Davis-Bacon Act: Labor Now Verifies Wage Data, but Verification Process
Needs Improvement (Letter Report, 01/11/99, GAO/HEHS-99-21).

Pursuant to a legislative requirement, GAO reviewed the: (1) Department
of Labor's response to the House Appropriations Committee's directive
that it verify a random sample of employers' wage data submissions and
select a sample of submissions for on-site data verification; and (2)
likely effect of these efforts on the accuracy and timeliness of
Davis-Bacon Act wage determinations.

GAO noted that: (1) in response to a Committee directive and GAO's
recommendation, Labor has implemented a program to verify wage survey
data submitted on standardized wage data forms by construction
contractors and interested third parties, such as contractor
associations and trade unions; (2) to verify these data, Labor has
developed procedures to select samples of these forms for telephone
verification that differ depending on whether the forms are submitted by
contractors or third parties; (3) in addition, Labor has hired a private
accounting firm to conduct on-site verification reviews; (4) as of
September 30, 1998, the accounting firm had issued final reports for 9
of the 85 geographic area surveys scheduled for audit from April 1997 to
June 1998 and had identified errors in wages reported in about 70
percent of the wage data forms reviewed; (5) in both the telephone and
on-site verification processes, all data--regardless of the entity that
submitted them--are verified only with the contractors; (6) even though
Labor has identified and corrected numerous errors in the wage data
submitted, its verification efforts will have limited impact on the
accuracy of the wage determinations and will increase the time required
to issue them; (7) specifically, errors the accounting firm identified
and corrected in all nine area surveys averaged 76 cents per hour; (8)
but, because Labor was only able to correct the limited number of wage
data forms verified, which contain a small portion of the wage rates
submitted, on average, changes to these wage determinations will be less
than 10 cents per hour, according to Labor officials' estimates; (9) the
extent to which correcting the errors found through verification will
improve the accuracy of wage determinations is limited by: (a) the
Committee directive to use a random sample of wage data forms for
verification, given the characteristics of the wage data with respect to
the universe being sampled; and (b) the procedures Labor uses to
implement this directive; (10) for example, in its procedures, Labor
assumes that data from contractors that refuse access to supporting
documentation are correct and includes the wages in calculating wage
determinations; and (11) while the time needed for verification reduced
timeliness of wage determinations, telephone verification added less
time to the process than did on-site verification--an estimated average
of 2 weeks as compared with an average of 211 days for the 30 area
surveys for which the auditor completed preliminary reports.

--------------------------- Indexing Terms -----------------------------

 REPORTNUM:  HEHS-99-21
     TITLE:  Davis-Bacon Act: Labor Now Verifies Wage Data, but 
             Verification Process Needs Improvement
      DATE:  01/11/99
   SUBJECT:  Income statistics
             Data integrity
             Minimum wage rates
             Labor law
             Payroll records
             Congressional oversight
             Construction contracts
             Wage surveys

             
******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO report.  Delineations within the text indicating chapter **
** titles, headings, and bullets are preserved.  Major          **
** divisions and subdivisions of the text, such as Chapters,    **
** Sections, and Appendixes, are identified by double and       **
** single lines.  The numbers on the right end of these lines   **
** indicate the position of each of the subsections in the      **
** document outline.  These numbers do NOT correspond with the  **
** page numbers of the printed product.                         **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
** A printed copy of this report may be obtained from the GAO   **
** Document Distribution Center.  For further details, please   **
** send an e-mail message to:                                   **
**                                                              **
**                                            **
**                                                              **
** with the message 'info' in the body.                         **
******************************************************************


Cover
================================================================ COVER


Report to Congressional Committees

January 1999

DAVIS-BACON ACT - LABOR NOW
VERIFIES WAGE DATA, BUT
VERIFICATION PROCESS NEEDS
IMPROVEMENT

GAO/HEHS-99-21

Davis-Bacon Act

(205376)


Abbreviations
=============================================================== ABBREV

  CBA - collective bargaining agreement
  OIG - Office of Inspector General
  RSPR - Regional Survey Planning Report
  WHD - Wage and Hour Division

Letter
=============================================================== LETTER


B-280302

January 11, 1999

Congressional Committees

The Davis-Bacon Act was enacted in 1931 to protect communities and
workers from the economic disruption caused by contractors hiring
workers from outside the local geographic area, where wages are
lower, thus obtaining federal construction contracts by underbidding
contractors who pay local wage rates.  The act requires that
employers pay locally prevailing wage rates, including fringe
benefits, to laborers and mechanics employed on the more than $40
billion of federal construction projects a year.\1 Since 1931, the
Congress has enacted many statutes that extend Davis-Bacon's
prevailing wage provisions to construction projects for which the
federal government provides only partial funding, such as the
construction of local schools.  The Department of Labor, which
administers the act, surveys construction contractors and other
interested third parties to obtain information on wages paid to
workers in each construction craft and uses the data submitted on
these survey forms to determine local prevailing wage rates.\2

In previous reports and testimony, we expressed concern that Labor's
procedures for determining prevailing wage rates were vulnerable to
the use of inaccurate or fraudulent data.\3 The use of such data
could lead to the payment of wages that are either lower than what
workers should receive or higher than the actual prevailing wages,
which would inflate federal construction costs at the taxpayers'
expense.  We also noted that basing wage determinations on old wage
data raises questions about whether the wage results still reflect
current conditions.\4 In 1996, we recommended that Labor verify wage
data it receives by obtaining appropriate documentation from a sample
of participating employers or conducting a limited number of on-site
reviews of employer wage data.\5

The House Appropriations Committee has also expressed concerns about
the accuracy and timeliness of Davis-Bacon wage determinations.  The
Committee's reports on appropriations for the Departments of Labor,
Health and Human Services, and Education and related agencies for
fiscal years 1998 and 1999 directed Labor to ensure that a portion of
the funds appropriated for the wage survey program be used to
randomly sample all data submissions to verify their accuracy.\6
Moreover, the reports specified that Labor select a sample of all
wage data submissions for on-site data verification against actual
payroll records.  In response to a request in the Committee reports,
we reviewed these verification activities to determine

  -- what Labor has done in response to the Committee's directive
     that it verify a random sample of employers' wage data
     submissions and select a sample of submissions for on-site data
     verification and

  -- the likely effect of these efforts on the accuracy and
     timeliness of Davis-Bacon wage determinations. 

We are sending this report to you because of your committees'
oversight responsibilities for Labor (see list of addressees at the
end of this letter).  We conducted our review between June and
October 1998 in accordance with generally accepted government
auditing standards.  For information on our methodology, see appendix
I. 


--------------------
\1 The $40 billion dollar estimate does not include the cost of
construction projects covered by the prevailing wage laws for which
others such as state and local governments, rather than the federal
government, are responsible for the contract costs. 

\2 "Interested parties" may include the employers or contractors;
contractor associations; construction workers; labor unions; and
federal, state, and local agencies. 

\3 For a list of selected reports and testimony we have issued on
this topic, see Related GAO Products at the end of this report. 

\4 Labor reported that in 1996 the average age of a wage
determination was 7 years.  In addition, when the wage determination
is issued, it can be based on survey data that are already out of
date.  For example, in 1998 Labor was verifying some data for wages
reported to have been paid in 1993 and had not yet issued a wage
determination as of September 30, 1998. 

\5 Davis-Bacon Act:  Process Changes Could Raise Confidence That Wage
Rates Are Based on Accurate Data (GAO/HEHS-96-130, May 31, 1996). 

\6 U.S.  Congress, House of Representatives, Departments of Labor,
Health and Human Services, and Education and Related Agencies
Appropriations Bill, 1998, H.  Rept.  105-205, 105th Cong., 1st
sess., and U.S.  Congress, House of Representatives, Departments of
Labor, Health and Human Services, and Education and Related Agencies
Appropriations Bill, 1999, H.  Rept.  105-635, 105th Cong., 2nd sess. 


   RESULTS IN BRIEF
------------------------------------------------------------ Letter :1

In response to a Committee directive and our recommendation, Labor
has implemented a program to verify wage survey data submitted on
standardized wage data forms by construction contractors and
interested third parties, such as contractor associations and trade
unions.  To verify these data, Labor has developed procedures to
select samples of these forms for telephone verification that differ
depending on whether the forms are submitted by contractors or third
parties.  In addition, Labor has hired a private accounting firm to
conduct on-site verification reviews.  As of September 30, 1998, the
accounting firm had issued final reports for 9 of the 85 geographic
area surveys scheduled for audit from April 1997 to June 1998 and had
identified errors in wages reported in about 70 percent of the wage
data forms reviewed (see fig.  4).  In both the telephone and on-site
verification processes, all data--regardless of the entity that
submitted them--are verified only with the contractors. 

Even though Labor has identified and corrected numerous errors in the
wage data submitted, its verification efforts will have limited
impact on the accuracy of the wage determinations and will increase
the time required to issue them.  Specifically, errors the accounting
firm identified and corrected in all nine area surveys averaged 76
cents per hour.  But, because Labor was only able to correct the
limited number of wage data forms verified, which contain a small
portion of the wage rates submitted, on average, changes to these
wage determinations will be an average of 10 cents per hour,
according to Labor officials' estimates.  The extent to which
correcting the errors found through verification will improve the
accuracy of wage determinations is limited by (1) the Committee
directive to use a random sample of wage data forms for verification,
given the characteristics of the wage data with respect to the
universe being sampled, and (2) the procedures Labor uses to
implement this directive.  For example, in its procedures, Labor
assumes that data from contractors that refuse access to supporting
documentation are correct and includes the wages in calculating wage
determinations.  While the time needed for verification reduced the
timeliness of wage determinations, telephone verification added less
time to the process than did on-site verification--an estimated
average of 2 weeks as compared with an average of 211 days for the 30
area surveys for which the auditor submitted preliminary reports. 
The verification efforts completed to date may, however, have a
significant impact on improving the accuracy of future wage
determinations by deterring the submission of fraudulent and
inaccurate data, educating contractors on how to complete wage data
forms, and providing Labor with information to use in its long-term
reengineering efforts.  We are recommending specific changes to
Labor's verification procedures that should increase their impact on
the accuracy of the wage determinations while reducing the time and
the cost to collect this information. 


