NBER WORKING PAPER SERIES
CHINA'S GDP GROWTH MAY BE UNDERSTATED
Hunter Clark
Maxim Pinkovskiy
Xavier Sala-i-Martin
Working Paper 23323
nber/papers/w23323
NATIONAL BUREAU OF ECONOMIC RESEARCH
1050 Massachusetts Avenue
Cambridge, MA 02138
April 2017
We are very grateful to Michelle Jiang, Lauren Price and Rachel Schuh for superb research
assistance. This paper reflects solely the opinions of the authors and not necessarily of the Federal
Reserve Bank of New York, the Federal Reserve System, or the National Bureau of Economic
Research.
NBER working papers are circulated for discussion and comment purposes. They have not been
peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies
official NBER publications.
sections of text, not to exceed two paragraphs, may be quoted without explicit permission
provided that full credit, includingnotice, is given to the source.
China's GDP Growth May be Understated
Hunter Clark, Maxim Pinkovskiy, and Xavier Sala-i-Martin
NBER Working Paper No. 23323
April 2017
JEL No. F0
ABSTRACT
Concerns about the quality of China’s official GDP statistics have been a perennial question in
understanding its economic dynamics. We use data on satellite-recorded nighttime lights as an
independent benchmark for comparing various published indicators of the state of the Chinese
economy. Using the methodology of Pinkovskiy and Sala-i-Martin (2016a and b), we exploit
nighttime lights to compute the optimal weights for various Chinese economic indicators in a best
unbiased predictor of Chinese growth rates. Our computations of Chinese growth based on
optimal weightings of various combinations of economic indicators provide evidence against the
hypothesis that the Chinese economy contracted precipitously in late 2015, and are consistent
with the rate of Chinese growth being higher than is reported in the official statistics.
Hunter Clark
Federal Reserve Bank of New York
33 Liberty St, New
York, NY 10045, USA
Hunter.Clark@ny.frb
Maxim Pinkovskiy
Federal Reserve Bank of New York
33 Liberty Street
New York, NY 10045
maxim.pinkovskiy@ny.frb
Xavier Sala-i-Martin
Department of Economics
Columbia University
420 West 118th Street, 1005
New York, NY 10027
and NBER
xs23@columbia
1Introduction
WeareinterestedinunderstandingrecentgrowthratesinChinaandwhethertheycoincidewiththeocially
publishedstatistics.Thereisalargeliteraturedoubtingtheaccuracyandveracityofociallypublished
ChineseGDPdata,oftenclaimingthatocialgrowthrateconsiderablyoverstatesactualgrowth.Ina
seminalarticle,Rawski(2001)arguesthattheChineseeconomymighthavegrownat2%orlessperyear
during1997-2001insteadofthe7.1%asociallyclaimed.Otherresearchers,inparticularAdamsandChen
(1996),MaddisonandWu(2007)havesharedthisskepticismbycomparingocialChinesegrowthestimates
togrowthratesofvariousinputsintoproduction,suchasenergy,steelandcement.Ontheotherhand,Holz
(2013)andPerkinsandRawski(2008)1haveclaimedthattheChinesedataaregenerallyaccurate.Recently,
ChinesestatisticshavebeenatthesourceofadditionalcontroversyfollowingWikileaks’publicationofthe
premier,LiKeqiang(thenstillaprovincialgovernor),admittingtoanAmericandiplomatthathemonitored
provincialeconomicactivitybythesimplearithmeticaverageofthegrowthratesofelectricityproduction,
railroadfreightandbankloans,andthattheocialstatisticswereman-madeandforreferenceonly.
TheEconomisthasbeenreportingtheLiKeqiangindextothisday.However,alloftheaboveproposalsfor
measuringthegrowthratesoftheChineseeconomydependonassumptionsabouttherelationshipsbetween
variousmacroeconomicproxiesandeconomicactivity,manyofwhicharediculttoevaluate.
Inthispaper,weattempttotranscendtheproblemofunderstandingwhichtheoryisrightbyusing
avariablethatcanactasanimpartialrefereeforthemacroeconomicproxies.PinkovskiyandSala-i-
Martin(2016aandb)showthatifwecan…ndavariablewhosemeasurementerrrorisindependentofthe
measurementerrorsofthemacroeconomicvariablesthatforecasterstypicallyaggregatetopredicteconomic
activity,wecana)testthequalityofthesevariables,andb)obtaintheoptimalweightsonthesevariablesina
bestunbiasedlinearpredictorofunobservedtrueincome.Wearguethatsuchavariableissatellite-recorded
nighttimelights(Elvidgeetal.1993,Hendersonetal.2012,PinkovskiyandSala-i-Martin2016aandb).
