The Current Population Survey—tracking unemployment in the United States for over 75 years

The Current Population Survey (CPS) has been conducted for more than three-quarters of a century.¹ From the outset, the main purpose of the survey has been to gather information on the employment status of the U.S. population, with an emphasis on the measurement of unemployment. CPS data have been used by policymakers and others to gauge both the degree of labor market weakness during recessions and the strength of the job market in economic expansions. More than 900 monthly reports on national employment and unemployment have been issued since the survey began in March 1940.

The survey also has been used to provide a wealth of information on a wide range of other subjects—some related to the labor market and some unrelated—through supplemental questions to the basic survey instrument. Over the years, supplements to the CPS have been used to collect data on topics ranging from income and worker displacement to tobacco use and participation in the arts.

The main objective of the CPS, however, has always been to measure unemployment and other aspects of labor market performance. This article summarizes some of the major developments in achieving this goal over the past three-quarters of a century.

1930s—what is unemployment and how pervasive is it?

Mass unemployment became a major problem in the United States during the Great Depression, yet there was remarkably little statistical information on the extent or nature of joblessness across the nation. Efforts to measure unemployment were stymied both by disagreement over how to define the problem—that is, what it meant to be “unemployed”—and by the lack of a tool that could gather reliable data on a recurring basis. It was not until the end of the 1930s that a monthly household survey yielding timely and reliable information on unemployment was put in place. The survey, called the Monthly Report of Unemployment (now called the CPS), embodied two significant breakthroughs that made measuring unemployment on an ongoing basis possible: the activity concept and probability sampling.²

Activity concept

Many early surveys relied on a person’s stated interest in work to determine whether or not that person was unemployed. These surveys often started by identifying the number of “gainful workers”—people who had a usual occupation.³ The surveys then proceeded to sort the subset of individuals in this group who were not currently working into the unemployed category on the basis of whether a person said they were “able” and “willing” to work. But many researchers questioned the meaningfulness of these criteria.

Asking people without work if they were able and/or wanted to work was fraught with value judgments, as the questions were largely hypothetical. They failed to stipulate a range of factors people would consider before deciding whether to enter the job market, like the nature of the job, hours, location, and pay, among others. Also, whether a person not currently working wanted or was able to take a job—even if liberally interpreted as indicative of current labor market attachment—arguably could vacillate from day to day with changing perceptions about the state of the job market or changes in one’s current living situation. There was also concern about how cultural norms and current economic conditions might bias answers to such questions toward socially acceptable responses, especially during a period of widespread unemployment. In the end, measuring unemployment by inquiring about ability and/or wanting/willingness to work was deemed much too subjective.

Instead, researchers settled on a more objective way to ascertain one’s employment status—by simply finding out what job-market-related activities people were doing during a prescribed period. Under this activity concept, a person who was not working but was actively seeking a job would be classified as unemployed. Classification based on activity is inherently more objective because actions are overt and demonstrable, rather than merely indicating that one wants a job.⁴

Probability sampling

A more widely accepted measurement of unemployment required not only a more objective concept, but also more timely and accurate data. During the 1930s, statisticians were experimenting with probability sampling, using sample surveys that could accurately reflect population totals. Not only were the surveys representative of the population, but the data could be gathered and made available relatively quickly, unlike a complete tally such as a census.⁵

Toward the end of the decade, the Work Projects Administration (WPA) tested a monthly probability-based sample survey to produce national estimates of unemployment using the activity concept. Although there was skepticism that a small sample survey could produce reliable estimates representative of the entire population, advances in sample design and the use of trained interviewers to collect data helped ensure the sample-survey approach would work. The first operational survey, which collected data for March 1940, was intended to coincide with the 1940 Census, which contained similar labor-related questions.⁶ When the data from the two surveys were compared, it was determined that the sample survey produced more accurate statistics on unemployment, in part because survey interviewers were trained specifically on labor force concepts.⁷

Figure 1. Unemployment rate and timing of changes to Current Population Survey measurement, 1940–2017

