Supplement to the User's Manual for the 2012 CFPS Data
Contents
1. level of education attained in 2012 edu2012 and the highest level of school attended in 2012 sch2012 3 6 Employment Status Adult Dataset Each adult s employment status employ is classified as employed unemployed or economically inactive at the time of interview 3 7 Cognitive Function in the Adult and Child Datasets Cognitive function was measured through a word recall test and a number series test in 2012 The number series test involved an adaptive two stage design in which scores in Stage I test items determine the difficulty level of Stage II test items We provide summary scores for both the word recall test and the number series test in the adult and child datasets Due to its complex design we have constructed two summary scores for the number series test a Guttman score based on original questions and a W score based on the Rasch model of Item Response Theory For more information about the number series test see the CFPS technical report CFPS 31 in Chinese Table 3 summarizes the aforementioned individual level composite variables 3 Table 3 Composite Variables in the Adult and Child Datasets Variable Name Variable Label Personal Income INCOME Personal income INCOME_adj Personal income Adjusted Educational Level edu2012 Highest level of education attained in 2012 sch2012 Highest level of education attended in 2012 Cognitive Function NS_G NS_W NS_WSE IWR1 I2. respondents reside The first variable refers to the Census Bureau s definition of whether a community was classified as urban or rural in 2012 urban12 The other two variables are based on respondents self report and interviewer s observation urbancomm is a dichotomous indicator of urban versus rural and typecomm further distinguishes among four types of communities city town village and suburb Table 2 Composite Variables in the Family Dataset Variable Name Variable Label Family Income WAGE _1 Wage income WAGE 2 Wage income comparable with year 2010 wage_1_adj Wage income Adjusted wage_2_adj Wage income Adjusted comparable with year 2010 6 foperate_1 foperate_2 ftransfer_1 ftransfer_2 fproperty_1 fproperty_2 FELSE_1 FELSE_2 FINCOME1 FINCOME2 fincome1_adj fincome2_adj fincome1_per fincome2_per fincome1_per_adj fincome2_per_adj fincper_p fincperadj_p Family Expenditure PCE FOOD DRESS HOUSE DAILY MED TRCO EEC OTHER EPTRAN EPWELF MORTAGE EXPENSE Other typecomm urbancomm urban12 Business income Business income comparable with year 2010 Transfer income Transfer income comparable with year 2010 Property income Property income comparable with year 2010 Other income Other income comparable with year 2010 2011 2012 net family income 2011 2012 net family income comparable with year 2010 2011 2012 net family income Adjusted 2011 2
3. 012 net family income Adjusted comparable with year 2010 2011 2012 net family income per capita 2011 2012 net family income per capita comparable with year 2010 2011 2012 net family income per capita Adjusted 2011 2012 net family income per capita Adjusted comparable with year 2010 Percentile of family income per capita Percentile of family income per capita Adjusted Residents consumption expenditure Sum Expenditure on food Adjusted Expenditure on clothing yuan Expenditure on housing Adjusted Expenditure on family equipment amp daily necessities Adjusted Medical and fitness expenditure Expenditure on communication and transportation Adjusted Expenditure on education and entertainment Other expenditure on consumption Transfer expenditure Welfare expenditure imputed Mortgage on housing Estimate Total family expenditure Type of village residential community simplified Urban or rural according to community type Urban area Census Bureau s definition in year 2012 3 4 Personal Income in the Adult and Child Datasets We have constructed both unadjusted income and adjusted income_adj personal income variables The main difference between the two is that we carried out imputations for persons whose wage income was 0 or missing for adjusted personal income 3 5 Educational Level in the Adult and Child Datasets Two composite variables on education are available the highest
4. 1012 Panel weight individual level nationally representative subsample Family Dataset fswt_natcs12 Cross sectional weight family level total sample fswt_rescs12 Cross sectional weight family level nationally representative subsample fswt_natpn1012 Panel weight family level total sample fswt_respn1012 Panel weight family level nationally representative subsample Total sample includes all families or individuals in the CFPS The nationally representative subsample is a subsample from the total sample where only parts of the five large provinces subsamples and the full 20 small provinces subsamples are included For details about the sampling strategy of CFPS users may refer to the User s Manual of CFPS 2010 http www isss edu cn cfps EN Documentation js 236 html Cross sectional weights are used to compute national estimates in 2012 Panel weights should be used when conducting longitudinal analysis and they are available only for the families and individuals that participated in both waves 12
