The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Technical Description of the
Quarterlf Data on Weekly
Earnings from the Current
Population Survey
U.S. Department of Labor
Bureau of Labor Statistics
January 1982
Bulletin 2113
l-Ml
1982
For sale by the Superintendent of Documents, U.S. Government Printing Office
Washington, D.C. 20402 - Price $2.25
Iddtoidal Description of tod
Ctoartdrl^ Data ©n W©akS^
Earnings fr©m the Cyrrant
SPapygatm Syrwa^f_________
U.S. Department of Labor
Raymond, J. Donovan, Secretary
Bureau of Labor Statistics
Janet L. Norwood, Commissioner
January 1982
Bulletin 2113
-
In 1979, the Bureau of Labor Statistics, working
through the Bureau of the Census, began to collect data
on weekly earnings by demographic groups through
out the year rather than just once a year as had previ
ously been the case. These data, which are now pub
lished on a quarterly basis, are collected through the
Current Population Survey (CPS). This bulletin pro
vides background information on the CPS and describes
how the new quarterly earnings data are collected and
processed and how they compare with the earnings data
previously collected once a year. The merits and limi
tations of the data, particularly with regard to sampling
error, also are discussed.
This bulletin was prepared in the Office of Current
Employment Analysis by Earl F. Mellor. Material in
this publication is in the public domain and may, with
appropriate credit, be reproduced without permission.
© © P fie in iS s
Page
The Current Population Survey (CPS) ..................................................................................................................
Monthly collection of earnings data in the CPS ..............................................................................................
Processing of the d a ta ................................................................................................................
Comparability of the new and old series ..................................................................................
Comparability to establishment data ........................................................................................
Reliability of the data ................................................................................................................
Definitions of terms used in reports on usual weekly earnings ..............................................
Tables:
1. Parameters used for the estimation of standard errors of weekly and hourly
earnings data .............................................................................................................
2. Standard errors of estimates of quarterly levels, by selected characteristics ............
3. Standard errors of estimates of annual average levels, by selectedcharacteristics . .
4. Standard errors of estimated percentages—quarterly averages .................................
5. Standard errors of estimated percentages—annual averages .....................................
6. Standard errors of median usual weekly earnings of full-time workers and of
families, by selected characteristics, May 1978, second quarter 1979, and
annual average 1980 ....................................................................................................
v
1
1
2
3
3
4
^
9
11
12
13
13
14
T@©tol(gafl ©@@eripti©[n) ©if tlh©
Quarterly Date ©n W®®kDy
Earnings from the Currant
F©pulati©n Surw@y
tablishments conducted by the BLS. For example, while
the Current Employment Statistics survey (also known
as the establishment survey) provides a large body of
monthly data on employment and earnings by industry
and geographic area, it yields no information on the
distribution of earnings or on the characteristics of the
earners in terms of their age, race, or family status. The
only way to obtain demographically oriented earnings
information at relatively little cost is through a house
hold survey such as the CPS.
After conducting special tests, regular collection of
current earnings data through the CPS was begun in
January 1979. A question on the usual weekly earnings
of all wage and salary workers and one on the hourly
earnings of workers paid by the hour were incorpo
rated into the monthly CPS questionnaire. However,
in order to minimize the burden upon the respondents,
these questions are asked of only one-fourth of the wage
and salary workers in the monthly sample—those in the
“outgoing” rotation groups. The resulting data are then
averaged over a 3-month period for publication on a
quarterly basis.
The CPS sample consists of households in eight sub
samples or rotation groups. Each group is interviewed
for 4 consecutive months, dropped from the sample for
the following 8 months, and then returned to the sam
ple for a final 4 months. The sample is so designed that
2 of the 8 rotation groups are dropped each month—one
group for 8 months, the other group, permanently. The
set of questions pertaining to weekly and hourly earn
ings is asked only of those households which are in
these two outgoing rotation groups. The outgoing ro
tation groups were selected for the earnings queries in
order to minimize the possibly negative impact of these
questions on the future cooperation of respondents.
Since earnings questions are sensitive in nature, it was
feared that, if they were asked monthly in all house
holds, they might jeopardize respondent cooperation in
answering the basic labor force and employment ques
tions in the surveys. The primary purpose of the CPS
is to gather data on the labor force and employment
status of the population, and other survey questions
which might interfere with this purpose are kept to a
The Current Population Survey (CPS)
The CPS uses a scientifically selected national sam
ple of households, with coverage in all 50 States and
the District of Columbia. The survey is conducted each
month by the Bureau of the Census for the Bureau of
Labor Statistics, and provides information on the labor
force, the employed, and the unemployed, including
such characteristics as age, sex, race and Hispanic ori
gin, family relationship, marital status, occupation, and
industry attachment. The information is collected by
trained interviewers from a sample of about 60,000
households.
In addition to the basic questions on employment and
unemployment, the CPS questionnaire for some specific
months contains supplemental questions on other sub
jects. For example, during March, information on in
come and work experience during the previous calen
dar year is collected. Each May, supplemental questions
are asked about multiple jobholding, union representa
tion, receipt of premium pay for work in excess of 40
hours a week, and work schedules. Over the 1967-78
period (except 1968), a question was also asked each
May regarding usual weekly earnings. Beginning in May
1973, questions pertaining to hourly earnings also were
added. The data obtained each May through these ques
tions have been published by BLS in news releases,
Monthly Labor Review articles, and Special Labor Force
Reports. In addition, a large amount of unpublished data
has been transmitted on request to users both inside and
outside the Federal Government.
Monthly <s@ii®cti®n of earnings dlatta m the CPS
Over the 1967-78 period, the demand for demographically oriented earnings data grew considerably. Such
data were needed, among other things, to keep track
of the differences in earnings among the various popu
lation groups—men and women, whites and blacks,
young and old, etc. In addition, there was a great de
mand for data on the earnings of families, particularly
those with two or more workers and those in which
some of the members were unemployed.
Earnings data of this type, particularly those for fami
lies, cannot be derived from the various surveys of es
1
the higher paying white-collar occupations, mentioned
above, the maximum ceiling is $99.99—the amount
coded by the interviewer onto the questionnaire for any
worker reporting hourly earnings of $100 or more.
Entries outside these ranges and those with missing
digits are treated as nonresponses. For example, the en
try for a private household worker reporting an hourly
wage of $30 or more or a craft worker reporting $80
or more per hour would be treated as a nonresponse.
