Several major changes affect the Current Population Survey (household) data
— additional details
From the Employment Situation news release, February 7, 2003,
USDL 03-46.
———————————————————————————————————————————————-
Revisions to Household Survey Data
Several major changes affect the Current Population Survey (household) data being released today. These include the use of new population controls that reflect Census 2000 results and new
information about net migration, the use of new questions about race and Hispanic ethnicity, the introduction of new industry and
occupational classifications, improvements in seasonal adjustment procedures, and the annual update of seasonal adjustment factors. For all data series, these changes affect the comparability of the
January 2003 estimates with those for earlier months. Additional information is provided beginning on page 5.
———————————————————————————————————————————————-
- 5 -
Revisions to Household Survey Data
Effective with the release of data for January 2003, a number of
changes affect estimates from the Current Population Survey (CPS), or
household survey. These changes were undertaken to benchmark the survey
data to more current estimates of the U.S. population; to adopt new
standards for data on race, ethnicity, industry, and occupation; and to
improve seasonal adjustment procedures. In addition, the data reflect
updated seasonal adjustment factors and revisions to historical seasonally
adjusted data. The following describes these changes in more detail and
provides information regarding their impact on the survey estimates. As
discussed below, a one-month change in weighting procedures makes data for
January not strictly comparable with data for both prior and subsequent
months.
—Beginning in January 2003, population controls based on the results of
Census 2000 were used in the monthly CPS estimation process, and data for
the 3 prior years were revised to reflect these new controls. (Previously,
estimates for January 1990 through December 2002 were based on 1990 Census
population controls adjusted for the estimated undercount.) The Census
2000-based controls increased the size of the civilian noninstitutional
population by over 3 million. As a result, they also increased the
estimates of employment and unemployment. Since the increases were roughly
proportional, however, the overall unemployment rate did not change
significantly.
—In addition to the shift to Census 2000-based controls, the U.S. Census
Bureau introduced another large upward adjustment (+941,000) to the CPS
population controls in January 2003 as part of its annual update of
population estimates. BLS does not anticipate revising historical data
to reflect this additional adjustment. This accords with usual practice
regarding intercensal population changes. Thus, the population levels
for January 2003 (and beyond) and the levels of employment and unemploy-
ment are not strictly comparable with those for earlier months. The ad-
ditional adjustment to the population taken in January 2003 raised the
number of employed by about 575,000 and the number of unemployed by about
40,000. The adjustment had a negligible effect on the overall unemploy-
ment rate and other ratios.
—The questions on race and Hispanic origin in the CPS were modified in
January 2003 to comply with the new standards for maintaining, collecting,
and presenting federal data on race and ethnicity. A major change under
those standards is that respondents may select more than one race when
responding to the survey. Respondents continued to be asked a separate
question to determine if they are Hispanic, which is considered an
ethnicity rather than a race. Under the new standards, the question on
Hispanic ethnicity was reworded and placed before the question on race.
Persons who report that they are Hispanic or Latino also are classified
separately by the race (or races) they consider themselves to be. Based
on the evidence obtained from a special supplement to the CPS in May
2002, the new questions have little effect on the unemployment rate over-
all and for most major worker groups. An exception was the jobless rate
for Hispanics; the May 2002 results indicate that their unemployment rate
may be somewhat higher based on the new questions.
- 6 -
—To accommodate the race categories under the new guidelines, the Employ-
ment Situation news release now presents data for persons who report they
are white (and no other race), black or African American (and no other
race), and Asian (and no other race). Results from the May supplement
indicate that the unemployment rates for these groups are comparable to
the rates for these categories using the old classifications that allowed
only one race to be identified. The published labor force estimates for
the groups, however, are smaller because those estimates no longer include
persons who report that they belong to more than one race. In addition,
under the old classification scheme, Asians were included in a single
category with Pacific Islanders. There are now two separate categories:
(1) Asians and (2) Native Hawaiian and other Pacific Islanders. Data for
three race categories—Native Hawaiian and Pacific Islanders, Native
Americans, and persons of more than one race—are included in the total
employment and unemployment estimates in this release but are not shown
separately because of their small sample sizes. Estimates are presented
separately for Hispanics, who can be of any race.
