Calibrating census microdata against a gold standard employment

Calibrating census microdata against a gold standard (employment survey): women in the workforce, Mexico 1990 and 2000 * * * Robert Mc. Caa, Albert Esteve , Rodolfo Gutierrez and Gabriela Vasquez, Minnesota Population Center paper at: www. hist. umn. edu/~rmccaa/mxflfp. doc Calibrate, v. 1864. a. trans. . to graduate a gauge of any kind with allowance for its irregularities. The Oxford English Dictionary Online (Oxford: Clarendon Press, 2001) www. ipums. org/international 1

Today’s Presentation » 1. The Infomercial: IPUMS-International a. Preserving the world’s census microdata b. And making them usable » 2. The paper: Calibrating censusmicrodata a. Census vs. employment surveys b. Female labor force participation(FLFP) c. vs. ENEU (“national”urban survey): 1990 & 2000 d. vs. ENE (national survey): 2000 e. Conclusion: Mexican censusmicrodata on FLFP are better than commonly thought www. ipums. org/international 2

** Infomercial ** www. ipums. org/international 3

IPUMS-International goals » 1. Inventory the world’s census microdata a. historical b. contemporary » 2. Preserve endangered microdata and metadata a. contract preservation with repositories b. archive validated copies * * * » 3. Integrate census microdata and metadata of selected countries on all continents using UN, ECE, and other standards » 4. Disseminate resulting database without charge with full access to all who agree to non-disclosure www. ipums. org/international 4

IPUMS-International partners by stage final stage (data in development/dissemination) = darkest middle stage (signed agreement) = medium green first stage (verbal agreement, signing pending) = lightest www. ipums. org/international 5

I N T I E P R U N A M T S I O N A L For census 2000 (and earlier, historical censuses): 42 OSI (official statistical institutes) agree to produce a new statistical product: scientifically anonymized, integrated census microdata samples made up of unidentifiable individuals. . . OSI Partners by Region & Year Joined Year Africa Amer Arab Asia Euro Total 1999 0 2 0 0 1 3 2000 1 3 0 1 2 7 2001 1 5 0 2 3 11 2002 2 16 1 2 5 27 2003 5 16 1 2 17 42 2004 ? 16 ? +1? 17 ? www. ipums. org/international 6

L A T I N I P U M A S M E R I C A Round 1 partners: 1960 INEGI-Mexico 1960 1970 plus new 10% samples: 1970 DANE-Colombia 1964 1972 IBGE-Brazil 1960 1970 Round 2, 2003 -2008: census round: 1960 s 1970 s Argentina 1960 1970 Bolivia 1976 1992 Chile 1960 1970 Costa Rica 1963 1973 Dominican Republic 1960 1970 Ecuador 1962 1974 El Salvador 1961 1971 Guatemala 1964 1973 Honduras 1961 1974 Nicaragua 1971 1995 Panama 1960 1970 Paraguay 1962 1972 Peru 1981? Puerto Rico 1960? 1970 Venezuela www. ipums. org/international 1961 1971 1980 1990 & 1993 1991 2000 2010 2000 1980 s 1980 2001 1982 1984 1981 1982 1992 1981 1988 1990 s 1991 2000 s 2001 1992 2000 1993? 1990 2002 1994 2001 2002 1980 1982 1993 1980 1981 1990 1992 2003 1990 2000 2002 1985 1980 2004 2001 2002 2000 2001 7

2004? -2009? I P U M S E U R O P E Round 1: 3 partners: INSEE-France 1962, 1968, 1975, 1982, 1990, 1999 CSO-Hungary INE-Spain 1970, 1980, 1990, 2000 1981, 1991, 2001 Round 2, 2004 -2009: +14 Signed: Austria, Belarus, Bulgaria, Czech Republic, France, Germany, Greece, Hungary, Netherlands, Portugal, Romania, Slovenia, Spain, and the UK 9 Negotiating: *Belgium, Finland, Ireland, Italy, *Poland, *Russia, Switzerland, Turkey, Ukraine *= included in proposal; likely to sign before project www. ipums. org/international 8 begins