Editing. The next processing step is a detailed editing
procedure either to compute missing data or to assign
a record to be allocated for nonresponse. For example:
(1) If there is no response to the usual weekly earnings
question (25D), but there are entries for usual hours
worked (25A) and hourly earnings (25C), the product
of the two is given a range check and a valid result
would be entered as the usual weekly earnings. (2) If
there is no entry to the question on whether the worker
is paid an hourly rate (25B), but an hourly rate is en
tered in 25C, a yes is inserted into 25B. (3) If the worker
is reported to be paid an hourly rate, but there is no
value in 25C, the reported usual weekly earnings value
is divided by the number of hours usually worked. The
quotient is given a range check, and a valid result is
entered in 25C.
Although original plans called for estimating usual
weekly earnings for those workers paid hourly rates by
multiplying usual weekly hours by the hourly rate,
evaluation of special tabulations indicated that such a
procedure would result in a slight understatement of
average weekly earnings. This understatement results
primarily from the omission of tips, commissions, piece
work, or overtime pay at premium rates from such a
calculated weekly amount. Hence, it was decided to
use the reported usual weekly earnings figure for BLS
tabulations on weekly earnings.
Allocation. Blank items which cannot be filled in dur
ing the editing stage and out-of-range items are allo
cated. Allocation is performed by matching the record
of a person with no entry for an item with the record
of a person of similar demographic and other charac
teristics who has an entry for the item in question. The
value on the donor record is then inserted onto the
record requiring allocation. This procedure has long
been used in processing other data series from the CPS
as well as in the processing of the decennial censuses.
The computer program for the processing of the weekly
and hourly earnings data contains four levels of alloca
tion, with the highest level—a level-1 match—having
the largest number of characteristics for matching the
two records, and the lowest level—a level-4
match—having the least number of characteristics.
There are 12 characteristic items used in determining a
level-1 match (occupation alone has 44 categories),
while a level-4 match has only 4 characteristic items,
minimum. Also for this reason, the new questions on
earnings are positioned on the questionnaire following
the questions on labor force participation and employ
ment status. The segments of the questionnaires used
for the former once-a-year collection and for the cur
rent monthly collection of earnings data are shown
below.
May supplement
(Old series May 1973-78)
These questions were asked of wage and salary work
ers, including self-employed persons whose business was
incorporated. Questions refer to a worker’s sole or pri
mary job.
50. How many hours per week does . . . USUALLY
work at this job?
51. How much does . . . USUALLY earn per week at
this job before deductions?
52. Is . . . paid by the hour on this job?
(If yes, ask 53.)
53. How much does . . . earn per hour?
Monthly basis questionnaire
(New series beginning January 1979)
These questions are asked only of wage and salary
workers. All self-employed persons are excluded. Ques
tions refer to a worker’s sole or primary job.
25A. How many hours per week does . . . USUALLY
work at this job?
25B. Is . . . paid by the hour on this job?
(If yes, ask 25C; if no, skip to 25D.)
25C. How much does . . . earn per hour?
25D. How much does . . . USUALLY earn per week at
this job BEFORE deductions? Include any overtime
pay, commissions, or tips usually received.
Processing ©f the data
Range of earnings checked. The results from each in
terview are checked to determine if the reported earn
ings are within a reasonable range and if entries con
tain all digits. The ranges for usual weekly earnings are
based both on the occupation and the hours usually
worked. For full-time workers, the floor is $20 a week
for private household workers and $30 for others. Ceil
ings are set at $999, the highest value which can be
coded onto the questionnaire. (Higher values can be
entered in the questionnaire, but the maximum ma
chine-readable entry is $999.) For part-time workers,
the range is $20 to $999 for workers in professional,
technical, managerial, and administrative occupations,
and $1 to $749 for other workers.
In-range entries for hourly workers have a floor of
50 cents and ceilings which range from $29.99 to $99.99
based on the occupation of the earners. For workers in
2
1979
1980
1981
question on usual weekly earnings. Since the question
on whether or not one is paid by the hour appears be
fore any questions requesting amounts earned, there
may be some effect on the number of workers report
ing being paid hourly rates. In addition, nonresponses
to this question are now being allocated. As a result,
the new data show a substantially larger number of
workers paid hourly rates relative to the May series.
All averages and aggregations are affected to some
extent by the allocation process. (The old May series
contained no allocated data.) Special tabulations indi
cate that allocation for nonresponses tends to raise the
earnings averages. The increase in median usual weekly
earnings of full-time workers resulting from allocation
averged $2.66 in 1979 and $3.27 in 1980. The increase
in median hourly earnings of workers paid hourly rates
was about 5 cents in both 1979 and 1980.
In terms of historical continuity, it appears that the
averages for the second quarter, of which May is the
midmonth, are at least roughly comparable with the
data previously obtained in May. However, in addition
to the possible discontinuities brought about by the
changes discussed above, another factor must be con
sidered. Relative to the earnings averages for May, the
averages for the second quarter might be influenced by
the influx of youth into low-paying jobs in June. This
may pull the value for the second quarter average be
low that for May alone. Special tabulations of weekly
earnings data for 1979 show that the second quarter
median was about $4 less than the May median for full
time workers. In 1980 and 1981, differences between
the second quarter average and May were not statisti
cally significant. Hourly earnings data also may be af
fected, but apparent differences of 4 cents found in each
of the 3 years were not statistically significant. The net
effect of all these factors is not known.
13.3
16.3
13.6
14.0
14.3
12.0
13.5
14.1
13.5
13.3
12.4
13.6
Comparability to establishment data
most of which have less detail (occupation has only 4
categories).
If the characteristics of a worker whose earnings en
tries require allocation do not match at level 1 with
those of any other worker for whom valid earnings in
formation has been reported, an attempt is made for a
match at the next level. Upon a failure to match at level
2, successive levels are tried until a match is found.
During the first 30 months of processing these data,
only 18 records requiring allocation could not be
matched with a donor record at any level. For these
cases, values approximating earnings averages for re
cent months were inserted for the missing items.
During 1979, 1980, and the first half of 1981, the pro
portion of weekly earnings items which had to be al
located for full-time workers ranged from a low of 13.5
percent in March 1980 to a high of 20.0 percent in May
1979. Over the 30-month period, the rate averaged 16.8
percent. Weekly earnings allocation rates for each quar
ter and annual averages are shown below:
1............................. .........
I I ............................ .........
I l l ........................... .........
IV ....................................
Annual............................
