Revisions Made to IPUMS-CPS

Below is a list of significant changes made to the IPUMS-CPS data, documentation, and extract system. The next regular IPUMS-CPS update is tentatively scheduled for September 1, 2010.

June 25, 2010. Posted new versions of all IPUMS-CPS samples.

  • The previous 2001 CPS has been replaced with the 2001 SCHIP expansion file, which contains data on 98,015 households and 206,639 persons--about 50 percent more households and 70 percent more persons than the previous version. The summary health insurance variables are now available for 2001, although other variable availability is unchanged.
  • A new variable, CPI99, gives the Consumer Price Index for all urban consumers (CPI-U) to convert any dollar amount into constant 1999 dollars. For more information on how to use this variable, see this page.
  • YRIMMIG, previously a three-digit code of intervalled years, is now a four-digit year that represents the latest possible year that the respondent could have immigrated to the U.S.
  • Five variables are now available on the source of unearned income (INCUNERN) in the 1966 and 1967 samples: GOTSS, GOTINT, GOTDIVID, GOTRENT, and GOTWELFR. Because data on the amount of unearned income does not distinguish among the sources of unearned income, these variables contain information available nowhere else for these samples. However, users are advised that these variables may be inconsistent with INCUNERN.
  • PROPTAX is now available for the 1992-2003 samples.
  • WKSUNEM1 and WKSUNEM2 contained incorrect data in the 1988 sample. They now contain the correct codes.
  • Miscellaneous clarifications and corrections to documentation.

March 2, 2010. Posted new versions of all IPUMS-CPS samples.

  • New summary health insurance variables are available. Constructed by the University of Minnesota's State Health Access Data Assistance Center (SHADAC), these variables smooth over several temporal inconsistencies in the CPS's measurement of health insurance and make several improvements, most notably with the imputation of verified insurance data for the 1988-1999 period. They largely mirror the measurement of health insurance in the 2008-onward American Community Survey and therefore provide a convenient set of summary variables to easily analyze trends in insurance coverage across time. IPUMS-CPS thanks SHADAC staff for providing these variables and for their assistance in developing documentation.
  • New variables have also been posted for the 2009 CPS that were not available when the data were originally released. They include the alternative poverty measures along with a new variable for 2009 (STIMULUS) describing individuals' income from the economic stimulus package passed in 2008. Finally, the after-tax variables are now available as well: Newly-released household variables are STAMPVAL, HOUSRET, and PROPTAX; Person variables are ADJGINC,DEPSTAT, EITCRED, EMCONTRB, FEDRETIR, FEDTAX, FEDTAXAC, FICA, FILESTAT, MARGTAX, PMVCAID, PMVCARE, STATETAX, STATAXAC, and TAXINC. Additionally, STAMPVAL is no longer censored by the Census Bureau at $9,999; the new topcode is $19,999. CAPGAIN and CAPLOSS are no longer part of the alternative poverty model and are not included in the 2009 CPS PUMS file.
  • Alternative poverty measures are also now available for the 2009 CPS. These measures were developed by the Census Bureau to incorporate information on medical out-of-pocket costs, geographic differences in the cost of living, and alternative methods of adjusting poverty thresholds for inflation.  For details, see the alternative poverty measures page.
  • The main education variable has been streamlined to parallel as much as possible its measurement in IPUMS-USA data. EDUCREC has been replaced with EDUC, which contains all information in the original variables. HIGRADE contains additional detail on grade attendance (EDUC, in contrast, informs only about completion) and has been retained. Although EDUC99 is redundant with EDUC, it is also retained because it contains simpler coding that may be of use to users analyzing only 1992-onward data.
  • Negative replicate weight values (REPWT and REPWTP) are no longer automatically recoded to zero, and more information about working with replicate weights in the CPS is now available on our replicate weights page.

September 30, 2009. Added replicate weights, recently released by the Census Bureau, to the 2005-2009 CPS samples. Replicate weights can be used to estimate standard errors of sample statistics directly from the data. For more information, please see the REPWTP variable description. Users should be aware that selecting replicate weights will add 160 additional variables to extracts.

