Description of Survey Sample and Estimating Procedure

The estimates presented in this report are derived from a probability
sample of approximately 7,800 manufacturing establishments which were
selected from the 1992 Census of Manufactures excluding administrative
records. The sample was restricted to 108 industry classifications. Of
these, 91 represented four digit SIC levels, 14 three digit SIC levels, and
3 two digit SIC levels. The two- and three-digit SIC's constituent four-
digit SIC's that were being sampled separately.  These industries, based on
earlier detailed studies, had been shown to be engaged extensively in
business with the Federal Government.  Most of the industries fall in Major
Groups 34, 35, 36, 37, or 38. In addition, 65 plants not initially in the
Census Bureau file but identified by the Department of Army as having
significant amounts of shipments to the Federal Government, were treated as
separate strata and included in the sample with certainty. These plants
were subsequently located in the Census Bueau files and were classified
accordingly. Because of this slight amount of duplication in the frame, a
trivial amount of bias was introduced in the estimates. Any duplicate
sample cases were identified and one of the duplicates deleted.

In addition to the Department of Army adds, additional establishments were
identified for certainty inclusion in the panel. These included in-scope
establishments of particular companies surveyed in the Census Bureau's
Manufacturers' Shipments, Inventories, and Orders (M3) survey and all
establishments with total employment of 500 or more. All other
establishments in the frame were subsequently assigned sampling
probabilities that were proportional to their total value of shipments.  No
establishment, regardless of its shipments, was assigned a probability less
than 0.005.  Establishments chosen for the survey were assigned weights
equal to the reciprocal of their probabilities of selection. Individual
establishment data were inflated by their sampling weights to develop the
industry estimates used in the final estimation procedure.

All establishments included in the survey were instructed to report total
employment, cost of materials, and value of shipments. These data were to
be identical to the data reported for these items on the 1992 Census of
Manufactures filed with the Census Bureau for that plant. In addition, the
establishments were instructed to indicate by means of a check box inquiry
the range of value of shipments made by the establishments to the Federal
Government in 1992. Those establishments indicating shipments of $1 million
or more to the Federal Government were asked to report the value of such
shipments classified by the specific Government agency to which the
products were billed. The remaining plants (those indicating value of
shipments of less than $1 million to the Federal Government) were not
required to complete this portion of the inquiry. In order to estimate for
the latter respondents, a linear approximation to the mean value of the
range (which the respondent checked) was employed. The following mean
values were used to estimate shipments for establishments that shipped less
than $1 million to the Federal Government:

  Up to $100,000, estimated at $42,000
  $100,000 to $249,999, estimated at $150,000
  $250,000 to $499,999, estimated at $350,000
  $500,000 to $999,999, estimated at $700,000

These estimates are included in the "other Government agency not
specifically known" category.

Government employment by agency was derived for each plant as a straight
proration of its total employment by the ratio of Government shipments data
by agency total shipments. A Government employment figure is shown in the
various tables.

Simple weighted estimates X|ai of shipments by agency "a" for industry "i"
were formed.  These estimates were then adjusted as follows to provide
final survey estimates X||ai of shipments for agency "a", industry "i":

Where

X||ai = (Yi / Y|i) X|ai

X||ai = the estimate of Government shipments for industry "i" to agency
        "a".
Y|i =   the simple weighted estimate of total value of shipments for
        industry "i" obtained from the sample survey.
Yi =    the value of shipments for industry "i" obtained from the 1992
        Census of Manufactures.
X|ai =  the simple weighted estimate of Government shipments for industry
        "i" to agency "a" obtained from the sample survey.

A similar estimation was used for employment data.