\
            United States
            Environmental Protection
            Agency
Office of
Inspector General
401 M Street, SW
Washington, DC 20460
            Report of Audit
          REPORT  ON CERCLIS REPORTING
          AUDIT NO. E1SFF9-15-0023-0100187

                    MARCH 12,  1990




-------
                        TABLE OF CONTENTS
SCOPE AND OBJECTIVES

SUMMARY OF FINDINGS

ACTION REQUIRED


BACKGROUND


FINDINGS AND RECOMMENDATIONS

     1.
CERCLIS Report Program Documentation Needs
Significant Improvement

Improvements in CERCLIS Report Library and Change
controls Are Needed

Formal Testing Procedures for CERCLIS Reports Are
Needed
TECHNICAL EXHIBITS

GLOSSARY

APPENDIX A  OSWER'S Response to the Draft Report

APPENDIX B  Distribution
                                                      1


                                                      2

                                                      4


                                                      5


                                                      7
11



15


19


43


44


43

-------
         UNITED STATES ENVIRONMENTAL PROTECTION AGENCY

                     WASHINGTON, DC 20460
                                                           OF
                                                   THE INSPECTOR GENERAi.
MEMORANDUM

SUBJECT:  Report on CERCLIS Reporting
          Audit Report No. E1SFF9-15-0023-0100187
TO:
FROM:     Kenneth A. Konz 'I
          Acting Assistant: Inspector General for Audit
Don R. Clay, Assistant Administrator
Office of Solid Waste and Emergency Response
SCOPE AND OBJECTIVES

We have completed an audit of the accuracy, reliability, and
consistency of standard CERCLIS  (Comprehensive Environmental
Response and Liability Information System) reports.  The
objective of the audit was to determine whether reporting was
being performed accurately and reliably.

The audit was conducted primarily at EPA Headquarters.  We also
visited Regions 2 and 3 to obtain background information and
users' views on working with CERCLIS.  We interviewed
Headquarters Superfund program staff to determine how they
interacted with CERCLIS and whether they had encountered any
problems in working with the system.  Our work included:

     examination of documentation for report programs;

     verification of controls over report program libraries;

     comparison of related reports for consistency;

     review of quality control procedures over report
     programming;

     verification of the accuracy of five reports by
     duplicating report selection logic provided to us by
     CERCLIS officials; and

     determination of whether data listed in the Superfund
     Progress Report was supportable (adequacy of its audit
     trail).

-------
Field work began March  1989 and was completed September 1989.
The audit reviewed CERCLIS reporting and related controls as they
existed  from June through September 1989.

We performed the audit  in accordance with Government Auditing
Standards  (1988 Revision), issued by the Comptroller General of
the United states.  In  the course of our review, we identified
and reviewed internal controls related to the development and
modification of CERCLIS reports programming.  The weaknesses we
found are included in this report along with our recommendations
to strengthen  internal  controls.  No other issues came to our
attention which were significant enough to warrant expanding the
scope of our review.

SUMMARY  OF FINDINGS

While our review found  that the audit trail for the Superfund
Progress Report was adequate, we also found that management of
CERCLIS  report development and modification was deficient due to
the absence of good controls over documentation, software
changes, and testing.   Although employment of strict procedures
that would ensure the existence of these controls has long been
an accepted practice within data processing, both government and
private  industry, CERCLIS management did not ensure that such
procedures were put into effect.  As a result, material errors
have arisen within CERCLIS reports and any information currently
being reported by the system must be considered suspect and
employed cautiously.

Except for the SCAP (see background section) reports, which
report plans and accomplishments, CERCLIS reports were not
heavily  utilized.  In fact, the OIG audit of the Agency's
Superfund Annual Report for FY 1988 demonstrated the significant
differences between the information being reported at the
national level, based on CERCLIS, and the information being
maintained in the regions.

The reason for continued regional support for the CERCLIS SCAP
reporting was that users' budgets were tied to entry of strategic
data (see Background section) into the system.  Regions were not
given their funds, via  the Advice of Allowance process, unless
and until they fulfilled their SCAP data entry requirements.

Specifically, we found  that CERCLIS report development and
modification practices  were deficient in the following areas:

1.  Documentation at the programming level was poor.  Documenting
the specific steps of a program and the rationale behind these
steps is vital to the ability of future programmers to be able to
correctly modify existing programs.  With only one exception, all

-------
of the program source code  we examined contained no documenting
comments and did not agree with descriptions of report logic
prepared by contractors.  It will certainly be very difficult for
future programmers to make changes to the reports.  It may even
be necessary to completely rewrite programs rather than simply
change them due to a lack of knowledge of how they work.  In
addition, without a good understanding of how these programs
select and process records, it may become impossible to resolve
questions regarding what the reports mean and how they relate to
other reports.  In fact, it may become impossible to rely on them
at all.

2.  Control over report programs was not adequate.  Programmers
could change reports without approval.  Control over programming
was so weak that the correct source code for one production
report could not be found.  The result is that reports produced
by these programs cannot be relied upon because users cannot be
certain that the correct and error-free version of a program is
the one being executed.  If users cannot rely on these reports,
they will either not be used or be used with considerable risk.
In general, as noted above, with the exception of the SCAP
reports, CERCLIS reporting is not being utilized for day to day
management decision making.

Information listed in related reports was inconsistent because of
the same lack of control.  A comparison of reports which would
appear to users to list identical items showed significant
differences.  Whether the reason, in each instance, was error or
inadequate labeling of the reports, the result is the same; loss
of faith in the reliability of the system and, whenever possible,
avoidance of use of information produced by it.

3.  Formal testing and acceptance procedures for new report
programming or changes to old reports did not exist.  If CERCLIS
officials did not receive negative comments on a report from
field users, it was assumed that all was well.  The number of
coding errors found in the five reports reviewed demonstrates
that this methodology was seriously flawed.  In fact, one report,
the SCAP-18 (FY89 FINANCIAL SUMMARY - TOTAL FUNDING) understated
prior years obligations for one region by $500 million.

We provided a draft report, dated December 13, 1989, to the
Assistant Administrator for the Office of Solid Waste and
Emergency Response.  The AA's complete response to that draft is
included in this report as Appendix A.  According to the
response,  all recommendations were implemented by December 29,
1989.   The response included a general comment to which we would
like to respond.
     See Glossary at the end of the report,

-------
The AA stated that  "There  is a recurring theme in the draft
report that CERCLIS reports are not heavily utilized."  He
pointed out that CERCLIS reports were used for strategic decision
making through use  of the  SCAP reports, for FOIA requests, and
for day to day monitoring  by regional program managers using
CERCLIS data on their local area networks.

The statement was made in  the SUMMARY OF FINDINGS that, except
for the SCAP reports, CERCLIS reports were not heavily utilized.
We reached this conclusion after having spoken to users in the
regions and Headquarters.  In fact, only two of the SCAP reports
were used consistently:  14B and 16.  They have to do with
reporting on accomplishments and plans.  We also pointed out that
the only reason the SCAP reports were used to any great extent
was that Regions were not  given their funds via the Advice of
Allowance process,  unless  and until they fulfilled their SCAP
data entry requirements.

With regard to FOIA requests, while they are important, it is not
cost-effective to build a  large mainframe system like CERCLIS
just to process them.

In discussing CERCLIS with regional program managers, we found
that they did not use CERCLIS to perform their duties.  The local
area networks referred to  in the AA's response are completely
separate systems, called collectively HASTELAN.  Their primary
link with CERCLIS is the uploading of new data to mainframe
CERCLIS.   Even in  the case of WASTELAN, not all program managers
said that they would use it for day to day monitoring of
activities.  It should be  kept in mind that CERCLIS, and not
WASTELAN, reporting was being audited.
ACTION REQUIRED

As all recommendations have been implemented and there are no
unresolved issues, no additional responses to this report are
necessary.  The audit is considered closed.  A Special Review to
verify that all recommendations have been implemented adequately
will be performed in approximately one year from the date of this
report.

We appreciated the cooperation and assistance of your staff.
Should you or your staff have any questions, please have them
contact Sheldon Kantrowitz, Acting Chief, ADP and Statistics
Unit, on 382-7603.

-------
BACKGROUND

The Comprehensive  Environmental Response, Compensation, and
Liability Information System  (CERCLIS) is a data base that was
developed to aid EPA Headquarters and regional personnel with
Superfund site, program, and  project management.  CERCLIS is the
required and sole  source of Superfund planning and accomplishment
data.  As such, information reported from it serves as the
primary basis  for  strategic decision-making for the multi-billion
dollar Superfund program.  A  region will not be issued monies
through the Advice of Allowance process unless it satisfies all
of its CERCLIS data entry responsibilities.

Data entry responsibilities are at the regional level.  Any user
can access any of  the standard reports from the on-line menu.
Headquarters relies on CERCLIS as the repository of data on plans
and accomplishments.

CERCLIS has been designed to  meet the reporting requirements of
Superfund Comprehensive Accomplishments Plan (SCAP), Strategic
Planning and Management System  (SPMS), and Superfund Progress
Report (SPR).  Site managers  monitor activities at assigned sites
and develop quarterly SCAP plans in cooperation with other staff.
Regional Information Management Coordinators (IMCs) review the
quarterly SCAP plans.  The site managers identify and report
changes in site plans as they occur and coordinate changes with
the IMCs to ensure compliance with each region's overall SCAP.
Regional managers  also ensure that changes have been recorded
correctly in CERCLIS.

The SCAP is a  quarterly report of each region's plans and
accomplishments.   The Office  of Emergency and Remedial Response
(OERR)  and the Office of Waste Programs Enforcement (OWPE) use
SCAP to allocate resources to regions and to evaluate regional
performance.