1979
1980
1981
15.9
19.2
16.9
17.0
17.2
15.2
16.8
18.0
16.8
16.7
15.8
16.6
-
These rates are lower than those from the May series on
weekly earnings. The proportion of workers for whom
earnings data were not available averaged 20.0 percent
between May 1973 and May 1978.
The proportion of hourly earnings items which had to
be allocated for workers paid hourly rates ranged from
a low of 10.8 percent in March 1980 to a high of 17.2
percent in May 1979. Over the 30-month period, the rate
averaged 13.6 percent. Hourly earnings allocation rates
for individual quarters and annual averages are:
I
......................
I I ....................................
III ...................................
IV ....................................
Annual............................
Data on weekly and hourly earnings have long been
collected from business establishments in the Current
The not available rate for hourly earnings averaged 15.0 Employment Statistics (CES) survey, and results have
been published by the BLS for many years. These re
percent over the May 1973-78 period.
sults, which are published for major industry groups
Comparability of {the rs@w and ©Id series
and for detailed individual industries, differ from those
As already noted, the new quarterly earnings series available from the CPS because of differences in defi
exclude all self-employed workers, whereas the May nition, coverage, and computation. Most of the house
series did include those self-employed workers whose hold data tabulated on weekly earnings pertain to all
businesses were incorporated (from a legal standpoint, wage and salary workers who usually work full time,
they were the paid employees of a corporation). These and those on hourly earnings are limited to employees
workers have been classified as wage and salary work (full and part time) paid hourly rates. These earnings
ers in the CPS since 1967. The change is expected to are generally expressed as medians. Furthermore, the
produce slightly different earnings medians than would weekly earnings data from the CPS refer to earnings
have been produced if the incorporated self-employed usually received. The establishment data pertain to all
were included.
production workers in mining and manufacturing, con
Another change from the May series is that the new struction workers in the construction industry, and nonquestions on hourly earnings are asked prior to the supervisory workers in other industries, but only within
3
ment Statistics survey includes 39 percent of employ
ees (the CPS sample includes less than one in a thou
sand households), the CES data have smaller sampling
error; thus the degree of industry detail, much at the
4-digit SIC level, is much greater than that available
from the CPS. Nonsampling error, too, should be
smaller in the CES because employers maintain better
records and generally can provide more accurate infor
mation than households.
the private, nonagricultural economy. Both full- and
part-time workers are included in the establishment data,
but there are numerous exclusions, as indicated above,
based on occupation. No distinction is made between
workers paid hourly rates and others. Averages are ex
pressed as means and refer to actual payroll figures for
a specific reference week including paid overtime. An
other difference is in the handling of workers with two
or more jobs. Multiple jobholders are counted sepa
rately in each job they hold in the establishment data
(provided the jobs are production or nonsupervisory),
whereas they are counted only once—according to their
primary job—in the household data.
The extent of earnings differences between the estab
lishment and household surveys varies by industry. For
some of the major industry groups and some of the
more detailed groups for which comparisons can be
made, the differences are small. They are as low as 1
or 2 percent in mining, manufacturing, and in transpor
tation and public utilities, industry groups in which defi
nitions and coverage of the two surveys are similar.
Earnings differences are greatest in the trade, services,
and construction industry groups. For example, accord
ing to the establishment data, mean weekly earnings in
trade, in 1980, were $176 per week, while the CPS data
showed an earnings median of $222—a difference of 21
percent. The CPS universe in this case is limited to
full-time workers, thus excluding many low-paid parttime workers in trade. Also, the CPS data refer only
to those employees whose primary job is in trade, while
the establishment data include any multiple jobholders
whose second job is in trade. Furthermore, the CPS
data include more of the highly paid professional and
technical workers and managerial and administrative
staff. The points made here also apply to the services
group, in which establishment data show mean earnings
17 percent below the CPS median.
For the construction industry, on the other hand, the
mean weekly earnings from the establishment payroll
data are 14 percent higher than the CPS median. This
may be explained, in part, by the inclusion of only ac
tual construction workers in the payroll data, whereas
the CPS data relate also to office workers employed in
the industry. The CPS is also more likely to pick up
more of the lower paid workers in construction (i.e.,
those working for private individuals and small firms)
than is the case for the establishment survey.
Because the establishment data are collected from
each reporting unit in the aggregate form (total em
ployees, total production workers, total hours paid for
production workers, and total payroll for production
workers), information on the characteristics of workers
and their families, earning distributions, and median
earnings are not available. Nevertheless, the establish
ment survey has its own distinct advantages relative to
the CPS. Because the coverage of the Current Employ
Reliability o f the data
In any sample survey, variations in the data can oc
cur by chance because a sample, rather than the whole
population, is surveyed. The standard error is a measure
of variations that occur by chance because the sample
may not be completely representative of the universe.
The chances are about 68 out of 100 that an estimate
from the sample would differ from a complete census
by less than the standard error. The chances are about
95 out of 100 that it would be less than twice the stand
ard error, and about 99 out of 100 that it would be less
than 2-1/2 times the standard error. All statements of
comparison appearing in the text of BLS publications
with CPS earnings data are generally significant at the
90-percent level (1.6 standard errors or more).
If other factors are held constant, the relative size of
the standard error is inversely related to the number of
persons sampled. Since only one-quarter of the wage
and salary workers in the sample are asked the earnings
questions, the standard error is such that publication of
the earnings data on a monthly basis was not deemed
advisable. Similarly, quarterly data—and even annual
average data—for very small population groups may
be subject to relatively large standard errors. Thus, the
user is cautioned against drawing conclusions from rela
tively small differences among numbers for small popu
lation groups without first comparing such differences
with the standard errors associated with them.
Because of the large standard errors associated with
small numbers, percentages and medians derived from
a base of less than 100,000 persons for the quarterly
data and less than 50,000 for the annual average data
are not shown. There is little chance that summary
measures based on smaller numbers would reveal any
useful information. Small values in tables showing per
cent distributions are also subject to relatively large
sampling errors and should be used with caution. Spe
cifically, most values of less than 1.0 percent have a
relative error of 25 percent or more. Such cell values
are shown to permit the user to make combinations
having adequate sample size.
Calculating standard errors. The Bureau of the Census
has developed parameters for estimating standard er
rors. In order to derive standard errors that would be
applicable to a large number of estimates and could be
prepared at a moderate cost, a number of approxima4
tions were required. Therefore, instead of providing an
individual standard error parameter for each estimate,
generalized parameters are provided for various types
of characteristics. As a result, the parameters and the
standard errors shown in the tables give only an ap
proximate indication of the order of magnitude of the
standard error of an estimate rather than its precise
value.