Webelievethatthisisareasonableassumptionbecauseerrorsinnighttimelightscomefromvariationin
weatherpatternsaswellasfromvariationinsatellitequalityovertime.Ontheotherhand,errorsinGDP
andotherocialseriescomefrommisreportingbyindividuals,…rmsandotherinstitutions,aswellasfrom
methodologicalchoicesofstatisticaloces.Therefore,thereislittlereasontobelievethattheseerrorshave
anythingtodowitheachother.
UnderourcrucialassumptionandusingdataacrossChineseprovincesandovertimefortheperiod2004-
2013,aregressionoflognighttimelightsonthemacroproxiesaswellasonprovinceandyear…xede¤ects
yieldscoe…cientsontheproxiesthatareproportionaltotheproxies’optimalweights(seethemathematical
1ThisisthesameRawskiasRawski(2001).frameworkinSection2.1).We…ndthattheLiKeqiangvariablesaresigni…cantpredictorsofnighttime
lightsgrowth,butthattheyreceiveradicallydi¤erentweights.Inparticular,bankloansreceiveconsiderably
morethatone-halfofthetotalweight,whilerailroadfreightreceivesmuchlessthanone-third.Electricity
productionalsotypicallyreceivesalargeandstatisticallysign…cantcoecient,butitmaybebiasedupward
ifourcoreidenti…cationassumptionfails.FormallytestingthatthecoecientsonthethreeLiKeqiang
variablesareequalrejectsthisnullhypothesisinmostspeci…cations,andthefailurestorejectthatwe
…ndaredrivenbylargestandarderrorsratherthansimilarcoecients.Sincethegrowthrateofloansis
largerandmorestableovertimethanthegrowthratesofelectricityand(especially)freight,this…nding
isimportantforpredictingtrueunobservedChineseGDPgrowth.Moreover,we…ndthataddinglogGDP
tothetheseparatecomponentsoftheLiKeqiangindexdoesnotmateriallychangethecoecientson
theLiKeqiangvariables,andthecoecientonlogGDPisonlymarginallysigni…cant.Wealso…ndthat
conservativeadjustmentsforthepotentialcorrelationbetweenthemeasurementerrorinnighttimelightsand
themeasurementerrorinelectricitydonotchangeourconclusionsconcerningtheimportanceofloansin
explainingnighttimelights.WeperformsimilarregressionsincludingothervariablesbesidestheLiKeqiang
componentsthathavebeenusedbyanalyststopredictChinesegrowth(e.g.retailsales,passengertrac
andoorspaceunderconstruction,amongothers)and…ndthatbankloanscontinuetobeimportantin
explainingnighttimelights,andtherefore,inpredictingChineseeconomicgrowth.
Foreachregressionthatwerun,weconstructforecastsforthepathofChinesegrowthratesbytaking
weightedaveragesofthenationalgrowthratesofthemacroeconomicproxiesasimpliedbytheregression
coecients.Aremainingproblemisthattheregressioncoecientsareproportional,butnotidentical,to
thevariables’optimalweights,andanyinterceptisunidenti…ed.Hence,wenormalizetheresultinggrowth
pathsbyregressingtheocialGDPgrowthpathontheweightedaverageforthe2004-2012periodandusing
thepredictedvaluesofthissecondregression(inandoutofsample)asourpredictedGDPgrowthseries.We
obtainstandarderrorsforthepredictedpathsofChineseGDPgrowthbyparametricbootstrappingfromthe
asymptoticdistributionoftheestimatedcoecientsfromtheprovincial-levelregression.We…ndthatGDP
growthin2015Q4,atimewhenthe…nancialpresswasawashwithstoriesabouta“hardlanding”ofthe
Chineseeconomy,issomewhathigher,butquiteclosetotheociallyreportedrate,witha95%con…dence
intervalthatprecludesgrowthratesconsistentwithasharpslowdownoftheChineseeconomyatthistime.
Wealsoobtainthis…ndinginourmostcomprehensivespeci…cationsthatincludedi¤erentsetsofmacro
proxiescorrespondingtothevariousindicesusedbymarketparticipantstopredicttheChine