   BACKGROUND
------------------------------------------------------------ Letter :2

The Davis-Bacon Act and related legislation require employers on
federally funded construction projects valued in excess of $2,000 or
federally assisted projects to pay their workers no less than the
prevailing wage rate.  According to its regulations, the Department
of Labor determines the wages "prevailing for the corresponding
classes of laborers and mechanics employed on projects of a character
similar to the contract work in the .  .  .  city, town, village, or
other civil subdivision of the State in which the work is to be
performed." The Department of Labor's Wage and Hour Division (WHD) is
responsible for making these wage determinations. 

In addition to making these determinations, Labor is responsible for
ensuring that employers covered by the law pay at least the mandated
wage.  While some of the same staff work on both responsibilities,
Labor estimates that in fiscal year 1997, it used the equivalent of
51 full-time staff and spent about $5.5 million on the process of
determining Davis-Bacon prevailing wages at both its Washington,
D.C., headquarters and five regional offices.\7

To determine the prevailing wage rates, Labor periodically conducts
surveys, called "area surveys," to collect data on wages and fringe
benefits paid to workers in similar job classifications on comparable
construction projects in a specific geographic area.\8 The agency
solicits information from employers and third parties, such as
representatives of unions and contractor associations.  As shown in
figure 1, after receiving and analyzing the data, Labor issues wage
determinations for a series of job classifications such as
electricians, carpenters, and drywallers in specific geographic areas
varying in size from a section of a county to an entire state.\9 For
example, the prevailing wage determination for the Washington, D.C.,
metropolitan area in 1996 included individual wage rates for 143
different construction crafts.  For a more complete description of
the wage determination process, see appendix II. 

   Figure 1:  Flowchart of Wage
   Determination Process

   (See figure in printed
   edition.)

Labor relies on the voluntary participation of contractors and other
interested parties in conducting the wage survey.  While
participation is voluntary, failure to supply truthful answers can
have serious consequences:  it is a crime under federal law (18
U.S.C.  1001) to knowingly submit false data to the government, and
it is a crime under federal law (18 U.S.C.  1341) to use the U.S. 
mail for fraudulent purposes. 

Contractors and interested third parties generally submit data on
wage survey forms, called WD-10s, which are supplied by WHD.\10 As
shown in the sample forms WD-10 in figure 2, regardless of whether
data are submitted by contractors or third parties, a separate WD-10
is submitted for each construction project and for each contractor
that employed workers on that project.  Wage rate determinations are
issued for different job classifications, such as drywallers and
electricians.  In fiscal year 1997, WHD issued 1,860 individual wage
rate determinations, which were based on information obtained from 43
area wage surveys.\11

In accordance with its regulations, Labor determines an area's
prevailing wage rate on the basis of the 50-percent rule, which
states that the prevailing wage will be the wage paid to the majority
of workers employed in a specific job classification.  If the same
rate is not paid to a majority of those workers in the
classification, the prevailing wage will be the average of the wages
paid, weighted by the total number of workers employed in the
classification.  See the prevailing wage formula in figure 2 for an
example of this calculation using the two hypothetical forms WD-10. 



   Figure 2:  Examples of Wage
   Data Forms (WD-10s) Submitted
   for an Area Survey

   (See figure in printed
   edition.)



   (See figure in printed
   edition.)

Note:  The hypothetical example chosen is simplified for ease of
discussion.  It illustrates a case in which the majority of workers
employed in a specific job classification are paid the same wage
rate.  In many other cases, a majority does not exist and Labor
calculates a weighted average. 

We previously reported that Labor's wage determination process
contained internal control weaknesses that contributed to a lack of
confidence in the resulting wage determinations.  These weaknesses
included limitations in Labor's verification of the voluntarily
submitted wage and fringe benefit data.  We recognized, however, that
accurately reported wage data are not sufficient to ensure the
accuracy of the wage determinations.  For example, in a previous
report, we concluded that reporting bias resulting from the voluntary
nature of the wage surveys may also reduce the accuracy of the wage
determinations. 

In 1995, a congressional committee heard specific allegations that
inaccurate and fraudulent wage data were submitted and used to
determine prevailing wage rates in Oklahoma City.  Both GAO and
Labor's Office of Inspector General (OIG) then received congressional
requests to review Labor's wage determination process.  Labor
responded to the allegations by introducing a policy to verify a
sample of wage data forms received from third parties, but it did not
extend this verification process to forms submitted by
contractors.\12

Before that, Labor had contacted both contractors and third parties
to obtain clarification about data that were inconsistent or unclear,
but it had not attempted to verify data that were not obviously
inaccurate. 

In May 1996, we recommended that Labor obtain appropriate
documentation or conduct a limited number of on-site inspection
reviews to verify a sample of wage data submissions.  Our
recommendation was intended to improve data reliability and increase
confidence in the accuracy of wage data in the short term while Labor
continued its longer-term efforts to address larger weaknesses in the
wage determination process.  We expected that verification would also
increase the accuracy of future wage determinations by reducing
errors through educating both contractors and third parties about how
to complete wage data forms and deterring the submission of
fraudulent data. 

In March 1997, Labor's OIG issued a report on its audit of a
judgmental sample of wage data collected for use in calculating
prevailing wage rate determinations issued in calendar year 1995.\13
The audit did not find evidence of fraud or deliberate misreporting
of wage data.  However, OIG determined that inaccurate data submitted
by both employers and third parties were frequently used in
prevailing wage determinations and that access to payroll records was
the most important factor to successfully verifying wage data.  OIG
echoed our suggestion that verification efforts be viewed as
temporary steps until more fundamental reforms in Labor's survey
methodology could be made. 


--------------------
\7 This amount excludes $3.75 million that the Congress appropriated
specifically for Davis-Bacon program reengineering efforts in fiscal
year 1997, as well as some costs incurred by states conducting area
wage surveys. 

\8 Fringe benefits considered in computing the prevailing wage rate
include, but are not limited to, holiday and vacation pay, health
insurance, and pension benefits. 

\9 Labor conducts wage surveys and issues wage determinations for
four basic types of construction--building, residential, heavy, and
highway.  For determining prevailing wage rates for highway
construction, both the survey and its analysis are often done by
individual state agencies.  Labor also uses a contractor to complete
both the survey and preliminary analysis for a limited number of area
surveys.  Labor includes both the state- and contractor-completed
surveys among those sent to the accounting firm for on-site
verification.  The contractor-completed surveys are also subject to
telephone verification, but the state-completed surveys are not. 

\10 In an effort to obtain the greatest participation, Labor staff
sometimes complete the WD-10 form with information provided over the
telephone by employers. 

\11 This number does not include ongoing modifications to wage
determinations when there are automatic or negotiated changes to the
wage rate specified in the collective bargaining agreement or wage
determinations limited to specific projects rather than for entire
geographic areas. 

\12 Third parties generally provide about one-third of all wage
survey forms. 

\13 Department of Labor, OIG, Inaccurate Data Were Frequently Used in
Wage Determinations Made Under the Davis-Bacon Act, Final Report
#04-97-013-04-420 (Mar.  10, 1997). 


   LABOR HAS IMPLEMENTED A PROGRAM
   TO VERIFY WAGE DATA
------------------------------------------------------------ Letter :3

In response to the House Appropriations Committee's directive and our
recommendation, Labor has implemented a program to verify wage survey
data submitted by construction contractors and interested third
parties, such as contractor associations and trade unions.  To verify
these data, Labor established procedures to select samples of wage
data forms for telephone verification that differ depending on the
entity that submitted the form.  In addition, Labor has hired a
private accounting firm to conduct on-site verification reviews.  In
both the telephone and on-site verification process, all
data--regardless of which party submitted them--are verified only
with the contractors. 


      LABOR'S TELEPHONE
      VERIFICATION PROCEDURES
      DIFFER DEPENDING ON WHETHER
      THE DATA ARE SUPPLIED BY
      CONTRACTORS OR THIRD PARTIES
---------------------------------------------------------- Letter :3.1

In response to our recommendation, in June 1996 Labor expanded its
telephone verification process begun the previous year from one of
verifying wage data submitted by third parties to one that also
verifies wage data submitted by contractors.  However, Labor verifies
a different percentage of wage data forms submitted by third parties
than that for data forms submitted by contractors.  In addition,
regardless of who submitted the wage data form, Labor asks for
supporting documentation only from contractors, not from third
parties. 

Labor verifies a larger percentage of wage data forms submitted by
interested third parties than for those submitted by contractors, as
shown in figure 3.  For wage data submitted by third parties, wage
analysts must select every tenth--10 percent--WD-10 wage data form
submitted (and no fewer than two WD-10 forms) for verification with
the contractor.\14 In contrast, for wage data submitted by
contractors, Labor requires that regions select every fiftieth--2
percent--wage data form (but no fewer than five) for telephone
verification.  Regional wage officials told us that they always use
the sample size required by Labor's national office for contractors,
but they exceed it for third parties.  For example, wage analysts in
one region told us that they verify 100 percent of wage data forms
submitted by third parties by requiring that all such forms be signed
by contractors who paid the wages that were reported.  A senior wage
analyst in another region told us that they conduct telephone
verification of almost all third-party wage data forms. 

   Figure 3:  Selecting Wage Data
   Forms for Verification

   (See figure in printed
   edition.)

Labor's procedures require that wage analysts verify data only with
the contractors, not with third parties, even for data submitted by
third parties.  The procedures require the regional offices to send
letters to the contractors selected for verification requesting that
supporting payroll documents be mailed to Labor.  (See app.  III for
a sample of the letter Labor sends to contractors.) Wage analysts
contact the contractors selected for verification by telephone to
verify wage data regardless of whether the contractor has provided
the requested documents.  Wage analysts told us that they generally
do not receive the documents requested from the contractor and,
therefore, rely on the contractor's verbal assurance that the data
are correct.  When the wages reported by the contractor, either with
documentation or with oral confirmation, are different from those
originally submitted, Labor replaces the wage data submitted on the
WD-10.\15 When the information provided by the contractor does not
agree with the data submitted by a third party, regional wage
analysts told us that they always take the word of the contractor
over the information supplied by the third party.  Unlike the process
Labor uses with contractors, Labor seldom notifies third parties that
the wage data forms they submitted have been selected for
verification and does not ask them for documentary evidence to
support the data they provided. 