September 11, 2009. Posted new data for 2009. Noteworthy changes include the following:

  • Seven new disability variables: DIFFHEAR, DIFFEYE, DIFFREM, DIFFPHYS, DIFFMOB, DIFFCARE, and DIFFANY. These mirror the 2008 American Community Survey data, which will be released in late October.
  • Industry codes (IND) now use the 2007 Census Bureau classification.There are no major substantive differences in the meanings of the categories, although some labels have changed. Only one code is different: industry codes of 6675 (Internet publishing and broadcasting) in the 2003-2008 CPS data have been shifted to 6672 (Internet publishing and broadcasting and web search portals) in the 2009 CPS data.
  • Documentation has been updated for the health insurance variables PRIVWHO1, PRIVWHO2, GRPWHO1, and GRPWHO2, which give the line number (LINENO) of the policyholder if one is covered by a privately purchased or employer-based insurance policy. Although previous documentation stated otherwise, the universe for these variables does not include policyholders themselves, so these variables will not have the same value as LINENO for policyholders.

All other variables from the 2008 CPS remain the same, with the exception of non-cash benefits and tax variables as well as alternative poverty measures. These will be released in the IPUMS-CPS shortly after the Census Bureau releases them, most likely in early 2010.

June 19, 2009. Posted new data for all years with thirteen new poverty variables:

  • CONTPOV, available from 1968 onwards, gives one's total family income (FTOTVAL) expressed as a percentage of the poverty threshold for that family (CUTOFF). It is comparable to the POVERTY variable in IPUMS-USA.
  • Twelve experimental poverty measures are now available from 2002 onwards. These measures were developed by the Census Bureau to incorporate information on medical out-of-pocket costs, geographic differences in the cost of living, and alternative methods of adjusting poverty thresholds for inflation. For details, see the alternative poverty measures page.

The new versions also include three corrections to the FAMUNIT coding. In each of the 1962-1994 samples, FAMUNIT has generally changed for less than 0.5 percent of the cases. In each of the 1995-2008 samples, FAMUNIT has changed for 1-2 percent of the cases. Because FAMSIZE and NFAMS incorporate information from FAMUNIT, all persons in households where someone has a new FAMUNIT code may have different values for FAMSIZE and NFAMS, even if their own FAMUNIT code has not changed. In practice, though, very few persons have different FAMSIZE and NFAMS values in the 1962-1994 data. Details follow:

  • In households where the head is not the first person record, the first person record and any persons linked via MOMLOC, POPLOC, or SPLOC to the first person record were grouped with the primary family (FAMUNIT = 1) because of a programming error. From 1976 to 1993 and from 2006 to 2008, the head was always listed first in the household, and this particular correction does not affect these samples. In other samples, though, unrelated individuals are occasionally listed first; and they and their own relatives were mistakenly given FAMUNIT codes of 1.
  • FAMUNIT codes of 1 are now reserved exclusively for persons with a RELATE code of less than 1100 and any individuals linked to them by birth, marriage, or adoption. Thus household fragments now receive FAMUNIT codes of 2; they received codes of 1 before. This correction affects all samples.
  • A programming error included unmarried partners of the head in the primary family (FAMUNIT = 1). This particular correction affects only the 1995-2008 samples, since the unmarried partner relationship category is available only in these years.

Finally, there are minor changes to three other variables:

  • The original 1963 CPS data include some nonsense codes for total family income (FTOTVAL). This variable is no longer available for 1963 from IPUMS-CPS.
  • Where the original CPS data has both the head and the head's spouse as the same SEX, the IPUMS imposes an edit that codes the head as male and the spouse as female. The correction allows this edit to be carried out for one household in the 1973 CPS where it did not occur before.
  • The categories of EDUC99 have been altered to conform to the recent IPUMS-USA revision of this variable. Codes of 0-11 remain identical, the code of 12 is now an empty placeholder, and former codes of 12-17 have been shifted to 13-18. This change applies to the codes only; there has been no substantive change in the data.