The SPMS is a  quarterly report of each region's accomplishments
with respect to certain measures and targets.  The Administrator
and others use SPMS to assess progress in each EPA program.
Superfund SPMS targets and measures are incorporated in the SCAP.

The SPR is a monthly report of accomplishments to date at
individual sites and for the  Superfund program as a whole.  The
information reported in the SPR is very similar to and overlaps
in many cases  with SCAP information.  SPR is used for general
information on the status to  date of the Superfund program.

CERCLIS has three  main groups of data elements:  Headquarters
required data  elements with national definitions, non-required
data elements with national definitions, and non-required data
elements with  definitions that each region is free to set, such
as intermediate subevents and milestones.  CERCLIS consists of

-------
two primary data bases:  one site-specific data base  (also named
CERCLIS) and a second non-site-specific data base  (CERHELP).  A
third data base, CERTRAN, is used for logging transactions.  The
Update reports (UPDT) provide information drawn from the CERTRAN
data base.

-------
FINDINGS AND RECOMMENDATIONS

1.
     Improvement
nentation Neec
Our review of documentation for six reports currently available
to users on the CERCLIS Reports Menu showed that documentation at
the computer program level needed significant improvement.  We
found discrepancies between write-ups of programming logic
(Report Specification Forms), write-ups contained within the
official CERCLIS Reports Library, and the source code1  itself.
The source code and the Report Specification Forms were used by
programmers to understand and maintain  (correct and modify) the
reports.  The Reports Library write-ups were prepared for and
employed by users as a tool for understanding the information
provided by each report.  The source code contained no comments
relating what information each of the program variables contained
and how processing proceeded.  CERCLIS management was not
ensuring that the contractors who were developing report
programming provide and maintain adequate documentation.  Lack of
adequate documentation can lead to difficulty in learning how a
particular program works and/or the inability to modify it.  It
may be necessary to completely rewrite a program rather than
simply modify it.  In fact, when documentation does not match
output, both programmers and users tend to lose confidence in a
system.

National Bureau of Standards (NBS) Special Publication 500-106,
Guidance on Software Maintenance  (December 1983), section 8.1.5,
addresses the necessity for the comments in computer source code.
It emphasizes the importance of making programs readable;
providing information on program history, purpose, input/output
requirements, formats, etc.; and helping the programmers who have
to maintain programming understand aspects of the code that are
not clear.  Sections 4.2 and 10.2.4 discuss the necessity for
good documentation and enforcement of documentation standards.
Documentation is described as communication which "should include
design specifications, code comments, programmer notebooks, and
other documentation.1*  Coding and documentation ". . .standards
must be continually enforced via technical review and examination
of all work performed by the software maintenance staff."
Without adequate program documentation, computer program
maintenance becomes more difficult and programmer error more
frequent.

We found discrepancies and inconsistencies between computer
program logic (Report Specification Forms), write-ups contained
          See Glossary at the end of the report.

                                7

-------
 in the official CERCLIS Reports Library Manual, and the computer
 program source language.  Source code contained no comments and
 did not relate the  information each of the programming variables
 contained and how processing proceeded.  One exception was
 Enforcement Report  22  (ENFR-22).  This program had adequate
 comments.  The technical details of these findings are presented
 in exhibits 1-1 through 1-5.

 Although SCAP  (Superfund Comprehensive Accomplishments Plan)
 reports are the most critical and the most widely used within
 CERCLIS, documentation for them was as bad as for any of the
 other reports reviewed.  The preparation of write-ups (Report
 Specification 'Forms) for SCAP reports was not initiated until FY
 1989.  Not all were completed as of June 1989.  None of the
 source code for these  reports had comments describing the fields
 used and the flow of the logic.  This, we were told, was planned
 for a later time.   We  were also told that the rush to produce a
 multitude of reports during the two prior fiscal years resulted
 in documentation being a secondary priority.

 The other reports reviewed represented, although secondary to the
 strategic SCAP reports, some of the most widely used and relied
 upon CERCLIS products.  They included the financial, remedial,
 removal, and enforcement areas.  The same poor documentation
 practices and discrepancies were found in each instance.

 Due to the lack of  documentation that adequately describes how
 report programming  works, it will be very difficult for future
 programmers to make changes to the reports.  It is likely that
 programmers are experiencing difficulties at this time because of
 the poor quality of the program documentation.  Inconsistent
 documentation exacerbates the problem.  It may even be necessary
 to completely rewrite  programs rather than simply change them due
 to a lack of knowledge of how they work.  In addition, without a
 good understanding  of  how these programs select and process
 records, it may become impossible to resolve questions regarding
what the reports mean  and how they relate to other reports.  In
 fact, it may become impossible to rely on them at all.

User documentation  was also inadequate.  Write-ups in the Reports
 Library were developed so that users might understand what
 information was being  listed in each of the reports.  To the
 extent that this information was not reliable, the potential for
 error increased.

We found that management was not ensuring that contractors
provide and maintain sufficient and accurate documentation  for
report programming,  in this regard, CERCLIS officials were
relying exclusively on their contractors without any effort at
verification.
                                8

-------
CONCLUSIONS

CERCLIS management had not given adequate attention to program
documentation.  Accurate and timely documentation of computer
programs  is essential for efficient system maintenance.  It is
too  late  to develop  this type of documentation after the original
programmers have  left or so much time has passed that they no
longer have an  adequate grasp of how programs work.  Given that
most of the life  of  a program or system is dedicated to
maintenance, good programming documentation is essential as the
means by  which  the mechanics of a program are communicated to
future maintainers.  It is less expensive to write documentation
than to have to completely rewrite programs at a later date.  In
addition, the integrity of CERCLIS will be in question as long as
its  documentation is insufficient or inaccurate.

DRAFT RECOMMENDATIONS

In response to  our position paper detailing the findings
discussed herein, the Director, OERR, stated that the following
steps were being  considered for implementation:

          When  reports are examined for inclusion in
          the national library, comments will be
          inserted into the code and the CERCLIS
          Reports Library will be modified to reflect
          the changes.  A three way check will be
          instituted between Report Specification Form,
          the CERCLIS Reports Library, and the
          documented code.

          Signoff from the EPA program office (report
          owner)  will be required for all report
          specifications, reports library
          documentation, and test/sample reports.

          In the  future all report documentation will
          undergo comprehensive review where it will be
          evaluated  for adherence to selection criteria
          standards  as well as documentation standards
          prescribed in the National Bureau Standards
          Guidance on Software Maintenance and
          Documentation and FIPS PUB 38.

We recommended that  the Assistant Administrator for OSWER
implement and monitor the actions listed in the OERR response to
our position paper on CERCLIS documentation.

-------
SUMMARY OF OSWER'S COMMENTS!

OERR implemented each of the three steps outlined in its response
to our position paper on CERCLIS documentation by December 29,
1989.  OERR additionally requested that the OSWER Information
Management Staff develop a monitoring plan or procedures to
implement the monitoring portion of our recommendation.

OUR EVALUATION OF OSWER'S COMMENTSj.

The decision to implement the three courses of action should help
to resolve CERCLIS documentation problems.  We do not see the
need, however, to create any special procedures to support those
actions and to finalize the recommendation.  The three steps
presented by OERR are a normal part of system maintenance.  As
long as CERCLIS is active, they must be performed on a daily
basis.   No further recommendations are being made.
                               10

-------
2.   Improvements in CERCLIS Report Library and Chance Controls
     Are Needed

Our review found that CERCLIS report library and change controls
were not adequate,  inconsistencies and inaccuracies arose in
CERCLIS reports because its management did not ensure that
controls over program report libraries  and changes to programs
within those libraries were adequate.  Procedures for providing
these controls were never formalized.  Reliance was placed on
verbal communications and meetings between CERCLIS managers and
contractor representatives.  The contractor who was given the
responsibility of controlling access and changes to those files
was found not -to be performing that function.  Documentation
tracking the disposition of programming changes and identifying
individual responsibilities was not complete;  it ended at the
point at which changes were submitted to a contractor for
implementation.  Contractors should have maintained documentation
describing each step in the change process, naming each
individual involved, annotating pertinent dates, and detailing
the testing methodology and results of testing.  Poor library
controls result in the dissemination of erroneous and
inconsistent data and, potentially, in loss of confidence in both
the system and Superfund program.

Adequate library controls ensure that only authorized,
tested, and the correct version of programs are executed.
Lack of control severely impacts the integrity of a system.
Procedures should exist whereby changes to production
programs must be approved formally, and the process of
designing, coding, testing, and implementing modifications is
fully documented and controlled.

Reliable reporting also ensures that reports purporting to
provide similar data be consistent and reconcilable.
If there are valid reasons for the existence of reports which
appear to contain related data and yet are materially different,
then their differences should be made obvious to readers by means
of such tools as headings, labels and footnotes.

In the course of our audit work, we examined and tested change
control procedures, reviewed controls over program libraries,  and
compared apparently related reports for consistency.  As a result
of this effort and information gathered during performance of
other audit steps, we concluded that CERCLIS librarian and change
control functions needed improvement.
     See Glossary at the end of the report.

                                11

-------
Changes to the production program library3 can and have been made
by programmers without approval because the responsible
contractor was not providing the necessary controls and
oversight.  In fact, this contractor did not even have copies of
the source code for two programs which had been in production
since the beginning of FY 1989.  Likewise, when a change to one
report was made, no attention was given to changing related
reports.  The responsibility for this type of error was shared by
both the contractor performing the coding changes and the first
contractor, responsible for controlling the production library
itself.  There were also no controls to ensure that a particular
version of the source code for a production report was the
correct one and that all changes to the report over a period of
time had been documented.  In one instance, our auditors were
given access to the wrong source code for a production report.
Contract staff did not know where the correct version of the
source code was.