5. C F g S China Family Panel Studies Supplement to the User s Manual for the 2012 CFPS Data Qiong Wu Lihong Dai Yahong Cui Wenjia Zhang Cong Zhang Jia Chen Hongwei Xu Yu Xie 03 11 2015 Contents Vin trOGUCti OM ses cesce ede eetiees reiri ae ied ee NN OANT ae ete nde ES 1 1 1 Adult Dataset seinen en Ln ee a ee Sa 1 1 2 Child Dataset nininini ei iii 2 1 3 Family Dataset aonana eee ee ee ee et A EE ASA A SAS 2 1 4 Family Roster Dataseboncnenorenninin ienai aiaa ne eee 2 1 5 Cross Wave ID Dataset 0 0 0 eecesceseceneceseeeseeeseceseeeaeeeaeeeaeesaeeeaeseaeeeaeeeaeeeaeeeaeeeaeees 3 Changes in Data Structure from CFPS 2010 00 00 cccecesesseteeceeceeeeaeeaeeeeceeesaeeaeeeeeaeeas 4 2 1 Combining Face to face and Telephone Interview Data 00 0 cee eeeeeeeeneeeneee 4 2 2 Combining Long and Short Forms of Questionnaires in Individual level Data 4 Composite Variables ici cise eee erie tie eee et ie te ete te ee ee 6 3 1 Family Income in the Family Dataset 0000 00 eeeeseeeeeeeeceeeeeeeeeeetneeeeeeaees 6 3 2 Family Expenditure in the Family Dataset 0 00 0 eee eeeeeeeeeeeeeeeeeeeteneteeeeeees 6 3 3 Community Type in the Family Adult and Child Datasets 00 0 eee 6 3 4 Personal Income in the Adult and Child Datasets cccccccccsssssseeeeeeeeesesees 7 3 5 Educational Level in the Adult and Child Datasets cccccccsssssseeeeeeeeeeeees 8 3 6 Employment Status Adult Da
6. WR2 IWR DWR Other Employ Number Series Test Guttman scale 0 15 Number Series Test W score Number Series Test Standard error of W score Immediate Word Recall 1st trial Immediate Word Recall 2nd trial Immediate Word Recall Combined Ist and 2nd Delayed Word Recall Employment status Note Detailed information on constructing these composite variables will be available in corresponding technical reports 4 Best Variables Similar to 2010 we have also constructed a number of variables ending with best in the publicly released 2012 data The CFPS research team recommends these as the best among all other related variables when used in secondary data analysis These best variables were created under the following circumstances 1 when information on a variable came from multiple sources for example when information on gender and age was recorded in both the family roster questionnaire proxy report and the individual questionnaire self report and might not be identical in the two sources 2 when suspicious extreme values were identified during the data cleaning process To deal with inconsistency across different data sources we have adopted the basic principle that self reports are preferred to proxy reports unless we have other reasons to believe that the former lack credibility To deal with suspicious values for a given variable we have borrowed information from other related variables For instance some ass
7. could not be followed in 2012 n 136 are retained in the family roster for future tracking 1 5 Cross Wave ID Dataset The cross wave ID dataset includes all CFPS members gene members core members and non core member in any wave regardless of their interview status in any particular wave Compared with the family roster dataset the cross wave ID dataset includes more individuals e g those deceased cases and those who were not interviewed in a particular wave The main purpose of the cross wave ID dataset is to record the basic demographic information on all the CFPS individual participants and their interview status in every wave 2 Changes in Data Structure from CFPS 2010 Two important changes were made in the field operation of the 2012 follow up survey compared to the 2010 baseline causing changes to the 2012 data structure 2 1 Combining Face to face and Telephone Interview Data In CFPS 2010 all the interviews were done in person In CFPS 2012 telephone interviews were conducted whenever face to face interviews could not be completed Telephone interviews account for about 0 89 2 8 and 0 44 of all the family adult and child interviews respectively The content of the telephone interview was quite similar to that of the face to face interview except that certain infeasible questions such as cognitive tests were dropped in the former In addition certain complex questions were also simplified in the telephone inte
8. dents might receive only the long form 4 questionnaire for most of the observations in the individual dataset or both the short and long form questionnaires for non coresident family members successfully followed up in their current residence or only the short form questionnaire for non core members whose data were not publicly released and non coresident family members who could not be tracked The proportions of those in which only either the long or short questionnaire were used in the adult dataset were 87 9 and 7 2 respectively and the remaining 4 9 were interviewed using both questionnaires The corresponding figures are 95 6 2 8 and 1 6 respectively for the child dataset Since the short form questionnaires were mostly answered by proxy respondents questions on such topics as subjective test and cognitive test were not included Most of the questions in the short form questionnaire were directly taken from the long form questionnaire so the same variable names were adopted When we combined the two sources of data we kept the self reported data from the long form questionnaire if both types of questionnaire were available for the same respondent There were two conditions under which we retained the data from the short form questionnaires when the long form questionnaires were also available 1 when a variable had a missing value in the long form but not in the short form questionnaire and 2 when a variable was measured