May 19, 2009. Posted new data for 2008. In the initial 2008 CPS release, several variables were withheld until the release of official reports on alternative income and poverty measures. These variables have since been released by the Census Bureau and are included in the current IPUMS-CPS release. Newly-released household variables are STAMPVAL, HOUSRET, and PROPTAX. Person variables are ADJGINC, CAPGAIN, CAPLOSS, DEPSTAT, EITCRED, EMCONTRB, FEDRETIR, FEDTAX, FEDTAXAC, FICA, FILESTAT, MARGTAX, PMVCAID, PMVCARE, STATETAX, STATAXAC, and TAXINC.

February 23, 2009. Posted new versions for 1962-1967. The only variables affected were OCC, IND, OCCLY, and INDLY. Persons not in the universe previously had values of either 0, 98, 99, or 999. All persons not in the universe now have a value of 99 in all of these variables.

October 9, 2008. Posted new versions for 1963-2008 to include the original CPS variable FTOTVAL and make improvements to the OCCLY variable for 2003-2008.

September 16, 2008. Posted new data for 2007 and 2008:

2007: In the initial 2007 CPS release, several variables were withheld until the release of official reports on alternative income and poverty measures. These variables have since been released by the Census Bureau and are included in the current IPUMS-CPS release. Newly-released household variables are STAMPVAL, HOUSRET, and PROPTAX. Person variables are ADJGINC, CAPGAIN, CAPLOSS, DEPSTAT, EITCRED, EMCONTRB, FEDRETIR, FEDTAX, FEDTAXAC, FICA, FILESTAT, MARGTAX, PMVCAID, PMVCARE, STATETAX, STATAXAC, and TAXINC.

2008: All variables for the 2008 CPS are the same as those available in the 2007 CPS, except for the non-cash benefits and after-tax variables mentioned above, which are again being withheld for several months. These will be released in the IPUMS-CPS shortly after the Census Bureau releases them, most likely in early 2009.

November 12, 2007. Released a new data extract system with the Attach Variables feature. This allows researchers to create variables specifying the characteristics of respondents' spouses, mothers, fathers, and household heads. New online help for the Case Selection and Attach Variables features is also available via hyperlinks from the data extract system. In addition, we have improved the look and functionality of our variable descriptions and codes pages.

October 24, 2007. Posted new data for 2005, 2006, and 2007:

2005: In April 2007, the Census Bureau released a new 2005 data file, with revisions to the health insurance variables. The Census Bureau documents this in a User Note. The new data are part of the current IPUMS-CPS release.

The most significant changes are to the following employment-based insurance variables: GROUPWHO1 and GROUPDEP (1.3% of cases modified); and GROUPTYP (.9% of cases modified).    Another group of health insurance variables (PHINSUR, PHIOWN, VERIFY, COVERGH, COVERPI, and EMCONTRB) had approximately .7% of their values changed. The rest of the changes to the health insurance variables in 2005 amount to less than .5% of cases modified, except for PAIDGH and GROUPOUT, which were not changed at all.

In addition, the Census Bureau modified the following variables (the percent of cases changed to the 'Not applicable' category appears in parentheses after each variable name): UNION (3.7%), EARNWEEK (3.7%), and HOURWAGE (35.3%). Additional information on these changes will appear in a future revision note.

2006: In the initial 2006 CPS release, several variables were withheld until the release of official reports on alternative income and poverty measures. These variables have since been released by the Census Bureau and are included in the current IPUMS-CPS release. Newly-released household variables are STAMPVAL, HOUSRET, and PROPTAX. Person variables are ADJGINC, CAPGAIN, CAPLOSS, DEPSTAT, EITCRED, EMCONTRB, FEDRETIR, FEDTAX, FEDTAXAC, FICA, FILESTAT, MARGTAX, PMVCAID, PMVCARE, STATETAX, STATAXAC, and TAXINC.