Even though information from contractors who participate in the
verification process sometimes leads to changes in the wage data,
Labor includes in the prevailing wage calculation data reported by
contractors who refuse to participate in the verification process,
thereby assuming these data to be accurate.  Labor does not keep
records that would allow us to assess how often this occurs.  One of
Labor's regional offices, however, provided data showing that in the
18-month period from April 1, 1997, to September 30, 1998, wage
analysts in that region were unable to verify data by telephone for
41--57 percent--of the 72 WD-10 forms submitted by contractors that
were selected for verification.  Labor's procedures allow it to
assume that these data are correct and to include them in the wage
calculation.  However, this assumption is questionable because, of
the remaining 31 forms that were verified by telephone, analysts
found errors in data submitted in 9 of the forms, or 29 percent. 


--------------------
\14 Labor also requires its wage analysts to verify with the
contractor all third-party data submitted by a union that reports
wages for workers employed by a nonunion firm.  Labor's policy does
not include any similar procedures to verify data submitted by third
parties that are not unions, such as contractor associations and
federal and state agencies.  However, Labor's procedures require wage
analysts to verify any wage data that on the surface are
questionable, regardless of the source of the data. 

\15 Labor does not keep records, however, that would show how often
errors are found in telephone verification or how often changes are
made. 


      LABOR USES A PRIVATE
      ACCOUNTING FIRM TO CONDUCT
      ON-SITE VERIFICATION
---------------------------------------------------------- Letter :3.2

In April 1997, Labor began a process of on-site wage data
verification under a contract awarded to a private accounting firm. 
As of September 30, 1998, Labor had paid the firm a total of $1
million for on-site verification for fiscal years 1997 and 1998, and
had awarded a new contract to the same firm for $500,000 in fiscal
year 1999.\16

The contract requires the accounting firm to review a sample of
payroll records to verify wage survey data and interview employers to
obtain information to assist in Labor's efforts to reengineer the
wage determination process.  Labor selects what it describes as a
10-percent sample of the wage data forms submitted by both
contractors and interested third parties for a specific area wage
survey.  It reaches this percentage by selecting every tenth data
form for verification.  In fact, the percentage of WD-10s selected
for on-site verification usually exceeds 10 percent.  This is
because, after selecting every tenth data form submitted, Labor adds
to the set of forms the auditor will review all other usable WD-10s
that could be examined at that contractor's office, such as the forms
for other projects for which data were reported.\17 As a result, 10
percent is the minimum sample size; the actual sample size varies
from one survey to another.\18 For example, for the 9 surveys for
which final audit reports were completed, the actual percentage of
WD-10s selected for on-site verification ranged from 10 to 56
percent. 

As with the telephone verification conducted by WHD wage analysts,
auditors verify the data on-site only with the contractor who
employed the workers, even when the data were submitted by third
parties.  The WHD regional office mails the contractors selected for
verification a letter notifying them that an auditor will contact
them to request a visit to their establishments.  (See app.  III for
a sample letter sent to contractors.) The contractor is asked to make
payroll records available to the auditor to confirm that the data
reported on the WD-10 are complete and correct.  While the
contractor's cooperation with the auditor is requested, it is
voluntary for contractors whose wage data cover private construction
projects, because Labor is not authorized to require contractors to
provide records for such projects.\19 In contrast, contractors on
federal projects are required by law to grant access to payroll
records related to the federal projects.  Labor, however, does not
specify this in its letters to these contractors.  Labor is concerned
that doing so might discourage contractors from participating in
future Davis-Bacon surveys, which could reduce the number of survey
respondents and thus affect the accuracy of wage determinations. 

The auditor compares the data reported on the WD-10 with the payroll
records for the reported project.  Discrepancies between the original
WD-10 submitted and the payroll records or contractor's testimony are
recorded by the auditor.  After completing the audit of the area wage
survey data, the auditor submits a preliminary report to Labor, which
includes a list of all discrepancies and a list of contractors that
did not participate in the verification process.\20 Labor reviews the
preliminary report and makes follow-up telephone calls with
contractors as necessary.  After Labor reviews the auditor's
findings, the accounting firm submits a final report reflecting
changes on which Labor and the firm have agreed.  Regional wage
analysts told us that only after Labor receives the final report do
the analysts incorporate appropriate changes to wage data,
recalculate wage data, and forward recommended wage rate
determinations to Labor's national office for final review and
issuance. 

During the 15-month period from the beginning of the on-site
verification process in April 1997 to June 30, 1998, Labor sent 85
surveys to the auditors for on-site verification.\21 As of September
30, 1998, the auditors had completed audits for 30 of the 85 area
surveys and had issued final reports for 9 of these. 

The auditors reported finding errors in both the wages and the number
of workers reported in the majority of wage data forms they reviewed. 
Specifically, for the nine surveys for which they had completed final
reports, the auditors found errors in wage rates reported in about 70
percent of all wage data forms reviewed.  Labor has issued new wage
determinations based on four of these nine surveys (see fig.  4). 

   Figure 4:  Status of 85 Area
   Surveys Subject to On-Site
   Verification, as of September
   30, 1998

   (See figure in printed
   edition.)


--------------------
\16 This $1 million is in addition to the $10 million Labor estimated
it spent for fiscal years 1997 and 1998 wage determination activities
other than verification. 

\17 The data are submitted on a separate WD-10 form for each
contractor and project.  Thus, a contractor who employed workers on
more than one project will be identified as the contractor on more
than one WD-10 form.  (See #1 in fig.  2.)

\18 Multiple WD-10 forms may be submitted for the same contractor and
project if data are submitted for the project by both contractors and
third parties. 

\19 Legislative proposals for providing such authority have been
introduced in the Congress in recent years, but none has been enacted
to date. 

\20 The report also describes the contractors' responses to five
specific questions, including the contractors' experience using the
WD-10 to submit wage data, and their interest in providing and
ability to provide wage data electronically.  This information is
collected as part of the second purpose of the audit:  to gather
information to help Labor improve the wage determination process. 

\21 This represented an unusually large number of surveys because it
included 42 surveys that were begun before the beginning of the
on-site verification process in April 1997 and 43 that were initiated
in the following 15 months. 


   VERIFICATION WILL HAVE LIMITED
   IMPACT ON ACCURACY OF
   PREVAILING WAGE DETERMINATIONS
   AND WILL REDUCE TIMELINESS
------------------------------------------------------------ Letter :4

Verification efforts conducted to date will have a limited impact on
the accuracy of prevailing wage rate determinations and will increase
the time required to issue them.  The extent to which the
verification process improves the accuracy of Labor's prevailing wage
determinations will be limited by the congressional directive to use
a random sample of wage data forms to select wage data for
verification and the procedures Labor uses to implement the
directive.  In addition, on-site verification has added time to the
wage determination process, increasing the likelihood that data used
will be outdated.  The on-site verification process has, however,
provided information that Labor is using as it tries to improve the
process for future wage determinations. 


      ERRORS IDENTIFIED, BUT
      IMPACT ON ACCURACY OF WAGE
      DETERMINATIONS IS LIMITED
---------------------------------------------------------- Letter :4.1

Although Labor has identified and corrected numerous errors in the
wage data submitted, it has been able to correct only the limited
number of wage data forms verified.  Since this represents only a
small portion of the total number of data forms submitted, these
corrections have only a minimal impact on the accuracy of the data
used to calculate wage determinations.  As a result, even though we
found that errors the auditors identified in all nine area surveys
averaged 76 cents per hour, Labor officials estimate that the changes
to wage determinations will amount to an average of 10 cents per
hour. 

Furthermore, both the Committee directive to use a random sample of
wage data forms to select wage data for verification and the
procedures Labor uses to implement the directive also limit the
extent to which errors found will improve the accuracy of wage
determinations.  While a random sample is often assumed to be the
most effective approach to selecting a sample, it is not the best
approach for verification in this situation.  It would be impractical
to verify a large enough random sample of wage forms to ensure that
verification would have an impact on the accuracy of the wage
determination.  Moreover, the procedure Labor uses to verify wage
data (1) does not take into account whether the data it selects for
verification are likely to be used in calculating wage
determinations, (2) assumes that data from contractors that refuse
access to supporting documentation are correct, and (3) does not
attempt to verify data with third-party submitters when contractors
are unable to provide or refuse access to supporting documentation. 


         VERIFICATION SAMPLE NOT
         LARGE ENOUGH TO ENSURE
         IMPACT ON ACCURACY OF
         WAGE DETERMINATIONS
-------------------------------------------------------- Letter :4.1.1

Although the House Committee directed Labor to use a random sample to
select wage data forms to verify the accuracy of wage data, Labor
does not select a large enough number of data forms to ensure that
the errors found will improve the accuracy of wage determinations,
nor would it be feasible to do so.\22 Although random sampling is
sometimes the best approach to selecting data, in some circumstances
other sampling strategies are more effective.  A random sample would
allow Labor to assume that data found to be in error were
representative of all data submitted, and Labor could adjust the
prevailing wage rate rather than adjusting the data on only the
WD-10s selected for verification.  However, the sample size needed
for this approach would require Labor to verify most of the wage data
submitted.  Conversely, a carefully chosen judgmental sample would
allow Labor to select wage data forms for which correcting errors
found would have the greatest effect on the accuracy of the wage
determinations. 

To select a representative random sample that would ensure the
accuracy of the data used to determine prevailing wage rates, Labor
would have to sample workers within each job classification rather
than wage data forms, which often combine job classifications, as it
does now.  Labor currently determines the amount of wage data to be
verified by selecting a uniform percentage of WD-10s for each area
survey, ranging from 2 percent to 10 percent.  However, because Labor
determines multiple prevailing wage rates, one for each job
classification, it must select a sample of wage data from every job
classification within a survey to ensure a representative sample for
all prevailing wage rates.  Since wage data forms often include data
for multiple job classifications, sampling wage data forms alone does
not ensure representativeness within specific job classifications. 