We found it difficult to verify the implementation of changes
because the paper trail ended once requested changes were
submitted to contractors.  Furthermore, there were no formal
procedures for informing users that requested changes had been
made.  In the case of emergency fixes, no attempt was made to
notify users at all.

Our work on reporting inconsistency revealed material
discrepancies in selected financial, SCAP, and enforcement
reports.  In total, twelve CERCLIS reports were reviewed.

An examination of two reports, SCAP-13 and SCAP-14A, led us to
the conclusion that certain items listed in them were related and
that they could be reconciled to each other.  In response to a
position paper on this subject, which had been issued to the
Director, OERR, we were informed that our comparison of the two
reports was not valid because the ". . .reports calculate
mutually exclusive SCAP categories and address separate
functions."

It is our opinion that reports which appear to be related and
show apparently inconsistent information do as much damage to the
credibility of a system as reports which are truly related but
which do not reconcile due to error.  Users had the same reaction
to both apparent and actual inconsistencies.  That reaction is
lack of trust in the reports and abandonment of them.  The case
of SCAP-13 and SCAP-14A is to the point in that two categories
appearing on the reports, planning estimates  (targets) for
Preliminary Assessment Completions and Screening Site Inspection
Completions, reconciled to the same numbers.  All other
categories, for both plans and accomplishments, differed.  The
     See Glossary at the end of the report.

                                12

-------
logical conclusion would be that one, or both, of the reports was
in error and that it would be imprudent to use either report.

A detailed technical discussion of these items is contained in
exhibits 2-1, Control over the Production Reports Library; 2-2,
Control over the Source Code Library; and 2-3, Reporting
Inconsistencies.

As a result of the findings discussed above, there could be no
assurance that the correct and authorized version of a program
was in production or that a program was reporting results
consistent with what was reported in other programs.  During our
audit, a number of users told us that errors and inconsistent
reporting were the principal problems they faced in using
CERCLIS.  As a result, they did not rely on CERCLIS data.  Except
for the SCAP (see background section) reports, which report plans
and accomplishments, CERCLIS reports were not heavily utilized.
The reason for continued regional support for the CERCLIS SCAP
reporting was that users' budgets were tied to entry of strategic
data into the system.  Regions were not given their funds, via
the Advice of Allowance process, unless and until they fulfilled
their SCAP data entry requirements.  The data reported by the
SCAP reports were also used as measures of success or failure.

The confusion and lack of faith that are engendered cause users
to abandon a system and, at times, create their own personal
ones.  Most significantly, inconsistencies and errors could
potentially cause loss of confidence in the Superfund program
itself.  In the instance of the source code for a production
program that was not found, modifications will have to be made by
completely rewriting the program.  This represents a significant
waste of computing resources.

Reliance placed by CERCLIS management on informal procedures and
verbal communication was the primary cause for the conditions
discussed here.  Management did not feel the need to closely
monitor and control the development and change control process
for report programming.  Thus, CERCLIS officials never became
aware that its contractors were not ensuring the integrity of the
reporting function and that many of the complaints received from
users concerning errors and inconsistencies could be laid at the
door of the system's own contractors.

DRAFT RECOMMENDATIONS

In response to a position paper which we issued detailing the
findings discussed herein, the Director, OERR, stated that the
following steps were being considered for implementation:
                                13

-------
          signoff from the EPA program office  (report
          owner) will be required for all report
          modifications prior to reinstatement to the
          production reports menu.

          Source programs and load modules will reside
          in a centralized library.4

          Program offices and reports librarian must
          approve changes to the production program
          library.  Librarian will be responsible for
          making sure the library contains the latest,
          approved copy of each program.  Librarian
          will use the time and date stamp of the
          source code to verify the latest version.

          Any changes to source code will be recorded
          within the program in the form of comments.
          Additionally, a report change log must be
          updated each time a report is modified.

          When a change is made to one program report
          developers will consult the reports librarian
          to see if related or affected programs need
          to be changed.

          Users will be notified of modifications to a
          report via electronic mail.

We recommended that the AA for OSWER implement the steps listed
in the OERR response to our position paper on CERCLIS library and
change controls.

SUMMARY OF OSWER'S COMMENTS;

OERR implemented the six steps outlined in its response to our
position paper on CERCLIS documentation by December 29, 1989.

OUR EVALUATION OF OSWER*S COMMENTS;

The decision to implement the six courses of action should help
to resolve the CERCLIS system development and maintenance
problems discussed in this section of the report.  No further
recommendations are being made.
     See Glossary at the end of the report.

                                14

-------
 3.
Formal Testing Procedures for CERCLIS Reports Are Needed
Our examination of  five  CERCLIS reports disclosed that formal
testing procedures  that  included  fixing of responsibility,
ensuring  adequate authorization,  and the use of test transactions
were  not  in effect.  This absence of procedures had given rise to
a  significant  number of  coding errors.  CERCLIS managers had not
ensured that report program testing was adequate because they
believed  that  report content was  exclusively the responsibility
of users.  If  users did  not complain, then it was assumed that
the report in  question was acceptable.  Lack of program quality
assurance procedures leads to unreliable reports which, in turn,
causes users to abandon  use of a  management information system.

Effective testing is necessary to ensure that programming is
reliable.  Without  an adequate level of reliability, users lose
confidence in  and eventually abandon an information system.  NBS
Special Publication 500-106, Guidance on Software Maintenance
(pages 11 and  43),  describes the  place of testing within the
system life cycle process and provides guidelines for its
performance.

Configuration  management is a term which broadly refers to the
identification of the items that  make up a system and formal
controls  over  any changes to it.  One of the goals of
configuration  management is to help maintain the integrity of a
system throughout its life.  The  OSWER System Life Cvcle
Management Guidance. Part 3: Practice Paper. Configuration
Management states:

          A rigorous and disciplined configuration
          management function will maintain system
          integrity for  these systems, and all other
          systems as well, in the following ways. . .:

          Helps ensure that an adequate review of
          requested changes to the system, and approval
          by an authorized organization and individual,
          takes place before the  system is changed. . .
          Helps ensure effective control over changes
          to the software and release of changes to
          users....
We uncovered the reporting errors being discussed here by the
method of developing our own programming which duplicated the
select logic purportedly being used for the selected reports.
limited our review to only five reports because of the time-
consuming nature of our methodology and the need to perform a
substantial amount of additional work in order to confirm the

                                15
                                                          We

-------
discrepancies that were  revealed.  The reports selected represent
a broad  spectrum:  financial,  removal, SCAP, and enforcement.

Although all of  the  errors were serious, we were especially
concerned with the financial and SCAP reports because so much of
the  planning is  dependent on the information contained within
them.  The  financial report  (FINC-4A) did not list all applicable
sites  and did not process negative transactions (e.g.,
deobligations) properly.  The  resulting variances for Region 2
amounted to over $2.5 million.  One of the SCAP reports (SCAP-18)
understated prior year obligations for one region by $500
million.  Detailed discussions of our findings are in exhibits
3-1  through 3-4.  A  complete discussion of the probjems
concerning  SCAP-18 is included in exhibit 2-3B.

We found that formal testing and acceptance procedures for new
report programming or changes  to old reports did not exist.  If
CERCLIS  officials did not receive negative comments on a report
from the field,  the  assumption was that all was well.  A data
base of  test transactions was  not being used.  We were informed
that a test data base was being developed and would be
implemented.   Furthermore, responsibility for SCAP and
enforcement (OWPE) report testing was relegated to the program
(user) offices responsible for developing report requirements and
not  to CERCLIS management.

The  errors  that were revealed  by our review of just five of the
reports  available gave ample indication of the dangers inherent
in running  systems that  have not been subjected to adequate
testing  procedures.   Users cannot rely on reports which have not
been tested and which contain  so many errors.  Furthermore,
errors known to exist in some  reports can cause users to suspect
all  reports and to completely  abandon a system.  When that
happens,  people usually  start  developing their own personal
information systems.  These private systems represent wasted
resources and duplication of efforts.  Furthermore, should EPA
management  or Congress rely on these reports, there is a
likelihood, given the high error rate, that any decisions made
would  be  erroneous ones.

Conclusions

Testing procedures for CERCLIS report programs must be improved
if users  are to gain confidence in the system and the data it
provides.   There must be a review and approval process put in
place  that  certifies that programming meets requirements and is
reliable  before it is entered  into production.  CERCLIS
management  must be directly involved in the approval process and
must continually monitor it if it is to succeed.
                                16

-------
DRAFT RECOMMENDATIONS

In response to the position paper which we issued on Reporting
Consistency, the Director, OERR, stated that the following
actions would be taken immediately:

          OERR/MSDS will review reports usage analysis
          and solicit user comments to identify all
          reports critical to end of the year reporting
          and FY90 planning.  These reports will be
          given highest priority for review and
          correction.  As problems are identified,
          users will be informed.

          Each of these critical reports will be placed
          on a fast track for systems analysis, program
          confirmation and sign-off, reprogramming,
          verification of code against select logic,
          testing, approval by program office and
          submission to the National Library.

          All reports not identified as critical to end
          of the year reporting or FY90 planning will
          be removed from the CERCLIS National Reports
          Menu and made available only through a
          special menu on the production system until
          such time as each report can be verified,
          tested and released onto the National Reports
          Menu.  As part of this complete audit of
          CERCLIS reports, reports will also be
          identified for deletion or combination.

In response to a position paper which we issued detailing the
findings discussed herein, the Director, OERR, stated that the
following steps were being considered for implementation:

          A report change request form and a new report
          request form will be the only mechanisms by
          which existing reports can be modified or new
          reports requested, respectively.  Report
          testers will use these documents to ensure
          that modifications have been implemented
          correctly and that new reports meet original
          specifications.

