《世界银行-使用重新加权和贫困预测模型校正电话调查贫困估计中的抽样和无响应偏差（英）-2023.12..docx》由会员分享，可在线阅读，更多相关《世界银行-使用重新加权和贫困预测模型校正电话调查贫困估计中的抽样和无响应偏差（英）-2023.12..docx（43页珍藏版）》请在课桌文档上搜索。
1、PaUOLnnValnsoosQ.2-qndPolicyResearchWorkingPaper10656CorrectingSamplingandNonresponseBiasinPhoneSurveyPovertyEstimationUsingReweightingandPovertyProjectionModelsP ZOfnV alnsoosQ.2-qndKexinZhangShinyaTakamatsuNobYoshidaworldBankgroupxt-rPovertyandEquityGlobalPracticeDecember2023PolicyResearchWorkingP
2、aper10656AbstractTo monitor the evolution of household living conditions during the COVID-19 pandemic, the World Bank conducted COVID-19 High-Frequency Phone Su,eys in around 80 countries. Phone sunfeys are cheap and easy to implement, but they have some major limitations, such as the absence of PoV
3、Crty data, sampling bias due to incomplete telephone coverage in many developing countries, and frequent nonresponses to phone interviews. To overcome these limitations, the World Bank conducted pilots in 20 countries where the SUrVey ofWellbeing via Instant and Frequent Tracking, a rapid povem, mon
4、itoring tool, was adopted to estimate poverty rates based on 10 to 15 simple questions collected via phone interviews, and where sampling weightswere adjusted to correct the sampling and nonresponse bias.This paper examines whether reweighting procedures andmethodology can eliminate the bias in povc
5、rtr estimation based on the COVID-19 High-Frequency Phone Surveys. Experiments using artificial phone survey samples show that (i) reweighting procedures cannot fully eliminate bias in povert estimates, as previous research has demonstrated, but (ii) when combined with SUrVey of Wellbeing Via Instan
6、t and Frequent Tracking PoVerty projections, they effectively eliminate bias in poverty estimates and otherstatistics.hispaperisaproductofthePovertyandEquityGlobalPractice.ItispartofalargereffortbytheWorldBanktoprovideopenaccesstoitsresearchandmakeacontributiontodevelopmentpolicydiscussionsaroundthe
7、world.PolicyResearchWorkingPapersarealsopostedontheWebatTheauthorsmaybecontactedatkzhang2worldbank.org.TbePolieyResearchWorkingPaperSeriesdisseminatestfjefindingsofuforkinprogresstoencouragetbeexchanfofideasaboutdeivlopmentissues.Anobecihfeoftheseriesisto般thefindingoutquickfy,evifthepresentationsare
8、lessthanJulbfpolished.ThepaperscanythenamesOftbeauthorsandshouldbecitedaccordingly.Thefinding,interpretations,andconclusionsexpressedinthispaperareentircbfthoseoftheaut!,wrs.TheydonotnecessarilyrfpnsvnttheviewsoftheInternationalBankforKecoftstnictionandDert,omenf/WorldBa欣anditsafftiaterPSWtoeliminat
9、esamplingbias,asetofassumptions,likestrongignorability,needtobesatisfied,whichcannotbeeasilytested,forwhichnon-PSWservesasacomplementforweightingpurposes.Second,non-PSWapproachesmatchthemeansofkeyindicatorsbetweenaphone/webSUrVeyandareferencesurey,butthereisnoguaranteethatthedistributionsoftheseindi
10、catorsarealsomatched.PSWapproaches,bycontrast,matchthedistributionofpropensityscores.Sincebothapproacheshavetheirownstrengths,conductingbothPSWandnon-PSWadjustmentsisreasonable.ThispaperinvestigateswhetherreweightingcancorrectthebiasofpovertyprojectionsproducedbytheSWIFTmethodology.Theperformanceofr
11、eweightingtechniquesdiffersbydataandtargetindicatorsthatwerematched,andthereisagreementintheliteraturethatreweightingtechniquesreducethebiasesintargetstatisticsyetdonoteliminatethem(Lee(2006)andDrezeandSomanchi(2023).DrezeandSomanchi(2023)createdbiasedsamplesbydroppingpoorerhouseholdsfromahouseholdS
12、UrVeyandtestedwhetheranon-PSWreweightingtechnique(maximumentropyreweighting,ormaxentropy)canreducebiasesinpovertyratesandmeanhouseholdexpenditures.1thoughthebiasesinpovertyrateestimatesandmeansofhouseholdexpendituresdeclined,substantialproportionsremained.However,existingliteraturelacksanassessmento
13、fhowwellreweightingtechniquescanreducethebiasesofpovcrtrprojectionsproducedbySWIFToranyotherPOvCrtyprojectionmethod.UsingphoneorWCbsurveystoestimatepovertynecessitatestheuseofPoVertyprojectionmethods.DrezeandSomanchi(2023)usedactualconsumptionandincomedataandshowedthatalargebiasinthepovertyrateandme