Labor would also have to select a sufficient number of workers within
each job class to meet the statistical criteria for appropriately
projecting from sampled cases to all the wage data.  However, data
submitted on the number of workers within a job class can be small,
often fewer than 10.  As a result, Labor would need to select a
sample size equal or close to the total number of workers.  For
example, we calculated that the sample size required for a
statistically representative sample would require that Labor verify
all data submitted for 40 of 45 job classifications in one fiscal
year 1997 area survey in order to be within 50 cents per hour of the
correct wage.  For all job classifications, Labor would have had to
verify the wages of more than 5 times the number of workers it
verified, 439 rather than 78 of the 664 workers for whom wages were
reported.  (See app.  IV.)


--------------------
\22 In addition, the sample Labor draws is systematic but does not
meet the criteria to be considered random.  See app.  IV for more
information. 


         LABOR SELECTS WAGE DATA
         FOR VERIFICATION THAT
         WILL HAVE LITTLE OR NO
         IMPACT ON ACCURACY OF
         WAGE DETERMINATIONS
-------------------------------------------------------- Letter :4.1.2

Using a random sample does not allow Labor to judgmentally select for
verification wage data that will have the greatest potential impact
on accuracy.  For example, Labor verifies wage data that it does not
expect to, and does not, use to calculate prevailing wage
determinations.  In addition, Labor does not consider the cost of
travel and staff time in selecting wage data forms to verify. 

Of the 30 area surveys for which on-site verification preliminary
reports had been completed as of September 30, 1998, 29 included
verification of wage data that would not be used in calculating
prevailing wage rates.  The data verified would not be used for two
reasons.  First, in some instances, wage determinations would be
based on wage rates included in collective bargaining agreements, not
on the wage data reported--whether the data were accurate or not.  In
general, when the same wage is paid to 50 percent or more of the
workers employed in a job classification, the wage rate is the same
because it is specified in a collective bargaining agreement that
covers the specified job classification.\23 When this occurs, Labor
does not determine the prevailing wage rates on the basis of data
reported on the WD-10s.  Instead, it uses the collective bargaining
rate as the prevailing wage rate.  Second, in some instances, Labor
knew it would not use the wage data because it had received
insufficient data within the specific job classification to allow it
to issue a wage determination.  Labor's procedures require that it
receive responses from a minimum of 25 percent of the contractors and
third parties it contacts for data and wages covering a minimum of
three workers from two contractors to determine the prevailing
wage.\24 When Labor receives too few responses, it does not issue a
new wage determination. 

For example, for one of the four area surveys for which Labor has
issued prevailing wage rate determinations, none of the verified
wages were used to determine the prevailing wage rates.  For this
specific survey, the rate specified in the collective bargaining
agreement was used for 34 of the 36 individual job classifications. 
Data for the remaining two job classifications were not used because
Labor did not receive sufficient data for these two job
classifications from the survey responses.  Thus, although the
on-site verification for this one survey cost about $40,000, and it
required 5 weeks to complete the preliminary report, none of the
results were used. 

Labor also does not balance the benefits against the costs of
verifying specific wage data forms when selecting its sample.  For
example, selecting wage data forms with wages reported for the
greatest number of workers within a specific job classification has
the potential for greater benefit in improving the accuracy of the
wage determination than does selecting and verifying wages for a
smaller number of workers.  Using the sample WD-10s in figure 2, if
Labor's random sample resulted in verifying submission 2, which
includes data for only 2 of a total of 82 workers for whom wages were
used in the calculation, correcting even large errors would have
little impact on the prevailing wage determination because only these
2 wage rates could be adjusted.  On the other hand, even a small
error in wage rates reported for the 80 workers would have a direct
effect on the wage determination.  Using the hypothetical WD-10s in
figure 2, if the fringe benefits reported on submission 1 were
incorrectly omitted from submission 2, correcting the wages would
increase the prevailing wage rate from $18.64 to $18.71, an increase
of 7 cents.  However, as figure 5 shows, if the workers did not
receive the fringe benefits on submission 1, which includes data on
80 workers, correcting the error would reduce the prevailing wage
rate from $18.64 to $16.00, a decrease of $2.64. 

   Figure 5:  Calculation
   Comparing Effect of Correcting
   Errors on Simplified
   Hypothetical Wage Data Forms

   (See figure in printed
   edition.)

Note:  For the purpose of illustration, we use only two forms WD-10
with widely differing numbers of workers reported in the job
classification.  In actual practice, the 50-percent rule would apply;
the prevailing wage rate would not be calculated but would be the
rate paid to those on submission 1 because more than 50 percent of
the workers are paid the same wage, to the penny. 

Labor also does not consider the cost of accessing payroll records
when selecting wage data forms for verification.  To do on-site
verification, the accounting firm's auditors must contact individual
contractors and visit their administrative offices to review payroll
records; these offices may be located far from one another.  For
example, as shown in figure 6, for the area wage survey covering
Lawrence and Greene counties in Pennsylvania, contractors' offices
were located in six states and were as far away as Texas and
Massachusetts.  Using Labor's systematic selection process, auditors
would attempt to visit the contractor's administrative office for
submission 1 in our example, regardless of the great distance from
the other sites and the small number of drywallers whose reported
wages were to be verified. 

   Figure 6:  Geographic
   Dispersion of Contractors in
   Pennsylvania Area Survey

   (See figure in printed
   edition.)


--------------------
\23 Labor obtains data on the WD-10 about whether the workers are
covered by a collective bargaining agreement (see #3 in fig.  2). 

\24 When Labor receives data from an entity it has not contacted, it
counts these data in the response rate as though it had requested the
data. 


         LABOR ASSUMES DATA ARE
         ACCURATE WHEN CONTRACTORS
         REFUSE OR ARE UNABLE TO
         PROVIDE SUPPORTING
         DOCUMENTATION
-------------------------------------------------------- Letter :4.1.3

Labor assumes that data from contractors that are unable to provide
or refuse access to supporting documentation are correct by including
them in the wage rate calculations.  A more reasonable assumption
would be that data from contractors who refuse access have a greater
chance of being inaccurate than data from contractors who provide
access.  Labor did not have comprehensive data on the reasons
contractors would not provide access to the payroll records necessary
to verify reported wage data.  Both Labor's wage analysts and the
on-site accounting firm auditors reported that contractors had many
reasons for not participating, such as not wanting to devote the
necessary resources to access the records or that the records were no
longer available.  It is also possible, however, that wage data may
be fraudulent or carelessly reported, because contractors who
knowingly submit fraudulent data may be unlikely to voluntarily
submit to an audit of their payroll records out of fear of
prosecution for committing a federal crime. 

As shown in figure 7, 27 percent of the contractors selected for
verification either refused or were unable to provide on-site
auditors access to some or all of the payroll records required for
verification.  For the 30 on-site audits for which Labor has received
preliminary reports, auditors were unable to visit 20 percent, or 59,
of the 293 contractors selected for on-site verification either
because the contractors would not participate in on-site verification
or the accounting firm was unable to schedule an acceptable time for
auditors to visit.  Another 7 percent of the contractors denied or
were unable to provide access to some or all of the necessary payroll
records after the auditors arrived at the contractors' offices.\25
Labor's OIG found in its verification review that access to payroll
records was the most important factor in successfully verifying wage
data. 

   Figure 7:  Access Provided by
   Contractors to Payroll Records

   (See figure in printed
   edition.)

While Labor does not have legal authority under the Davis-Bacon Act
to access payroll records for workers involved in private
construction, it does have authority to access such records for
federally funded or assisted construction work covered by the act. 
Labor officials told us that they do not exercise this right because
to do so might result in reduced accuracy of future wage
determinations if it discouraged contractors from voluntarily
providing wage data for future surveys. 


--------------------
\25 Labor sometimes asks government contracting agencies to provide
certified payroll records, but it reported limited success in
obtaining these records. 


         VERIFYING DATA ONLY WITH
         CONTRACTORS LIMITS
         ACCURACY OF WAGE
         DETERMINATIONS
-------------------------------------------------------- Letter :4.1.4

Labor's procedure of verifying wage data only with contractors also
limits the accuracy gains achievable from the verification process
and could actually result in decreased accuracy.  For example, our
review of on-site verification reports found wage data, including
fringe benefit data, submitted by third parties that auditors were
unable to verify through the contractor.  The contractor either did
not have records on fringe benefits paid or refused the auditors
access to any payroll records.  In some cases in which the contractor
did not verify the accuracy of the fringe benefits, the auditors
recorded $0 under the fringe benefits as though the reported data
were inaccurate.  These workers, however, were covered by a
collective bargaining agreement.  Especially for fringe benefits paid
under collective bargaining agreements, unions often have
documentation to verify amounts paid.  In fact, regional wage
analysts told us that, in some cases, unions may be the only source
of data on fringe benefits.  By not seeking documentation from the
third party, the verification process may have erroneously reduced
the amount of wages and fringe benefits paid and thus contributed to
a less accurate wage determination. 


      VERIFICATION PROCESS ADDS
      TIME TO ISSUANCE OF WAGE
      DETERMINATIONS
---------------------------------------------------------- Letter :4.2

As would be expected, verification efforts have increased the time
required to issue wage determinations after the area survey has been
completed.  Labor does not collect data on the amount of time
required to complete telephone verification, but Labor officials who
administer the wage determination process estimated that telephone
verification added an average of 2 weeks to the process.  Telephone
verification can be accomplished relatively quickly because Labor can
conduct telephone verification as wage data are being submitted.  In
addition, it does not require travel, which would add time and
expense. 