Date	Unemployment rate
Mar 1940	15.4
Apr 1940	15.0
May 1940	14.4
Jun 1940	14.5
Jul 1940	15.8
Aug 1940	15.2
Sep 1940	12.2
Oct 1940	13.0
Nov 1940	13.2
Dec 1940	12.7
Jan 1941	13.7
Feb 1941	12.9
Mar 1941	12.1
Apr 1941	11.6
May 1941	10.1
Jun 1941	10.7
Jul 1941	10.3
Aug 1941	9.7
Sep 1941	8.2
Oct 1941	6.9
Nov 1941	6.8
Dec 1941	6.5
Jan 1942	7.9
Feb 1942	7.4
Mar 1942	6.5
Apr 1942	5.5
May 1942	4.6
Jun 1942	5.0
Jul 1942	4.8
Aug 1942	3.8
Sep 1942	3.0
Oct 1942	2.8
Nov 1942	2.9
Dec 1942	2.7
Jan 1943	2.7
Feb 1943	2.6
Mar 1943	2.1
Apr 1943	1.8
May 1943	1.7
Jun 1943	2.3
Jul 1943	2.4
Aug 1943	1.8
Sep 1943	1.6
Oct 1943	1.4
Nov 1943	1.3
Dec 1943	1.3
Jan 1944	1.5
Feb 1944	1.3
Mar 1944	1.3
Apr 1944	1.2
May 1944	1.3
Jun 1944	1.6
Jul 1944	1.6
Aug 1944	1.2
Sep 1944	1.1
Oct 1944	0.8
Nov 1944	0.9
Dec 1944	0.9
Jan 1945	1.2
Feb 1945	1.2
Mar 1945	1.1
Apr 1945	1.0
May 1945	1.0
Jun 1945	1.6
Jul 1945	1.7
Aug 1945	1.5
Sep 1945	3.1
Oct 1945	2.9
Nov 1945	3.3
Dec 1945	3.7
Jan 1946	4.3
Feb 1946	4.9
Mar 1946	4.9
Apr 1946	4.1
May 1946	4.0
Jun 1946	4.4
Jul 1946	3.8
Aug 1946	3.4
Sep 1946	3.5
Oct 1946	3.3
Nov 1946	3.3
Dec 1946	3.6
Jan 1947	4.2
Feb 1947	4.3
Mar 1947	4.0
Apr 1947	4.1
May 1947	3.3
Jun 1947	4.1
Jul 1947	4.1
Aug 1947	3.4
Sep 1947	3.1
Oct 1947	2.8
Nov 1947	2.7
Dec 1947	2.8
Jan 1948	3.4
Feb 1948	3.8
Mar 1948	4.0
Apr 1948	3.9
May 1948	3.5
Jun 1948	3.6
Jul 1948	3.6
Aug 1948	3.9
Sep 1948	3.8
Oct 1948	3.7
Nov 1948	3.8
Dec 1948	4.0
Jan 1949	4.3
Feb 1949	4.7
Mar 1949	5.0
Apr 1949	5.3
May 1949	6.1
Jun 1949	6.2
Jul 1949	6.7
Aug 1949	6.8
Sep 1949	6.6
Oct 1949	7.9
Nov 1949	6.4
Dec 1949	6.6
Jan 1950	6.5
Feb 1950	6.4
Mar 1950	6.3
Apr 1950	5.8
May 1950	5.5
Jun 1950	5.4
Jul 1950	5.0
Aug 1950	4.5
Sep 1950	4.4
Oct 1950	4.2
Nov 1950	4.2
Dec 1950	4.3
Jan 1951	3.7
Feb 1951	3.4
Mar 1951	3.4
Apr 1951	3.1
May 1951	3.0
Jun 1951	3.2
Jul 1951	3.1
Aug 1951	3.1
Sep 1951	3.3
Oct 1951	3.5
Nov 1951	3.5
Dec 1951	3.1
Jan 1952	3.2
Feb 1952	3.1
Mar 1952	2.9
Apr 1952	2.9
May 1952	3.0
Jun 1952	3.0
Jul 1952	3.2
Aug 1952	3.4
Sep 1952	3.1
Oct 1952	3.0
Nov 1952	2.8
Dec 1952	2.7
Jan 1953	2.9
Feb 1953	2.6
Mar 1953	2.6
Apr 1953	2.7
May 1953	2.5
Jun 1953	2.5
Jul 1953	2.6
Aug 1953	2.7
Sep 1953	2.9
Oct 1953	3.1
Nov 1953	3.5
Dec 1953	4.5
Jan 1954	4.9
Feb 1954	5.2
Mar 1954	5.7
Apr 1954	5.9
May 1954	5.9
Jun 1954	5.6
Jul 1954	5.8
Aug 1954	6.0
Sep 1954	6.1
Oct 1954	5.7
Nov 1954	5.3
Dec 1954	5.0
Jan 1955	4.9
Feb 1955	4.7
Mar 1955	4.6
Apr 1955	4.7
May 1955	4.3
Jun 1955	4.2
Jul 1955	4.0
Aug 1955	4.2
Sep 1955	4.1
Oct 1955	4.3
Nov 1955	4.2
Dec 1955	4.2
Jan 1956	4.0
Feb 1956	3.9
Mar 1956	4.2
Apr 1956	4.0
May 1956	4.3
Jun 1956	4.3
Jul 1956	4.4
Aug 1956	4.1
Sep 1956	3.9
Oct 1956	3.9
Nov 1956	4.3
Dec 1956	4.2
Jan 1957	4.2
Feb 1957	3.9
Mar 1957	3.7
Apr 1957	3.9
May 1957	4.1
Jun 1957	4.3
Jul 1957	4.2
Aug 1957	4.1
Sep 1957	4.4
Oct 1957	4.5
Nov 1957	5.1
Dec 1957	5.2
Jan 1958	5.8
Feb 1958	6.4
Mar 1958	6.7
Apr 1958	7.4
May 1958	7.4
Jun 1958	7.3
Jul 1958	7.5
Aug 1958	7.4
Sep 1958	7.1
Oct 1958	6.7
Nov 1958	6.2
Dec 1958	6.2
Jan 1959	6.0
Feb 1959	5.9
Mar 1959	5.6
Apr 1959	5.2
May 1959	5.1
Jun 1959	5.0
Jul 1959	5.1
Aug 1959	5.2
Sep 1959	5.5
Oct 1959	5.7
Nov 1959	5.8
Dec 1959	5.3
Jan 1960	5.2
Feb 1960	4.8
Mar 1960	5.4
Apr 1960	5.2
May 1960	5.1
Jun 1960	5.4
Jul 1960	5.5
Aug 1960	5.6
Sep 1960	5.5
Oct 1960	6.1
Nov 1960	6.1
Dec 1960	6.6
Jan 1961	6.6
Feb 1961	6.9
Mar 1961	6.9
Apr 1961	7.0
May 1961	7.1
Jun 1961	6.9
Jul 1961	7.0
Aug 1961	6.6
Sep 1961	6.7
Oct 1961	6.5
Nov 1961	6.1
Dec 1961	6.0
Jan 1962	5.8
Feb 1962	5.5
Mar 1962	5.6
Apr 1962	5.6
May 1962	5.5
Jun 1962	5.5
Jul 1962	5.4
Aug 1962	5.7
Sep 1962	5.6
Oct 1962	5.4
Nov 1962	5.7
Dec 1962	5.5
Jan 1963	5.7
Feb 1963	5.9
Mar 1963	5.7
Apr 1963	5.7
May 1963	5.9
Jun 1963	5.6
Jul 1963	5.6
Aug 1963	5.4
Sep 1963	5.5
Oct 1963	5.5
Nov 1963	5.7
Dec 1963	5.5
Jan 1964	5.6
Feb 1964	5.4
Mar 1964	5.4
Apr 1964	5.3
May 1964	5.1
Jun 1964	5.2
Jul 1964	4.9
Aug 1964	5.0
Sep 1964	5.1
Oct 1964	5.1
Nov 1964	4.8
Dec 1964	5.0
Jan 1965	4.9
Feb 1965	5.1
Mar 1965	4.7
Apr 1965	4.8
May 1965	4.6
Jun 1965	4.6
Jul 1965	4.4
Aug 1965	4.4
Sep 1965	4.3
Oct 1965	4.2
Nov 1965	4.1
Dec 1965	4.0
Jan 1966	4.0
Feb 1966	3.8
Mar 1966	3.8