Standard errors of estimated numbers are obtained from
the following formula:
Tables 4 and 5 provide estimates of standard errors
of selected percentages cross-tabulated by the size of
the base of percentages. For example, if 10.0 percent
of a group of workers numbering 10,000,000 were re
ported with quarterly average earnings of $500 or more
per week, the standard error of the 10.0 percent value
is 0.52 percentage point. The 95-percent confidence in
terval (2 standard errors) around the 10.0-percent value
ranges from 9.0 to 11.0 percent. For percentages or
bases (of percentages) not shown in the tables, linear
interpolation of shown values should be used. The es
timates in tables 4 and 5 are only for data covering all
workers and full-time workers of both sexes or men
only, for husband-wife families, and for families main
tained by men (no spouse present). Estimates of stand
ard errors for other groups of individuals and for fami
lies maintained by women (no spouse present) may be
found by multiplying shown values by the factors listed
below each table. The values in tables 4 and 5 were
generated from the formula shown above by using re
spective B values of 2,970 and 861.
The standard error of an estimated median depends
upon the shape of the distribution (particularly in the
vicinity of the median value) as well as the size of its
base. A method to estimate the approximate standard
error of most medians is as follows: (1) Determine, us
ing either the standard error formula or the standard
error table and multiplication factors, the standard er
ror of 50 percent. (2) Determine the percent of the to
tal distribution contained in the published dollar inter
val containing the median. (3) Divide the result of step
(1) by that of step (2) and multiply the result by the
dollar width of the interval containing the median.
For example, in the second quarter of 1979, the me
dian usual weekly earnings of the 43,863,000 men who
usually work full time was $295. (1) The standard er
ror of 50 percent on a base of 43,863,000 is:
ax = "^/A x2 + Bx
where x is the size of the estimate and A and B are the
parameters in table 1 associated with particular
demographic characteristics. Values shown in table 1
are to be used only for usual weekly and hourly
earnings data, and should not be used for other
data series from the Current Population Survey. For
example, in the first quarter of 1981, there were an
estimated 9,528,000 employed full-time wage and salary
workers 25 years and over who were reported earning
$500 or more per week. From table 1, the appropriate
parameters are A = -0.0000300 and B = 2970. Using
the formula, the approximate standard error of an esti
mate of 9,528,000 is:
a x - y / - .0000300 (9,528,000)2 + 2970 (9,528,000)= 159,921
or about 160,000. An alternative method of estimating
standard errors of numbers is to use standard error
tables. Tables 2 and 3 provide standard errors for major
demographic groups. The figures were generated using
using A and B parameters provided in table 1. The
figures given in tables 2 and 3 are to be used as shown
in the following example: In 1980, an estimated
8,121,000 full-time workers had weekly earnings in
the $300 to $349 range. Table 3 should be used be
cause it covers annual average data. Linear interpo
lation in the second column shows the standard error
of an estimate of 8,121,000 is about 81,000.
Standard errors of estimated percentages, computed by
using sample data for both numerator and denomina
tor, are derived from the following formula:
0,50x0,50 = 0.0041 or 0.41 percent
V 2970*43,863,000
(2) The interval containing the $295 median (the $250 to
$299 interval, which has an estimated 6,085,000
workers) includes 13.87 percent of all men who usually
work full time.
6,085,000
43,863,000 = 0.1387 or 13.87 percent
B (P) (1—P)
(3) Dividing the 0.41 percent by 13.87 percent, and
multiplying the result by the $50 width of the interval,
yields a standard error of $1.48.
In some cases, the calculated result varies according
to the extent of rounding of numbers in each of the
three steps. The examples of standard errors of medi-
in which y is the base of the percentage (the
denominator), P is the percentage expressed as a
decimal (e.g., 10.5 percent = 0.05), and B is the
parameter (found in table 1) associated with the charac
teristic of the numerator.
5
weekly earnings of black men were $229 in the first quar
ter of 1979 and $239 in the first quarter of 1980. The
standard errors of these estimates were $3.62 and $3.56,
respectively. Using the formula, the standard error of the
year-to-year change is:
ans appearing in table 6 were computed using more sig
nificant digits than were used in the calculations shown
above.
For some items, the median may lie very close to an
interval boundary. In this case, the standard error (and
confidence limits) will differ depending on which side
of the median a significance test is needed, if (1) the
value of the median plus or minus the standard error
yields a total which lies within the adjoining interval,
and (2) the number of workers per dollar width of the
adjoining interval is different from that within the in
terval containing the median. Otherwise the estimated
standard error is the same on both sides of the median.
The standard error for a difference between two sample
estimates is approximately equal to:
CT(x_y) = ^/(3.62)2 + (3.56)2 -2(3) (3.62) (3.56) = $4.25
The standard error of a ratio, or a percent change of
numbers or medians, can be calculated from the fol
lowing formula:
where x and y are the respective numbers of medians
and p is the correlation coefficient (as indicated above)
between x and y. For example, between the first
quarters of 1979 and 1980, median weekly earnings of
blacks 35 to 44 years old rose 14.1 percent—from $206
to $235. Estimated standard errors of these medians
were $5.06 and $5.66, respectively. Using the formula,
the standard error of the change is:
0(x-yrV°x + °y
where ax and a, are the standard errors of the estimates x
and y; the estimates can be of numbers, percents, or medi
ans. This will represent the standard errors quite accu
rately for the difference between estimates of separate and
uncorrelated characteristics. (If there is high positive
correlation between the two characteristics, the formula
will overestimate the actual standard error.) For example,
the average 1979 median usual weekly earnings of blacks
was $204 and that of Hispanics was $197, a difference of
$7. Standard errors of the respective medians were $1.47
= 0.0347
and $2.09. Using the formula, the standard error of the
difference is about $2.56.
or 3.5 percent.
CT(x_y) = \/(1.47)2 + (2.09)2 =-\/6.529 = $2.56
quarterly data, based on monthly information from onefourth of the sample (or a total of six rotation groups),
are, on the average, almost as reliable as data based on
the May surveys of all eight rotation groups. Lower
nonresponse rates and improved editing of the new se
ries have resulted in estimates of standard errors simi
lar to those for the old May data. Annual average data
are about twice as reliable (i.e., have about one-half the
standard error) as the quarterly figures. Estimates of
standard errors for median earnings of selected popu
lation groups from the May 1978 survey, the second
quarter of 1979, and the annual averages for 1980 are
shown in table 6.
