Sampling Design for 2010 CGSSYongjin JinYongjin JinSchool of StatisticsSchool of StatisticsRenmin University of China Renmin University of China Email: Email: Website: Website: OutlineBackground1Discussion2Sampling Design3The Final Result4OutlineBackground1Discussion2Sampling Design3The Final Sample4Background1To fully understand social development of urban and rural areas 1. Survey ObjectivesBackground1To fully understand social development of urban and rural areas 2To make comparisions between urban and rural groups1. Survey ObjectivesBackground1To fully understand social development of urban and rural areas 2To make comparisons between urban and rural groups3To reflect regional discrepancy of the development in China, 1. Survey ObjectivesBackground 2. Target PopulationThree-Stage stratified probability samplingA respondents selected randomly from the ultimate unit (household)3. Sampling MethodAll the household in 31 provinces (excluding Hong Kong, Macao and Taiwan) OutlineBackground1Discussion2Sampling Design3The Final Sample4DiscussionCategoryCategoryCategory 1Category 1Certainty uniteCertainty uniteSelf-representativeSelf-representative必选层必选层必选层必选层Category 2Category 2Selected uniteSelected unite抽选层抽选层抽选层抽选层DiscussionstratificationSelf-representativeImportant unites36 Municipalities, provincial capitals and sub-provincial cities.DiscussionstratificationSelf-representative36 Municipalities, provincial capitals and sub-provincial cities.ConsiderationEconomic levelEducational levelDegree of opennessDiscussionstratificationSelf-representative36 Municipalities, provincial capitals and sub-provincial cities.indexGDPNumber of teachersFDIActual used foreign investmentConsiderationEconomic levelEducational levelDegree of opennessDiscussionstratificationSelf-representative36 Municipalities, provincial capitals and sub-provincial cities.indexGDPNumber of teachersFDIActual used foreign investmentConsiderationEconomic levelEducational levelDegree of opennessTop FiveTop Five：Beijing, Shanghai, Guangzhou, Shenzhen, Beijing, Shanghai, Guangzhou, Shenzhen, TianjingTianjingfactor analysisDiscussionstratificationselected categoryOther regions excluding the self-representativesDiscussionstratificationselected categoryOther regions excluding the self-representativesConsiderationDistrict level（ urban） and county level（ rural）Futher stratificationDiscussionstratificationselected categoryOther regions excluding the self-representativesindexDensity of populationProportion of non-agricultural populationGDP per capitaConsiderationDistrict level（and country levelDiscussionstratificationselected categoryOther regions excluding the self-representativesindexDensity of populationProportion of non-agricultural populationGDP per capitaConsiderationDistrict level and country levelDistrict level : 19 District level : 19 stratumstratumCountry level : 31 Country level : 31 stratumstratumcluster analysisSampling stageAdministrative Divisions of ChinaProvince LevelDistrict/County LevelTownship Level: SubdistrictTownship Level: Township/TownNeighborhood CommitteeVillage CommitteeUrban AreaRural AreaDiscussionstageDiscussionstagesSelf-representativeSubdistrictor streetNeighborhood committeehouseholdselected stratumDistrict- or county-level unitsNeighborhood committee Villagers committeehouseholdDiscussionsample size and allocationTotal sample size: 12 000 unitsDiscussionsample size and allocationTotal sample size: 12 000 unitsOperational definitionNeighborhood committee consisted of urban householdVillagers committee consisted of rural householdSome consideringLarger discrepancy among urban individuals Urban-rural ratioDiscussionsample size and allocationTotal sample size: 12 000 unitsConsidering operabilityNeighborhood committee consisted of urban householdVillagers committee consisted of rural household7 200Urban household4 800Rural householdConsidering actual situationLarger discrepancy among urban individuals Urban-rural ratioDiscussionsample size and allocationStage140 PSUsstreets1. Self-representativeDiscussionsample size and allocationStage 2Stage1 402 SSUsNeighborhood committee40 PSUsstreets1. Self-representativeDiscussionsample size and allocationStage3Stage 2Stage140225 USUsHousehold（household member) 402 SSUsNeighborhood committee40 PSUsstreets1. Self-representative40 x 2 x 25 = 2000 housholds40 x 2 x 25 = 2000 housholdsDiscussionsample size and allocationStage1100 PSUsDistrict, county2. Selected stratum Discussionsample size and allocationStage 2Stage1 1004 SSUsNeighborhood committee (Villagers committee)100 PSUsDistrict, county2. Selected stratum Discussionsample size and allocationStage3Stage 2Stage1100425 USUsHousehold（household member) 1004 SSUsNeighborhood committee (Villagers committee)100 PSUsDistrict, county2. Selected stratum 100 x 4 x 25 = 10000 100 x 4 x 25 = 10000 HouseholdsHouseholdsOutlineBackground1Discussion2Sampling Design3The Final Sample4Sampling DesignSelf-representativeu Beijing, Shanghai, Guangzhou, Shenzhen, Tianjin;u The samples are all urban households;u Primary sampling units (PSUs) are sub-districts in the five cities above;uWe use PPS to select 40 sub-districts as the PSUs in this stratum.Sampling DesignStage 1Stage 2Stage 3PPSSelects 40 city sub-districts in five citiesPPS Selects 2 neighborhood committees from each selected sub-districtsSystematic samplingSelects 25 households as the ultimate sampling unitsu Sampling frame of the second stage: list of neighborhood committees in the sub-districts selected in the first stage;uIn the third stage, the households are ranged by house number;u Well sample 50 households in each neighborhood committee in case of the nonresponce;uDeliver the questionnaire in batches based on the response rate.Sampling DesignSample Distribution of Self-representativeArea PSUNeighborhood committeesTarget sample sizealternate sample sizeSelf-representative40402=8040225=200040250=4000Sampling DesignSelected stratumuAccording to The China 2009 Statistical Yearbook , there are 792 districts and 2,003 counties left in the selected category. uBased on the population density, the rate of non-agricultural population and per capita Gross Regional Product, stratify the unites in selected category into fifty strata ( 19 district strata and 31 county strata).uSelect 2 PSUs in each stratum, resulting a total of 100 PSUsSampling DesignDeducting the 2,000 urban households in the self-represent, the urban-rural ratio in 抽选层 should be 5,200:4,800, i.e. the ratio of neighborhood committees to villagers committees is about 208:192.Challenge how to determine the distribution of samples between urban and rural areas in each PSUs Deducting the 2,000 urban households in self-representative, the urban-rural ratio in selected category should be 5,200:4,800We can calculate that the ratio of neighborhood committees to villagers committees is about 208:192 According to the urban-rural ratio of 6:4, urban sampling size should be 7,200, and rural area should be 4,800.Sampling DesignOur thoughts:uDivide the districts and counties into several intervals according to the level of urbanization(the rate of non-agricultural population ). uFor a PSU with a higher level of urbanization, weshould allocate more neighborhood committees to that PSU . Sampling DesignAllocation criteriaNeighborhood committeeVillagers committee95% and above4050%95%3115%50%2215% and below13Allocation Criteria of Second UnitsSampling Design Construct second sampling frameDetermine the number of neighborhood/villagers committees sampled in PSUSelects neighborhood / villagers committees with PPSSample 25 households in each selected SSU,Sampling of Second UnitsuConsidering the nonresponce, enlarge the sample size same as self-representative.OutlineBackground1Discussion2Sampling Design3The Final Resulte4The Final Resultcategory Target sample sizePSUSSUTSUSelf-represent2,0004040*2=8040*2*25=2,000Selected stratum10,000100208+192=40025*400=10,000Total 12,00014048012,000The Final Sample Distribution