On-site verification, however, adds much more time--months rather
than weeks--to the process because (1) it requires travel and (2) in
order to identify all wage data forms related to a specific
contractor and more efficiently manage travel, it does not begin
until after the survey cutoff date for wage forms has passed.  In
fact, regional wage analysts told us that they do not send the
surveys to the accounting firm for on-site verification until the
telephone verification and all preliminary analysis have been
completed such that wage determinations are ready to be forwarded to
the national office for review and publication.  Our analysis of the
30 area surveys for which the auditors submitted preliminary reports
shows that the time between when Labor sent the area survey data to
the accounting firm for verification and when Labor received the
firm's preliminary report ranged from 36 to 408 days, with an average
of 211 days.  However, Labor officials told us they cannot begin
final calculations until they receive the final report from the
auditors, which incorporates the results of discussions with Labor. 
Other Labor activities, such as reviewing the results of on-site
verification audits and making any necessary adjustments to wage
determinations before issuance, also add time to the wage
determination process, but Labor has no data to estimate the amount
of time these activities take.  Thus, while the impact verification
is having on timeliness is greater than the time elapsed between when
Labor forwards the surveys to the auditors and receives the
preliminary report, the total time required is not available.  WHD
officials told us that they expect this delay will decrease over
time, attributing some of it to the time required for WHD staff and
accounting firm auditors to master the verification process. 


      VERIFICATION PROCESS
      FACILITATES LONG-TERM
      EFFORTS TO IMPROVE WAGE
      SURVEY PROCESS
---------------------------------------------------------- Letter :4.3

While the effect of the verification process on the accuracy of data
used in wage determinations has been minimal, these efforts may have
a greater impact over the long term by deterring contractors and
third parties from submitting inaccurate or fraudulent data,
educating contractors about wage data form procedures, and assisting
Labor in its reengineering efforts.  Labor officials stated that they
had focused their verification procedures on identifying and
deterring fraud rather than on ensuring the accuracy of the wage
determinations.  But they also told us that the value of verification
as a deterrent to the submission of fraudulent wage data must be
balanced with its potential to deter voluntary participation in
future Davis-Bacon Act surveys, which could, conversely, reduce the
accuracy of the wage determinations. 

Verification efforts may help educate contractors about how to
complete wage survey forms properly.  But Labor's procedure of not
including third-party submitters in the verification process limits
the potential for verification to improve the accuracy of future wage
determinations.  Third parties do not benefit from the potential
educational value that verification has because they are not informed
of any errors identified on the WD-10s they submitted, nor do they
learn how to properly complete them. 

Through its verification efforts, Labor has also obtained information
that it is using in its long-term efforts to reengineer the wage
determination process.  Labor included in the on-site verification
process questions to contractors about the wage survey form and its
terms, such as whether the contractors had difficulty understanding
and completing the survey form.  Of those contractors who reported
confusion or difficulty with the form, many identified the "peak
week" and the number of workers employed during the peak week as
major sources of confusion.\26 In addition, the accounting firm found
in the course of its on-site payroll verification that contractors
and third parties that submitted wage data had difficulty completing
the form, including accurately identifying the job classification of
workers.  The auditors found that these difficulties affected the
accuracy of the wage data reported.  For example, for the nine area
surveys for which the auditors completed final reports, they
identified errors in wage data for 38 percent of all contractors
visited caused by misidentification of the peak week.  Labor is also
redesigning its wage reporting form, which responds to concerns
raised by contractors during on-site verification.  Labor is also
piloting a statewide survey using wage data for "total man-hours" in
place of wage data for the peak week.\27


--------------------
\26 The "peak week" refers to the work week in which the contractor
employed the largest number of workers in a particular craft or
classification for work on a specific project. 

\27 Total man-hours differs from the peak week.  Total man-hours does
not count the number of workers, but the number of total hours worked
within a specific job classification for the entire length of the
project.  In contrast, peak week counts the number of workers within
a job classification for only one week, regardless of the number of
hours worked and the wages paid them. 


   CONCLUSIONS
------------------------------------------------------------ Letter :5

Without accurate and timely data, Labor cannot determine prevailing
wage rates that correctly reflect the labor market.  While obtaining
accurate wage data through Labor's voluntary surveys will not ensure
that wage rate determinations are accurate, inaccurate data guarantee
inaccurate wage determinations.  We recognize that achieving
100-percent accuracy is not possible.  However, inaccurate prevailing
wage determinations could lead to the payment of wages that are
either lower than what workers should receive or higher than the
actual prevailing wages, which would inflate construction costs at
the taxpayers' expense.  A system to verify wage data submitted by
contractors and third parties is necessary to ensure that inaccurate
data do not have a negative effect on the prevailing wage
determination. 

As directed by the House Appropriations Committee, Labor has
implemented a process to verify wage data submitted by both
contractors and third parties.  This process allows it to identify
and correct errors it finds in wage data reported.  It may also have
a positive impact on the accuracy of wage data obtained in future
wage surveys by educating contractors on the proper completion of
wage data forms.  In addition, this process has helped Labor obtain
information that will assist it in reengineering efforts.  For
example, errors in wages reported often occur because of confusion by
contractors and third parties over how to report workers and wages
for the peak week.  Labor is exploring the alternative use of
"man-hours" rather than peak week, which may be easier for
contractors and third parties to report. 

The process Labor is using, however, is unnecessarily costly, in
terms of both money and time.  On-site verification is a costly
approach to verifying wage data, and it delays the issuance of wage
rate determinations by months, especially when compared with
telephone verification that is combined with supporting payroll
records submitted upon request.  On-site verification requires a
cadre of auditors to travel to worksites around the country to review
payroll records.  While Labor's OIG found that access to payroll
records was essential to successfully verifying wage data, the
process need not require that the contractor be contacted in person
rather than by telephone.  Obtaining copies of payroll records by
mail as part of the telephone verification is significantly less
costly and takes less time. 

Relying on a random rather than a judgmental sample limits the
accuracy gains achievable through verification.  While a random
sample is often assumed to be better than a judgmental sample, it is
actually less effective in selecting wage data forms to verify that
will have an impact on the accuracy of the wage determinations. 
Labor does not gain the most significant benefits of a random
sample--that is, being able to assume that errors found in verified
wage data forms are representative of all wage data forms and adjust
wage rates accordingly--because it is not feasible to verify a
sufficiently large number of wage data forms.  In contrast, a
carefully designed judgmental sample to select contractors for
verification could consider the likelihood that the data will be
used, the number of workers within a job classification, and the
geographic dispersion of contractors. 

The impact of Labor's verification procedures on the accuracy of the
wage determinations is also limited by the action it takes when
documentation cannot readily be obtained from a contractor.  In our
view, at least two aspects of Labor's verification procedures
contribute to limiting accuracy.  First, contractors may refuse
access to supporting documentation for many legitimate reasons--such
as the time required--but contractors who refuse to provide the
supporting documentation are more likely than those who provide
access to have submitted fraudulent or carelessly reported data. 
Labor, however, (1) accepts the data and uses it as if documentation
had supported it and (2) allows federal contractors to deny access to
the supporting documentation even though Labor has the legal
authority to access their records.  While discarding all such data
might have negative consequences on Labor's ability to issue wage
determinations, accepting all such data may contribute to inaccurate
wage determinations.  Labor's approach does not achieve the needed
balance in deciding which data to include and exclude. 

Second, Labor sometimes eliminates or revises data inappropriately
when it does not seek supporting documentation from third parties
that have submitted it.  Wage data submitted for a project by a third
party are generally verified against the payroll or oral testimony of
the contractor associated with that project.  The third party that
submitted the original WD-10 is not provided the opportunity to
validate the information it submitted; final corrections are
generally provided only by the contractor.  As a result, data
supplied by third parties may be eliminated or revised inaccurately,
even though, in some cases, only the third party, not the contractor,
can provide supporting documentation, for example, for benefits
provided by a union.  Supporting documentation provided by third
parties could improve the completeness and thus the accuracy of the
data used. 


   RECOMMENDATIONS
------------------------------------------------------------ Letter :6

To reduce the cost of verification and increase the benefits, we
recommend that the Secretary of Labor direct the WHD Administrator to
revise verification procedures to maximize the expected value to be
gained from verification.  Specifically, Labor should

  -- increase the use of telephone verification--while decreasing
     on-site verification audits--and increase efforts to obtain
     payroll documentation from all selected submitters;

  -- change the procedures used to select wage data for verification,
     using a judgmental sample of wage data forms based on the
     potential impact of the data on prevailing wage rate
     determinations rather than using a random sample; and

  -- revise verification procedures to take more appropriate action
     when documentation cannot readily be obtained from a contractor,
     such as not using data when supporting documentation is
     requested but not provided, requiring documentation where
     possible, and giving third parties an opportunity to provide
     supporting documentation for data they submitted. 


   AGENCY COMMENTS
------------------------------------------------------------ Letter :7

We provided a draft of this report to the Department of Labor for its
review and comment.  It generally agreed with our recommendations and
agreed to implement them by revising the verification process.  Labor
also stated that our report was generally helpful and that some of
our recommendations would decrease costs and improve timeliness. 
However, the Department took issue with some of our conclusions
concerning the accuracy of survey data submissions by contractors and
the use of data from contractors who refuse auditors access to
supporting payroll records.  Labor also provided technical comments
and corrections, and we have revised our report as necessary. 

With regard to the accuracy of the survey data, Labor stated that
despite the many errors found by both the on-site auditors and
Labor's OIG, the limited revisions to wage determinations that
resulted from correcting these errors demonstrated that the wage
determinations closely approximated the accurate prevailing wage
rate.  We disagree, because the small number of data submissions
verified is not a valid representative sample of all data submissions
used to calculate the revised wage determinations.  Furthermore, the
OIG's report does not support Labor's conclusion.  It states that: 
"The errors we discovered did not materially change many of the wage
decisions because the data we sampled often represented a small
portion of the responses for an individual WH
survey.  .  .  .  If we had conducted more payroll reviews, we
believe more exceptions would have been identified and would have
revealed more material errors in published wage decisions."

With regard to the use of data when contractors refuse access to the
supporting payroll records, Labor disagreed with our conclusion that
contractors unable or unwilling to provide auditors access to payroll
records are more likely to have submitted fraudulent data than those
who provide records.  Labor based its conclusions that contractors
try to provide accurate and complete information on the data
verification that has been done to date by both Labor's OIG and WHD. 
Basing conclusions about contractors unwilling or unable to provide
access to payroll records on verification of data from contractors
that do provide access is not logical or convincing.  We continue to
believe that employers submitting fraudulent or unsubstantiated wage
data forms are unlikely to voluntarily provide access to payroll
records for review.  Because all verification efforts conducted to
date, including those of the OIG, have relied on voluntary access to
payroll records, the absence of fraud in verified wage data
submissions provides no evidence that contractors who denied access
did not submit fraudulent data. 