          For new reports, report testers will use
          report documentation, specifically,
          specification forms and report layouts, to
          validate the successful development of the
          report.  The report will be executed using
          test criteria and the results will be
          compared against the initial documentation.

                                17

-------
          As each new report is developed, it will be
          added to the CEREVAL and CERTEST system
          directly after programming and initial
          testing of the report has been completed.
          CEREVAL will be accessible to all CERCLIS
          users while CERTEST will be limited to
          independent testers only.

          At the same time  independent report
          developers are testing a report, the program
          office or report  owner will be evaluating the
          report to verify  validity of the report.
          Reports librarian members will solicit
          comments from report requestors/users
          concerning the acceptance of the report.
          Report users can  also use the CERCLIS Hotline
          to report any problems they have encountered
          with the report output.

The response also indicated that when the reports we identified
as requiring modification in our position paper on reporting
inconsistency would be removed from the production menu, they
would be added to the CEREVAL and CERTEST systems.  CEREVAL will
contain reports currently being used by report requesters but
that are still undergoing testing.  CERTEST will contain reports
currently undergoing testing by independent testers.  At some
later date, all remaining production reports would be evaluated
for selection criteria consistency and mathematical accuracy.

We recommended that the AA  for OSWER:

     1.   Implement the three actions quoted above which were
          included in the response to our position paper on
          CERCLIS report testing.

     2.   Immediately evaluate the five reports cited in this
          discussion in the same way the reports cited  in
          Finding 2 were evaluated.

     3.   Abandon the concept of allowing users to access reports
          for which testing has not been completed.

SUMMARY OF OSWER'S COMMENTS:

OERR implemented each of the recommendations by December 29,
1989.

OUR EVALUATION OF OSWER'S COMMENTS:

The decision to implement all recommendations should help to
resolve the CERCLIS report  quality assurance problems discussed
herein.  No further recommendations are being made.

                                18

-------
                        TECHNICAL EXHIBITS

                        TABLE OF CONTENTS
PROGRAM DOCUMENTATION
     EXHIBIT 1-1
     EXHIBIT 1-2
     EXHIBIT 1-3
     EXHIBIT 1-4
     EXHIBIT 1-5
SCAP-14B
RMDL-21
RMVL-18
FINC-4A
ENFR-22 AND ENFR-13
20
22
24
25
26
LIBRARY AND CHANGE CONTROLS
     EXHIBIT 2-1
     EXHIBIT 2-2
     EXHIBIT 2-3
TESTING
     EXHIBIT 3-1
     EXHIBIT 3-2
     EXHIBIT 3-3
     EXHIBIT 3-4
Control over the Production Reports Library  27
Control over the Source Code Library         29
Reporting Inconsistencies                    30
FINC-4A
RMVL-18
SCAP-14B
ENFR-22
35
38
39
41
                                19

-------
     SCAP-14B;
     Report
                                      EXHIBIT 1-1

FY89 Targets and Accomplishments Site Summarv
The Report Specification Form was to complement the information
listed  in the CERCLIS Reports Library.  In the Reports Library
listing there was a summary of "Data Elements Used.1'  The
following CERCLIS fields were included in the summary of "Data
Elements Used" for the SCAP-14B in the CERCLIS Reports Library,
but were not referred to in the Report Specification Form:

             • Clll      ENTRY-CITY
               C137      ENTRY-SITE-INCIDENT-CATEGORY
               C226      ENTRY CLASSIFICATION
               C309      ENTRY-FINAL-NPL-UPDATE-NO
               C315      ENTRY-FMS-SS-ID
               C1719     ENF-ACT-NEG-LIT-OUTCOME
               C3124     SUBEVENT-ACTUAL-START

We also noted that the Report Specification Form included
reference in several instances to CERHELP field CP210, "T-A-FUND-
FINANCED-CEILING."  This field was not referred to in the listing
for SCAP-14B in the reports library.  Similarly, CERCLIS fields
2141, EVEN-ACTUAL-COMPLETION, 2903, ENF-FIN-TYPE and 2907, ENF-
FIN-AMOUNT were not referred to in the reports library.

A review of the source code for the SCAP-14B disclosed that other
than a  general and brief four line program description, there
were no other comments.  The variables, their meaning and use
within  the program, were not explained.  The general flow of the
program was not supported by comments either.

The first page of the source code included a listing of changes
made in FY 1989.  There was no evidence based on this listing
alone that the listing was all inclusive;  that is, that no other
changes had been made.  While the name of the original program's
author was noted, no indication of who wrote and compiled the
changes was made.  The CSC managers informed us that comments
would be added after the next round of changes to the SCAP.

The following fields were used by the source program, but were
not listed in the reports library:

                    CERCLIS

                    C2141     EVENT-ACTUAL-COMPLETION
                    C2801     ENF-MS-MILESTONE
                    C2903     ENF-FIN-TYPE
                    CERHELP
                    CP209
              T-A-SCAP-SPMS-FLAG

               20

-------
                    CP1204    TARGET-NUMBER
                    CP2203    TARGET-SITE-EPA-ID
                    CP2204    TARGET-SITE-NAME
                    CP2205    TARGET-SITE-OPERABLE-UNIT
                    CP2206    TARGET-SITE-EVENT-CODE
                    CP2207    TARGET-SITE-ENF-ACTIVITY
                    CP2209    T-A-SITE-EVENT-LEAD

C2801, CP1204, and CP209 were not found in the Report
Specification Form either.

There was no information at all in the Report Specification Form
on how CERHELP and CERCLIS records for enforcement were to be
matched.  A description of such a matching was provided for the
Event (fund) side only.
                                21

-------
                                                      EXHIBIT 1-2
     RMDL-21: Sites on the Proposed 175RA List
We examined the write-up of report logic (Report Specification
Form) provided to us by the CSC contractors.  Our examination
disclosed two inaccuracies.  The SUBEVENT field, C3101, was
described as having two characters while it actually had three.
The fields listed as C2128 and C2129 were C3128 and C3129 in the
data base.  In addition, we found the document to be inadequate
as a source of programming specifications.  What information was
to go in what column or row of the report was not specified.  It
was not clear from the write-up whether the report listed all
regions and had regional break logic or if it was run for one
region at a time.

We found it extremely difficult to understand, from the Report
Specification Form, the flow of the program and how records were
to be selected.  After describing a match of a subset of records
between CERHELP and CERCLIS, it then described additional fields
to be selected from CERCLIS.  Only after trial and error testing
and time-consuming examination of source code (which lacked
comments describing variables and processing) did it become
apparent that the processing logic was actually more complex and
additional records, not fields, were required to be selected.  In
duplicating the logic, we found that we selected more records
than were shown on the report, namely, multiple occurrences of
the same event for the same site that were in the data base.
This was particularly true when codes RA and RD were involved
because their selection was not limited by the first start
indicator code.  Omission of these records was not documented  in
the write-up.

Examination of the source code showed the logic to be
significantly different from that indicated by the Report
Specification Form.  According to the code, after an initial
match on a key (EPA ID, OP UNIT, and EVENT) between CERHELP and
CERCLIS, additional CERCLIS records were selected for specific
types of generic EVENTS for those previously matched records
(same EPA ID and OP UNIT).

Processing, per the source code, involved multiple passes through
the data base.  The additional records selected had to match the
EPA ID and OP UNIT of previously matched records.  Secondly,
records were selected for certain generic EVENTS (generic meaning
that the sequence number, the third character, was not
considered).  These were CO, RI, FS, RA and RD.  For CO and RD,
second passes through the data base were made if the initial
selections did not result in the finding of any CO or RD records
that were first starts (first start codes of A or B);  records of
any start status were then selected.  In the case of an RA EVENT,
should a first start code of A or B not be found, a subevent of

                                22

-------
SV was sought.  If there were multiple occurrences of a single
generic event, such as RD1, RD2, etc., only the first, in
ascending order of magnitude, would appear in the report.  None
of this special processing was clearly delineated in the write-
up.

Examination of the description for RMDL-21 in the Reports Library
disclosed that the listing of data elements used in the report
included eleven elements that were not used in the report based
upon the Report Specification Form and our own experience in
duplicating the logic.  The data elements so listed were:

                    CO110     ENTRY-STREET
                    C0204     ENTRY-RPM-OSC-NAME
                    C0111     ENTRY-CITY
                    C0112     ENTRY-ZIP
                    C0305     ENTRY-STATUS
                    C2108     EVENT-TYPE-SORT
                    C2130     EVENT-CURRENT-PLAN-START
                    C2131     EVENT-CURRENT-PLAN-COMPLETION
                    C3202     FINANCIAL-TYPE
                    C3230     FINANCIAL-AMOUNT
                    C3206     FINANCIAL-FMS-FLAG

On the other hand, there were data elements not listed in the
reports library description that were listed in the Report
Specification Form.  These were:

                    C2101     EVENT-TYPE
                    C2132     EVENT-CURRENT-PLAN-START-FYQ
                    C2133     EVENT-CURRENT-PLAN-COMPLETION-FYQ
                    C3101     SUBEVENT-ID
                    C312 4     SUBEVENT-ACTUAL-START
                    C3125     SUBEVENT-ACTUAL-COMPLETION
                    C3128     SUBEVENT-CURRENT-START-FYQ
                    C3129     SUBEVENT-CURR-COMPLETION-DATE-FYQ

The documentation did not include a listing of the applicable
CERHELP data elements.
                                23

-------
                                                      EXHIBIT 1-3

     RMVL~18i   Planned  Starts. Completions. Ongoings, and
     Obligations  for Region

 In the write-up of  report  logic  (Report Specification Form)
 provided to us, there were two locations in which instructions
 indicated that  CERCLIS  records were to be selected for C305, KPL
 indicator, equal  to 'D1,  'P1,  'F1,  'N', or  »R'.  In two other
 places, C305 was  to be  selected  for any value  not equal to  '0'.
 The  results would not be the same because there is a value  of 'S'
 for  which records would not be selected in the first case,  but
 would be selected in the second.