9. e 2012 wave since the community survey is scheduled to be carried out in every other wave A cross wave individual level dataset file has been added listing all the individuals who ever participated in the CFPS as well as their vital records survey participation status respondent type and family ID in each wave Table 1 lists the sample size and number of variables for the 2012 datasets A brief introduction to the structure of each dataset is given below Table 1 Sample Size and Number of Variables in Each of the Five CFPS 2012 Datasets Datasets Sample Size Number of Variables Adult 35719 1740 Child 8621 811 Family Roster 55014 319 Family 13315 685 Cross Wave ID 61423 18 1 1 Adult Dataset The adult dataset includes adult respondents aged 16 and above who were either repeat gene members from the 2010 baseline survey or new adult family members gene or core members added in 2012 These respondents were interviewed either face to face IWmode 1 or by phone IWmode 0 using long form longform 1 and or short form shortform 1 questionnaires Note that although most respondents received only one form of the questionnaire it is possible for certain respondents to have both the long and the short form questionnaires see Section 2 2 for more The statistics in this report are based on version 5 0 or 5 1 of CFPS2012 data details The adult dataset includes 34 425 original gene members from the 2010 baseline survey 23 new g
10. ene members in 2012 and 1 271 core members who are immediate relatives of the gene members but are not gene members themselves 1 2 Child Dataset The child dataset includes children aged between 0 and 15 who were either repeat gene members from the 2010 baseline survey or new child family members gene or core members added in 2012 Data on children under 10 were collected through proxy reports by their adult guardians whereas data on children aged 10 or above were collected through both self reports and proxy reports by their guardians Again the interviews were completed through either face to face or phone interviews The child dataset includes 7 254 original gene members from the 2010 baseline survey 1 264 new gene members added in 2012 and 103 core members 1 3 Family Dataset The family dataset treats family as the unit of analysis It covers the original families interviewed in 2010 as well as the new families that were split from original families since 2010 due to changes in marital status children s economic independence and other reasons It includes 12 625 original families and 690 newly split families 1 4 Family Roster Dataset The family roster dataset consists of individual level information on all the family members who were listed in the 2012 family rosters The 2012 family roster dataset includes 55 014 observations from 13 453 households It should be noted that these 55 014 observations do not represent 55 014 uni
11. et values were initially deemed too high or too low and we then looked at related variables such as types and total areas of real estate in house facilities and family income and family expenditure to assess the validity of the recorded values Detailed information on property data cleaning can be found in the CFPS 2012 technical report China Family Panel Studies the Property Values in CFPS 2012 and CFPS 2010 CFPS 29 in Chinese 10 5 Variable Coding 5 1 Occupational and Industry Codes CFPS 2012 collected respondents detailed work information including family agriculture work agriculture work for other families non agriculture waged job non agriculture self employment private business and unpaid help with family business Most of the work related variables contain responses recorded in verbal descriptions which had to be recoded for data release The detailed procedure on how the original verbal descriptions of responses were transcribed into standard and numeric occupational codes and industry codes can be found in the technical report CFPS 8 in Chinese The design of the work module in CFPS 2012 is different from that in CFPS 2010 For the convenience of cross year analysis we have created another composite variable Occupation of main job in 2012 job2012mn_occu Detailed information on constructing this variable can be found in the CFPS 2012 technical report China Family Panel Studies the Occupation of Cu
12. in different ways between the long and short form questionnaires e g different question wording In the latter case we would retain both variables and assign a missing value to the one that was not directly collected from the targeted respondent 3 Composite Variables We have constructed a number of composite variables to facilitate data analysis 3 1 Family Income in the Family Dataset We have constructed multiple variables related to family income that suit different research needs including total family income family income per capita and source specific income e g wage income business income transfer income property income etc These composite income variables are listed in Table 2 Because of several changes in the survey questions regarding income in CFPS 2012 as opposed to those in CFPS 2010 we have constructed another set of 2012 income variables that are comparable to those in 2010 We recommend using these comparable income variables in cross wave comparative analysis 3 2 Family Expenditure in the Family Dataset Family expenditure variables include total family expenditure and specific types of expenditure Total family expenditure consists of transfer welfare house mortgage and regular consumption of items such as food clothes etc see Table 2 3 3 Community Type in the Family Adult and Child Datasets In CFPS 2012 we provide three composite variables that capture the types of communities where