—Adjustments were made to both the second-stage and composite weighting
procedures to adapt them to the new race/ethnic classification system.
These adjustments also should enhance the stability over time of national
and state/substate labor force estimates for demographic groups. Com-
posite weights could not be calculated for the January 2003 data, however,
because that procedure requires the use of both the current and the pre-
vious month's information. Because the race/ethnic groups differ between
December and January, compositing could not be done for January. The
effect of compositing is different each month; thus, January estimates
could not be adjusted to make them comparable with those for other months.
The effect on the national unemployment rate is probably negligible.
—In January 2003, the CPS adopted the 2002 Census industry and occupa-
tional classification systems derived from the 2002 North American
Industry Classification System and the 2000 Standard Occupational
Classification system. These new classification systems create breaks
in the time series for industry and occupational data at all levels of
aggregation. As a result, the former industry and occupational cate-
gories have been discontinued. Employment and unemployment estimates
using the new industry and occupational classifications were developed
for 2000-02 by recoding previously collected information. Still, it
will take several additional years under the new classification systems
to collect enough industry and occupational data to allow for seasonal
adjustment.
Additional information about these changes and their impact can be found
in an article in the February 2003 issue of Employment and Earnings and on
the BLS Web site at (http://www.bls.gov/cps/).
Revision of seasonally adjusted data
At the end of each calendar year, the Bureau of Labor Statistics reesti-
mates the seasonal factors for the household data series by including
another full year of data in the estimation process. Based on this annual
reestimation, BLS issues projected factors for the first 6 months of the
new year, as well as revised estimates of historical seasonally adjusted
data for the last 5 years. This year, seasonally adjusted data for January
1998-December 2002 were subject to revision.
With the release of January 2003 data, the CPS also began using the X-12
ARIMA software for seasonal adjustment of time series data. This replaces
the X-11 ARIMA, used since 1980. The seasonal adjustment of CPS series was
reviewed to determine if additional series could be adjusted and if the cur-
rent adjustments were sound, particularly in light of the changes made to
race, ethnic, industry, and occupation classifications. The seasonal adjust-
ment of a number of series was discontinued or modified as a result of this
review.
- 7 -
As previously discussed, seasonal adjustment of series on industry and
occupation were discontinued. The new industry classification system also
led to a modification of the procedure for seasonally adjusting total CPS
employment and, indirectly, labor force levels and the unemployment rate.
Previously, the seasonally adjusted estimate of household employment was
obtained by adding up the directly-calculated seasonally adjusted estimates
for eight age-sex-industry series. These eight series consisted of teenage
men, teenage women, men age 20 and over, and women age 20 and over in
agriculture and in nonagricultural industries. The new industrial
classification system decreased the size of agriculture substantially by
moving some industries to the nonagricultural sector. When the smaller
agriculture estimate was disaggregated by sex and age, several of the
components became extremely small. For this reason, only total employment
for the four age and sex groups is being directly seasonally adjusted, and
the sum of these four groups constitutes the seasonally adjusted total
employment level. This change in procedure does not appreciably affect the
measures. Seasonally adjusted series for agriculture and nonagricultural
industries are now being obtained by direct adjustment.
Table B
summarizes the effects of the revisions on the overall unemployment
rate since January 2002. The rate was revised in 8 months, in each case by
0.1 percentage point, due primarily to the updating of seasonal adjustment
factors. Revised seasonally adjusted data for major labor force series since
December 2001 appear in table C.
Additional information about the new seasonal adjustment procedures can
be found in an article in the February 2003 issue of Employment and
Earnings. The publication also will contain the new seasonal adjustment
factors for major series for the January-June 2003 period and revised data
for the most recent 12 months for all regularly published tables containing
seasonally adjusted household survey data.
Changes in data presentation
Due to the changes reported above, revisions or additions have been
made to the A tables in this release. Several series have been affected by
conceptual changes or new standards for presenting data; new series have
been added and others will no longer be published. The most important
changes to the tables are described below.