 The  Form also included  a listing of the CERCLIS fields used in
 the  report.  That listing  included fields which were neither
 described in the  write-up  as to  their use nor  could they be found
 in the actual report:

          C3229   FINANCIAL-EVENT-BUDGET-SOURCE
          C3201   FINANCIAL-Id
          C3218   FINANCIAL-PLANNED-OBL-FYQ
          C3202   FINANCIAL-TYPE
          C3225   FINANCIAL-FUND-PRIORITY-STATUS '

 It could not be discerned  from the Report Specification Form how
 records selected  from CERHELP were to be used  or how they related
 to records selected from CERCLIS.

 The  Form provided no explanation of how records were segregated
 by category within  the  report.

 A comparison of the Report Specification Form  with the
 description for RMVL-18 in the CERCLIS reports library disclosed
 that while the reports  library showed that data elements C2122
 and C2123, EVENT-ORIGINAL-PLAN-START-FYQ and EVENT-ORIG-PLAN-
 COMPLETION-FYQ, respectively, were used, the Form indicated that
 C2132 and C2133,  EVENT-CURRENT-PLAN-START-FYQ  and EVENT-CURRENT-
 PLAN-COMPLETION-FYQ, respectively, were used instead.

 We then examined  the source code.  As we noted before in
 discussing other  reports,  necessary comments were omitted.  There
were no comments  describing variables or processing.  After
 further examination, we found that this source code could not be
 for the executable  program, load module, now in production.  This
matter is discussed in more detail in exhibit  2-2.
                                24

-------
    FINC-4A;
    Reoion
                                        EXHIBIT 1-4

Remedial/Removal Site Specific Funding Report bv
A review of the Report Specification Form for the FINO4A report
revealed these problems:

     Source program, load program and JCL libraries were not
     listed in the specifications.

     Under the heading "SELECT CRITERIA" and in the "EVENTS"
     select Iqgic it read ". . .AND C3203 CONTAINS 'FA1. . ."
     when it should have read "AND POSITIONS 3 AND 4 OF C3204
     CONTAIN 'FA1. . .."  C3203 is the Document Control Number
     while C3204 is the FHS Account Number.

     Under the section entitled "BREAKS" the LINE specification
     showed C3204 and then DOCUMENT CONTROL NUMBER.  C3204 is the
     FMS ACCOUNT NUMBER field.  It was unclear whether the report
     broke on account number or document control number.

     Under the heading "SELECT CRITERIA", in the "EVENTS" select
     logic, it read ". . .and C305 NE 0. . .."  Under the heading
     "PROCESSING LOGIC" in section B subsection 1 the last
     sentence read "IF NATION REPORT ALSO C305 NE O".  The select
     logic implied that the logic "C305 NE 0" occurred in all
     cases and the processing logic implied it occurred when
     running only a national report, i.e., all regions.

The FINC-4A source program code was not adequately documented.
There was no explanation of the program's purpose, no comments
explaining the purpose of data groups and individual variables,
and no comment lines explaining the processes that manipulate and
transform data.
                                25

-------
                                                      EXHIBIT 1-5
    ENFR-22:  FY89 Enforcement Planned Dollars
Select logic in the ENFR-22 source program did not natch the
select logic for ENFR-22 in the CERCLIS Reports Library manual.
The program source code indicated that fewer records would be
selected than indicated in the Reports Library select logic
description.
    ENFR-13t
    Dollars
FY89 Enforcement Obligations and Approved Planned
Documentation in the Reports Library Manual had an error in the
select logic for planned obligations.  The documentation read ".
.  .and C2903 EQ D. . ." when it should have read ". . .and C2903
EQ G. .  .."  C2903 is the Enforcement Activity Financial Type
field.
                                26

-------
                                                      EXHIBIT 2-1
    Control over the Production Reports Library
CERCLIS contractor Sycom was supposed to be in charge of the
reports library.  Its staff was supposed to have the exclusive
ability to write source code to the production load library.
Report programming was being developed by CERCLIS contractor CSC.
Source code developed or modified by CSC was to be turned over to
Sycom for entry into the appropriate production load library.  We
found, however, through both discussion and personal observation,
that CSC staff could and did make changes to source code and
immediately update the production load library.  Such a change
was made when one of our auditors referred an error on report
FINC-4A to one of the OERR staff members.  That staff member
referred the problem to a CSC programmer who made the change to
both the source and load modules.  Sycom was not involved at all.

It should also be noted that when the change was made to report
FINC-4A, no similar change was made to related FINC reports.
This type of process can and did lead to report inconsistency;
see exhibit 2-4, section A.

During the course of the audit, we requested access from an OERR
staff member to CERCLIS source code libraries.  At that time, we
were particularly interested in examining the source code for
report RMDL-21, which is actually two programs:  RA175F and
RA175RP, and RMVL-18.  A Sycom staff member thereupon contacted
us and told us we would find the source code we were seeking in
one of the following source libraries:

          NTSH.NATL.PROD.PLEX.COBOL
          NTSH.NATL.PROD.COPY.COBOL
          NTSH.NATL.PROD.SCHM.COBOL

Examination of the contents of the files revealed that the files
we were looking for were not there.

We then contacted the Sycom staff member who said that he guessed
CSC had not turned over the programs to Sycom as yet.  We then
verified that the two programs had been in the production reports
library since at least the beginning of the fiscal year.
RMVL-18 was entered into the production load library in September
1988 and the RMDL-21 in October 1988.
                     i
We thereafter received a call from a CSC supervisor who said that
the programs we wanted were in two files his staff controlled:

          GCMNTSH.SIW.SOURCE.COBOL
          NTSH.ROAR.COBOL

The source programs we were looking for were in these two files.

                                27

-------
(We later found that the version of the source code for RMVL-18
to which we were given access was not even the correct version
for the RMVL-18 report in the production load library.)
                               28

-------
                                                      EXHIBIT 2-2
    Control over the Source Code Library
Examination of the source library control procedures employed by
CERCLIS contractor CSC disclosed that there were no such
procedures in effect.  There could be no assurance that all
program modifications were recorded in the source code and there
was no log, manual or automated, in which all changes and the
current correct version of the source code could be identified.
We were informed by a CSC supervisor that librarian software was
being sought that would provide the necessary controls.

While examining the source code for report RMVL-18, we found that
this source code could not be for the executable program, load
module, now in production.  First, according to the working
storage section of the code provided to us, the fourth line of
the report (the heading) could be one of the following three:

     RMVL-18:  QUARTERLY REMOVAL APPROVED PLANS
     RMVL-18:  QUARTERLY REMOVAL ALTERNATE PLANS
     RMVL-18:  QUARTERLY REMOVAL APR/ALT PLANS

The actual heading of the report was, line 4,

     RMVL-18:  PLANNED STARTS, COMPLETIONS, ONGOINGS, AND

Line 5 was different as well.

As further verification, we checked the date the executable
program in the production library (load module) was compiled with
the date of compilation noted on the source code.  The
date on the source code was February 1988; that of the load
module, September 8, 1988.

It should be noted that a CSC supervisor thought this was the
correct version of the code and that Sycom did not have a copy of
it at all.
                                29

-------
                                                      EXHIBIT 2-3

     ReportingInconsistencies  (Region  2 data only)

A.   Financial  Reports

We  reviewed  the  following  reports:

      FINC-4A - RMDL/RMVL SITE  SPECIFIC FUNDING REPORT (Region
      Level)

      FINC-4B - RMDL/RMVL SITE  SPECIFIC FUNDING REPORT (STATE
      LEVEL)  for  New  York,  New  Jersey, Puerto Rico and the Virgin
      Islands,  i.e.,  States in  Region 2)

      FINC-5A - REMEDIAL SITE SPECIFIC FUNDING REPORT (Region
      Level)

      FINC-6A - REMOVAL SITE SPECIFIC FUNDING REPORT  (Region
      Level)

We  found that  the grand total  for the FINC-4A Report was not
correct and  that negative  financial transactions were not being
deducted from  obligation and outlay amounts, which in turn
affected site  totals.   We also found that sites with NPL
indicators of  "S" were not being selected.  See Finding 3.
These problems were  brought to the attention of CERCLIS program
personnel.   They had a contract programmer correct the FINC-4A
Report and move  it into production.

Because the  FINC-4A  was corrected without regard to correcting
any like reports, the problems originally discovered in the
FINC-4A report still existed in FINC-4B, FINC-5A and FINC-6A and
thus these related reports were no longer reporting information
consistently.  There had been  no follow-up by Superfund or
contractor personnel to correct reports which used the same logic
as FINC-4A.

These problems resulted:

     Total region obligations  varied by $2,631,204 and region
     outlays were incorrect by $1,872,300 when a total was
     computed  from the four state level reports (FINC-4B) for
     Region  2.    The  combined totals from the remedial report
      (FINC-5A)  and the removal report  (FINC-6A) showed the same
     results.  The reason  the  totals differed was that Albert
     Steel Drum  (SSID«02L4, EPAID=NJD000525154) and Chemical
     Insecticides (SSID=0294,  EPAID=NJD98048653) were not on the
     FINC-4B,  FINC-5A AND  FINC-6A reports.  The sites were
     missing because the code  value "S" which represents "HAS
     SCAP PLAN REMEDIAL ACTIVITIES" was omitted for the "S/I NPL
     INDICATOR"  field (C305) in the report programs.

                                30

-------
     Site obligation totals for Region 2 on the FINC-5A and
     FINC-4B were incorrect for two sites because negative
     amounts were not subtracted from the site totals.

