Recent occupation concepts applied to historical Census data

Peter B. Meyer

US Bureau of Labor Statistics, Office of Productivity and Technology[1]

March 1, 2008

Preliminary and incomplete. Please do not quote or cite yet.

Introduction

Key aspects of people's lives, and their economic effect on others,follow from their occupations. So when analyzing changes over time in earnings distributions, licensing, union membership, and technological sophistication of a population, it can be analytically helpful to make comparisons within and between occupations, rather than directly comparing individuals without taking their jobs into account. National census data have a wide spectrum of respondents and their jobs so detailed comparisons over time are possible. The purpose of this paper is to trace the general boundaries of the occupation data available over time from the United States censuses, and to highlight some conceptual issues in comparing occupation categories over the history of the census data.

Census of Population practices and occupation collection overview

A U.S. Census of Population has been conducted each ten years since 1790, because the U.S. Constitution required there be a count of persons by locality to help define representative political districts and fair taxation across the states. The census data would support changes decennially (every ten years) so that as citizens moved and grew in number, the political districts electing the House of Representatives would adapt. The framers of the Constitution also required that any direct taxes imposed by Congress would fall fairly and proportionately on the populations of the states. For the first century, the counts were collected by marshals, from the national law enforcement arm of the courts.

The Constitution did not call for collecting occupation or industry data, but it was repeatedly suggested that it should (including by James Madison). In 1820, there was a question about how many persons, including slaves, worked in three general economic sectors, agriculture, commerce, or manufactures. These are by current conceptions industry information, not occupational information. In the 1840 census, respondents were to be asked the same questions within these seven (industry) categories: mining; agriculture; commerce; manufactures and trades; navigation of the oceans; navigation of canals, lakes, and rivers; and learned professions and engineers. Again it appears slaves were included (Hunt, 1909). Starting in 1850 and ever since then, many respondents were asked more precisely for their profession, trade, or occupation.

In the 1850 census, "profession, occupation, or trade" was requested but only of free males over 15. 323 occupations were presented in the results, summarized under ten general headings: commerce; trade; manufactures; mechanic arts and mining; agriculture; law, medicine, and divinity; other pursuits requiring education; government, civil service; domestic servants; and other occupations.

The 1860 census was very similar to 1850, but now all free persons over 15, now including females, were asked an occupation. An alphabetical list of 584 occupations was presented in the results, according to Hunt (1909).[2] The 1850 and 1860 results were poor, in some unspecified sense, in Hunt's view. (p470)

In 1870, General Francis Walker took charge as the Superintendent of the Census. Marshals continued to do the collection. The inquiry of occupation now went to all person over 10 years of age. Starting with the 1880 census, marshals no longer collect the data. The President appointed a Superintendent of the Census, and supervisors and enumerators were hired. Hunt (1909) reported that quality improved with the 1890 and 1900 Censuses. The Bureau redefined the occupational category system each decade. The category system became elaborated in detail about manufacturing in the early 1900s. It was a large scale activity. Hunt reports there were more than 25 million copies of instructions and forms ("schedules") for the 1900 population census of pop, and several hundred supervisors, and 53,000 enumerators. The task was rushed.

"The International Institute of Statistics [recommended at its 1893 session] an international classification of occupations ... and in 1907 named a commission to prepare a technological glossary, in English, French, and German, of the designations of industries and occupations which have been employed in the censuses of leading countries, accompanied by a brief but exact description of the character of the work covered by each term so used ... [The commission's] first report is to be made ... [in] Paris, in May, 1909."

Until this time, the Census infrastructure was temporary, existing just for a couple of years before and after each decade’s canvass. In 1902, Congress established a permanent Bureau to conduct this and other surveys and analyses. This was expected to reduce loss of knowledge and skills between Censuses and sustain a staff of professionals. Other countries had done this around the same time.

A key analytical advance in 1910 was to ask two new questions along with the occupation question which give the basic job facts. One was what “industry” the employer was in. The other asked whether the respondent was an employee, or self employed, or an employer (Anderson Conk, 1978). The existence of these related questions allow the occupation question to break off from the sources of revenue and ask about the worker’s own function or tasks.

Alva Edwards was in charge of the occupational statistics from 1910 to 1940, and gradually reduced the use of industry information to classify the occupation. He had a strong preference for an occupation system to give information about the skills and intelligence of the worker. In the absence of measurable data about this he was sometimes reduced to making inferences from demographic information about the workers. He acknowledged that “certain specific occupations which technically are skilled occupations were classified as semiskilled because the enumerators [found] so many children, young persons, and women as pursuing these occupations as to render the occupations semiskilled, even though each of them did contain some skilled workers” (Edwards, 1917). Looked at from currently standard concepts, in which occupation is a structural, functional classification of workers, Edwards improved the analytical precision of the occupation categories by helping excise industry concepts from them, but clouded them by de facto including demographic or social-economic elements. For a more thorough discussion see Anderson Conk (1978, 1980).