IPUMS-International principles: » 1. Respect absolute anonymity and confidentiality » 2. Preserve all original data, except adjustments to insure » » » privacy (top codes, blurrings, masking, re-ordering, etc. ) 3. Harmonize codes using international standards occupation: ISCO, HISCO (detailed, general) education: ISCED “ “ family: IPUMS, etc. “ “ 4. Enhance with constructed variables (momloc, etc. ) 5. Disseminate free of charge to reseachers who need the data and agree to abide by conditions of use (please see Table 5) » Of 346 applicants, 197 accepted; denial of access rate = 43% » Usage by country (% of projects): Mexico (43), USA (38), Kenya (31), Colombia (29), Vietnam (26), China (9— 1982 census, May-Aug 2003). www. ipums. org/international 9

» Integrate (harmonize), not standardize I N I T P E G U R M A Si T E S 1. retain all original detail 2. harmonize every digit » How is this possible? Composite codes (multiple digits, 111) Not serial (1, 2, 3, . . ) (see next slide) » Why? Researcher confidently understands uses as much detail as needed www. ipums. org/international 10

Composite coding scheme: employment status integrated codes Codes in original data www. ipums. org/international 11

IPUMS-International Employment Status variable comparability discussion: Mexico 1990, 2000 “In 1990, the employment status question refers to ‘Principal Activity’ and therefore under-reports secondary economic activity by students, housewives, family-workers, the semi-retired, and others. “The 2000 Census sought to overcome deficiencies in reporting work status for people whose primary activity was not work (students, housewives, retirees, etc. ), but who in fact were working according to international definitions. A second question, introduced for the first time in 2000, sought to capture this secondary economic activity. For strict comparability with earlier Mexican censuses, this recovered activity (codes 1101 -1106) should be considered ‘inactive’. " www. ipums. org/international 12

2000 census--two questions on LFP: 1: 1: “Lastweek, week did (NAME). . . ” (NAME). . . ? ” 2: “Besides (. . . ), . . . ” did (NAME). . . ? ” Question 1: Question 2: Last week (Name): Did you help in a Did you work? 27. 5% family business? Sell some product? Had work? 0. 4 Look for work? 0. 3 Looked? 0. 3 Make some product to sell? Q. Are 1&2: you a combined student? student/ wrkd 0. 5 Help on a farm or with livestock? housewife/ wr 3. 7 housewife? retired/ wrkd 0. 0 Or in exchange for retired? pay did you do other/ wrkd 0. 4 permanently some other noincapacitated? reply/wrkd 0. 0 www. ipums. org/international activity? Did you not work? 13

** Web-site ** www. ipums. org/international 14

** Paper ** www. ipums. org/international 15

Today’s Presentation, part 2 » 1. The project: IPUMS-International a. Preserving the world’s census microdata b. And making them usable » 2. The paper: Calibrating censusmicrodata a. Census vs. employment surveys b. Female labor force participation c. vs. ENEU (“national” urban survey): 1990 & 2000 d. vs. ENE (national survey): 2000 e. Conclusion: Mexican censusmicrodata on FLFP are better than commonly thought www. ipums. org/international 16

Calibration test Mexico 1990, 2000: FLFP census microdata vs. employment surveys Employment surveys: date from the late 1980 s; many probing questions finely tuned instrument administered by trained interviewers Urban (ENEU)—quarterly from 1987, 16 cities in 1990, rising to 47 in 2000 --lacks national coverage; National (ENE)—from 1988; annual since 1995 Census microdata : strength: national coverage back to 1960 weakness: untrained interviewers, one question on LFP omits many working women, particularly informal workers Purpose of paper: calibrate census microdata w/ employment surveys www. ipums. org/international 17

! r ! ! n fo o o esmicrodata samples Table 1. Selected s l g p n i am 00 m of Mexico, 1960 2000 0 o s 2 C 0% & 1 0 Year Type Sample Size % pop. 9 w 9 e 1 , N 70 1960* Census 502, 702 1. 5 19 1970* Census 1980 Census 480, 265 1. 0 No sample available due to earthquake damage 1990* Census 802, 774 1. 0 1990, ENEU (urban survey) 172, 233 0. 2 2000 * Census 10, 099, 182 2000 ENEU (urban survey) 562, 471 2001 ENE (national survey) 588, 912 (*integrated in IPUMS-International) www. ipums. org/international 10. 0 0. 6 18

The problem (table 2) Mexico’s “global” female labor force participation rate (12 -64 years) microdata 1990 2000 survey (ENEU): 34. 8% 43. 3% national census: 20. 6% 32. 9% 14. 2 10. 4 A solution Control for survey (ENEU) sampling frame: 16 cities in 1990 survey (16 cities): 34. 8% 41. 7% census (16 cities): 29. 0% 40. 2%* 5. 8 1. 5 www. ipums. org/international * includes responses to LFP questions 1 (“activity”) & 2 (“verification”). 19