- 8 -
Table A-1. Employment status of the civilian population by sex and age.
Estimates of employment for agriculture and nonagricultural industries no
longer appear in this table. These series have been modified to reflect the
new industrial classification system and seasonal adjustment practices; they
now appear in "Table A-5. Employed persons by class of worker and part-time
status."
Table A-2. Employment status of the civilian population by race, sex,
and age. Estimates presented in this table have been modified to comply
with the new standards for presenting data on race. Hence, data for
January 2003 forward are presented for persons who report that they are
white (and no other race), black or African American (and no other race),
and Asian (and no other race). Race-specific unemployment rates for 16- to
19-year-old men and women are no longer presented. Estimates for persons
whose ethnicity is identified as Hispanic or Latino, formerly presented in
this table, are now presented in "Table A-3. Employment status of the
Hispanic or Latino population by sex and age."
Table A-3. Employment status of the Hispanic or Latino population by
sex and age. This new table presents estimates for persons of Hispanic or
Latino ethnicity. The detail has been expanded to include not seasonally
adjusted data for the major sex and age groups.
Table A-4. Employment status of the civilian population 25 years and
over by educational attainment. This table replaces old table A-3.
Table A-5. Employed persons by class of worker and part-time status.
This new table contains employment estimates by class of worker and persons
at work part time, previously shown in old table A-4. As noted above,
employment estimates for the revised total agriculture and related
industries and for total nonagricultural industries are now presented in
this table. Estimates for unpaid family workers (in both agriculture and
related industries and nonagricultural industries) and for private
household workers will no longer be seasonally adjusted. Employment
estimates by marital status now are presented in "Table A-6. Selected
employment indicators," and employment estimates by occupation are
presented in "Table A-10. Employed and unemployed persons by occupation,
not seasonally adjusted."
Table A-6. Selected employment indicators. This new table contains
employment estimates by sex and expanded age detail. It also contains
employment estimates by marital status and full- and part-time status.
Seasonally adjusted estimates for women who maintain families no longer
will be presented.
Table A-7. Selected unemployment indicators, seasonally adjusted.
This table replaces old table A-9 and has been expanded to include greater
age detail for unemployment estimates. It also contains estimates by
marital status and full- and part-time status. Seasonally adjusted
estimates for women 55 years and over and for women who maintain families
no longer will be presented; alternatively, not seasonally adjusted data
will be presented.
Data for unemployed persons by reason for unemployment are now
published in table A-8 and data for unemployed persons by duration of
unemployment now appear in table A-9.
Table A-10. Employed and unemployed persons by occupation, not
seasonally adjusted. This new table contains employment and unemployment
estimates for major occupational groups based on the 2002 Census
occupational classification system derived from the 2000 Standard
Occupational Classification system.
- 9 -
Table A-11. Unemployed persons by industry, not seasonally adjusted.
This new table contains unemployment estimates for major industry groups
based on the 2002 Census industry classification system derived from the
2002 North American Industry Classification System.
Table A-12. Alternative measures of labor underutilization, and Table
A-13. Persons not in the labor force and multiple jobholders by sex, not
seasonally adjusted replace old tables A-8 and A-10, respectively, with no
changes to format.
Changes to LABSTAT
LABSTAT, the electronic method for retrieving data series published in
the "A" tables of this release, also has undergone change. All LABSTAT
series IDs associated with the household survey data have been revised. A
substantial number of series related to the former occupation, industry,
and race classifications have been discontinued. Similarly, new series
have been added for Asians and for the current occupational and industry
classifications. Users can access data via the most requested series
option without the new LABSTAT codes. Users also can access historical
data for the household series contained in the "A" tables of this release
at (http://www.bls.gov/cps/cpsatabs.htm). Users who access the data
series via series report or FTP can obtain a crosswalk to the new LABSTAT
codes at (ftp://ftp.bls.gov/pub/special.requests/lf/).
Last Modified Date: April 2, 2003
