Recent occupation concepts applied to historical U S

Recent occupation concepts applied to historical U. S. Census data Peter B. Meyer US Bureau of Labor Statistics (but none of this represents official measurement or policy; Views and findings are those of the author not the agency) RC 33 2008, Naples, Sept 3, 2008 Outline 1. Brief history of U. S. Census occupations 2. Can standardize to recent definitions? 3. Issues: working wives, native Americans, slaves, children, others 1

An ideal occupation variable. . . Would extend over a long period using the same category system That enables comparison over time to see causes and effects of: q q Unionization Licensing Technological change And more Examining individuals, holding occupation constant; Or within and between occupations Can we assign current occupation concepts to past Censuses? Efforts include occ 1950, covering 1850 -present (by Sobek) and occ 1990, covering 1960 -present (by Meyer and Osborne). These put harmonize the occupations for those respondents listed as having occupations at the time. But: Who was counted as having an occupation? How did that change? 2

U. S. Census occupation history (the historical process recording the data) n Census starts in 1790, for political districting and taxation. q n 1850: Free male respondents first asked for their “Profession, occupation, or trade” by U. S. marshals. Occupations were not categorized. q n n n n n Most Native Americans (Indians) were not counted. 1850 s: International conferences on occupation collection in Censuses 1860: All free respondents asked for occupation; household head, usually male, is counted distinctively. 1870: Slave category disappears. 1870: Classification of occupations into 338 categories. Since 1870: The category system changed every decade since then. 1880: Data collectors now political appointees not judicial aw enforcers 1902 -10: Now permanent civil service bureau collects data and categorizes into occupations. Quality improves. 1940: Switch to “labor force” definitions and concepts, de-gendered Since 1940: Relatively stable definitions and practices. Since 1970: With each new system “dual-coded” data are now available. 3

Many working women did not have a recorded occupation in 1900 n n My main source: Bose (2001) Many women in 1900 who would now count as employed did not count as employed in the 1900 Census. q q q Working on family farm (est. 16. 9% of US women age 15 -64) Taking in boarders; taking care of other people’s children (3. 9%) Outwork, e. g. sewing at home, or running a shop (3. 1%) Her corrected figures are not far from the comparable measures available from Australia which used two similar occupation concepts Other sources: Sobek (1997), Goldin (2001), Abel and Folbre (1983), Deacon (1985) n 4

More changes in boundary of who has an “occupation” n n Indians – slow transition into occupation, since the early Census (for more see Snipp and Hacker) Slaves, till 1870 – not recorded as having an occupation Children – minimum age for recorded occupation jumps around, varying from 10 to 16 until 1910 People who were disabled, retired, students, institutionalized, or seasonal workers were less likely using post-1940 “labor force” definitions to have an occupation. (Anderson, 1980, p. 24) 5

In other cases, boundary did not change n n n n Unemployed men – May report an occupation Non-citizens, border-crossers -- Household location matters; employer's location does not matter. Working illegally or avoiding tax - Can have a Census occupation, without a legal implication, but until 1880 Census enumerators were law enforcement officials. Volunteers or hobbyists -- Not counted as in an occupation unless they report as self-employed Apprentices – yes, if paid. If not paid, might be conceived of as students. Homeless; traveling; or can't locate -- Can have an occupation based on information from others or remote location. Refused to answer -- Historically a small category. Enumerator may receive information from others. Possibly a growing category. 6

Rough magnitude of boundary changes 7

Tentative conclusions n n n Can potentially do a good job matching current job categories back to 1970 using “dual-coded” data sets. It is realistic to apply current occupation categories back to 1940 Before 1930, might adjust for adult women in homebased economy and Indians Before 1870, occ data was not categorized and there were slaves In 1850 maybe only 35% of population would have a Census occupation; now over 60%. With more research, it is feasible to get better at this. 8

The information coders have 9

Tangential motivation: Example of standardizing n n In 1960 Census, “lawyers and judges” was one category Later, “lawyers” and “judges” were separate We can impute which 1960 ones are judges for standardizing comparisons to later data. In 1970 -1990 these variables predict who’s called a judge: q q q q Employed in public sector, especially in state government Older Employed in state government High salary income Low business income Educated less than 16 years Employed at time of survey 10