Because the earnings questions are limited to only
one-fourth of the sample, the publication of reliable data
on a monthly basis is not practical. Most month-tomonth changes would be meaningless, given the stand
ard errors associated with the data. However, the BLS
may publish a limited amount of earnings data from a
single month’s sample for cases in which earnings are
cross-tabulated with special information collected only
1 month a year—for example, data on union status,
which are currently collected each May. Such data
a(x -y ) " V Gx + a y ~2p(Jx a y
where p represents the correlation between the two esti
mates, and ctx and oy are estimated standard errors of
levels, percents, or medians. Values of p for comparing
annual averages of consecutive years and identical quar
ters of consecutive years are:
Total, white, or b la ck........................
H ispanic.............................................
Families
0.30
.45
0.35
.55
In other cases, p is zero, resulting in a formula identical to
that in the preceding paragraph. For example, median
/5.60\ /5.66\T
\206/ \2 3 5 /J
Reliability of the new and old series compared. The
The $7 difference in medians is more than 2 times
the value of the standard error of the difference. Hence,
one can conclude with at least 95-percent confidence
that the usual weekly earnings median of blacks was
higher than that of Hispanics in 1979.
The standard error of a change is approximately equal
to:
Persons
/5.66 V
-2 (.3)
\235/
6
Association, Proceedings of the Social Statistics Section,
have standard errors about twice as large as the quar
terly data.
1979.
Defirniitoinis ©if terms used m reports ©o usual
weekly earnings
Nonsampling error. In any survey work, the results
Family. A family is a group of two or more persons
also are subject to errors of response and nonreporting
in addition to sampling variability. Such nonsampling
errors can be attributed to failure to represent all house
holds with the sample or all persons within sample
households (undercoverage), differences in the inter
pretation of questions, inability or unwillingness on the
part of respondents to provide accurate information,
errors made in collection such as in recording or cod
ing the data, errors made in processing the data, and
errors made in estimating values for missing data
(allocation).
The standard error information provided in tables 1
through 6 relate primarily to the magnitude of the sam
pling error; however, they also partially measure the
effect of some nonsampling errors in response and enu
meration. They do not, however, measure any system
atic biases in the data. An example of a systematic bias
would be the reporting of after-tax, rather than gross,
earnings—a systematic downward bias. The full extent
of nonsampling error is unknown, and is not, of course,
measurable as is sampling error.
A special test to gauge the accuracy in reporting of
earnings data was conducted in January 1977. Informa
tion relating to the earnings of nearly 4,000 workers
was obtained either personally from them or from other
members of their households. With their permission,
the same information was subsequently obtained by mail
from the specific employers. Median hourly earnings
for workers paid at an hourly rate was $3.53 on the
basis of the household reports and $3.64 on the basis of
the employer reports, for a difference of 11 cents or 3
percent. There were, however, relatively larger differ
ences between the two sets of data in terms of the pro
portion of workers falling in specific earnings intervals,
particularly at the low end of the earnings distribution.
Median weekly earnings (excluding tips or commissions)
were $170.24 on the basis of the household reports and
$179.50 on the basis of the employer reports, for a dif
ference of $9.26 or 5 percent. Also, as might be ex
pected, the difference between the two sets of data was
greater where the household information was obtained
from proxy respondents (wives, husbands, mothers, etc.)
than where it was obtained from the actual workers.
Probably because the earnings of men are more likely
to be obtained from proxy respondents, the average dif
ference in their earnings, relative to what was reported
by employers, was somewhat greater than that for
women. For more information on this test, see “Com
paring Earnings Data From the CPS and Employer
Records,” by Larry Carstensen and Henry Woltman,
of the Bureau of the Census, in American Statistical
residing together who are related by blood, marriage,
or adoption. The counts of persons by family status in
reports on earnings refer to their position in “primary”
families only; that is, all persons in the household who
are related to the householder. The universe used to
compare the total weekly wage and salary earnings for
families excludes all families in which there is no wage
or salary earner or in which the husband, wife, or other
person maintaining the family is either self-employed
or in the Armed Forces.
Householder. A householder is the person (or one of
the persons) in whose name the housing unit is owned
or rented. For purposes of presenting earnings data, the
term is never applied to either husbands or wives in
married-couple families. As used here, the term “house
holder” relates only to men or women with no spouse
present who are maintaining families.
Employed persons. The universe of employed persons
comprise (a) all those who during the survey reference
week did any work at all as paid employees, worked
in their own business, profession, or farm, or worked
15 hours or more as unpaid workers in an enterprise
operated by a member of the family; and (b) all those
who were not working but who had jobs or businesses
from which they were temporarily absent because of
illness, bad weather, vacation, labor-management dis
pute, or personal reasons, whether or not they were
paid by their employers for the time off, or whether or
not they were seeking other jobs.
Each employed person is counted only once. Those
who held more than one job are counted in the job at
which they worked the greatest number of hours dur
ing the survey reference week.
Wage and salary workers. Wage and salary workers
are employed persons who receive wages, salaries, com
missions, tips, payment in kind, or piece rates. The group
includes employees in both the private and public sec
tors but excludes self-employed persons.
Full-time workers. Full-time workers usually work 35
hours or more per week at their principal job.
Part-time workers. Part-time workers are those who
usually work fewer than 35 hours per week at their
principal job.
Usual weekly earnings. Data on usual weekly earnings
are provided from responses to the question “How much
does ... USUALLY earn per week at this job before
deductions? Include any overtime pay, commissions, or
7
is determined by the response to item 25C, “How much
does ... earn per hour?”
Change in constant dollars. The Consumer Price Index
for Urban Wage Earners and Clerical Workers (CPIW) is used to convert current dollars to constant dollars.
Survey reference week. The survey reference week is
the calendar week, Sunday through Saturday, which
includes the 12th of the month.
Hispanic origin. This term refers to persons who are
of Mexican, Puerto Rican, Cuban, Central or South
American, or other Hispanic origin or descent. Persons
of Hispanic origin may be of any race; hence, they are
included among the numbers for both whites and blacks.
tips usually received.” The response is for the worker’s
sole or principal job. The term “usually” is as perceived
by the respondent. If the respondent asks for a defini
tion of “usually,” interviewers are instructed to define
the term as more than half the weeks worked during
the past 4 or 5 months. The usual weekly earnings of
families is determined by aggregating the usual weekly
earnings of all family members 16 years and older who
were employed as wage and salary workers during the
survey reference week.