          The General Motors/Central Foundry Division site
          (SSID=02A6, EPAID=NYD091972554) showed a zero
          obligation amount for DCN KJ0029, Account 8TFA02K986,
          but the correct amount was a negative $367.  The site
          total was correct.

          The Pomona Oak Wells site (SSID=02D3,
          EPAID=NJD980769350), showed a zero obligation amount
          for DCN KCS389, Account 7QFA02KLD3;  the correct amount
          is a negative $45,000.  The negative $45,000 amount was
          not reflected in the site total.  The total reported
          for obligations for this site on the two reports was
          $1,200,000, but should have been $1,155,000.

     The site total for obligations for the Pomona Oak Wells site
     was still incorrect on the FINC-4A Report, although the
     program had been modified based on our earlier findings.
     The site total reported was $480,000.  The correct total
     should have been $1,155,000.

B.  SCAP-15 and SCAP-18

We compared SCAP-15, "FY89 FINANCIAL SUMMARY - APPROVED FUNDING",
and SCAP-18, "FY89 FINANCIAL SUMMARY  -  TOTAL FUNDING", to
verify the consistency of reported funding amounts.  As implied
by the titles, the major difference between the reports was that
SCAP-15 showed Approved Funds and SCAP-18 showed Total Funds
(Approved and Alternate Funding).  The reports had identical
column and row names (labels) with each having nine columns of
funding data.

The SCAP-15 report column totals and row totals were verified to
be mathematically correct.  The SCAP-18 report output, however,
was found not to be consistent with SCAP-15 and to have errors.
These problems were relayed to EPA staff responsible for SCAP
reports.  We suggested that the SCAP-18 Report be taken off the
CERCLIS Report Menu until the errors could be corrected.  The
SCAP-18 report was, for a time thereafter, listed as unavailable
on the CERCLIS menu,  it has since been returned to the menu and
its problems resolved.  The problems found when we compared the
two reports were:

     Prior Years Obligation Column  -  Logically prior year
     obligation totals should have been the same on both reports.
     Remedial obligations were identical on the SCAP-15 and
     SCAP-18 reports but the subtotal amount was $100,000,000
     less on SCAP-18.  The Remedial RA amount was $300,000,000

                                31

-------
less than the  same amount on the SCAP-15 report and the
Regional Grand Total amount was $500,000,000 less.  These
errors were caused by truncation  of numbers.

Total Remaining Plan Obligation Column - In the SCAP-15
Report  (Approved Funding), $25,000 was reported for
remaining headquarters enforcement funding.  The SCAP-18
report showed  zero dollars for this amount.  If the approved
funding report showed $25,000 as a valid amount then a
logical conclusion could be drawn that the total funding
report  (a report of all funding) should have shown $25,000
or a greater amount.
         t
Unexpended Obligations As Of 89/4 Column - The Remedial
subtotal amount on the SCAP-18 Report did not agree with the
SCAP-15 subtotal amount although both reports had identical
remedial line  item amounts from which the subtotals were
calculated.  The SCAP-15 amount showed the correct sum,
$67,804.9 (dollars in thousands).  SCAP-18 reported a sum of
$32,195.0 thousand which is a difference of $35,609.9
thousand.  The SCAP-18 Remedial RA line item amount was
$300,000,000 less than the SCAP-15 Remedial RA amount.  The
Regional Grand Total line amount was $400,000,000 less on
the SCAP-18 report.  The Remedial RA and Regional Grand
Total differences resulted from truncation errors.

FY 89 Planned/Actual Obligations and Commitments, 89/3
Column - The Enforcement line item amount was minus $28.5
thousand on the SCAP-15 report while the same total on the
SCAP-18 report was a positive $28.5 thousand.  However, the
Regional Grand Total was $31,793.9 thousand on both reports.

FY 89 Planned/Actual Obligations and Commitments, 89/4
Column - The same problem that the Total Remaining Plan
Obligation Column on the Headquarters Enforcement line was
experiencing was occurring in this column.  The Approved
Funding Report (SCAP-15) showed $25.0 thousand while the
Total Funding Report (SCAP-18) showed zero dollars.  The
total funding  amount should have been equal to or greater
than the approved funding amount.  This problem was carried
over to the Annual Funding Column, which is a sum of the
quarterly amounts under the FY 1989 Planned/Actual
Obligations and Commitments title.

Regional Allowance (TFAY9A EXTRAMURAL) Cumulative Obligation
Authority By Quarter - The 89/3, 89/4 and FY 1989 funding
columns reflected truncated left most numbers on the Total
Remedial line and the Regional Grand Total line on the SCAP-
See Glossary on last page of the report.

                           32

-------
      18  Report.   This  caused  a  discrepancy of  $100,000,000  in
      each of  six  cells of  information.

      Prior Years  Outlays - On the Regional Response Outlays row,
      the dollar amount on  SCAP-18 was truncated causing the
      correct  amount to be  in  error by $100,000,000.

 C.   ENFR-13 AND ENFR-22

 We  reviewed two enforcement reports:

      ENFR-13:  FY89 Enforcement Obligations and Approved Planned
              ' Dollars, By Region, State, and  Sitename
      ENFR-22:  FY89 Enforcement Planned Dollars,
               By Region,  Activity/Event Type, and Funding Status

 We  found that FY  1989  planned approved obligation dollars for
 Region 2 were reported differently on enforcement reports ENFR-13
 and  ENFR-22.  Report ENFR-13  showed the Region 2 grand total as
 $6,297,000.  Report ENFR-22 showed a grand total of $4,779,000,
 which is a difference  of $1,518,000.  Several  sites listed on
 ENFR-13  were missing from  ENFR-22.

 It should also be noted that  the select logic  in the ENFR-22
 source program did  not match  the select logic  for ENFR-22 in the
 CERCLIS  Reports Library manual.  The program source code
 indicated that fewer records  would be selected than indicated  in
 the  Reports Library description.

 We discussed the  findings  with  the Chief, Regional Planning
 Section,  Enforcement Branch,  OWPE.  He had no  immediate
 explanation of why  the totals in question were different.  He
 stated that ENFR-22  and ENFR-11 were reports used for showing
 planned  obligation  dollars.   ENFR-13 was not used by headquarters
 but might be used in the regions.  He had been planning to
 propose  that several reports  including ENFR-13 be taken off the
 CERCLIS  menu because they  were  no longer needed.

 In a  later conversation, the  Chief, Regional Planning Section,
 Enforcement Branch,  OWPE,  added that the select logic in the
 ENFR-22  source program was developed for selecting acceptable
 planned  obligations.   Records with planned approved obligation
 dollars  that did  not meet  the select criteria  were not considered
valid records.  He  said that  changes in report procedures for  FY
 1990 would cause  records that did not meet the established  select
criteria  to be printed on  a separate error report.

D.  SCAP-l AND FINANCE-4A

The site  obligation  and outlay  totals for SCAP-l and FINC-4A
reports were checked for consistency and mathematical accuracy.
In one instance,   EPA ID NY002249563 and SSID 02S7, we found that

                                33

-------
the SCAP-1 showed a rounded amount of $1,739 thousands for
outlays to date  (August l, 1989).  The FINC-4A showed a listing
totaling $1,697,518 for the same site.

Our own duplication of the FINC-4A logic revealed two negative
transactions totaling $41,929 that, apparently, while not being
listed on the FINC-4A, were part of the calculation of total
outlays for the  site on that report.  On the other hand, they
played no part in arriving at the total shown on the SCAP-1.

E.  SCAP-13 And  SCAP-14

Comparisons were made between SCAP-13 (SCAP/SPMS Planning
Estimates and Measures Summary Report) and SCAP-14A (FY 1989
Targets and Accomplishments Summary Report) for consistencies in
counts.  Planning estimates on SCAP-13 were compared against
SCAP/SPMS Targets on SCAP-14A.  Accomplishments on SCAP-13 were
compared against accomplishments on SCAP-14A.  Planning estimates
(targets) for Preliminary Assessment  (PA) Completions and
Screening Site Inspection (SSI) Completions on the Activity
Measure/SCAP SPMS Target side of the reports reconciled to the
same numbers.

The like appearance of the two reports and what they appeared to
be showing, led  us to the conclusion that a comparision of data
between them would be appropriate.  Our finding that two
categories actually matched supported that assumption.  Further
discussion with  CERCLIS officials revealed that the two reports
should not have  been compared because the record selection
criteria differed.

Reports which appear to be related and show apparently
inconsistent information do as much damage to the credibility of
a system as reports which are truly related but which do not
reconcile due to error.  The reports did not have detailed
titles, labels or formulas to explain what was being counted as
an accomplishment on each report.  It could only be assumed,
because the accomplishment counts were different for a number of
comparable categories, that the selection criteria for
accomplishments  was different for each report.  For example, it
was unclear on SCAP-13 whether the accomplishment counts for FY
89 included sites already started or completed during the fiscal
year or if a projection based on planning estimates was used.

we found two problems with SCAP-13 that might have also
contributed to differences.  In the total column under
accomplishments  for SSI Completions for State, we found a count
of 29.  The related quarterly columns when added together totaled
30.  In a number of other instances, quarterly columns showed
zeros while the  FY 1989 Totals had numbers greater than zero,
i.e., the FY total could not be verified by the quarterly counts.
                                34

-------
                                                      EXHIBIT 3-1

    FINO4A;  Remedial/Removal Site Specific Funding Report by
    Region

Several significant problems were identified in the FINC-4A
report for Region 2 when it was compared with a duplicate report
developed within the OIG:

Two sites on the OIG duplicate report did not appear on the FINC-
4A production report.  The sites were:

     Albert Steel Drum, SSID=02L4, EPAID=NJD000525154

     Chemical Insecticide, SSID=0294, EPAID=NJD980484653

After we referred this problem to an OERR staff member who in
turn referred it to a CSC programmer, the programmer discovered
that the problem was caused by the code "S" missing as a value
for the NPL INDICATOR field in the production report code logic.
Code "S" was used for a site that "HAS SCAP PLAN REMEDIAL
ACTIVITIES" (CERCLIS Data Dictionary, Page 31, Dated 11/28/88).
Region 2 grand totals were, as a result, understated by the
following dollar amounts: Obligations - $2,631,204, Outlays -
$1,872,300.  The CSC programmer corrected the program source
code, recompiled the program and placed it into production.