The 1950 category system appears to have a good reputation among academics. Matt Sobek of the IPUMS.org project extended its reach to the entire span of 1850-2000, with jobs matched as well as possible from 1950 to the category system of the respondent’s year. Since 1950 the classification has continued to be revised every decade, and in major ways in 1980 and 2000.

Recent occupational classification practices

In recent decades the answers of respondents to the occupation and industry questions in the decennial Census have been encoded into a three-digit occupation code and a three-digit industry code, which are made available in the public-use micro samples. A monthly survey conduced by the same Bureau, called the Current Population Survey, has questions and procedures which are similar to the decennial Census. This data is used to derive unemployment statistics. The following description of practices for the Current Population Survey is similar to the practices for the decennial Census. The respondent (or a proxy respondent like a family member or neighbor) is asked approximately these questions[3] in this order:

  • (class of employer) Is this person’s employer a private for-profit company, a nonprofit, self-employed, federal government, state government, or local government?
  • What is the name of the person’s employer?
  • (industry) What kind of business or industry is it; what do they make or do?
  • Is it mainly in manufacturing, retail trade, wholesales rate, or something else?
  • (occupation) What kind of work does this person do? (For example: plumber, typist, farmer)
  • What are the person’s usual activities or duties at the job? (For example: typing, keeping account books, filing, selling cars, operating printing press, laying brick.)

The answers to these questions, four of which are open-ended text, along with the respondent’s city, state, sex, age, and years of education are made available to specialized “coders” in the Census Bureau’s NationalProcessingCenter in Jeffersonville, Indiana. I visited that office and interviewed some of the specialists.

The open-ended answers are in the respondent’s handwriting, digitally scanned from the decennial Census form, or were typed by the CPS interviewer onto a computer then downloaded to the coder’s computer. The coders follow carefully documented procedures to assign a three-digit industry code and an occupation code to the respondent. The open-ended answers are not made available in the public-use samples, and may not be available any more in any form.

The respondent may have given a job title, but in the more challenging cases has given a too-brief description of the tasks. A coder may not be able to assign occupation and industry codes based on the documented procedures, in which case the respondent’s data is forwarded (“referred”) electronically to a specialist called a “referralist.” This occurred in 17% of cases in a large 1997-8 sample (Couper and Conrad, 2001, p. 10). The referralist can match the employer name to a giant list of known employers called the Business Registry (which is the same as an earlier list called the SSEL), and may do a variety of kinds of research on the employer or the words used in the answers to the “kinds of work” and the “activities or duties”. The referralist may look up the employer on the web, and also can refer to a number of books on occupations. Referralists told me that there is some tendency for respondents to over-inflate the importance or prestige of their jobs.

If the job is hard to classify exactly, there are occupation categories for a best-match, usually with the text "not elsewhere classified" abbreviated in the title. For example, in the 1990 categories we see “Engineers, n.e.c.”, “Managers and administrators, n.e.c.,” “ Therapists, n.e.c.” and a dozen others like that. This technique helps keep some categories precise while making room to fit everyone, and by stretching these residual categories it is easier to extend a long time series of a particular category system.

Classifications of the Census data

The main classifications are in Table 1. In all cases I know of, the category systems were determined in advance and the respondent’s self description of the work activity was used by a Census Bureau employee to assign them into one of the categories. When there are multiple numbers of classifications listed in table 1, it's because the sources don't agree. One reason for that is that sometimes industry information was compounded with occupational information. For example Anderson Conk (1980) uses the example of “foreman and overseers (street railway)” which was distinguished from “foreman and overseers (not specified).”

Histories of the occupation statistics treat top officials of the census as having had strong inclinations which affect the way the jobs are reported (notably Anderson Conk, 1980, but also the earlier work). Francis Walker was in charge in 1870 and 1880 and he tended to see the “sector” (industry) as important information to encode in the occupation classification, although since 1920 there is a separate “industry” variable which characterizes the employer’s field of activity. Carroll Wright was in charge for 1890 and 1900. Alba Edwards was an increasingly important official from 1910 to 1940. He tended to elevate the importance of differences between jobs where workers used their heads versus and those where workers used their hands.[4]

In Census-associated secondary surveys of business establishments, employers could offer job titles they used, and not use the decennial census classification. For example, in a 1880 census special survey manufacturing establishments (Weeks, 1884), employers were invited to submit lists of the jobs at their establishment, and these were not standardized. Meyer (2004) standardized these into a set of 300 jobs across all establishments, but it does not match the HISCO classification or the classification in the 1880 census of households.

Table 1. US Census occupational classifications

Sources: For 1850-60 categories, Hunt (1909). For 1870-1940 categories, Anderson Conk (1980), p. 23. For more recent categories, For the phrasing of the question, Wright with Hunt (1900).