Apr 1966	3.8
May 1966	3.9
Jun 1966	3.8
Jul 1966	3.8
Aug 1966	3.8
Sep 1966	3.7
Oct 1966	3.7
Nov 1966	3.6
Dec 1966	3.8
Jan 1967	3.9
Feb 1967	3.8
Mar 1967	3.8
Apr 1967	3.8
May 1967	3.8
Jun 1967	3.9
Jul 1967	3.8
Aug 1967	3.8
Sep 1967	3.8
Oct 1967	4.0
Nov 1967	3.9
Dec 1967	3.8
Jan 1968	3.7
Feb 1968	3.8
Mar 1968	3.7
Apr 1968	3.5
May 1968	3.5
Jun 1968	3.7
Jul 1968	3.7
Aug 1968	3.5
Sep 1968	3.4
Oct 1968	3.4
Nov 1968	3.4
Dec 1968	3.4
Jan 1969	3.4
Feb 1969	3.4
Mar 1969	3.4
Apr 1969	3.4
May 1969	3.4
Jun 1969	3.5
Jul 1969	3.5
Aug 1969	3.5
Sep 1969	3.7
Oct 1969	3.7
Nov 1969	3.5
Dec 1969	3.5
Jan 1970	3.9
Feb 1970	4.2
Mar 1970	4.4
Apr 1970	4.6
May 1970	4.8
Jun 1970	4.9
Jul 1970	5.0
Aug 1970	5.1
Sep 1970	5.4
Oct 1970	5.5
Nov 1970	5.9
Dec 1970	6.1
Jan 1971	5.9
Feb 1971	5.9
Mar 1971	6.0
Apr 1971	5.9
May 1971	5.9
Jun 1971	5.9
Jul 1971	6.0
Aug 1971	6.1
Sep 1971	6.0
Oct 1971	5.8
Nov 1971	6.0
Dec 1971	6.0
Jan 1972	5.8
Feb 1972	5.7
Mar 1972	5.8
Apr 1972	5.7
May 1972	5.7
Jun 1972	5.7
Jul 1972	5.6
Aug 1972	5.6
Sep 1972	5.5
Oct 1972	5.6
Nov 1972	5.3
Dec 1972	5.2
Jan 1973	4.9
Feb 1973	5.0
Mar 1973	4.9
Apr 1973	5.0
May 1973	4.9
Jun 1973	4.9
Jul 1973	4.8
Aug 1973	4.8
Sep 1973	4.8
Oct 1973	4.6
Nov 1973	4.8
Dec 1973	4.9
Jan 1974	5.1
Feb 1974	5.2
Mar 1974	5.1
Apr 1974	5.1
May 1974	5.1
Jun 1974	5.4
Jul 1974	5.5
Aug 1974	5.5
Sep 1974	5.9
Oct 1974	6.0
Nov 1974	6.6
Dec 1974	7.2
Jan 1975	8.1
Feb 1975	8.1
Mar 1975	8.6
Apr 1975	8.8
May 1975	9.0
Jun 1975	8.8
Jul 1975	8.6
Aug 1975	8.4
Sep 1975	8.4
Oct 1975	8.4
Nov 1975	8.3
Dec 1975	8.2
Jan 1976	7.9
Feb 1976	7.7
Mar 1976	7.6
Apr 1976	7.7
May 1976	7.4
Jun 1976	7.6
Jul 1976	7.8
Aug 1976	7.8
Sep 1976	7.6
Oct 1976	7.7
Nov 1976	7.8
Dec 1976	7.8
Jan 1977	7.5
Feb 1977	7.6
Mar 1977	7.4
Apr 1977	7.2
May 1977	7.0
Jun 1977	7.2
Jul 1977	6.9
Aug 1977	7.0
Sep 1977	6.8
Oct 1977	6.8
Nov 1977	6.8
Dec 1977	6.4
Jan 1978	6.4
Feb 1978	6.3
Mar 1978	6.3
Apr 1978	6.1
May 1978	6.0
Jun 1978	5.9
Jul 1978	6.2
Aug 1978	5.9
Sep 1978	6.0
Oct 1978	5.8
Nov 1978	5.9
Dec 1978	6.0
Jan 1979	5.9
Feb 1979	5.9
Mar 1979	5.8
Apr 1979	5.8
May 1979	5.6
Jun 1979	5.7
Jul 1979	5.7
Aug 1979	6.0
Sep 1979	5.9
Oct 1979	6.0
Nov 1979	5.9
Dec 1979	6.0
Jan 1980	6.3
Feb 1980	6.3
Mar 1980	6.3
Apr 1980	6.9
May 1980	7.5
Jun 1980	7.6
Jul 1980	7.8
Aug 1980	7.7
Sep 1980	7.5
Oct 1980	7.5
Nov 1980	7.5
Dec 1980	7.2
Jan 1981	7.5
Feb 1981	7.4
Mar 1981	7.4
Apr 1981	7.2
May 1981	7.5
Jun 1981	7.5
Jul 1981	7.2
Aug 1981	7.4
Sep 1981	7.6
Oct 1981	7.9
Nov 1981	8.3
Dec 1981	8.5
Jan 1982	8.6
Feb 1982	8.9
Mar 1982	9.0
Apr 1982	9.3
May 1982	9.4
Jun 1982	9.6
Jul 1982	9.8
Aug 1982	9.8
Sep 1982	10.1
Oct 1982	10.4
Nov 1982	10.8
Dec 1982	10.8
Jan 1983	10.4
Feb 1983	10.4
Mar 1983	10.3
Apr 1983	10.2
May 1983	10.1
Jun 1983	10.1
Jul 1983	9.4
Aug 1983	9.5
Sep 1983	9.2
Oct 1983	8.8
Nov 1983	8.5
Dec 1983	8.3
Jan 1984	8.0
Feb 1984	7.8
Mar 1984	7.8
Apr 1984	7.7
May 1984	7.4
Jun 1984	7.2
Jul 1984	7.5
Aug 1984	7.5
Sep 1984	7.3
Oct 1984	7.4
Nov 1984	7.2
Dec 1984	7.3
Jan 1985	7.3
Feb 1985	7.2
Mar 1985	7.2
Apr 1985	7.3
May 1985	7.2
Jun 1985	7.4
Jul 1985	7.4
Aug 1985	7.1
Sep 1985	7.1
Oct 1985	7.1
Nov 1985	7.0
Dec 1985	7.0
Jan 1986	6.7
Feb 1986	7.2
Mar 1986	7.2
Apr 1986	7.1
May 1986	7.2
Jun 1986	7.2
Jul 1986	7.0
Aug 1986	6.9
Sep 1986	7.0
Oct 1986	7.0
Nov 1986	6.9
Dec 1986	6.6
Jan 1987	6.6
Feb 1987	6.6
Mar 1987	6.6
Apr 1987	6.3
May 1987	6.3
Jun 1987	6.2
Jul 1987	6.1
Aug 1987	6.0
Sep 1987	5.9
Oct 1987	6.0
Nov 1987	5.8
Dec 1987	5.7
Jan 1988	5.7
Feb 1988	5.7
Mar 1988	5.7
Apr 1988	5.4
May 1988	5.6
Jun 1988	5.4
Jul 1988	5.4
Aug 1988	5.6
Sep 1988	5.4
Oct 1988	5.4
Nov 1988	5.3
Dec 1988	5.3
Jan 1989	5.4
Feb 1989	5.2
Mar 1989	5.0
Apr 1989	5.2
May 1989	5.2
Jun 1989	5.3
Jul 1989	5.2
Aug 1989	5.2
Sep 1989	5.3
Oct 1989	5.3
Nov 1989	5.4
Dec 1989	5.4
Jan 1990	5.4
Feb 1990	5.3
Mar 1990	5.2
Apr 1990	5.4
May 1990	5.4
Jun 1990	5.2
Jul 1990	5.5
Aug 1990	5.7
Sep 1990	5.9
Oct 1990	5.9
Nov 1990	6.2
Dec 1990	6.3
Jan 1991	6.4
Feb 1991	6.6
Mar 1991	6.8
Apr 1991	6.7
May 1991	6.9
Jun 1991	6.9
Jul 1991	6.8
Aug 1991	6.9
Sep 1991	6.9
Oct 1991	7.0
Nov 1991	7.0
Dec 1991	7.3
Jan 1992	7.3
Feb 1992	7.4
Mar 1992	7.4
Apr 1992	7.4
May 1992	7.6
Jun 1992	7.8
Jul 1992	7.7
