WID.worldWORKINGPAPERSERIESN2017/12

ExtremeandPersistentInequality:

NewEvidenceforBrazilCombiningNationalAccounts,

SurveysandFiscalData,2001-2015

MarcMorgan

August2017

1

ExtremeandPersistentInequality:NewEvidenceforBrazilCombiningNationalAccounts,SurveysandFiscalData,2001-2015

MarcMorgan1,2

Firstversion:14July2017

Thisversion:10August2017

AbstractThis research describes the methods used and the results obtained in the estimation ofinequality across the entire income distribution in Brazil over the last fifteen years. Weproduceanewseriesofdistributionalnationalaccounts(DINA)andfiscalincome,combiningannual and nationally representative household survey datawith detailed information onincometaxdeclarationsrecentlyreleasedbytheFederaltaxoffice, inaconsistentmannerwith macroeconomic totals. Our results provide a sharp upward revision of the officialestimates of inequality in Brazil, while the decreasing inequality trends are reversedaccordingtoourbenchmarknationalincomeseries.Theconcentrationofincomeatthetopisstriking,withtheTop1%incomeshareincreasingto28%bytheendoftheperiod,fromaninitialshareof25%.TheTop10%increasedtheirshareofincomefromabout54%to55%ofpre-taxnationalincomeandcaptured61%oftotalgrowth.TheBottom50%sharerosefrom11%to12%,experiencinghighergrowththanthetopdecile,butcapturingonly18%oftotalgrowthduetoitsextremelylowcommandofincome.Whileelitesandthepoormadegains,theMiddle40%ofthedistributiondecreaseditssharefromabout34%to32%,postinglessgrowththantheaverageforthewholeeconomy.Thesqueezedmiddleisaproductofitsrelatively low shareof incomeandpoor growthperformance. Thus, inequality among thebottom 90% declined at the expense of growing concentration at the top. While labourincomeinequalitydidregisteradeclineaccordingtoourcorrectedseries,itwasinsufficientto mitigate the deduced concentration of capital resources and reverse the growingconcentrationofnationalincomeamongelitegroups.

1PhDstudentattheParisSchoolofEconomicsandLEcoledesHautesEtudesenSciencesSociales.2ResearchFellow,WorldInequalityLab,ParisSchoolofEconomics.Contact information:marc.morgan@psemail.eu. IgratefullyacknowledgefundingfromtheFundacinRamnAreces.

2

1.IntroductionFromaregionhistoricallycharacterizedbyhighandpersistentlevelsofincomeinequalitysinceatleastthelate19thcentury(Williamson,2015)Brazilisnoexceptiontobeingunderthespotlightinthisdomain.Inanyofficialreportonincomedistributionbyaninternationalorganization,Brazilusuallyfeaturesnearthesummitoftheinequalityrankings,asmeasuredbyhouseholdsurveydata,alongsidesomeofitsregionalcounterpartssuchasColombia,andSouth Africa.Whilemost studies on income inequality in developing countries use eithersurvey-based measures or tax-based measures of inequality (when available), this papercombinesnationalaccountswithnationallyrepresentativehouseholdsurveydata(fromthenational statistics office) with detailed tabulations on income tax declarations (recentlyreleased by the federal tax office) in a consistent manner to produce two new series ofinequality for Brazil a series of pre-tax fiscal income inequality and a series of nationalincome inequality across the entire distribution.We thus construct a set of distributionalnationalaccounts(DINA)forBrazil.Wecomparetheseserieswiththeseriesestimatedfromthe raw survey data (prior to any correction) and with a new corrected series of labourincomeinequality,wherewecombinethesamedatasourcestodistributelabourincomes. The lastfifteenyears inBrazilarean interestingperiodtostudybecause it iswhenBrazilianeconomicgrowthpickedupagainafterhavingbeen througha stagnantpreviousdecade, fuelled by a boom in commodity exports and finance. Average adult incomesexpanded by 18% in the worlds 8th largest economy and 3rd largest parliamentarydemocracy.Politically,itisalsoaninterestingperiodduetothecomingtopowerofthefirstleft-leaninggovernmentinBrazilsincetheearly1960s,withtheelectionofLuladaSilvatothe presidency and his Workers Party (PT) compatriots into congress in late 2002. Newinnovationswerebroughtabout in the fieldof socialpolicy to tacklepoverty, suchas theBolsaFamilia conditionalcashtransferprogram,andahighershareof fiscal receiptswerededicatedtosocialspendingastheshareoftotalgovernmentspendingintheeconomyroseovertheperiod.ThePTgovernmentalsobroughtaboutnegotiationsthatincreasedtherealvalueof theminimumwagebymore than50%during theirmandate.Thesevariablesarestrictly tied to the evolution of income inequality. Moreover, put into a historicalperspective, the present can be instructive about inequality in the past given the distinctpoliticalregimesthatcanbeidentifiedbetweenthepresentandpast. Our focus on pre-tax inequality is in contrast tomost studies that concentrate ondisposable(i.e.net)incomeinequality.Whilethisfocusisnecessarytoassesstheroleofthestate in the redistribution of income in Brazil, its sole use detracts from the marketdistributionofincomeinthecountry,whichcanbeseenastheprecursortothesecondary(disposable)distributionofincome.Thusbydealingwiththeoriginalmarketdistributionofincome under the DINA framework, this paper provides a different angle from which toanalyseincomeinequalityinBrazil.ButitalsoconsiderstheimpactthatBrazilscelebratedcash transfer programs have on the distribution of income, if we factor them into theestimatesusingasimpleandtransparentmethod.