We have, however, clarified our conclusion on the use of data
submitted by contractors or third parties that do not cooperate with
verification efforts to allow Labor analysts to use judgment in
deciding when to exclude such data from its wage determination
calculations.  For example, we agree that Labor should consider
including data from contractors that routinely cooperated with data
verification efforts in the past and whose data were determined to be
generally accurate.  Another factor to consider would be the possible
adverse impact of discarding specific data on Labor's ability to
issue a wage determination. 

Finally, in agreeing to select data using a judgmental sample, Labor
stated that it intends to continue selecting some data for
verification using a systematic sample, albeit fewer than it does
now.  To the extent that data selected randomly represent a small
segment of all data verified, Labor's proposed approach is consistent
with our recommendation. 


---------------------------------------------------------- Letter :7.1

We are sending copies of this report to the Secretary of Labor,
appropriate congressional committees, and other interested parties. 

Please call me at (202) 512-7014 or Larry Horinko at (202) 512-7001
if you or your staffs have any questions about this report.  Other
major contributors were John Carney, Robert G.  Crystal, Lise Levie,
Ann P.  McDermott, Elizabeth T.  Morrison, and Ronni Schwartz. 

Carlotta C.  Joyner
Director, Education and
 Employment Issues

List of Addressees

The Honorable Ted Stevens
Chairman
The Honorable Robert C.  Byrd
Ranking Minority Member
Committee on Appropriations
United States Senate

The Honorable Arlen Specter
Chairman
The Honorable Tom Harkin
Ranking Minority Member
Subcommittee on Labor, Health and Human
 Services, and Education
Committee on Appropriations
United States Senate

The Honorable C.W.  Bill Young
Chairman
The Honorable David R.  Obey
Ranking Minority Member
Committee on Appropriations
House of Representatives

The Honorable John Edward Porter
Chairman
The Honorable David R.  Obey
Ranking Minority Member
Subcommittee on Labor, Health and Human
 Services, Education and Related Agencies
Committee on Appropriations
House of Representatives


OBJECTIVES, SCOPE, AND METHODOLOGY
=========================================================== Appendix I

The House Committee on Appropriations in its reports on
appropriations for the Departments of Labor, Health and Human
Services, and Education and related agencies for fiscal years 1998
and 1999 asked us to (1) review the Department of Labor's efforts to
verify a random sample of employers' wage data submissions and select
a sample of submissions for on-site data verification and (2)
determine the likely effect of these efforts on the accuracy and
timeliness of Davis-Bacon wage determinations. 

To describe Labor's actions, we interviewed Department of Labor
officials in the Wage and Hour Division (WHD) at both headquarters
and the five regional offices responsible for determining prevailing
wage rates.  At the southeast regional office in Atlanta, Georgia, we
interviewed WHD officials and reviewed the wage determination process
in more detail, including reading relevant documentation and
reviewing the process used to create a computerized database of wage
data forms.  We also obtained a draft of WHD's procedures for
telephone and on-site verification and task orders for on-site
verification.  In addition, we interviewed representatives of the
private accounting firm contracted to conduct on-site payroll
verification, who provided time schedules, cost data, and other
information about the firm's on-site reviews. 

To determine the likely effect of Labor's verification efforts, we
obtained and analyzed data from a number of different sources.  These
data include the following: 

  -- all available WHD preliminary analyses (forms WD-22) for all
     wage surveys sent to the contractor for verification for the
     18-month period from the beginning of on-site verification in
     April 1997 through September 1998;

  -- all preliminary and final reports completed by the accounting
     firm for on-site audits as of September 30, 1998; and

  -- electronic records of wage data forms (forms WD-10) maintained
     by WHD under contract with Computer Data Systems, Inc.,
     concerning the area surveys for which the accounting firm had
     issued final reports for on-site verification. 

Using these and other data provided by WHD, we conducted several
analyses. 

  -- To obtain the average error amount in wage rates and the
     percentage of wage data forms with errors for the nine area
     surveys for which the accounting firm issued final reports, we
     identified the dollar value of errors on each WD-10 by job
     classification.  We determined the dollar value of errors in
     wage rates by calculating the absolute value of the difference
     between the sum of the reported wages and fringe benefits and
     the sum of the verified wages and fringe benefits.  We weighted
     the amount of error by the lower of the number of employees
     reported or the number of employees verified.  Because the
     auditors were not consistent in their analysis and reporting, as
     necessary we made assumptions about individual wage rates, such
     as when an average wage rate was reported but individual wage
     rates were verified.  We then calculated the average of the
     absolute value of error for all workers. 

  -- To determine the percentage of contractors providing full and
     partial access to payroll records, we analyzed data in the 30
     preliminary reports summarizing the results of on-site
     verification audits.  Specifically, we counted the number of
     selected contractors the auditors reported as refusing access
     and those the auditors failed to access despite persistent
     efforts over a matter of weeks or months.  For those contractors
     who allowed the auditors access to the workplace, we identified
     the number of contractors unable or unwilling to provide access
     to the payroll records necessary to verify the wage rates
     reported on the selected forms WD-10. 

  -- To determine the amount of time between when Labor sent area
     survey data to the auditor for on-site verification and when
     Labor received the auditor's preliminary report, we relied on
     data provided by WHD for the 30 surveys with preliminary reports
     completed by the auditor.  We computed the time elapsed between
     the date Labor sent the survey to the auditor for on-site
     verification and the date the regional office received the
     preliminary report. 

In addressing the second objective, we recognize that while accurate
wage data are necessary, they are not sufficient to ensure that wage
determinations accurately reflect the prevailing wage that a
contractor would have to pay to obtain construction workers from the
local area at the market wage.  Other issues must be considered to
improve the wage determination process, such as the time lag between
obtaining the wage surveys and issuing the wage determinations. 
Labor is exploring options to reengineer its wage determination
process in the long term, which we will review at a later date.  We
did not attempt to assess the accuracy of the prevailing wage
determinations that result from these surveys, which was outside the
scope of this review.  We also did not verify the results of on-site
audit reviews; we focused on problems with procedures used rather
than contract compliance. 


LABOR'S WAGE DETERMINATION PROCESS
UNDER THE DAVIS-BACON ACT
========================================================== Appendix II

The Davis-Bacon Act requires that workers employed on federal
construction contracts valued in excess of $2,000 be paid, at a
minimum, wages and fringe benefits that the Secretary of Labor
determines to be prevailing for corresponding classes of workers
employed on projects that are similar in character to the contract
work in the geographic area where the construction takes place. 

To determine the prevailing wages and fringe benefits in various
areas throughout the United States, Labor's WHD periodically surveys
wages and fringe benefits paid to workers in four basic types of
construction (building, residential, highway, and heavy).\28

Labor has designated the county as the basic geographic unit for data
collection, although Labor also conducts some surveys setting
prevailing wage rates for groups of counties.  Wage rates are issued
for a series of job classifications in the four basic types of
construction, so each wage determination requires the calculation of
prevailing wages for many different trades, such as electrician,
plumber, and carpenter.  For example, in 1996 the prevailing wage
rates for the Washington, D.C., metropolitan area included wage rates
for 143 different construction trade occupations.  Because there are
over 3,000 counties, more than 12,000 surveys could be conducted each
year if every county in the United States was surveyed.  In fiscal
year 1997, Labor issued 1,860 individual wage rate determinations
based on 43 area wage surveys.  As shown in figure 1, Labor's wage
determination process consists of four basic stages: 

  -- planning and scheduling surveys of employer wages and fringe
     benefits in similar job classifications on comparable
     construction projects;

  -- conducting surveys of employers and third parties, such as
     representatives of unions or industry associations, on
     construction projects;

  -- clarifying and analyzing respondents' data; and

  -- issuing the wage determinations.\29


--------------------
\28 Heavy construction is a catch-all grouping that includes projects
not properly classified under the other three types of construction: 
for example, dredging and sewer projects. 

\29 A wage determination is the listing of wage and fringe benefit
rates for each classification of workers that the WHD administrator
has determined to be prevailing in a given area for a type of
construction.  Each wage determination involves establishing
prevailing wage rates for many occupations. 


   STAGE 1:  PLANNING AND
   SCHEDULING SURVEY ACTIVITY
-------------------------------------------------------- Appendix II:1

Labor annually identifies the geographic areas that it plans to
survey.  Because Labor has limited resources, a key task of Labor's
staff is to identify those counties and types of construction most in
need of a new survey.\30 In selecting areas for inclusion in planned
surveys, the regional offices establish priorities based on criteria
that include

  -- the need for a new survey based on the volume of federal
     construction in the area;

  -- the age of the most recent survey; and

  -- requests or complaints from interested parties, such as state
     and county agencies, unions, and contractors' associations. 

If a type of construction in a particular county is covered by a wage
determination based on collective bargaining agreements (CBA) and
Labor has no indication that the situation has changed such that a
wage determination should now reflect nonunion rates, an updated wage
determination may be based on updated CBAs.  The unions submit their
updated CBAs directly to the national office. 


--------------------
\30 In 1996, the average age of a wage survey was more than 7 years. 


      THE REGIONAL SURVEY PLANNING
      REPORT SHOWS WHERE FEDERALLY
      FINANCED CONSTRUCTION IS
      CONCENTRATED
------------------------------------------------------ Appendix II:1.1

Planning begins in the third quarter of each fiscal year when the
national office provides regional offices with the Regional Survey
Planning Report (RSPR).  The RSPR provides data, obtained under
contract with the F.W.  Dodge Division of McGraw-Hill Information
Systems, showing the number and value of active construction projects
by region, state, county, and type of construction and giving the
percentage of total construction that is federally financed.\31 Labor
uses the F.W.  Dodge data because they comprise the only continuous
nationwide database on construction projects.  Labor supplements the
F.W.  Dodge data with additional information provided to the national
office by federal agencies regarding their planned construction
projects.  The RSPR also includes the date of the most recent survey
for each county and whether the existing wage determinations for each
county are union, nonunion, or a combination of both. 