Aug 1992	7.6
Sep 1992	7.6
Oct 1992	7.3
Nov 1992	7.4
Dec 1992	7.4
Jan 1993	7.3
Feb 1993	7.1
Mar 1993	7.0
Apr 1993	7.1
May 1993	7.1
Jun 1993	7.0
Jul 1993	6.9
Aug 1993	6.8
Sep 1993	6.7
Oct 1993	6.8
Nov 1993	6.6
Dec 1993	6.5
Jan 1994	6.6
Feb 1994	6.6
Mar 1994	6.5
Apr 1994	6.4
May 1994	6.1
Jun 1994	6.1
Jul 1994	6.1
Aug 1994	6.0
Sep 1994	5.9
Oct 1994	5.8
Nov 1994	5.6
Dec 1994	5.5
Jan 1995	5.6
Feb 1995	5.4
Mar 1995	5.4
Apr 1995	5.8
May 1995	5.6
Jun 1995	5.6
Jul 1995	5.7
Aug 1995	5.7
Sep 1995	5.6
Oct 1995	5.5
Nov 1995	5.6
Dec 1995	5.6
Jan 1996	5.6
Feb 1996	5.5
Mar 1996	5.5
Apr 1996	5.6
May 1996	5.6
Jun 1996	5.3
Jul 1996	5.5
Aug 1996	5.1
Sep 1996	5.2
Oct 1996	5.2
Nov 1996	5.4
Dec 1996	5.4
Jan 1997	5.3
Feb 1997	5.2
Mar 1997	5.2
Apr 1997	5.1
May 1997	4.9
Jun 1997	5.0
Jul 1997	4.9
Aug 1997	4.8
Sep 1997	4.9
Oct 1997	4.7
Nov 1997	4.6
Dec 1997	4.7
Jan 1998	4.6
Feb 1998	4.6
Mar 1998	4.7
Apr 1998	4.3
May 1998	4.4
Jun 1998	4.5
Jul 1998	4.5
Aug 1998	4.5
Sep 1998	4.6
Oct 1998	4.5
Nov 1998	4.4
Dec 1998	4.4
Jan 1999	4.3
Feb 1999	4.4
Mar 1999	4.2
Apr 1999	4.3
May 1999	4.2
Jun 1999	4.3
Jul 1999	4.3
Aug 1999	4.2
Sep 1999	4.2
Oct 1999	4.1
Nov 1999	4.1
Dec 1999	4.0
Jan 2000	4.0
Feb 2000	4.1
Mar 2000	4.0
Apr 2000	3.8
May 2000	4.0
Jun 2000	4.0
Jul 2000	4.0
Aug 2000	4.1
Sep 2000	3.9
Oct 2000	3.9
Nov 2000	3.9
Dec 2000	3.9
Jan 2001	4.2
Feb 2001	4.2
Mar 2001	4.3
Apr 2001	4.4
May 2001	4.3
Jun 2001	4.5
Jul 2001	4.6
Aug 2001	4.9
Sep 2001	5.0
Oct 2001	5.3
Nov 2001	5.5
Dec 2001	5.7
Jan 2002	5.7
Feb 2002	5.7
Mar 2002	5.7
Apr 2002	5.9
May 2002	5.8
Jun 2002	5.8
Jul 2002	5.8
Aug 2002	5.7
Sep 2002	5.7
Oct 2002	5.7
Nov 2002	5.9
Dec 2002	6.0
Jan 2003	5.8
Feb 2003	5.9
Mar 2003	5.9
Apr 2003	6.0
May 2003	6.1
Jun 2003	6.3
Jul 2003	6.2
Aug 2003	6.1
Sep 2003	6.1
Oct 2003	6.0
Nov 2003	5.8
Dec 2003	5.7
Jan 2004	5.7
Feb 2004	5.6
Mar 2004	5.8
Apr 2004	5.6
May 2004	5.6
Jun 2004	5.6
Jul 2004	5.5
Aug 2004	5.4
Sep 2004	5.4
Oct 2004	5.5
Nov 2004	5.4
Dec 2004	5.4
Jan 2005	5.3
Feb 2005	5.4
Mar 2005	5.2
Apr 2005	5.2
May 2005	5.1
Jun 2005	5.0
Jul 2005	5.0
Aug 2005	4.9
Sep 2005	5.0
Oct 2005	5.0
Nov 2005	5.0
Dec 2005	4.9
Jan 2006	4.7
Feb 2006	4.8
Mar 2006	4.7
Apr 2006	4.7
May 2006	4.6
Jun 2006	4.6
Jul 2006	4.7
Aug 2006	4.7
Sep 2006	4.5
Oct 2006	4.4
Nov 2006	4.5
Dec 2006	4.4
Jan 2007	4.6
Feb 2007	4.5
Mar 2007	4.4
Apr 2007	4.5
May 2007	4.4
Jun 2007	4.6
Jul 2007	4.7
Aug 2007	4.6
Sep 2007	4.7
Oct 2007	4.7
Nov 2007	4.7
Dec 2007	5.0
Jan 2008	5.0
Feb 2008	4.9
Mar 2008	5.1
Apr 2008	5.0
May 2008	5.4
Jun 2008	5.6
Jul 2008	5.8
Aug 2008	6.1
Sep 2008	6.1
Oct 2008	6.5
Nov 2008	6.8
Dec 2008	7.3
Jan 2009	7.8
Feb 2009	8.3
Mar 2009	8.7
Apr 2009	9.0
May 2009	9.4
Jun 2009	9.5
Jul 2009	9.5
Aug 2009	9.6
Sep 2009	9.8
Oct 2009	10.0
Nov 2009	9.9
Dec 2009	9.9
Jan 2010	9.8
Feb 2010	9.8
Mar 2010	9.9
Apr 2010	9.9
May 2010	9.6
Jun 2010	9.4
Jul 2010	9.4
Aug 2010	9.5
Sep 2010	9.5
Oct 2010	9.4
Nov 2010	9.8
Dec 2010	9.3
Jan 2011	9.1
Feb 2011	9.0
Mar 2011	9.0
Apr 2011	9.1
May 2011	9.0
Jun 2011	9.1
Jul 2011	9.0
Aug 2011	9.0
Sep 2011	9.0
Oct 2011	8.8
Nov 2011	8.6
Dec 2011	8.5
Jan 2012	8.3
Feb 2012	8.3
Mar 2012	8.2
Apr 2012	8.2
May 2012	8.2
Jun 2012	8.2
Jul 2012	8.2
Aug 2012	8.1
Sep 2012	7.8
Oct 2012	7.8
Nov 2012	7.7
Dec 2012	7.9
Jan 2013	8.0
Feb 2013	7.7
Mar 2013	7.5
Apr 2013	7.6
May 2013	7.5
Jun 2013	7.5
Jul 2013	7.3
Aug 2013	7.2
Sep 2013	7.2
Oct 2013	7.2
Nov 2013	6.9
Dec 2013	6.7
Jan 2014	6.6
Feb 2014	6.7
Mar 2014	6.7
Apr 2014	6.3
May 2014	6.3
Jun 2014	6.1
Jul 2014	6.2
Aug 2014	6.2
Sep 2014	5.9
Oct 2014	5.7
Nov 2014	5.8
Dec 2014	5.6
Jan 2015	5.7
Feb 2015	5.5
Mar 2015	5.5
Apr 2015	5.4
May 2015	5.5
Jun 2015	5.3
Jul 2015	5.2
Aug 2015	5.1
Sep 2015	5.0
Oct 2015	5.0
Nov 2015	5.0
Dec 2015	5.0
Jan 2016	4.9
Feb 2016	4.9
Mar 2016	5.0
Apr 2016	5.0
May 2016	4.7
Jun 2016	4.9
Jul 2016	4.9
Aug 2016	4.9
Sep 2016	5.0
Oct 2016	4.9
Nov 2016	4.6
Dec 2016	4.7
Jan 2017	4.8
Feb 2017	4.7
Mar 2017	4.5
Apr 2017	4.4
May 2017	4.3
Jun 2017	4.3
Jul 2017	4.3
Aug 2017	4.4
Sep 2017	4.2
Oct 2017	4.1
Nov 2017	4.1
Dec 2017	4.1
Source: 1940–47 figures are based on authors' calculations using U.S. Census Bureau data; 1948–2017 data are from the U.S. Bureau of Labor Statistics.