In addition, the latest release of the 2006 CPS includes revisions to the health insurance variables (see the Census Bureau's User Note). The largest changes are to the following employment-based insurance variables: GROUPWHO1 and GROUPDEP (1.3% of cases modified); and GROUPTYP (1% of cases modified). Another group of health insurance variables (PHINSUR, PHIOWN, VERIFY, COVERGH, and COVERPI) had approximately .7% of their values changed. The rest of the changes to the health insurance variables in 2006 amount to less than .5% of cases modified, except for PAIDGH, GROUPOUT, CHAMPVA, and MOCAID, which were not changed at all.

The Census Bureau also modified the following variables (the percent of cases changed to the 'Not applicable' category appears in parentheses after each variable name): UNION (3.7%), EARNWEEK (3.7%), and HOURWAGE (35.3%). Additional information on these changes will appear in a future revision note.

2007: For the 2007 March CPS, the Census Bureau dropped the Hurricane Katrina questions from the survey (IPUMS-CPS variables KATEVAC, KATEVAC2, and KATPRIOR) as well as the variable indicating country of previous residence (COUNTRY). All other variables for the 2007 CPS are the same as those available in the 2006 CPS, except for the non-cash benefits and after-tax variables mentioned above, which are again being withheld for several months. These will be released in the IPUMS-CPS shortly after the Census Bureau releases them.

May 1, 2007. Posted new version of the 1994 sample: a correction was made to the MBPL variable.

February 2, 2007. Posted new versions of the 2005 and 2006 samples: a correction was made to the CUTOFF variable.

December 1, 2006. Posted new data for 1963, 1966, and 1967. All weighting variables in these samples were adjusted by a constant. The samples now produce accurate population estimates. More information is available on the sample notes page.

October 19, 2006. Posted new data for 2004, 2005, and 2006:

2004: In the initial March 2004 release, several variables were withheld until the release of official reports (see User Note #1 on the Bureau of Labor Statistics' CPS downloads page). These variables have since been released by the Census Bureau and are included in the October 2006 IPUMS-CPS release. Newly-released household variables are STAMPVAL, HOUSRET, and PROPTAX. Person variables are ADJGINC, CAPGAIN, CAPLOSS, DEPSTAT, EITCRED, EMCONTRB, FEDRETIR, FEDTAX, FICA, FILESTAT, MARGTAX, PMVCAID, PMVCARE, STATETAX, and TAXINC.

In addition, the Census Bureau released alternative weight variables, which have been included in the IPUMS-CPS as HHWT04 and PERWT04 (see our CPS Weights note for more information).

2005: In August 2006, the Census Bureau released a new 2005 data file. The new dataset includes all variables listed above, plus FEDTAXAC and STATAXAC. In addition, changes were made to the weight variables and several health insurance variables. The Census Bureau documents this along with their March CPS downloads (Endnote 16). The IPUMS-CPS variables affected by these changes are HHWT, PERWT, PAIDGH, HIMCAID, HIMCARE, HICHAMP, PHINSUR, PHIOWN, CAID, CHAMPUS, CHAMPVA, MILITVA, GROUPOWN, GROUPDEP, PRIVOWN, PRIVDEP, GROUPTYP, PRIVTYP, GRPWHO1, GRPWHO2, PRIVWHO1, PRIVWHO2, SCHIP, VERIFY, GROUPOUT, PRIVOUT, OUT, COVERGH, COVERPI, KIDPRIV, and KIDCAID. The weight variables were recalculated for the entire sample. For PRIVTYP, about 25% of people who were coded 'Family plan' or 'Self only covered' in the previous version of the dataset are coded 'NIU' in the new data. Aside from the weights and PRIVTYP, the changes to any given variable affect less than 1 percent of person records in the 2005 CPS.

2006: For the 2006 March CPS, several migration questions were added to the survey in order to assess the impact of Hurricane Katrina. The new IPUMS-CPS Katrina variables are KATEVAC, KATEVAC2, and KATPRIOR. In addition, several new veteran variable were introduced. These variables expand the information collected on a veteran's period(s) of service. For more information, see the variable descriptions for the new veteran variables: VET1, VET2, VET3, VET4.