Using this information, the regional offices, in consultation with
the national office, designate the counties and type of construction
to be included in the upcoming regional surveys.  Although Labor
usually designates the county as the geographic unit for data
collection, in some cases more than one county is included in a
specific data-gathering effort. 

The regional offices determine the resources required to conduct each
of the priority surveys.  When all available resources have been
allocated, the regional offices transmit to the national office for
review their schedules of the surveys they plan to do:  the types of
construction, geographic area, and time periods that define each
survey. 

When Labor's national office has approved all regional offices'
preliminary survey schedules, it assembles them in a national survey
schedule that it transmits to interested parties, such as major
national contractor and labor organizations, for their review and
comment.  The national office transmits any comments or suggestions
received from interested parties to its affected regional offices. 
Organizations proposing modifications of the schedule are asked to
support their perceived need for alternative survey locations by
providing sufficient evidence of the wages paid to workers in the
type of construction in question in the area where they want a survey
conducted. 


--------------------
\31 The F.W.  Dodge data consider a project to be active from the
time on-site work begins (ground breaking) until it is released to
and accepted by the owner. 


      EACH REGIONAL OFFICE OBTAINS
      A FILE OF ACTIVE PROJECTS
      THAT MATCH ITS SURVEY
      OBJECTIVES
------------------------------------------------------ Appendix II:1.2

The target date for establishing the final fiscal year survey
schedule is September 15.  Once the national office has established
the final schedule, each regional office starts to obtain information
it can use to generate lists of survey participants for each of the
surveys it plans to conduct.  Each regional office then contacts
Construction Resources Analysis at the University of Tennessee, which
applies a model to the F.W.  Dodge data that identifies all
construction projects in the start-up phase\32 within the parameters
specified in the regional office's request and produces a file of
projects that were active during a given time period.  The time
period may be 3 months or longer, depending on whether the number of
projects active during the period is adequate for a particular
survey.  F.W.  Dodge provides information on each project directly to
the regional offices.  The F.W.  Dodge reports for each project
include the location, type of construction, and cost; the name and
address of the contractor or other key firm\33 associated with the
project; and, if available, the subcontractors.\34


--------------------
\32 F.W.  Dodge defines the start-up phase as one in which the
construction will commence within 60 days. 

\33 Other examples of key firms would be the owner or architect of
the project. 

\34 A subcontractor is an employer that has a contractual agreement
with the project's prime employer.  On a typical construction
project, most employees working on the job will be employees of
subcontractors. 


      ANALYSTS SCREEN PROJECTS TO
      DETERMINE THOSE TO BE
      SURVEYED
------------------------------------------------------ Appendix II:1.3

When the F.W.  Dodge reports are received by the regional offices,
Labor analysts screen them to make sure the projects meet four basic
criteria for each survey.  The project must

  -- be of the correct construction type,

  -- be in the correct geographic area,

  -- fall within the survey time frame, and

  -- have a value of at least $2,000. 

In addition to obtaining files of active projects, Labor analysts are
encouraged to research files of unsolicited information that may
contain payment evidence submitted in the past that is within the
scope of a current survey. 


   STAGE 2:  CONDUCTING SURVEYS OF
   PARTICIPANTS
-------------------------------------------------------- Appendix II:2

When the regional offices are ready to conduct the new surveys, they
send a WD-10 wage reporting form to each contractor (or employer)
identified by the F.W.  Dodge reports as being in charge of one of
the projects to be surveyed, together with a transmittal letter that
requests information on any additional applicable projects the
contractor may have.  Every WD-10 that goes out for a particular
project has on it a unique project code, the location of the project,
and a description of it.  Data requested on the WD-10 include a
description of the project and its location, in order to assure the
regional office that each project for which it receives data is the
same as the one it intended to have in the survey (see examples in
fig.  2).  The WD-10 also requests the contractor's name and address;
the value of the project; the starting and completion dates; the wage
rate, including fringe benefits, paid to each worker; and the number
of workers employed in each classification during the week of peak
activity for that classification.  The week of peak or highest
activity for each job classification is the week when the most
workers were employed in that particular classification.  The survey
respondent is also asked to indicate which of four categories of
construction the project belongs in. 

In addition, about 2 weeks before a survey is scheduled to begin,
regional offices send WD-10s and transmittal letters to a list of
third parties, such as national and local unions and industry
associations, to encourage participation.  Labor encourages the
submission of wage information from third parties, including unions
and contractors' associations that are not the direct employers of
the workers in question, in an effort to collect as much data as
possible.\35 Third parties that obtain wage data for their own
purposes may share it with Labor without identifying specific
workers.  For example, union officials need wage information to
correctly assess workers' contributions toward fringe benefits. 
Third-party data generally serve as a check on data submitted by
contractors if both submit data on the same project.  Regional
offices also organize local meetings with members of interested
organizations to explain the purpose of the surveys and how to fill
out the WD-10. 

Because the F.W.  Dodge reports do not identify all the
subcontractors, both the WD-10 and the transmittal letter ask for a
list of subcontractors on each project.  Subcontractors generally
employ the largest portion of on-site workers, so their
identification is considered critical to the success of the wage
survey.  Analysts send WD-10s and transmittal letters to
subcontractors as subcontractor lists are received. 

Transmittal letters also state that survey respondents will receive
an acknowledgment of data submitted and that the respondent should
contact the regional office if one is not received.  Providing an
acknowledgment is intended to reduce the number of complaints that
data furnished were not considered in the survey.  Labor analysts
send contractors who do not respond to the survey a second WD-10 and
a follow-up letter.  If they still do not respond, analysts attempt
to contact them by telephone to encourage them to participate. 


--------------------
\35 Labor officials said that third-party data submissions generally
account for about one-third of all wage survey submissions.  The
percentage of survey respondents that are third parties can be
substantial for surveys of metropolitan areas.  Staff estimated that
third-party participation may have been as high as 50 percent for one
survey of metropolitan building construction.  There is little or no
third-party participation in surveys of rural areas, staff said. 


   STAGE 3:  CLARIFYING AND
   ANALYZING RESPONDENTS' DATA
-------------------------------------------------------- Appendix II:3


      ANALYSTS REVIEW THE DATA
      SUBMITTED AS THEY RECEIVE
      THEM
------------------------------------------------------ Appendix II:3.1

As the Labor wage analysts receive the completed WD-10s in the
regional offices, they review and analyze the data.  Labor's training
manual guides the analyst through each block of the WD-10, pointing
out problems to look for in data received for each one.  Analysts are
instructed to write the information they received by telephone
directly on the WD-10 in a contrasting color of ink, indicating the
source and the date received.  They are instructed to draw one line
through the old information so it is still legible. 

Labor's wage analysts review the WD-10s to identify missing
information, ambiguities, and inconsistencies that they then attempt
to clarify by telephone.  For example, an analyst may call a
contractor for a description of the work done on a project in order
to confirm that a particular project has been classified according to
the correct construction type.  An analyst may also call a contractor
to ask about the specific type of work that was performed by an
employee in a classification that is reported in generic terms, such
as a mechanic.  In that situation, the analyst would specify on the
WD-10 whether it is a plumber mechanic or some other type of mechanic
to make sure that the wages reported are appropriately matched to the
occupations that are paid those rates. 

Similarly, because of variations in area practice, analysts may
routinely call to find out what type of work the employees in certain
classifications are doing.  This is because in some areas of the
country some contractors have established particular duties of
traditional general crafts--for example, carpenters--as specialty
crafts, which are usually paid at lower rates than the general craft. 


      VERIFYING THIRD-PARTY DATA
------------------------------------------------------ Appendix II:3.2

See letter portion of this report for a description of the
verification process. 


      DATA ARE RECORDED AND
      TABULATED
------------------------------------------------------ Appendix II:3.3

When an analyst is satisfied that any remaining issue with respect to
the data on the forms WD-10 for a particular project have been
resolved, the data are recorded and tabulated.  The analyst enters
them into a computer, which uses the data to generate a Project Wage
Summary, form WD-22a, for reporting survey information on a
project-by-project basis.  The WD-22a has a section for reporting the
name, location, and value of each project; the number of employees
who were in each classification; and their hourly wage and fringe
benefits.  It also has a section for reporting the date of completion
or percentage of the project completed, whichever is applicable. 


      ANALYSTS DETERMINE WHETHER
      DATA ARE ADEQUATE
------------------------------------------------------ Appendix II:3.4

At least 2 weeks before the survey cutoff date, the response rate for
the survey is calculated to allow time to take follow-up action if
the response rate is determined to be inadequate.  For example, WHD
operational procedures specify that if data gathered for building or
residential surveys provide less than a 25-percent usable response
rate or less than one-half of the required key classes of workers,\36
the analyst will need to obtain data from comparable federally
financed projects in the same locality.\37

If an analyst has no data on occupations identified by Labor as key
classifications of workers for the type of construction being
surveyed, Labor's procedures require him or her to call all the
subcontractors included in the survey who do that type of work and
from whom data are missing, to try to get data.  If the analyst still
cannot obtain sufficient data on at least one-half of the required
key classes, consideration must be given to expanding the scope of
the survey geographically to get more crafts represented.  If the
overall usable response rate for the survey is 25 percent or more,
data on three workers from two contractors are sufficient to
establish a wage rate for a key occupation. 

After the survey cutoff date, when all valid data have been recorded
and tabulated, the final survey response rate is computer-generated. 
Typically, it takes a WHD analyst 4 months to conduct a survey. 


--------------------
\36 Labor defines key classes of workers as those determined
necessary for each of the four types of construction surveys. 

\37 Since 1985, regulation has prohibited, to the extent practicable,
the use of wages for federal construction in determining prevailing
wages. 