View Chart Data

The first unemployment figures were released in an internal memorandum, just a few weeks after data collection.⁸ (See figure 1 for the full history of CPS-based unemployment rates.) At this point, to be classified as unemployed, a person had to be without work and seeking a job. Also included among the unemployed in the earliest years of the survey were people who were enrolled in work relief programs. The unemployment rate—the number of unemployed people divided by the number of people in the labor force (the sum of employed and unemployed people)—was still high at about 15 percent in March 1940. This was well below the peak reached during the Great Depression, later estimated to have been around 25 percent.⁹ At the time the first unemployment figures were released, widespread unemployment was still a concern. However, war in Europe had broken out, and the formal entry of the United States into World War II was less than 2 years away. Ironically, the survey that had initially been built to measure unemployment would quickly be pressed into service to help manage the opposite problem for which it was designed—labor scarcity.

1940s—World War II and shifting focus from labor surplus to labor shortages

By 1942, with U.S. wartime mobilization well underway, the Monthly Report of Unemployment documented various industrial production issues associated with the war effort. These included economic dislocations, as people, mostly men, left the labor force to report for military service. As the production of war-related materiel surged and the need for depression-era work relief programs quickly faded, the WPA disbanded, and the Census Bureau took charge of the household survey in August 1942. The name of the survey was then changed to Monthly Report on the Labor Force, a title that more broadly reflected the evolving data needs of policymakers.¹⁰

To underscore new wartime concerns about labor scarcity, special questions were added to the CPS about the “potential labor supply,” later referred to as the “labor reserve.” As the jobless rate fell to around 1 percent, concerns shifted away from high unemployment to discussion about how unemployment could not get below some “irreducible minimum.” At this time, the household survey data revealed the growing participation of women and youth in the labor force as a critical part of industrial production. Indeed, the majority of the labor reserve was made up of women, mostly “housewives.” With the end of the war, however, large-scale cutbacks in the production of war-related supplies, plus the return of soldiers to civilian life, raised concerns about the reemergence of the problem that had given root to the survey—mass unemployment—which never materialized.¹¹

Apart from revamping the questionnaire in 1945 to better capture marginal or secondary employment activities, the Census Bureau made no major changes to the survey in its early years, and the unemployment definition remained the same.¹² However, in 1947–48, the name of the survey was changed to the Current Population Survey (its current name), a title that better reflected the expanding role of the survey in collecting more than just labor force data.¹³ The survey’s first major interagency review, conducted in 1948 by the Bureau of the Budget’s Interagency Committee on Labor Supply, Employment, and Unemployment Statistics, made no major recommendations for change.¹⁴

It is worth noting that the seeds of the first regular CPS supplement were sown in 1948. Two questions were added in April of that year to gather information on the earnings and income received in calendar year 1947. The income-related questions expanded dramatically in scope and detail over the years, and this yearly supplement would eventually become known as the Annual Social and Economic Supplement (ASEC).¹⁵ It could be argued that the success and popularity of these two (and other early) supplemental questions helped pave the way for future supplements on a much wider array of topics.¹⁶

1950s—back to business cycles, and survey improvements

With the Great Depression and World War II over, the U.S. economy entered a period of long-term economic growth, and the unemployment rate hovered around 4 percent during the early 1950s. This long-term trend, however, was interrupted by several business-cycle contractions, resulting in a demand for more and higher quality data on the job market. All of this kept the household survey concepts and methods, as well as other economic indicators, under review by interagency committees. This oversight, in turn, contributed to some minor definitional changes during the decade.

The most important changes stemming from these reviews were made in January 1957, following the recommendations from an interagency subcommittee that had met in 1954–55.¹⁷ The subcommittee recommended that two small groups of people—specifically, those “with a job but not at work” due to a temporary layoff for less than 30 days and those waiting to begin a new job within 30 days—be shifted from the employed to the unemployed classification. This change was made because their attachment to an employer was too tenuous to warrant inclusion among those who had a job. Additionally, for those on temporary layoff, the subcommittee suggested that classifying them among the unemployed would improve the macroeconomic sensitivity of the unemployment measure. The committee also believed that these new definitions were more consistent with estimates from the unemployment insurance (UI) system. (The differences in the number of employed and unemployed due to these definitional shifts were not large enough to constitute a break in series.)¹⁸

One conceptual issue that surfaced in discussions of unemployment during the 1940s and 1950s was the classification of people who were not looking for work because of the “belief that no work was available in their line of work or community.” Technically, these people did not meet the activity test for unemployment. Although the survey questionnaire typically did not ask people why they did not look for work, interviewers were instructed to classify individuals who volunteered this information as unemployed. This subjectivity would later be debated in comprehensive reviews of labor statistics in the early 1960s and the late 1970s.¹⁹

In terms of understanding the data collected from the CPS, one facet of time-series data on the labor force and its components that challenged early analysts was the underlying seasonality in the measures. Large fluctuations in employment and unemployment levels during the course of the year made it difficult to detect underlying economic movements in the data. (Employment in agriculture at the time was a much larger share of overall employment than it is today: about 1 in 5 workers in the early 1940s compared with fewer than 1 in 50 workers now.) The large seasonal swings in survey data led analysts at the WPA to abandon the calculation of the unemployment rate and replace it with a ratio of unemployment to the population. The goal was to produce a more stable metric for monthly comparisons.²⁰ What was really needed, however, was seasonally adjusted data on unemployment (which first became available in the mid-1950s). The purpose of seasonal adjustment was to remove regular seasonal variations in the data so that underlying changes in data trends, especially month-to-month comparisons, could be observed. At first, computations needed to “smooth” the data were done by hand, and this limited the number of series that could be adjusted for seasonal patterns. With the advent of computers, however, the volume of seasonally adjusted data from the survey expanded greatly.²¹

Another issue that persisted through the period had to do with which agency should be in charge of the survey. Although the Census Bureau had taken over the household survey in 1942, not everyone had been satisfied with that decision. A number of agencies, including the Bureau of Labor Statistics (BLS), were considered as a new home. After much debate, an agreement was reached to better coordinate both the release and analysis of data on the labor market. In 1959, the analysis and dissemination of household survey data was transferred to BLS—the CPS was, after all, a survey of labor—but the Census Bureau retained responsibility for collecting the data.²²

In terms of actual measures of labor market performance, the CPS estimates indicated some troubling signs in the late 1950s. Data showed that the unemployment rate was still above 5 percent at the end of 1959, nearly 2 years after the end of the last recession. Moreover, it had not returned to levels observed in previous recoveries. This relatively high unemployment rate was viewed by some against the backdrop of continued automation in the workplace. With the introduction of mainframe computers, innovation again fueled fears that manual jobs would be lost to machines. These concerns were sufficient for congressional hearings to be held on the subject in both 1955 and 1960.²³

By the end of the 1950s, CPS concepts and methods had been thoroughly reviewed and were considered capable of capturing data essential for evaluating the state of the job market and, by extension, the overall economy. To be sure, there was still some dissent on definitional issues regarding the measurement of unemployment. But there was nothing that amounted to a wholesale indictment of survey concepts or the integrity of the economists and statisticians who analyzed and published the data. As the decade came to a close, that was about to change.

1960s—is unemployment really this high?

Unemployment began to grow again after the onset of a recession in April 1960. The rising unemployment rate again piqued public interest in its measurement. In particular, the September 1961 issue of Reader’s Digest featured an article that claimed definitions and methods had been modified to “magnify the unemployment problem,” thus justifying the need for more government spending and regulation. The author, James Daniel, alleged that BLS deliberately manipulated the data—a claim that garnered a great deal of attention. This increased attention led to a third major review of the sources of labor market data.²⁴

Gordon Committee review

In November 1961, the President’s Committee to Appraise Employment and Unemployment Statistics, chaired by Robert A. Gordon, was charged with evaluating the collection and tabulation of the full range of labor market data being gathered at the time, including the labor force concepts and published analyses of CPS data. After nearly a year of deliberation, the Gordon Committee issued a 412-page report that concluded that the system of reporting the official employment and unemployment data was in fact objective, scientifically reliable, and free from deliberate manipulation. The Committee also supported and reaffirmed the activity-based concept used to measure unemployment. However, it did suggest several noteworthy changes to the CPS questionnaire.²⁵

With respect to the measurement of unemployment, one problem the Committee found was the potential for interviewers to include among the unemployed those who volunteered (mentioned in passing) that they were not looking for work because they believed there were no jobs available for them (although there was little information on the extent to which this happened). More specifically, the Committee determined that “reliance on subjective attitudes and volunteered information should be minimized.”²⁶

