov.debtportfolio,comparedwith48%for

ABC-thesavingscouldhelpPSBCnarrowitsROEdecreaseby

160bpseach

yearonaveragein2023E-25E,vs.139bps

atABC.nICBCandCCBontheirlocalgov.

mix,asICBChas38%exposuretolocalgov.bonds,comparedwith62%forCCB.

The

morelocalgov.

bonds,themorecapitalandtaxsavingstomitigateearningspressurecausedbylocalgov.

debt.

WecompareICBCwithCCBasthetwostocksgenerallytradeinthesamedirection,andasinvestorsseefew

differences

betweenthetwo.However,

wethinklocalgov.debtcanwidenthedivergence.Ifwe

lookatthesebanks’reporteddataandapply4July20238GoldmanSachsChinaFinancialServicesthesameassumptionsasourmarginlosstest,weestimateICBCcouldseePPOPROA

decreaseby7bpseach

yearonaveragein2023E-25E,comparedwith4bpsforCCB.Ifweincludetaxandcapitalsavings,themarginlossonthebottomlinecouldnarrowmoreforCCB,

withROE

decreasinglessby125bps

each

yearonaveragein2023E-25E,comparedwith120bpsfor

ICBC.Exhibit21:PSBC/ABChas13%/20%localgov.

debtexposureas%ofIEA,withROEadjustedformarginlossandbondbene?tsat13.8%/11.7%onaveragein2023-25E.Exhibit22:ICBC/CCBhas38%/62%localgov.

bondmix,withROEadjustedformarginlossandbondbene?tsat12.2%/11.3%onaveragein2023-25E.Adj.ROEafter/beforeincludingbenefitsAdj.ROEafter/beforeincludingbenefits16.014.012.010.08.012.61.313.71.713.81.613.91.512.21.211.81.211.71.311.31.214.012.010.08.010.91.212.01.411.71.411.41.36.06.012.112.212.411.310.910.610.510.110.610.49.810.04.04.02.02.00.00.02023E2024EPSBC2025E2023E2024EABC2025E2023E2024ECCB2025E2023E2024EICBC2025EAdj.ROEaftermargindilutionROEchangeonbenefitsAdj.ROEafterbenefitsAdj.ROEaftermargindilutionROEchangeonbenefitsAdj.ROEafterbenefitsSource:GoldmanSachsGlobalInvestmentResearch,CompanydataSource:GoldmanSachsGlobalInvestmentResearch,CompanydataTheauthorwould

liketothankZihan

WangandClaireOuyangfor

theircontributionstothisreport.4July20239GoldmanSachsChinaFinancialServicesAppendixBasedonGSmacroteamestimatesfor

gov.

debt,wesumupof?ciallocalgov.

debtofRmb34.9tn(USD5.0tn)andimplicitonshorelocalgov.

debtofRmb58.9tn(USD8.4tn)(incl.Rmb13.5tn(USD1.9tn)

ofbondandRmb45.4tn(USD6.5tn)loansandshadowcredits),togettothetotalonshorelocalgov.

debtexposureofRmb~94tn(USD13.4tn).Exhibit23:Localgov.

debt(onshore)isestimatedtobeRmb94tn(USD13.4tn),asof2022.EstimatesfromGSmacroteam.Source:GoldmanSachsGlobalInvestmentResearch,MOF,

Wind,Bloomberg,CEIC4July202310GoldmanSachsChinaFinancialServicesDisclosureAppendixRegACI,ShuoYang,

Ph.D.,

herebycertifythatalloftheviewsexpressedinthisreportaccuratelyre?ect

mypersonalviewsaboutthesubjectcompanyorcompaniesanditsortheirsecurities.Ialsocertifythatnopartofmycompensationwas,

isorwillbe,directlyorindirectly,relatedtothespeci?crecommendationsorviewsexpressedinthisreport.Unlessotherwisestated,theindividualslistedonthecoverpageofthisreportareanalystsinGoldmanSachs’GlobalInvestmentResearch

division.GSFactorPro?leTheGoldmanSachs

FactorPro?leprovidesinvestmentcontextforastock

bycomparingkey

attributestothemarket(i.e.ourcoverageuniverse)anditssectorpeers.Thefourkey

attributesdepictedare:Growth,FinancialReturns,