Information used when coding n “what kind of work" “most important activities or duties" employer name “what kind of industry” home city and state years of education age sex n before 1994, had income too n This information is available when choosing “industry” and “occupation” n n n • Tens of thousands of job titles are mapped to a code in a reference book they have, if industry matches what is expected. • Some cases may be "autocoded" by software; coder checks. • Coder with two years experience should assign 94 codes per hour with 95% “accuracy”, which is checked. • Cases not meeting the rules go to “referralist” (specialist) 11 • They have 9+ years of experience.

Problems faced by referralists n Having to hurry n Ambiguity; too little information from respondent q q q n “computer work" for “kind of work” "water company" for industry or employer "surveyor" occupation "boot" vs "boat" in handwriting exaggeration (example: dot com businesses) Referralists confer with each other routinely, but sometimes make different choices from one another 12

Can standardize occupation categories over time that look like recent ones? IPUMS (Matt Sobek) defined occ 1950 for 1850 -present n Meyer and Osborne (2005) defined 1990 -based classification from 1960 to present. Plan to improve, correct, and extend that. n Let us look at recent Census practices occupation variable. n Census “coders” in a single location assign 3 -digit industry and occupation codes n They follow carefully documented practices. n I interviewed four experienced ones. n They work just north of Louisville, in Jefferson, Indiana 13

Census year Notes; the question Census asked 1790 -1840 Censuses No specific occupation question. Census collected by U. S. marshals. 1850 "Profession, occupation, or trade of each male person over 15 years of age" and "Number of slaves" 1860 "Profession, occupation, or trade of each person, male and female, over 15 years of age" and "Number of slaves" (without work details) 1870 Slave category disappears. "Profession, occupation, or trade of each person, male or female" 1880 Census enumerators are no longer marshals. "Profession, occupation, or trade of each person, male or female" over age 10 and, of those, months unemployed during the census year. Separately, months at school. 1890 ""Profession, trade, or occupation" and, of those, months unemployed during the census year 1900 Hunt (1909) refers to 1893 international standard. Bose (2001) estimated dramatic undercount of women in 1900 and presumably before. 1910 1920 Permanent Census Bureau started 1902. Most Indians now included, but most were not included in 1910. 14

Category of persons Relation to occupation categories Indians (Native Americans) Excluded, by Constitution; over time, most descendants have become U. S. citizens or showed up in Census data, e. g. could have occupations Slaves Counted separately; did not have occupations 1860; then that disappears legally and as a data category Adult women Not counted in 1850. Occupations dramatically undercounted, per Bose (2001), especially if it were informal and through an ethnic enclave. By 1900 Census definition 22. 5% employed; by 2000 definition, 46. % did. Children or students Minimum age for an occupation has jumped around from 15 to 10 to 14 to 16. "Student" is recorded in some years but is not by usual definition an occupation. Retired, unemployed Sometimes these are given categories in the occupation category system Non-citizens, border-crossers Household location matters; employer's location doesn't. Institutionalized In prison, hospital, disabled, charity shelter. Can have an occupation but usually don't. Working illegally or avoiding tax Can be categorized in a Census occupation. No link to law enforcement. However before 1880, Census enumerators were law enforcement officials. Volunteers or hobbyists Not counted as in an occupation unless they report as self-employed Apprentices Yes, if paid. If not paid, might be conceived of as students. Homeless; can't locate; traveling Can have an occupation based on information from others or remote location. Refused to answer Historically a small category. Can have an occupation if enumerator receives information from others (I believe). Possibly a growing category. 15

Newly Imputed Judges Using that information in a logistic regression, can assign some to new “judges” category. 1960 1970 1980 1990 Lawyers 1971 2570 5082 7603 Judges 82 123 298 331 Seems to be 80% accurate. In this case, can estimate a time series of earnings separately for judges and lawyers more accurately. Let us go back to look at the earlier period to see how far it is from this data-processing world. 16