The Committee also recommended other refinements to the CPS concept of unemployment—such as an explicit timeframe to define current job-search activity and an emphasis on the specific steps taken to find work. It also advocated collecting additional information on people not in the labor force. Following these recommendations, BLS and the Census Bureau developed questions to measure the new criteria and designed a research sample, called the Monthly Labor Survey (MLS), to test these questions. The MLS was administered beginning in April 1964. After more than 2 years of testing, many Gordon Committee recommendations were implemented in January 1967.²⁷

Notably, the unemployment concept, which had generally included people who had no job but were seeking work, was tightened by incorporating two new criteria: an availability test and an explicit 4-week job-search period. The availability test required a person actually be available for work at the time of the survey to be counted as unemployed. The 4-week job-search window introduced a defined period during which job search must have occurred. The latter criterion was introduced to ensure consistency across all respondents—previously, the job-search window was not specified in the questionnaire.²⁸

The Gordon Committee wanted the unemployed to have tested the job market in a reasonably recent period, but they did not feel strongly about what that period should be. Addressing this recommendation, BLS settled on a period longer than 1 week because job search could include periods of inactivity and waiting. The 4-week period was chosen because it indicated a recent search that lined up rather well with the CPS measurement of employment (which classified as employed anyone who worked at least 1 hour for pay or profit in the previous week) and also allowed for the ebb and flow of job search.²⁹

At the Committee’s recommendation, a question was also added about job-search methods: “What has…been doing in the last 4 weeks to find work?” The purpose of the question was to ensure that those included among the unemployed were people making specific efforts to find work.³⁰

In addition, new interviewer instructions excluded from the unemployed those who volunteered that they were not looking for work because they were discouraged—whittling away at some of the remaining subjectivity that existed in the measurement of unemployment. New questions were also introduced at the Committee’s recommendation to capture data on the labor force attachment of people who were neither working nor looking for work. These questions included asking whether these individuals wanted to work now and the reason they were not currently looking for work.³¹

With the implementation of the Gordon Committee’s recommendations in 1967, the unemployed now included those who had no job during the survey reference week, but who had actively looked for work in the 4 weeks prior to the survey and were available to take a job if one were offered (people on temporary layoff continued to be classified as unemployed whether or not they were looking for work). People who wanted a job but did not look for work because they thought no work was available for them would no longer be included among the unemployed, but instead were counted among those not in the labor force, as “discouraged workers.”

Committee weighs the unemployment rate as a measure of hardship

Early on in the history of the survey, some saw the unemployment rate as an indicator of economic hardship.³² Some critics thought the official estimate of unemployment did not adequately measure hardship. They argued that the estimate was too high and that not all who wanted or looked for work were truly experiencing hardship. It was claimed that many unemployed people, such as housewives, students, or recent high school graduates, were not “in need” of a job. After weighing this consideration, the Gordon Committee concluded that unemployment could be only an indirect measure of hardship, because although work is related to personal income, hardship is more directly associated with family income. The Committee determined that the CPS provides a wealth of related statistics that reveal the relationship between joblessness and hardship, including estimates of family income from the CPS March supplement (now known as the ASEC). The Committee stressed that no one number can capture the intricacies of economic hardship encompassed in labor market difficulties.³³

The income data collected in the ASEC became even more important beginning in 1965, when Mollie Orshansky at the Social Security Administration developed the first official poverty line, the level of income below which a person is considered poor.³⁴ Coupled with this threshold, CPS ASEC data were used to publish a key metric of economic hardship—the poverty rate.³⁵

Unemployment in states and local areas

Beginning in the 1970s, CPS data were used to achieve another goal established by the Gordon Committee—estimating comparable unemployment figures for subnational areas. While state and local estimates of unemployment were already being published, these estimates were not produced by BLS. Further, they were based mostly on administrative data from the UI system, and they did not correspond to the unemployment concept used in the CPS.

The Committee was critical of the validity of these estimates, and independent studies confirmed biases and inaccuracies in estimation procedures. In 1972, the Office of Management and Budget (OMB) transferred the responsibility for subnational unemployment estimates to BLS. In November 1973, BLS began using CPS data, with other inputs, to produce the state and local area unemployment estimates. The objective of this innovation was twofold: (1) to introduce more conformity between the national CPS estimates and the subnational estimates and (2) to achieve more consistency in procedures among the states.³⁶

1970s—a new economic conundrum: stagflation

Before the 1970s, conventional economic theory suggested that the unemployment rate and inflation had an inverse relationship. During the 1970s, this relationship appeared to “break.” After increasing gradually through the 1950s and 1960s, inflation increased rapidly in the early 1970s—driven in part by the October 1973 oil crisis—while the unemployment rate rose markedly. The unemployment rate reached 9.0 percent in May 1975, when inflation hovered around 10 percent, contradicting the economic theory of the day.³⁷ This economic puzzle gave birth to a new concept—stagflation—a phenomenon where inflation and unemployment rise in tandem. Stagflation, in turn, drove the economist Arthur Okun to invent the “discomfort factor,” later called the “misery index”—the sum of the unemployment rate and the annual inflation rate.³⁸

Politics and the unemployment rate

With inflation and unemployment both running high in the 1970s, economic indicators were receiving more attention from both the public and President Nixon’s administration. In February 1971, tensions surfaced between the administration and BLS. At issue was a discrepancy between the way BLS characterized a 0.2-percentage-point decline in the unemployment rate—as “marginally significant” during a press briefing—and the Secretary of Labor’s characterization that it was “of great significance.” A similar disagreement followed the next month when the Secretary called a similar decline in the unemployment rate “heartening,” but BLS described it as “sort of mixed” in light of other labor market indicators. Consequently, the administration terminated BLS press briefings, prompting allegations of politicization, congressional hearings and reports, and investigations by statistical associations and the Industrial Relations Research Association. Following these frictions, in July 1971, a third incident stirred tensions between the administration and BLS: a technical error in seasonal adjustment of the June data led to a relatively large correction, reducing the originally published decline in the unemployment rate by half.³⁹

This friction between the administration and BLS culminated in an OMB directive for a reorganization of federal statistical agencies, which led to further accusations of politicization. To maintain the statistical system’s credibility, OMB issued a series of directives during the 1970s and early 1980s to combat the potential for politicization.⁴⁰ By 1985, several guidelines regarding Principle Federal Economic Indicators, including the unemployment rate, had been codified in OMB Statistical Policy Directive 3. This directive established new procedures for announcing news-release schedules in advance, restricting prerelease access to data, prohibiting related policy commentary until 1 hour after the statistical release, and announcing changes in data collection, analysis, or estimation methods in advance.⁴¹

Alternative measures of labor underutilization—expanding on the unemployment rate

For much of the survey’s history, some have suggested a need for a range of alternative indicators of unemployment and labor underutilization that would paint a more complete picture of joblessness and labor-market-related hardship.⁴² In 1976, BLS addressed this concern. Commissioner Julius Shiskin introduced seven alternative measures—U-1 through U-7—because “no single way of measuring unemployment can satisfy all analytical or ideological interests.”⁴³ While public debate about labor force concepts and definitions stirred, these new alternatives packaged CPS data in different ways to meet varying viewpoints about what should and should not be included in the definition of unemployment. Measures U-1 through U-4 were introduced to assuage those still convinced that the Reader’s Digest article from 1961 was correct in that unemployment was defined too broadly. For their part, measures U-6 and U-7 were introduced to satisfy those who believed the official definition of unemployment (identified at the time as U-5) overlooked people who had stopped looking for work and those who were “partially unemployed.”⁴⁴ In the end, regardless of how narrowly or broadly one defined unemployment, none of these alternatives had (or has) an advantage in terms of cyclical analysis. This remains true even after the 1994 CPS redesign introduced a new range of alternative indicators. The new measures, U-1 through U-6, all exhibit very similar trends during both recessions and recoveries.⁴⁵