In the obligation column two separate negative obligation amounts
were reported in our duplicate program, but not in the FINC-4A
production program.  In the case of General Motors/Central
Foundry Div. (SSID=02A6, EPAID=NYD091972554), the negative
amount, although not appearing, was correctly subtracted from the
site total.  In the case of Pomona Wells (SSID=02D3,
EPAID=NJD980769350), a negative transaction of $45,000 not only
did not appear on the report, but was not subtracted from the
site total.

The FINC-4A production report for Region 3 showed no major
problems.  We observed that two sites had the same SSID (SSID =
03Z3), Colwell Lane Site and Whitemarsh Twp Drum Dump.

Region 4 had sites missing from the FINC-4A report due to the
same problem that caused missing sites on the Region 2 copy of
the report;  see the discussion above.  The missing sites for
Region 4 caused the Region's grand total to be understated by the
following amounts: Commitments - $285,942, Obligations -
$1,289,430 and Outlays - $207,056.  The missing sites were:

     Bay Drum, SSID=04L5, EPAID-FLD088783865

     B M F Industries Petroleum Products, SSID-0410,
     EPAID=ALD990831968

                                35

-------
 Some sites  that  had obligations  and outlays with negative amounts
 did  not show correct site  totals.  The sites  involved were:

      The 62nd Street Dump  (SSID=0452) site showed an outlay site
      total  that  was $22,044  less than the OIG duplicate report
      outlay site total.  The OIG duplicate report showed an
      outlay transaction  of $22,044 subtracted from the site
      total.   We  concluded  that the CERCLIS FINC-4A program
      doubled the negative  transaction amount  in arriving at the
      site total.

      The Martin  Marietta-SODY  (SSID=0467) site showed an outlay
      site total  with $1,734  less than the duplicate report site
      total.   The CERCLIS FINC-4A report program was tripling the
      amount of the  only  negative transaction  of $578.

      Three  sites on the  CERCLIS  FINC-4A report had negative
      transaction amounts that were not reflected in site totals:
          Canton  Plating  and  Bumper Works, Inc.  (SSID=04E6) - a
          $452.00 negative outlay.

          Coleman Evans Wood  Preserving Co.  (SSID=0441) - a
          $968.00 negative outlay.

          Zenith  Chemical  (SSID=04Z5) - a negative obligation of
          $1,273.00.

The Golden Strip  Septic Tank  site  (SSID=04Z4) showed a site
obligation total  of zero  on the FINC-4A report.  The OIG
duplicate program showed  a site total of minus $300,000.  This
total was composed of one positive and two negative transactions
of $300,000.  Superfund finance staff expressed  the opinion that
one of the negative obligation amounts was a duplicate entered by
the region.

By developing transaction listings, we found records that had
incomplete information or questionable field values.  While these
could have been examples, in  some  instances of poor quality
control over entered data, we discuss them here  because of their
impact on the quality of  reporting and because they may indicate
the need for improved data entry edits.  The first two items are
applicable to the entire country (all regions);  the balance
concerns only Region 2:

We found a number of financial records with  missing SSID's
(CERCLIS field C315).  We had been told that this field was
critical to a good deal of processing.  We submitted a national
listing of these  records to the OERR Finance and Administrative
Management Section.

                                36

-------
We found several account numbers with a character "1" in the
first position and the remainder of the field blank.  The
obligation dates on these records were for fiscal years 1981 and
1982.  The obligations were not reported on the FINC-4A report.
We turned over a national list of these account numbers to the
OERR Finance and Administrative Management Section.

In examining Region 2 transactions, we found that:

     Ringwood Mines site (SSID - 0262) showed an outlay of
     $34,000 with no Account Number/DCN, no obligating date and
     no obligating document number.  A copy of this transaction
     was given to the OERR Financial and Administrative
     Management Section.

     Montclair/West Orange Radium Site showed an event as "DAI"
     (design assistance) for commitment and "RD2fl for obligation.
     SSID '0216* showed two different site names:
     Field1  and 'Suffolk County ANG Base'.
'Olean Well
                                37

-------
                                                      EXHIBIT 3-2

     RMVL-18;   Planned Starts. Completions. Ongoings, and
     Obligations  for Region

 A  comparison of the results of our duplication of the select
 logic  contained in available documentation with the actual report
 disclosed  significant discrepancies.  The comparison was made for
 Region 2 only.  In the ongoing sites category, the duplication
 listed 18  event transactions not contained in the CERCLIS report
 and  the CERCLIS report had 7 not contained in the duplication.
 In the planned  completions category, the standard report had
 three  more sites  than the duplication.  Further research revealed
 that the additional records that appeared in the duplication did
 so properly, according to the select logic contained in the
 Report Specification Form and that the questionable transactions
 appearing  in the  CERCLIS report should not have been so listed.
 Listings for the  28 event transactions in question were developed
 to verify  whether, according to the select criteria, these
 entries should  or should not have appeared on the standard
 CERCLIS report.   In each case, we verified that, based on the
 select criteria presented to us, our results were more accurate.

 The  events that did not appear on the official CERCLIS report
 were all of the proper event type, NPL type, fund priority, and
 financial  type  that would warrant their inclusion.  Our
 examination of  the ones that our duplication did not select
 generally  lacked  the correct fund priority and/or financial type.

 Finally, perusal  of the official CERCLIS version of RMVL-18
 disclosed  that  monetary totals did not accurately reflect the sum
 of the amounts  in each of the three categories.  These totals
were for approved spending.  There were no totals for alternate
 spending.

The summary of  planned quarterly approved obligations on the
 final page showed an amount of $500,000 for "CONTINGENCY."  There
was no explanation anywhere of what this amount represented or
where  it came from.  In addition, the number of sites reported in
this summary did  not reflect the detailed listings and was,
therefore,  confusing.
                                38

-------
     SCAP-14B:
     Report
                                      EXHIBIT 3-3

FY89 Targets and Accomplishments Site Summary
Due to the complexity of this report, we limited our review to
only two sections:  RI/FS FIRST STARTS and COST RECOVERY REFERRAL
ACT, SECT 107 RMVLS.  Our findings were referred to an OERR staff
member working on the SCAP reports.

     (1)  RI/FS First Starts.

          We found that if a FY 1988 planned activity was
          accomplished in FY 1989, it would be included in RI/FS
          FIRST STARTS for FY 1989 (this fact was not noted in
          the documentation), but would not be included in the
          total for planned starts.  CSC contractors informed us
          that they thought this bug was being worked on.

          We found, in the case of Region 3, that there were
          seven instances in which the activity LEAD codes did
          not print on the SCAP-14B although they were in the
          data base.  We developed a special report of LEAD codes
          to verify this fact.  In the case of Region 4, one of
          the actual start dates did not print although it was in
          the data base.

     (2)  COST RECOVERY REFERRAL ACT. SECT 107 RMVLS

          Comparison of the official SCAP-14B and a duplicate
          report developed by OIG staff, disclosed that there
          were two Region 2 sites that were not listed on the
          official version, although they apparently should have
          been.  These were Bridgeport Rental, EPA ID
          NJD053292652, and Lone Pine LF, EPA ID NJD980S055425.
          Both had records with SV activities that included a
          SPMS target status of P, financial codes of F, remedy
          types that were either VM or VO or vp (both applicable
          records were remedy types VD), and financial amounts
          over $200,000.  As such, they should have appeared on
          the CERCLIS SCAP-14B report.

          We also noted, for Region 2, that a "plan start11
          scheduled for the first quarter of the next fiscal year
          may have appeared because a current year accomplishment
          date existed.  According to the Report Specification
          Form, the record with the FY 1990 plan date should not
          have been selected.  In addition, its selection caused
          the total of plan dates to be inaccurate because the FY
          •90 date was not added into the total.
                                39

-------
 In  the  case  of Region  3, we found that an FY 1990 plan
 date had been selected, inaccurately.  This time there
 was no  actual start date that might occasion the
 selection of the  1990  date.  In addition, no "Totals"
 line was printed  for the section.

 For Region 4, we  again found that FY 1990 plan dates
 were selected, although such selection was not
 indicated by the  Report Specification Fora, and that
 the plan start date total was inaccurate due to failure
 to  add  FY 1990 plan dates into that total.

 As  an additional  test, we interactively subjected the
 six Region 4 sites that had only target dates (only
 CERHELP planning  records were selected;  no matching
 CERCLIS records were selected), to additional
 processing to determine why there were no matching
 records for  them.  We  found that there were no SV
 transactions for  these sites that amounted to over
 $200,000, even when combining multiple like records.
 The question must then be raised as to why there would
 be  any  reason to  print these records if one of the
 requirements for  this  section is that planned or actual
 financial amounts be over $200,000.  Their presence
 could confuse or  mislead.