Hourly earnings. Data on hourly earnings, as obtained
in the CPS, relate only to employed wage and salary
workers who are reported as being paid by the hour at
their principal job. (Those for whom yes is entered in
item 25B of the questionnaire.) The amount of earnings
8
Table 1.
Parameters used for the estimation of standard errors of weekly and hourly earnings data
For quarterly average data
Characteristic
For annual average data
A
parameter
parameter
A
parameter
-.0000115
2327
-.0000033
675
-.0000232
-.0001905
-.0000300
-.0001931
-.0013655
-.0002661
2970
2970
2970
2970
2970
2970
-.0000067
-.0000553
-.0000087
-.0000560
-.0003961
-.0000772
861
861
861
861
861
861
B
B
parameter
PERSONS
A ll r a c e s , w h ite , b la c k
Both sexes:
Part-time workers ..............................................................................................................................
All workers and full-time workers:
Total or w h ite ..............................................................................................................................
5-year age groups or te e na g e ...............................................................................................
25 years and over ...................................................................................................................
B la c k ............................................................................................................................................
5-year age groups or te e na g e ...............................................................................................
25 years and over ...................................................................................................................
Men:
Part-time workers ..............................................................................................................................
All workers and full-time workers:
Total ............................................................................................................................................
5-year age groups or te e na g e ...............................................................................................
20 years and over ...................................................................................................................
25 years and over ...................................................................................................................
White ..........................................................................................................................................
5-year age groups or te e na g e ...............................................................................................
20 years and over .................................................................................................................
25 years and over ...................................................................................................................
B la c k ............................................................................................................................................
5-year age groups or te e na g e ...............................................................................................
20 years and over ...................................................................................................................
-.0000242
2327
-.0000070
675
-.0000491
-.0003948
-.0000552
-.0000642
-.0000554
-.0004537
-.0000621
-.0000719
-.0004285
-.0030458
-.0005524
2970
2970
2970
2970
2970
2970
2970
2970
2970
2970
2970
-.0000143
-.0001145
-.0000160
-.0000186
-.0000161
-.0001316
-.0000180
-.0000209
-.0001243
-.0008835
-.0001602
861
861
861
861
861
861
861
861
861
861
861
Women:
All workers, full- and part-time workers:
Total or w h ite ..............................................................................................................................
5-year age groups or te e na g e ...............................................................................................
B la c k ............................................................................................................................................
5-year age groups or teenage...............................................................................................
-.0000217
-.0001821
-.0001738
-.0012239
2327
2327
2327
2327
-.0000063
-.0000528
-.0000504
-.0003550
675
675
675
675
.0010500
2401
.0003200
739
(3)
(3)
(3)
5712
4081
3026
(3)
(3)
(3)
1758
1256
931
-.0000345
4488
-.0000010
1302
- 0000232
-.0000491
-.0000217
2970
2970
2327
- 0000067
-.0000143
-.0000063
861
861
675
.0010500
2401
.0003200
739
H is p a n ic o rig in
(levels only)1
Both sexes, men, women:
All workers, full- and part-time workers
.......................................................................................
H is p a n ic o rig in
(percents and medians only)2
All workers, full- and part-time workers:
Both sexes ................................................................................................................................
Men ............................................................................................................................................
Women .............................................................................................................................
G e o g ra p h ic
All characteristics for total, men only, and women only ............................................................
F A M IL IE S
A ll ra c e s , w h ite , b la c k
All families .........................................................................................................................................
Husband-wife families and families maintained by men (no spouse p re se n t)..........................
Families maintained by women (no spouse present) ..................................................................
H is p a n ic o rig in
(levels only)1
All families, husband-wife families, and families maintained by either men or women
(no spouse present) ......................................................................................................................
See footnotes at end of table.
9
Table 1.
Continued— Parameters used for the estimation of standard errors of weekly and hourly earnings data
For quarterly average data
Characteristic
For annual average data
A
parameter
B
parameter
A
parameter
B
parameter
(3)
(3)
(3)
4081
4081
3026
(3)
(3)
(3)
1256
1256
931
F A M IL IE S — Continued
H ispanic origin
(percents and medians only)2
All families ........................................................................................................................................
Husband-wife families, and families maintained by men (no spouse present) ........................
Families maintained by women (no spouse present) ..................................................................
1 Use these parameters for calculating standard errors of estimated
levels (numbers of persons with a given characteristic) only.
2 Use these parameters for calculating standard errors of estimated
percentages and median earnings only.
3The A parameter is not used to calculate standard errors of percents
and medians.
NOTE: Parameters in this table were used to generate the numbers in
table 2-5. These parameters are valid only for weekly and hourly earn
ings data tabulated from the 2-rotation group sample and should not be
used for other items from the Current Population Survey. In the case of a
cross-tabulation involving items with different parameters, the larger
parameter should be used.
SOURCE: U.S. Department of Commerce, Bureau of the Census,
memoranda prepared in the Statistical Methods Division (SMD), Mar. 11,
1980, and Nov. 12,1980. Additional B parameters for the standard errors
of family earnings tabulated by labor force characteristics of family
members are available in an SMD memorandum of Jan. 29, 1980.
Table 2. Standard errors of estimates of quarterly levels, by selected characteristics
(Numbers in thousands)
Both sexes
Estimated level
Part-time
workers
Men
Total or full-time workers
All races
or white
Black
Part-time
workers
Women
Total or full-time workers
All races
White
Black
Total, full- or part-time
workers
All races
Black
or white
10..................................
5 0 ..................................
75 ..................................
100..................................
150..................................
5
11
13
15
19
5
12
15
17
21
5
12
15
17
21
5
11
13
15
19
5
12
15
17
21
5
12
15
17
21
5
12
15
17
21
5
11
13
15
19
5
11
13
15
19
200..................................
250..................................
300...................................
500..................................
750..................................
22
24
26
34
42
24
27
30
38
47
24
27
30
38
46
22
24
26
34
42
24
27
30
38
47
24
27
30
38
47
24
27
29
37
45
22
24
26
34
42
21
24
26
33
41
1,000..................................
1,500..................................
2,000..................................
2,500..................................
3,000...................................