13. que individuals This is because the dataset lists the information of those who split from their original families twice once as a member of the original family and once as a member of the new family A dichotomous variable co_al2_p is coded 1 if a pid individual level unique ID is financially connected with a fid12 family ID in 2012 and O otherwise By CFPS definition an individual can belong to only one family at any given time point so for any pid co_al2_p 1 for only one fid12 Therefore when the family roster dataset is restricted to individuals with co_al2_p 1 there would be no duplicate individual records The rationale for retaining such duplicated individual records in the family roster data is that this better preserves the dynamic changes in family compositions between 2010 and 2012 Researchers may refer to the forthcoming document entitled Decomposing Family Roster Data and Constructing Family 2 Relations in the CFPS in Chinese for details The family roster dataset also recorded 639 deaths between 2010 and 2012 It includes 40 757 gene members 75 8 1 354 new gene members 2 5 2 713 new non gene members 5 0 5 488 people who were not living in the household physically but were still financially connected to the family 10 2 and 2 672 new gene members 5 0 living in the family By study design a small number of gene members who were abroad in prison or for other reasons
14. rrent Main Job CFPS 30 in Chinese 5 2 Address Codes Similar to the 2010 baseline survey address information in all the 2012 datasets consists of 3 level codes province code district county code and village community code The province code follows the national standard providing real province identifier the district county code and village community code are both recoded pseudo values so that the true county and community names are kept anonymous for confidentiality For those who are interested in examining the possible contextual effect at the county level a county level dataset which includes a number of macro variables e g GDP total population employment rate education levels etc is available upon further request and application The detailed information can be found in the CFPS 2010 technical report China Family Panel Studies the Conversion of District County Data CFPS 23 in Chinese 11 6 Weighting Variables CFPS 2012 provides both cross sectional and panel weights at the family and individual levels Table 4 presents the complete list of weighting variables for CFPS 2012 Table 4 Weighting variables in CFPS 2012 Variable Name Variable Label Adult and Child datasets rswt_natcs12 Cross sectional weight individual level total sample rswt_rescs12 Cross sectional weight individual level nationally representative subsample rswt_natpn1012 Panel weight individual level total sample rswt_respn
15. rview to maintain a high response rate In order to make the publicly released data more user friendly we have merged the telephone interview data and the face to face interview data and have added a dichotomous variable Wmode indicating whether the interview was conducted face to face 1 or over the telephone 0 It should be noted that although the face to face telephone interviews shared the same goal the actual survey questions and hence variables were sometimes different When combining the telephone and face to face interview data we compared each question from both sources We adopted the same variable name if the two sources employed exactly the same question and otherwise kept two different variables 2 2 Combining Long and Short Forms of Questionnaires in Individual level Data In CFPS 2010 all the adult respondents were administered the same long form questionnaire In CFPS 2012 a short form adult questionnaire was added to supplement the long form questionnaire so that when a gene or core member was away from home some basic information about him her would be collected through proxy report by a local family member using the short form questionnaire Meanwhile we would try to track the non coresident member at his her current residence and if we succeeded a self report would be collected using the long form questionnaire Also non core family members were interviewed using the short form questionnaire In short CFPS respon
16. taset 00 0 0 ei eeeeseeseeeseeeseeeseeeseeeaeeeseeeneeeeeeeaees 8 3 7 Cognitive Function in the Adult and Child Datasets 0 ec eecceeeeeeseeeeeeeeee 8 Best Variables oreinen cece ra seeds nde ie EEEE ON tee den ANENE ened nie dee enters 10 Variable Coding i233 3c nih oie Miele il ie coal lal aan ate cen ie A 11 5 1 Occupational and Industry Codes 0 0 0 0 ceeecesesesseeseeeseeeseeseceeceseeeeeeeeaeseeeenaeeees 11 S2 Address Codes ozini ve ti ae eee eee ce cee ee ee AA 11 Weighting Variables uiaiia ne iii ee ene N ee ea 12 1 Introduction The CFPS baseline survey in 2010 consists of five basic datasets community family roster family adults and children In 2012 the CFPS carried out the first full scale follow up survey which targeted all 14 960 families and 57 155 members from the 2010 family roster who were considered CFPS gene members In addition individuals newly born to the 57 155 gene members would also be considered the CFPS gene members while individuals newly married to the existing CFPS gene members would be considered the CFPS core members By design gene members are followed regardless of their relationships with their original families whereas core members are followed as long as they maintain their family relations with gene members The 2012 questionnaires share similar structures to those in 2010 although some specific contents have been modified For example no community dataset is available in th
Download Pdf Manuals
Related Search