      PREVAILING WAGE RATES ARE
      COMPUTER-GENERATED
------------------------------------------------------ Appendix II:3.5

Once all the valid project data have been entered, the prevailing
wage rate for each classification of worker can be generated by
computer.  If there is a majority of workers paid at a single rate in
a job classification, that rate prevails for the classification.  The
wage rate needs to be the same, to the penny, to constitute a single
rate.  If there is no majority paid at the same rate for a particular
classification, a weighted average wage rate for that occupation is
calculated. 

The prevailing wage rate for each occupation is compiled in a
computer-generated comprehensive report for each survey, called the
Wage Compilation Report, form WD-22.  The WD-22 lists each occupation
and the wage rate recommended for that occupation by the regional
office.  The form indicates whether the rate is based on a majority
or a weighted average, and provides the number of workers for which
data were used to compute each wage rate.  The regional offices
transmit survey results to the national office, which reviews the
results and recommends further action if needed. 


   STAGE 4:  ISSUING THE WAGE
   DETERMINATIONS
-------------------------------------------------------- Appendix II:4

When all its recommendations have been acted upon, the national
office issues the wage determination.  These determinations are
final.  There is no review or comment period provided to interested
parties before they go into effect.  Access to wage determinations is
provided both in printed reports available from the U.S. 
Superintendent of Documents and on an electronic bulletin board. 
Modifications to general wage determinations are published in the
Federal Register. 


   LABOR'S APPEALS PROCESS
-------------------------------------------------------- Appendix II:5

An interested party may seek review and reconsideration of Labor's
final wage determinations.  The national office and the regional
offices accept protests and inquiries relating to wage determinations
at any time after a wage determination has been issued.  The national
office refers all the complaints it receives to the relevant regional
offices for resolution.  Most inquiries are received informally by
telephone, although some are written complaints.  Regional office
staff said that a majority of those with concerns appear to have
their problems resolved after examining the information (collected on
a form WD-22a) for the survey at issue, because they do not pursue
the matter further.  If an examination of the forms does not satisfy
the complainant's concerns, the complainant is required to provide
information to support his or her claim that a wage determination
needs to be revised.  The national office modifies published wage
determinations in cases in which regional offices, on the basis of
evidence provided, recommend that it do so, such as when it has been
shown that a wage determination was the result of an error by the
regional office.  However, some of those who seek to have wage rates
revised are told that a new survey will be necessary to resolve the
particular issue that they are concerned about.  For example, if the
wage rates of one segment of the construction industry were not
adequately reflected in survey results because of a low rate of
participation in the survey by that segment of the industry, a new
survey would be necessary to resolve this issue. 


      AN INTERESTED PARTY MAY
      APPEAL A DECISION OF LABOR'S
      WHD ADMINISTRATOR
------------------------------------------------------ Appendix II:5.1

Those who are not satisfied with the decision of the regional office
may write to the national office to request a ruling by Labor's WHD
Administrator.  If the revision of a wage rate has been sought and
denied by a ruling of Labor's WHD Administrator, an interested party
has 30 days to appeal to the Administrative Review Board for review
of the wage determination.  The board consists of three members
appointed by the Secretary of Labor.  The Solicitor of Labor
represents WHD in cases involving wage determinations before the
Administrative Review Board.  A petition to the board for review of a
wage determination must be in writing and accompanied by supporting
data, views, or arguments. 




(See figure in printed edition.)Appendix III
SAMPLE LETTERS SENT TO CONTRACTORS
SELECTED FOR VERIFICATION
========================================================== Appendix II



(See figure in printed edition.)


METHODOLOGICAL ISSUES IN LABOR'S
SAMPLING PROCEDURES
========================================================== Appendix IV

In reviewing Labor's process of sampling wage data for verification,
we identified problems with its sampling methodology that are
primarily technical in nature.  Specifically, although the Congress
directed that Labor use a random sample in selecting wage data to
verify, and Labor describes its sample as being "random," the
selection method used does not meet the criteria for randomness. 
Randomness would require that each WD-10 have an equal opportunity
for being selected.  However, while Labor uses a systematic sample
that does not target any specific wage data for verification, it
fails to meet the criteria for randomness because not all wage
submissions have an opportunity of being selected.  Labor uses a
systematic sample, organizing WD-10s by project and by contractor
prior to selection and then selecting them based on a fixed interval. 
Labor does not require that the first WD-10 selected be based on a
number chosen purely by chance.  For example, to select a sample for
telephone verification of data submitted by contractors, Labor
procedures direct the wage analyst to select the 50th, then the
100th.  However, because the data are organized prior to selection,
the first 49 WD-10s are predetermined on the basis of the specific
project and contractor involved.  Therefore, the WD-10s for those
projects and contractors do not have any chance of being selected for
verification. 

Labor officials in the national office told us that because of this,
they do not know whether they have selected enough data for telephone
and on-site verification to ensure the accuracy of the data used, or
whether they have selected more data than needed and are wasting
resources.  As a result, they do not know the extent to which data
used to calculate wage rates have been verified, if at all.  For
example, using the hypothetical wage data forms in figure 2, Labor
would know the number of wage forms it had selected but would not
know whether it was verifying wages for drywallers, electricians, or
painters.  If Labor had selected only one of the two wage data forms
for verification, it would disregard the fact that one form reported
wages for 80 drywallers and the other form reported wages for 2; it
would merely report that it had verified 50 percent of the WD-10s. 
In one of the on-site audit reports we examined, although Labor
sampled 42 percent of WD-10s, the on-site auditor reviewed 28 percent
of workers' wages and fringe benefits out of all wage data submitted
for workers employed in that geographic area (390 out of 1,369).\38
This resulted in a review of 100 percent of data used in calculating
prevailing wage determinations for job classes such as stone masons,
and no verification of data used in other job classes, such as
painters. 

To select a random sample that would ensure the accuracy of the data
used to determine prevailing wage rates, Labor would have to sample
workers within each job classification rather than sample wage data
forms as it does now, and it would have to select a sufficient number
of workers within each classification.  We calculated the sample size
required for a statistically representative sample in order to be
within 50 cents per hour of the correct wage for one area survey (see
table IV.1).  The table shows that Labor would need to select a
sample size equal or close to the total number of workers, because
data reported on the number of workers by job classification can be
small. 



                         Table IV.1
          
           Sampling Sizes, by Job Classification,
            Required to Ensure a Representative
              Sample for Selected Area Survey

                    Number of workers
            ----------------------------------
                                                 Number of
                                      Required  additional
                          Selected         for     workers
                           for on-  representa    for whom
Job          Listed on        site        tive   data need
classifica    form WD-  verificati      random       to be
tion                22          on    sample\a    verified
----------  ----------  ----------  ----------  ----------
Accoustica           6           2           6           4
 l Ceiling
 Mechanic
Backhoe              3           0           3           3
 Operator
Boilermake         140          26          30           4
 r
Brick               11           5          11           6
 Mason
Carpenter           37           6          30          24
Carpenter            8           0           8           8
 -Form
Carpenter           12           5          12           7
 Helper
Case and             3           0           3           3
 Cabinet
 Installer
Cement               1           0           1           1
 Mason
Crane                1           0           1           1
 Operator
Dozer                2           0           2           2
Drywall              5           0           5           5
 Finisher
Drywall              1           0           1           1
 Finisher
 Helper
Drywall              9           0           9           9
 Mechanic
Drywall              3           0           3           3
 Mechanic
 Helper
Electricia          33           3          30          27
 n
Electricia          25           3          25          22
 n Helper
Exterminat           2           0           2           2
 or
Fence                2           0           2           2
 Erector
Front End            2           0           2           2
 Loader
Glazier              7           0           7           7
Glazier              5           0           5           5
 Helper
HVAC                16           0          16          16
 Helper
HVAC                38           0          30          30
 Mechanic
Ironworker          13           1          13          12
 ,
 Structura
 l
Laborer,           127          22          30           8
 Unskilled
Landscaper           4           0           4           4
Lather               2           0           2           2
Mason                8           0           8           8
 Tender
Metal Stud           4           0           4           4
 Framer
Millwright           2           0           2           2
Painter             20           0          20          20
Plasterer           13           0          13          13
Plumber             19           0          19          19
Plumber             15           0          15          15
 Helper
Roofer               9           0           9           9
Sheetmetal           9           0           9           9
 Worker
Sheetmetal          18           0          18          18
 Worker -
 Metal
 Bldg.
Sign                 2           0           2           2
 Installer
Softfloor           10           0          10          10
 Layer
Softfloor            3           0           3           3
 Layer
 Helper
Sprinkler            4           4           4           0
 Fitter
Sprinkler            2           1           2           1
 Fitter
 Helper
Tile                 3           0           3           3
 Setter
Truckdrive           5           0           5           5
 r -Tri-
 Axle
                   664          78         439         361
----------------------------------------------------------
\a Calculations are based on a confidence interval of plus or minus
50 cents per hour. 



(See figure in printed edition.)Appendix V

--------------------
\38 To obtain these figures, for one area survey we calculated the
number of workers with wages reported that were verified as a
percentage of the total number of workers with wages reported that
were used to calculate wage determinations. 


COMMENTS FROM THE DEPARTMENT OF
LABOR
========================================================== Appendix IV



(See figure in printed edition.)



(See figure in printed edition.)



(See figure in printed edition.)



(See figure in printed edition.)



(See figure in printed edition.)

RELATED GAO PRODUCTS

Davis-Bacon Act:  Process Changes Could Address Vulnerability to Use
of Inaccurate Data in Setting Prevailing Wage Rates
(GAO/T-HEHS-96-166, June 20, 1996). 

Davis-Bacon Job Targeting (GAO/HEHS-96-151R, June 3, 1996). 

Davis-Bacon Act:  Process Changes Could Raise Confidence That Wage
Rates Are Based on Accurate Data (GAO/HEHS-96-130, May 31, 1996). 

Addressing the Deficit:  Budgetary Implications of Selected GAO Work
for Fiscal Year 1996 (GAO/OCG-95-2, Mar.  15, 1995). 

Davis-Bacon Act (GAO/HEHS-94-95R, Feb.  7, 1994). 

The Davis-Bacon Act Should Be Repealed (GAO/HRD-79-18, Apr.  17,
1979). 


*** End of document. ***