Counting the labor force—Levitan Commission recommendations

In addition to fueling tension between the administration and BLS, the high unemployment rates during the recessions of the early 1970s once again revived doubts about the relevancy and objectivity of the unemployment rate.⁴⁶ Despite previous reviews, a fourth major review of labor market statistics from various surveys was organized in 1976 as the National Commission on Employment and Unemployment Statistics, chaired by Sar A. Levitan. Changing economic trends, such as women’s increasing participation in the labor force and new social policy initiatives, helped stir this new debate over the adequacy of labor force concepts and methods. Again, the main debate regarding CPS data centered on how broadly unemployment should be defined and whether it should be a measure of economic hardship.⁴⁷

After 2 years of deliberation, the Levitan Commission found the main labor force definitions were sound and praised the general integrity and objectivity of BLS. While making only minor recommendations about the concepts of employment and unemployment, the Levitan Commission did recommend more substantive changes to the measurement of labor market discouragement. The Commission also recommended that the scope of the survey be expanded beyond the civilian population, to include the Armed Forces. The argument was that the Armed Forces should be represented in labor force statistics because, after the draft was eliminated in 1973, workers could choose between military and civilian employment. Further, the Commission found a need for even more demographic detail on people who faced challenging labor market conditions.⁴⁸

Overall, the Commission found no major problems with the definition of unemployment as it stood, but recommended changes at the margins to further tighten the measure. At the time of the review, interviewers were instructed to count as unemployed people who volunteered they were participating in a paid work experience program or were enrolled in a job training program but were not seeking work. (While there were no estimates of the number of program participants who were classified as unemployed, evidence at the time suggested that few program participants volunteered this information.) As with previous reviews, the Commission reverted to the activity test and recommended that people in work experience programs who were receiving a form of pay be counted among the employed. Additionally, the Commission recommended that people in job training programs be treated as were other students—that is, classified as not in the labor force if they were not actively seeking work. To adhere more strictly to the activity concept, these changes again modified interviewer instructions, but they did not alter the definition of unemployment. The unemployed were still people without jobs who were actively seeking a job and were available to begin work.⁴⁹

During this review, the issue of how to classify discouraged workers arose again. Like the Gordon Committee, the Levitan Commission decided it was appropriate to exclude these people from the count of the unemployed. After analyzing trends in labor market movements under changing economic conditions, the Commission concluded that the measure was not as strong an indicator of potential labor supply as some had previously argued it was. Further, the measure of discouraged workers was not considered to be as objective as the measure of unemployment.⁵⁰

After confirming that the discouraged should not be counted among the unemployed, the Commission zeroed in on the reliability of the discouraged-worker measure. It pointed out that the current construct did not test an individual’s availability for, or their willingness to, work. As a result, the Commission recommended the concept of discouraged workers be refined to include only those who had demonstrated some attachment to the labor market, as evidenced by a job search in the recent past.⁵¹

Levitan Commission deliberates unemployment and hardship

Like the Gordon Committee, the Levitan Commission debated the idea of unemployment as a measure of hardship. In their deliberations, they recognized a complex relationship between labor force status and economic hardship. They pointed out that, since the Gordon Committee last weighed the issue in the early 1960s, income-support programs had further weakened the relationship between unemployment and economic hardship. Additionally, they considered that employment did not necessarily guarantee that a family could live above the poverty threshold. Because of this complex relationship, the Commission agreed that the unemployment rate’s greatest value was as a cyclical indicator, but that accompanying data could be used to measure hardship. In this regard, they recommended an annual report linking labor market status to poverty with the use of ASEC income data and the poverty threshold.⁵² BLS subsequently issued the Linking Employment Problems to Economic Status report that linked earnings and income data for 1979 to three specific labor market experiences that could cause people to experience economic hardship: employment with low wages, unemployment, and low labor force participation. This report was later replaced by A Profile of the Working Poor.⁵³

1980s—double-dip recessions and displacement

In the early 1980s, there were back-to-back recessions associated with especially large job losses in manufacturing. The unemployment rate reached a high of 10.8 percent in November 1982. The double-dip recession, along with structural factors such as automation and import competition, disproportionately affected manufacturing industries (most notoriously steel mills). This fueled a fear that workers with long tenures and relatively high pay would lose their jobs and be unable to find similar work.

This concern triggered interest in data on workers who were displaced from their jobs, prompting the collection of such data in a biennial supplement to the CPS. This supplement tallied workers who lost or left jobs because their plant or company closed or moved, because there was insufficient work for them to do, or because their position or shift was abolished. A key metric calculated from these data was the reemployment rate, or the percentage of workers who were displaced but were working again at the time of the survey. Reemployment rates from the first survey conducted in 1984 underscored the upheaval that was taking place in manufacturing. While 60.1 percent of all displaced workers who lost or left their jobs since January 1979 had been reemployed in January 1984, that number was considerably lower for those in primary metal manufacturing (the aforementioned steel mills), at 45.7 percent. (It should be noted that displaced workers who were reemployed may not have found work in the same industry.)⁵⁴

Implementing the Levitan Commission’s recommendations

Also in the 1980s, several of the Levitan Commission’s recommendations were implemented. Beginning in 1983, selected labor force statistics included the resident Armed Forces (although they were subsequently removed in the mid-1990s because of data-quality concerns).⁵⁵ Since the Armed Forces are by definition employed, the unemployment rate including the resident Armed Forces was usually about 0.1 percentage point lower than the civilian rate.⁵⁶ In 1984, the CPS began classifying unpaid job training participants who were not actively searching for work as not in the labor force and paid job training participants as employed. (See earlier discussion of Commission recommendations.) However, this information was volunteered by so few respondents that there was no measurable effect on labor force estimates.⁵⁷

Over time, technological advancements continued to change the way people work, as well as the way labor market data were collected and analyzed. Just as computers became more prevalent in all facets of industry in the 1980s, more widespread use of personal computers changed how statistical surveys could be administered and the amount of data that could be collected. As a result, in the late 1980s, computer-assisted telephone interviewing from centralized telephone centers was introduced. Soon thereafter, BLS and the Census Bureau began to develop a fully automated computerized CPS questionnaire that would make its debut in 1994.⁵⁸

1990s—survey redesign

The U.S. economy contracted again in the early 1990s, as business and consumer confidence suffered several setbacks.⁵⁹ While the increase in the unemployment rate was not as pronounced as in the downturns of the mid-1970s or early 1980s, the unemployment rate continued to climb for 15 months following the official end of the recession, reaching a high of 7.8 percent in June 1992. This was an atypical pattern, as historically the unemployment rate’s peak tended to coincide more closely with the end of the recession.⁶⁰ Moreover, this recession was characterized by a shift toward “permanent” and more widespread job losses, with white-collar jobs hit hard—particularly in finance, insurance, and real estate industries—areas relatively unaffected in previous recessions.⁶¹

Meanwhile, the CPS was on track to undergo a survey redesign in 1994. One goal of the redesign was to enhance the quality of the data by implementing fully automated computer-assisted interviewing. In line with this goal, the CPS became the first monthly federal government survey with fully automated data collection.⁶² Another focus of the redesign was to revise the questionnaire to obtain more precise and consistent estimates of the labor force. Cognitive research and testing were used extensively to determine question wording and sequence in the new survey instrument. In addition, the CPS redesign implemented many of the remaining recommendations of the Levitan Commission, including modifications to the discouraged-worker measure, which had been deemed too subjective. Finally, an auxiliary, but still important, reason for the redesign was to collect more information about the U.S. population of working age.⁶³

The same technological advancements that made computer-assisted interviewing possible in the redesign also fueled a growing demand for more data. With computer use on the rise, and the expansion of the Internet, researchers were now equipped with the technologies necessary to conduct their own data analyses, and their demand for survey microdata (individual data records) grew. Public-use microdata files had been available since the late 1960s on magnetic tapes, but in the 1990s microdata became available online.⁶⁴ For the first time, people interested in conducting their own analyses could access CPS microdata through the Internet.