48
59
68
76
83
54
66
76
85
93
53
63
72
79
85
48
59
68
75
82
54
66
76
84
92
54
66
76
84
92
50
59
65
69
71
48
59
68
75
82
46
56
63
69
74
5,000...................................
7,500...................................
10,000...................................
15,000..................................
20,000...................................
107
130
149
180
205
119
145
165
198
224
100
107
102
—
—
105
127
144
187
192
117
140
157
183
199
116
138
155
179
193
64
—
—
—
—
105
127
145
173
195
85
—
—
—
—
25,000..................................
30,000..................................
40,000..................................
50,000..................................
75,000..................................
100,000....... ...........................
226
244
273
296
331
343
244
261
286
301
304
255
207
219
236
—
—
—
209
212
201
—
—
—
199
198
174
—
—
-
_
—
—
—
—
—
211
224
—
_
—
—
—
—
NOTE: Dashes indicate the standard error is not applicable because
the number of workers having the indicated characteristic is less than
that listed in the estimated level column,
10
242
—
—
-
—
—
—
—
—
Table 3. Standard errors of estimates of annual average levels, by selected characteristics
(Numbers in thousands)
Estimated level
Part-time
workers
Total or full-time workers
All races
or white
Women
Men
Both sexes
Part-time
workers
Black
Total or full-time workers
All races
White
Black
Total, full- or part-time
workers
All races
Black
or white
10..................................
5 0 ..................................
7 5 ..................................
100...................................
150..................................
3
6
7
8
10
3
7
8
9
11
3
7
8
9
11
3
6
7
8
10
3
7
8
9
11
3
7
8
9
11
3
7
8
9
11
3
6
7
8
10
3
6
7
8
10
200...................................
250...................................
300...................................
500...................................
750...................................
12
13
14
18
22
13
15
16
21
25
13
15
16
20
25
12
13
14
18
22
13
15
16
21
25
13
15
16
21
25
13
14
16
20
24
12
13
14
18
22
12
13
14
18
22
1,000...................................
1,500...................................
2,000..................................
2,500..................................
3,000..................................
26
32
37
41
45
29
36
41
46
50
28
34
39
42
46
26
32
36
41
44
29
35
41
45
50
29
35
41
45
49
27
32
35
37
38
26
32
36
41
44
25
30
34
37
40
5,000...................................
7,500...................................
10,000...................................
15,000...................................
20,000...................................
57
70
80
97
110
64
78
89
107
121
54
58
55
—
—
57
68
78
92
103
63
75
85
98
107
62
75
84
96
104
35
57
69
78
93
105
46
—
—
—
—
25,000..................................
30,000...................................
40,000...................................
50,000..................................
75,000...................................
100,000...................................
122
131
147
160
179
186
132
141
154
162
164
138
112
118
126
—
112
114
108
—
—
—
107
106
93
—
—
—
114
121
130
—
—
—
_
—
—
—
—
—
_
—
—
—
—
—
—
—
NOTE: Dashes indicate the standard error is not applicable because
the number of workers having the indicated characteristic is less than
11
that listed in the estimated level column,
—
—
—
—
_
—
—
—
—
—
Table 4.
Standard errors of estimated percentages—quarterly averages
Estimated percentages
(thousands)
0.5 or
99.5
1 or
99
2 or
98
3 or
97
5 or
95
10 or
90
15 or
85
20 or
80
25 or
75
30 or
70
50
5 0 ...................................
75...................................
100...................................
150...................................
200...................................
1.72
1.40
1.22
.99
.86
2.42
1.98
1.71
1.40
1.21
3.41
2.79
2.41
1.97
1.71
4.16
3.39
2.94
2.40
2.08
5.31
4.34
3.76
3.07
2.66
7.31
5.97
5.17
4.22
3.66
8.70
7.11
6.15
5.02
4.35
9.75
7.96
6.89
5.63
4.87
10.55
8.62
7.46
6.09
5.28
11.17
9.12
7.90
6.45
5.58
12.19
9.95
8.62
7.04
6.09
250...................................
300...................................
500...................................
750...................................
1,000...................................
.77
.70
.54
.44
.38
1.08
.99
.77
.63
.54
1.53
1.39
1.08
.88
.76
1.86
1.70
1.31
1.07
.93
2.38
2.17
1.68
1.37
1.19
3.27
2.98
2.31
1.89
1.63
3.89
3.55
2.75
2.25
1.95
4.36
3.98
3.08
2.52
2.18
4.72
4.31
3.34
2.72
2.36
4.99
4.56
3.53
2.88
2.50
5.45
4.97
3.85
3.15
2.72
1,500...................................
2,000...................................
2,500...................................
3,000..................................
5,000..................................
.31
.27
.24
.22
.17
.44
.38
.34
.31
.24
.62
.54
.48
.44
.34
.76
.66
.59
.54
.42
.97
.84
.75
.69
.53
1.33
1.16
1.03
.94
.73
1.59
1.38
1.23
1.12
.87
1.78
1.54
1.38
1.26
.97
1.93
1.67
1.49
1.36
1.06
2.04
1.77
1.58
1.44
1.12
2.22
1.93
1.72
1.57
1.22
7,500...................................
10,000...................................
15,000...................................
20,000..................................
25,000..................................
.14
.12
.10
.09
.08
.20
.17
.14
.12
.11
.28
.24
.20
.17
.15
.34
.29
.24
.21
.19
.43
.38
.31
.27
.24
.60
.52
.42
.37
.33
.71
.62
.50
.44
.39
.80
.69
.56
.49
.44
.86
.75
.61
.53
.47
.91
.79
.64
.56
.50
.99
.86
.70
.61
.54
30,000..................................
50,000..................................
75,000..................................
100,000..................................
.07
.05
.04
.04
.10
.08
.06
.05
.14
.11
.09
.08
.17
.13
.11
.09
.22
.17
.14
.12
.30
.23
.19
.16
.36
.28
.22
.19
.40
.31
.25
.22
.43
.33
.27
.24
.46
.35
.29
.25
.50
.39
.31
.27
NOTE: For part-time workers, women, and families maintained by
women, multiply the above values by 0.89. For Hispanics: Multiply by
1.39 for both sexes; by 1.17 for men, all families, husband-wife families,
Table 5.
and families maintained by men; and by 1.01 for women and families
maintained by women.
Standard errors of estimated percentages—annual averages
Base of estimated percentage
(thousands)
5 0 ....................................
75 ....................................
100 ....................................
1 5 0 ....................................
200 ....................................