3

Anumberofreasonsmayalsomotivatewhywecombinefiscaldataandsurveydatatomeasurethedistributionofincome.Themostobviousoneisalmostexclusiverelianceonhousehold self-reported surveys to assess income distributions in Brazil. While the trueincomedistribution isunobserved,household surveys canapproximateapersonal incomedistributionbyexpandingthefrequenciesofarepresentativesampleofthepopulation.Theproblemwithsurveysisthattheytendnottoincludecompleteinformationontheveryrichinthestudiedcountry.Despiterandomsampling,theirincomeiseithernotwellmeasuredorarenotobserved,duetothereluctanceoftherichest individualstodisclosealloftheirincomesources,particularlytheirassets.Additionally,therichmayrefusetoengageinthetime-consuming task of answering a comprehensive household survey, assuming thatinterviewers manage to enter the gated communities in which they live. Moreover,statisticians may intentionally remove extreme observations, so as to top-code thedistribution. Surveys are thusprone toover-represent theextentof labour incomeat thetop of the distribution and underestimate the extent of capital income distributed tohouseholdscomparedtowhatthenationalaccountswouldimply. Incometaxdata,ontheother hand, better captures richer individuals, as filing a declaration is obligatory abovespecifiedincomethresholds,andinmanycasesthereisthirdpartyreporting.Althoughnoteverybody declares income to the fiscal authorities and some people can be tempted tounder-declare their income inorder topay less tax,wecanbequiteconfident in thinkingthat the people appearing in tax data actually exist (as they are well identified by fiscalprocedures)andearnatleastwhattheydeclare. This paper is thus among the first to use personal income tax records to studydistributive issues, but it is not the only one.Medeiros et al. (2015) andGobetti &Orair(2016)havealso contributed to show that taxdata conveyadifferentpictureofBrazilianinequalitythanthatwhichwaspreviouslyreleasedusingsurveys.However,ourstudyisthefirsttogenerateaseriesofdistributionalnationalaccounts(DINA).Thisisimportantfortwomainreasons.First,bybeinganchoredtothenationalaccounts,DINAallowustodistributetheproceedsofgrowthasmeasuredanddiffusedbyofficialsources.Thiswasnotpossibleinprevious studies because the income concept used had less of a direct relation withmacroeconomicgrowth. Second, andmore importantly,DINAallowus to see towhom inthehierarchyincomeoriginallyflows.Whileitmaybearguedthatfiscalincome(i.e.incomereceivedbyhouseholdssubjecttoassessmentbythetaxauthorities)ordisposableincome(i.e. incomeindividualsactuallydisposeofaftergovernmenttaxesandtransfers)aremorerelevant, the distribution of national income gives us insights into the distribution ofeconomicresources,includingcorporations,pensionfunds,insurancefundsandrealestate.It is thusmore intricately connectedwith the distribution of capital and power. This canoftenbemorecrucialthanthedistributionofdisposableincomeasitprecedesitintheflowoffundsandcanthusdeterminethesocietyscapacityforredistribution. We present our main results in the form of income shares because the help tostratifytheincome-generatingpopulationintoincomeclasses,sothatatop,forexample,maybevisible,asopposedtobeingconfoundedinasyntheticindicatorliketheGini.Suchan

4

indicatorissyntheticinthatitsummarizeswithonenumberthebetween-groupdispersionof income across thewhole population. It is difficult to understand how such an abstractindicatorhasbeenconstructedandwhatitreallymeans.Weknowisthatwhenitiscloserto0thedistribution itdescribes ismoreequal,whilewhenit iscloserto1thedistribution ismoreunequal.Yetwhenpresentedwithanumberlike0.44or0.65it isnoteasytograsp.Distribution tables depicting income shares, on the other hand, are a lot easier tounderstandastheirconstructionisstraightforward(theaverageincomeofagivenfractileinthedistributiondividedbytheaverageincomereceivedbytheadultpopulation)andtheirinterpretation is transparent (an incomeshareof50% for theTop10%of thedistributiongivesusaclearsenseofhowthepieisdivided).Agroupthatreceiveshalfofalldistributedincomewhenitonlyrepresentsone-tenthofthepopulationisamoreconcreteandvisibleclaim on income concentration than saying that the Gini is 0.60, as the latter is withoutreferencetoanyparticularsocialgroupinthehierarchy.AssuchwithanindexliketheGiniweareunabletoobservetheinequalitybetweenthetopandthebottomofthehierarchyorbetween themiddle and the bottom or themiddle and the top or within the top.Moreimportantlywhenpresentingincomelevelsincashterms(insteadofpercentages)itmakesitpossible for peo