14、anhouseholdexpenditureremainsevenafterreweightingbutdidnotassessifreweightingcombinedwithpovcm,projectionmethodsiseffectiveinreducingthebias.Infoct,ourstudyfindsthattheperformanceofreweightinginestimatingpovertjrcanbeimprovedwhencombinedwithpovertyimputationmethodssuchasSWIFT.ExperimentsThispaperexa
15、mineswhether,andifso,towhatextent,acombinationofreweightingtechniquesandtheSWIFTpovertyprojectionmethodOlOgycaneliminatesamplingbiasesinpovertyestimatesbasedonbiasedsurveysamples.Toseethis,wcfirstusereferencehouseholdsurveysinRwanda,StLucia,andUgandaandconstructsubsamplesbyselectinghouseholdswithatl
16、eastamobileorlandlinephone.Thesesamplesaresubjecttosamplingbiasbyconstruction.WithoutreweightingandSWIFTpovertyprojections,thepovertyratesinthesesubsamplcsofphoneownersarclowerthanthoseinthefullsamples.WethenexaminewhetherreweightingandSWIFpovertyprojectionscancorrectfortheabovementionedbias.Phonean
17、dwebSUrVeydatacollectionsfacesamplingandnonresponsebiases,buttheabovementionedexperimentsonlyfocusonsamplingbiasesthatarisefromunevenphoneownership.TounderstandrheabilityoftheSWIFTandreweightingtechniquestoadjustfbrnonresponsebias,thispaperconductsanadditionalexperimentusingthesampleofEthiopiaHigh-F
18、requencyPhoneSurveyround7(HFPS7),whichisasubsampleofthe2018/19EthiopiaSocioeconomicSUrVeyround4(ESS4).SincethissubsampleofESS4includesonlyphoneowners,itissubjecttosamplingbias.Also,sinceitincludesonlyhouseholdsofESS4thatrespondedtotheHFPS7,itisalsosubjecttononresponsebias.Usingthissubsample,Weconduc
19、tthesameanalysisasabovetoassesswhetherreweightingtechniquesandSWIFTcancorrectthebiasinpovertyestimatesarisingfrombothsamplingandnonresponsebiases.Thispaperisorganizedasfollows.SectionIIintroducestheSWIFTpovertyprojectionmethodOlOgyandreweightingtechniques,andSectionIIIdisplaysresultsfromfourexperime
20、ntalstudies(SaintLucia,Rwanda,Uganda,andEthiopia).SectionIVconcludeswithanassessmentofacombinationofreweighringandtheSWIFT-basedpovertyprojectionmodelineliminatingthebiasinpovertyrestimationbasedonthef()urcasestudies.II. SWIFTandReweightingMethodologyII.1.SllHFTpovertyprojectionmetbodoloSWIFTisanapp
21、licationofSuney-to-Sureyimputationtechniques(S2S)tomonitorpovertrrapidly.SWIFTtrainsanimputationmodelinatrainingdatasetbyregressinghouseholdcxpcnditurcs/incomcsonpovertyproxies.Householdexpendituresandpovemrratesarethenimputedinanotherdataset,calledoutputdata,“bypluggingpovertyproxiesoftheoutputdata
22、intothemodel.Figure1illustratestheprocess.TherearetwokeyassumptionsinthestandardSvnFTmethodology.First,therelationshipbetweenhouseholdincomeorexpenditureandpovertycorrelatesinrheoutputdatacanbeexpressedinequation(1):lnyh0=Xhoo+UhO(1)wherel0N(0,0)whereIny0referstoanaturallogarithmofthehouseholdincome
23、orexpenditureofhouseholdbintheoutputdatao.Xfioisa(Ze1)vectorofpovertycorrelatesofhouseholdhntheoutputdata,o.0isa(k1)vectorofcoefficientsofpovertycorrelates(x0).UfioreferstoaresidualandisoftenassumedtofollowanormaldistributionOfN(0,0O)JTheoutputdataincludesthePOVertyproxydata冲。%=butdocsnotinclude1Thi
24、snormaldistributionandlinearity-canberelaxed.Forthesakeofexposition,normaldistributionisassumed.householdexpendituresZnyl0)=1whicharetobeimputed.Forthesakeofexposition,therelationshipisassumedtobelinear,butthisconditioncanberelaxed(Yoshidaetal.,2022a).Figure11llustrationofbowtheSI11FTworksDatasetJny
25、krtoh.Relationship如乂%;=M嫄-M九饶MModelstabilityholdsImputationb:.i7i(,uImputeddata11nyL.xk)fc.SNote:Authors*illustration.ThesecondkeyassumptionisthattherelationshipbetweenhouseholdexpendituresandPOVertyproxiesinthetrainingdataalsofollowsequation(1).Thisassumptioniscalled“modelStabiIity“implyingthatthem
26、odeldocsnotchangefromthetrainingandoutputdata.SWIFTestimatesparametersinequation(15),(0,趣),andtheirdistributions,usingthetrainingdataset,drawthem(0r,偷)randomlyfromtheirestimateddistributions,andsubstitutesthemintoequation(1)toimputehouseholdexpendituresforallhouseholdsintheoutputdata.SWIFTrepeatsthisimputation(typically20-100times),resultingin20to100vectorsofhouseholdexpenditures)intheoutputdata(witheachvectorincludingtheexpenditurefcrallhouseholds).Povertyrandinequalitymeasuresareestimated