 In  reviewing Region 4  data, we noticed that an actual
 start date for McDonald Farm (EPA ID 'GAD981929748')
was printed  on the CERCLIS SCAP-14B, November 30,  1988.
 It  was  not printed on  the OIG duplication.  Further
examination  of the records for this EPA ID within  the
data base revealed that there was no record or
combination  of records that met all of the code and
date selection criteria and had financial value over
$200,000.  The actual  start date should not, per the
documentation, have appeared on the report.
                      40

-------
                                                      EXHIBIT 3-4
     ENFR-22:  FY89 Enforcement Planned Dollars
In examining the ENFR-22 program, which used a SAS data base
version of CERCLIS data and SAS programming, we found coding
errors in a program segment that calculated fiscal year and FY
quarter variables from, in one instance, the current date (the
SAS system variable TODAY) and, in another, the Events and
Enforcement Activity Completion Dates.  The fiscal year and
quarter variables were then concatenated6 resulting in combined
fiscal year/quarter variables  (Example: "894").  The variable
"FYQA" (based -on the completion dates) was then used in the
select logic to compare against the Event or Enforcement Planned
Financial FY/QTR Dates and the other variable, "FYQ" (based on
the current date), was used in comparison with Event or
Enforcement Planned Start Dates.  The results developed when the
fiscal year/quarter variables were incorrectly calculated were
unpredictable.

Errors arose in the way in which quarters were determined.  If
the month was July, the program logic would set the FY quarter to
"3," when it should have been set to "4."  This placed four
months in the third quarter (April, May, June and July).  If the
month was October, the quarter was set to "4," when it should
have been set to one.  For example, October, 1989 would normally
be represented as "901," i.e., FY 90, quarter 1 (the program
increased the fiscal year by 1 if the month was greater than or
equal to 10).  The erroneous coding, however, resulted in the
fiscal year/quarter "904," FY 90, quarter 4.  The effect of this
error would depend on the time of year the report was being run
and/or the Actual Completion dates contained in the file.  The
erroneous section of the source code read as follows:

     IF Y >=4 AND Y <= 7 THEN
          QTR=3;
     ELSE
     IF Y >=7 AND Y <= 10 THEN
          QTR»=4;

     That section should have read:

     IF Y >- 4 AND Y <= 6 THEN
          QTR»3;
     ELSE
     IF Y >-7 AND Y <* 9 THEN
          QTR*4;
     See Glossary at the end of the report,
                                41

-------
When we executed an OIG duplication of the program with the
corrected code, we found that a site was selected that had not
been selected on previous executions.  The site, Chemical Control
Corporation  (EPAID-NJD000607481), had an alternate planned
obligation amount of $300,000.  The grand total planned
obligation dollar amount for Region 2 changed from $5,736,000 to
$6,036,000.

A discussion with officials of the contract firm of BOOZ, Allen
and Hamilton confirmed the errors and we received assurance that
the necessary corrections would be made.
                               42

-------
  !•   source code
  2.   program  library
 3.  production librarv
 4.   load module
5.  truncation
6.   concatenated
 Glossary

 A computer program written in a
 high level (english-like)  language.
 It must be changed to a machine
 language format (load module)
 before it can be processed
 by a computer.

 A special file which is composed of
 one or more individual programs.  It
 has its own directory which records
 the programs contained within  it.
 In IBM mainframe systems,  load
 modules must be in a program library
 to be executed.   Source code does
 not have to be in a program library,
 but often is in a large multiuser
 system.

 This is a program library  which is
 composed of fully tested programs
 that perform the day to day
 processing and reporting for a
 system.   It may be differentiated
 from a  development or test library
 which is composed of programs  not
 fully tested.

 A  computer program which has
 been translated  into a form which
 enables  it  to  be processed by  a
 computer.   Programs in this form
 cannot be  read.   They also cannot  be
 converted back to  source code  format
 so that  they can be read and
 modified.

 The  failure to include in
 calculations and/or show in computer
 output the significant digits  of a
 number.  The significant digits are
 the left most  ones  in  numbers
greater than zero.

Combined end-to-end  two strings of
text or two variable values to form
a single string.
                                43

-------
                   STATES ENv-RGiVY.ElM A. PRO'1C '.O'
                        WASHINGTON  D C  20460
                      FEB   9(990
MEMORANDUM
SUBJECT:  OIG  Draft Report,  "Report on CERCLIS Reporting,"
          Audit Number  E1SFF9-15-0023
FROM:
TO:
        j   Don R.  Clay
        kpAssistant  Administrator

           Kenneth A. Konz
           Acting  Assistant  Inspector General  for Audit
           Office  of  the  Inspector General  (A-109)
     The purpose  of  this  memorandum  is  to  respond to the  findings
and recommendations  contained  in  the subject draft  report.  While
we generally concur with most of the  draft report, we have  restated
the  findings  and  recommendations  which  we   feel require  our
comments .
Summary of Findings  (page  2)

     Except  for the  SCAP  reports,  which  report  plans  and
     accomplishments,   CERCLIS  reports  were  not   heavily
     utilized.
Response

     There  is  a recurring theme in the draft  report  that CERCLIS
reports are not heavily utilized.  We feel that is not an accurate
reflection of CERCLIS based upon our experience.  CERCLIS has been
designed  to meet the  reporting  requirements of  the  Superfund
Comprehensive Accomplishments Plan (SCAP),  the Strategic Planning
and Management System (SPMS),  and the superfund Progress Report
(SPR).   CERCLIS also  is used to  fulfill  hundreds of  Freedom of
Information Act (FOIA) requests received  annually by Headquarters
and Regional Offices.  Additionally,  CERCLIS data residing on the
Headquarters and Regional Office local area networks are used by
program managers for the in-depth analysis  required for planning
and understanding the overall progress of  the Superfund program.
We  believe  these  are  not   inconsequential uses  of  CERCLIS
information and reports.
                                                           frmtiA e* KtCfcttd f*f*t

-------
                              - 2 -
Summary of Findings  (page 2)
     In fact, the OIG audit of the Agency's Superfund Annual
     Report   for   FY  1988   demonstrated   the   significant
     differences between the information being reported at the
     national  level,  based on CERCLIS, and the information
     being maintained by the regions.

Response

     We believe- the wording of this sentence is inconsistent with
the  information in  the audit and implies  that  a  serious problem
exists with data quality  between Headquarters and the Regions.  The
summary finding addresses the utilization of CERCLIS reports.  We
suggest that this statement be reworded or  eliminated in the final
report.


Recommendation  (page 9)

     We  recommend  that  the  Assistant Administrator for  OSWER
     implement and monitor the actions listed in the OERR response
     to our position paper on CERCLIS documentation.


Response

     We accept this  recommendation.  The Director,  OERR, previously
submitted three steps that were under consideration to improve the
CERCLIS documentation in  his response to  the position papers.
These steps were  included  in  the draft report with an evaluation
by auditors that if  implemented, these steps should improve CERCLIS
documentation   to   a   satisfactory   level.    OERR  adopted  the
implementation portion of this recommendation before December 29,
1989, and we request that this be included in the  final  report.

     We also request  that  this recommendation be  reworded  in the
final report  to request that OSWER  develop a monitoring plan or
procedures.  The monitoring of this recommendation will be provided
by the OSWER Information Management staff  (IMS).  Although the new
procedure  will  stay  in place,  we  do not want to  keep this
recommendation open indefinitely.  We will submit a  time frame to
implement  the monitoring  portion  of  this recommendation   in our
response to the final report.
                                 45

-------
Recommendation  (page 14)

     We recommend that the AA for OSWER  implement the steps listed
     in the OERR response to our position paper on CERCLIS library
     and change controls.


Response

     We accept  this recommendation.  Again,  the  Director,  OERR,
previously  submitted six  steps that were being  cdnsidered  for
implementation  to  improve  the integrity of CERCLIS.   These steps
were included in the draft  report with an evaluation by your office
that if implemented,  the  integrity of  CERCLIS  reporting and the
reliance by users on the data reported should materially improve.
We implemented  these recommendations by December 29, 1989, and we
request that this be included in the final report.

Recommendation  (page 18)

     We recommend that the AA for OSWER:

     1.   Implement  the three  actions  quoted  above  which were
          included in the response to our position paper on CERCLIS
          report testing.
Response

     We  accept   this  recommendation.     As  in  the  previous
recommendations,  the Director, OERR,  previously submitted these
three steps, along with several others, that were being considered
for implementation to improve CERCLIS report testing.   These steps
were included in the draft report.  Since December 29,  1989, these
steps have  been  implemented and we request that  this  be  included
in the final report.
R^fnmiriendation  (page  18)

     We recoamend that the  AA for OSWER:

     2.   Immediately evaluate  the five  reports  cited in  this
          discussion  in  the same way the  reports cited in Finding
          2 were evaluated.

Response

     The five reports listed  in the draft report were evaluated by
December 1989,  and  we request that this be  included in the final
report.
                                46

-------
                              - 4 -

Recommendation (page 18)

     We recommend that the AA for OSWER:

     3.   Abandon the concept of allowing users to access reports
          for which testing has not been completed.


Response

     By Decemb'er 29, 1989, we implemented this recommendation and
thus we  request  that this be  included  in the final report.   we
would like to clarify a point concerning this recommendation.  By
placing  untested  reports  in  the CEREVAL  data  base,  only  the
requestor or testers have access to these reports, not all users.


     Thank you for the opportunity to provide our comments to this
draft  report.    If  you have any  questions, please  contact Jose
Acevedo, OSWER Audit Follow-up Coordinator, at 382-4510.
cc:  Christian Holmes
     Mary Gade
     Henry Longest
     Asa Frost
     Thad Juszczak
     Sonya Stelmack
     Jose Acevedo
     Jim Maas
                               47

-------
                                                       APPENDIX B

                       REPORT DISTRIBUTION

Director, Office of Emergency and Remedial Response (OS-200)

Director, Office of Program Management and Technology  (OS-110)

Director, Office of Information Resources Management (PM-211)

Office of the Comptroller (PM-225)

Associate Administrator for Regional Operations and
        State/Local Relations (A-101)

Agency Followup Official (PM-225), Attn:  Director, Resource
        Management Division

Inspector General (A-109)
                                48

-------