In a search for more objectivity and precision, the measurement of labor force concepts was further sharpened as part of the redesign effort. One such concept was that of discouraged workers, or people outside the labor force who wanted a job but had not looked for work in the past 4 weeks because they believed no jobs were available for them. Until the redesign, individuals were not required to have looked for work at any time in order to be counted among the discouraged. Individuals also were not asked if they were available to work. In line with the Levitan Commission’s recommendation for a more concrete, objective definition of discouraged workers, the availability test was made more explicit by asking whether an individual was free to begin work in the prior week. Moreover, individuals would have to have looked for work in the past year as a way of demonstrating some attachment to the job market. (See earlier discussion of Commission recommendations.)

In addition to tightening the definition of discouraged workers, the redesign introduced a broader group of people who were not in the labor force but were attached to the job market. This group—which came to be known as people marginally attached to the labor force—included not only discouraged workers, but also people who cited nonlabor-market-related reasons for not currently looking for work (such as the need for childcare or school attendance).

Other changes introduced with the 1994 redesign included shifting a small group of people waiting for a job to begin—those who did not look for work in the 4 weeks preceding the survey—from unemployed to “not in the labor force.” This change was made to conform more strictly to an activity-based measure. Refinements were also made to the measurement of several other labor force concepts, including hours of work, people on temporary layoff, reasons for working part time, and earnings.⁶⁵ In addition, some of the alternative measures of labor underutilization were redefined to take advantage of new information collected in the redesigned survey.

2000s—economic expansion, natural disaster, and the Great Recession

For much of the 1990s, there was relatively little debate about the accuracy and relevance of the unemployment measure used in the CPS. Perhaps this should not be surprising, since unemployment was not as high for most of this period as it had been in earlier years of the survey. During the record-long expansion of the 1990s, the unemployment rate fell to around 4 percent, a level not seen since the late 1960s. While the economy contracted again in early 2001, the downturn was shallow by historical standards, with the unemployment rate climbing to 6.3 percent in June 2003. Yet, as in the early 1990s, the job market was slow to recover, again with a “jobless recovery” and a lagged peak in the unemployment rate.⁶⁶

Hurricane Katrina

In 2005, the CPS was used to quickly obtain information on the labor market in the aftermath of Hurricane Katrina, one of the deadliest and most calamitous hurricanes to hit the United States. The storm severely affected the Gulf Coast region, leaving many homeless and without work. Katrina separated thousands of workers from their livelihoods, and the effects rippled across the nation as evacuees from the storm relocated throughout the country. From October 2005 to October 2006, the CPS included questions to identify evacuees and to capture the impact of their diaspora on the labor market. The addition of these questions marked the first time the CPS was modified specifically to gather data on the labor market impact of a natural disaster.⁶⁷ The questions helped show how many evacuees returned to the area after Katrina, how many were displaced for long periods, and how many were still relocated to new areas. These new data also provided insights into the unemployment picture among those affected by the storm. For example, jobless rates were much lower for evacuees who returned to their homes than for those who relocated.⁶⁸

The Great Recession

In December 2007, the U.S. economy was affected by another crisis: the long-running housing bubble had burst, leading to the onset of what would become the most severe recession in the United States since the Great Depression.⁶⁹ The CPS provided a wide array of information on the deteriorating labor market conditions during the Great Recession. During this downturn, both the unemployment rate and the number of unemployed rose to levels not seen since the deep downturn of the early 1980s; the unemployment rate peaked at 10.0 percent in October 2009. The median duration of unemployment rose to 25.2 weeks in June 2010, and the number of people unemployed for 6 months or longer (27 weeks or more) reached unprecedented levels—almost half of the unemployed were jobless for 6 months or longer. Further, the proportion of employed people who worked part time for economic reasons, often referred to as involuntary part-time workers, had also reached a record level at 6.7 percent in March 2010.⁷⁰ The labor force participation rate, already trending down since 2000, began a sharp decline during the contraction. This led critics again to question whether the unemployment rate is an accurate reflection of the employment situation, especially in an economy with dwindling labor force participation.⁷¹ There also arose a renewed interest in alternative measures of labor underutilization.

2010s—a slow recovery

Following the end of the Great Recession,⁷² the jobless rate trended down from late 2009 through the end of 2017. At the time this article was written, however, several labor market indicators still had not returned to their prerecession levels. Although the number of people unemployed and the unemployment rate had dropped below levels experienced before the recession, the problem of long-term unemployment persisted, and involuntary part-time work remained somewhat elevated. More than 8 years after the end of the recession, about one-quarter of the unemployed were looking for work for 6 months or longer, still high by historical standards, and about 5 million people who wanted full-time work had to settle for part-time work. This, coupled with a prolonged period of low labor force participation rates,⁷³ resulted in continued interest in alternative measures of labor underutilization, since the broader measures were associated with larger groups of people experiencing labor market difficulties.⁷⁴ The question of how broadly unemployment ought to be defined also reentered the public discourse, as some observers once again challenged the relevance of the official unemployment rate.⁷⁵

The CPS—standing the test of time

For more than three-quarters of a century, the CPS has been a vital tool for providing information on the state of the job market. While initially designed to obtain data on the employment status of individuals (such as whether or not individuals were employed or unemployed), survey questions have been expanded over the years to capture additional labor market data. Because the CPS is a survey of the population, its data can be cross-classified by a full range of demographic characteristics—including age, gender, race, and educational attainment, to name a few—allowing researchers and policymakers to examine and track the labor market experiences across specific population groups. And it was recognized early on that supplemental questions could be added to the survey on a wide range of topics, some unrelated to the job market.

Despite the vast amount of labor market data the CPS captures, people continue to focus mostly on the unemployment rate. Some critics have revived the argument that the unemployment rate is flawed because it is not an all-encompassing measure of labor market conditions and hardship, although it was never intended to be so. Some call for a broader definition of unemployment. In fact, it has been suggested that the job-search requirement for unemployment is far too stringent—ironically, circling the discussion back to the measurement issues that resulted in the survey in the first place.

These arguments about the relevance and reliability of labor force concepts are not new or even particular to the current economic climate. During the survey’s history, such questions led to several wholesale reviews of CPS concepts. In fact, the definition of unemployment has been tightened at the margins after each review. However, these reviews also concluded that, fundamentally, the measure was objective and reliable. Hence, the basic concept has remained largely intact—the unemployed are people who are not working, but are currently looking for work and are available to take a job—and the historical comparability of the unemployment rate has been preserved. In the end, it is only by studying the full range of information collected on the job market—not just from the CPS, but also from various other data sources—that one can accurately assess the health of the labor market and the overall state of the U.S. economy.