Census year / Number of categories / The question asked, or other notes
1790-1840 Censuses / No specific occupation question
1850 / 323 / "Profession, occupation, or trade of each male person over 15 years of age" and "Number of slaves" (without further detail on their activities)
1860 / 584 / "Profession, occupation, or trade of each person, male and female, over 15 years of age" and "Number of slaves" (without further detail on their activities)
1870 / 338 / "Profession, occupation, or trade of each person, male or female"
1880 / 276 or 265 / "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. Full list of occupations is at
1890 / 218 / ""Profession, trade, or occupation" and, of those, months unemployed during the census year
1900 / 303
1910 / 428 or 432
1920 / 572 or 574 / Full list of occupations is at
1930 / 534 / Full list of occupations is at
1940 / 228, 221, or 451 / Full list of occupations is at
1950 / 287 / Full list of occupations is at
1960 / 296 / Full list of occupations is at
1970 / 441 / Full list of occupations is at
1980 / 504 / "(a) What kind of work was this person doing? (at least two words) (b) What were this person's most important activities or duties?" Full list of occupations is at
1990 / 504 / Full list of occupations is at
2000 US Census (1% sample) / 510 / Full list of occupations is at

People at the margins of occupation definition

Some occupational categories have a clear continuity over time, such as those of doctors, barbers, carpenters, laborers, or public officials. But most of the population was not categorized in an occupation, or was considered in an occupation in some decades but not others. A key principle was that a person's activity was almost always conceived of as an occupation if and only if the person received pay for their time. To clarify this through experience it is helpful to examine some of the sets of people who were sometimes thought of as marginally employed or outside the system of occupations.

Indians. Indians, now called native Americans, were not defined by the Constitution as citizens, and did not have representation in the U.S. and were at first excluded from the Census. The sovereignty of North American Indian nations has been redefined over time. By 1890, there was an ethnic category in the census for “civilized Indians,” apparently meaning more or less the ones who had settled into fixed dwellings and who were willing to be included in the American population. J. David Hacker reports that perhaps 90% of Indians were excluded from the U.S. Census until 1920. In 1890, 1900, and 1910 there were special censuses of Indians, including occupation, but the results were separate from the main Census of Population.[5] This means on the order of 250,000 persons were not in the main occupation counts until 1920.

Slaves. The Constitution required a count of free citizens and a count of slaves. The subject of slavery was very divisive, and after difficult bargaining the framers agreed to count slaves as three-fifths of a person for purposes of a district's representation and taxation, though slaves could not vote and did not themselves pay taxes. The census did not inquire about the occupations of slaves, nor was "slave" defined as an occupation. Some slaves were specialized and skilled however. One can see this as a tourist to George Washington's plantation, where among the slaves there were metalworking and leather specialists, clothing makers, food processors, housekeepers, and farm workers. The standard census form ("schedule") did not ask for this information in the 1850 and 1860 censuses. After that, slavery was abolished legally and as a data category.

Wives and mothers at home. If not employed, these women do not have occupations. This principle is of long standing. But because they were expected to be dependents, women in past decades were less likely to be recorded as having an occupation even if they were employed than they are now. My information on this research comes from Anderson (2002). Bose (2001) evaluated how then-conventional roles for women affected the way the 1900 census data reported female-headed households and formally-unemployed housewives. “Bose estimates that in 1900, 46.4% of women aged 15-64 'worked' in the formal and informal home based economy, compared with a 1994 rate of 58.8%. According to the 1900 census definition, 22.5% ofwomen 'worked'.” (Anderson, pp. 507) Thus using 2000 definitions, twice as many women were working in 1900 than the 1900 census reported. Thus many women in 1900 who were not recorded in an occupation category would, now, be recorded as having an occupation if they submitted the same data to a current census.

Children and students. There is usually a lower bound on persons for whom an occupation can be recorded. Based on the instructions to the enumerators, it was age 15 in the 1850 and 1860 Censuses, there was no age in the 1870 Census, then it was age 10 in 1880. There was no lower age in 1890. In 1940 the lower bound was changed from 10 to 14. In recent years I believe the lower bound age is 16. Being a student has not been defined as an occupation in any year. A youth or student with a part-job is now to be recorded as having an occupation if the job takes (I believe) twenty or more hours per week.

Retired or unemployed. Since the 1970 census, there is classification for the unemployed who have not worked in five years but would be willing to work. A respondent who does not have a job but has had one less than five years ago and seeks one now may describe themselves as having an occupation which is the customary job or a previous job. A retired person, that is one who does not intend to return to a previous line of work, is now conceived of as outside the labor force, and does not have a census-defined occupation.

Non-citizens and border-crossers. The census is a survey of people who live in the United States. Likewise, persons crossing borders between home and work, legally or otherwise, are counted in the census and their occupations recorded if they live in the U.S. and not if they live in another country. (These numbers can be large. It has been estimated recently is that 11% of Mexico's population lives in the U.S.) In all cases it does not matter whether the employer or the workplace is in the United States. I do not know how far these principles go back. For redistricting purposes, I believe only citizens count.