Table 3. Urban Females, 1990 (aged 12 -64) Total Education Less than 6 years Completed primary Completed middle Post-middle (10+) Marital Status Married (all types) Not in union Structure Survey Census 62, 248 63, 929 FLFP Srvy 34. 8 Rates Cnss 29. 0 20. 1 21. 1 37. 9 42. 2 20. 9 34. 7 20. 4 23. 9 21. 7 34. 8 24. 3 19. 3 27. 6 31. 3 53. 1 48. 2 51. 8 50. 4 49. 6 27. 7 21. 3 41. 4 36. 9 www. ipums. org/international 20

Table 5. Females 2000: Urban (limited to same 16 cities as 1990) Total Education Less than 6 years Completed primary Completed middle Post-middle (10+) Marital Status Married (all forms) Not in union Structure FLFP Rates Survey Census Srvy Cnss 124, 051 1, 073, 222 41. 7 40. 2 14. 9 30. 8 28. 5 25. 8 15. 935. 3 28. 732. 1 19. 747. 0 36. 051. 1 51. 9 48. 1 52. 935. 1 34. 0 47. 148. 9 47. 3 www. ipums. org/international 31. 0 30. 1 41. 0 51. 9 21

Table 5. Logistic Regression: Source (Females 1990, 2000; same 16 cities as in ENEU 1990) Female labor force participation ENEU (indicator) vs. Census Variable B 1990: -. 2212 2000: . 0860 S. E. Wald df Sig R. 0135 267. 2291 1. 0000 -. 0412. 0067 163. 1781 1. 0000. 0100 Exp(B). 8015 1. 0898 Model source effect taking into account age, marital status and education. Interpretation If for both sources weights are considered correct and slight structural differences are taken into account: 1990 census under-reported 20% of FLFP. 2000 census “over-reports” FLFP by 9%. www. ipums. org/international 22

Today’s Presentation » 1. The project: IPUMS-International a. Preserving the world’s census microdata b. And making them usable » 2. The paper: Calibrating censusmicrodata a. Census vs. employment surveys b. Female labor force participation c. vs. ENEU (“national” urban survey): 1990 & 2000 d. vs. ENE (national survey): 2000 e. Conclusion www. ipums. org/international 23

Table 6 a. Females 2000: National Total Education Less than 6 years Completed primary Completed middle Post-middle (10+) Marital Status Married (all forms) Not in union Structure FLFP Rates Survey Census Srvy Cnss 212, 890 3, 431, 891 39. 8 32. 9 22. 0 38. 0 16. 5 23. 5 27. 630. 9 30. 036. 4 16. 841. 2 25. 652. 4 23. 3 25. 0 36. 8 49. 9 54. 7 54. 836. 3 27. 6 www. ipums. org/international 45. 3 45. 243. 9 39. 3 24

Table 6 b. Females 2000: National Limited to municipios in ENE Total Education Less than 6 years Completed primary Completed middle Post-middle (10+) Marital Status Married (all forms) Not in union Structure Survey Census 22. 0 38. 0 16. 5 23. 5 22. 330. 9 29. 636. 4 30. 841. 2 27. 352. 4 FLFP Rates Srvy Cnss 39. 8 35. 7 25. 4 26. 7 40. 5 55. 5 54. 7 54. 336. 3 30. 1 www. ipums. org/international 45. 3 45. 743. 9 42. 2 25

Postscript * * * * www. ipums. org/international 26

Fig 1. LFP by sex and marital status Mexico 1990 and 2000 (national figures) Females 1990 Males www. ipums. org/international 2000 27

Marriage and education strongly affect FLFP (Mexico 1990 and 2000, national figures) Not 1990 Married www. ipums. org/international 2000 28

Reflections Mexican census microdata may be more informative, than commonly thought—even about FLFP Mexican census microdata on FLFP display remarkable coherence in time and space “Chorus of calamity” on Mexican FLFP may overlook enormous changes in education weakening power of patriarchy over married women real advances of women in the workforce 2000 microdata tell the story Calibrate me! weigh strengths and weaknesses of sources. www. ipums. org/international 29

Thank you * * * * rmccaa@umn. edu www. ipums. org/international 30