0.5 or
99.5
1 or
99
2 or
98
3 or
97
5 or
95
10 or
90
15 or
85
20 or
80
25 or
75
0.93
.76
.65
.53
.46
1.31
1.07
.92
.75
.65
1.84
1.50
1.30
1.06
.92
2.24
1.83
1.58
1.29
1.12
2.86
2.34
2.02
1.65
1.43
3.94
3.21
2.78
2.27
1.97
4.69
3.83
3.31
2.71
2.34
5.25
4.29
3.71
3.03
2.62
5.68
4.64
4.02
3.28
2.84
6.01
4.91
4.25
3.47
3.01
6.56
5.36
4.64
3.79
3.28
30 or
70
50
250
300
500
750
1,000
....................................
....................................
....................................
....................................
....................................
.41
.38
.29
.24
.21
.58
.53
.41
.34
.29
.82
.75
.58
.47
.41
1.00
.91
.71
.58
.50
1.28
1.17
.90
.74
.64
1.76
1.61
1.24
1.02
.88
2.10
1.91
1.48
1.21
1.05
2.35
2.14
1.66
1.36
1.17
2.54
2.32
1.80
1.47
1.27
2.69
2.45
1.90
1.55
1.34
2.93
2.68
2.07
1.69
1.47
1,500
2,000
2,500
3,000
5,000
....................................
....................................
....................................
....................................
....................................
.17
.15
.13
.12
.09
.24
.21
.18
.17
.13
.34
.29
.26
.24
.18
.41
.35
.32
.29
.22
.52
.45
.40
.37
.29
.72
.62
.56
.51
.39
.86
.74
.66
.60
.47
.96
.83
.74
.68
.52
1.04
.90
.80
.73
.57
1.10
.95
.85
.78
.60
1.20
1.04
.93
.85
.66
7,500
10,000
15,000
20,000
25,000
....................................
....................................
....................................
....................................
....................................
.08
.07
.05
.05
.04
.11
.09
.08
.07
.06
.15
.13
.11
.09
.08
.18
.16
.13
.11
.10
.23
.20
.17
.14
.13
.32
.28
.23
.20
.18
.38
.33
.27
.23
.21
.43
.37
.30
.26
.23
.46
.40
.33
.28
.25
.49
.43
.35
.30
.27
.54
.46
.38
.33
.29
30,000
50,000
75,000
100,000
....................................
....................................
....................................
....................................
.04
.03
.02
.02
.05
.04
.03
.03
.08
.06
.05
.04
.09
.07
.06
.05
.12
.09
.07
.06
.16
.12
.10
.09
.19
.15
.12
.10
.21
.17
.14
.12
.23
.18
.15
.13
.25
.19
.16
.13
.27
.21
.17
.15
NOTE: For part-time workers, women, and families maintained by
women, m ultiply the above values by 0.89. For Hispanics: M ultiply by
1.43 for both sexes; by 1.21 for men, all families, husband-wife families,
12
and families maintained by men; and by 1.04 for women and families
maintained by women.
Table 6. Standard errors of median usual weekly earnings of full-time workers and of families, by selected characteristics,
May 1978, second quarter 1979, and annual average 1980
Second quarter
1979
May 1978
Annual average
1980
Characteristic
Median
Standard
error
Median
Standard
error
Median
Standard
error
Both sexes:
16 years and over . . ...................................... ..................................
16 to 24 years .............................................................. ..................
25 years and over .......................................................... ..............
White ..................................................................................................
Black ..................................................................................................
Hispanic .................................... ........................................................
$227
162
246
232
181
174
$0.97
1.11
1.32
1.03
3.04
4.14
$242
176
267
247
205
193
$0.98
1.36
1.36
1.05
2.95
4.11
$266
1:91
292
273
219
2:14
$0.71
.72
.76
.75
1.48
4.97
Men:
16 years and over ............................................................................
16 to 24 years ................................................................................
25 years and over .........................................................................
White ..................................................................................................
Black ..................................................................................................
Hispanic .............................................................................................
272
185
294
279
213
201
1.37
2.27
1.44
1.44
3.70
5.96
295
199
319
302
237
222
1.48
2.15
1.46
1.49
3.79
5.13
322
214
346
329
247
238
.82
1.22
.87
.86
2.20
2.77
Women:
16 years and over ............................................................................
16 to 24 years ................................................................................
25 years and over .........................................................................
White ..................................................................................................
Black ..................................................................................................
Hispanic .............................................................................................
166
142
175
167
156
141
.74
1.25
1.34
.79
2.21
4.51
183
153
. 193
184
172
153
.87
1.60
1.03
.92
2.96
4.17
204
171
217
206
190
177
.61
.81
.67
.65
1.35
2.33
n.a.
n.a.
n.a.
n.a.
n.a.
n.a.
401
294
490
2.38
2.85
2.92
437
312
539
1.33
1.60
1.90
Families maintained by women (no spouse present) ..........................
n.a.
n.a.
205
3.72
227
1.80
Families maintained by men (no spouse present) ...............................
n.a.
n.a.
332
7.51
365
5.97
PERSONS
FAMILIES
Married-couple families ..........................................................................
One e a rn e r.........................................................................................
Two or more earners ............................................................ ..........
j
n.a. = not available.
☆ U.S. GOVERNMENT PRINTING OFFICE -.1982 0 - 3 6 1 - 2 7 0 (4 9 4 1)
13
Byresiu off Labor Statistics
Regional Offices
VII
Region 5V
1371 Peachtree Street, N.E.
Atlanta, Ga. 30367
Phone: (404) 881-4418
Regions
and ViiS
911 Walnut Street
Kansas City, Mo. 64106
Phone: (816) 374-2481
Region II
Suite 3400
1515 Broadway
New York, N.Y. 10036
Phone: (212) 944-3121
Region V
9th Floor
Federal Office Building
230 S. Dearborn Street
Chicago, III. 60604
Phone: (312) 353-1880
Regions IX and X
450 Golden Gate Avenue
Box 36017
San Francisco, Calif. 94102
Phone: (415) 556-4678
Region SSS
3535 Market Street
P.O. Box 13309
Philadelphia, Pa. 19101
Phone: (215) 596-1154
Region VI
Second Floor
555 Griffin Square Building
Dallas, Tex. 75202
Phone: (214) 767-6971
Region S
1603 JFK Federal Building
Government Center
Boston, Mass. 02203
Phone: (617) 223-6761
@FedFRASER