NARA expects to digitally release the 1950 population census schedules for researcher use on April 1, 2022, which is 72 years after the official 1950 census.  This is the 39th in a series of blog posts on the 1950 census.


Edwin D. Goldfield was a career statistician with the Bureau of the Census and was the "program coordinator" for the 1950 census. In 1992, he wrote "Innovations in the Decennial Census of Population and Housing:  1940-1990" for the Committee on National Statistics of the National Research Council.  We have posted a copy here, courtesy of the Census Bureau.


Mr. Goldfield identified eight innovations involving the 1950 census, some of which we have discussed in previous blog posts.  Most of the innovations were methods of evaluating the accuracy of the information received.


(1) Change in Enumeration of College Students.  Goldfield wrote:  "Prior to 1950, college students had been enumerated at their parental homes.  Beginning in 1950, college students have been enumerated where they are residing at or in the vicinity of the college.  The change was made not only because it was in accord with the "usual place of residence" rule but because it was expected to yield a more complete enumeration.  In planning for this revised procedure, the Bureau mailed questionnaires to all large educational institutions early in 1948.  This survey disclosed that many of the students did not live on the college campus; therefore educational institutions were not set up as separate enumeration districts.


(2) Experiment with Use of Mail.  Goldfield wrote:  "Although the 1950 census standard enumerative procedure was, like 1940 and all preceding censuses, door-to-door personal interview, there was included a test of a list/leave self-enumeration procedure whereby the enumerators listed addresses and left a questionnaire for households to fill in and mail back to the census district offices.  The testing thus included also the use of household questionnaires.  The success of this and later tests led to the use of mail techniques in parts of the country in 1960."


(3) Missed Persons Form.  Goldfield wrote: "This was a form that requested persons to fill it out and send it to the census office if they believed they had not been counted in the census.  Quantities of this form were printed, distributed in the form of cards, reproduced in newspapers, and announced on radio.  Later called "Were You Counted?" the forms have been used in subsequent censuses.  In 1980, the forms were credited with adding 67,000 persons to the census count, at an estimated cost of $267,000.  Although the numbers are not very large, the program is relatively cost-effective compared with other wrap-up procedures." 


(4) Post-Enumeration Survey (PES).  Goldfield wrote:  "This first PES consisted of two samples, one of areas to measure completeness of coverage of housing units, the other a list sample of enumerated households to measure completeness of coverage of persons within enumerated units and to evaluate content.  The probability sample of 3,500 small areas was recanvassed and the prelistings compared with the original census listings.  The list sample of 22,000 enumerated households was reinterviewed and the reinterviews compared with the original enumerations.  The 1950 PES estimated the gross census undercount to be 3.4 million people and the net undercount to be 2.1 million.   Most of the undercount was in dwellings that were themselves missed.  According to the PES, the estimated net undercount was 1.4 percent of the enumerated population.  The results are biased because people are missed even in the more intensive PES and there is likely to be a high correlation between being missed in the census and being missed in the PES.  Demographic analytical studies indicated that the true net undercoverage was substantially greater, amounting to perhaps 5 to 5.5 million people.  Post-enumeration surveys in later censuses were improved over this pioneering one, with efforts to increase independence between census and PES.  The fundamental problem is not totally overcome."


(5) CPS-Census Match.  Goldfield wrote:  "A matching study was conducted to determine the differences between the information in the monthly Current Population Survey and obtained for the same households in the census.  Census schedules were selected from 5,000 enumeration districts were microfilmed during the census processing, and the microfilm copy was matched with the April 1950 CPS questionnaires.  Response variability and response bias were greater for the census than for the CPS.  Labor force participation estimates were significantly hire in the CPS.  The CPS-census match was repeated in succeeding censuses."

  • The Current Population Survey (CPS) - begun in 1940 as the "Monthly Record of Unemployment" - is today the primary source of monthly labor force statistics in the United States but also asks varying supplemental inquires on matters such as child support, volunteerism, health insurance coverage, school enrollment, and more.


(6) Access to IRS Records.  Goldfield wrote:  "Presidential executive orders in 1944 and 1961 authorized access to Internal Revenue Service tax return records by the Census Bureau for statistical purposes.  Subsequent amendments to the census law (U.S. Code, Title 13) and to the IRS law (Title 26) confirm and supersede the executive orders."


(7)  Record Checks.  Goldfield wrote:  "Several record checks were made, on a sample basis, to evaluate the accuracy of selected content items.  For example, age reported in the census was checked against birth certificates, against 1920 census records, and against Social Security and Veterans Administration records.  Other record checks were on veteran status matched against VA files, income matched against Internal Revenue Service files, and wage and salary income and industry matched against Social Security records."


(8)  Public Announcement of Census Undercount.  Goldfield wrote:  "Several years after the 1950 census, the Bureau made public an estimate of the national net undercount in the census.  Later a retroactive estimate for 1940 was made and estimates have been produced for subsequent censuses.  The latest revised consistent estimates, based on demographic analysis, for the national net undercount rates are:  1940:  5.4 percent; 1950:  4.1 percent; 1960:  3.1 percent; 1970:  2.7 percent; 1980:  1.2 percent; and 1990:  1.8 percent."